Abstract
Background
Preterm birth (PTB) is a leading cause of neonatal mortality, particularly in sub-Saharan Africa where 40% of global neonatal deaths occur. We identified and combined demographic, clinical, and psychosocial correlates of PTB among Kenyan women to develop a risk score.
Methods
We used data from a prospective study enrolling HIV-negative women from 20 antenatal clinics in Western Kenya (NCT03070600). Depressive symptoms were assessed by study nurses using the Center for Epidemiologic Studies Depression Scale (CESD-10), intimate partner violence (IPV) with the Hurt, Insult, Threaten, Scream scale (HITS), and social support using the Medical Outcomes Survey scale (MOS-SSS). Predictors of PTB (birth < 37 weeks gestation) were identified using multivariable Cox proportional hazards models, clustered by facility. We used stratified k-fold cross-validation methods for risk score derivation and validation. Area under the receiver operating characteristic curve (AUROC) was used to evaluate discrimination of the risk score and Brier score for calibration.
Results
Among 4084 women, 19% had PTB (incidence rate: 70.9 PTB per 100 fetus-years (f-yrs)). Predictors of PTB included being unmarried (HR:1.29, 95% CI:1.08–1.54), lower education (years) (HR:0.97, 95% CI:0.94–0.99), IPV (HITS score ≥ 5, HR:1.28, 95% CI:0.98–1.68), higher CESD-10 score (HR:1.02, 95% CI:0.99–1.04), lower social support score (HR:0.99, 95% CI:0.97–1.01), and mild-to-severe depressive symptoms (CESD-10 score ≥ 5, HR:1.46, 95% CI:1.07–1.99). The final risk score included being unmarried, social support score, IPV, and MSD. The risk score had modest discrimination between PTB and term deliveries (AUROC:0.56, 95% CI:0.54–0.58), and Brier Score was 0.4672. Women considered “high risk” for PTB (optimal risk score cut-point) had 40% higher risk of PTB (83.6 cases per 100 f-yrs) than “low risk” women (59.6 cases per 100 f-ys; HR:1.6, 95% CI:1.2–1.7, p < 0.001).
Conclusion
A fifth of pregnancies were PTB in this large multi-site cohort; PTB was associated with several social factors amenable to intervention. Combining these factors in a risk score did not predict PTB, reflecting the multifactorial nature of PTB and need to include other unmeasured factors. However, our findings suggest PTB risk could be better understood by integrating mental health and support services into routine antenatal care.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12884-024-07058-x.
Keywords: Risk score, Preterm birth, Kenya, Maternal depression, Birth outcomes
Introduction
Preventable neonatal deaths account for nearly half of all deaths among children under five, totaling 2.4 million per year [1]. Many (43%) global neonatal deaths occur in in low- and middle-income countries (LMICs) of sub-Saharan Africa (SSA) [1], where gains in neonatal survival have been slowest [2, 3]. Preterm birth (PTB) contributes to neonatal deaths, particularly in low-resource settings of Africa where 40% of all neonatal deaths occurred among preterm births between 2012 and 2016 [4–6]. High attendance to maternal child health (MCH) care has improved maternal and child health survival across SSA, increasing the proportion of under 5 deaths attributable to the neonatal period [7]. New approaches to alleviating persisting disparities in neonatal morbidity and mortality are needed to achieve “last mile” gains in MCH.
Multiple factors are associated with increased risk of PTB, including maternal age (women < 18 and > 35 are at higher risk for PTB) [8], inadequate antenatal care attendance [9], and comorbidities (e.g., diabetes) [10]. Psychosocial factors such as depression [11], anxiety [12], and chronic stress [13, 14] increase risk of adverse perinatal outcomes, and risk may be further exacerbated by experiencing violence by an intimate partner (IPV) [15, 16]. Identifying risk factors for adverse perinatal outcomes can help prioritize pregnant women for enhanced monitoring in antenatal care, yet independent risk factors alone may be insufficient in screening for heightened risk when some factors may be additive.
Clinical prediction tools, such as “risk scores” or “calculators” are increasingly used in a range of healthcare settings to identify clients at increased risk of adverse health outcomes [17]. These tools offer a low-resource, efficient approach to aid healthcare workers in evaluating a client’s risk factors, assigning a score to differentiate “high” versus “low” overall risk for the outcome, and prioritizing “high risk” clients for enhanced monitoring and services. They may have utility in resource-limited settings to prioritize constrained healthcare worker time and services. While some risk factors for PTB are well documented, to date there is no risk scoring tool to identify pregnant women in SSA settings with high risk for PTB who may benefit from enhanced antenatal care. A few risk scores for PTB exist to date, yet these were predominantly developed in high-income settings, have low-to-moderate predictive capacity, and rely on factors less relevant in SSA settings such as smoking status [18–20]. One PTB prediction tool developed in Ethiopia included factors measured at the time of delivery, making it less useful for early detection and monitoring [21]. The persisting high risk of PTB in Kenya combined with high antenatal care attendance (> 90%) [22], make Kenya an ideal setting for development and validation of a risk scoring tool to predict PTB.
We aimed to derive and validate a novel clinical tool using a combination of maternal factors feasibly obtained within routine maternal child health care to identify pregnancies at higher risk for PTB among pregnant women in Western Kenya.
Methods
Study design and participants
This study was nested this analysis in the PrEP Implementation for Mothers in Antenatal Care study (PrIMA) conducted among pregnant women in Western Kenya to compare two models for pre-exposure prophylaxis implementation in a cluster randomized controlled trial (NCT03070600) [23]. We screened and enrolled women attending antenatal care (ANC) in 20 MCH clinics in Siaya and Homa Bay counties between January 15, 2018, and July 31, 2019. Women were eligible for the PrIMA study if they were pregnant, HIV-uninfected, ≥ 15 years old and were able to provide consent.
