Stillbirth (fetal death at ≥28 weeks) and early neonatal death (death within seven days of life) together define perinatal mortality, a major contributor to global child mortality concentrated in low- and middle-income countries (LMICs). Household surveys such as the Demographic and Health Surveys (DHS) are primary data sources in many LMICs, but survey-derived estimates of stillbirth and early neonatal mortality are vulnerable to underreporting, misclassification (especially between stillbirths and deaths on days 0–1), recall bias, and age-at-death heaping. These data quality issues limit the usefulness of survey data for monitoring trends and targeting interventions.
The present study applies an adjustment framework previously proposed to improve the use of survey data by correcting common reporting artifacts. The authors aim to estimate trends in stillbirth, early neonatal, and perinatal mortality and to assess sociodemographic inequalities in perinatal mortality across 17 LMICs with at least four standard DHS surveys that include reproductive calendar data.
The analysis included 93 DHS surveys from 17 countries conducted between 1990 and 2023. The pooled dataset covered 1,019,253 reported pregnancies from 733,181 women. Analyses included pregnancies reaching at least seven months’ gestation to align with the standard definition of stillbirth used in the study.
To address survey data quality limitations, the study applied a modified Gompertz–Makeham (mGM) model. The mGM model was designed to adjust observed reporting patterns for likely omissions and misclassifications. Specific adjustments targeted:
The approach reallocates reported events to their most plausible timing and outcome based on modelled patterns and predefined quality indicators. The model was applied survey-by-survey, producing adjusted age-specific risks and aggregated perinatal measures for trend and differential analyses.
Age-specific risks of stillbirth and early neonatal death were estimated using life-table methods combined with the mGM adjustment. Survey-specific estimates produced by the mGM were pooled and examined over time. The study evaluated associations between perinatal mortality and maternal and household characteristics using the mGM outputs and pooled random-effects meta-regression. Covariates were included both to adjust perinatal mortality estimates and to account for survey sampling design. A sensitivity analysis using alternate model assumptions was conducted to probe robustness at the country level.
Adjusted estimates indicated that, overall, stillbirth, early neonatal, and perinatal mortality rates generally declined across the study countries, although magnitude and timing of declines varied substantially between settings. The pooled annual declines reported were:
The authors note heterogeneity in country trajectories: while most countries experienced reductions, the pace and consistency of improvement were uneven.
The analysis identified persistent inequalities in perinatal mortality across socioeconomic and demographic groups. Lower perinatal mortality was associated with:
Conversely, maternal age of 30 years or older was associated with higher perinatal mortality risk in the pooled analyses. These associations reflect within- and between-country disparities that persisted despite overall declines in rates.
The authors ran sensitivity analyses using alternate assumptions to test the robustness of their findings at the country level. The reported sensitivity checks supported the general direction and magnitude of trends in adjusted estimates. The paper also references prior validation work showing that the mGM adjustment increases estimated stillbirth and perinatal mortality compared with unadjusted survey figures by reallocating misclassified or omitted events.
Key limitations reported by the authors include:
The authors acknowledge that these factors could affect the absolute level of adjusted estimates and interpretation of differences between countries.
Applying statistical adjustments for survey data quality issues can improve estimates of stillbirth and early neonatal mortality derived from household surveys. Although most countries in the sample showed declines in perinatal mortality, progress was uneven and substantial socioeconomic disparities persisted, disproportionately affecting rural, less educated, and poorer populations. Strengthening maternal and newborn health services, addressing inequities, and improving civil registration and vital statistics and survey data quality are essential steps to accelerate progress toward global targets for reducing perinatal mortality.
The authors make data and code available through public repositories referenced in the original study, and they note that their approach may help maximize the value of household survey data for monitoring and planning in settings where routine death registration is incomplete.