Protocol for an Observational Study on the Eects of Giving Births from Unintended Pregnancies on Later Life Physical and Mental Health

2025-04-15 0 0 530.01KB 30 页 10玖币

侵权投诉

Protocol for an Observational Study on the Eﬀects of Giving

Births from Unintended Pregnancies on Later Life Physical

and Mental Health

Samrat Roy

Department of Statistics and Data Science, University of Pennsylvania

email: roysa@wharton.upenn.edu

Marina Bogomolov

Industrial Engineering and Management, Technion - Israel Institute of Technology

email: marinabo@technion.ac.il

Ruth Heller

Department of Statistics and Operations Research, Tel-Aviv University

email: ruheller@tauex.tau.ac.il

Amy M. Claridge

Child Development and Family Science, Central Washington University

email: claridgea@cwu.edu

Tishra Beeson

Department of Health Sciences, Central Washington University

email: Tishra.Beeson@cwu.edu

Dylan S. Small

Department of Statistics and Data Science, University of Pennsylvania

email: dsmall@wharton.upenn.edu

June 2022

Abstract

There has been increasing interest in studying the eﬀect of giving births to unin-

tended pregnancies on later life physical and mental health. In this article, we provide

the protocol for our planned observational study on the long-term mental and physical

health consequences for mothers who bear children resulting from unintended preg-

nancies. We aim to use the data from the Wisconsin Longitudinal Study (WLS) and

examine the eﬀect of births from unintended pregnancies on a broad range of outcomes,

including mental depression, psychological well-being, physical health, alcohol usage,

and economic well-being. To strengthen our causal ﬁndings, we plan to address our

research questions on two subgroups, Catholics and non-Catholics, and discover the

arXiv:2210.05169v3 [stat.AP] 1 May 2023

“replicable” outcomes for which the eﬀect of unintended pregnancy is negative (or, pos-

itive) in both subgroups. Following the idea of non-random cross-screening, the data

will be split according to whether the woman is Catholic or not, and then one part of

the data will be used to select the hypotheses and design the corresponding tests for the

second part of the data. In past use of cross screening (automatic cross screening) there

was only one team of investigators that dealt with both parts of the data so that the

investigators would need to decide on an analysis plan before looking at the data. In

this protocol, we describe our analysis plan for carrying out automatic cross screening

to study the eﬀects of unintended pregnancy. In addition, we describe plans to carry

out a novel ﬂexible cross-screening in which there will be two teams of investigators

with access only to one part of data and each team will use their part of the data to

decide how to plan the analysis for the second team’s data. In addition to the above

replicability analysis, we also discuss the plan to test the global null hypotheses, in

order to identify outcomes which are aﬀected by unintended pregnancy for at least one

of the two subgroups of Catholics and non-Catholics.

1 Background and Motivation

Unintended pregnancies are prevalent in the United States (U.S.) and worldwide. A preg-

nancy may be considered unintended when the pregnancy was not wanted, or when it was

mistimed or earlier than desired (CDC, 2021). 45% of all U.S. pregnancies were unintended

in 2011 (Finer and Zolna (2016)), with higher rates of unintended pregnancies among ado-

lescents and young adults (Postlethwaite et al. (2010)), among women living at or below

poverty (Finer and Zolna (2016)), among unmarried individuals (Musick (2002)) and among

non-Hispanic black and African American women (Finer and Zolna (2016)). Unintended

pregnancy rates have been estimated at 44%, worldwide (Bearak et al. (2018)).

There has been an escalated interest in studying the eﬀect of unintended pregnancies

on the later life mental and physical health of the women (Bahk et al. (2015), Herd et al.

(2016), Barton et al. (2017)). This eﬀect can either be attributed to the termination of

the unintended pregnancy, or to the continuation of the unintended pregnancy to term.

While the former has been discussed by a robust body of literature (see Herd et al. (2016)

and references therein), there has not been much developments on the latter one. Birthing

individuals with unintended pregnancies tend to report less emotional attachment to their

baby in pregnancy (Pakseresht et al. (2018)), needing more time to accept their pregnancy.

After birth, parents of unintended pregnancies are less likely to initiate breastfeeding (Mark

and Cowan (2022)), tend to report less expression of aﬀection (Hayatbakhsh et al. (2011))

and issues in attachment formation (Miller et al. (2009)). Also, they are more likely to report

parenting stress at 6 months, one year postpartum, and to engage in less eﬀective parenting

strategies (Miller et al. (2009), East et al. (2012)). Thus, a robust large-scale study on the

physical and mental health consequences of the mothers who continued their unintended

pregnancies to term is of critical importance. Most of the existing literature in this regard,

considered the births only after Roe v. Wade (1973)1, which protected the liberty of a

1see https://en.wikipedia.org/wiki/Roe_v._Wade

pregnant woman to choose to have an abortion, and 40% of the post-Roe v. Wade (1973)

unintended pregnancies were terminated. Thus, the existing literature, that consider the

births only after Roe v. Wade (1973), can hardly capture the actual eﬀect of unintended

pregnancies on later life physical and mental health. This gap in the literature, along with

the recent overturning of Roe v. Wade by the Supreme Court2, raises the question, “What

are the long-term mental and physical health consequences for mothers who bear children

resulting from unintended pregnancies?”.

In this study, we address the above question by using data from Wisconsin Longitudinal

Study (WLS) (Herd et al. (2014)), wherein the respondents were the women who graduated

from Wisconsisn High School in 1957, and the survey data on various aspects of those respon-

dents’ life course were collected in 1957,1964,1975,1992,2004 and 2011. Herd et al. (2016)

used the same data and analyzed the later life physical and mental health consequences

of the women who gave births to unintended pregnancies. As they pointed out, the WLS

data has some advantageous features which alleviate the aforementioned drawbacks in the

existing literature. First, unlike the previous work on the eﬀects of unwanted pregnancies,

WLS respondents had experienced nearly all their pregnancies before the 1973 Roe v. Wade

decision. Thus, most, if not all, of these women did not have the opportunity to terminate

an unintended pregnancy and hence this data is more reliable for inferring the actual eﬀect

of giving births to unintended pregnancies on later life physical and mental health. Second,

WLS data consists of a wide range of covariates, including the family background, adoles-

cent characteristics, educational and occupational achievement and aspirations, which could

potentially confound the relationship between unplanned pregnancies and later-life mental

and physical health outcomes. We use these variables to create matched set of treatment

and control individuals and then compare the physical and mental health outcomes within

each matched set. This is a standard practice as performing a randomized control study

is highly unfeasible in this case, and we need to depend solely on the observational data.

Finally, this data, unlike the other relevant ones, tracks the information longitudinally at

multiple time points, and this facilitates the study of later life physical and mental health

consequences.

We suggest our own statistical design in order to answer the research questions. Our sug-

gested design addresses potential biases by planning to carry out both a replicability analysis

and a sensitivity analysis. The replicability analysis will be possible since we address the

research questions on two subgroups, Catholics and non-Catholics. More speciﬁcally, we

aim to discover the outcomes for which the eﬀect of unintended pregnancy is negative in

both Catholic and non-Catholic subgroups, as well as the outcomes for which the eﬀect is

positive in both subgroups. These outcomes are usually referred to as “replicable” ﬁndings

(Bogomolov and Heller (2022)). When treatments are not randomly assigned, the evidence

that the treatment is the cause of its ostensible eﬀects is strengthened by showing that peo-

ple who receive the treatment for diﬀerent reasons experience similar eﬀects (Rosenbaum

(2015)). Catholics and non-Catholics may have had unintended pregnancies for somewhat

diﬀerent reasons. The Catholic Church opposes birth control while most other faiths do not.

2see https://www.npr.org/2022/06/24/1102305878/supreme-court-abortion-roe-v-wade-decision-overturn

Consequently, Catholic women who have unintended pregnancies are relatively more likely

to have them because their religious beliefs forbid birth control while non-Catholic women

are relatively more likely to have unintended pregnancies because they chose not to use birth

control. We employ the idea of directional replicability, introduced in Bogomolov and Heller

(2018), in order to achieve this goal (see Section 4for more details). We use non-random

cross-screening, wherein, the data is split according to whether the woman is Catholic or

not. One part of the data is used to select the hypotheses to be tested based on the second

part of the data, and then to design the corresponding tests. The scenario, wherein, the

same team of investigators deals with both parts of the data, is referred to as “Automated

cross-screening”. In such cases, to guarantee family wise error rate (FWER) control, the

investigators need to decide before seeing the data how they will use one part of the data

to design the analysis of the second part. On the contrary, when there are two teams of

investigators, each having access only to one part of the data, then each team can see their

part of the data and decide how to plan the analysis for the second team’s data. This is

referred to as “Flexible cross-screening”. In Section 4, we provide a detailed description on

how we use these non-random cross-screening methods to identify the replicable outcomes

with positive (or, negative) treatment eﬀect in both the Catholic and the non-Catholic sub-

groups. In addition to the above replicability analysis, we also discuss the plan to test the

global null hypothesis in Appendix A, that is intended to identify the outcomes which are

aﬀected by unintended pregnancy for at least one of the two subpopulations of Catholics

and non-Catholics.

