Changes in preterm birth and stillbirth during COVID-19 lockdowns in 26 countries

Approximately 10% of babies are born preterm (<37 weeks), equating to nearly 15 million preterm births annually and making PTB the leading cause of infant mortality. While many PTBs are spontaneous, some are medically indicated to prevent adverse outcomes such as stillbirth. Early reports during the first wave of COVID-19 lockdowns showed widely varying changes in PTB (from substantial reductions to increases) and concurrent reports of increased stillbirths in some settings, but most studies used facility-based data and did not account for pre-pandemic trends, hampering interpretation and comparability. Given potential mechanisms by which lockdowns could reduce PTB (e.g., decreases in infections and air pollution) or alternatively signal gaps in maternity care (potentially increasing stillbirths), the International Perinatal Outcomes in the Pandemic (iPOP) study aimed to rigorously evaluate whether initial COVID-19 lockdowns were associated with changes in PTB and stillbirth rates using population-based, harmonized data and interrupted time series analyses across multiple countries, considering country income level and PTB subtypes, and accounting for pre-lockdown trends.

Prior single- and multi-center studies during early COVID-19 lockdowns reported heterogeneous PTB changes: notable reductions in some high-income countries (Australia, Israel, parts of Europe), increases in Nepal, Uruguay and California, and little or no change in national data from Canada, Spain, Sweden and the USA. Some studies from LMICs (Nepal, Nigeria) and HICs (UK, Italy) reported increased stillbirths, but few examined PTB and stillbirth concurrently. Meta-analyses generally suggested no overall global change, with subgroup signals: decreased PTB and increased stillbirth in HICs. However, comparisons were limited by methodological differences: many studies were facility-based, few accounted for temporal trends, and population coverage varied, with a living review showing differences between single-centre versus regional/national studies. This literature underscores the need for standardized, population-level analyses that consider pre-pandemic trends and evaluate PTB alongside stillbirth.

Design and data sources: Aggregated, monthly, population-based data (capturing >90% of births) and non-population-based datasets were harmonized from 26 countries, covering 52,067,596 births (January 2015–July 2020). Eighteen population-based datasets contributed 98.6% of births; 26 non-population-based datasets (facilities, pooled facilities, or a demographic surveillance site) contributed 1.4%. Data were standardized via a common protocol and analyzed within the SAIL Databank. Country income level (World Bank) was recorded. Lockdown definition: Initial lockdown onset was defined as the first date a country’s Oxford COVID-19 Government Response Tracker Stringency Index exceeded 50; if after the 15th of the month, lockdown was assigned to the following month. Outcomes were assessed for the first four months following initial lockdown. Outcomes: Monthly rates were computed as follows—preterm birth (<37 weeks/100 births), very preterm (<32 weeks/100 births), spontaneous preterm (<37 weeks with spontaneous onset/100 births), and stillbirth (≥22 weeks per 1,000 births). Where only live births were available (Chile, Peru, USA) or different lower gestational limits applied (NSW Australia and Wales ≥24 weeks), analyses reflected those constraints. Data quality: Completeness, proportion missing gestational age, outliers, and deviations of observed versus expected births (forecasted via Poisson time series) during lockdown were assessed. Population-based datasets with ≥10% deviation in total births during lockdown were excluded from the primary population-based analysis and treated as non-population-based. Statistical analysis: For each population-based dataset and outcome, a weighted interrupted time series (ITS) model was fit to pre- and post-lockdown monthly log-odds, weighting by monthly total births (imputed months down-weighted). Models included month (seasonality) and long-term trend; five functional forms (linear, square, quadratic, logarithmic, second-order polynomial) were compared via AIC and the best selected. A ‘pre-lockdown model’ refit to pre-lockdown data forecast expected odds for each of the first four lockdown months; the impact of lockdown was summarized as the observed-to-forecasted odds ratio (OR) per month. For non-population-based datasets, linear regression on log-odds including month (categorical), year (continuous), and year-squared (non-linear trends) generated the counterfactual forecasts and corresponding ORs for months 1–4. Meta-analysis: Monthly ORs from each dataset were pooled via random-effects meta-analysis, separately for population-based and non-population-based data. Population-based analyses were stratified by country income level (high vs upper-middle, where possible). For non-population-based data, a three-level meta-analysis accounted for within-country facility dependence. Heterogeneity was assessed using I². Sensitivity analyses: (1) Restricting to live births among datasets that also provided all-births data; (2) restricting to births ≥28 weeks (WHO international comparison threshold); (3) excluding Brazil and the USA (together >70% of births) from population-based meta-analyses. Analyses used R v4.1.1.

• