Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article &

Our papers are 100% unique and written following academic standards and provided requirements. Get perfect grades by consistently using our writing services. Place your order and get a quality paper today. Rely on us and be on schedule! With our help, you'll never have to worry about deadlines again. Take advantage of our current 20% discount by using the coupon code GET20


Order a Similar Paper Order a Different Paper

Please read each article attached. For each article:

  1. Provide a paragraph summarizing each article
  2. Provide a paragraph discussing the limitations of each study
  3. Provide a paragraph reviewing the article & providing suggestions for improvements

There should be 3 paragraphs addressing each point mentioned above for each article. DO NOT COPY AND PASTE the article, Please write the critique in your own words.

Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article &
American Journal of Epidemiology© The Author(s) 2018. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http:// creativecommons.org/licenses/by-nc/4.0), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected] Vol. 187, No. 11 DOI: 10.1093/aje/kwy142 Advance Access publication: September 10, 2018 Original Contribution Associations of Religious Upbringing With Subsequent Health and Well-Being From Adolescence to Young Adulthood: An Outcome-Wide Analysis Ying Chen and Tyler J. VanderWeele* *Correspondence to Dr. Tyler J. VanderWeele, Department of Epidemiology, Harvard T. H. Chan School of Public Health, Kresge Building, 677 Huntington Avenue, Boston, MA 02115 (e-mail: [email protected]). Initially submitted November 14, 2017; accepted for publication June 29, 2018. In the present study, we prospectively examined the associations of religious involvement in adolescence (includ- ing religious service attendance and prayer or meditation) with a wide array of psychological well-being, mental health, health behavior, physical health, and character strength outcomes in young adulthood. Longitudinal data from the Growing Up Today Study were analyzed using generalized estimating equations. Sample sizes ranged from 5,681 to 7,458, depending on outcome; the mean baseline age was 14.74 years, and there were 8–14 years of follow-up (1999 to either 2007, 2010, or 2013). Bonferroni correction was used to correct for multiple testing. All models were controlled for sociodemographic characteristics, maternal health, and prior values of the outcome vari- ables whenever data were available. Compared with no attendance, at least weekly attendance of religious services was associated with greater life satisfaction and positive affect, a number of character strengths, lower probabilities of marijuana use and early sexual initiation, and fewer lifetime sexual partners. Analyses of prayer or meditation yielded similar results. Although decisions about religion are not shaped principally by health, encouraging service attendance and private practices in adolescents who already hold religious beliefs may be meaningful avenues of development and support, possibly leading to better health and well-being. health; lifecourse; outcome-wide analysis; prayer or meditation; religious service attendance; religious upbringing; well-being Abbreviations: GUTS, Growing Up Today Study; NHSII, Nurses’Health Study II; STI, sexually transmitted infection. America is highly religious ( 1,2). Religious beliefs and prac- tices are likely shaped by a number of factors, the most promi- nent of which may be religious upbringing in early life ( 3,4). It is a common practice for parents to raise their children based on their own religious beliefs ( 5). There has, however, been a con- tinuing decline in religiosity for decades, for the most part due to lower rates in younger generations ( 6,7). Despite the general trends of declining religious participation, there is still consider- able intergenerational religiouscontinuity in the United States ( 4). For instance, recent estimates of the rates of intergenera- tional transmission of religious affiliation were 82% in Jews, 85% in Muslims, 62% in Evangelical Protestants, and 43% in Catholics, and 59% of parents who attended religious ser- vices at least weekly had children who reported frequent ser- vice attendance ( 4). Empirical research suggests that religion is associated with better health and well-being in adults ( 8). For instance, thereis a gradient relationship between frequent religious service attendance and lower mortality risk, even in the most rigorous studies ( 9–14). In other studies, religious involvement has also been linked to a wide range of other outcomes, such as greater psychological well-being, character strengths, reduced mental illness, and healthier behaviors ( 8,15,16). Religious teachings often concern practices related to living a healthy lifestyle and also sometimes explicitly consider character or respect for the body as an integral part of the beliefs ( 15). Individuals engage in religion in a variety of ways, such as public participation, religious affiliation and identity, private practices, and religious coping ( 15). There have only been a limited number of studies in which investigators have com- pared the health associations of multiple forms of religious participation within the same study. Results from studies in adults generally suggest that religious attendance shows the strongest health associations in community samples, whereas 2355Am J Epidemiol.2018;187(11):2355–2364 religious coping is a prominent predictor for recovery and sur- vival in clinically ill populations ( 13,15,17). To date, prior studies have mostly been conducted in adults. However, research has increasingly suggested that religion may confer lifecourse influences and that religion may have even more profound health effects at younger ages ( 18,19). Existing evidence in adolescents suggests that religious involvement may protect against certain behaviors and promote positive practices ( 20–23). These studies are, however, subject to certain limita- tions. Specifically, much of the prior work is cross-sectional. There is often limited control for baseline characteristics, and reverse causation often cannot be ruled out. For example, an observed inverse association between service attendance and depression may be confounded by prior depression status, because depression may affect subsequent service attendance ( 24). In addition, different aspects of religious involvement are often examined in separate studies and a limited number of outcomes are investigated, so that existing evidence re- mains scattered across studies. It may be important to exam- ine multiple health and well-being outcomes simultaneously within the same study ( 25,26). To provide additional insights into the role of religious upbring- ing, we used an outcome-wide analytic approach ( 26) to prospec- tively examine the associations of religious involvement in adolescence with a wide array of psychological, mental, behav- ioral, physical health, and character strengths outcomes in young adulthood. The 2 aspects of religious participation that were examined were frequency of religious service atten- dance (a form of public participation) and frequency of prayer or meditation (a form of private practice). The inde- pendent associations of service attendance and prayer or meditation across outcomes were also examined in a second- ary analysis. We hypothesized that both frequent service attendance and prayer or meditation are each associated with greater psychological, mental, behavioral, and physical health and character strengths outcomes. Drawing upon prior literature in adults ( 13,15,17), we expected that service attendance would have stronger associations with various outcomes than would prayer or meditation. METHODS We used longitudinal data from the Nurses’Health Study II (NHSII) and the Growing Up Today Study (GUTS). NHSII was initiated in 1989, and it enrolled 116,430 nurses aged 25–42 years. In 1996, NHSII participants with children between 9 and 14 years of age were invited to have their children participate in another cohort of GUTS. A total of 16,882 children completed the questionnaires about their health. NHSII and GUTS partici- pants continue to be followed up with mailed or Web-based questionnaires annually or biennially ( 27,28). This study was approved by the Brigham and Women’s Hospital Institutional Review Boards. Religious participation wasfirst assessed in the GUTS 1999 questionnaire wave; therefore, this year was considered as base- line for the present study. The outcome variables were assessed in the most recent waves, either the 2010 wave (for participants aged 23–30 years) or the 2013 or 2007 wave (if data were not availablein the 2010 wave). Of respondents to the 1999 questionnaire (n=12,410), those with missing data on the exposure (n=1,597 on religious service attendance andn=1,621 on prayer or medi- tation) or the outcome variable (nranged from 3,355 to 5,124 for analyses on service attendance and from 3,341 to 5,108 for analy- ses on prayer or meditation, depending on the outcome) were removed from each analysis involving those variables. Missing data on the covariates were imputed from the previous ques- tionnaire year; if no such data were available, the mean values (for continuous variables) or values of the largest category (for categorical variables) of the nonmissing data were used for imputation. This yielded samples of 5,689–7,458 indivi- duals (up to 1,329 were siblings) for analyses on service atten- dance and 5,681–7,448 individuals (up to 1,325 were siblings) for analyses on prayer or meditation, depending on the out- come. Compared with participants who were lost to follow-up in the 2010 questionnaire wave, those who remained in the cohort were older and healthier, had a higher socioeconomic status, and were more likely to report frequent religious partici- pation at baseline; in addition, a higher percentage was female (Web Table 1, available at https://academic.oup.com/aje ). Web Table 2 shows the timing of the assessment of all vari- ables. The exposure variables (service attendance and prayer or meditation) were assessed in the GUTS 1999 questionnaire wave (participants aged 12–19 years). To reduce the possibil- ity of reverse causation, prior values of the outcome variables assessed in wave 1998 or 1999 were used as a covariate when- ever available. Exposure assessment Religious service attendance. Frequency of religious service attendance (1999 wave) was measured using the question,“How often do you go to religious meetings or services?”Response op- tions ranged from 1 (never) to 5 (more than once per week). Re- sponses were grouped into 3 categories: never, less than once per week, and at least once per week ( 29). Prayer or meditation. Frequency of prayer or meditation (1999 wave) was assessed with the question,“How often do you pray or meditate?”Response categories ranged from 1 (never) to 4 (once per day or more). Outcome assessment A wide array of psychological well-being (life satisfaction, positive affect, self-esteem, emotional processing, and emotional expression), character strengths (frequency of volunteering, sense of mission, forgiveness of others, and being registered to vote), physical health (number of physical health problems and overweight/obesity), mental health (depression, anxiety, and probable posttraumatic stress disorder), and health behavioral (cigarette smoking, frequent binge drinking, marijuana use, other illicit drug use, prescription drug misuse, number of life- time sexual partners, early sexual initiation, history of sexually transmitted Infections (STIs), teen pregnancy, abnormal Pap test results) outcomes were assessed (waves 2010, 2013, or 2007). See Web Table 3 and the Web Appendix for details on each measurement. Am J Epidemiol.2018;187(11):2355–2364 2356Chen and VanderWeele Covariates assessment Sociodemographic characteristics. Sociodemographic co- variates included participant age (in years), sex (female or male), race (white or nonwhite), and geographic region (West, Mid- west, South, or Northeast) (GUTS 1999). Maternal covariates included maternal age (in years; NHSII 1999), race (white or nonwhite; NHSII 1999), marital status (NHSII 1997), subjec- tive socioeconomic status in the United States and in the com- munity (both rated on a scale from 1 to 10), and pretax household income (<$50,000, $50,000–$74,999, $75,000–$99,999, or ≥$100,000; NHSII 2001). We also considered census-tract col- lege education rate (used as a continuous variable) and median income (<$50,000, $50,000–$74,999, $75,000–$99,999, or ≥$100,000; NHSII 2001). Maternal depression. The 5-item Mental Health Index ( 30) was used to measure maternal depressive symptoms over the past 4 weeks (NHSII 1997). As in prior work, a score less than 53 was considered to be an indicator of probable depression ( 31). Maternal smoking. The mothers also reported their current smoking status (NHSII 1997). The response categories were yes and no. Prior values of the outcome variables. To reduce the possi- bility of reverse causation, we adjusted for prior values of the outcome variables whenever data were available. Specifically, adjustments were made for prior depressive symptoms, weight status, smoking, drinking, marijuana use, other drug use, prescrip- tion drug misuse, number of lifetime sexual partners, history of early sexual initiation, history of STIs, and history of pregnancy (GUTS 1998 or 1999). Statistical analyses All statistical analyses were performed using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina) (Pvalues were calculated based on 2-sided tests). Distributions of participant characteristics in the full analytic samples werefirst exam- ined. Next,χ 2test and analysis of variance test were used to examine the associations of service attendance and prayer or meditation with covariates separately. In primary analyses, multiple generalized estimating equa- tions werefirst used to regress each health and well-being outcome on religious service attendance in separate models, with adjustment for clustering by sibling status. Continuous outcomes were standardized (mean=0, standard deviation, 1) to facilitate comparison of effect estimates across outcomes. Bonferroni correction was used to correct for multiple testing. For all analyses, we adjusted for sociodemographic character- istics, maternal health, and prior values of the outcome vari- ables whenever available. Next, we reanalyzed the primary sets of models with prayer or meditation as the exposure. Lastly, we included service attendance andprayer or meditation simulta- neously in the models to examine their independent associa- tions across outcomes. Sensitivity analyses were performed to assess the robust- ness of the observed associations to unmeasured confounding ( 32,33). Specifically, we calculated E-values ( 33), which indi- cate the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome on the risk ratio scale to fully account for anobserved exposure-outcome association, above and beyond the measured covariates. RESULTS Descriptive analyses In the full analytic sample, participants were predominantly white, a higher percentage was female, and most had a high family socioeconomic status (Web Tables 4 and 5). The mean baseline age was 14.74 (standard deviation, 1.66) years. Nearly 60% of the participants attended religious services at least weekly, and 36% reported prayer or meditation at least once per day. Participant characteristics by frequency of service atten- dance are shown in Table 1, and characteristics by frequency of prayer or meditation are shown in Web Table 6. Religious service attendance, health, and well-being Compared with never attendance, at least weekly service attendance was subsequently associated with greater life sat- isfaction and positive affect, greater volunteering, greater sense of mission, more forgiveness, and lower probabilities of drug use and early sexual initiation (Table 2). It was also possibly associated with fewer depressive symptoms and lower probabilities of probable posttraumatic stress disorder, cigarette smoking, prescription drug misuse, history of STIs, and abnormal Pap test results, although the associations did not reachP<0.05 after correction for multiple testing. In comparison, there was little difference between less than weekly and never attendance of ser- vices except for in the character outcomes. When service attendance and prayer or meditation were simul- taneously included in the model, the associations of service atten- dance with outcomes were mostly attenuated (Web Table 7), which may be due to the correlation between service attendance and prayer or meditation (r=0.60). Nevertheless, the associa- tions of service attendance with volunteering, forgiveness, mari- juana use, early sexual initiation, and the number of lifetime sexual partners remained at the levelP<0.05 even after correc- tion for multiple testing. Prayer or meditation, health, and well-being Compared with never praying or meditating, at least daily practice was associated with greater positive affect, emotional processing, and emotional expression; greater volunteering, greater sense of mission, and more forgiveness; lower likeli- hoods of drug use, early sexual initiation, STIs, and abnormal Pap test results; and fewer lifetime sexual partners (Table 3). It was also possibly associated with greater life satisfaction and self-esteem, greater likelihood of being registered to vote, fewer depressive symptoms, and a lower risk of cigarette smoking, although the associations did not reach a level ofP<0.05 after correction for multiple testing. Somewhat unexpectedly, com- pared with never praying or meditating, at least daily practice was possibly associated with more, rather than fewer, physical health problems. A comparison of less than daily praying or meditating with never showed little difference, with only a few exceptions. For example, prayer or meditation was positively associated with volunteering, sense of mission, and forgiveness Am J Epidemiol.2018;187(11):2355–2364 Religious Upbringing and Health and Well-Being2357 Table 1.Distribution of Participant Characteristics by Frequency of Religious Service Attendance at Study Baseline (n=10,813), Growing Up Today Study, 1999 Participant CharacteristicFrequency of Religious Service Attendance a PValue Never (n=1,703)Less Than Once per Week (n=2,922)At Least Once per Week (n=6,188) % Mean (SD) % Mean (SD) % Mean (SD) Sociodemographic factors Age, years b 15.03 (1.66) 14.94 (1.67) 14.56 (1.64)<0.001 Male sex 44.86 40.97 40.16 0.002 White race 90.04 93.69 94.07<0.001 Geographic region<0.001 West 27.97 15.31 11.34 Midwest 27.44 32.43 39.23 South 9.25 13.15 16.47 Northeast 35.34 39.11 32.97 Mother’s age, years b 44.91 (3.62) 44.37 (3.53) 43.81 (3.52)<0.001 Mother’s race (white) 95.42 97.53 97.72<0.001 Mother married 87.61 89.32 94.88<0.001 Mother’s subjective SES in the Uniteed States b 7.15 (1.35) 7.13 (1.32) 7.14 (1.28) 0.90 Mother’s subjective SES in the community b 6.87 (1.61) 6.98 (1.56) 7.08 (1.53)<0.001 Pretax household income<0.001 <$50,000 11.74 12.67 13.67 $50,000–$74,999 23.40 21.97 24.83 $75,000–$99,999 20.73 21.43 23.60 ≥$100,000 44.13 43.94 37.90 Census tract college education rate, b% 33.45 (16.77) 32.81 (16.40) 30.33 (15.81)<0.001 Census tract median income<0.001 <$50,000 22.91 23.44 28.18 $50,000–$74,999 45.53 46.34 48.03 $75,000–$99,999 22.39 22.93 18.29 ≥$100,000 9.17 7.29 5.49 Maternal health Maternal depression 11.14 12.14 9.21<0.001 Maternal current smoking 9.87 9.31 5.37<0.001 Prior health status or prior health behaviors Prior depressive symptoms b 1.26 (0.62) 1.24 (0.58) 1.16 (0.57)<0.001 Prior overweight or obesity 21.09 19.18 19.68 0.30 Prior cigarette smoking 24.85 22.59 12.70<0.001 Prior alcohol drinking 13.79 12.16 4.97<0.001 Prior marijuana use 21.97 17.46 7.03<0.001 Prior drug use other than marijuana 8.93 5.50 2.40<0.001 Prior prescription drug misuse 8.83 8.36 5.30<0.001 Prior number of lifetime sexual partners b 0.40 (1.10) 0.27 (0.89) 0.10 (0.55)<0.001 Prior history of early sexual initiation 11.86 8.02 3.24<0.001 Prior history of sexually transmitted infections 0.62 0.40 0.05<0.001 Prior history of teen pregnancy 0.75 0.61 0.26 0.006 Abbreviations: SD, standard deviation; SES, socioeconomic status. aAnalysis of variance orχ 2tests were used to examine the mean (SD) levels of the characteristic or proportion of individuals within each religious service attendance category with that characteristic. bRanges of the participant characteristics were as follows: age, 12–19 years; mother’s age, 35–54 years; mother’s subjective SES in the United States, 1–10; mother’s subjective SES in the community, 1–10; census tract college education rate, 0%–85%; prior depressive symptoms, 0–4; and prior number of lifetime sexual partners, 0–6. Am J Epidemiol.2018;187(11):2355–2364 2358Chen and VanderWeele in a monotonic fashion; compared with never praying or medi- tating, doing so 1–6 times per week was related to greater emo- tional expression, fewer depressive symptoms, and fewer sexualpartners. When prayer or meditation and service attendance were simultaneously included in the model, the associations of at least daily versus never praying or meditating with emotional Table 2.Religious Service Attendance in Adolescence and Health and Well-Being in Young Adulthood (n=5,689–7,458 a), Growing Up Today Study, 1999 to 2007, 2010, or 2013 Health and Well-Being OutcomeReligious Service Attendance Comparison Less Than Once per Week vs. Never At Least Once per Week vs. Never RR b βc 95% CIPValue Threshold RR b βc 95% CIPValue Threshold Psychological well-being Life satisfaction 0.04−0.05, 0.12 0.13 0.05, 0.21<0.0019 d Positive affect 0.09 0.01, 0.17<0.05 0.18 0.10, 0.25<0.0019 d Self-esteem 0.05−0.03, 0.12 0.07−0.00, 0.14 Emotional processing 0.04−0.04, 0.12 0.03−0.05, 0.10 Emotional expression 0.04−0.04, 0.12 0.04−0.03, 0.12 Character strengths Frequency of volunteering 0.13 0.06, 0.20<0.0019 d 0.28 0.21, 0.35<0.0019 d Sense of mission 0.11 0.03, 0.19<0.01 0.28 0.20, 0.35<0.0019 d Forgiveness of others 0.33 0.24, 0.41<0.0019 d 0.69 0.61, 0.77<0.0019 d Registered to vote 1.04 1.01, 1.07<0.01 1.03 1.01, 1.06<0.05 Physical health No. of physical health problems 0.10 0.02, 0.18<0.05 0.02−0.05, 0.09 