The remainder of the document is organized as follows. Section 2discusses the match-

ing method that we use to prepare the matched set of treated and control individuals for

both the Catholics and non-Catholics subgroups. In Section 3, we summarize the later-life

mental, physical and economic outcomes on which we test the eﬀect of unintended preg-

nancies. Section 4and Appendix Adiscuss the details of our proposed testing design for

replicability analysis and global null respectively. Finally we conclude with some simulation

studies in Appendix B, which is followed by some additional tables in Appendix C.

2 Risk-set matching

We employ Risk Set Matching (Li et al. (2001); Zubizarreta et al. (2014)) separately for

both the Catholic and non-Catholic women, and deal with the potential confounding in each

subgroup. Note that diﬀerent women gave births to unintended pregnancies at diﬀerent

years. Thus, unlike the randomized experiments, where all the subjects are assigned to

treatment or control at some ﬁxed time point, in this case, there is no such ﬁxed time

of treatment assignment. While some women can give birth to unintended pregnancies

now (and are assigned to the treatment group now), others can give births to unintended

pregnancies years later (and will be assigned to the treatment group then) or, may not

have it at all. Also, the individuals not giving birth to any unintended pregnancies yet

(thus currently assigned to the control group), may end up giving births to unintended

pregnancies later and then get assigned to the treatment group. Risk-set matching, proposed

by Li et al. (2001), enables us to deal with the aforementioned temporal structure of the

treatment assignment. The matching algorithm acts sequentially over time, and at each

time point it pairs two individuals from two diﬀerent groups. The ﬁrst group of subjects

are the ones who just got treated for the ﬁrst time at that time point, while the second

group consists of the ones who have not been treated yet. While pairing the individuals, the

matching is performed based on the observed covariates prior to that time point, ensuring

that the two individuals looked similar in terms of covariates till the point when one of

them received treatment for the ﬁrst time, and the other one did not receive the treatment

yet. The term ‘risk-set’ relates to the Cox’s proportional hazards model, where the two

individuals will have ‘similar’ time-dependent covariates and thus will be at similar ‘risk’ of

receiving the treatment just before the moment one of them actually receives it.

What follow next, are the description of the covariates based on which the women were

matched and then the detailed steps of the risk-set matching.

Covariates: The women are matched on some time-varying covariates and some ﬁxed

covariates as listed below.

Childhood measures: We consider the following four childhood measures as ﬁxed co-

variates for matching. 1) High School percentile rank : woman’s high school percentile rank

on the basis of high school grades, derived as, 100−[rank in class/(no. of students in class×

100)]3. 2) IQ: derived from an administrative measure, namely, the Henmon-Nelson Test of

Mental Ability4, that was administered to all high school students in Wisconsin. The year

women took the test varied over time, but most scores employed in the WLS are from the

student’s junior year or are adjusted to reﬂect what their junior scores would be. These

scores were then renormed to IQ equivalents on the basis of the percentile distribution of

scores that were observed among all Wisconsin high school juniors. 3) Parental socioe-

conomic status: factor-weighted SES scores for parents 5that are derived based on the

following variables: father’s education, mother’s education, father’s occupation, and the

4-year average of parental income derived from Wisconsin tax records 6. 4) Population of

town: 1957 population of the town in which the woman attended her high school. Matching

on these variables ensures that two women had similar childhood measures before one of

them gave birth to an unintended pregnancy for the ﬁrst time and the other one did not.

Adulthood measures: Here we consider the following ﬁve time-varying covariates for

matching: 1) Number of children: this is a time-varying covariate which is used to match

the women based on the number of children they had prior to the time point when one of

them gave birth to an unintended pregnancy for the ﬁrst time and the other one did not

3see https://www.ssc.wisc.edu/wlsresearch/documentation/waves/?wave=wls5764&module=aedu

4see https://www.ssc.wisc.edu/wlsresearch/documentation/appendices/G/cor652.asc

5see https://www.ssc.wisc.edu/wlsresearch/documentation/waves/?wave=wls5764&module=tax

6see https://www.ssc.wisc.edu/wlsresearch/documentation/appendices/L/cor689.asc

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

ProtocolforanObservationalStudyontheEectsofGivingBirthsfromUnintendedPregnanciesonLaterLifePhysicalandMentalHealthSamratRoyDepartmentofStatisticsandDataScience,UniversityofPennsylvaniaemail:roysa@wharton.upenn.eduMarinaBogomolovIndustrialEngineeringandManagement,Technion-IsraelInstituteofTechnology...

展开>> 收起<<

Protocol for an Observational Study on the Eects of Giving Births from Unintended Pregnancies on Later Life Physical and Mental Health.pdf

共30页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Protocol for an Observational Study on the Eects of Giving Births from Unintended Pregnancies on Later Life Physical and Mental Health

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: