Conflict violence reduction and pregnancy outcomes: A regression discontinuity design in Colombia.

BACKGROUND: The relationship between exposure to conflict violence during pregnancy and the risks of miscarriage, stillbirth, and perinatal mortality has not been studied empirically using rigorous methods and appropriate data. We investigated the association between reduced exposure to conflict violence during pregnancy and the risks of adverse pregnancy outcomes in Colombia. METHODS AND
FINDINGS: We adopted a regression discontinuity (RD) design using the July 20, 2015 cease-fire declared during the Colombian peace process as an exogenous discontinuous change in exposure to conflict events during pregnancy, comparing women with conception dates before and after the cease-fire date. We constructed the cohorts of all pregnant women in Colombia for each day between January 1, 2013 and December 31, 2017 using birth and death certificates. A total of 3,254,696 women were followed until the end of pregnancy. We measured conflict exposure as the total number of conflict events that occurred in the municipality where a pregnant woman lived during her pregnancy. We first assessed whether the cease-fire did induce a discontinuous fall in conflict exposure for women with conception dates after the cease-fire to then estimate the association of this reduced exposure with the risks of miscarriage, stillbirth, and perinatal mortality. We found that the July 20, 2015 cease-fire was associated with a reduction of the average number of conflict events (from 2.64 to 2.40) to which women were exposed during pregnancy in their municipalities of residence (mean differences -0.24; 95% confidence interval [CI] -0.35 to -0.13; p < 0.001). This association was greater in municipalities where Fuerzas Armadas Revolucionarias de Colombia (FARC) had a greater presence historically. The reduction in average exposure to conflict violence was, in turn, associated with a decrease of 9.53 stillbirths per 1,000 pregnancies (95% CI -16.13 to -2.93; p = 0.005) for municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of FARC-related conflict events and a decrease of 7.57 stillbirths per 1,000 pregnancies (95% CI -13.14 to -2.00; p = 0.01) for municipalities with total number of FARC-related violent events above the 75th percentile of FARC-related events. For perinatal mortality, we found associated reductions of 10.69 (95% CI -18.32 to -3.05; p = 0.01) and 6.86 (95% CI -13.24 to -0.48; p = 0.04) deaths per 1,000 pregnancies for the 2 types of municipalities, respectively. We found no association with miscarriages. Formal tests support the validity of the key RD assumptions in our data, while a battery of sensitivity analyses and falsification tests confirm the robustness of our empirical results. The main limitations of the study are the retrospective nature of the information sources and the potential for conflict exposure misclassification.
CONCLUSIONS: Our study offers evidence that reduced exposure to conflict violence during pregnancy is associated with important (previously unmeasured) benefits in terms of reducing the risk of stillbirth and perinatal deaths. The findings are consistent with such beneficial associations manifesting themselves mainly through reduced violence exposure during the early stages of pregnancy. Beyond the relevance of this evidence for other countries beset by chronic armed conflicts, our results suggest that the fledgling Colombian peace process may be already contributing to better population health.

Introduction

Over 135 million people today need humanitarian assistance, the vast majority due to armed conflicts [1]. These conflicts have disproportionately affected low- and middle-income countries (LMICs) since the mid-20th century, with devastating consequences for health and development [2,3]. There has been much interest in recent years, by researchers and policymakers alike, in understanding how conflicts affect different health outcomes. Pregnant women and children have received special attention for the study of these effects, due to their high vulnerability to conflict violence [2,4,5]. In LMICs, such vulnerability is often compounded by pervasive healthcare inequities that contribute to higher rates of maternal and infant mortality than in richer countries [6,7]. Indeed, several studies have found that exposure to conflict during pregnancy is associated with short- and long-term adverse outcomes for mothers and their live-born children, both in terms of health and human capital, particularly for poor individuals [8-13].

Yet the aforementioned body of evidence suggesting a causal relationship between exposure to conflict violence and pregnancy outcomes is, in general, weakened by great methodological challenges. Since it is impossible to randomize conflict exposure, there are many possible confounders, and good quality information from contexts of armed conflict is scarce. This includes often patchy information about pregnancy histories in many countries, leading most research to focus on pregnancy outcomes related to live births, and, hence, relying on selected samples. A recent systematic review could find only a few analyses of the associations between conflict exposure and miscarriages (2 studies), stillbirth risk (5 studies), and perinatal mortality (2 studies) [14-20]. All these studies are observational and suffer from data and/or methodological limitations that make causal inference difficult [13] (five studies are before-and-after analyses without an adequate statistical control for observable and unobservable confounders [14-17,19]. A further study is a retrospective cohort analysis that includes a multivariate regression model to identify associations [20], whereas another study relies on before-and-after comparisons within a simple fixed effects panel analysis [18]).

Our study provides evidence of the association between conflict exposure and pregnancy outcomes, taking advantage of a natural experiment created by specific features of the conflict and peace process in Colombia. Colombia has endured one of the longest civil conflicts in history, with over 200,000 direct fatalities and millions of nonfatal victims (mainly civilians) during more than 5 decades [21]. In 2012, the government began a period of peace talks with the country’s largest rebel guerrilla, the Fuerzas Armadas Revolucionarias de Colombia (FARC). These conversations, known as the Havana talks, concluded in November 2016 with the signing of a definitive peace agreement between FARC and the government. During the Havana talks, FARC declared unilateral cease-fires on 5 occasions (in 2012, 2013, twice in 2014, and in 2015) before a final, bilateral cease-fire was agreed in August 2016. Due to the fact that these cease-fires mandated a cessation of violent activities by FARC specifically, but not by other rebel groups, the actual reduction in violence should have occurred primarily (if not only) in the geographic areas where FARC used to be active. For our empirical purposes, a cease-fire introduces a plausibly exogenous change in exposure to violence for pregnant women. We are then able to exploit this exogenous change, combined with rich information about pregnancy histories, to determine how the risks of miscarriage, stillbirth, and perinatal death are associated with a reduction in the intensity of conflict violence to which a pregnant woman is exposed in her municipality of residence. Our hypothesis is that reductions in the exposure to armed conflict during pregnancy are associated with reductions in the rates of miscarriages, stillbirths, and perinatal mortality.

Methods

Ethics

This study was granted institutional review board (IRB) ethical approval by the Research and Institutional Ethics Committee of the School of Medicine of the Pontificia Universidad Javeriana (Minutes No. 10/2018), Colombia. The study protocol is presented in S1 Protocol.

Pregnancy cohorts and outcomes

This study included all pregnant women in Colombia between January 1, 2013 and December 31, 2017. We constructed daily single pregnancy cohorts for each of the 1,826 days in the study period. We defined cohorts by the probable conception date. All women were followed until pregnancy ended in either a live birth or fetal death. Newborns were followed since birth until their seventh day of life. We constructed each pregnancy cohort based on birth and death certificates from the Ministry of Health’s Single Registry of Enrollees. All live births were identified from birth certificates, and all fetal deaths and deaths before the seventh day of life were identified from death certificates. We merged these 2 databases using an anonymized identifier. The probable date of conception was estimated using the difference between the date of delivery (or death in the case of stillbirths) and the gestational age reported in the birth or death certificates. These data sources were provided by the Colombian Ministry of Health to the Clinical Research Institute of Universidad Nacional de Colombia for use in our research (see Section A in S1 Text for a full description of each information source and the construction of the final database).

Since competing risks of loss in early pregnancy may affect pregnancy outcomes, we followed recent recommendations and used perinatal mortality as our main outcome, defined as death at any point between week 22 of pregnancy (day 154) and the seventh day after birth [22]. We were able to examine fetal deaths also in terms of stillbirth (fetal death after 22 weeks of pregnancy) and miscarriage (fetal death before week 22 of pregnancy) [23]. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (S1 Checklist).

Exposure to conflict violence

Our measure of individual exposure to conflict violence is the total number of conflict events that occurred in a woman’s municipality of residence during her pregnancy. Conflict data were obtained from the Memory and Conflict Observatory database (National Center for Historical Memory, NCHM). This database combines information on conflict-related events from several independent sources, being a leading data source for research about the Colombian conflict due to its reliability and completeness [24]. We downloaded the publicly available NCHM dataset from http://centrodememoriahistorica.gov.co in December 2018, including information about all types of conflict-related events (e.g., murders, kidnappings, sexual violence, land mines, and others) by location. We identified all conflict events that occurred on each day, in each Colombian municipality, between 2013 and 2018 (we included 2018 because women whose pregnancy began at the end of 2017 were also potentially exposed to events in 2018; we excluded from the analysis the events for which it was impossible to determine the date or place of occurrence, representing only 1.11% of all events reported by the NCHM).

We determined exposure to conflict violence for each pregnant woman included in our final dataset, based on the occurrence of conflict events in the municipality where the woman resided during her pregnancy. Thus, in our analytical setting, women “exposed to conflict violence” in their municipality of residence may have been exposed to violent events directly (by suffering violence themselves) or indirectly (by witnessing or knowing of acts of violence committed on others), with potential physical and/or psychological harmful consequences. With the daily information about pregnancy cohorts, we could identify the total number of conflict-related events to which each woman was exposed from the beginning to end of each pregnancy. A woman’s municipality of residence was presumed to be the one reported on the birth certificate for live births or death certificate for fetal deaths. Wherever possible, we also cross-verified the municipality of residence using official information from health insurance enrolment and healthcare provision databases to identify whether, during pregnancy, the woman made contact with the health system in the municipality that was assigned as that of residence.

Cease-fires during the Havana talks

FARC were involved in 7 cease-fires during the Havana talks. Only the last 2 cease-fires before the peace accord, on July 20, 2015 and August 29, 2016, were declared for an indefinite period and lasted more than just a few months, holding until the date of the final peace accord itself (Fig B in S1 Text). The July 20, 2015 cease-fire was declared unilaterally by FARC, followed in August 29, 2016 by a definitive cease-fire by the Colombian government, close to the signing of the peace accord. The other 5 cease-fires were very limited in duration (lasting no more than 4 months), and violence returned to the municipalities soon after these ended. Therefore, we focused on the July 20, 2015 cease-fire for our empirical analyses. This is the first cease-fire that can be expected to have introduced a sudden and permanent reduction in the levels of FARC violence across municipalities, since it was the definitive cease-fire declared and adhered to by FARC. Nonetheless, we also assessed whether the August 29, 2016 cease-fire (which involved both the government and FARC) generated a sharp and lasting reduction in violence. The latter analysis was again performed using regression discontinuity (RD) analysis (see Results).

FARC presence across municipalities

Several groups other than FARC were responsible for armed violence in certain areas during the study period (Fig C in S1 Text). The cease-fires involving FARC could be expected to have affected violence levels mainly in the areas where FARC had more presence historically. To identify those areas, we grouped all municipalities into categories of FARC-related conflict intensity, based on the distribution of the total number of conflict events involving FARC for the years 2000 through to 2017, thus identifying those municipalities with persistent FARC presence over the past 17 years. We created 4 categories of municipalities: (1) M-p90: municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of these events; (2) M-p75: municipalities with total number of FARC-related events above the 75th percentile; (3) M-zero: municipalities with no conflict event recorded during the 2000 to 2017 period (whether involving FARC or not); and (4) M-other: municipalities that did not belong to the other 3 groups (Fig 1). We chose these categories to achieve the objective of identifying municipal heterogeneity with respect to FARC-related violence levels, in order to investigate the relationships between said heterogeneity and the local responses of violence levels to the cease-fire, without compromising sample sizes for estimation purposes.

Fig 1

Distribution of categories of municipalities in Colombia according to FARC presence between 2000 and 2017.

Classification of Colombian municipalities according to the distribution of the total number of conflict events involving FARC for the years 2000 through to 2017. Four categories of municipalities: (1) M-p90: municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of these events (red); (2) M-p75: municipalities with total number of FARC-related events above the 75th percentile (orange); (3) M-zero: municipalities with no conflict event recorded during the entire 2000–2017 period (whether involving FARC or not) (white); and (4) M-other: municipalities that did not belong to any of the other 3 municipality groups (yellow). FARC, Fuerzas Armadas Revolucionarias de Colombia. Base layer of the map from Geoportal DANE (https://geoportal.dane.gov.co/servicios/descarga-y-metadatos/descarga-mgn-marco-geoestadistico-nacional/).

Estimation methodology

We adopted a quasi-experimental RD approach, a methodological design that is well suited to permit robust statistical inference in study settings like ours [25,26]. The cease-fires involving FARC can be regarded as introducing exogenous changes (i.e., discontinuities) in the number of conflict events in the municipalities where women lived during their pregnancies. This setting creates a natural experiment allowing the comparison of outcomes between women with conception dates before and after a cease-fire date (the threshold). The sudden, discontinuous change in exposure to conflict violence implies that assignment to different levels of conflict violence is as good as random for women close to either side of the threshold [25,26]. If the underlying assumptions of the RD design are met (which we can test with the data available), we can use the discontinuous change induced by a cease-fire to estimate the association between exposure to violence and pregnancy outcomes [27-29]. We used the conception date for each cohort as the running variable and the date of the July 20, 2015 cease-fire as the discontinuity threshold (Section C in S1 Text).

We implemented our RD approach through a sequence of estimations steps, each of these subject to formal tests to assess result robustness. First, we examined whether the July 20, 2015 cease-fire was in fact associated with a discontinuity in the total number of conflict events to which women were exposed during pregnancy, separately for events involving (a) FARC; and (b) any armed group (not necessarily FARC). We did so for all municipalities and for each of the categories that we created based on FARC-related events, anticipating that municipalities with greater FARC presence (M-p90 and M-p75) would have seen larger changes in violence associated with the July 20, 2015 cease-fire than municipalities where FARC had less/no presence or where there was no conflict violence prior to the cease-fire (M-zero and M-other). We repeated this process for the August 29, 2016 cease-fire.

Second, for the categories of municipalities where we identified a discontinuity in violence (measured by either FARC-related events or total conflict events), we estimated the association between the implementation of the cease-fire and the risks of perinatal mortality, stillbirth, and miscarriage. We performed these estimations using a nonparametric RD approach with optimal bandwidth [30,31]. This estimation strategy involves approximating the regression functions above and below the threshold (i.e., July 20, 2015) by means of weighted polynomial regressions of order 1 (also called local linear regressions), with weights computed by applying a triangular kernel function on the distance (in days) of each observation to the threshold [32-34]. Similarly to a randomized experiment with imperfect adherence, exposure to conflict took a nondeterministic form (women exposed and not exposed to conflict were on both sides of the threshold) [25]. The parameter of interest was the intention-to-treat estimate.

Third, we tested the validity of the basic RD assumptions. The RD estimates could be biased if the key assumptions of no-manipulation and continuity of baseline variables do not hold in our data. We evaluated the no-manipulation assumption graphically and statistically (McCrary tests) [27]. We also tested statistically for the presence of any discontinuities in several observable baseline characteristics between women with conception dates before and after the cease-fire. These key baseline characteristics (extracted from the birth or death certificates) included age, education level, health insurance status, marital status, and number of children.

Lastly, we conducted a battery of sensitivity analyses and falsification tests for our main results. We reran all our RD estimations using alternative bandwidths and a parametric approach with first-, second-, and third-order polynomial specifications. The order of the polynomial refers to the highest exponent associated with the running variable of the discontinuous regression (i.e., the distance in days from each observation to the cease-fire date) [27,35]. We also estimated placebo effects by reestimating our models using solely the samples of women living in the municipalities where no discontinuities in conflict events were found (M-zero and M-other), and hence where no associations between cease-fire implementation and pregnancy outcomes should be expected.

For all RD estimations, we report p-values from local linear regressions with 95% confidence intervals (CIs). The analyses were performed using Stata 15 (StataCorp LLC, College Station, Texas, US).

Results

A total of 3,661,022 birth certificates and 1,441,140 death certificates were issued between January 2013 and December 2017 in Colombia (Fig A in S1 Text). After identifying single pregnancies with conception dates between January 1, 2013 and December 31, 2017, the total cohort included 3,254,696 pregnant women. Table 1 presents the characteristics of the cohort for the full sample and separately for women in M-p90 and M-p75 municipalities. Women in M-p90 and M-p75 municipalities had higher pregnancy-related mortality rates than the national average.

Table 1

Baseline characteristics of all pregnant women in Colombia between January 1, 2013 and December 31, 2017.

	Full sample (n = 3,254,696)	M-p90 (n = 1,288,771)	M-p75 (n = 1,732,875)	M-other (n = 1,469,251)	M-zero (n = 52,570)
Sociodemographic characteristics
Age—mean (SD)	24.85 (6.60)	25.32 (6.65)	25.06 (6.67)	24.61 (6.51)	24.67 (6.66)
Age—category n (%)
<18 y	426,709 (13.11)	149,135 (11.57)	219,233 (12.65)	200,191 (13.63)	7,285 (13.86)
18 to 34	2,521,718 (77.48)	1,003,727 (77.88)	1,339,213 (77.28)	1,142,234 (77.74)	40,271(76.60)
35 to 39	245,862 (7.55)	109,010 (8.46)	139,404 (8.04)	102,496 (6.98)	3,962 (7.54)
>39 y	60,407 (1.86)	26,899 (2.09)	35,025 (2.02)	24,330 (1.66)	1,052 (2.00)
Health insurance n (%)
Contributory	1,390,665 (42.73)	692,376 (53.72)	813,811 (46.96)	560,725 (38.16)	16,129 (30.68)
Subsidized	1,703,726 (52.35)	518,915 (40.27)	824,155 (47.56)	844,328 (57.47)	35,243 (67.04)
Other	160,262 (4.92)	77,458 (6.01)	94,882 (5.48)	64,183 (4.37)	1,197 (2.28)
Education n (%)
Primary or less	460,928 (14.73)	141,260 (11.33)	231,539 (13.91)	217,215 (15.37)	12,174 (23.53)
Secondary	731,012 (23.36)	252,348 (20.25)	368,319 (22.13)	349,885 (24.76)	12,808 (24.76)
Higher than secondary	1,937,422 (61.91)	852,698 (68.42)	1,064,527 (63.96)	846,146 (59.87)	26,749 (51.71)
Married n (%)	513,775 (15.79)	219,873 (17.06)	273,650 (15.79)	231,274 (15.74)	8,851 (16.84)
Number of children—mean (SD)	1.83 (1.21)	1.74 (1.12)	1.89 (1.20)	1.86 (1.21)	2.04 (1.29)
Prenatal care—mean (SD)	6.47 (2.53)	6.63 (2.63)	6.45 (2.61)	6.48 (2.43)	6.49 (2.45)
Pregnancy outcomesRates per 1,000 pregnancies (95% CI)
Miscarriage	45.65	72.45	61.48	28.61	12.55
	(45.43 to 45.88)	(72.00 to 72.89)	(61.12 to 61.84)	(28.34 to 28.88)	(12.35 to 12.75)
Stillbirth	8.00	10.13	9.45	6.61	3.82
	(7.90 to 8.10)	(9.95 to 10.31)	(9.30 to 9.59)	(6.48 to 6.75)	(3.37 to 4.26)
Perinatal death	11.86	13.7	13.15	10.65	4.85
	(11.74 to 11.98)	(13.49 to 13.91)	(12.97 to 13.32)	(10.48 to 10.82)	(4.26 to 5.45)

Data are means (SD). M-p90: pregnant women in municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of these events. M-p75: pregnant women in municipalities with total number of FARC-related violent events above the 75th percentile of the distribution of these events. M-zero: pregnant women in municipalities with no conflict event recorded during the 2000 to 2017 period (whether involving FARC or not). M-other: pregnant women in municipalities that did not belong to any of the other 3 municipality groups. The number (percentage (%)) of missing values for each variable is age = 0 (0.00%), health insurance = 43 (0.00%), education = 125,334 (3.85%), married = 0 (0.00%), number of children = 29 (0.00%), prenatal care utilization = 174,147 (5.35%), and pregnancy outcomes = 0 (0.00%).

95% CI, 95% confidence interval; FARC, Fuerzas Armadas Revolucionarias de Colombia; SD, standard deviation.

On average, the women in our study were exposed to 3.76 (95% CI 3.75 to 3.77) conflict events of any type during pregnancy, with the most frequent types of violent events being selective murder (1.53 events occurred during pregnancy, on average) and sexual violence (0.98 events) (Table 2). Women in M-p90 municipalities were exposed to a higher number of total conflict events on average during pregnancy (7.64; 95% CI 7.62 to 7.66) than women in M-p75 municipalities (6.31; 95% CI 6.29 to 6.32) (mean difference 1.34 (95% CI 1.31 to 1.36; p < 0.001)).

Table 2

Exposure to conflict violence events during pregnancy.

	Full sample (n = 3,254,696)	M-p90 (n = 1,288,771)	M-p75 (n = 1,732,875)	M-other (n = 1,469,251)	M-zero (n = 52,570)
Total events1	3.764 (8.555)	7.644 (11.930)	6.308 (10.987)	0.899 (1.851)	NA
Total FARC-related events	0.417 (2.730)	0.922 (4.219)	0.765 (3.702)	0.021 (0.210)	NA
Type of conflict event2
Terrorist attack	0.030 (0.171)	0.075 (0.264)	0.056 (0.231)	0.000 (0.010)	NA
Act of war	0.414 (1.955)	0.877 (2.942)	0.740 (2.624)	0.044 (0.282)	NA
Attack on populations	0.000 (0.013)	0.000 (0.020)	0.000 (0.018)	0.000 (0.000)	NA
Selective murder	1.535 (3.385)	3.096 (4.742)	2.408 (4.279)	0.561 (1.361)	NA
Kidnapping	0.061 (0.369)	0.123 (0.539)	0.102 (0.487)	0.015 (0.133)	NA
Child recruitment	0.108 (0.433)	0.249 (0.636)	0.199 (0.574)	0.006 (0.082)	NA
Massacre	0.004 (0.065)	0.004 (0.059)	0.007 (0.082)	0.001 (0.040)	NA
Forced disappearance	0.226 (1.497)	0.438 (2.316)	0.353 (2.011)	0.085 (0.394)	NA
Damage to property	0.293 (1.292)	0.644 (1.946)	0.523 (1.728)	0.033 (0.212)	NA
Sexual violence	0.976 (3.807)	1.891 (4.891)	1.712 (5.070)	0.142 (0.641)	NA
Land mine	0.116 (0.973)	0.247 (1.499)	0.208 (1.319)	0.011 (0.150)	NA

Data are means (SD). M-p90: pregnant women in municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of these events. M-p75: pregnant women in municipalities with total number of FARC-related violent events above the 75th percentile of the distribution of these events. M-zero: pregnant women in municipalities with no conflict event recorded during the 2000 to 2017 period (whether involving FARC or not). M-other: pregnant women in municipalities that did not belong to any of the other 3 municipality groups.

1Note: Most of the conflict events taking place after the start of the peace talks with FARC involved the Colombian armed forces, as the latter continued to fight the other rebel armed groups.

2Note: See Section B in S1 Text for an official definition of conflict events.

FARC, Fuerzas Armadas Revolucionarias de Colombia; NA, not applicable; SD, standard deviation.

Exposure to conflict events during pregnancy showed a sustained falling trend throughout the study period, despite a few spikes (Fig D in S1 Text). Instead of relying on visual inspection, we use RD estimation to identify discontinuities in violence intensity around the cease-fires declared during the study period. We found that the July 20, 2015 cease-fire was significantly associated with a reduction in the average number of conflict events to which a woman was exposed during pregnancy, both with respect to FARC-related and total events (i.e., not necessarily FARC related) (Fig 2, Fig E in S1 Text). Specifically, this cease-fire was associated with a decrease in exposure to violence during pregnancy of 0.24 total events (95% CI −0.33 to −0.08; p < 0.001) and 0.03 FARC-related events (95% CI −0.05 to −0.01; p = 0.01), on average, for women with conception dates after the cease-fire (Fig 2). Moreover, we found that these reductions in conflict exposure associated with the cease-fire occurred only for pregnant women who lived in M-p90 and M-p75 municipalities. By contrast, repeating these analyses for the August 29, 2016 cease-fire reveals no statistically significant changes in exposure to conflict events during pregnancy (Fig 2).

Fig 2

Association between the July 20, 2015 and August 28, 2016 cease-fires and the exposure to total conflict events during pregnancy: Colombia and categories of municipalities (RD plots).

Association between the July 20, 2015 and August 28, 2016 cease-fires and the exposure to total conflict events during pregnancy (i.e., events related to any armed group). Local linear regressions and 95% confidence interval. A, B, C, and D show the estimates of the July 20, 2015 cease-fire (red lines). E, F, G, and H show the estimates of the August 28, 2016 cease-fire (blue lines). A and E include women from all municipalities in Colombia. B and F only include women in M-p90 municipalities. C and G only include women in M-p75 municipalities. D and F only include women in M-other municipalities. We do not show graphs for women in M-zero municipalities as these women were not exposed to conflict events. M-p90: municipalities with total number of FARC-related violent events above the 90th percentile of the distribution of these events; M-p75: municipalities with total number of FARC-related events above the 75th percentile of the distribution of these events; M-zero: municipalities with no conflict event recorded during the entire 2000 to 2017 period (whether involving FARC or not); and M-other: municipalities that did not belong to any of the other 3 municipality groups. The results indicate that the July 20, 2015 cease-fire was associated with a statistically significant discontinuity in the number of conflict events only for women in M-p90 and M-p75 municipalities, while the August 28, 2016 cease-fire was not associated with in any statistically significant changes in exposure to conflict events for pregnant women. ITT estimates by RD analysis using local linear regression and optimal bandwidth of 56 days. 95% CI, 95% confidence interval; FARC, Fuerzas Armadas Revolucionarias de Colombia; ITT, intention-to-treat; RD, regression discontinuity.