Overweight/obesity 0.98 0.89, 1.08 1.01 0.92, 1.10 Mental health Depressive symptoms−0.03−0.11, 0.05−0.12−0.19,−0.04<0.01 Depression diagnosis 0.90 0.76, 1.06 0.87 0.75, 1.01 Anxiety symptoms 0.03−0.05, 0.11−0.04−0.11, 0.04 Anxiety diagnosis 1.01 0.84, 1.22 0.89 0.75, 1.07 Probable PTSD 0.87 0.67, 1.13 0.72 0.57, 0.93<0.01 Health behaviors Cigarette smoking 0.99 0.88, 1.11 0.85 0.76, 0.96<0.01 Frequent binge drinking 1.05 0.95, 1.17 0.97 0.87, 1.07 Marijuana use 0.99 0.93, 1.04 0.83 0.78, 0.88<0.0019 d Any other illicit drug use 0.92 0.75, 1.13 0.67 0.55, 0.81<0.0019 d Prescription drug misuse 1.02 0.90, 1.15 0.84 0.74, 0.95<0.01 Number of lifetime sexual partners−0.02−0.09, 0.04−0.28−0.34,−0.21<0.0019 d Early sexual initiation 0.91 0.78, 1.06 0.65 0.55, 0.77<0.0019 d History of STIs 0.99 0.82, 1.20 0.79 0.66, 0.95<0.05 Teen pregnancy 0.81 0.47, 1.37 0.76 0.45, 1.28 Abnormal Pap test results 0.87 0.75, 1.02 0.82 0.71, 0.95<0.01 Abbreviations: CI, confidence interval; PTSD, posttraumatic stress disorder; RR, risk ratio; STIs, sexually transmitted infections. aThe full analytic sample was restricted to those who had valid data on religious service attendance. The actual sample size for each analysis varied depending on the number of missing values for each outcome under investigation. Missing data on the covariates were imputed from previ- ous questionnaire years; if no such data were available, missing data were imputed as the mean values (continuous variables) or values of the larg- est category (categorical variables) of the nonmissing data. All models were controlled for participants’age, race, sex, geographic region, and prior health status or prior health behaviors (prior depressive symptoms, overweight/obesity, smoking, drinking, marijuana use, other drug use, prescrip- tion drug misuse, number of sexual partners, early sexual initiation, history of sexually transmitted infections, history of teen pregnancy), as well as their mother’s age, race, marital status, socioeconomic status (subjective socioeconomic status, household income, census tract college education rate, and census tract median income), depression, and smoking. bThe effect estimates for the outcomes of probable PTSD, any other illicit drug use, and teen pregnancy were odds ratios; these outcomes were rare (prevalence<10%), so the odds ratios would approximate the RRs. The effect estimates for other dichotomized outcomes were RRs. cAll continuous outcomes were standardized (mean=0, standard deviation, 1), andβwas the standardized effect size.dP<0.05 after Bonferroni correction (thePvalue cutoff for Bonferroni correction=0.05/26 outcomes=0.0019). Am J Epidemiol.2018;187(11):2355–2364 Religious Upbringing and Health and Well-Being2359 processing, emotional expression, volunteering, sense of mission, forgiveness, drug use, number of sexual partners, and history of STIs still held (Web Table 7); associations were attenuated, though some remained, when instead controlling for young adult, rather than adolescent, service attendance (Web Table 8). Sensitivity analyses for unmeasured confounding To assess the robustness of the observed associations to unmeasured confounding, we calculated E-values ( 33) for the associations of religious service attendance (at least weekly vs. never) and prayer or meditation (at least daily vs. never) with various outcomes (Table 4). In the present study, there is evidence suggesting that some of the observed associations were likely robust to unmeasured confounding. This is espe- cially true with the character outcomes, drug use, and sexual behaviors. For example, as noted in Table 4, an unmeasuredconfounder would need to be associated with both service attendance and volunteering by risk ratios of 1.90 each to fully explain away the observed association of at least weekly (vs. never) attendance of services with volunteering and by 1.72-fold each to shift the lower confidence limit for the estimate to include the null value, above and beyond the measured covariates. DISCUSSION There is growing interest in promoting protective factors that lead to better health, beyond the traditional approach that focuses on reducing risk factors for diseases ( 34). Once risk factors are established, it can be difficult to restore a healthy state. It may be more effective to promote and maintain health and well-being starting in early life ( 35). Results from the pres- ent study suggest that religious involvement in adolescence may Table 3.Prayer or Meditation in Adolescence and Health and Well-Being in Young Adulthood (n=5,689–7,448 a), Growing Up Today Study, 1999 to 2007, 2010, or 2013 Health and Well-Being OutcomePrayer or Meditation Comparison Less Than Once per Week vs. Never 1–6 Times per Week vs. Never Once per Day or More vs. Never RR b βc 95% CIPValue ThresholdRR b βc 95% CIPValue ThresholdRR b βc 95% CIPValue Threshold Psychological well- being Life satisfaction 0.05−0.04, 0.13 0.10 0.02, 0.17<0.05 0.12 0.04, 0.20<0.01 Positive affect 0.07−0.01, 0.15 0.11 0.04, 0.18<0.01 0.16 0.08, 0.23<0.0019 d Self-esteem 0.01−0.07, 0.09 0.10 0.02, 0.18<0.05 0.08 0.00, 0.15<0.05 Emotional processing0.03−0.05, 0.12 0.10 0.02, 0.18<0.05 0.13 0.06, 0.21<0.0019 d Emotional expression0.08 0.00, 0.17<0.05 0.13 0.06, 0.21<0.0019 d 0.15 0.07, 0.22<0.0019 d Character strengths Frequency of volunteering0.14 0.07, 0.22<0.0019 d 0.27 0.20, 0.34<0.0019 d 0.36 0.29, 0.43<0.0019 d Sense of mission 0.14 0.05, 0.22<0.0019 d 0.21 0.13, 0.28<0.0019 d 0.43 0.36, 0.51<0.0019 d Forgiveness of others0.37 0.29, 0.46<0.0019 d 0.60 0.52, 0.68<0.0019 d 0.83 0.75, 0.91<0.0019 d Registered to vote 1.01 0.99, 1.04 1.01 0.99, 1.04 1.03 1.00, 1.05<0.05 Physical health Number of physical health problems0.10 0.02, 0.18<0.05 0.02−0.05, 0.10 0.08 0.01, 0.15<0.05 Overweight/ obesity1.02 0.92, 1.13 0.99 0.90, 1.10 1.00 0.91, 1.10 Mental health Depressive symptoms−0.07−0.16, 0.01−0.15−0.22,−0.07<0.0019 d −0.09−0.16,−0.01<0.05 Depression diagnosis0.93 0.78, 1.10 0.95 0.80, 1.12 0.88 0.74, 1.03 Anxiety symptoms 0.02−0.06, 0.10 0.00−0.08, 0.07 0.04−0.03, 0.11 Anxiety diagnosis 1.00 0.82, 1.23 1.00 0.82, 1.21 0.96 0.79, 1.16 Probable PTSD 0.72 0.53, 0.97<0.05 0.93 0.72, 1.21 0.94 0.73, 1.22 Table continues Am J Epidemiol.2018;187(11):2355–2364 2360Chen and VanderWeele be one such protective factor for a range of health and well- being outcomes ( 20). Consistent with prior literature, our results suggest associa- tions of frequent religious participation in adolescence with greater subsequent psychological well-being, character strengths, and lower risks of mental illness and several health behaviors ( 36–38). For instance, congruent with prior meta-analyses of mostly cross-sectional adolescent studies on religion and health behaviors ( 37,38), we found reduced probabilities of drug use and several sexual behaviors among religiously observant ado- lescents. Also, consistent with results from a prior meta-analysis of religion and forgiveness ( 39), we found a positive association of religious involvement with forgiveness in early life. Likewise, the effect size between religious involvement and depressive symptoms in the present study is similar to that from a meta- analysis (β=−0.09, 95% confidence interval:−0.11,−0.08) in which investigators integrated evidence across ages ( 40). In our study, there was little association between religious involvement and anxiety, which is in fact consistent with results from other prior longitudinal studies of adult populations ( 15)andcontrasts with results from cross-sectional studies ( 16). Our study adds to prior literature by providing evidence from longitudinal data withconfounding control and also control for baseline values of the outcome variables. Contrary to our expectation, ourresults suggest that frequent prayer or meditation may be associated with more physical health problems. To our knowledge, the association between religion and adolescent physical health has not been well- studied; we are not aware of any prior longitudinal work in this area using community samples of adolescents ( 41). There is, however, evidence from clinical populations that individuals with chronic conditions are more likely to use private religious practices to cope with illness ( 41,42). Because of the lack of available data, we did not control for baseline physical health. The inverse association between prayer or meditation and physical health in the present study may in part be due to reverse causation. Those who already have physical health pro- blems may be more likely to pray. It is also conceivable that those with religious beliefs may sometimes avoid medical care because of these beliefs or potentially thinking that the prayer will suffice for healing. Service attendance is generally the strongest religious/spiritual predictor of health in nonclinical adult samples ( 8,13,17,36). In contrast, we found that compared with service attendance, prayer Table 3.Continued Health and Well-Being OutcomePrayer or Meditation Comparison Less Than Once per Week vs. Never 1–6 Times per Week vs. Never Once per Day or More vs. Never RR b βc 95% CIPValue ThresholdRR b βc 95% CIPValue ThresholdRR b βc 95% CIPValue Threshold Health behaviors Cigarette smoking 0.98 0.86, 1.11 0.99 0.88, 1.12 0.89 0.78, 1.00<0.05 Frequent binge drinking0.97 0.87, 1.09 1.00 0.90, 1.10 0.91 0.82, 1.01 Marijuana use 0.99 0.93, 1.05 0.92 0.87, 0.97<0.01 0.75 0.71, 0.80<0.0019 d Any other illicit drug use0.91 0.74, 1.12 0.75 0.62, 0.92<0.01 0.56 0.46, 0.69<0.0019 d Prescription drug misuse0.90 0.79, 1.02 0.88 0.78, 0.99<0.05 0.72 0.64, 0.82<0.0019 d Number of lifetime sexual partners−0.05−0.12, 0.02−0.13−0.20,−0.07<0.0019 d −0.40−0.46,−0.34<0.0019 d Early sexual initiation1.05 0.89, 1.24 0.84 0.71, 1.00 0.70 0.59, 0.84<0.0019 d History of STIs 0.90 0.68, 1.18 0.83 0.64, 1.08 0.60 0.47, 0.78<0.0019 d Teen pregnancy 0.87 0.50, 1.52 0.64 0.36, 1.15 0.88 0.52, 1.48 Abnormal Pap test results0.82 0.70, 0.98<0.05 0.95 0.81, 1.11 0.74 0.63, 0.88<0.0019 d Abbreviations: CI, confidence interval; PTSD, posttraumatic stress disorder; RR, risk ratio; STIs, sexually transmitted infections.aThe full analytic sample was restricted to those who had valid data on frequency of prayer or meditation. The actual sample size for each analy- sis varied depending on the number of missing values for each outcome under investigation. Missing data on the covariates were imputed from pre- vious questionnaire years; if no such data were available, missing were imputed as the mean values (continuous variables) or values of the largest category (categorical variables) of the nonmissing data. All models were controlled for participants’age, race, sex, geographic region, and prior health status or prior health behaviors (prior depressive symptoms, overweight/obesity, smoking, drinking, marijuana use, other drug use, prescrip- tion drug misuse, number of sexual partners, early sexual initiation, history of sexually transmitted infections, history of teen pregnancy), as well as their mother’s age, race, marital status, socioeconomic status (subjective socioeconomic status, household income, census tract college education rate, and census tract median income), depression, and smoking. bThe effect estimates for the outcomes of probable PTSD, any other illicit drug use, and teen pregnancy were odds ratios; these outcomes were rare (prevalence<10%), so the odds ratios would approximate the RRs. The effect estimates for other dichotomized outcomes were RRs. cAll continuous outcomes were standardized (mean=0, standard deviation, 1), andβwas the standardized effect size.dP<0.05 after Bonferroni correction (thePvalue cutoff for Bonferroni correction=0.05/26 outcomes=0.0019). Am J Epidemiol.2018;187(11):2355–2364 Religious Upbringing and Health and Well-Being2361 or meditation had more robust associations with a number of outcomes, including emotional processing, emotional expression, number of physical health problems, prescription drug misuse, his- tory of STIs, and abnormal Pap testresults. The exceptions to this were for life satisfaction, positive affect, probable posttraumatic stress disorder, cigarette smoking, and early sexual initiation,for which the associations with service attendance were stronger. In adolescent populations, service attendance may be a marker of parental service attendance patterns that may not persist into later life, whereas private religious practices may more closely corre- spond to their own service attendance patterns later in life ( 43). Adjustment of the prayer or meditation analyses for young adult Table 4.Robustness to Unmeasured Confounding (E-Values a) for Assessing the Causal Associations Between Religious Upbringing in Adolescence and Health and Well-Being in Young Adulthood (n=5,681–7,458 a), Growing Up Today Study, 2007, 2010, or 2013 Health and Well-Being OutcomeReligious Service Attendance Prayer or Meditation For Effect Estimate b For CI Limit c For Effect Estimate b For CI Limit c Life satisfaction 1.50 1.28 1.47 1.25 Positive affect 1.64 1.44 1.58 1.38 Self-esteem 1.33 1.00 1.36 1.07 Emotional processing 1.20 1.00 1.50 1.28 Emotional expression 1.23 1.00 1.56 1.35 Frequency of volunteering 1.90 1.72 2.12 1.93 Sense of mission 1.90 1.71 2.32 2.11 Forgiveness of others 3.15 2.88 3.68 3.37 Registered to vote 1.21 1.11 1.21 1.08 Number of physical health problems 1.16 1.00 1.36 1.10 Overweight/obesity 1.11 1.00 1.00 1.00 Depressive symptoms 1.47 1.25 1.39 1.13 Depression diagnosis 1.56 1.00 1.53 1.00 Anxiety symptoms 1.23 1.00 1.23 1.00 Anxiety diagnosis 1.50 1.00 1.25 1.00 Probable posttraumatic stress disorder 2.12 1.36 1.32 1.00 Cigarette smoking 1.63 1.25 1.50 1.03 Binge drinking 1.21 1.00 1.43 1.00 Marijuana use 1.70 1.53 2.00 1.81 Any other illicit drug use 2.35 1.77 2.97 2.26 Prescription drug misuse 1.67 1.29 2.12 1.74 Number of lifetime sexual partners 1.90 1.73 2.23 2.06 Early sexual initiation 2.45 1.92 2.21 1.67 History of sexually transmitted infections 1.85 1.29 2.72 1.88 Teen pregnancy 1.96 1.00 1.53 1.00 Abnormal Pap test 1.74 1.29 2.04 1.53 Abbreviation: CI, confidence interval. aSee VanderWeele and Ding ( 33) for the formula for calculating E-values. bThe E-values for effect estimates are the minimum strength of association on the risk ratio scale that an unmea- sured confounder would need to have with both the exposure and the outcome to fully explain away the observed as- sociations of religious service attendance (at least weekly vs. never) and prayer or meditation (at least daily vs. never) with various health outcomes as shown in the last column of Tables 2and 3, conditional on the measured covariates. For example, an unmeasured confounder would need to be associated with both religious service attendance and for- giveness of others by risk ratios of 3.15 each, above and beyond the measured covariates, to fully explain away the observed association between at least weekly religious service attendance and forgiveness of others. cThe E-values for the limit of the 95% CI closest to the null denote the minimum strength of association on the risk ratio scale that an unmeasured confounder would need to have with both the exposure and the outcome to shift the confidence interval to include the null value, conditional on the measured covariates. For example, an unmeasured confounder would need to be associated with both religious service attendance and forgiveness of others by 2.88-fold each, above and beyond the measured covariates, to shift the lower limit of the CI for the observed association between at least weekly service attendance and forgiveness of others to include the null value. Am J Epidemiol.2018;187(11):2355–2364 2362Chen and VanderWeele service attendance only partially attenuated the associations, per- haps suggesting some independent effect. Adolescents are particularly vulnerable to heightened inter- est in pursuing thrill-seeking behaviors ( 18). The behavioral norms and patterns formed in this period may, in fact, exert profound in- fluences over the lifecourse ( 18). For example, the initiation of smoking is more likely to occur in adolescence than in other stage of life, if it happens at all. Adolescents who have initiated smoking are also likely to continue smoking into adulthood ( 44). Therefore, if a resilience factor can protect adolescents from initi- ating smoking, it may reduce their lifetime health risk substan- tially ( 18). The present study adds to prior evidence suggesting that religious involvement in adolescence may serve as one such protective factor in not only reducing smoking but also in main- taining psychological well-being,developing character strengths, and reducing certain behaviors, as well as possibly also reducing depression. The beneficial effects of religious involvement in adolescence may function through a number of mechanisms. For instance, religion provides directives or personal virtue to help maintain self-control and develop negative attitudes toward certain behaviors ( 45). Some religious groups promote beliefs that create meaning and practices that foster active coping, such as practicing forgiveness and meditation, which could help youth actively cope with stress ( 45). Moreover, peer religious youth groups may be an important source of social support and adult role modeling, and they may be an avenue to direct peer influ- ences on behavioral choices. Religious congregations could also connect adolescents to networks and resources in the broader community ( 41,45). The present study advances beyond prior literature in a number of ways. First, we took an outcome-wide analytic approach to provide a broad picture of the roleof religious participation dur- ing adolescence in relation to a wide range of health and well- being outcomes within the same sample, which helps synthesize previously scattered evidence on individual health outcomes in separate studies. Second, the longitudinal design and the follow- up periods of 8–14 years help establish temporal ordering for as- sessing causality. Third, the longitudinal data along with the adjustment for baseline values of the outcome variables help reduce the possibility of reverse causation, which has been identi- fied as a major threat to assessing causal effects of religious prac- tice ( 8). The present study also used sensitivity analyses to assess the robustness of the associations to unmeasuredconfounding, which provides further evidence for assessing causality. Our study is, however, subject to certain limitations. First, religious involvement was measured with 2 single items that were widely used in adults. These measures did not consider developmental characteristics of adolescents. For instance, adolescents’decisions on religious participation are likely shaped by both parents and peers. It may, therefore, be important to assess influences from both (e.g., pressure by parents to attend religious services and participation in peer religious youth groups) to facilitate understanding in a developmentally relevant frame- work ( 46). Second, the results may be subject to residual con- founding by parental religiousness(e.g., parental church attendance and parental religious affiliation) for which information was not available. However, we adjusted for baseline maternal depres- sion and smoking status, whichhave both been linked to reli- gious participation and to child outcomes ( 16,47). Results from the sensitivity analysis also suggest that a number of the observedassociations are relatively robust to potential unmeasured con- founding. As a further limitation, GUTS participants were mostly white, and their mothers all worked in the nursingfield. Therefore, results of this study may not be generalizable to other populations. There is evidence that religion is an important social deter- minant of health over the lifecourse ( 18 ). Religious participa- tion in adulthood is, in many cases, a function of religious upbringing in early life ( 18). Intergenerational transmission of religious values and practices occurs largely through parental modeling and is likely facilitated by close parent-child relation- ships ( 48). Although decisions about religion are not shaped principally by health, for adolescents who already hold reli- gious beliefs, encouraging service attendance and private prac- tices may be meaningful avenues of development and support, possibly leading to better health and well-being. ACKNOWLEDGMENTS Author affiliations: Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Ying Chen, Tyler J. VanderWeele); and Human Flourishing Program, Institute for Quantitative Social Science, Harvard University, Cambridge, Massachusetts (Ying Chen, Tyler J. VanderWeele). This work was funded by the Templeton Foundation (grant 52125) and the National Institutes of Health (grant ES017876). The National Institutes of Health supports the Nurses’Health Study II (grant UM1CA176726) and the Growing Up Today Study (grants R01HD045763, R01HD057368, R01HD066963, R01DA033974, K01DA023610, and K01DA034753). We thank the Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School for their support in conducting this study. The funding agencies had no role in the data collection, analysis, or interpretation, nor were they involved in the writing or submission of this publication. Conflict of interest: none declared. REFERENCES 1. Gallup. In depth: religion. http://www.gallup.com/poll/1690/ religion.aspx . Accessed September 15, 2017. 2. Lugo L, Stencel S, Green J, et al.US Religious Landscape Survey. Washington, DC: Pew Research Center; 2008. 3. Cornwall M. The determinants of religious behavior: a theoretical model and empirical test.Soc Forces. 1989;68(2): 572–592. 4. Bengtson VL.Families and Faith: How Religion Is Passed Down Across Generations. New York, NY: Oxford University Press; 2013. 5. Astley J, Francis LJ.Critical Perspectives on Christian Education: A Reader on the Aims, Principles and Philosophy of Christian Education. Leominster, UK: Gracewing; 1994. 6. Voas D, Chaves M. Is the United States a counterexample to the secularization thesis?Am J Sociol. 2016;121(5): 1517–1556. Am J Epidemiol.2018;187(11):2355–2364 Religious Upbringing and Health and Well-Being2363 7. Uecker JE, Regnerus MD, Vaaler ML. Losing my religion: the social sources of religious decline in early adulthood.Soc Forces. 2007;85(4):1667–1692. 8. VanderWeele TJ. Religion and health: a synthesis. In: Peteet JR, Balboni MJ, eds.Spirituality and Religion Within the Culture of Medicine: From Evidence to Practice. New York, NY: Oxford University Press; 2017:357–402. 9. Li S, Stampfer MJ, Williams DR, et al. Association of religious service attendance with mortality among women.JAMA Intern Med. 2016;176(6):777–785. 10. Hummer RA, Rogers RG, Nam CB, et al. Religious involvement and US adult mortality.Demography. 1999; 36(2):273–285. 11. McCullough ME, Hoyt WT, Larson DB, et al. Religious involvement and mortality: a meta-analytic review.Health Psychol. 2000;19(3):211–222. 12. Gillum RF, King DE, Obisesan TO, et al. Frequency of attendance at religious services and mortality in a US national cohort.Ann Epidemiol. 2008;18(2):124–129. 13. Musick MA, House JS, Williams DR. Attendance at religious services and mortality in a national sample.J Health Soc Behav. 2004;45(2):198–213. 14. Chida Y, Steptoe A, Powell LH. Religiosity/spirituality and mortality: a systematic quantitative review.Psychother Psychosom. 2009;78(2):81–90. 15. George LK, Ellison CG, Larson DB. Explaining the relationships between religious involvement and health. Psychol Inq. 2002;13(3):190–200. 16. Koenig HG, King DE, Carson VB, et al.Handbook of Religion and Health. 2nd ed. New York, NY: Oxford University Press; 2012. 17. VanderWeele TJ, Yu J, Cozier YC, et al. Attendance at religious services, prayer, religious coping, and religious/ spiritual identity as predictors of all-cause mortality in the Black Women’s Health Study.Am J Epidemiol. 2017;185(7): 515–522. 18. Idler EL.Religion as a Social Determinant of Public Health. New York, NY: Oxford University Press; 2014. 19. Dupre ME, Franzese AT, Parrado EA. Religious attendance and mortality: implications for the black-white mortality crossover.Demography. 2006;43(1):141–164. 20. Rew L, Wong YJ. A systematic review of associations among religiosity/spirituality and adolescent health attitudes and behaviors.J Adolesc Health. 2006;38(4):433–442. 21. Wallace JM, Jr., Brown TN, Bachman JG, et al. Religion, race and abstinence from drug use among American adolescents. Ann Arbor, MI: University of Michigan; 2003. https://deepblue. lib.umich.edu/bitstream/handle/2027.42/137867/occ58.pdf? sequence=1&isAllowed=y . Accessed September 15, 2017. 22. Nonnemaker JM, McNeely CA, Blum RW. Public and private domains of religiosity and adolescent health risk behaviors: evidence from the National Longitudinal Study of Adolescent Health.Soc Sci Med. 2003;57(11):2049–2054. 23. Nonnemaker J, McNeely CA, Blum RW. Public and private domains of religiosity and adolescent smoking transitions.Soc Sci Med. 2006;62(12):3084–3095. 24. Maselko J, Hayward RD, Hanlon A, et al. Religious service attendance and major depression: a case of reverse causality? Am J Epidemiol. 2012;175(6):576–583. 25. VanderWeele TJ. On the promotion of humanflourishing. Proc Natl Acad Sci U S A. 2017;114(31):8148–8156. 26. VanderWeele TJ. Outcome-wide epidemiology. Epidemiology. 2017;28(3):399–402. 27. Field AE, Camargo CA, Jr., et al. Overweight, weight concerns, and bulimic behaviors among girls and boys.JAm Acad Child Adolesc Psychiatry. 1999;38(6):754–760.28. Solomon CG, Willett WC, Carey VJ, et al. A prospective study of pregravid determinants of gestational diabetes mellitus. JAMA. 1997;278(13):1078–1083. 29. VanderWeele TJ, Li S, Tsai AC, et al. Association between religious service attendance and lower suicide rates among US women.JAMA Psychiatry. 2016;73(8):845–851. 30. Pan A, Lucas M, Sun Q, et al. Bidirectional association between depression and type 2 diabetes mellitus in women. Arch Intern Med. 2010;170(21):1884–1891. 31. Ware JE, Snow KK, Kosinski M, et al.SF-36 Health Survey: Manual and Interpretation Guide. Boston, MA: Health Institute, New England Medical Center; 1993. 32. Ding P, VanderWeele TJ. Sensitivity analysis without assumptions.Epidemiology. 2016;27(3):368–377. 33. VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the E-value.Ann Intern Med. 2017; 167(4):268–274. 34. Resnick MD. Protective factors, resiliency, and healthy youth development.Adolesc Med. 2000;11(1):157–165. 35. Morgan A, Ziglio E. Revitalising the evidence base for public health: an assets model.Promot Educ. 2007;14(2 suppl):17–22. 36. Wong YJ, Rew L, Slaikeu KD. A systematic review of recent research on adolescent religiosity/spirituality and mental health.Issues Ment Health Nurs. 2006;27(2):161–183. 37. Weaver AJ, Flannelly KJ, Strock AL. A review of research on the effects of religion on adolescent tobacco use published between 1990 and 2003.Adolescence. 2005;40(160):761–776. 38. Yeung JW, Chan YC, Lee BL. Youth religiosity and substance use: a meta-analysis from 1995 to 2007.Psychol Rep. 2009; 105(1):255–266. 39. Davis DE, Worthington EL, Hook JN, et al. Research on religion/spirituality and forgiveness: a meta-analytic review. Psycholog Relig Spiritual. 2013;5(4):233–241. 40. Smith TB, McCullough ME, Poll J.Religiousness and depression: evidence for a main effect and the moderating influence of stressful life events.Psychol Bull. 2003;129(4):614–636. 41. Cotton S, Zebracki K, Rosenthal SL, et al. Religion/spirituality and adolescent health outcomes: a review.J Adolesc Health. 2006;38(4):472–480. 42. Rippentrop AE, Altmaier EM, Chen JJ, et al. The relationship between religion/spirituality and physical health, mental health, and pain in a chronic pain population.Pain. 2005; 116(3):311–321. 43. Hackney CH, Sanders GS. Religiosity and mental health: a meta-analysis of recent studies.J Sci Study Relig. 2003;42(1): 43–55. 44. US Department of Health and Human Services.Preventing Tobacco Use Among Youth and Young Adults: A Report of the Surgeon General. Atlanta, GA: US Department of Health and Human Services; 2012. https://www.ncbi.nlm.nih.gov/books/ NBK99237/pdf/Bookshelf_NBK99237.pdf . Accessed September 15, 2017. 45. Smith C. Theorizing religious effects among american adolescents.J Sci Study Relig. 2003;42(1):17–30. 46. Cotton S, McGrady ME, Rosenthal SL. Measurement of religiosity/spirituality in adolescent health outcomes research: trends and recommendations.J Relig Health. 2010;49(4): 414–444. 47. Li S, Okereke OI, Chang SC, et al. Religious service attendance and lower depression among Women-a Prospective Cohort Study.Ann Behav Med. 2016;50(6):876–884. 48. Bao WN, Whitbeck LB, Hoyt DR, et al. Perceived parental acceptance as a moderator of religious transmission among adolescent boys and girls.J Marriage Fam. 1999;61(2): 362–374. Am J Epidemiol.2018;187(11):2355–2364 2364Chen and VanderWeele Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article & Citation: Flor E, George F, Kibria BMG (2020) Changes in Mobility during COVID-19 as a Response to Government Imposed Restrictions: A Multiple Regression Analysis for the Top Five Populous U.S States. Int J Clin Biostat Biom 6:032. doi.org/10.23937/2469-5831/1510032 Accepted: November 12, 2020: Published: November 14, 2020 Copyright: © 2020 Flor E, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. DOI: 10.23937/2469-5831/1510032 • Page 1 of 8 • Changes in Mobility during COVID-19 as a Response to Government Imposed Restrictions: A Multiple Regression Analysis for the Top Five Populous U.S States Eitan Flor, Florence George and BM Golam Kibria * Department of Mathematics and Statistics, Florida International University, Miami, Florida, USA Abstract With the rise of the COVID-19 pandemic, several chang - es occurred in the lifestyle and habits of human life. These include various voluntary and mandatory governmental restrictions, limiting social interaction by encompassing so- cial distancing, travel limitations, social gatherings, person - al mobility, as well as closures and reduced capacities in sectors such as retail, restaurants, and the entertainment industry. The purpose of the restrictions ideally intended to reduce the transmission rate of COVID-19 by limiting the overall movement of individuals, thus preventing the spread of the virus. As a result, this study seeks to identify whether regulatory restrictions posed an overall significant impact on mobility in the United States by conducting multiple linear regression analysis studies on predicting average statewide mobility (in kilometers) based on the factors of daily cas- es, daily deaths, and imposed governmental restrictions. By identifying the significant changes in mobility across the continental United States, a baseline can be established in order to evaluate upon the efficacy of government-imposed restrictions and extend to further implementation of policies to minimize mobility and disease spread simultaneously. Additionally, with increasing concerns about a second wave or outbreak of COVID-19, this study will seek to establish inferences to re-evaluate and improve upon the existing regulations, control measures, and disease mitigation tech- niques used to combat the spread of COVID-19 and the potential for any other similar diseases or epidemics in the future. Keywords COVID-19, Coronavirus, Linear regression model, Mobility Index OriGinal articlE *Corresponding author: BM Golam Kibria, Department of Mathematics and Statistics, Florida International University, Miami, Florida, USA Check for updates Introduction The COVID-19 disease (Coronavirus 2019) is caused by and attributable to the virus known as Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). The first instance of the disease introduced itself in late De- cember 2019 in the Hubei province of China and ever since has drastically proliferated to reach a pandem - ic status accruing over 10 million cases and 500,000 deaths globally as of present-day. The United States alone accounts for over 2.5 million of those cases and over 125,000 deaths [1]. While the origins of COVID-19 are unknown, it is presumed to have ties with the form of a human to animal (zoonotic) interaction; possibly identified as chiropteran origin [2]. COVID-19 continues to pose a severe public health catastrophe with far more significant consequences, like the ability to damage and halt economic growth permanently. Juxtaposed with a second wave or outbreak on the near horizon, the motive of this study was to use publicly available and anonymous cell phone GPS data as a direct measure of human mobility for the period of March 10, 2020, to May 28, 2020, to form a basis on evaluating the govern - mental restrictions that took place in the United States in mid to late March 2020. The scope of this study was narrowed down to the five most populous states (Cal - ifornia, Florida, New York, Pennsylvania, and Texas) as we believe these states would comprise the majority influence in decision making and changes in future reg - ulations for imposed restrictions for individual states as well on the national scale. International Journal of Clinical Biostatistics and Biometrics Flor et al. Int J Clin Biostat Biom 2020, 6:032 ISSN: 2469-5831 Volume 6 | Issue 2 Open Access Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 2 of 8 • Unlike defined physical processes, the evaluation of human psychological behavior in this context spanning human responses to imposed government mandates is challenging to analyze due to a lack of a quantifiable metric. For this reason, it was hypothesized that the ide- al metric to investigate in terms of evaluating the over - all response of such restrictions would be in the form of a distance measure or Mobility Index as defined in the context of this study. The mobility data originates from the anonymous and commercial collection of cel- lular GPS data and is publicly available from Descartes Labs. As a result of GPS data collection, several metrics or mobility measures can be defined such as the maxi- mum mobility from the initial location of an individual for a given day (M max ), bounding mobility (M bb) and the convex hull (M ch) all in terms of distance traveled [3]. To investigate the validity of imposed restrictions, the de- cided metric of interest to represent as the Mobility In- dex was the maximum distance traveled on a given day (M max ), as the fundamental notion behind any COVID-19 related restriction is to minimize the distance people typically travel. Due to anonymity restrictions of per- sonal data, the direct data values of maximum distance are not accessible; thus, the median of the maximum distance produced over a random set of individuals that reside in the same county is used instead; hence for the capital M in the naming convention of all mobility-re - lated metrics. Using the median values for mobility will also provide a more accurate depiction of the actual val- ues as compensation for abnormalities taking place in errors that can manifest as GPS malfunctioning, GPS in - consistencies, and the accidental capture of abnormally high or low traveling distances in individuals ( Figure 1).To the best of our knowledge, the published litera- ture on the fitting of Mobility data for COVID-19 is not available yet. The objective of this paper is to fit several regression models on mobility index of the five states and investigate the effect of Government restriction on mobility index during the COVID-19 pandemic. In addi - tion, we have developed a combined regression model using data of the top five populous states. The organization of this paper is as follows: The data sources, data descriptions, data cleaning and processing are given in section 2. The statistical models and data analysis are provided in section 3. Finally, some con - cluding remarks are added in section 4. Materials and Data The primary source of data used in this experiment to collect mobility information was made publicly avail - able by Descartes Labs, a founded company from re - searchers and scientists from Los Alamos National Lab - oratory, and focuses on large-scale computing, artificial intelligence, and satellite imagery and create solutions targeting Data Modeling and Analytics [4]. The hosting and online access of the data utilized www.data.world, a Public Benefit Corporation that provides a cloud-na - tive solution for hosting publicly available and open- source data repositories along with API and software integration for tools such as Python, R, SQL, and Excel. To visit and explore the direct data source, navigate to https://github.com/descarteslabs/DL-COVID-19 , and for more information on Descartes Labs, please visit https://www.descarteslabs.com/company/#about . Ad- ditional secondary data sources were used to compile detailed information relating to the statewide data of Figure 1: Visualization of anonymous mobility metrics obtained from GPS data and Mobility Index illustrated as M max as from Figure 1 of Warren M, et al. [ 3]. ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 3 of 8 • daily confirmed cases and deaths. These secondary data sources originated from the Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE) [1], Worldometer [5], and lastly from the Institute of Health Metrics and Evaluation (IHME) [6] and were re - spectively combined into further columns of Cases and Deaths for each date in the time period between March 10, 2020, and May 28, 2020. The overall process of extracting the primary data (State, Date, and Mobility) involved utilizing SQL que- ries to filter out the Descartes Labs data set in group- ings for the individual states of California, Florida, New York, Pennsylvania, and Texas. Additionally, another partition was applied on the date in the form of a range in order to maintain consistent statewide data from March 10, 2020, to May 28, 2020. Once completed and initially filtered under the above two constraints, an API call was created from www.data.world to be read and manipulated further via the Python programming language (Python 3.8.1). Likewise, a Python script was created to preprocess the dataset by managing and or- ganizing the filtered data into data frames, which will be combined with the data from the secondary sources and represented as the remaining variables (Daily Cases and Daily Deaths). Additionally, to construct a variable for the effect or contribution of government-imposed restrictions, a logical comparison was implemented to make a binary coded variable that would take assigned values of 0 (No Restriction present) or 1 (Restriction present) based upon the respected date values from the data found in the Tracking Involuntary Government Restrictions (TIGR) Dataset [7]. The Cases and Deaths data were compiled from JHU CSSE [1], Worldometer [5], and IHME [6] and integrated into the same Python script, adding on to the initial data frame. At this point, the dataset for analysis is completed and requires trans- formation for analysis and migration to R. As a result, a random subset of the data was taken for each of the individual states (20%) via Python’s random sample() function and flattened as averages to yield distinctive data records for each date in the range as depicted be- low in Figure 2. Finally, a conversion of the data's for- mat (DataFrame to CSV) took place, which enabled the migration directly to RStudio (Version 3.6.3) for per- Figure 2: Final stage of data preprocessing before appending external data. ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 4 of 8 • Figure 3: Initial data format and manipulation process leading to the reduced dat aset. Table 1: Example of final data (random 5% subset from 395 records total) with appended external data after preprocessing and manipulation stages. Date Mobility Index RestrictionCasesDeaths State 2020-04-23 3.094429 11945104 California 2020-03-11 15.197200 050 Texas 2020-04-09 3.085364 1112848 Florida 2020-03-13 19.293300 0170 Texas 2020-04-21 8.616700 177427 Texas 2020-04-10 3.462533 1114248 Florida 2020-03-25 5.687200 15103 Florida 2020-04-22 1.142000 15713661 New York 2020-03-25 0.541833 12764 Pennsylvania 2020-04-05 0.064750 1138828 California 2020-04-15 6.579100 199626 Texas 2020-03-10 5.646600 050 Pennsylvania 2020-04-23 6.227700 1107260 Florida 2020-04-24 10.093500 177727 Texas 2020-04-12 1.331300 143524 Texas 2020-03-14 11.411000 0263 Florida 2020-03-17 3.032273 0330 Pennsylvania file, dpylr [ 11] for data manipulation, lindia [12] for cre - ating regression diagnostic plots along with verifying lin - ear model assumptions, and knitr [13] for printing and exporting the results of our analyses. A sample of the forming statistical modeling and analysis. A variety of packages were utilized in RStudio namely broom [8] for summarizing the model results, ggplot2 [9] as a graphi - cal visualization tool, readr [10] to process the CSV data ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 5 of 8 • finalized data after the preprocessing and manipulation stage is depicted in below in Table 1. The above Figure 2 and Figure 3 serve as an example of the data processing stages for an individual state for graphical purposes. It is important to note that this pro- cess is repeated for each of the states (California, Flori - da, New York, Pennsylvania, and Texas). Once the data processing stage was completed, external data of Cases, Deaths, and Restriction were appended to the data for final use. An example of the finalized data is shown be- low in Table 1 . Statistical Models and Data Analysis To analyze the data, we consider the following multi - ple linear regression models: 0 11 2 2 3 3 Y XX X ββ β β ε=++ + + (3.1) where Y = Mobility Index, 1X = Restriction, 2X = Cases and 3X = Deaths. We assume that ε has a nor- mal distribution with zero mean and unit variance. We also assume that the regressors, 12, XX and 3X are in - dependent. The five implemented models for the five individual states follow equation 3.1. Before proceeding in performing analysis on any of the models, baseline evaluations were conducted on each of the models and included Shapiro-Wilk normality tests, calculations of Variance Inflation Factors, as well as exploratory plots of Standardized residuals to investi- gate the validity of the normality assumptions and ver - ify multicollinearity was not present in the model itself. Based on these results, each state model, except for Florida, underwent a square root transformation in the dependent variable (Mobility Index) to satisfy the Shap - iro-Wilk normality test or removal of an insignificant re - gressor(s). A recurring theme displayed that the Deaths regressor proved statistically insignificant in several of the models (under α = 0.05), and hence, was removed from the model where necessary. As a result, all model descriptions in the following sections will refer to the revised versions of the regression models. Individual state models In the final fitted multiple linear regression models of California, New York, and Texas the associated re - gressors of Restriction and Cases show statistical signif - icance at α = 0.05. It is important to note that the ad- justed R-squared values (0.42, 0.59, 0.34) demonstrate a weak, moderate, and weak fit respectively. For the re - maining final fitted multiple linear regression models of Florida and Pennsylvania, all the associated regressors of Restriction, Cases, and Deaths show statistical signif - icance at α = 0.05. Similarly, it is important to note that the adjusted R-squared value (0.48, 0.35) demonstrate a moderate and weak fit respectively for the models. Table 2: Final individual state linear models (summary output). State Model Summary CoefficientsStandard Error t StatP valueR 2 adj California 0.42 (Intercept) 2.08422130.125291 16.6350440.000000 Restriction -1.2576390.166272 -7.5637540.000000 Cases 0.0002970.000074 4.0404770.000127 Florida 0.48 (Intercept) 10.2528860.715398 14.3317320.000000 Restriction -3.7013720.916664 -4.0378730.000129 Cases -0.0037100.000747 -4.9668050.000004 Deaths 0.0444320.010651 4.1714970.000081 New York 0.59 (Intercept) 2.2477810.178481 12.5939650.000000 Restriction -0.9542150.223721 -4.2652050.000058 Cases -0.0002740.000115 -2.3853300.019590 Pennsylvania 0.35 (Intercept) 2.2477810.178481 12.5939650.000000 Restriction -0.9542150.223721 -4.2652050.000058 Cases -0.0002740.000115 -2.3853300.019590 Deaths 0.0038970.000914 4.2613070.000058 Texas 0.34 (Intercept) 13.7385010.932479 14.7333100.000000 Restriction -7.6856141.197682 -6.4170760.000000 Cases 0.0026270.000804 3.2685790.001625 ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 6 of 8 • R-squared value (0.69) demonstrates a moderate fit for the model. Additionally, government restriction proves to be a significant factor for Mobility Index in both the individual state models and combined model. The importance of the imposed government restric - tion factor may possibly benefit and stem from the in - clusion of two main effects: The direct restrictions on personal mobility and the restrictions in place on social distancing. The CDC recommends social distancing to control COVID-19 spread and government restrictions are one of the most influential factors in controlling the implementation of social distancing standards which in effect explains the limited mobility of individuals. The performance of the categorical state factors can be partly explained as Texas and Florida have an overall higher mobility compared to the other three states of California, New York, and Pennsylvania. Both Texas and Florida have shown cases spiking only recently, while An important observation from the results as seen in Table 2 are the consistent negative coefficient values for the Restriction variable across all individual state mod - els which serves as an indicator that a negative linear re - lationship exists between the factors of Mobility Index and governmental restrictions. This finding supports and extends on the primary basis that government re - strictions play a critical role in reducing the typical mo- bility of individuals. Combined state model (All states) In this section, we will investigate a combined mod - el using the State as a categorical variable. The regres - sion analyses for all states are presented in Table 3. It is observed that the associated regressors of restriction, cases and in part components of the categorical state factor (Florida and Texas) show statistical significance at α = 0.05. It is important to note that the adjusted Figure 4: Final combined state linear model (Normal QQ Plot).Table 3: Final combined state linear model (summary output). Term Model Summary (R 2 adj = 0.69) Coefficients Standard Error t StatP value (Intercept) 2.16617390.0942412 22.98541780.0000000 Restriction -0.76297980.0888960 -8.58283270.0000000 Cases -0.00010790.0000170 -6.33826700.0000000 Florida 0.92141050.0814441 11.31340520.0000000 New York 0.17672990.0989717 1.78566010.0749352 Pennsylvania 0.02087300.0808681 0.25811190.7964574 Texas 1.51917840.0811243 18.72655270.0000000 ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 7 of 8 • results. The issue could be further exacerbated due to the dependencies on a variety of factors such as the varying duration of the incubation period for an individ - ual, age of the individual, and any pre-existing comor- bidities [14]. Additionally, the models created for this project involved at most four regressors to estimate the Mobility Index value, and in further research, additional regressor variables can be explored to better the fit of the model and explain more in the overall variability. It is also important to note that each state has varying population densities and geographic sizes; as a result, the descriptive statistics and current data for these individual states may not be the best for direct use or judgment in analysis and a method for standardization or normalization of the Mobility Index distance can im- prove upon the validity of the models and may remove the need of transformation in the model entirely. An- other consideration involves the decreased or decay - ing effect of restrictions as time progresses. Due to the implementation of policies and related logistics to poli - cymaking, restrictions have not been officially deemed as lifted and are instead related to an “easing” state, in which case it would be beneficial to consider this effect. Acknowledgements Authors are grateful to the anonymous referees and the Editor-in-Chief for their valuable comments and suggestions, which certainly improved the quality and presentation of the paper. We wish to dedicate this pa- per to all of those who have been lost or greatly affect- ed by the COVID-19 pandemic . References 1. Dong E, Du H, Gardner L (2020) An interactive web-based New York and Pennsylvania had a significantly higher number of cases from the very beginning. On the other hand, California has many IT companies and Hollywood film industries for which employees are either transi - tioning to work remotely, already working from home, or the industry closed completely due to COVID-19. It is evident from Figures 4 and 5 that the normality and constant variance assumptions have been met to some extent. Summary and Concluding Remarks Upon constructing and identifying the ideal model out of the initial and final models for each of the individ - ual states of California, Florida, New York, Pennsylvania, and Texas, it is clear that the overall trend for mobility is decreasing as government restrictions came to imple - mentation; hindering on social gatherings, personal mo- bility, restaurant capacity/usage, as well as in the gener - al business domain whether it be entertainment or the retail industry. The model that demonstrated the most success was the Combined State model which displays a relatively high adjusted R-squared value compared to all of the individual models. Some of the limitations and challenges identified during this study were the repeated insignificance of the deaths regressor, the lack of additional regressor variables, and biased comparison of statewide mobility without normalization. The possible cause in the pattern of the insignificant deaths regressor could have an asso - ciation with a delay in the number of confirmed deaths from the instance of infection to the reporting stage as no exact method or central guidelines have been estab- lished in reporting death data, leading to inconsistent Figure 5: Final combined state linear model (Residuals vs. Fitted Plot). ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 DOI: 10.23937/2469-5831/1510032 • Page 8 of 8 • 9. H Wickham (2016) ggplot2: Elegant graphics for data anal - ysis. Springer-Verlag, New York. 10. Hadley Wickham, Jim Hester, Romain Francois, R Core Team, R Studio, et al. (2018) readr: Read rectangular text data. 11. Hadley Wickham, Romain François, Lionel Henry, Kirill Müller, R Studio (2020) dplyr: A grammar of data manip - ulation. 12. Yeuk Yu Lee, Samuel Ventura (2017) lindia: Automated lin- ear regression diagnostic. 13. Xie Yihui (2014) Knitr: A comprehensive tool for reproduc - ible research in R. In: Victoria Stodden, Friedrich Leisch, Roger D Peng, Implementing Reproducible Computational Research. Chapman & Hall/CRC. 14. Teodoro Alamo, DG Reina, Pablo Millán (2020) Data-driven methods to monitor, model, forecast and control Covid-19 pandemic: Leveraging data science, epidemiology and control theory. dashboard to track COVID-19 in real time. Lancet Infect Dis. 2. Lu R, Zhao X, Li J, Niu P, Yang B, et al. (2020) Genomic characterisation and epidemiology of 2019 novel coronavi - rus: Implications for virus origins and receptor binding. The Lancet 395: P565-P574. 3. Warren Michael S, Skillman Samuel W (2020) Mobility changes in response to COVID-19. arXiv. 4. Hao Zhu, Thomas Travison, Timothy Tsai, Will Beasley, Yihui Xie, et al. (2020) kableExtra: Construct complex table with 'kable' and pipe syntax. 5. https://www.worldometers.info/ 6. Institute for Health Metrics and Evaluation (IHME) (2020) COVID-19 Projections. University of Washington, Seattle, WA. 7. Rex W Douglass (2020) Crowd-sourced COVID-19 dataset Tracking Involuntary Government Restrictions (TIGR). 8. David Robinson, Alex Hayes, Simon Couch, Indrajeet Patil, Derek Chiu, et al. (2020) broom: Convert statistical objects into tidy tibbles. ISSN: 2469-5831 Flor et al. Int J Clin Biostat Biom 2020, 6:032 Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article & DOI: 10.23937/2469-5831/1510022 Citation: Etikan I, Babatope O, Bala K, İlgi S (2019) Child Mortality: A Comparative Study of Some Developing Countries in the World. Int J Clin Biostat Biom 5:022. doi.org/10.23937/2469-5831/1510022 Accepted: September 28, 2019: Published: September 30, 2019 Copyright: © 2019 Etikan I, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. • Page 1 of 6 • Abstract Background: According to the United Nations Sustainable Development Goals, countries all over the world are expect- ed to create a healthy living environment for the populace most especially the vulnerable population of which children are examples. Despite the reducing trend in the under-five mortality rate in some developing nations, some nations still have a high record of under-five mortality. Methods: This study adopted the use of a One-way Analy- sis of Variance (ANOVA) to evaluate any significant differ- ence in the under-five mortality rate of five major developing countries namely Brazil, Bangladesh, Turkey, South Africa and Nigeria owing to their similarity in economy potentials and population. The significant level of 0.05 was consid- ered for the statistical test. The under-five mortality rate figures were secondary data obtained from the UNICEF database for the period between 1990 through 2017. Results: The test result shown that there was a significant difference in the under-five mortality rates of the five coun- tries under considerations (p < 0.05). A post-hoc analysis us- ing Tukey’s method revealed significant pairwise tests com- parison between Bangladesh VS. Brazil, Bangladesh VS. Nigeria, Bangladesh VS. Turkey, Brazil VS. Nigeria, Brazil VS. South Africa, Nigeria VS. South Africa, Nigeria VS. Tur- key, and South Africa VS. Turkey. Conclusion: Nigeria has the highest under-five mortality rate (163.17 ± 38.42) while Brazil has the lowest under-five mortality rate (21.92 ± 15.27). Nigeria’s average under-five mortality rate is more than twice the average rate of Brazil and almost twice of Turkey’s under-five mortality rate within the period under review (1990-2017). Child Mortality: A Comparative Study of Some Developing Countries in the World İlker Etikan *, Ogunjesa Babatope, Kabiru Bala and Savaş İlgi Department of Biostatistics, Near East University, Cyprus *Corresponding author: İlker Etikan, Department of Biostatistics, Faculty of Medicine, Near East University; Nicosia- TRNC, Cyprus, Tel: +90-392-223-64-64 RESEaRch aRticlE Check for updates one of the important constituents of the vulnerable population. According to the 2017 World Bank report on population growth, there are about 1,953 billion people between ages 0-14 years accounting for 26% of the global population [1]. Even though there is a report- ed demographic shift that is increasing the aged 65 + cohort, the ages 0-14 years cohort is facing an increas- ing risk with an estimated decline to 21% of the global population by 2050. Child mortality is a central case study among many researchers that revolves around children health outcome. Hence, different national and internation- al stakeholders, as well as policymakers, continual- ly strive to see to the global reduction of childhood deaths worldwide. It is also in this vein the United Nations (UN) developed the United Nations Develop- ment Goals (MDGs) which was later revamped into the Sustainable Development Goals with the number three (3) agenda still prominently dedicated to im- proving maternal and child health [2 ]. Though consid- erable milestone progress has been made in regards to reducing child mortality rate globally, however, the rate is still higher in Southern Asia and Sub-Saharan Africa with an average of 4 dead cases among every 5 deaths of children before their 5 th year birthday [3 ,4 ]. There are several types of indices that are used to describe child mortality. They include prenatal, perina- tal, neonatal, infancy and under-5 child mortality [5]. Prenatal mortality refers to the death of a child before delivery. The perinatal mortality is defined as the death of a child that occurs within the first week of delivery. The neonatal death describes the death of a child be- fore the 28 days after delivery. The infant mortality rate Introduction The study on the vulnerable population has contin - ued to remain an important research theme in both health and social economic researches. Children remain Etikan et al. Int J Clin Biostat Biom 2019, 5:022 ISSN: 2469-5831 International Journal of Clinical Biostatistics and Biometrics Volume 5 | Issue 2 Open Access Etikan et al. Int J Clin Biostat Biom 2019, 5:022 DOI: 10.23937/2469-5831/1510022 • Page 2 of 6 • ran Region still recorded the highest cases of under-five mortality with a record death of 76 per 1000 live births [8]. Risk Factors of Child Mortality There are several factors that are responsible for child mortality. Ria, et al. [9 ] in their study on caus- es and contributors to infant mortality in North India identified social and verbal autopsy drivers as a major contributor to child death. The social autopsy borders on social, health models and behavioral agents. These include factors stemming from transportations, living conditions of people as well as health systems that are operational. According to Suwal [10], malaria, malnutrition,diarrhea, and respiratory infections are major risk factors responsible for children fatality. Regarding neonatal death, Adeyele and Ofoegbu [11] posited that genetically propelled malfunctions, mea- sure, and level of antennal care support, and quality of postpartum treatments play a role in the morbidity and mortality of children. Other factors responsible for child mortality include maternal literacy level, poverty, early marriage, place of abode, pneumonia, congenital abnormalities, preterm birth complica- tions, nutritional and breastfeeding practices, access to medic care support, food insecurity, early preg- nancy, poor hygiene practices, poor water access and so on [12- 14]. Methodology Data Source The data for the study was extracted from the Unit- ed Nations Children Fund (UNICEF) under-five mortality rate datasets portal [15]. The under-five mortality rate for Nigeria, India, Turkey, South Africa and Brazil be- is used to describe the likelihood of a child’s death per 1000 live birth between the time of birth and exactly a year old. The under-five year child mortality is the most common form and widely used metric to define child mortality [6]. The Under-five year child mortality is ba- sically defined as the probability of a child dying before reaching the 5 th year birthday anniversary and is usually expressed per 1000 live birth. Global Overview of Child Mortality Rate According to the UN Inter-agency Group for Child Mortality Estimation [7], an estimated 6.3 million deaths occurred among children and young adolescents. In this recorded death, 2.5 million dead cases were recorded among newborn deliveries; 1.6 million deaths occurred among age bracket 1-11 months; 1.3 million deaths oc- curred among age bracket 1-4 years; 600 deaths from age bracket 5-9 years and children between age 10-14 years recorded 400,000 thousand deaths. The report indicated that the world has indeed improved on its re - cord of child mortality compared to past records. About 5.5 million children under the age of five-years-old died in 2017 in contrast to about 12.6 million dead cases in this age cohort in 1990. This translated to a rate reduc- tion from 93 deaths per 1000 live births in 1990 to about 39 deaths per 1000 live births in 2017 Figure 1. Geographically in 2017, the Northern America and Europe region performs better on their under-five mortality rate record by recording the lowest rate of 6 deaths per 1000 live births in 2017 [7]. The Central and Southern Asia recorded 43 deaths per 1000 live births while the Eastern and South-Eastern Asia had 16 deaths per 1000 live births. Latin America and the Caribbean reported 18 deaths per 1000 live births while Oceania recorded 23 deaths per 1000 live births. The Sub-Saha - Mortality rates 1990 1995 2000 2005 2010 2017 1990 1995 2000 2005 2010 2017 Deaths per 1,000 100 75 50 250Under-five mortality rate Neonatal mortality rate Mortality rate among children aged 5-14 years 93 3912.6 11.3 9.88.3 7.0 5.4 15 71.71.6 1.41.2 1.1 0.9 37 185.0 4.5 4.03.5 3.1 2.5 Deaths (in millions) Number of deaths Under-five deaths Neonatal deathsDeaths among children aged 5-14 years Figure 1: Child mortality decline (1990-2017) Source: [ 7]. ISSN: 2469-5831 Etikan et al. Int J Clin Biostat Biom 2019, 5:022 DOI: 10.23937/2469-5831/1510022 • Page 3 of 6 • a group classification deviation (α i) and the random ef- fect (e ij). The ANOVA makes a comparison of the varia- tion between samples (Sum of squares for groups: SSB) relative to the variation within samples (sum of squares for Error: SSE). Mathematically, the equation for the one-way ANO - VA can be stated as follows: ( ) () () 22 2 = 1 = 1 = 1 = n nn i ii yi y yi y yi yi − −+ − ∑ ∑∑  (2) The equ (2) can be written as: The total sum of Squares (SSTO) = Regression sum of squares (SSR) + Error sum of squares(SSE) The following degrees of freedom namely n-1, 1, and n-2 are associated with the SSTO, SSR, and SSE respec - tively. Where n = sample size. In general, the Table 1 below gives a summary of the One-way ANOVA F-test The p-value is derived when the comparison of the F Cal is made with the F-distribution table with their ap- propriate degrees of freedom. If the null hypothesis is rejected, a post-hoc test is required to determine a pair of groups that are responsible for the test significance. The Tukey’s method, Bonferroni’s and Scheffe’s tests are some of the common ANOVA post-hoc tests [ 19]. Result According to Table 2 above, the mean under-five mortality rate of Bangladesh is 77.995 while for Brazil is 31.921. Nigeria have the highest a mean under-five tween 1991 through 2017 was used in this study. Method The One-way Analysis of Variance (ANOVA) statisti- cal methodology was adopted in this study. InVivoStat statistical software which was based on an r-program - ming language was used in the analysis of this study. ANOVA is multiple comparisons and parametric method used to test a mean effect difference for more than two groups. It could be considered as an advanced exten- sion of the t-test independent sample distribution. Orig- inally invented by Sir Ronald A. Fisher; a versatile and renowned Statistician to investigate treatment effects in agricultural experiments, the method is now widely used in Economics, Medicine and social sciences field [16]. This statistical methodology assumes that obser- vations sets under consideration are independent, nor- mally distributed and the condition of homogeneity of variance (equal variance) is satisfied [17, 18]. This meth- od generally seeks to test the null hypothesis that: H 0: µ1 = µ 2 = µ3 =……..= µk versus the alternative hypothesis that: H 1: k ∃ 1 ≤ i, l ≤ k: µ i ≠ µ l ( at least one of the pair mean is not equal) The linear model for the One-way ANOVA is given as follows: X = i ij ij e µα++ (1) The “I” entails the group membership while the “j” subscript denotes class membership (from the value of 1 to n). The Xij is equal to three distinct components namely; the overall mean of the experimental units( µ), Table 1: Summary of one-way ANOVA test. Sources of VariationDFSum of Squares (SS) Mean Squares (MS)F Regression 1 ( ) 2 1 = n i SSR yi y = − ∑  = 1 SSR MSR = MSR Fcal MSE Residual Error n-2 ( ) 2 1 = n i SSE yi yi = − ∑  = 2 SSE MSE n− Total n-1 () 2 1 = n i SSTO yi y = − ∑ The F Cal is the test statistic. Table 2: Descriptive statistics of the countries under-five mortality rate. MeanNStd Dev Std error Categorization Factor levels Bangladesh 77.992834.42 6.5 Brazil 31.922815.27 2.89 Nigeria 163.172838.95 7.36 South Africa 63.912815.17 2.87 Turkey 35.112819.1 3.61 ISSN: 2469-5831 Etikan et al. Int J Clin Biostat Biom 2019, 5:022 DOI: 10.23937/2469-5831/1510022 • Page 4 of 6 • model is significant which entails that the means of the under-five mortality of at least a pair of the countries under evaluation are significantly different. Thus, a post-hoc test using the Tukey’s Method will be used to further evaluate this significant difference as a result of not accepting the null hypothesis of equal means the under-five mortality rate of the countries. The post-hoc comparison indicated that the means of the under-five mortality rate between Bangladesh versus Brazil; Bangladesh versus Nigeria; Bangladesh mortality rate of 163.171, South Africa have 63.913 and Turkey have mean under-five mortality rate of 35.107 Figure 2. The normality test using the Shapiro Wilk statistic shown that the under-five mortality rate of the coun - tries is normally distributed (p > 0.05). This normality property of the data can be visualized as indicated in Figure 3 below. Table 3 gave the F statistic of the ANOVA distribu – tion. Since the p-value < 0.0001, it show that the ANOVA Under-Five Mortality Rate BANGLADESH BRAZILNIGERIA CountrySOUTH AFRICA TURKEY 200 150 100 50 Figure 2: Box-plot of the countries under-five mortality rate. Normal probability plot Theoretical Quantiles Sample Quantiles -2 -1 0 1 2 50 0 -50 Figure 3: Probability plot of the under-five mortality rate. ISSN: 2469-5831 Etikan et al. Int J Clin Biostat Biom 2019, 5:022 DOI: 10.23937/2469-5831/1510022 • Page 5 of 6 • 6. Ahmad OB, Lopez AD, Inoue M (2000) The decline in child mortality: A reappraisal. Bulletin of the World Health Orga- nization 78: 1175-1191. 7. United Nations Inter-Agency Group for Child Mortality Esti- mation [UN IGME] (2018) Levels & trends in child mortality. 8. Bereka SG, Habtewold FG, Nebi TD (2017) Under-five mortality of children and its determinants in Ethiopian So- mali regional state, Eastern Ethiopia. Health Science Jour- nal 11. 9. Rai SK, Kant S, Srivastava R, Gupta P, Misra P, et al. (2017) Causes of and contributors to infant mortality in a rural community of North India: Evidence from verbal and social autopsy. BMJ Open 7. 10. Suwal JV (2001) The main determinants of infant mortality in Nepal. Social Science and Medicine Journal 53: 1667- 1681. 11. Adeyele IT, Ofoegbu DI (2013) Infant and child mortality in Nigeria: An impact analysis. International Journal of Eco- nomics Practices and Theories 3: 122-132. 12. Dawit G Ayele, Temesgen T Zewotir, Hemry Mwambi (2017) Survival analysis of under-five mortality using cox and frailty models in Ethiopia. Journal of Health, Population and Nutrition 36. 13. Adewemimo Adeyinka, Henry D Kalter, Jamie Perin, Alain K Koffi, John Quinley, et al. (2017) Direct estimates of cause-specific mortality fractions and rates of under-five deaths in the northern and southern regions of Nigeria by verbal autopsy interview. PLoS One 12: e0178129. 14. Akinyemi JO, Bamgboye EA, Ayeni O (2015) Trends in neonatal mortality in Nigeria and effects of bio-demograph - ic and maternal characteristics. BMC Pediatrics 15: 36. 15. https://data.unicef.org/country/bgd 16. Aczel AD (1989) Complete business statistics, (7 th edn), Mc Grawill Irwin, USA. 17. Yoosun Jamie Kim, Cribbie Robert (2018) ANOVA and the variance homogeneity assumption: Exploring a better gatekeeper. British Journal of Mathematical and Statistical Psychology 71: 1-12. versus Turkey; Brazil versus Nigeria; Brazil versus South Africa; Nigeria versus South Africa; Nigeria versus Tur- key; and South Africa versus Turkey were all statistically significant (p < 0.05) [ 20]. Conclusion It can, therefore, be inferred that among the five countries under consideration, Nigeria has the highest under-five mortality rate (163.17 ± 38.42) followed by Bangladesh (77.99 ± 34.42). However, Brazil has the lowest under-five mortality rate (21.92 ± 15.27) close- ly followed by Turkey (35.12 ± 19.10). Nigeria’s average under-five mortality rate between 1990 through 2017 is more than twice the average rate of Brazil and almost twice of Turkey’s under-five mortality rate within the period under review. Therefore, there is a need for the Nigeria government to improve the country’s health care systems by implementing policies and regulations to cater for maternal and child health. This improvement is urgently needed if the country will be able to fulfill the United Nations Sustainable Development Goals in the area of child health by the year 2030. References 1. The World Bank (2018) A changing world population. 2. https://www.un.org/sustainabledevelopment/sustain- able-development-goals 3. Kazembe L, Clarke A, Kandala N (2012) Childhood mor- tality in sub-Saharan Africa: Cross-sectional insight into small-scale geographical inequalities from census data. BMJ Open 2. 4. United Nations Children’s Fund (2018). Under-five mortal- ity. 5. Ali Kazemi Karyani, Zhila Kazemi, Faramarz Shaahmadi, Zohreh Arefi, Zahra Meshkani (2015) The main determi- nants of under 5 mortality rate (U5MR) in OECD countries: A cross-sectional study. Int J Pediatr 3: 14. Table 3: ANOVA table. Sums of squares Degrees of freedom Mean square F-value p-value Country 317843.7 479460.92112.56< 0.0001 Residuals 95304.83 135705.962 Difference Lower 95% CI Upper 95% CI Std error p-value Comparison Bangladesh vs. Brazil 46.0732.03 60.12 7.10.0001 *** Bangladesh vs. Nigeria -85.18-99.22 -71.13 7.10.0001 *** Bangladesh vs. South Africa 14.080.04 28.13 7.10.2799 Bangladesh vs. Turkey 42.8928.85 56.93 7.10.0001 *** Brazil vs. Nigeria -131.25-145.29 -117.21 7.10.0001 *** Brazil vs. South Africa -31.99-46.04 -17.95 7.1 0.0001 *** Brazil vs. Turkey -3.19-17.23 10.86 7.10.9915 Nigeria vs. South Africa 99.2685.21 113.3 7.1010.0001 *** Nigeria vs. Turkey 128.06114.02 142.11 7.10.0001 *** South Africa vs. Turkey 28.8114.76 42.85 7.10.0008 ** Note: *p < .05; **p < .01; ***p < .001 indicates a statistical significance. ISSN: 2469-5831 Etikan et al. Int J Clin Biostat Biom 2019, 5:022 DOI: 10.23937/2469-5831/1510022 • Page 6 of 6 • 20. MO Jaiyeola, SO Oyamakin, JO Akinyemi, SA Adebowale, AU Chukwu, et al. (2016) Assessing infant mortality in Ni- geria using artificial neural network and logistic regression models. British Journal of Mathematics & Computer Sci- ence 19: 1-14. 18. Cochran WG (1947) Some consequences when the as- sumptions for the analysis of variance are not satisfied. Biometrics 3: 22-38. 19. Eva Ostertagová, Oskar Ostertag (2013) Methodology and application of one-way ANOVA. American Journal of Me- chanical Engineering 1: 256-261. ISSN: 2469-5831 Etikan et al. Int J Clin Biostat Biom 2019, 5:022 Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article & R E S E A R C H A R T I C L E Open Access Socio-economic inequality and HIV in South Africa Njeri Wabiri 1*and Negussie Taffa 2 Abstract Background: The linkage between the socio-economic inequality and HIV outcomes was analysed using data from a population-based household survey that employed multistage-stratified sampling. The goal is to help refocus at- tention on how HIV is linked to inequalities. Methods: A socio-economic index (SEI) score, derived using Multiple Correspondence Analysis of measures of own- ership of durable assets, was used to generate three SEI groups: Low (poorest), Middle, and Upper (no so poor). Dis- tribution of HIV outcomes (i.e. HIV prevalence, access to HIV/AIDS information, level of stigma towards HIV/AIDS, perceived HIV risk and sexual behaviour) across the SEI groups, and other background characteristics was assessed using weighted data. Univariate and multivariate logistic regression was used to assess the covariates of the HIV outcomes across the socio-economic groups. The study sample include 14,384 adults 15 years and older. Results: More women (57.5%) than men (42.3%) were found in the poor SEI [P<0.001]. HIV prevalence was highest among the poor (20.8%) followed by those in the middle (15.9%) and those in the upper SEI (4.6%) [P<0.001]. It was also highest among women compared to men (19.7% versus 11.4% respectively) and among black Africans (20.2%) compared to other races [P<0.001]. Individuals in the upper SEI reported higher frequency of HIV testing (59.3%) compared to the low SEI (47.7%) [P< 0.001]. Only 20.5% of those in poor SEI had “good access to HIV/AIDS information ”compared to 79.5% in the upper SEI (P<0.001). A higher percentage of the poor had a stigmatizing attitude towards HIV/AIDS (45.6%) compared to those in the upper SEI (34.8%) [P< 0.001]. There was a high personal HIV risk perception among the poor (40.0%) and it declined significantly to 10.9% in the upper SEI. Conclusions: Our findings underline the disproportionate burden of HIV disease and HIV fear among the poor and vulnerable in South Africa. The poor are further disadvantaged by lack of access to HIV information and HIV/AIDS services such as testing for HIV infection. There is a compelling urgency for the national HIV/AIDS response to maximizing program focus for the poor particularly women. Background The debate on the link between poverty and HIV infec- tion in sub-Saharan Africa has continued for almost two decades without definite consensus. A large body of lit- erature in the early years of the HIV epidemic indicated that relative wealth was associated with a higher risk of HIV infection [1,2]. Owing to the relative abundance of disposable income, individuals and households in the higher income groups were more likely to be engaged in risky multiple concurrent sexual partnerships. As the epidemic matured, those in the poorer income brackets began to become equally at risk of HIV infection, mainly due to the expansion of sexual networks and also due to the increasing transactional nature of sex. During the second decade of HIV epidemic, the lost economic op- portunities and cost of caring for the sick and orphaned became severe among poorer households and communi- ties. This socio-economic impact of HIV/AIDS led to HIV becoming strongly associated with poverty [3,4]. These assertions were, however, context specific. Around 2005, it emerged that socio-economic inequality and vulnerabil- ity, rather than just poverty were most strongly associated with HIV occurrence in sub-Saharan Africa. Piot, Greener, & Russell (2007) and Temah (2008), reported that African countries with greatest Gini Coefficient Index were hardest hit by the epidemic, and most of these countries were found in the Southern African region reaffirming that HIV/AIDS is a disease of inequality rather than of poverty. One reason behind the debate is rooted in methodo- logical shortcomings to measure income and poverty at * Correspondence: [email protected] 1Epidemiology and Strategic Information Unit, Human Sciences Research Council, Private Bag X41, Pretoria 0001, Gauteng, South AfricaFull list of author information is available at the end of the article © 2013 Wabiri and Taffa; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited. Wabiri and Taffa BMC Public Health 2013, 13:1037 http://www.biomedcentral.com/1471-2458/13/1037 the individual and household levels. Population-based Demographic and Health Surveys (DHS), AIDS Impact Surveys (AIS) and other cluster sample surveys use house- hold ownership of items such as radios, refrigerators, phones and the availability of social amenities such as water, electricity and toilets to indicate levels of poverty. It is now believed that possession of these assets and amenities inadequately discriminates poor and non-poor households [5,6].Wealth and social status in rural Africa are still expressed in land, cattle and agricultural owner- ship, despite a more urban lifestyle becoming common in African populations. The cross-sectional nature of data sources that were used to link poverty and HIV infection is another source of controversy because such studies are faced with the dilemma of inferring causality between poverty and HIV infection. Nevertheless, derived socio-economic status or inequality measures are widely used in epidemiological research and are crucial not only for studies focusing on the social determinants of health, but also for the vast majority of observational health research [7]. Previous studies on poverty and HIV infection in South Africa have revealed mixed results. Booysen and Summerton [8] analysed the 1998 DHS data and found that socio-economic inequality, including gender, did not have significant association with HIV infection and sexual risk behaviour. Steinberg et.al [9] reported that households worst hit by HIV were also least served by basic social services such as water and sanitation showing that poor people are most affected by HIV/AIDS in South Africa. However, longitudinal surveillance data from rural Kwazulu-Natal indicated differences in household edu- cational attainment to be a much stronger factor for HIV acquisition than income and expenditure [10]. Using the 2008 South African National HIV prevalence, incidence, behaviour and communication survey [11], this study examines the association of socio-economic in- equality as measured by the household asset index, with key HIV-related outcomes such as HIV prevalence, HIV risk perception, sexual behaviour and utilization of HIV testing services. The study hopes to clarify the above mixed results reported on the link between socio- economic inequality and HIV in South Africa. Data source The study used data from the 2008 South African National HIV prevalence, incidence, behaviour and com- munication survey [11]. This is a cross-sectional population- based household survey conducted every 3–4years using a multi-stage stratified sampling approach (by province, geo-type and predominant racial groups). Sam- pling frames were based on enumeration areas (EA) used in the national census, with updates to reflect changes in the socio-demographic profile of the country sincethe last census in 2001. A total of 1000 EAs were selected as the primary sampling units, 15 households within EAs formed the secondary sampling unit and four eli- gible individuals selected within households formed the final sampling unit. All household members in the se- lected households were eligible to participate, including those living in hostels. People staying in educational insti- tutions, old-age homes, hospitals and uniformed-service barracks, as well as homeless people, were excluded from the survey. A household was defined as a group of people living, cooking and eating together. Dried blood spot (DBS) specimens were used for HIV antibody testing. An algorithm of three HIV enzyme immunoassays was used to test for HIV antibodies [12]. Full details of the survey methodology, including sample weighting, fieldwork procedures and quality control mea- sures and ethical approval aare described elsewhere [6,12]. Methods Deriving study measures Based on the multistage stratified sampling described above, this study draws on data collected from adults aged 15 years and above. Data are drawn from three face- to-face questionnaires: a household-level questionnaire; a youth aged 15–24 years questionnaire; and an adults aged 25 years and above questionnaire. The socio-economic index (SEI) measures were de- rived from 32 items, in the household questionnaires, related to measures of household-living standards, such as infrastructure and housing characteristics (source of drinking water, access to electricity, main source of en- ergy for cooking, and type of toilet used) and household ownership of durable assets (presence of a working re- frigerator, radio, television, cell phone and landline phone). Quantiles were generated using the multiple cor- respondence analysis (MCA) [6,11]. Other studies that have used MCA to generate socio-economic index mea- sures included the works of:- Asselin and Anh [13] in Vietnam; Ki et.al in Senegal [14]; Ndjanyou [15] and Njong [16] both for the Cameroon case. Booysen et. al utilised MCA to construct wealth indices for seven sub- Saharan African countries [17], while Cleary et.al used MCA to generate socio-economic status in assessing equity in the use of antiretroviral treatment in the pub- lic health care system in urban South Africa [18]. Three socio-economic index groups were used instead of the more widely used five groups due to the skewed distribution of the quintiles; the 5th quintile had only 0.6% of the total adults which meant that the frequency was too low frequency for meaningful analysis. Also the socio-economic class differences in the rural com- munities are narrow because of similar income gener- ation activities at that level [19]. Hence, it was more Wabiri and TaffaBMC Public Health2013,13:1037Page 2 of 10 http://www.biomedcentral.com/1471-2458/13/1037 realistic to use three socio-economic index groups to differentiate the households. The Chronbach alpha coefficient for the resulting SEI was 0.7726. The large positive MCA weight of 1.067 for possession of a fixed telephone line (see Additional file 1: Table S1), was a notable sign of high socio-economic status. Absence of toilet (MCA weight of 3.190) and electricity (MCA weight of 3,061) were significantly associated with low socio-economic index or status. The working telephone (Std.Dev. of 0.463), sanitary services (Flush toilet) (Std.Dev of 0.467), and access to electricity (Std.Dev. of 0.412) also confirms that they are important in differentiating SEI scores among the household (See Additional file 1: Table S1). A stigma score was constructed using factor analysis of five variables:‘would you buy food from a shopkeeper who lives with HIV’,‘care for HIV infected family member’, ‘disclose one’s HIV status with at least one family member’ and‘relate to a teacher who lives with HIV’. The reliability and internal consistency of the resulting index was assessed using Chronbach alpha score and had a value of 0.6005. All the five variables had high correlation with the stigma score. The information access score to estimate the level of access to HIV/AIDS information was constructed using factor analysis of five variables namely; frequency of use of TV, radio, newspaper, magazine and internet assuming that they constitute important sources of information on HIV/AIDS. The new variable on information access had a reliability index- Chronbach alpha score of 0.6497, show- ing that the variables used were consistent in explaining the information access construct. HIV risk perception was measured from a personal risk assessment scale ranging between 1 and 4 (1 being low risk and 4 being high HIV risk perception). The survey respondents were asked to rate themselves on the risk of becoming infected with HIV based on four choices: -‘l will definitely not get infected’;‘I probably won’t get infected’;‘I’m probably going to get infected’; and‘I’m definitely going to get infected with HIV’.Re- sponse options such as I’m probably going to get infected; and I’m definitely going to get infected with HIV, were arbitrarily taken to imply high risk perception. Descriptive and regression analysis Analysis was done in Stata version 11.0 (College Station, Texas, United States), taking into account the complex multilevel sampling design and participant non-response. STATA software (svy) commands were used to obtain the estimates of proportions and confidence intervals (95% CI). Summary indices for descriptive analysis are weighted bpercentages, and un-weighted counts are pro- vided. In invariable analysis, the distribution of the study outcomes- HIV testing, HIV risk Perception and HIVprevalence- across population groups were compared using the Rao-Scott F statistic to determinePvalues [20]. Multivariate logistic regression analysis, using backward fitting, was used to identify factors associated with HIV testing, HIV prevalence and HIV risk Perception. The independent variables include socio-economic index, education, stigma score, information access score and selected background characteristics. Clustering was not accounted for given that the large number of primary sampling units (1000) in the study is comparable to respondent number, thus diminishing such effects. Results Of the 15,000 households sampled, only 13,440 (89.3%) were occupied; 80.8% of whom were interviewed (10,856/ 13,440). Non-response was largely due to refusal (9.3%, 1252/13,440) or no household member at home after four repeat visits (7.0%, 946/ 13,440). The study sample N = 14,384 of adults 15–64 years, out of the 23,112 cases in the survey. Descriptive analysis results Background characteristics across the socio-economic index (SEI) Overall, 40.1% of the 14,384 adults (15–64 years) fall in the poor SEI group, 42.5% in middle and remaining 17.4% in upper SEI [P < 0.001] (Table 1). More women (57.5%) compared to men (42.5%) were found in the poor SEI group [P < 0.001]. All but three provinces (Gauteng 35.3%, Western Cape 22.2% and Kwazulu-Natal 17.9%), had less than 10% of respondents who belonged to the upper (not-so-poor) SEI group [P < 0.001]. Limpopo (18.8%), Eastern Cape (18.8%) and Kwazulu-Natal (23.0%) had the largest percentage of respondents belonging to the poor SEI [P < 0.001]. The largest percentage of respon- dents in poor SEI (57.5%) lived in the rural tribal land (or rural formal settlement) [P < 0.001]. More than 30% of respondents in poor SEI were urban residents, equally dis- tributed in the formal and informal settlementt [P > 0.05]. More than 70% of those in the middle and 92% of the upper SEI group lived in formal urban settlements (P < 0.001). Eighty four per cent of respondents in the poor SEI had no formal education or completed only up to primary level compared to 31% in the upper SEI group [P < 0.001]. On the other hand, 68.2% of those in upper SEI and 36.3% of the middle SEI respectively had completed matric exam or tertiary education. Over 95% of people in the poor SEI group were black Africans while other races formed close to 47% of those belonging to the upper SEI [P < 001. HIV testing and prevalence across socio-economic index groups Respondents in upper SEI reported higher percentage of HIV testing (59.3%) in the past followed by those in the Wabiri and TaffaBMC Public Health2013,13:1037Page 3 of 10 http://www.biomedcentral.com/1471-2458/13/1037 Table 1 Distribution of Socio-economic index (SEI) among adults (15–65 years) by selected background characteristics Characteristic Socio-economic index (SEI) groups Poor(low) Middle Upper % [95% CI] % [95% CI] % [95% CI] Sex*** Male 42.3[40.2,44.4] 45.5[43.3,47.8] 45.1[41.6,48.6] Female 57.7[55.6,59.8] 54.5[52.2,56.7] 54.9[51.4,58.4] N = 4308 N = 5067 N = 2892 Age (years)*** 15-2437.1[35.6,38.6] 31.3[29.9,32.8] 25.1[23.2,27.2] 25-4948.8[46.9,50.8} 55.8[53.7,57.8] 54.2[51.2,57.2] 50+14.1[12.9,15.4] 12.9[11.614.3] 20.6[18.4,23.0] N = 4308 N = 5067 N = 2892 Race*** African 96.2[95.2,97.0] 78.3[75.0,81.3] 35.7[30.0,41.8] White 0.5[0.3,1.0] 8.0[6.4,9.8] 35.8[30.8,41.1] Coloured 3.2[2.5,4.1] 11.5[9.7,13.7] 17.9[15.0,21.3] Indian 0.1[0.0,0.1] 2.2[1.3,3.6] 10.6[8.5,13.0] N = 4300 N = 5060 N = 2881 Province*** Western Cape 4.3[3.2,5.8] 13.8[11.4,16.6] 22.4[19.2,26.0] Eastern Cape 18.8[15.5,22.6] 8.4[6.4,11.0] 9.6[6.8,13.3] Northern Cape 1.4[1.1,1.8] 2.4[1.9,2.9] 2.8[2.1,3.7] Free State 4.9[3.9,6.2] 8.2[6.3,10.8] 3.7[2.6,5.4] KwaZulu-Natal 23.0[18.9,27.6] 18.1[14.6,22.3] 17.9[14.5,22.0] North West 10.6[8.5,13.1] 7.8[6.2,9.6] 3.8[2.6,5.5] Gauteng 9.0[6.9,11.7] 28.7[23.3,34.8] 35.3[30.0,41.0] Mpumalanga 9.3[7.4,11.5] 6.7[5.0,9.0] 2.9[1.8,4.5] Limpopo 18.8[15.9,22.0] 5.9[4.4,7.8] 1.8[1.1,2.7] N = 4308 N = 5067 N = 2892 Geographical location*** Urban formal 16.6[13.5,20.4] 72.9[68.4,77.1] 92.4[88.8,94.9] Urban informal 16.0[12.9,19.8] 7.8[5.6,10.8] 7.8[0.3,1.4] Rural informal 57.5[52.5,62.4] 14.6[11.7,18.2] 1.9[0.7,5.0] Rural formal 9.8[7.6,12.5] 4.6[3.3,6.3] 5.1[3.2,8.0] N = 4308.0 N = 5067.0 N = 2892.0 Education*** No schooling 8.0[6.9,9.2] 2.3[1.8,3.1] 0.9[0.5,1.7] Less than matric 75.5[73.5,77.4] 56.0[53.5,58.5] 31.0[27.7,34.5] Matric 13.7[12.2,15.3] 30.5[28.3,32.8] 41.1[38.1,44.1] Above Matric 2.8[2.1,3.8] 11.1[9.5,13.0] 27.1[24.0,30.4] N = 4084 N = 4882 N = 2792 Wabiri and TaffaBMC Public Health2013,13:1037Page 4 of 10 http://www.biomedcentral.com/1471-2458/13/1037 middle (56.6%) and the poor (47.7%) respectively [P < 0.001]. Although men generally reported testing for HIV infection less frequently than women, this difference was more sub- stantial among the poor-a difference of 24.4% compared to a difference of 11.7% between men and women who were tested for HIV infection in the upper SEI group. Among the poor SEI group, only 33.6% of men reported HIV testing compared to 58% women [P < 0.001]. The rate of HIV testing was 52.8% and 64.5% for men and women respectively in the upper SEI [P < 0.001]. Other results (not shown) indicate that HIV testing was higher among whites (66.8%) although more than 50% of all other races also reported having ever been tested [P < 0.001]. HIV testing reports showed less disparity by major geographic locations except in rural informal (tribal) areas. On average, 56.0% of people in urban and formal rural areas reported to have been tested for HIV infection compared to 46.3% for those living in urban informal [P < 0.001]. Majority of respondents in the upper SEI did not receive HIV test from public health facilities. Accordingly, 89.0% of the poor, 64.5% of those in themiddle and only 31.1% of the upper SEI group ever sought HIV test from government hospital or clinic [P < 0.001]. More than 97.0% of those who sought HIV test in public facilities expressed satisfaction in the service across all socio-economic index groups. HIV prevalence was highest among the poor (20.8%) compared to those in the middle (15.9%) and upper SEI (4.6%) respectively [P < 0.001] (Table 2). HIV prevalence was significantly higher among women than men (19.7% versus 11.4% respectively); among black Africans (20.3%) compared to other races (3.4% among coloured and <1.0% amongwhites);andamongurbaninformalsettlement residents (28.5%) compared to those living in rural areas (18.2%) or urban formal areas (12.7%). Among the poor, HIV prevalence was almost twice as high among women compared to men (25.4% versus 14.0%) (Table 2). It was 5 times higher among black Africans when compared to coloured people (21.5% versus 4.4%), but 20 times higher when compared to whites or Indians who had almost no HIV positive test results [P < 0.001]. Similar patterns of race and gender differentials were seen when HIV Table 1 Distribution of Socio-economic index (SEI) among adults (15–65 years) by selected background characteristics (Continued) Gross income before tax per annum*** No Income 35.8[31.6,40.2] 18.0[15.1,21.2] 12.6[10.1,15.5] < 6000 23.7[20.4,27.2] 14.0[11.8,16.5] 5.2[3.8,7.1] 6000-24000 27.4[24.3,30.9] 25.5[22.7,28.6] 16.4[13.2,20.1] 24000-96000 11.4[9.3,13.9] 29.7[26.7,33.0] 30.9[26.9,35.2] >96000 1.7[1.0,3.1] 12.8[10.6,15.3] 34.9[30.9,39.2] N = 1518 N = 2551 N = 1781 ***Significant at 0.1%. Table 2 HIV prevalence, HIV testing, HIV risk perception, HIV information access and HIV stigma among men and women (15-65 years) across social economic index groups Social economic index groupsHIV prevalence HIV testing HIV risk perception (High)HIV information score (Low access)HIV stigma score (High stigma) n % [95% CI] n % [95% CI] n % [95% CI] n % [95% CI] n % [95% CI] Overall*** Poor 3470 20.8[18.8,22.9] 4078 47.7[45.6,49.9] 4063 40.0[37.9,42.1] 4051 79.5[77.5,81.5] 4066 45.6[43.2,48.0] Middle 3981 15.9[14.0,18.1] 4877 56.6[54.2,59.0] 4873 26.0[23.7,28.5] 4859 37.4[35.3,39.6] 4865 34.4[31.9,37.0] Upper 2110 4.6[2.6,7.9] 2789 59.3[56.3,62.2] 2789 10.9[8.5,13.8] 2774 20.1[17.4,23.1] 2787 34.8[31.7,38.1] Men*** 3859 11.4[9.7,13.3] 4771 44.3[42.1,46.6] 4784 24.2[22.3,26.2] 4761 48.4[45.8,51.0] 4767 41.3[38.9,43.6] Poor 1296 14,0[11.5,16.9] 1573 33.6[30.3,37.0] 1581 33.2[30.2,36.3] 1571 75.8[72.3,78.9] 1565 48.6[45.2,52.0] Middle 1588 11.7[9.0,15.0] 1973 49.9[46.5,53.4] 1978 22.6[19.8,25.7] 1967 35.8[32.6,39.0] 1972 37.1[33.5,40.8] Upper 912 4.6[1.9,10.5] 1181 52.8[48.1,57.4] 1181 8.0[5.6,11.2] 1178 20.6[16.6,25.1] 1186 37.1[32.7,41.7] Women*** 5883 19.7[18.1,21.4] 7065 60.8[59.1,62.6] 7034 32.7[30.6,34.8] 7019 53.2[50.9,55.6] 7047 37.0[35.1,39.0] Poor 2174 25.4[22.9,28.1] 2505 58.0[55.3,60.7] 2482 45.0[42.3,47.6] 2480 82.3[80.0,84.3] 2501 43.4[40.7,46.2] Middle 2393 19.4[16.9,22.2] 2904 62.2[59.2,65.1] 2895 28.8[25.8,32.1] 2892 38.8[36.2,41.4] 2893 32.2[29.5,35.0] Upper 1198 4.5[2.8,7.1] 1608 64.5[60.8,68.1] 1608 13.2[9.8,17.6] 1596 19.7[16.4,23.5] 1601 33.0[29.0,37.2] ***Significant at 0.1%. Wabiri and TaffaBMC Public Health2013,13:1037Page 5 of 10 http://www.biomedcentral.com/1471-2458/13/1037 prevalence data was examined among those in the middle and upper SEI groups. HIV prevalence followed the national pattern of high peak rate among 29–40 year olds across all the three socio-economic groups, but maintained socio-eco- nomic gradient of being highest among the poor and lowest among the upper SEI group (i.e. 30.5%, 22.3% and 6.4% respectively). HIV risk perception across socio-economic index groups There was high personal HIV risk perception among the poor which declined significantly higher up in the socio- economic ladder. About, 40% among the poor, 26% among those in the middle SEI group and 10.9% of those in the upper SEI group believed that they were at high risk of HIV infection [P < 0.001] (Table 2). HIV risk per- ception was 4 times higher among the poor compared to the upper SEI. High HIV risk perception was reported more among women (32.7%) than men (24.3%) for all socio-economic groups (P < 0.001) although this differ- ence was much pronounced among the poor (33.2% for men versus 45% for women) and the upper SEI group (8.0% for men and 13.2% for women) than it was for those in the middle (22.6% for men and 28.8% for men). HIV-related stigma across socio-economic index groups In general, 61.1% of the respondents had non-stigmatiz- ing attitude (low-stigma score) towards HIV/AIDS, al- though women had better attitudes (63.0%) compared to men (58.7%) [P < 0.01]. A higher percentage of those in the poor SEI group had a high stigmatizing attitude to- wards HIV/AIDS (45.6%) compared to those in middle and upper SEI group, 34.4% and 34.8% respectively [P < 0.001]. HIV/AIDS information across socio-economic index groups Only 20.5% of the poor SEI group had what could be labelled as“good access to HIV/AIDS information” compared to 79.9% in the upper SEI group [P < 0.001] (Table 2). Further, 74.8% of total respondents (63.6% for the poor and 82.9% for the upper SEI; P <0.001) listen to the radio and 76.7% watch television almost daily (49.2% for the poor and 95.5% for the upper SEI; P <0.001) (Additional file 1: Table S2). However, the frequencies of internet use and newspaper and maga- zine reading were much less although there was great disparity by socio-economic status. Only 41.2% read newspapers almost daily (23.4% among the poor SEI and 58% among the upper SEI groups; P < 0.001), 32.5% read magazines almost daily (19% among the poor SEI and 46% among the upper SEI groups [P < 0.001]; and 13.4% surfed internet almost daily [P < 0.001]) (Additional file 1: Table S2). Sexual behaviour across the socio-economic index groups Sexual risk behaviour was assessed using the reported num- ber of sex partners (regular and non-regular), the number who reported that their recent sexual partner had other sex- ual partner(s) and the report of condom use during last sex- ual encounter. The survey did not enquire if respondents already knew their sero-status before anonymous HIV test was done although it is assumed that a good percentage of them were known HIV positives. However, 55.0% of HIV positive individuals reported to have had only one regular sex partner during the year preceding the survey compared to only 29.0% of those who were found to be HIV negatives (P < 0.001). Seventy percent of the HIV positives and 44.4% of the HIV negatives with high HIV risk perception believed that their sexual partner had other sex partner(s) (P < 0.001). Only 2.8% of the total respondents reported that they had more than one regular sex partner during the last 12 months. There was a statistically significant but mar- ginal difference between the poor (2.7%) and the upper SEI group (1.8%) [P < 0.001]. Similarly, only 4.8% of the respondents said that they had one or more non-regular sex partner during the past 12 months. Majority of respondents (96.9%), across all the social- economic index groups, believed that condoms are easy to find at any time when one is in need to use. Women believed more so (98.0%) than men (95.0%) [P < 0.01]. Regression analysis results Logistic regression results depicted in Tables 3 and 4 con- firmed the statistical association indicated by the descriptive measures. Educational level was significantly associated with HIV risk perception, decreasing with increase in level ofeducation,andthisappliestoallSEIgroups(Table3). Those with less than secondary school educational level perceived themselves to be at higher risk of acquiring HIV infection (OR = 1.46, p < 0.001) compared to those with ter- tiary level and above. Those in urban informal areas (mostly urban poor) had significantly increased the odds of high HIV prevalence compared to those in urban formal areas (mostly the urban non-poor) (OR = 2.74, P < 0.000, Table 3). The HIV risk perception is all high among those in urban informal (urban poor) compared to urban formal (non- poor) (OR = 2.34, p < 0.001, Table 3). Overall, the poor were 5 times more likely (OR = 5.46, p < 0.001; Table 3) to perceive themselves as being at high risk of acquiring HIV infection compared to those in the upper SEI group, while those in the middle SEI groups were 3 time more likely (OR = 2.88, p < 0.001) to have similar HIV risk perceptions . This is more pronounced among black Africans with odds of 6.69 compared to other races. Among the poor SEI groups, being female (AOR = 2.21, P < 0.001) and black African (AOR = 7.92, P < 0.001) sig- nificantly increased the odds of high HIV prevalence (Table 4). As the black African population moves up in Wabiri and TaffaBMC Public Health2013,13:1037Page 6 of 10 http://www.biomedcentral.com/1471-2458/13/1037 SEI, the odds of infection increase compared to other races. Being in the middle SEI group and living in urban informal settlement was significantly associated with high odds of HIV infection (AOR = 1.61, P < 0.001), 2-times higher than those living in formal urban areas. Among the poor and middle SEI groups, being a black African female and living in rural formal (Tribal) or urban informal areas, increased the odds of perceiving self to be at risk of acquiring HIV infection (AOR = 1.4, p < 0.01; Table 4). Nonetheless, the poor were less likely to test for HIV infection (OR = 0.63, P < 0.001; Table 3) and this is more pronounced among black Africans compared to other races (OR = 0.72, p < 0.05; Table 3). Those with tertiary education were more likely to test across all socio-economic index groups. More females compared to men underwent HIV testing across all the SEI groups, but this was 3 times higher among the poor (AOR = 2.81, P < 0.001; Table 4). The odds of high stigma is roughly the same across all SEI groups (AOR = 0.76; p < 0.01; Table 4). For those in upper SEI group, the key factor for high HIV prevalence was being a black African (AOR = 49, P < 0.001, Table 4). Discussion The findings showed high HIV prevalence among the poor in general and specifically among women, black African race and individuals with low educational status.The poor also felt more susceptible to HIV infection compared to those in upper SEI group. These study results are in agreement with current global thinking around the bidirectional relationship between socio- economic inequality and poor health outcome, in this case, HIV/AIDS [9,12,21-23]. In particular, the study points to the assumption that the poor in South Africa would have dual challenges of vulnerability (particu- larly women) and lack of opportunities to make better life choices due to limited education and HIV/AIDS services (such as information on and testing for HIV infection). Relative economic opportunities among black South Africans, referred to as“relative wealth”by Fox (2012), on the other hand were strongly associated high HIV prevalence. Magadi [24] observed similar results in her analysis of DHS survey data from 20 countries in sub-Saharan Africa in which the urban poor were noted to have significantly higher odds of HIV infection than their urban non-poor counterparts. Some researchers argue that the“social history of AIDS and the way it was represented”in the early years gave a legacy to the continued stigma towards the disease [25] although it was believed to diminish in the era of ARV scaling [24]. Our analysis indicated a fairly high level of stigma at a time (in 2008) when nearly 40% of eligible South Africans were receiving ARV treatment [26]. More rigorous studies are indispensable to fully understand the Table 3 Un-adjusted odds ratio for HIV prevalence (Model 1); HIV testing (Model 2), and risk perception (Model 3) and by background characteristics Model 1: HIV prevalence Model 2: HIV testing Model 3: HIV risk perception Un-adjusted odds ratio (OR)Std. err Un-adjusted odds ratio (OR)Std. err Un-adjusted odds ratio (OR)Std. err Sex:Female vs. Male 1.91*** 0.19 1.95*** 0.11 1.52*** 0.10 Socio-economic Index(Upper SEI a) Low (poor) SEI 5.48*** 1.68 0.63*** 0.05 5.46*** 0.81 Middle SEI 3.97*** 1.23 2.88*** 0.40 Race:Blacks vs. Other Races 18.48*** 3.10 0.72*** 0.04 6.69*** 0.70 Location(Urban Formal a) Urban Informal 2.74*** 0.33 2.38*** 0.28 Rural Informal 1.53*** 0.19 0.65*** 0.05 2.37*** 0.22 Rural Formal1.81*** 0.27 Education(Secondary a) No schooling 0.29*** 0.04 1.75*** 0.26 Less Secondary 1.36*** 0.16 0.48*** 0.03 1.46*** 0.13 Tertiary 0.38*** 0.09 1.79*** 0.21 0.57*** 0.08 Stigma Score: High vs. Low 0.80*** 0.07 0.65*** 0.04 0.88** 0.06 Information Access: High vs. Low 0.56*** 0.05 1.60*** 0.10 0.47*** 0.04 N 9,742 11730 11748 aReference group; ***significant at .1%; **significant at 1%. Wabiri and TaffaBMC Public Health2013,13:1037Page 7 of 10 http://www.biomedcentral.com/1471-2458/13/1037 reasons behind continued stigma and discrimination more than two decades into the epidemic and when liv- ing with HIV does not anymore mean“death sentence”as it used to be.Stringer et al. [27] found that having had more than two lifetime sexual partners was a marker of high per- sonal risk perception for HIV infection, although this perception did not predict HIV sero-status among women Table 4 Adjusted odds ratio for HIV prevalence (Model M1), HIV testing (Model M2) and HIV risk (Model M3) perception within social economic groups by background characteristics Model M1: HIV prevalenceModel M2: HIV testingModel M3: HIV risk perception Adjusted odds ratio (AOR)Std. err Adjusted odds ratio (AOR)Std. err Adjusted odds ratio (AOR)Std. err Poor SEI (n = 3,190) (n = 3,947) (n = 3,933) Sex:Female vs. Male 2.21*** 0.30 2.81*** 0.28 1.59*** 0.14 Race:Blacks vs. Other Races 7.92*** 2.38 2.74*** 0.63 Geographical Location (Urban Formal a) Urban Informal Rural Informal 0.66** 0.12 0.75** 0.10 Rural Formal1.44** 0.27 Education(Secondary a) No schooling 0.35*** 0.07 Less secondary 0.61*** 0.08 Stigma Score: High vs. Low 0.76*** 0.07 0.77*** 0.07 Information Access: High vs. Low0.75** 0.09 Middle SEI (n = 3,726) (n = 4,753) (n = 4,748) Sex:Female vs. Male 1.66*** 0.30 1.71*** 0.17 1.33** 0.15 Race:Blacks vs. Other Races 10.42*** 2.50 0.83* 0.09 4.10*** 0.65 Geographical Location (Urban Formal a) Urban Informal 1.61** 0.30 1.43* 0.27 Education(Secondary a) No schooling 0.24*** 0.07 Less secondary 0.49*** 0.05 Tertially 0.54* 0.18 1.77*** 0.31 Stigma Score: High vs. Low 0.75* 0.12 0.76*** 0.07 Information Access: High vs. Low 0.76* 0.12 0.68*** 0.08 Upper SEI (n = 1,926) (n = 2,727) (n = 2,728) Sex:Female vs. Male 1.72*** 0.24 Race:Blacks vs. Other Races 49.02*** 27.68 5.72*** 1.42 Geographical Location (Urban Formal a) Urban Informal 2.87* 1.65 Education(Secondary a) No schooling 0.23* 0.19 Less secondary 2.10* 0.84 0.40*** 0.07 Tertially 2.26*** 0.44 0.54* 0.19 Stigma Score: High vs. Low 0.42* 0.21 0.79* 0.11 aReference group; ***significant at .1%; **significant at 1% ; *significant at 5%. Wabiri and TaffaBMC Public Health2013,13:1037Page 8 of 10 http://www.biomedcentral.com/1471-2458/13/1037 in Zambia. The survey used in this study did not en- quire if respondents already knew their sero-status be- fore anonymous HIV test was done although it is assumed that a good percentage of them were known HIV positives. It is plausible to assume that sexual risk perception is strongly linked to one’ssexualrisk behaviour [28]. The extent to which this high risk per- ception serves as a motivational factor for adaptation of protective behaviour or remains subdued due to individ- uals’socio-economic vulnerability needs to be further investigated. In our analysis, the poor had limited access to HIV/ AIDS information. A study on public communications and its role in reducing and eliminating health disparities by Viswanath (2006) indicated that those in low socio- economic status (SES) also tend to gain less from the information flows than their counterparts of higher SES [29]. It is thus important to note that inequalities in ac- cess to mass media also follow the pattern of existing inequalities in HIV/AIDS service delivery and marginalize the poor and vulnerable. One’s economic status created synergy with gender and level of educational attainment to significantly influ- ence HIV-related outcomes in this study. This again is in agreement with many studies in sub-Saharan Africa which highlighted the disadvantage of being a woman living among communities with high socio-economic inequality [22,23]. Majority of people who did not test for HIV infection did so because of misconceptions about the disease which is closely related to their attitude towards the disease. The expansion of a routine and intensified campaign for HIV testing would contribute immensely towards breaking this link between misconcep- tions and stigma. Our analysis also highlighted the fact that perceived level of service quality may not significantly limit demand for HIV testing in public health facilities. Rather, poor access to these facilities could be seen to be an issue for poor people. Finally, we believe that the socio-economic index that we constructed was able to discriminate inequalities by race, province, geographic location (urban and rural), level of income and educational attainments. The house- hold asset or wealth status generated in this analysis was cut-down to three quintiles to minimize erroneous infer- ences based on extremely skewed socio-economic profiling. We also found instances of over-and-under estimation of poverty in some settings in a manner that is not reflective of the reality on the ground even after adjustment. These weaknesses mirror the usual critic on an asset index for its inability to clearly distinguish“poor households from the poorest ones”. Rustein (2008) suggests developing research instruments based on variables that appropri- ately describe economic situations both in the urban and rural area and adequately discriminate differenteconomic groups among residents and calculating a composite index of the two. Conclusions Using a simplified socio-economic index profile, this study was able to underline the disproportionate distribution of HIV disease burden and fear among the poor in South Africa. The poor were further disadvantaged by lack of access to HIV information and HIV/AIDS services such as testing for HIV infection. Our socio-economic index profiling could not make a clear discrimination within the“middle class and wealthy”mainly because of weak- nesses in measures of living standards. There is a compelling urgencyforthenationalHIV/AIDS response to maximize program focus for the poor particularly women. EndnotesaStudy activities were approved by the Human Science’s Research Council’s Research Ethics Committee (REC 2/ 13/10/07) and Human Subjects Review (IRB # 00006347) from the Centre for Disease Control and Prevention’s Global AIDS Programme. bWeighting of the sample by age, race group and province was applied to ensure the study estimates are representative of the general population. Additional file Additional file 1:MCA Weights and Variance of the Variable modalities as Table S1 and Sources of HIV/AIDS information by Socio-economic index as Table S2. Competing interests Both authors declare that they have no competing interests. Authors’contributions NW, NT conceived of the study, and participated in its design and coordination. NW, NT wrote the draft manuscript. NW, NT performed the statistical analysis. Both authors read and approved the final manuscript. Acknowledgements Financial support for survey data used in the study was from the President’s Emergency Plan for AIDS Relief (PEPFAR) through the US Centers for Disease Control and Prevention (HHS/CDC) under the terms of Cooperative Agreement No. 1 U2G PS 000570. The South African National HIV prevalence, incidence, behaviour and communication survey team that managed, coordinated and supervised field data collection. Author details 1Epidemiology and Strategic Information Unit, Human Sciences Research Council, Private Bag X41, Pretoria 0001, Gauteng, South Africa. 2United States Center for Disease Control and Prevention (CDC), QED GrS2oup LLC, Windhoek, Namibia. Received: 23 April 2013 Accepted: 30 October 2013 Published: 4 November 2013 Wabiri and TaffaBMC Public Health2013,13:1037Page 9 of 10 http://www.biomedcentral.com/1471-2458/13/1037 References1. Mishra V, Assche SB, Greener R, Vaessen M, Hong R, Ghys PD: HIV infection does not disproportionately affect the poorer in sub-Saharan Africa. AIDS 2007, 21(Suppl 7):S17 –S28. 2. Gillespie S: Poverty, food insecurity, HIV vulnerability, and the impacts of AIDS in Sub-saharan africa. IDS Bull 2008, 39:10–18. 3. UNAIDS: Report on the Global AIDS Pandemic, Joint United Nations Programme on HIV/AIDS. Geneva, Switzerland: UNAIDS; 2004. 4. Gonzalez L, Seley C, Martorano J, Garcia-Moreno I, Troncoso A: Infections and inequalities: anemia in AIDS, the disadvantages of poverty. Asian Pac J Trop Biomed 2012, 2:485 –488. 5. Bingenheimer JB: Wealth, wealth indices and HIV risk in east africa. Int Fam Plan Perspect 2007, 33:83–84. 6. Rutstein SO: The DHS Wealth Index: Approaches for Rural and Urban Areas. Washington, DC: Macro International Inc; 2008. http://pdf.usaid.gov/ pdf_docs/PNADN521.pdf. 7. Howe LD, Hargreaves JR, Huttly SR: Issues in the construction of wealth indices for the measurement of socio-economic position in low-incomecountries. Emerg Themes Epidemiol 2008, 5:3. 8. Booysen Fle R, Summerton J: Poverty, risky sexual behaviour, and vulnerability to HIV infection: evidence from South Africa. J Heal Popul Nutr 2002, 20:285 –288. 9. Steinberg M, Johnson S, Schierhout G, Ndegwa D: Hitting home: how households cope with the impact of the HIV/AIDS epidemic: a survey of households affected by HIV/AIDS in South Africa. Washington, DC: Henry J. Kaiser Family Foundation; 2002. 10. Barnighausen T, Hosegood V, Timaeus IM, Newell M-L: The socioeconomic determinants of HIV incidence: evidence from a longitudinal,population-based study in rural South Africa. AIDS 2007, 21(suppl 7):S29 –S38. 11. Shisana O, Rehle T, Simbayi L, Zuma K, Jooste S, Pillay-van-Wyk V, Mbelle N, Van Zyl J, Parker W, Zungu N, Pezi S, SABSSM III Implementation Team T: South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2008 A Turning Tide Among Teenagers? Cape Town, South Africa: HSRC press; 2009:120. 12. Temah Tsafack C: Socio-Economic Inequalities and HIV/AIDS Epidemic: Evidence from Sub-Saharan Africa. France: Doctoral dissertation, CERDI, University d ’Auvergne- Clermont-Ferrand I; 2008. 13. Asselin L, Anh TV: Multidimensional Poverty Measurement with Multiple Correspondence Analysis .In Quant Approaches to Multidimens Poverty Meas. Edited by Kakwani N, Silber J. London: Palgrave-Macmillan; 2008. 14. Ki JB, Faye S, Faye B: Multidimensional Poverty in Senegal: A Non- Monetary Basic Needs Approach. PMMA Working Paper-PEP; 2005. 15. Ningaye P, Ndjanyou L, Saakou GM: Multidimensional Poverty in Cameroon: Determinants and Spatial Distribution. Nairobi, Kenya: AERC Research Paper 211; 2011. 16. Njong MA: Multidimensional Spatial Poverty Comparisons in Cameroon. Nairobi, Kenya: AERC Research Workshop; 2007. 17. Booysen F, Van Der Berg S, Burger R, Von Maltitz M, Du Rand G: Using an Asset Index to Assess Trends in Poverty in Seven Sub-Saharan African Countries. World Dev 2008, 36(6):1113 –1130. 18. Cleary S, Silal S, Birch S, Carrara H, Pillay-van Wyk V, Rehle T, Schneider H: Equity in the use of antiretroviral treatment in the public health caresystem in urban South Africa. Health Policy (New York) 2011, 99:261 –266. 19. Onwujekwe O, Uzochukwu B, Eze S, Obikeze E, Okoli C, Ochonma O: Improving equity in malaria treatment: relationship of socio-economicstatus with health seeking as well as with perceptions of ease of using the services of different providers for the treatment of malaria in Nigeria. Malar J 2008, 7:5. 20. Bartholomew D, Steele F, Moustaki I, Galbraith J: The Analysis and Interpretation of Multivariate Data for Social Scientists. London: Chapman & Hall/CRC; 2002. 21. Piot P, Greener R, Russell S: Squaring the circle: AIDS, poverty, and human development. PLoS Med 2007, 4:1571 –1575. 22. Magadi MA: The disproportionate high risk of HIV infection among the urban poor in sub-Saharan Africa. AIDS Behav 2013, 17:1645 –1654. 23. Fox AM: The HIV-poverty thesis re-examined: poverty, wealth or inequality as a social determinant of HIV infection in sub-Saharan Africa? J Biosoc Sci 2012, 44:459–480. 24. Remien RH, Mellins CA: Long-term psychosocial challenges for people living with HIV: let ’s not forget the individual in our global response to the pandemic. AIDS 2007, 21:55–63. 25. Lynskey V: Risk perception and the stigma of HIV/AIDS: why routine testing will change how Americans view the disease. Interdiscip J Heal Ethics, Policy 2008, 7:7–11. 26. Adam M, Johnson L: Estimation of adult antiretroviral treatment coverage in South Africa. South African Med J 2009, 99:661 –667. 27. Stringer ME, Sinkala M, Kumwenda R, Chapman V, Mwale A, Vermun S: Personal risk perception, HIV know ledge and risk avoidance behaviour and their relationship to actual HIV sero-status in an urban africapopulation. J Acquir Immune Deficiciency Syndr 2004, 35:60–66. 28. Schoemaker J, Twikirize J: A life of fear: Sex workers and the threat of HIV in Uganda. Int J Soc Welf 2012, 21:186 –193. 29. Viswanath K: Public communications and its role in reducing and eliminating health disparities .In Examining Heal Disparities Res Plan Natl Institutes Heal Unfinished business. Edited by Thomson G, Mitchell F, Williams M. Washington, D.C: Institute of Medicine; 2006:215 –253. doi:10.1186/1471-2458-13-1037 Cite this article as: Wabiri and Taffa: Socio-economic inequality and HIV in South Africa. BMC Public Health 2013 13:1037. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Wabiri and Taffa BMC Public Health 2013, 13:1037 Page 10 of 10 http://www.biomedcentral.com/1471-2458/13/1037 Please read each article attached. For each article: Provide a paragraph summarizing each articleProvide a paragraph discussing the limitations of each studyProvide a paragraph reviewing the article & Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ijmf20 The Journal of Maternal-Fetal & Neonatal Medicine ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ijmf20 The risk of preterm birth among women with a history of leukemia or lymphoma Sonia T. Anand, Elizabeth A. Chrischilles, Rebecca J. Baer, Mary E. Charlton, Patrick J. Breheny, William W. Terry, Monica R. McLemore, Deborah A. Karasek, Laura L. Jelliffe-Pawlowski & Kelli K. Ryckman To cite this article: Sonia T. Anand, Elizabeth A. Chrischilles, Rebecca J. Baer, Mary E. Charlton, Patrick J. Breheny, William W. Terry, Monica R. McLemore, Deborah A. Karasek, Laura L. Jelliffe-Pawlowski & Kelli K. Ryckman (2021): The risk of preterm birth among women with a history of leukemia or lymphoma, The Journal of Maternal-Fetal & Neonatal Medicine, DOI: 10.1080/14767058.2021.1907332 To link to this article: https://doi.org/10.1080/14767058.2021.1907332 View supplementary material Published online: 08 Apr 2021.Submit your article to this journal Article views: 108View related articles View Crossmark data ORIGINAL ARTICLE The risk of preterm birth among women with a history of leukemia or lymphoma Sonia T. Anand a , Elizabeth A. Chrischilles a, Rebecca J. Baer b,c , Mary E. Charlton a, Patrick J. Breheny d, William W. Terry e, Monica R. McLemore f , Deborah A. Karasek g, Laura L. Jelliffe-Pawlowski c,g and Kelli K. Ryckman a,e aDepartment of Epidemiology, University of Iowa, Iowa City, IA, USA; bDepartment of Pediatrics, University of California San Diego, La Jolla, CA, USA; cCalifornia Preterm Birth Initiative, University of California San Francisco, San Francisco, CA, USA; dDepartment of Biostatistics, University of Iowa, Iowa City, IA, USA; eDepartment of Pediatrics, University of Iowa, Iowa City, IA, USA; fDepartment of Family Health Care Nursing, University of California San Francisco, San Francisco, CA, USA; gDepartment of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, CA, USA ABSTRACTObjective:Leukemia and lymphoma are top cancers affecting children, adolescents and young adults with high five-year survival rates. Late effects of these cancers are a concern in reproduct- ive-age patients, including pregnancy outcomes such as preterm birth. Our study aimed to evaluate whether diagnosis of leukemia or lymphoma prior to pregnancy was associated with preterm birth (<37 weeks gestation). Methods:We conducted a cross-sectional study using a population-based dataset from California with linked birth certificates to hospital discharge records and an Iowa-based sample that linked birth certificates to Surveillance, Epidemiology, and End Results (SEER) cancer registry data. Preterm birth was defined using birth certificates. We ascertained history of leukemia and lymphoma using discharge diagnosis data in California and SEER registry in Iowa. Results:Prevalence of preterm birth in California and Iowa was 14.6% and 12.0%, respectively, in women with a history of leukemia/lymphoma compared to 7.8% and 8.2%, respectively, in women without a cancer history. After adjusting for maternal age, race, education, smoking, and plurality, Women with history of leukemia/lymphoma were at an increased risk of having a pre- term birth in California (odds ratio (OR) 1.89; 95% confidence interval (CI) 1.56–2.28) and Iowa (OR 1.61; 95% CI 1.10–2.37) compared to those with no cancer history. Conclusion:In both California and Iowa, women with a history of leukemia or lymphoma were at increased risk for preterm birth. This suggests the importance of counseling with a history of leukemia/lymphoma prior to pregnancy and increased monitoring of women during pregnancy. ARTICLE HISTORYReceived 21 August 2020 Revised 7 March 2021 Accepted 17 March 2021 KEYWORDSPreterm birth; leukemia; lymphoma; gestational age; premature Introduction Leukemias and lymphomas, which are among the most frequent cancers affecting children, adolescents and young adults, are highly curable with five-year survival approximately 80% for most types [1]. Once young people reach reproductive age, they are often concerned about the negative impact of their prior cancer on their reproductive health including infertility or adverse pregnancy outcomes such as preterm birth. Preterm birth (PTB) is defined as having a baby born too early, specifically<37 weeks gestation. According to the World Health Organization, there are approximately 15 million babies every year who are born preterm, which is more than 1 in 10 babies [2].In the United States in 2018 alone, approximately 1 in every 10 newborns was born preterm; 2018 was the fourth straight year that the rate increased [3]. There are 1.1 million deaths due to preterm birth globally and it continues to be one of the single greatest con- tributors to infant mortality in the United States and to disability-adjusted life years worldwide [4,5]. Identifying women at higher risk for preterm birth is important to inform research that aims to improve pregnancy outcomes in women with preexist- ing conditions. There have been several studies of the effects of childhood, adolescent and young adult leukemia/ lymphoma on pregnancy outcomes [6–14]. The major- ity of these studies were retrospective cohort studies CONTACTKelli K. Ryckman [email protected], 145 N. Riverside Drive, S400 CPHB, Iowa City, Iowa, USA Supplemental data for this article is available online athttps://doi.org/10.1080/14767058.2021.1907332. 2021 Informa UK Limited, trading as Taylor & Francis Group THE JOURNAL OF MATERNAL-FETAL & NEONATAL MEDICINE https://doi.org/10.1080/14767058.2021.1907332 with less than 500 individuals with leukemia and lymphoma. The studies with more than 500 individuals used self-reported outcomes and/or compared out- comes to control groups that included individuals with another cancer-type [7,12,14]. Prior studies have yielded mixed results with some showing an increased risk of preterm birth among those with a history of leukemia/lymphoma [7–10] while others showed no significant risk [6,8,11–14]. We evaluated the relationship using birth-certifica- te-based measures of preterm birth linked with cancer information for two different large population-based US samples. Our primary objective was to evaluate the relationship of diagnosis of leukemia or lymphoma prior to pregnancy with preterm birth. Methods Data source, linkages, and study population We had access to two retrospective administrative data sources, which included women who gave birth from two different geographical areas: the State of California and the State of Iowa. From the State of California, we conducted secondary data analysis on an existing birth certificate-linked dataset with mother and infant hospital discharge information from one year prior to birth (mother only) to one year postdeliv- ery (mother and infant) for births that occurred in California from 2007 to 2012. We have used this data- set for previous studies of preterm birth [15,16]. This dataset is maintained by the California Office of Statewide Health Planning and Development (OSHPD). In Iowa, we developed a new dataset by linking Iowa birth certificates with Surveillance, Epidemiology, and End Results (SEER) cancer registry data covering the State of Iowa. Eligible participants included women who were Iowa residents age 44 or under at the time of a leukemia or lymphoma diagnosis between 1973 and 2018, linked to the first Iowa birth certificate after their cancer diagnosis date. Up to two randomly selected unexposed births were selected by matching on birth month and birth year to each Iowa exposed infant. The unexposed infant had to have a mother who was an Iowa resident and 18 years of age or older at delivery. For both California and Iowa, we only included live births. In both states, we included both singletons and multiples, and births with gestation between 20 and 44 weeks and maternal age of<45 years. Additionally, where possible, we included only the first pregnancy of mothers during the study period and first preg- nancy after diagnosis. This information was notcomplete for the California data. We also excluded women who did not have complete information for the primary outcome of preterm birth and had more than one type of cancer. Study variables The primary outcome variable in this study was pre- term birth. Preterm birth is defined as a gestational age at delivery less than 37 weeks. In both California and Iowa, data on gestational age were obtained from birth certificates. Women who had a gestational age at birth that was 37 weeks were categorized as hav- ing a term birth. The primary exposure variable of this study was a diagnosis of leukemia or lymphoma prior to birth. For California, to identify women with a history of leuke- mia and lymphoma, we used the following ICD-9 codes: 201.x-202.x, 203.1x, 204.x-208.x, V10.6, and V10.7. There have been two validation studies for these ICD-9 codes including V10 history codes although not specifically for pregnancy discharge data [17,18]. The sensitivity ranged from 80% to 90% and positive predictive values from 63% to 76% [17,18]. In Iowa, we used the 3rd edition of theInternational Classification of Diseases forOncology: C024, C098- C099, C111, C142, C379, C422, C770-C779, C420, C421, and C424. The comparison group were women with- out a prior history of cancer. We used birth certificate data from both Iowa and California to capture important covariates including maternal age, maternal race, maternal education, smoking during pregnancy, prior live births, plurality, and gestational hypertension. Maternal race was defined as non-Hispanic White, Asian, Black, Hispanic, or Other race; maternal education was trichotomized by years of education of<12 years, 12 years, and >12 years; smoking during pregnancy was dichotom- ized as either yes or no; prior live birth was catego- rized as 0, 1, 2, and 3 or more; plurality was dichotomized as having singleton and twins or more; and gestational hypertension was defined as a new onset of hypertension during pregnancy and included preeclampsia and eclampsia with Iowa using birth cer- tificate data and California using diagnoses codes of ICD-9 642.1-3, 642.4-642.7. Prior live births was defined as births that occurred prior to the diagnosis of cancer and/or those births that occurred prior to the study period. Overall, there was<5% missing from any one variable, except for prior live births, there was<13% missing in Iowa. Furthermore, Iowa’s SEER registry 2 S. T. ANAND ET AL. data included information about cancer diagnosis and treatment. The Iowa Cancer Registry data allowed us to cap- ture the following covariates: age at leukemia/lymph- oma diagnosis, time since diagnosis to birth (<3 years, 3-5 years, 6-8 years, and 9 or more years), cancer stage (local, regional, distant, or unstaged), cancer treatment (chemotherapy only, chemotherapy and radiation, radiation only, and neither chemotherapy nor radi- ation), hormone therapy (yes/no), immunotherapy (yes/no), chemotherapy (yes/no), and radiation (yes/no). Statistical analysis For both California and Iowa, we used Chi-square tests and t-tests to compare descriptive characteristics for categorical variables and continuous variables, respect- ively. To assess the relationship between preterm birth and leukemia/lymphoma, we used logistic regression models for the California data, and conditional logistic regression models to account for matching in the Iowa data. We also adjusted for potential confounders in the multivariate analyses. Additionally, we evaluated gestational hypertension as a potential mediator in the relationship between preterm birth and leuke- mia/lymphoma. With SEER data providing cancer treatment informa- tion in Iowa, we conducted an Iowa-only analysis to evaluate treatment effects. In both California and Iowa, we also conducted a sensitivity analysis to evalu- ate preterm birth among only singleton pregnancies. Additionally, among women with preterm births, we compared leukemia/lymphoma history for women with spontaneous preterm birth versus women with indicated preterm birth. We also assessed the relation- ship between preterm birth with each cancer, leuke- mia and lymphoma, separately (Supplementary Table A1). All analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC). Ap-value<.05 was considered as statistically significant. Methods and protocols for the study using California data were approved by the Committee for the Protection of Human Subjects within the Health and Human Services Agency of the State of California. De-identified data was provided to the researchers by the California Office of Statewide Health Planning and Development (Protocol # 12–09-0702) and determined not to qualify as human subjects research by the University of Iowa Institutional Review Board (IRB no.: 201602793). For Iowa data, the study was approved by the University of Iowa Institutional Review Board.Data was approved for linkage by the Iowa Department of Public Health (RA 3873) and by the University of Iowa Institutional Review Board (IRB no: 201811805). Results The descriptive characteristics of both the California cohort and the Iowa cohort can be seen inTable 1 and the flowcharts are shown inFigures 1and2.In California, a total of 1024 women had a history of leukemia or lymphoma and 2,468,625 women had no history of cancer coded in the dataset. In Iowa, there was a total of 515 women with a history of leukemia or lymphoma and 1009 unexposed selected with no history of cancer. In both states, women were mostly nonsmokers and had singleton pregnan- cies. Women with a history of leukemia/lymphoma gave birth at an older age and had a higher educa- tion level than women without a history of cancer in both California and Iowa. The racial and ethnic distri- butions of the two samples were consistent with known demographic differences between the two states. In California, where there is more racial and ethnic diversity, particularly notable differences between women with and without a history of leu- kemia or lymphoma were that women with a history of leukemia or lymphoma were less likely to be Hispanic (34.5% vs 50.6%, respectively) or Asian (7.3% vs 12.3%). The cancer and treatment characteristics from the Iowa SEER registry are shown inTable 2. As expected, due to differences in age-specific incidence patterns and treatment options for leukemia and lymphoma, those with a history of leukemia had a more distant history of their cancer (78.6% were diagnosed 9 or more years prior to pregnancy) compared with lymph- oma patients (32.4% were diagnosed 9 or more years prior). Only 9.5% of leukemia patients were diagnosed within 3 years before pregnancy compared with 23.1% of those with prior lymphoma. Lymphoma patients were more likely to receive radiation or chemoradia- tion than leukemia patients who were most likely to receive chemotherapy only. The prevalence of preterm birth in California among those without a history of cancer was 7.8%, compared to 14.6% in those with a history of leuke- mia/lymphoma (unadjusted odds ratio (OR) 2.03; 95% CI 1.71-2.42) (Table 3). The relationship between leuke- mia/lymphoma and preterm birth remained after adjusting for maternal age, race, education, smoking and plurality (OR: 1.89; 95% CI 1.56–2.28). In Iowa, the THE JOURNAL OF MATERNAL-FETAL & NEONATAL MEDICINE 3 Table 1.Descriptive characteristics by history of leukemia or lymphoma of California women who gave birth between 2007 and 2012 and Iowa women who gave birth between 1989 and 2018. CALIFORNIA IOWA Variable a Total (N¼2,469,649)Leukemia/ Lymphoma b(N¼1024) No Cancer (N¼2,468,625) P-value Total (N¼1524)Leukemia/ Lymphoma b(N¼515)No Cancer (N¼1009)pValue Preterm Birth<0.001 .016 Preterm birth (<37 weeks) 192524 (7.8%) 150 (14.6%) 192374 (7.8%) 145 (9.5%) 62 (12.0%) 83 (8.2%) No preterm birth (37weeks or more)2277125 (92.2%) 874 (85.4%) 2276251 (92.2%) 1379 (90.5%) 453 (88.0%) 926 (91.8%) Maternal Age at Birth<0.001 .002 <20 245360 (9.9%) 75 (7.3%) 245285 (9.9%) 92 (6.0%) 34 (6.6%) 58 (5.7%) 20-24 530999 (21.5%) 177 (17.3%) 530822 (21.5%) 359 (23.6%) 92 (17.9%) 267 (26.5%) 25-29 659507 (26.7%) 262 (25.6%) 659245 (26.7%) 505 (33.1%) 167 (32.4%) 338 (33.5%) 30-34 608744 (24.6%) 281 (27.4%) 608463 (24.6%) 400 (26.2%) 157 (30.5%) 243 (24.1%) 35-39 340692 (13.8%) 178 (17.4%) 340514 (13.8%) 138 (9.1%) 52 (10.1%) 86 (8.5%) 40-44 84347 (3.4%) 51 (5.0%) 84296 (3.4%) 30 (2.0%) 13 (2.5%) 17 (1.7%) Maternal Age (Continuous)<0.001 .003 mean and std 28.1 (6.3) 29.2 (6.2) 28.1 (6.3) 27.8 (5.4) 28.3 (5.4) 27.5 (5.3) median and IQR 28.0 (23.0, 33.0) 29.0 (25.0, 34.0) 28.0 (23.0, 33.0) 28.0 (24.0, 32.0) 28.0 (25.0, 32.0) 27.0 (23.0, 31.0) min and max (13.0, 44.0) (13.0, 44.0) (13.0, 44.0) (16.0, 43.0) (16.0, 43.0) (18.0, 43.0) Maternal Race/Ethnicity<0.001<.001 Asian 304811 (12.3%) 75 (7.3%) 304736 (12.3%) ccc Black 124112 (5.0%) 53 (5.2%) 124059 (5.0%) 55 (3.6%) 13 (2.5%) 42 (4.2%) Hispanic 1249865 (50.6%) 353 (34.5%) 1249512 (50.6%) 70 (4.6%) 9 (1.7%) 61 (6.0%) Other race 179609 (7.3%) 97 (9.5%) 179512 (7.3%) 48 (3.2%) 8 (1.6%) 40 (4.0%) Non-Hispanic White 611252 (24.8%) 446 (43.6%) 610806 (24.7%) 1351 (88.6%) 485 (94.2%) 866 (85.8%) Smoking History During Pregnancy0.016 .006 No smoking 2360594 (95.6%) 963 (94.0%) 2359631 (95.6%) 1298 (85.2%) 455 (88.3%) 843 (83.5%) Smoked during pregnancy 109055 (4.4%) 61 (6.0%) 108994 (4.4%) 215 (14.1%) 55 (10.7%) 160 (15.9%) Prior Live Births<0.001<.001 0 1157853 (46.9%) 555 (54.2%) 1157298 (46.9%) 433 (28.4%) 195 (37.9%) 238 (23.6%) 1 669706 (27.1%) 267 (26.1%) 669439 (27.1%) 495 (32.5%) 155 (30.1%) 340 (33.7%) 2 382400 (15.5%) 123 (12.0%) 382277 (15.5%) 257 (16.9%) 60 (11.7%) 197 (19.5%) 3 or more 258159 (10.5%) 79 (7.7%) 258080 (10.5%) 151 (9.9%) 32 (6.2%) 119 (11.8%) Maternal Education<0.001<.001 <12 years 636044 (25.8%) 135 (13.2%) 635909 (25.8%) 132 (8.7%) 29 (5.6%) 103 (10.2%) 12 years 625588 (25.3%) 215 (21.0%) 625373 (25.3%) 369 (24.2%) 101 (19.6%) 268 (26.6%) >12 years 1113706 (45.1%) 628 (61.3%) 1113078 (45.1%) 1013 (66.5%) 383 (74.4%) 630 (62.4%) Plurality<0.001 .178 Singleton 2425843 (98.2%) 988 (96.5%) 2424855 (98.2%) 1479 (97.0%) 504 (97.9%) 975 (96.6%) Twins and more 43806 (1.8%) 36 (3.5%) 43770 (1.8%) 45 (3.0%) 11 (2.1%) 34 (3.4%) aThe data source for gestational hypertension was from hospital discharge diagnoses for California and from birth certificate for Iowa. All other variables from both states came from birth certificates.bCalifornia: 324 Leukemia, 700 Lymphoma; Iowa: 126 Leukemia, 389 Lymphoma.cMaternal Race/Ethnicity: In Iowa, Asian was grouped with“Other race”due to<6 cell count. 4 S. T. ANAND ET AL. prevalence of preterm birth among women without a cancer history was 8.2%, compared to 12.0% in women with a history of leukemia/lymphoma (unadjusted OR 1.50; 95% CI 1.07–2.11) (Table 3) and this remained after adjusting for maternal age, race, education, smoking and plurality (OR: 1.61; 95% CI 1.10–2.37). There was no evidence of mediation through gestational hypertension as the odds ratio did not change in either Iowa (OR: 1.61 vs 1.56) or California (OR: 1.89 vs. 1.89) when this variable was added to the models (Table 3). In analyses aimed at examining the potential increased risk associated with cancer treatments in Iowa (Table 4), sample sizes were small and estimates imprecise. Although no statistically significant differen- ces in the odds of preterm birth by cancer treatment were observed, there was an indication that a history of radiation (14.8% had preterm birth) or chemo-radi- ation (16.9% had preterm birth) treatment was associ- ated with greater adjusted risk of preterm birth compared with treatment using chemotherapy alone (9.6% had preterm birth). In our analyses evaluating the relationship between preterm birth and each of the cancer types separately in both California and Iowa (Supplementary Table S1), increased risk was seen with each cancer type, and especially with lymphoma. This indicates that results may have been driven by those with a history of lymphoma. However, leukemia had a smaller sample size, which led to imprecise estimates. It has been shown that the risk for preterm birth is higher in multiples compared to singletons [19,20]. To address this, we conducted sensitivity analysis to assess the risk of preterm birth among women with a history of leukemia/lymphoma com- pared to women without a history of cancer using only singleton pregnancies. In both California and Iowa, the adjusted odds ratios did not appreciably change when restricted to only singletons (California: OR¼1.94 (95% CI 1.59–2.35); Iowa: OR¼1.77 (95% CI 1.19–2.64)). Additionally, we evaluated spon- taneous and indicated preterm birth only among those with a preterm birth. In California, there was no significant difference in odds of a leukemia/ lymphoma history between those with spontaneous preterm birth and those with indicated preterm birth (OR 0.81; 95% CI 0.52–1.26). In Iowa, due to inad- equate sample size in sub-group analyses, we were not able to differentiate between spontaneous and indicated preterm birth as the model would not converge. Total observations: 18367 Total after excluding observations that were not the first cancer: 17718 Maternal age <45 years of age: 17672 Total after excluding observations that were not the first birth: 13733 Only leukemia/lymphoma and unexposed subjects: 1560 Other cancers and their unexposed and those in the exposed group without a match: 12173 Total subjects in study: 1524 Excluded: - 31 with chronic hypertension - 5 with unknown preterm birth status Figure 1.Flowchart of Iowa dataset. Total observations: 3059186 Total after excluding observations that were not the first baby: 2531478 Maternal age <45 years of age: 2524811 Only leukemia/lymphoma and no cancer history: 2520574 Total subjects in study: 2469649 Excluded: - 50925 with chronic hypertension - 0 with unknown preterm birth status Figure 2.Flowchart of California dataset. THE JOURNAL OF MATERNAL-FETAL & NEONATAL MEDICINE 5 Discussion In two unique populations (Iowa and California), we found that there was a statistically significant increase in the odds of preterm birth among women with a prior history of leukemia or lymphoma compared to women without any prior history of cancer. Our find- ings are consistent with previous studies assessing the risk of preterm birth in women with a history leuke- mia/lymphoma (Supplementary Table S2). Although there was variability in outcome ascertainment between the studies, the findings were consistent in directionality and magnitude. The risk estimates for preterm birth in these other studies were consistent with ours and ranged from 1.50 to 2.60 for leukemia and 1.59 to 2.11 for lymphoma [6–10]. Specifically, in a study by Haggar et al., the risk for preterm birthamong leukemia patients was 1.72 (95% CI 1.18–2.41), and in another study by Anderson et al., there was an increased risk of 1.59 (95% CI 1.06–2.37) among Hodgkin lymphoma patients [8,10]. Although these studies assessed leukemia and lymphoma separately, they both found results of the same magnitude and direction as we did. There were, however, other studies that found no statistically significant relationship between preterm birth and leukemia/lymphoma with risks ranging from 0.4–1.54 [6,8,11–14]. Some of these used self-reported outcome data from questionnaires, were conducted mostly outside of the United States and included mostly women diagnosed between 15–39 years of age. Possible explanations for the increase in preterm births among women with a history of leukemia or Table 2.Cancer and treatment characteristics of Iowa study sample who were diagnosed between 1973 and 2018 by can- cer type. Variable Description Data Source a Leukemia (N¼126) c Lymphoma (N¼389) c Cancer Stage Local ICR 56 (14.4%) Regional 139 (35.7%) Distant 126 (100.0%) 80 (20.6%) Unstaged 114 (29.3%) Age At Cancer Diagnosis<5 ICR 37 (29.4%)– 5-9 36 (28.6%) 11 (2.8%) 10–14 22 (17.5%) 36 (9.3%) 15–19 11 (8.7%) 79 (20.3%) 20–24 9 (7.1%) 129 (33.2%) 25–29 7 (5.6%) 81 (20.8%) 30–44 S 53 (13.6%) Hormone Treatment None ICR 22 (17.5%) 239 (61.4%) Yes 104 (82.5%) 150 (38.6%) Immune Treatment None ICR 118 (93.7%) 376 (96.7%) Yes 8 (6.3%) 13 (3.3%) Time From Diagnosis To Delivery<3 years CALCULATED FROM ICR AND BC 12 (9.5%) 90 (23.1%) 3–5 years 8 (6.3%) 99 (25.4%) 6–8 years 7 (5.6%) 74 (19.0%) 9þyears 99 (78.6%) 126 (32.4%) Chemotherapy No ICR S 86 (22.1%) Yes 123 (97.6%) 303 (77.9%) Radiation Treatment No ICR 92 (73.0%) 182 (46.8%) Yes 34 (27.0%) 207 (53.2%) Cancer Treatment Breakdown b Chemotherapy only GROUPED FROM ICR 89 (70.6%) 161 (41.4%) Both Chemotherapy and Radiation 34 (27.0%) 142 (36.5%) Radiation only S 65 (16.7%) Neither Radiation nor Chemotherapy S 21 (5.4%) aData source: ICR- Iowa Cancer Registry.bA total of 93 women received surgery, typically coded as lymph node surgery.cS¼suppressed cells (<6 cell count). Table 3.Risk of preterm birth among women<45 years of age with leukemia/lymphoma who gave birth, by state. California Iowa NWith preterm birth N With preterm birth Leukemia/Lymphoma,N(%) 1024 150 (14.6%) 515 62 (12.0%) No Cancer,N(%) 2,468,625 192,374 (7.8%) 1,009 83 (8.2%) Unadjusted model, OR (95% CI) 2.03 (1.71, 2.42) 1.50 (1.07, 2.11) Model 1: adjusted for age, race, education, plurality, smoking OR (95% CI)1.89 (1.56, 2.28) 1.61 (1.10, 2.37) Model 2: adjusted for covariates in Model 1 plus gestational hypertension (OR (95%CI)1.89 (1.56, 2.29) 1.56 (1.05, 2.30) p<.05. 6 S. T. ANAND ET AL. lymphoma include the late-effects of cancer treatment. Chemotherapy drugs such as platinum agents and anthracyclines can cause renal impairments and cardi- otoxicities [21–24]. The renal impairments can lead to hypertension and preterm birth [25,26]. Additionally, obesity is a risk factor for preterm birth, and cortico- steroids, which are medications used in the treatment for cancer that can cause obesity [27–29]. The exact mechanism for corticosteroid and increases in body mass index are unknown, but a potential mechanism is that corticosteroids could cause changes in fat dis- tribution and metabolism and increase gluconeogene- sis [30,31]. Future studies, with carefully collected and longitudinal data on BMI before and during preg- nancy, are needed to investigate these potential mechanisms leading to preterm birth in women with a history of leukemia or lymphoma. Furthermore, radiotherapy, especially to the abdom- inopelvic region such as for some lymphoma patients, can potentially damage the vagina, uterus, and/or ova- ries and thus lead to vaginal stenosis and fibrosis, uterine vasculature and musculature damage, and pre- mature ovarian insufficiency [32–38]. This damage can lead to adverse birth outcomes such as preterm birth [35,36]. Potential limitations of our study include misclassifi- cation of cancer history, lack of details on cancer type and treatment characteristics in the California data, and insufficient power for separate analyses of spon- taneous and medically indicated preterm birth. The California study used ICD-9 codes including V10 his- tory codes from the birth discharge abstract (86% of patients) and any discharges in the year prior to birth (14% of patients) to determine our primary exposure of leukemia or lymphoma. This could have potentially led to misclassification of our exposure by missing leu- kemia/lymphoma history in some subjects. Also, since we did not have the cancer date of diagnosis in California, we could not distinguish between women who had a recent cancer diagnosis, including some who were actively being treated for leukemia/ lymphoma during pregnancy, and women with a diag- nosis longer before pregnancy. In Iowa, we also didnot have the power to further stratify by different age groups and time intervals between diagnosis and childbirth. Additionally, in both California and Iowa we were unable to assessin vitrofertilization, which is a known risk factor for preterm birth and a potential mediator of the relationship between cancer and pre- term birth, given the observation that cancer patients are more likely to receivein vitrofertilization. Also, another limitation was that in Iowa, we were only approved to match on birth month and year for obtaining our unexposed births and could not match on other important factors such as maternal and paternal age at childbirth. Finally, we did not have adequate statistical power to assess indicated versus spontaneous preterm birth. However, the damage to the uterus and vagina caused by cancer treatment and impairments such as uterine vasculature and mus- culature damage, uterine fibrosis and cervical shorten- ing can potentially lead to either spontaneous or medically indicated preterm birth [32–40]. Despite these limitations, our study yielded results similar to previously conducted studies and it should be noted that any missed cancer diagnoses in the control group would have led to a dampening of our risk estimates which further bolsters our findings. The replication of findings in two large samples– the whole States of Iowa and California–is a strength of this study. The racial/ethnic diversity of the popula- tions, one a primarily urban population and the other a mix of urban and rural, and the two distinct meth- ods of measuring leukemia/lymphoma history, enhance generalizability of the findings. Additional strengths include use of birth certificate data for the outcome of preterm birth for both states, which is an improvement on studies that relied on self-reported data. We were also able to ascertain complete infor- mation on cancer diagnoses and treatments from the Iowa SEER Cancer Registry. Overall, our study found that there was an increased risk of preterm birth among women with a history of leukemia or lymphoma in both Iowa and California. Though challenging, it would be beneficial for additional studies to be conducted based on data Table 4.Risk of preterm birth among Iowa women<45 years of age with leukemia/lymphoma who gave birth between 1989 and 2018 by cancer treatment. Unadjusted model Adjusted model a Cancer Treatment b Total With preterm birth,N(%) c OR (95% Confidence Interval (CI)) OR (95% Confidence Interval (CI)) Chemotherapy only 250 24 (9.6%) Ref Ref Both chemotherapy and radiation 176 26 (14.8%) 1.63 (0.90, 2.95) 1.72 (0.89, 3.33) Radiation only 65 11 (16.9%) 1.92 (0.89, 4.16) 1.62 (0.66, 3.96) Neither radiation nor chemotherapy 24 S 0.41 (0.05, 3.17) 0.33 (0.04, 2.77) Adjusted for time from diagnosis to delivery, diagnosis age, and cancer stage. bA total of 93 women received surgery, typically coded as lymph node surgery.cS¼suppressed cells (<6 cell count). THE JOURNAL OF MATERNAL-FETAL & NEONATAL MEDICINE 7 with verified pregnancy outcomes, cancer and its treatments, and further assessment of the risk of pre- term birth in leukemia and lymphoma patients separ- ately and with each subtype such as acute lymphoblastic leukemia and non-Hodgkin’s lymphoma. Future studies also need to address how leukemia or lymphoma impact fertility decisions and fecundity. Our study supports the importance of early identification of pregnant women and newborns at risk for compli- cations, which should then inform preventive interven- tions. Moreover, the ability to characterize and understand the contributors to adverse birth out- comes such as preterm birth and complications in newborns is important as it provides the potential for improved, tailored prenatal care as women can be provided more information on their risks, options, and opportunities to prepare for an early birth. Acknowledgements The authors express our sincere thanks to Dr. Paul Romitti and his team at the Iowa Registry for Congenital and Inherited Disorders (IRCID)/Iowa vital records for their assist- ance with the linkage to the birth certificate data. The authors also like to express our gratitude to Jason Brubaker at the Iowa Cancer Registry for his assistance on providing the SEER Registry data and for linkage with the birth certifi- cate data in Iowa. This research is supported by the National Cancer Institute (P30 CA086862-18S6). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health. Disclosure statement This research is supported by the National Cancer Institute under Grant number P30 CA086862-18S6. Dr. Kelli Ryckman and Dr. Sonia Anand received the grant during the conduct of the study. Additionally, Dr. Ryckman has a patent for Serum Screening and Lipid Markers Predicting Preterm Birth pending. All other authors report no conflict of interest. Author contributions Sonia T. Anand: Conceptualization, Formal analysis, Writing– original draft, Writing–review & editing; Elizabeth A. Chrischilles: Conceptualization, Project administration, Supervision, Writing–review & editing; Rebecca J. Baer: Data curation, Writing–review and editing; Mary E. Charlton: Conceptualization, Writing–review and editing; Patrick J. Breheny: Conceptualization, Writing–review and editing; William W. Terry: Conceptualization, Writing–review and editing; Monica R. McLemore: Writing–review and edit- ing; Deborah A. Karasek: Writing–review and editing; Laura L. Jelliffe-Pawlowski: Writing–review and editing; Kelli K. Ryckman: Conceptualization, Funding acquisition, Project administration, Supervision, Writing–review & editing. ORCID Sonia T. Anand http://orcid.org/0000-0002-5494-1821 Monica R. McLemore http://orcid.org/0000-0001- 6539-4256 References [1] Miller KD, Nogueira L, Mariotto AB, et al. Cancer treat- ment and survivorship statistics, 2019. CA A Cancer J Clin. 2019;69(5):363–385. [2] Organization WH. Preterm Birth 2020. [Available from: https://www.who.int/news-room/fact-sheets/detail/ preterm-birth. [3] Prevention CfDCa. Preterm Birth 2020. [Available from:https://www.cdc.gov/reproductivehealth/mater- nalinfanthealth/pretermbirth.htm. [4] Callaghan WM, MacDorman MF, Rasmussen SA, et al. The contribution of preterm birth to infant mortality rates in the United States. Pediatrics. 2006;118(4): 1566–1573. [5] Blencowe H, Cousens S, Chou D, et al.; Born Too Soon Preterm Birth Action Group. Born too soon: the global epidemiology of 15 million preterm births. Reprod Health. 2013;10 Suppl 1:S2. [6] Hartnett KP, Ward KC, Kramer MR, et al. The risk of preterm birth and growth restriction in pregnancy after cancer. Int J Cancer. 2017;141(11):2187–2196. [7] Signorello LB, Cohen SS, Bosetti C, et al. Female survi- vors of childhood cancer: preterm birth and low birth weight among their children. J Natl Cancer Inst. 2006; 98(20):1453–1461. [8] Haggar FA, Pereira G, Preen D, et al. Adverse obstetric and perinatal outcomes following treatment of ado- lescent and young adult cancer: a population-based cohort study. PLoS One. 2014;9(12):e113292. [9] Mueller BA, Chow EJ, Kamineni A, et al. Pregnancy outcomes in female childhood and adolescent cancer survivors: a linked cancer-birth registry analysis. Arch Pediatr Adolesc Med. 2009;163(10):879–886. [10] Anderson C, Engel SM, Mersereau JE, et al. Birth Outcomes Among Adolescent and Young Adult Cancer Survivors. JAMA Oncol. 2017;3(8):1078–1084. [11] Madanat-Harjuoja LM, Malila N, L€ ahteenm€ aki PM, et al. Preterm delivery among female survivors of childhood, adolescent and young adulthood cancer. Int J Cancer. 2010;127(7):1669–1679. [12] Reulen RC, Bright CJ, Winter DL, et al. Pregnancy and labor complications in female survivors of childhood cancer: The British Childhood Cancer Survivor Study. J Natl Cancer Inst. 2017;109(11):djx056. [13] Swerdlow AJ, Jacobs PA, Marks A, et al. Fertility, reproductive outcomes, and health of offspring, of patients treated for Hodgkin’s disease: an investiga- tion including chromosome examinations. Br J Cancer. 1996;74(2):291–296. [14] Reulen RC, Zeegers MP, Wallace WH, et al.; British Childhood Cancer Survivor Study. Pregnancy out- comes among adult survivors of childhood cancer in the British Childhood Cancer Survivor Study. Cancer Epidemiol Biomarkers Prev. 2009;18(8):2239–2247. 8 S. T. ANAND ET AL. [15] Baer RJ, Chambers CD, Ryckman KK, et al. Risk of pre- term and early term birth by maternal drug use. J Perinatol. 2019;39(2):286–294. [16] Smith CJ, Baer RJ, Oltman SP, et al. Maternal dyslipi- demia and risk for preterm birth. PLoS One. 2018; 13(12):e0209579-e. [17] Herman RA, Gilchrist B, Link BK, et al. A systematic review of validated methods for identifying lymph- oma using administrative data. Pharmacoepidemiol Drug Saf. 2012;21(Suppl 1):203–212. [18] Clarke CL, Feigelson HS. Developing an Algorithm to Identify History of Cancer Using Electronic Medical Records. EGEMS (Washington, DC). 2016;4(1):1209. [19] Kurdi AM, Mesleh RA, Al-Hakeem MM, et al. Multiple pregnancy and preterm labor. Saudi Med J. 2004; 25(5):632–637. [20] Stock S, Norman J. Preterm and term labour in mul- tiple pregnancies. Semin Fetal Neonatal Med. 2010; 15(6):336–341. [21] Barry E, Alvarez JA, Scully RE, et al. Anthracycline- induced cardiotoxicity: course, pathophysiology, pre- vention and management. Expert Opin Pharmacother. 2007;8(8):1039–1058. [22] Lipshultz SE, Adams MJ. Cardiotoxicity after childhood cancer: beginning with the end in mind. J Clin Oncol. 2010;28(8):1276–1281. [23] Lipshultz SE, Adams MJ, Colan SD, et al.; American Heart Association Congenital Heart Defects Committee of the Council on Cardiovascular Disease in the Young, Council on Basic Cardiovascular Sciences, Council on Cardiovascular and Stroke Nursing, Council on Cardiovascular Radiolo. Long- term cardiovascular toxicity in children, adolescents, and young adults who receive cancer therapy: patho- physiology, course, monitoring, management, preven- tion, and research directions: a scientific statement from the American Heart Association. Circulation. 2013;128(17):1927–1995. [24] Cardous-Ubbink MC, Geenen MM, Schade KJ, et al. Hypertension in long-term survivors of childhood can- cer: a nested case-control study. Eur J Cancer (Oxford, England: 1990). 2010;46(4):782–790. [25] Bramham K, Parnell B, Nelson-Piercy C, et al. Chronic hypertension and pregnancy outcomes: systematic review and meta-analysis. BMJ. 2014;348(apr15 7): g2301–g2301. [26] Catov JM, Nohr EA, Olsen J, et al. Chronic hyperten- sion related to risk for preterm and term small for gestational age births. Obstet Gynecol. 2008;112(2 Pt 1):290–296.[27] Cnattingius S, Villamor E, Johansson S, et al. Maternal obesity and risk of preterm delivery. JAMA. 2013; 309(22):2362–2370. [28] Lopez-Jaramillo P, Barajas J, Rueda-Quijano SM, et al. Obesity and preeclampsia: common pathophysio- logical mechanisms. Front Physiol. 2018;9:1838 [29] McDonald SD, Han Z, Mulla S, et al.; on behalf of the Knowledge Synthesis Group. Knowledge Synthesis G. Overweight and obesity in mothers and risk of pre- term birth and low birth weight infants: systematic review and meta-analyses. BMJ (Clinical Research ed). 2010;341(jul20 1):c3428–c., [30] Chow EJ, Pihoker C, Hunt K, et al. Obesity and hyper- tension among children after treatment for acute lymphoblastic leukemia. Cancer. 2007;110(10): 2313–2320. [31] Munck A, Naray-Fejes-Toth A. Glucocorticoid physi- ology. In: DeGroot LJ, Jameson JL, editors. Endocrinology. Philadelphia: Elsevier Saunders; 2006. [32] Emily G, Alka P, Jane M, et al. Reproductive health in female survivors of childhood cancer. Obstet Gynaecol. 2016;18(4):315–322. [33] Galuni c-Bili cL, Santek F. Infradiaphragmal radiother- apy in patients with lymphoma: volume definition and side effects. Acta Clin Croat. 2018;57(3):554–560. [34] Manzella A, Borba-Filho P, D’Ippolito G, et al. Abdominal manifestations of lymphoma: spectrum of imaging features. ISRN Radiol. 2013;2013:483069 [35] Chiarelli AM, Marrett LD, Darlington GA. Pregnancy outcomes in females after treatment for childhood cancer. Epidemiology (Cambridge, Mass.). 2000;11(2): 161–166. [36] Green DM, Hall B, Zevon MA. Pregnancy outcome after treatment for acute lymphoblastic leukemia dur- ing childhood or adolescence. Cancer. 1989;64(11): 2335–2339. [37] Bath LE, Critchley HO, Chambers SE, et al. Ovarian and uterine characteristics after total body irradiation in childhood and adolescence: response to sex steroid replacement. Br J Obstet Gynaecol. 1999;106(12): 1265– 1272. [38] Green DM, Sklar CA, Boice JD, Jr, et al. Ovarian failure and reproductive outcomes after childhood cancer treatment: results from the Childhood Cancer Survivor Study. J Clin Oncol. 2009;27(14):2374–2381. [39] Melin J, Heinavaara S, Malila N, et al. Risk factors for preterm delivery among early onset cancer survivors: A Finnish register-based study. Int J Cancer. 2019; 144(8):1954–1961. [40] Vink J, Feltovich H. Cervical etiology of spontaneous preterm birth. Semin Fetal Neonatal Med. 2016;21(2): 106–112. THE JOURNAL OF MATERNAL-FETAL & NEONATAL MEDICINE 9

Writerbay.net

We offer the best essay writing services to students who value great quality at a fair price. Let us exceed your expectations if you need help with this or a different assignment. Get your paper completed by a writing expert today. Nice to meet you! Want 15% OFF your first order? Use Promo Code: FIRST15. Place your order in a few easy steps. It will take you less than 5 minutes. Click one of the buttons below.


Order a Similar Paper Order a Different Paper