Study procedures
Data about demographics, pregnancy history, partner characteristics, and psychosocial factors were collected by study nurses through questionnaires administered in Kiswahili, Dholuo, or English languages with tablet-based REDCap surveys. Gestational age at enrollment was assessed by study nurses via ascertainment of last menstrual period (LMP) or fundal height. Study nurses collected data on pregnancy outcomes at the 6-week postpartum visit and abstracted data from medical records at study visits that aligned with antenatal and infant immunization visits recommended by the Kenya Ministry of Health. Visits occurred monthly during pregnancy, at 6 weeks postpartum, and three-monthly until 9 months postpartum.
Outcomes
The primary outcome was PTB, defined as a live or non-live birth < 37 weeks gestation. Gestational age was determined by study nurses at enrolment by ascertainment of last menstrual period (LMP) or fundal height (which was assessed in 2% of participants). In study settings where ultrasound is typically unavailable, LMP is used as a standard [24].
Potential predictors
We hypothesized multiple demographic, clinical, and social characteristics as potential predictors of adverse perinatal outcomes, including PTB, based on existing evidence (Supplementary Material Fig. 1). The Center for Epidemiologic Studies Depression Scale (CESD-10) was used to collect information about experience of depressive symptoms during pregnancy (enrollment visit). In this 10-item scale, higher severity of depressive symptoms is denoted by higher scores (absolute score range: 0–30) and a cut-point of 10 or greater denotes moderate-to-severe depressive symptoms (MSD, CESD-10 score ≥ 10) [25]. To identify differences between women with “any” versus “no” depressive symptoms, we evaluated mild-to-severe depressive symptoms (Mild-SD), defined as a cut-point of 5 or greater (CESD-10 score ≥ 5). The 18-item Medical Outcomes Study Social Support Survey (MOS-SSS) was used to collect social support data; higher scores describe higher social support (range: 18–90) [26]. To determine history of intimate partner violence (IPV), we used the 4-item Hurt, Insult, Threaten, Scream Scale with a cut-point of 10 or greater (absolute score range: 4–20) [27].
Experience of prior PTB was self-reported, as was diagnosis of sexually transmitted infection (prior 6 months). Syphilis diagnosis information was abstracted from the patient’s medical card at enrollment or serologically evaluated. Household crowding was used as a marker of socioeconomic status [28], defined as the ratio of people per room in a residence greater than the median (> 2 people/room). Maternal comorbidity was defined as self-report of diabetes, hypertension, or tuberculosis.
Overall, we evaluated the following potential predictors of PTB: Maternal age (years, continuous), educational attainment (years, continuous), regular employment (yes/no), being married or living with a partner (yes/no), household crowding (yes/no), financial support from partner (yes/no), maternal comorbidity (yes/no), recent STI (yes/no), recent syphilis infection (yes/no), multiparity (yes/no), prior preterm birth, IPV in last 2 weeks (yes/no), social support score (continuous), moderate-to-severe depressive symptoms (yes/no), mild-to-severe depressive symptoms (yes/no), depressive symptom score (continuous).
Statistical analysis -- Risk score derivation and validation
Participants were included in this analysis if they had a singleton birth, had information on gestational age at birth (implausible values of gestational age > 44 weeks were considered missing), did not acquire HIV during the study, and had depressive symptom information during pregnancy.
We used stratified k-fold cross-validation methods for risk score derivation and validation [29–32]. Cox proportional hazards models clustered by facility were used to identify potential predictors of PTB (p < 0.10), which were included in a multivariable model. A stepwise backwards Cox proportional hazards regression approach was used to identify the combination of variables that best predicted PTB, starting with the variables included in the multivariable model. Dichotomized versions of continuous variables were evaluated by identifying the optimal cut-point through signal detection receiver operating characteristic curve (ROC) analysis (Stata “cutpt” command) [33]. We used dichotomized versions if they were more predictive than the continuous version. This was the case for HITS score, where the most predictive cut-point was HITS score ≥ 5 versus < 5.
All variables were defined so that the direction of their association was a “risk factor” for PTB, except for continuous variables where higher scores were associated with lower risk of PTB (these variables had negative item-level scores). The model with lowest Akaike information criterion (AIC) from the models evaluated in stepwise analysis was identified as the best fit model for the risk score [34]. To calculate a score value for each predictor from the best fit model, we divided each factor’s regression coefficient by the absolute value of the lowest value coefficient, then rounded to the nearest integer. The final risk score for each woman was the sum of individual predictor scores.
To understand the extent to which the risk score differentiated women experiencing PTB from women without this experience (discrimination), we used ROC curve analysis to calculate area under the receiver operating characteristic curve (AUROC) for the risk score as a predictor of PTB [17, 32]. To understand how closely the predictions of the risk score match the observed outcomes in the data (calibration), we calculated a Brier score [17, 32]. Overall, useful risk scores have AUROC > 0.70 and Brier score < 0.25.
To internally validate the risk score, we first split the data into 10 equally sized random subsets (“folds”) that were stratified by PTB to ensure equal distribution of outcome risk in each fold. Meaningful differences were not found between participants in each fold. In 10 iterations using datasets comprised of 9 out of 10 folds each (90% of full dataset), we repeated all model derivation steps from the main risk score development process. Each of the 10 resulting risk scores were tested for their performance predicting PTB in the fold not used to derive that risk score (10% of full dataset). Estimated AUROC, sensitivity, specificity (at their respective optimal cut-points), and Brier score for each risk score were averaged to obtain overall measures of model performance in internal validation to which we compared the main risk score performance characteristics. Risk scores were dichotomized into the most predictive categories using the optimal cut-point identified through ROC curves and evaluated incidence of PTB by risk score group.
We imputed item-level scores for participants missing data in < 5 out of 10 depressive symptom scale items (11.8%, 492/4185, Supplementary Material Table 1) using the median score across the participant’s existing items (person-median imputation) [35]. We imputed item-level scores using person-median imputation among those missing < 8 out of 16 social support scale items (2.6%, 108/4185, Supplementary Material Table 2).
Sensitivity analyses
In sensitivity analyses, we evaluated the risk score among sub-groups stratified by trimester of enrollment to understand if the risk score was more predictive of PTB in particular trimesters. We additionally evaluated the risk score among groups stratified by adolescent and young adult women (< 25 years) versus adult women (≥ 25 years). Analyses were conducted using Stata 15.
Ethical considerations
We received approval from the Kenyatta National Hospital-University of Nairobi Ethics Research Committee and University of Washington Human Subjects Review Committee for the study protocol, informed consent forms, and data collection tools. The study received approval from Siaya and Homabay county Departments of Health and facility administrators. All participants provided written informed consent.
Results
Overall, 4084 (92%, 4084/4447) PrIMA study participants were included in this analysis. These participants had information about PTB outcome, did not acquire HIV during the study, had a singleton birth, and had depressive symptom information during pregnancy. Median maternal age was 24 years (interquartile range [IQR]: 21–28), median gestational age at enrollment in PrIMA was 24 weeks (IQR: 20–29), and median educational attainment was 10 years (IQR: 8–12) (Table 1). The majority of participants were multiparous (74.4%, 3035/4084) and married or living with their partner (84.8%, 3433/4084).
Table 1.
Baseline characteristics of PrIMA study participants included in analysis
| Overall (n = 4084) | Preterm birth (n = 780) | No preterm birth (n = 3304) | ||||
|---|---|---|---|---|---|---|
| Demographic characteristics | N | n (%) or Median (IQR) | n | n (%) or Median (IQR) | n | n (%) or Median (IQR) |
| Age (years) | 4082 | 24 (21, 28) | 780 | 24 (21, 28) | 3302 | 24 (21, 28) |
| Adolescents and young women (≤ 24 years) | 4082 | 2332 (57.1%) | 780 | 467 (59.9) | 3302 | 1865 (56.5) |
| Gestational age (enrollment, weeks) | 4084 | 24 (20, 29) | 780 | 24 (18, 28) | 3304 | 24 (20, 30) |
| Married or living with a partner | 4050 | 3433 (84.8%) | 777 | 652 (83.9) | 3273 | 2781 (85.0) |
| Completed education (years) | 4084 | 10 (8, 12) | 768 | 10 (8, 12) | 3235 | 10 (8, 12) |
| Regularly employed | 4032 | 603 (15.0%) | 772 | 118 (15.3) | 3260 | 485 (14.9) |
| Financial support from partner | 4084 | 3616 (88.5%) | 780 | 694 (89.0) | 3304 | 2922 (88.4) |
| Household crowding (≥ 2 people/room) | 4057 | 1960 (48.3%) | 777 | 378 (48.6) | 3280 | 1582 (48.2) |
| Pregnancy history & reproductive health factors | ||||||
| Multiparous | 4079 | 3035 (74.4%) | 779 | 581 (74.6) | 3300 | 2454 (74.4) |
| Prior preterm birth | 4084 | 42 (1.0%) | 780 | 6 (0.8%) | 3304 | 36 (1.1) |
| Maternal comorbidity | 4084 | 45 (1.1%) | 780 | 11 (1.4) | 3304 | 34 (1.0) |
| Trimester of initial antenatal care (ANC) visit | 4084 | 780 | 3304 | |||
| First | 610 (14.9%) | 121 (15.5) | 489 (14.8) | |||
| Second | 2092 (51.2%) | 443 (56.8) | 1649 (49.9) | |||
| Third | 1382 (33.8%) | 216 (27.7) | 1166 (35.3) | |||
|
Sexually transmitted infection (prior 6 months) |
4077 | 103 (2.5%) | 779 | 16 (2.1) | 3298 | 87 (2.6) |
| Syphilis infection (enrollment) | 4016 | 42 (1.0%) | 757 | 8 (1.1) | 3259 | 34 (1.0) |
| Psychosocial characteristics | ||||||
| Social support scoreb | 4084 | 75 (63, 88) | 780 | 73 (58.5, 86) | 3304 | 76 (64, 88) |
| Intimate partner violence c (HITS score ≥ 10) | 4079 | 318 (7.8%) | 778 | 48 (6.2) | 3301 | 270 (8.2) |
| CESD-10 score | 4084 | 5 (3, 9) | 780 | 6 (3, 10) | 3304 | 773 (23.4) |
|
Mild-to-severe depressive symptoms (CESD-10 ≥ 5) |
4084 | 2239 (54.8%) | 780 | 492 (63.1) | 3304 | 1747 (52.9) |
|
Moderate-to-severe depressive symptoms (CESD-10 ≥ 10) |
4084 | 975 (23.9%) | 780 | 202 (25.9) | 3304 | 5 (2, 9) |
*Among those with current partners
bWe evaluated social support using the 18-item Medical Outcomes Study social support score (MOS-SSS)
cWe evaluated intimate partner violence using the 4-item Hurt, Insult, Threaten, and Scream scale (HITS), defining intimate partner violence as scores of 10 and above (IPV: HITS score ≥ 10 = “Yes”, HITS score < 10 = “No”)
CESD-10: Center for epidemiologic studies depression scale-10
MSD: Moderate-to-severe depressive symptoms
About a quarter of women (23.9%, 975/4084) reported MSD during pregnancy (median CESD-10 score: 5, IQR: 3–9), and over 50% of women reported mild-SD (54.8%, 2239/4084). Median MOS-SSS score was 75, IQR: 63–88), and 8% reported IPV at enrollment (7.8%, 318/4084). Prior to imputing CESD-10 and MOS-SSS items for partial scores (missing < 50% of items) 36.9% of women had low social support (median social support score: 75, IQR: 63–88) and 24.6% of women had MSD (median CESD-10 score: 5, IQR: 3–9). Those with complete vs. partial CESD-10 data were not meaningfully different in most factors evaluated (Supplementary Material Table 3).