Associations between reduced conflict exposure and pregnancy outcomes

Fig 3 shows that the decreased exposure to conflict events during pregnancy, associated with the July 20, 2015 cease-fire, was associated with reductions of 9.53 (95% CI −16.13 to −2.93; p = 0.005) stillbirths and 10.69 (95% CI −18.32 to −3.05; p = 0.01) perinatal deaths per 1,000 pregnancies, for women in M-p90 municipalities. These figures represent relative reductions of 62.32% in stillbirths and 53.69% in perinatal mortality (Table A in S1 Text). For pregnant women in M-p75 municipalities, we estimated associated reductions of 7.57 (95% CI −13.14 to −2.00; p = 0.01) stillbirths and 6.86 (95% CI −13.24 to −0.48; p = 0.04) perinatal deaths per 1,000 pregnancies, corresponding to relative reductions of 55.09% in stillbirths and 40.19% in perinatal mortality. We did not find statistically significant associations with respect to miscarriages: Point estimates were −8.52 (95% CI −7.47 to 24.52; p = 0.30) in M-p90 municipalities and −3.27 (95% CI −9.78 to 16.32; p = 0.62) in M-p75 municipalities.

Fig 3

Association between the July 20, 2015 cease-fire and pregnancy outcomes.

ITT estimates of the July 20, 2015 cease-fire on the risks of miscarriage, stillbirth, and perinatal mortality, for pregnant women in M-p90 and M-p75 municipalities. Estimates obtained through RD analysis using local linear regression and bandwidth of 28 days (Section C in S1 Text). We observe statistically significant estimates on the risk of stillbirths and perinatal mortality, but no effects on the risk of miscarriage. The estimates are larger for women in M-p90 municipalities than for those in M-p75 municipalities. 95% CI, 95% confidence interval; ITT, intention-to-treat; RD, regression discontinuity.

Validity of the RD assumptions

Formal statistical tests support the validity of the underlying RD assumptions in our data, and hence of the estimated RD associations (Fig F in S1 Text). Both the histograms of the distribution of women according to conception dates (panel A) and the McCrary tests (panel B) showed no evidence of manipulation, i.e., there is no evidence of discontinuities in the density function of conception dates around the cease-fire (threshold). We also conducted graphical and statistical analyses to assess balance in baseline characteristics between pregnant women on each side of the threshold and again found no evidence of discontinuity for any of these variables (Figs G and H in S1 Text). Taken together, the results above offer strong statistical evidence that exposure to different levels of conflict violence during pregnancy is as good as random for women on both sides of the July 20, 2015 cease-fire [25,34].

Robustness checks

We tested the sensitivity of our main results to using several different combinations of functional forms and bandwidths. We found that our results of statistically significant reductions in both stillbirths and perinatal mortality, associated with the reduction in violence after the July 20, 2015 cease-fire, are robust to all these specification changes (Tables B and C in S1 Text).

We also estimated associations of the July 20, 2015 cease-fire with pregnancy outcomes for women in M-zero and M-other municipalities as placebo tests (as we did not find any discontinuities in exposure to conflict events in those municipalities). We found no evidence of associations for either stillbirths or perinatal mortality (Table 3). All these results provide further reassurance about the validity of our estimates.

Table 3

Associations between the July 20, 2015 cease-fire and fetal deaths and perinatal mortality.

Placebo tests using M-other and M-zero municipalities.

	ITT effect	95% CI	p-value	Functional form	Bandwidth	Observations in bandwidth
M-other municipalities
Stillbirth	0.91	−3.88 to 5.70	0.71	LLR	28	20,833
Stillbirth	0.69	−3.23 to 4.61	0.73	LLR	42	31,701
Stillbirth	0.98	−2.43 to 4.38	0.57	LLR	56	42,706
Perinatal mortality	5.16	−0.89 to 11.21	0.09	LLR	28	20,833
Perinatal mortality	3.72	−1.27 to 8.72	0.14	LLR	42	31,701
Perinatal mortality	3.39	−0.95 to 7.73	0.13	LLR	56	42,706
M-zero municipalities
Stillbirth	0.06	−0.06 to 0.18	0.32	LLR	28	778
Stillbirth	−2.52	−7.46 to 2.42	0.32	LLR	42	1,163
Stillbirth	−3.09	−9.16 to 2.97	0.32	LLR	56	1,585
Perinatal mortality	13.26	−4.10 to 30.61	0.13	LLR	28	778
Perinatal mortality	9.00	−5.63 to 23.64	0.23	LLR	42	1,163
Perinatal mortality	5.52	−8.91 to 19.94	0.45	LLR	56	1,585

** p < 0.001, * p < 0.05.

Results of the falsification tests using different bandwidths (28, 42, and 56 days) and LLRs We observe no statistically significant estimates of the cease-fire on pregnancy outcomes in M-other and M-zero municipalities (as expected). M-zero: pregnant women in municipalities with no conflict event recorded during the 2000 to 2017 period (whether involving FARC or not). M-other: pregnant women in municipalities that did not belong to the M-p90, M-p75, or M-zero municipality groups.

95% CI, 95% confidence interval; FARC, Fuerzas Armadas Revolucionarias de Colombia; ITT: intention-to-treat effect per 1,000 pregnancies; LLR, local linear regression.

Discussion

The relationship between conflict violence and the health outcomes of civilian populations has been examined from different perspectives by previous studies [36,37]. Nevertheless, the associations between conflict violence and the risks of perinatal mortality and stillbirth have not been evaluated previously with the necessary statistical robustness [13]. Our study employed a RD approach, using cease-fire dates as thresholds that could introduce exogenous discontinuities in conflict intensity in Colombia, to estimate the association between reduced exposure to conflict events during pregnancy and the risks of miscarriage, stillbirth, and perinatal mortality. We found that the July 20, 2015 cease-fire was associated with a decrease in the number of conflict events to which women were exposed during pregnancy, which, in turn, was associated with reduced risks of stillbirth and perinatal mortality.

To our knowledge, our study is the first to robustly estimate the association between exposure to conflict during pregnancy and the risks of stillbirth and perinatal mortality. In addition to methodological shortcomings, data limitations (particularly around pregnancy records) have prevented previous studies from establishing the aforementioned associations. For example, Wagner and colleagues reported that exposure to conflict increased mortality for 1 year olds, 5 year olds, and for reproductive-aged women, as well as the risk of children becoming orphans [5,36]. Further studies have investigated associations between exposure to conflict during pregnancy and other indicators including birth weight, height for age, and cognitive outcomes [8,10,38-40]. Unlike all those previous studies, in our analysis, we were able to reconstruct all pregnancy cohorts in Colombia, since the beginning of each pregnancy, using rich administrative data. This enabled us to measure conflict exposure during pregnancy for all women whose pregnancies began between January 1, 2013 and December 31, 2017, following them until the pregnancy ended in fetal death or a live birth. We therefore mitigated the possibility of survival bias influencing our results.

Strengths and limitations of this study

The validity of our RD approach and its findings are supported by specific historic features of the Colombian peace process. First, while the number of conflict events showed a generally declining trajectory during the entire period of the Havana talks, we found that only the July 20, 2015 cease-fire was associated with a sharp and sustained discontinuity in conflict exposure. This is in line with the fact that for the previous FARC-declared cease-fires, the start/end dates were known and were near in time to one another, and/or the Colombian army continued attacking FARC guerrillas, thus leading to minimal and unsustained changes in violence levels, followed by subsequent rises in these levels (see spikes in violence between 50 and 100 days before the July 20, 2015 cease-fire, Fig E in S1 Text). However, the Havana talks intensified markedly in June to July 2015, leading to the FARC’s definitive cease-fire and an accompanying reduction in attacks by the Colombian army against that armed group, in order to support the possibility of a definitive peace agreement. This sharp reduction in violent activities involving the key conflict actors after the July 2015 cease-fire was sustained, resulting in violence levels between FARC and the government that were already very low at the time of the August 29, 2016 bilateral cease-fire.

Second, we found that the July 20, 2015 cease-fire was associated with a higher reduction in conflict violence in municipalities where the FARC had greater presence historically (M-p90) than in municipalities with relatively lower FARC presence (M-p75), whereas no associated changes in violence levels were found in municipalities where FARC were not present or where there was no armed conflict. These (expected) patterns suggest a dose–response relationship between conflict exposure and pregnancy outcomes: The July 20, 2015 cease-fire was associated with greater reductions in stillbirths and perinatal mortality in M-p90 than in M-p75 municipalities and no changes in outcomes in municipalities with no FARC presence or where there were no conflict-related events.

There could have been, of course, several other changes in socioeconomic factors or events (e.g., local health interventions) taking place during the study period, which could influence pregnancy outcomes in general. However, to threaten the validity of our estimates, given the characteristics of our RD estimation design, events, or trends such as those described would need to have been implemented or changed suddenly, discontinuously in the neighborhood of the threshold (i.e., 1 month around the July 20, 2015 cease-fire), and, specifically, for the particular groups of women under comparison (e.g., those living in M-p90 municipalities), in such a way to then introduce discontinuities in the observed distribution of conception dates and/or baseline characteristics at the cease-fire threshold. Yet in our extensive statistical testing of both the no-manipulation and baseline variable continuity assumptions, we found consistent evidence that the distributions of observed conception dates and key observable confounders are continuous at the threshold, thus also supporting the continuity of unobservable and other observable confounders at the threshold [28,29].

The numerous sensitivity tests undertaken indicate that our estimates are robust and provide valuable insights for the interpretation of findings. Although our main estimation results for pregnancy outcomes are based on the same optimal RD bandwidth of 1 month [31], we tested various alternative bandwidth sizes of up to 2 months, with no material changes in our conclusions. Nevertheless, it is noteworthy that the RD point estimates for stillbirths and perinatal mortality tend to decrease in absolute value as the bandwidth increases from 14 to 56 days (Tables B and C in S1 Text). This suggests that exposure to violence very early in the pregnancy is a key driver of the risks of stillbirth and perinatal mortality.

One limitation of our study is that, with the data available, we were unable to undertake an in-depth investigation of possible mechanisms linking reductions in conflict exposure to reduced risks of stillbirth and perinatal mortality. In this regard, however, evidence from other studies suggest that maternal stress could be a crucial mechanism [20,41-44]. A dysfunction of the hypothalamic–pituitary–adrenal axis (and abnormal cortisol levels) has been identified as the key channel whereby higher levels of stress influence pregnancy outcomes [42,43,45,46]. Simple descriptive analysis of our data offers some support to this possible mechanism. The data on the types of conflict violence to which pregnant women were most vulnerable during the study period indicate that the highest exposure rates were related to very stressful and traumatic events (including events showing a direct relationship with sexual and reproductive health), particularly sexual violence and murder. In addition to reduced risk of suffering traumatic episodes themselves following the July 2015 cease-fire, pregnant women may have also experienced reduced maternal stress stemming from the decrease in violence occurring in the communities where they lived, as such “neighborhood effects” have also been argued to be associated with pregnancy outcomes [47,48]. Related to this, the absence of associations between conflict reduction and miscarriages may be explained by the fact that such fetal losses tend to occur very early in the pregnancy and are mainly caused by chromosomal abnormalities or other factors unrelated to violence exposure during pregnancy [49,50]. Changes in healthcare access due to reduced conflict violence may have played a role for improved pregnancy outcomes as well. However, due to data constraints, we were unable to determine the plausibility of a competing “care access” channel vis-à-vis maternal stress. Although in supplementary analyses we did not find evidence that the reduction in conflict exposure following the July 2015 cease-fire was associated with any changes in prenatal care utilization rates for pregnant women (either in M-p90 or M-p75 municipalities; Table D in S1 Text), we lack suitable data to investigate any other indicators of access to care during pregnancy, which could prove more relevant in the early stages of pregnancy than simple prenatal care utilization (e.g., prenatal care quality, access to essential drugs, or diagnostic tests).

We must note other limitations of this study. First, although Colombia’s administrative records have improved substantially in the last decade, some data quality issues may remain and possibly influence some empirical conclusions. One relevant example is that the number of miscarriages reported in our data may well be an underestimation of the true number during the study period, since miscarriages may occur without formal healthcare intervention. Unfortunately, there is no information available on the patterns of underreporting of miscarriages across Colombian municipalities, thus precluding an assessment a priori of whether this phenomenon is likely to have led to an underestimation of our RD associations between conflict exposure and the risk of miscarriage. Yet it does not seem plausible to expect that the July 2015 cease-fire induced sudden changes (e.g., within 1 month) in the patterns of underreporting of miscarriages in M-p90 or M-p75 municipalities, which would be the type of change necessary to influence our estimates (note that a systematic but unchanged pattern of underreporting of miscarriages before and after the cease-fire in a given municipality would not affect the validity of our RD estimates). Data constraints limit our ability to scrutinize this possibility further, however. Second, we measured exposure to conflict based on information about the woman’s place of residence taken primarily from birth and death certificates. Although we also conducted cross-checks using other administrative databases wherever feasible, we cannot rule out the possibility that some women migrated, at some point during their pregnancy periods, between municipalities with different intensities of conflict violence. Similarly, we cannot account in our empirical analysis for potential differences in the intensity of individual conflict exposure, or in the knowledge about conflict events, among pregnant women living in the same municipality. Unfortunately, we do not have detailed geographic information that would allow us to construct finer measures of heterogeneity in individual conflict exposure. Finally, in-depth examination of access and quality of care indicators could have provided valuable insights about possible pathways leading from reduced conflict violence exposure to improved pregnancy outcomes. As previously noted, however, we do not have information about indicators beyond prenatal care use that could be linked up with our dataset of individual pregnancies (such as indicators of quality of prenatal care and access to quality care during labor and delivery or postnatal follow-up for mother and child), and, therefore, we must leave such analyses for future research.

Conclusions

Our study provides rigorous evidence that reduced exposure to conflict violence during pregnancy is associated with important (yet previously unmeasured) benefits in terms of reduced risks of stillbirth and perinatal mortality. Beyond the relevance of this evidence for other countries beset by chronic armed conflicts, our results offer support to the view that the fledgling Colombian peace process may be already contributing to better population health.

STROBE Statement.

(PDF)

Click here for additional data file.

Effects of armed conflict during the first three months of life on mortality and the use of health services among infants under 1 year of age.

(PDF)

Click here for additional data file.

Comprehensive description of methods and complementary analyzes.

Fig A: Flow diagram of the data linkage process. Fig B: Cease-fires declared during the Havana talks. Fig C: Historic presence of armed groups in Colombia (2000 to 2017). Fig D: Trend in the number of conflict events to which pregnant women were exposed during pregnancy in Colombia between January 2013 and December 2017. Fig E: Effects of the July 20, 2015 and August 28, 2016 cease-fires on the exposure to FARC-related conflict events during pregnancy: Colombia and categories of municipalities (RD plots). Fig F: Tests of the RD no-manipulation assumption around the July 20, 2015 cease-fire threshold. Fig G: Tests of balance in baseline characteristics around the July 20, 2015 cease-fire for women in M-p90 municipalities. Fig H: Tests of balance in baseline characteristics around the July 20, 2015 cease-fire for women in M-p75 municipalities. Table A: Effects of the July 20, 2015 cease-fire on fetal deaths and perinatal mortality. Table B: Effects of the July 20, 2015 cease-fire on fetal deaths and perinatal mortality. LLR and parametric regressions by order of polynomial for M-p90 municipalities. Table C: Effects of the July 20, 2015 cease-fire on fetal deaths and perinatal mortality. LLR and parametric regressions by order of polynomial for M-p75 municipalities. FARC, Fuerzas Armadas Revolucionarias de Colombia; LLR, local linear regression; RD, regression discontinuity.

(DOCX)

Click here for additional data file.

17 Feb 2021

Dear Dr Buitrago,

Thank you for submitting your manuscript entitled "The Effects of Conflict Violence Reduction on Pregnancy Outcomes: Evidence from a Regression Discontinuity Design in Colombia" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by February 19, 2021.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Beryne Odeny

Associate Editor

PLOS Medicine

12 Apr 2021

Dear Dr. Buitrago,

Thank you very much for submitting your manuscript "The Effects of Conflict Violence Reduction on Pregnancy Outcomes: Evidence from a Regression Discontinuity Design in Colombia" (PMEDICINE-D-21-00816R1) for consideration at PLOS Medicine. We are in receipt of all reviewer comments. As promised, this is an official communication of the major revision decision.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also sent to four independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by May 03 2021 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Beryne Odeny,

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Please address the following editorial comments:

Please revise your title according to PLOS Medicine's style. Your title must be nondeclarative and not a question. It should begin with main concept if possible. "Effect of" should be used only if causality can be inferred, i.e., for an RCT. Please place the study design ("A regression Discontinuity Design”) in the subtitle (ie, after a colon).

Your study is observational and therefore causality cannot be inferred. Please remove language that implies causality such as “causal evidence” or “causal inference” or “causal effect” or “causes.” Please refer to associations instead.

- The Data Availability Statement (DAS) requires revision. For each data source used in your study:

a) If the data are owned by a third party but freely available upon request, please note this and state the owner of the data set and contact information for data requests (web or email address). Note that a study author cannot be the contact person for the data.

b) If the data are not freely available, please describe briefly the ethical, legal, or contractual restriction that prevents you from sharing it. Please also include an appropriate contact (web or email address) for inquiries (again, this cannot be a study author).

-If you could provide some data to support the validity of their proxy method for disease diagnosis, it would significantly enhance the quality of the manuscript

Please ensure that the study is reported according to the STROBE guideline, and include the completed [STROBE or other] checklist as Supporting Information. When completing the checklist, please use section and paragraph numbers, rather than page numbers. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).",Please report your study according to the relevant guideline, which can be found here: http://www.equator-network.org/

In the abstract Methods and Findings:

-Please ensure that all numbers presented in the abstract are present and identical to numbers presented in the main manuscript text.

-Please include the actual amounts of relevant outcomes

- Please quantify the main results (with both 95% CIs and p values).

- In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

Abstract summary - At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary.

Introduction section: please provide a clear statement of hypothesis at the end of the introduction.

Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

-In statistical methods, please refer to any post-hoc corrections to correct for multiple comparisons during your statistical analyses. If these were not performed please justify the reasons. Please refer to our statistical reporting guidelines for assistance (https://journals.plos.org/plosone/s/submission-guidelines.#loc-statistical-reporting)

In the Methods and Results section:

- Please provide both adjusted and unadjusted estimates.

- Please provide 95% CIs and p values for all HRs.

- When a p value is given, please specify the statistical test used to determine it.

For your figures and tables, please do the following:

- Please define the abbreviations such as FARC, m-P90, m-p75, M-oth, M-zero

Please include line numbers in your next draft.

Comments from the reviewers:

Reviewer #1: This well-written manuscript addresses the important topic of how reductions in conflict violence may affect pregnancy outcomes in low-resource settings with high levels of conflict violence. The authors report on data from over 3 million pregnant women in the country of Columbia between 2013 and 2017. They take advantage of a natural experiment that occurred when ceasefires with FARC occurred in 2015 and 2016 and resulted in dramatic reductions in conflict events. They utilize comprehensive administrative municipality-level data on conflict events that occurred and on adverse pregnancy outcomes (stillbirths, miscarriages, and perinatal mortality) to conduct a rigorous regression discontinuity analysis, and found that decreases in conflict events in municipalities heavily impacted by FARC-related conflict were associated with statistically significant reductions in stillbirths and perinatal mortality (but not miscarriages). The authors have done a good job of describing their methodology, assumptions, sensitivity analyses, and noting the limitations of their research. However, I had a few questions and comments as detailed below. Overall, I thought this was a fascinating analysis and an important contribution to the literature. Specific comments by section are given below.

Abstract:

1. Page 2 and throughout: Given that this necessarily an observational study, it seems like the authors should avoid language such as "the reduction in average exposure to conflict violence resulted in…." throughout the paper. There is not definitive evidence that the reduction in violence caused these outcomes. However, I do recognize that this may be the best evidence we can get on this question.

Introduction:

2. Page 5: Were there other events or trends going on simultaneously in the country during this period that could have also contributed to improved pregnancy outcomes? Campaigns to increase birth in a health facility with a skilled attendant? Increases in maternal education? Changes in the economy?

Methods:

3. Page 10: Which observable baseline characteristics of women before and after the ceasefire were tested statistically? These should be specified.

Results:

4. Page 11: Based on the characteristics in Table 1, I would like to know if any of the differences in other characteristics of the women are statistically significant among the municipality groups. I think it would be more informative to show all four municipality groups in these tables (not just the total, M-p90, and M-p75), so we can understand more about the differences between those municipalities who were most impacted by FARC-related conflict and those who were less impacted.

Discussion:

5. Page 15: The results of the sensitivity test indicating that exposure very early in pregnancy is the key driver are not explained clearly enough. It is not clear to me why increasing the RD bandwidth to sizes up to two months with no changes in results would lead to this conclusion.

6. Page 16: The discussion of mechanisms of effect is much appreciated. However, could another mechanism be related to improvements in economic activity and the socio-economic status of women due to the environment of reduced conflict? Poverty is associated with adverse health outcomes and perhaps improvements in economic status could be another mechanism.

7. Page 17: Another limitation is that we do not know for certain whether women were directly exposed or even aware of the conflict events that occurred in their municipalities.

Tables and figures:

8. I suggest including all four municipality groups in Tables 1 and 2, for full transparency and information.

9. Table 2: Please define "Selective murder". Overall, it would be helpful to define all the conflict events in footnotes to the table. What is the difference between a terrorist attack and a massacre?

10. Figure 2: The changing of the scales on the Y-axis for July 2015 and August 2016 is confusing and makes it appear that there were dramatic and similar decreases for both ceasefires. Use of the same scales would make clearer the reason for the lack of effects for the August 2016 ceasefire. This is explained in a footnote, but may be missed by many readers.

Reviewer #2: See attachment

Michael Dewey

Reviewer #3: Article Review

Thank you for the opportunity to read the manuscript, "The Effects of Conflict Violence Reduction on Pregnancy Outcomes: Evidence from a Regression Discontinuity Design in Colombia". This interesting piece of work uses a quasi-experimental design to explore the causal effect of exposure to violence on pregnancy outcomes. The study benefits from a large sample and an interesting use of administrative data. The sensitivity analyses conducted (e.g., bandwidths and a parametric approach with first-, second-, and third-order polynomial specifications, placebo effects) are a valuable strategy to validate the RD design.

Abstract

- In the methods and findings paragraph, the authors indicate: "We found that the July 2015 ceasefire reduced the average number of conflict events to which women were exposed during pregnancy in their municipalities of residence (absolute effect -0·20; 95% CI -0·33 to -0·08)". I suggest specifying the effect size index reported here. Is it a difference in means?

Introduction

- Page 4, end of the second paragraph. The authors mention previous research testing the associations between conflict exposure and miscarriages, stillbirth and perinatal mortality. It is stated that all previous evidence is based on observational data. Did any of the studies involve a quasi-experimental design (propensity score matching, interrupted series analysis? If so, I would discuss the limitations of previous quasi-experimental evaluations.

Methods and results

- Please add some details on how the "probable conception date" was calculated.

- The authors assert that data were extracted from birth/death certificates. However, Table 1 in the findings section describes demographic characteristics of participants such as age, education, marital status, health insurance, number of children, among others. It is not clear if all this information was extracted from the birth/death certificates or from a different source of administrative data. If a different source of data was included in the study, was that data affected by missingness? Did you explore patterns of potential missing data? Please clarify.

- The description of the RD design and the sensitivity analyses are well developed and offer to the reader enough elements to evaluate the quality of the outcomes.

- It might help to describe in more detail the variables analysed in the study, for example, exposure to conflict events such as selective murder or sexual violence. How are these concepts operationalised?

Discussion

- The discussion on the role of stress as a potential mechanism linking exposure to violence and stillbirth and perinatal mortality deserves more elaboration. Given the aims of the Journal, it would be useful to further elaborate the biological mechanisms underpinning the effects of violence and adversity exposure on maternal and birth outcomes.

Reviewer #4: Congratulations on both an expertly implemented study using a novel analytical tool (RD) and for writing up your work in an accessible and enjoyable paper. I think this work is publication worthy but there are some minor errors I think need to be attended to. These I have captured on my annotated document attached. Good luck with your publication and follow up studies in this under-researched field.

Any attachments provided with reviews can be seen via the following link:

[LINK]

Submitted filename: buitrago.pdf

Click here for additional data file.

Submitted filename: PMEDICINE-D-21-00816_reviewerannotated.pdf

Click here for additional data file.

Submitted filename: Review report_D-21-00816.pdf

Click here for additional data file.

4 May 2021

Submitted filename: Respuestas revisores_final.docx

Click here for additional data file.

26 May 2021

Dear Dr. Buitrago,

Thank you very much for re-submitting your manuscript "Conflict Violence Reduction and Pregnancy Outcomes: A Regression Discontinuity Design in Colombia" (PMEDICINE-D-21-00816R2) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by three reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Jun 02 2021 11:59PM.

Sincerely,

Beryne Odeny,

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1) Thank you for providing your STROBE checklist. Please replace the line numbers with paragraph numbers per section (e.g. "Methods, paragraph 1"), since there will be no line numbers in the final published paper.

2) Please provide the English translation of the Spanish protocol and include it as a Supplementary Information file

3) Please provide p-values for estimates in lines 308-310

4) Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references.

a) Please ensure that weblinks are accessible and dates of access have been updated. For example, the weblink for reference #1 is no longer accessible.

b) Please provide a weblink for reference #3

Comments from Reviewers:

Reviewer #1: The authors have done an excellent job of responding to all the concerns of the editors and the multiple reviewers. The paper has been considerably strengthened and makes an excellent contribution to the literature. I have no further comments.

Reviewer #2: The authors have addressed my points.

Michael Dewey

Any attachments provided with reviews can be seen via the following link:

[LINK]

1 Jun 2021

Submitted filename: Respuestas revisores_final.docx

Click here for additional data file.

2 Jun 2021

Dear Dr Buitrago,

On behalf of my colleagues and the Academic Editor, Dr. Sarah Stock, I am pleased to inform you that we have agreed to publish your manuscript "Conflict Violence Reduction and Pregnancy Outcomes: A Regression Discontinuity Design in Colombia" (PMEDICINE-D-21-00816R3) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Beryne Odeny

Associate Editor

PLOS Medicine