Women’s physical health around live births and pregnancy losses: a longitudinal study

Abstract

This study examines the long-term physical health impacts of pregnancy outcomes, comparing women who experienced live births to those who had pregnancy losses (miscarriages or stillbirths). While previous research has documented short-term links between pregnancy outcomes and physical health, fewer studies have explored these associations over the long term, particularly considering women’s preconception health. Data were drawn from the Understanding Society Survey [UK Household Longitudinal Study (UKHLS)] from 2009 to 2023. The sample included 2386 women who reported their first pregnancy and were observed over multiple time points, both before and after pregnancy. Physical health was assessed using the Physical Component Summary (PCS) from the 12-item Short Form Survey (SF-12) questionnaire. Linear fixed-effects models were used to analyze changes in physical health relative to pregnancy outcomes, adjusting for socioeconomic, demographic, and mental health covariates. The analysis revealed a sharp decline in physical health (PCS score) at the end of pregnancy for both groups, with a more pronounced decline among women experiencing pregnancy losses. Post-pregnancy, these women continued to report lower PCS scores compared to those with live births, particularly in the 2 years following pregnancy. The confounders did not fully explain the observed differences. Pregnancy losses are associated with a significant and lasting decline in women’s physical health, even after controlling for socio-demographic factors. Women who experience losses often report worse health than those transitioning to motherhood, with symptoms potentially including fatigue, pain, and cardiovascular diseases. These findings underscore the importance of long-term health monitoring and support for women following miscarriage or stillbirth.

Introduction

Several physiological factors have been identified as predictors of women’s physical health during and after pregnancy [1]. These include maternal smoking [2], obesity [3], advanced maternal age [4], low socioeconomic status [4, 5], and pre-existing conditions like hypertension and diabetes [2, 6]. Higher parity or birth order of the child also plays a role [7]. Furthermore, studies have shown that pregnancy outcomes, such as miscarriage or stillbirth, can significantly impact women’s health [8], with those experiencing pregnancy loss facing increased risks of poorer health outcomes, such as cardiovascular disease and chronic pain, compared to mothers of live births [9–11]. This suggests that pregnancy loss may result from pre-existing poor health conditions.

However, few studies have examined whether the association between maternal health and birth outcomes persists in the long term, possibly due to certain limitations. Most studies collect data only during or immediately after pregnancy, lacking preconception health information, which makes it difficult to discern whether poor health preceded or resulted from the pregnancy loss [12–14]. Additionally, much research focuses on recurrent pregnancy losses, which may have already compromised health, either through the physiological or psychological effects of prior losses [6, 15–22]. Lastly, most analyses do not assess long-term outcomes, typically focusing only on the short term, and when they do, they often focus on a specific subgroup of women experiencing recurrent pregnancy losses [20, 22, 23].

The evidence on the association between pregnancy outcomes and women’s health is mixed [14]. Some studies indicate that pregnancy loss is linked to increased risks of cardiovascular disease [9, 10, 16, 23], metabolic disorders [3, 6, 22], and chronic pain [4, 13, 19, 24, 25]. Others suggest no significant long-term impact on physical health, with effects limited to psychological distress [11, 14, 15]. This inconsistency may be due to variations in study design, population characteristics, and the timing of health assessments relative to pregnancy outcomes. Given the trend of delayed childbearing, increasing the risk of involuntary pregnancy end due to advanced maternal age [26], it is crucial to revisit this question using high-quality data and robust methods.

This study uses data from the Understanding Society Survey [UK Household Longitudinal Study (UKHLS)] to address key limitations in the existing literature. First, it documents the association between pregnancy outcomes (“live birth” or “pregnancy loss,” encompassing miscarriage and stillbirth) and women’s health before, during, and after the first pregnancy, shedding light on whether women experiencing pregnancy loss had poorer health prior to conception. Second, the study examines how pregnancy outcomes influence women’s physical health post-pregnancy, assessing whether those experiencing loss fare worse than their counterparts who become mothers.

By leveraging the longitudinal nature of the UKHLS dataset, which tracks health data years before and after a woman’s first pregnancy, the analysis provides a nuanced view of the long-term health effects of pregnancy outcomes. This methodology, which incorporates repeated observations, offers a novel contribution by exploring how the unadjusted relationships between pregnancy outcomes and health are influenced by adjustments for factors such as age, socioeconomic status, partnership status, mental health, and transitions to (another) childbirth.

Methods

Study sample

This study uses data from the Understanding Society Survey (UKHLS), a nationally representative longitudinal panel survey that provides detailed information on individuals and households in the UK. The dataset includes individual-level data on women, their partners—if co-resident—and other household members, offering insights into family dynamics.

The analysis focuses on pregnancies reported between 2009 and 2023 (Waves 1–13). The initial sample comprises 30 955 women (165 978 observations). After restricting the age range to 18–45 (reproductive age), the sample size reduces to 26 100 women. From this, 2414 women were selected as they were childless before the survey and reported their first pregnancy during the study period. After the exclusion of observations with missing values in the pregnancy outcome and physical health measures, the final analytical sample consists of 2386 women (257 reporting pregnancy loss), contributing a total of 19 153 observations over multiple years before and after their first reported pregnancy. Attrition rates were observed at 4.7%, with the highest attrition of 24.1% occurring between Waves 1 and 2, and the lowest at 1.2% between Waves 12 and 13.

The design specifically targets first reported pregnancies to avoid biases from prior childbirth experiences, ensuring a focus on the short- and long-term health effects of pregnancy outcomes. Induced abortions are excluded due to severe underrepresentation in the sample (2.3% of reported pregnancies, compared to an expected 18%–25% in high-income countries [27]), which may compromise the data reliability and the validity of analyses including these cases. As this study involves secondary data analysis of anonymized data, a formal ethical review was not required, adhering to confidentiality and ethical research standards.

Exposure and outcome

The primary explanatory variable is the outcome of the first pregnancy, categorized as either a live birth or a pregnancy loss (miscarriage or stillbirth). Voluntary terminations (abortions) are excluded. The study tracks women’s physical health across various time periods: 25+ months before pregnancy, 24–13 months before, 12–1 months before, 0–12 months after, 13–24 months after, 25–36 months after, 37–48 months after, 49–60 months after, and 61+ months after pregnancy. The reference period, 25+ months before, is selected to minimize the impact of fertility treatments or other interventions [28], often absent during early family planning phases [29].

The primary outcome measure is the Physical Component Summary (PCS) of the 12-item Short Form Survey (SF-12) questionnaire, assessing physical functioning, bodily pain, general health, and role limitations due to physical health. The SF-12, derived from the SF-36 Health Survey, is a validated and reliable tool widely used in large-scale studies [30–32]. In the analytical models, the PCS is standardized, with a mean of 0 and standard deviation of 1, enabling comparisons across populations and time periods.

In the robustness tests, cardiovascular conditions serve as a secondary dependent variable due to their frequent use in prior research on pregnancy outcomes and their comprehensive availability in the dataset.

Covariates

All models adjust for the year of conception to account for temporal trends in pregnancy outcomes and women’s health. Key covariates include demographic characteristics like women’s time-varying age (linear and quadratic terms) and partnership status (married, cohabiting, or single). Socioeconomic factors are captured through education (categorical) and family job class, using the UK’s National Statistics Socio-economic Classification (NSSEC-5), a five-category system that groups occupations based on employment relations and conditions. Family income is adjusted for inflation and expressed logarithmically to account for changes over time.

Mental health is controlled using the Mental Component Summary (MCS) from the SF-12, which assesses emotional well-being, vitality, and social functioning. MCS and PCS are correlated but capture distinct dimensions of health [33]. A time-varying control for the transition to (another) childbirth is included, representing the transition to a second child for the “live birth” group and the first for the “pregnancy loss” group. This variable is included because physical health problems can arise or worsen after becoming a mother, whether building on issues already faced by women who have become mothers or emerging anew for those experiencing motherhood for the first time. Full details of the analytical variables are provided in Table A.1 in the supplementary material.

Statistical analysis

The analysis estimates five linear probability models with individual-specific fixed effects (FE), which provide marginal effects directly interpretable as percentage differences in health outcomes relative to the average level in the sample. This approach also ensures comparability across models. The analysis begins by examining the association between pregnancy outcomes (“live birth” and “pregnancy loss”) and women’s health before and after pregnancy through event-study regressions, which compare outcomes at different time points relative to pregnancy, initially adjusting only for the year of conception (the “unadjusted” model). The use of FE models helps control for time-invariant individual characteristics, such as genetic predispositions or chronic health conditions, thus reducing bias and ensuring more accurate estimates by focusing on within-individual changes over time.

Subsequent models introduce covariates incrementally. Model 2 adjusts for maternal demographic characteristics (age and partnership status). Model 3 includes family socioeconomic characteristics, such as household income and education. Model 4 adds the MCS from the SF-12 to account for mental health, while Model 5 controls for the transition to another childbirth (for the “live birth” group) or the first childbirth (for the “pregnancy loss” group), allowing assessment of the full attenuation effect when considering all observed characteristics.

This approach tests whether health trajectories for women who experience live births and pregnancy losses differ before conception, offering insights into whether poorer preconception health could influence pregnancy outcomes. It also tracks post-pregnancy health to determine if and how health recovers to baseline levels, highlighting potential long-term impacts of pregnancy loss.

Results

Table 1 and Fig. 1 present descriptive characteristics of the study sample, comparing key variables across time periods before and after pregnancy for live birth and pregnancy loss outcomes. Prior to pregnancy end, physical health is similar for both groups, with no significant differences. Figure 1 presents PCS scores across four time intervals: 24–13 months before pregnancy, 0–12 months after, 13/24 months after, 25/36 months after. During the “1–12 months before” period, PCS scores are comparable for the two groups (∼53). In the “0–12 months after” period, the “pregnancy loss” group reports a lower PCS score than the “live birth” group (52.32 vs. 53.89). In the “13–24 months after” period, women who experienced pregnancy loss have a PCS score of 52.99, 0.90 points lower than mothers, suggesting potential disparities based on pregnancy outcome.

Figure A.4 in the supplementary material displays PCS across four time windows before the pregnancy ends. Although similar in earlier periods, in the “24–13 months before” period, the “live birth” group reports significantly lower PCS scores (51.84) than the “pregnancy loss” group (54.58), likely reflecting symptoms such as nausea or fatigue.

Table 2 presents the coefficients from the event-study analysis of the PCS scores (while all the covariates’ coefficients are exhibited in Table A.2 in the supplementary material). Model 1 investigates the physical health trajectories of women before and after their first pregnancy, comparing those with live births to those with involuntary pregnancy terminations. The results indicate that both groups exhibit a relatively stable profile of physical health during the “13–24 months before pregnancy end” period compared to the baseline measured in the “25+ months before pregnancy end” period. However, in the “12–1 months before” period, a significant decline in PCS is observed for both groups, with a sharper decline among women who experience miscarriage or stillbirth compared to their baseline. Specifically, the PCS is 0.13 standard deviations below the baseline for the “live birth” group in the “12–23 months after pregnancy end,” while it is 0.31 standard deviations below the baseline for the “pregnancy loss” group. The difference in PCS declines between the two groups is statistically significant, with the latter group experiencing a greater decline. The point estimate for the “pregnancy loss” group remains lower than that of the “live birth” group until the “61+ months after pregnancy end” period, although this difference is no longer statistically significant.

In Model 2, adjusting for partnership status does not significantly reduce the post-pregnancy decline in PCS. However, when parental socioeconomic characteristics are adjusted for in Model 3, the decline is attenuated to a larger extent compared to Model 2, reducing by about 20%, particularly in the period 0–13 months after pregnancy end for both women who report live births and those who do not. Model 4 reveals that the PCS coefficients in the later post-pregnancy periods are similar to those observed in Model 3, with one notable exception: for women with a pregnancy loss, the decline in PCS is sharper in the 0–12 months post-pregnancy period with a 0.24 standard deviation decrease, compared to a 0.15 standard deviation decrease in Model 2. The point estimates of the post-pregnancy levels of PCS for women experiencing a pregnancy loss consistently remain below those of women transitioning to motherhood across all models. This gap is statistically significant only in the “13–24 months after” period across all tested models and is more pronounced in the fully adjusted Model 5 (and Model 4) compared to Model 1. Figure 2 illustrates the results for Models 1, 4, and 5, clearly showing these differences in PCS trajectories, with a sharper and more persistent decline in physical health among women who experience a pregnancy loss.

Sensitivity analyses

Sensitivity analyses replicated Models 1, 3, and 5 using cardiovascular diseases as the dependent variable (Fig. A.1 in the supplementary material). Post-pregnancy cardiovascular disease risk rises for both mothers and those with pregnancy loss, indicating cardiovascular disease may partially explain the decline in physical health.

PCS models for male partners (Fig. A.2 in the supplementary material) show stable short-term physical health for both fathers and men whose partners experienced pregnancy loss, with a gradual decline over time, which vanished when accounting for the whole set of controls. This suggests that the sharper decline in women’s physical health may stem from pregnancy-specific symptoms, rather than broader family-related factors.

Mental health trajectories (MCS, Fig. A.3 in the supplementary material) reveal improved mental health during pregnancy for mothers, while those with pregnancy loss face a temporary decline, followed by recovery. This contrasts with the persistent decline in physical health.

Discussion

Using data from a nationally representative British survey, this study analyzed physical health trajectories for two groups of women: those who experienced pregnancy loss and those with a live birth during their first pregnancy. By adjusting for observed characteristics, this analysis traced health conditions before conception, not just during or shortly after pregnancy, as previous studies have done. This approach provides deeper insights into health disparities between women with different pregnancy outcomes and offers new evidence on how pregnancy loss may independently affect physical health in both the short and long term.

The unadjusted results show that both groups—those with live births and those without—reported similar PCS levels up to 12 months before pregnancy. One year before delivery, women who became mothers experienced worsening physical health. After pregnancy end (“0–12 months after”), both groups showed a decline in health, with limited recovery. Notably, women who experienced pregnancy loss consistently had lower PCS levels compared to those who transitioned to motherhood, and this difference persisted in later post-pregnancy periods. The gap in PCS between the groups was statistically significant in the “13–24 months after pregnancy end” period but became insignificant in later periods, likely due to sample size limitations affecting long-term statistical power.

The decline in physical health outcomes can largely be attributed to socioeconomic factors like household income and education, and to a lesser extent, demographic factors such as maternal age. In Model 4, the inclusion of a mental health indicator (MCS), which is correlated with but distinct from PCS, helped control for the interplay between physical and mental health during pregnancy. Even after adjusting for mental health, the PCS decline post-conception remained statistically significant for both groups.

These findings support the hypothesis that involuntary pregnancy termination is associated with worse physical health, beyond the well-documented mental health effects. Pregnancy loss, whether due to miscarriage or stillbirth, may be an independent risk factor for women’s physical health. Descriptive and multivariate analyses show that women who experience pregnancy losses do not consistently fare worse than those with live births prior to conception, suggesting pregnancy loss itself drives the health disparities. Medical complications related to miscarriage or stillbirth—such as infection, heavy bleeding, or retained tissue—may lead to longer-term health issues, including chronic pain and an increased risk of cardiovascular disease [6, 8].

This study makes two key contributions to the literature on pregnancy outcomes and women’s physical health. First, it helps reconcile conflicting findings from previous research. Prior studies have been inconclusive, with some showing associations between pregnancy loss and poorer health, while others do not. This inconsistency may arise from not accounting for preconception health, leading to potential reverse causality—where poorer health before pregnancy contributes to loss and worsens post-pregnancy outcomes. Additionally, differing control variables and smaller sample sizes in earlier research may have limited their ability to detect significant effects.

Second, while most existing research has focused on how health conditions influence the risk of pregnancy loss, this study highlights the potential long-term health consequences of involuntary pregnancy termination. The findings suggest that miscarriage or stillbirth may lead to more severe and prolonged physical health issues than those experienced by women who transition to motherhood, underscoring the lasting impact of pregnancy loss on women’s health.

As documented in previous research [26], the physical toll of repeated pregnancy loss may exacerbate existing health vulnerabilities, highlighting the importance of comprehensive medical follow-up and support for women experiencing these adverse outcomes. This study underscores the need for further research to better understand the full scope of health impacts associated with miscarriage and stillbirth.

This study has two main strengths. First, the use of longitudinal survey data allowed for the analysis of validated proxies of women’s health over an extended period, including preconception health, a novel contribution to this literature. This approach enabled the examination of health trajectories not only during and after pregnancy but also in the years leading up to conception. Second, the data permitted adjustments for a broad set of individuals’ characteristics, controlling for key socio-demographic factors that could influence the relationship between pregnancy outcomes and health.

The study has some limitations. First, excluding women who had previously carried pregnancies to term helped avoid the potential bias from the health consequences of prior pregnancies, but it limits the generalizability of the findings. Second, while the design adjusts for observed factors, it is not causal. Unobserved pre-existing conditions that increase the risk of pregnancy loss and impact women’s health may still be influencing the results. Although involuntary pregnancy loss is largely random, its association with pre-existing conditions has been documented, indicating that residual confounding may still be present. Third, miscarriage reporting in survey data, including UKHLS, faces challenges of underreporting and misclassification. Up to 40% of miscarriages may go unreported [34, 35], influenced by stigma, psychological distress, and ambiguous terminology [36, 37]. Although FE models mitigate some biases, underreporting may still affect the findings, especially for disadvantaged groups [38]. Fourth, the analysis focuses on first reported pregnancies in UKHLS, as retrospective pregnancy histories for pregnancies not ending in live births are unavailable. Women who experienced a prior pregnancy loss or abortion before joining the survey may be included without these events being accounted for. This limitation could bias estimates downward if prior losses reduced health, or upward if such women are underrepresented due to survey dropout. Fifth, the findings are specific to the British context and may not be generalizable to other settings with different healthcare systems or cultural norms around motherhood. Additionally, body mass index (BMI) is an important confounder but was not considered as it is measured in only one wave. Finally, the small sample size prevented a clear distinction between miscarriage and stillbirth, though the latter is likely associated with more severe health impacts due to its physical and psychological toll [4, 10, 16].

In conclusion, this study is the first to analyze the association between pregnancy outcomes and women’s health across the entire pregnancy period, incorporating both unadjusted and adjusted analyses. The findings highlight pregnancy loss as a potential risk factor for women, both in the short and long term. This highlights the need for further research to explore the mechanisms linking pregnancy loss to subsequent health outcomes and to better understand its role as an independent risk factor for women’s health. Expanding research in this area could inform targeted interventions aimed at mitigating the health impacts of pregnancy loss and supporting women’s health across the reproductive lifespan.

Acknowledgements

The author thanks the European Research Council and the Understanding Society team for the dataset and support.

Author contributions

The author designed the study, conducted the analyses, interpreted the findings, drafted the paper, revised it, and wrote the final manuscript.

Supplementary data

Supplementary data are available at EURPUB online.

Conflict of interest: None declared.

Funding

This study is co-funded by the European Union (European Research Council, BIOSFER, 101071773) and by the Understanding Society (EBAL900—Understanding Society Fellowship). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them.

Data availability

The data used in this study are available from the UK Data Service, under the Understanding Society dataset, subject to licensing and data-sharing agreements.

Ethical approval and consent to participate

As this study involves secondary data analysis of anonymized data, a formal ethical review was not required. The research adheres to confidentiality and ethical research standards.

Consent for publication

Not applicable. The study uses secondary data from an anonymized dataset, and no identifiable personal data are included.

References