Risk score to predict preterm birth
Overall, one in five (19%, 780/4084) women experienced PTB over 1099.3 person-years of follow-up for an incidence rate of 70.9 cases per 100 fetus-years (95% Confidence Interval [CI]: 66.1–76.1). Risk of PTB was higher among women who were unmarried and not living with a partner (HR: 1.3, 95% CI: 1.1–1.6, p = 0.0035) compared to married women. On average, every additional year of educational attainment was associated with lower risk of PTB (HR: 0.97, 95% CI: 0.94–0.99, p = 0.0383). Women experiencing IPV (HITS score ≥ 5) had higher risk of PTB (HR: 1.3, 95% CI: 0.9–1.7, 0.0697) compared to those with lower HITS scores (< 5). Women with higher CESD-10 scores were at higher risk of PTB (HR: 1.02, 95% CI: 0.98–1.68, p = 0.0632), as well as those with mild-to-severe depressive symptoms (HR: 1.5, 95% CI: 1.1–1.9, p = 0.0157), compared to those reporting low or no depressive symptoms. Those with low social support (MOS-SSS score ≤ 75; identified by Stata “cutpt” command) had higher risk of PTB (HR: 1.3, 95% CI: 0.90–1.84, p = 0.1625) (Table 2). Being unmarried, experiencing IPV, having low social support, and having mild-to-severe depressive symptoms comprised those included in the model with lowest AIC in stepwise backwards Cox regression (Table 2).
Table 2.
Correlates of preterm birth and item-level risk scores to predict preterm birth
| Baseline characteristics | Univariate analysis | Multivariate analysis | Stepwise | Coefficient | Risk score | |||
|---|---|---|---|---|---|---|---|---|
| Crude HR (95% CI) |
p-value | Adjusted HR (95% CI) | p-value | Adjusted HR (95% CI) |
p-value | β | ||
| Demographic characteristics | ||||||||
| Age (years) | 0.99 (0.97–1.01) | 0.2526 | ||||||
| Adolescents and young women (< 24 years) | 1.11 (0.92–1.35) | 0.2740 | ||||||
| Not married or living with a partner | 1.29 (1.08–1.54) | 0.0044 | 1.34 (1.11–1.62) | 0.0023 | 1.34 (1.12–1.62) | 0.0018 | 0.29 | 2 |
| Completed education (years) | 0.97 (0.94–0.99) | 0.0386 | ||||||
| Educational attainment ≥ 2.5 years | 0.87 (0.74–1.03) | 0.1087 | 0.99 (0.84–1.18) | 0.9432 | ||||
| Regularly employed | 1.00 (0.65–1.56) | 0.9901 | ||||||
| Financial support from partner | 0.88 (0.67–1.14) | 0.3366 | ||||||
| Household crowding (≥ 2 people/room) | 1.01 (0.97–1.06) | 0.5832 | ||||||
| Pregnancy history & reproductive health factors | ||||||||
| Multiparous | 0.97 (0.79–1.20) | 0.8003 | ||||||
| Prior preterm birth | 0.64 (0.34–1.20) | 0.1653 | ||||||
| Maternal comorbidity | 1.24 (0.79–1.94) | 0.3411 | ||||||
| Sexually transmitted infection (prior 6 months) | 0.87 (0.49–1.54) | 0.6289 | ||||||
| Syphilis infection (enrollment) | 1.08 (0.50–2.29) | 0.8511 | ||||||
| Psychosocial characteristics | ||||||||
| Social support scorea | 0.99 (0.97-1.00) | 0.0953 | ||||||
| Low social support (MOS-SSS ≤ 75)b | 1.29 (0.90–1.84) | 0.1625 | 1.19 (0.82–1.72) | 0.3629 | 1.19 (0.82–1.72) | 0.3641 | 0.17 | 1 |
| Intimate partner violence (HITS score ≥ 5)c | 1.28 (0.98–1.68) | 0.0697 | 1.22 (0.88–1.69) | 0.2392 | 1.22 (0.88–1.69) | 0.2400 | 0.19 | 1 |
| CESD-10 scored | 1.02 (0.99–1.04) | 0.0632 | ||||||
|
Mild-to-severe depressive symptoms (CESD-10 ≥ 5)e |
1.46 (1.07–1.99) | 0.0157 | 1.35 (0.98–1.87) | 0.0650 | 1.35 (0.98–1.87) | 0.0657 | 0.30 | |
|
Moderate-to-severe depressive symptoms (CESD-10 ≥ 10)f |
1.14 (0.82–1.60) | 0.4123 | 2 | |||||
aWe evaluated social support using the 18-item Medical Outcomes Study social support score (MOS-SSS)
bWe evaluated low social support as MOS-SSS score ≤ 75 based on the Stata cutpt function which identified the most predictive cut-point for a dichotomized variable
cWe evaluated intimate partner violence using the 4-item Hurt, Insult, Threaten, and Scream scale (HITS). In risk score development, we used a cutpoint of HITS score ≥ 5 vs. HITS score < 5 based on the most predictive cutpoint for preterm birth
dCESD-10: Center for epidemiologic studies depression scale-10
eMild-to-severe depressive symptoms was defined as CESD-10 ≥ 5 vs. CESD-10 < 5
fMSD: Moderate-to-severe depressive symptoms was defined as CESD-10 ≥ 10 vs. CESD-10 < 10
The ROC curves for the risk score and the multivariable model overlapped (Fig. 1), indicating equal predictive capacity. Median risk score among all women was 2 (IQR: 1–4, range: 0–6). The risk score had an AUROC of 0.56 (95% CI: 0.54–0.58). At the optimal cut-point (> 2.5 vs. ≤2.5), sensitivity was 55.6% (95% CI: 52.1–59.2) and specificity was 52.7% (95% CI: 51.0-54.4). The Brier score was 0.4672. Incidence of PTB differed between groups dichotomized by the optimal cut-point for the risk score, where incidence of PTB was 83.6 cases per 100 fetus-years among those with risk scores > 2.5, and 59.6 cases per 100 fetus-years among those with score ≤ 2.5 (HR: 1.4, 95% CI: 1.2–1.7, p-value < 0.001) (Table 3).
Fig. 1.
Receiver operating characteristic curve and cut-points of risk score
Table 3.
Area under the receiver operating curve overall and by trimester of enrollment
| N | AUROC | 95% CI | |
|---|---|---|---|
| Overall | 4084 | 0.561 | 0.540–0.582 |
| 1st Trimester | 406 | 0.573 | 0.507–0.639 |
| 2nd Trimester | 2059 | 0.551 | 0.523–0.579 |
| 3rd Trimester | 1619 | 0.579 | 0.544–0.614 |
| Adolescents and young adults (< 25 years) | 2332 | 0.547 | 0.519–0.574 |
| Adult women (≥ 25 years) | 1750 | 0.579 | 0.547–0.612 |
In sensitivity analyses, the discrimination of the risk score did not improve among groups defined by trimester of enrollment or in groups stratified by age group (< 25 years versus ≥ 25 years) (all AUCs < 0.58; Table 3).
Risk score validation
After dividing the dataset into 10 random and equal subsets, we did not detect meaningful differences in characteristics between women included in these validation subsets or the overall dataset, and stratification ensured incidence of PTB was equal across groups (Table 4). The AUROC, sensitivities and specificities (at their respective optimal cut-points), and Brier scores were similar across validation subsets and the overall dataset. The mean AUROC across iterations was 0.56 (95% CI: 0.49–0.63) and mean Brier score was 0.5765. In all risk scores from validation subsets, the incidence of PTB in the group with “high risk” scores were statistically significantly higher than the incidence of PTB in the group with “low risk” scores (data not shown). Averaged across validation subsets, the sensitivity at the optimal risk score cut-point was 61.5% and specificity was 49.7%. The risk score performance characteristics were similar for the main risk score and the average metrics across validation subsets, indicating strong internal validity (Table 3).
Table 4.
Discrimination Performance of Risk Score in Development and K-fold Cross-Validation cohorts
| Risk of preterm birth | Discrimination performance | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Cohort | N |
Women with high risk score
n (%) |
No. of cases | Fetus-years | Incidence rate per 100 fetus-years (95% CI) | AUC (95% CI) | Brier score | Sensitivity* | Specificity* |
| Overall (derivation) | 4084 | 2029 (48.9%) | 780 | 1099.3 | 70.9 (66.1–76.1) | 0.56 (0.54–0.58) | 0.4672 | 55.6% | 52.7% |
| 1 | 408 | 224 (53.9%) | 78 | 114.2 | 68.3 (54.7–85.2) | 0.58 (0.56–0.60) | 0.4730 | 67.9% | 49.1% |
| 2 | 408 | 234 (56.4%) | 78 | 111.7 | 69.8 (55.9–87.1) | 0.58 (0.55–0.59) | 0.4755 | 71.8% | 47.9% |
| 3 | 408 | 238 (78.0%) | 78 | 107.7 | 72.4 (58.0-90.4) | 0.57 (0.55–0.60) | 0.6545 | 85.5% | 23.2% |
| 4 | 409 | 204 (49.6%) | 78 | 113.2 | 68.9 (55.2–86.0) | 0.59 (0.56–0.61) | 0.4938 | 50.0% | 50.8% |
| 5 | 409 | 207 (49.8%) | 78 | 110.2 | 70.8 (56.7–88.4) | 0.58 (0.56–0.60) | 0.4719 | 57.7% | 51.7% |
| 6 | 408 | 223 (53.7%) | 78 | 110.7 | 70.4 (56.4–87.9) | 0.58 (0.56–0.60) | 0.4975 | 59.0% | 48.2% |
| 7 | 408 | 206 (49.6%) | 78 | 105.7 | 73.8 (59.1–92.1) | 0.57 (0.55–0.59) | 0.4436 | 65.4% | 53.3% |
| 8 | 408 | 216 (52.1%) | 78 | 108.4 | 71.9 (57.6–89.9) | 0.57 (0.55–0.60) | 0.4730 | 62.8% | 50.3% |
| 9 | 409 | 225 (54.1%) | 78 | 108.3 | 72.0 (57.7–89.9) | 0.58 (0.56–0.60) | 0.5061 | 59.0% | 47.1% |
| 10 | 409 | 201 (49.1%) | 78 | 109.2 | 71.4 (57.2–89.2) | 0.58 (0.56–0.60) | 0.7818 | 55.2% | 54.6% |
| Average (validation) | 4084 | -- | -- | -- | -- | 0.56 (0.49–0.63) | 0.5765 | 61.5% | 49.7% |
*Sensitivity and specificity at the optimal cut-point of each respective risk score
Discussion
In this large, multi-site, longitudinal study among Kenyan pregnant women, we determined risk factors for PTB and endeavored to develop and validate a risk score to identify women at higher risk of PTB. We envisioned that this tool could be implemented in routine antenatal care settings in sub-Saharan Africa to prioritize women for enhanced care throughout pregnancy with the aim of preventing PTB and ultimately improving neonatal survival. Overall, 19% of births were preterm -- higher than PTB estimates for sub-Saharan Africa as a whole (12%) [36, 37], but similar to other studies in Kenya [38]. Several pregnancy and psychosocial predictors of PTB were identified, including being unmarried, having lower educational attainment, experiencing IPV, having lower social support, and experiencing mild-to-severe depressive symptoms. Despite these associations, the combination risk score was not useful in discriminating between cases of PTB and non-cases (AUC:0.57; minimum “useful” AUC: ≥0.70). Our findings contribute new information to inform future risk prediction tool development for this important population, with the goal of preventing PTB to improve neonatal survival.
A few risk assessment tools for adverse perinatal outcomes have been developed globally with low-to-moderate prediction accuracy [18, 19, 21, 39]. Most tools were developed in US or Europe and have limited utility in sub-Saharan African antenatal care clinics since they include clinical factors not routinely assessed in these settings, such as fetal fibronectin in vaginal fluid [19], or behavioral factors less common in SSA settings like smoking [20]. One risk score developed among Ethiopian pregnant women to predict PTB performed well (AUROC: 0.816) and included some maternal factors readily collected in routine ANC (residence, gravidity) as well as others less-routinely collected in public sector ANC (hemoglobin < 11 mg/dL, early rupture of membranes, antepartum hemorrhage, pregnancy-induced hypertension) [21]. Some of the variables included in this risk score were obtained at delivery, which limits utility for predicting women at-risk for PTB for care during the antenatal period [21]. These variables often occur simultaneously with PTB and may have the same pathophysiology which make them less useful for predicting PTB. We attempted to develop a model for use in earlier antenatal care to identify women earlier in pregnancy who may be at risk for PTB, using factors easily collected by lower cadre healthcare workers in routine sub-Saharan African antenatal care settings.
The predictors of PTB in our analysis were consistent with factors identified in prior studies. Unmarried women are at increased risk globally for preterm, low birthweight, and small-for-gestational age births [40, 41]. Linkages between lower educational attainment and heightened risk for PTB are well-documented, particularly in high-income country settings [42, 43], however a recent analysis of Demographic and Health Surveys from > 170,000 reproductive-aged women from 36 sub-Saharan African countries found that women with higher educational achievement had higher risk for PTB [41]. Our data support lower educational attainment as a risk factor for PTB. Increasingly, psychosocial factors including social support, depressive symptoms, and intimate partner violence are understood as important factors associated with adverse perinatal outcomes like PTB [11, 16, 44]. Our study supported these factors as potential risk factors for PTB among Kenyan pregnant women. Our study did not identify wealth index, employment status, prior PTB, or parity/gravidity as correlates of PTB.
Overall, while our study identified predictors of PTB that are routinely and/or easily collected in ANC clinics in SSA, the effect sizes in combination were not strong enough to discriminate cases from non-cases in our risk score. Regardless of the poor predictive performance, women with risk scores indicating “high risk” still had 60% higher risk of PTB compared to women with lower scores. Unmarried women with lower educational attainment and those experience intimate partner violence, depressive symptoms, or low social support be prioritized for enhanced monitoring throughout antenatal care. Routine screening for intimate partner violence and depressive symptoms should be integrated into maternal child health in sub-Saharan Africa as an important component of addressing maternal psychosocial distress to reduce risk of PTB.
Limitations
We hypothesized that the poor performance of our risk score could be due to failure to account for selection bias in estimates of PTB, based on differential enrollment by exposure and outcome status among those enrolling later versus earlier in gestation. Therefore, a sensitivity analysis was conducted to evaluate discrimination of the risk score among groups defined by trimester of enrollment. This did not improve discrimination of the risk score (AUC for all trimesters: ≤0.59). Further, we hypothesized that the risk score may discriminate differentially among adolescent or young adult women (< 25) years compared to adult women (≥ 25 years), yet stratification by age did not improve discrimination (AUC for both age groups ≤ 0.59). Our risk score could have been improved by including other predictors that were not measured in our study, such as bleeding in pregnancy, hypertension, or indoor air quality [45].
Conclusion
Among a large cohort of pregnant Kenyan women, a fifth of pregnancies were preterm, and this outcome was higher among women who were unmarried, those with lower educational attainment, and those with mild-to-severe depressive symptoms. Our risk score that combined marital status, years of education, and psychosocial factors of intimate partner violence, social support, and depressive symptoms did not yield high discriminatory capacity for predicting preterm birth. However, our data suggest subgroups of women who may benefit from increased antenatal support (unmarried, lower social support, IPV), and that integrating depression screening into antenatal care with appropriate referral for counseling or psychotherapy could potentially reduce risk for preterm birth.
Electronic Supplementary Material
Below is the link to the electronic supplementary material.
Acknowledgements
The authors thank all study participants for their time and information. We also thank the full study team.
Author contributions
AL-conceptualized the idea for the article, had access to raw data, performed and verified data analyses, wrote the first draft, conducted the edits for the drafts, drafted the figures and tables, reviewed and approved final draft, and made the decision to submit for publication.JP-reviewed the data in the figures and tables, had access to raw data, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.FA- managed study operations and study team, had access to raw data, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.JD- reviewed drafts and provided substantial edits and contributed towards final draft, had access to raw data, reviewed and approved final draft.LG- reviewed drafts and provided substantial edits and contributed towards final draft, had access to raw data, reviewed and approved final draft.MM- performed data management, had access to raw data, accessed and verified the data, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.NN- reviewed drafts and provided substantial edits and contributed towards final draft, had access to raw data, reviewed and approved final draft.BO- reviewed drafts and provided substantial edits and contributed towards final draft, had access to raw data, reviewed and approved final draft.BR-oversaw statistical analysis, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.SW- performed data management, had access to raw data, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.JS- performed data checking of analysis, had access to raw data, reviewed drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft.JK- directed study operations, conceptualized the idea for the paper, reviewed all drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft, and made the decision to submit for publication.GJS- conceptualized the idea for the article, reviewed the data in the figures and tables, reviewed all drafts and provided substantial edits and contributed towards final draft, reviewed and approved final draft, and made the decision to submit for publication.Conflict of interest: The authors do not have conflicts to report.
Funding
This work was supported by the National Institute of Allergy and Infectious Disease (R01 AI125498 to GJS and P30AI027757), the Eunice Kennedy Shriver National Institute of Child Health and Human Development (F31HD101149 to AL, F32HD108857 to AL, R01HD100201 to JP and R01 HD094630 to GJS). The funding agencies had no role in the writing of the manuscript or the decision to submit it for publication.
Data availability
The datasets used during the current study are available from the corresponding author on reasonable request.
Declarations
Ethics approval and consent to participate
We received approval from the Kenyatta National Hospital-University of Nairobi Ethics Research Committee and University of Washington Human Subjects Review Committee for the study protocol, informed consent forms, and data collection tools. We also received approval from Siaya and Homabay county Departments of Health and facility administrators. All participants provided written informed consent. All study procedures were performed in accordance with guidelines and regulations for research involving human subjects. This trial was registered at ClinicalTrials.gov on January 15, 2018 (NCT03070600). The Community Advisory Board for HIV/AIDS related studies among the Kenyatta National Hospital and University of Washington collaboration provided feedback to the parent study protocol.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.You D, Hug L, Ejdemyr S, Beise J, Idele P, UN IGME. Levels & Trends in Child Mortality.; 2015. http://www.childmortality.org/files_v20/download/IGME Report 2015_9_3 LR Web.pdf.
- 2.Lawn J, Kerber K, Enweronu-Laryea C, Cousens S. 3.6 Million Neonatal Deaths-What Is Progressing and What Is Not? Semin Perinatol. Published online Oct. 2010;17. 10.1053/j.semperi.2010.09.011. [DOI] [PubMed]
- 3.Lawn JE, Blencowe H, Oza S, et al. Every Newborn: progress, priorities, and potential beyond survival. Lancet (London England). 2014;384(9938):189–205. 10.1016/S0140-6736(14)60496-7. [DOI] [PubMed] [Google Scholar]
- 4.Tamirat KS, Sisay MM, Tesema GA, Tessema ZT. Determinants of adverse birth outcome in Sub-Saharan Africa: analysis of recent demographic and health surveys. BMC Public Health. 2021;21(1):1–10. 10.1186/S12889-021-11113-Z/TABLES/3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hug L, You D, Blencowe H, et al. Global, regional, and national estimates and trends in stillbirths from 2000 to 2019: a systematic assessment. Lancet. 2021;398(10302):772–85. 10.1016/S0140-6736(21)01112-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nisar MI, Yoshida S. Population-based rates, risk factors and consequences of preterm births in South-Asia and sub-Saharan Africa: A multi-country prospective cohort study. J Glob Health. 2022;12:4011. 10.7189/JOGH.12.04011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lassi ZS, Salam RA, Das JK, Bhutta ZA. Essential interventions for maternal, newborn and child health: Background and methodology. Reprod Health. 2014;11(1):1–7. 10.1186/1742-4755-11-S1-S1/TABLES/3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Londero AP, Rossetti E, Pittini C, Cagnacci A, Driul L. Maternal age and the risk of adverse pregnancy outcomes: A retrospective cohort study. BMC Pregnancy Childbirth. 2019;19(1):1–10. 10.1186/S12884-019-2400-X/TABLES/4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Heaman MI, Newburn-Cook CV, Green CG, Elliott LJ, Helewa ME. Inadequate prenatal care and its association with adverse pregnancy outcomes: A comparison of indices. BMC Pregnancy Childbirth. 2008;8(1):1–8. 10.1186/1471-2393-8-15/TABLES/6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Farrar D, Simmonds M, Bryant M, et al. Hyperglycaemia and risk of adverse perinatal outcomes: systematic review and meta-analysis. BMJ. 2016;354. 10.1136/BMJ.I4694. [DOI] [PMC free article] [PubMed]
- 11.Grote NK, Bridge JA, Gavin AR, Melville JL, Iyengar S, Katon WJ. A Meta-analysis of Depression During Pregnancy and the Risk of Preterm Birth, Low Birth Weight, and Intrauterine Growth Restriction. Arch Gen Psychiatry. 2010;67(10):1012. 10.1001/archgenpsychiatry.2010.111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Rose MS, Pana G, Premji S. Prenatal Maternal Anxiety as a Risk Factor for Preterm Birth and the Effects of Heterogeneity on This Relationship: A Systematic Review and Meta-Analysis. Biomed Res Int. 2016;2016:1–18. 10.1155/2016/8312158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dunkel Schetter C, Tanner L. Anxiety, depression and stress in pregnancy: implications for mothers, children, research, and practice. Curr Opin Psychiatry. 2012;25(2):141. 10.1097/YCO.0B013E3283503680. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Borders AEB, Grobman WA, Amsden LB, Holl JL. Chronic Stress and Low Birth Weight Neonates in a Low-Income Population of Women. Obstet Gynecol. 2007;109(2, Part 1):331–8. 10.1097/01.AOG.0000250535.97920.b5. [DOI] [PubMed] [Google Scholar]
- 15.Alhusen JL, Bullock L, Sharps P, Schminkey D, Comstock E, Campbell J. Intimate partner violence during pregnancy and adverse neonatal outcomes in low-income women. J Womens Health (Larchmt). 2014;23(11):920–6. 10.1089/jwh.2014.4862. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Yaya S, Odusina EK, Adjei NK, Uthman OA. Association between intimate partner violence during pregnancy and risk of preterm birth. BMC Public Health. 2021;21(1):1–9. 10.1186/S12889-021-11625-8/FIGURES/1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology. 2010;21(1):128–38. 10.1097/EDE.0B013E3181C30FB2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Honest H, Bachmann LM, Sundaram R, Gupta JK, Kleijnen J, Khan KS. The accuracy of risk scores in predicting preterm birth—a systematic review. https://doi.org/101080/01443610410001685439. 2009;24(4):343–59. 10.1080/01443610410001685439 [DOI] [PubMed]
- 19.Stock SJE, Horne M, Bruijn M, et al. Development and validation of a risk prediction model of preterm birth for women with preterm labour symptoms (the QUIDS study): A prospective cohort study and individual participant data meta-analysis. PLOS Med. 2021;18(7):e1003686. 10.1371/JOURNAL.PMED.1003686. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kim JI, Lee JY. Systematic Review of Prediction Models for Preterm Birth Using CHARMS. https://doi.org/101177/10998004211025641. 2021;23(4):708–722. 10.1177/10998004211025641 [DOI] [PubMed]
- 21.Feleke SF, Anteneh ZA, Wassie GT, Yalew AK, Dessie AM. Developing and validating a risk prediction model for preterm birth at Felege Hiwot Comprehensive Specialized Hospital, North-West Ethiopia: a retrospective follow-up study. BMJ Open. 2022;12(9):e061061. 10.1136/BMJOPEN-2022-061061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Demographic and Health Survey - Kenya.; 2014. Accessed June 1, 2019. www.DHSprogram.com.
- 23.JC D. PrEP Implementation for Mothers in Antenatal Care (PrIMA): Study Protocol of a Cluster Randomised Trial. BMJ Open. 2019;9(3). 10.1136/BMJOPEN-2018-025122. [DOI] [PMC free article] [PubMed]
- 24.Macaulay S, Buchmann EJ, Dunger DB, Norris SA. Reliability and validity of last menstrual period for gestational age estimation in a low-to-middle-income setting. J Obstet Gynaecol Res. 2019;45(1):217–25. 10.1111/JOG.13801. [DOI] [PubMed] [Google Scholar]
- 25.Andresen EM, Malmgren JA, Carter WB, Patrick DL. Screening for depression in well older adults: evaluation of a short form of the CES-D (Center for Epidemiologic Studies Depression Scale). Am J Prev Med. 10(2):77–84. Accessed September 8, 2019. http://www.ncbi.nlm.nih.gov/pubmed/8037935 [PubMed]
- 26.Sherbourne CD, Stewart AL. The MOS social support survey. Soc Sci Med. 1991;32(6):705–14. 10.1016/0277-9536(91)90150-B. [DOI] [PubMed] [Google Scholar]
- 27.Rabin RF, Jennings JM, Campbell JC, Bair-Merritt MH. Intimate partner violence screening tools: a systematic review. Am J Prev Med. 2009;36(5):439–e4454. 10.1016/j.amepre.2009.01.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Melki IS, Beydoun HA, Khogali M, Tamim H, Yunis KA, National Collaborative Perinatal Neonatal Network (NCPNN). Household crowding index: a correlate of socioeconomic status and inter-pregnancy spacing in an urban setting. J Epidemiol Community Heal. 2004;58(6):476–80. 10.1136/jech.2003.012690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ellis RP, Mookim PG. Cross-Validation Methods for Risk Adjustment Models. Published online 2009.
- 30.Taylor JMG, Ankerst DP, Andridge RR. Validation of Biomarker-based risk prediction models. Clin Cancer Res. 2008;14(19):5977. 10.1158/1078-0432.CCR-07-4534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Zhang Z, Zhang H, Khanal MK. Development of scoring system for risk stratification in clinical medicine: a step-by-step tutorial. Ann Transl Med. 2017;5(21):436. 10.21037/atm.2017.08.22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Grant SW, Collins GS, Nashef SAM. Statistical Primer: developing and validating a risk prediction model. Eur J Cardio-Thoracic Surg. 2018;54(2):203–8. 10.1093/EJCTS/EZY180. [DOI] [PubMed] [Google Scholar]
- 33.Kiernan M, Kraemer HC, Winkleby MA, King AC, Taylor CB. Do logistic regression and signal detection identify different subgroups at risk? Implications for the design of tailored interventions. Psychol Methods. 2001;6(1):35–48. 10.1037/1082-989X.6.1.35. [DOI] [PubMed] [Google Scholar]
- 34.Vrieze SI. Model selection and psychological theory: a discussion of the differences between the Akaike information criterion (AIC) and the Bayesian information criterion (BIC). Psychol Methods. 2012;17(2):228–43. 10.1037/A0027127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bono C, Ried LD, Kimberlin C, Vogel B. Missing data on the Center for Epidemiologic Studies Depression Scale: A comparison of 4 imputation techniques. Res Soc Adm Pharm. 2007;3(1):1–27. 10.1016/J.SAPHARM.2006.04.001. [DOI] [PubMed] [Google Scholar]
- 36.Chawanpaiboon S, Vogel JP, Moller AB, et al. Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis. Lancet Glob Heal. 2019;7(1):e37–46. 10.1016/S2214-109X(18)30451-0/ATTACHMENT/3B7A79C8-B16F-410C-A7C7-3B3039631E52/MMC1.PDF. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Vogel JP, Chawanpaiboon S, Moller A-B, Watananirun K, Bonet M, Lumbiganon P. The global epidemiology of preterm birth. Published online 2018. 10.1016/j.bpobgyn.2018.04.003 [DOI] [PubMed]
- 38.Wagura P, Wasunna A, Laving A, Wamalwa D. Ng’ang’a P. Prevalence and factors associated with preterm birth at kenyatta national hospital. BMC Pregnancy Childbirth. 2018;18(1):107. 10.1186/s12884-018-1740-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Kayode GA, Grobbee DE, Amoakoh-Coleman M, et al. Predicting stillbirth in a low resource setting. BMC Pregnancy Childbirth. 2016;16(1):1–8. 10.1186/S12884-016-1061-2/FIGURES/3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Shah PS, Zao J, Ali S. Maternal marital status and birth outcomes: A systematic review and meta-analyses. Matern Child Health J. 2011;15(7):1097–109. 10.1007/S10995-010-0654-Z/TABLES/5. [DOI] [PubMed] [Google Scholar]
- 41.Alamneh TS, Teshale AB, Worku MG, et al. Preterm birth and its associated factors among reproductive aged women in sub-Saharan Africa: evidence from the recent demographic and health surveys of sub-Sharan African countries. BMC Pregnancy Childbirth. 2021;21(1):1–11. 10.1186/S12884-021-04233-2/TABLES/3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Ruiz M, Goldblatt P, Morrison J, et al. Mother’s education and the risk of preterm and small for gestational age birth: a DRIVERS meta-analysis of 12 European cohorts. J Epidemiol Community Health. 2015;69(9):826. 10.1136/JECH-2014-205387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Alterman N, Johnson S, Carson C, et al. Gestational age at birth and academic attainment in primary and secondary school in England: Evidence from a national cohort study. PLoS ONE. 2022;17(8):e0271952. 10.1371/JOURNAL.PONE.0271952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Szegda K, Markenson G, Bertone-Johnson ER, Chasan-Taber L. Depression during pregnancy: a risk factor for adverse neonatal outcomes? A critical review of the literature. J Matern Fetal Neonatal Med. 2014;27(9):960–7. 10.3109/14767058.2013.845157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Younger A, Alkon A, Harknett K, Jean Louis R, Thompson LM. Adverse birth outcomes associated with household air pollution from unclean cooking fuels in low- and middle-income countries: A systematic review. Environ Res. 2022;204:112274. 10.1016/J.ENVRES.2021.112274. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used during the current study are available from the corresponding author on reasonable request.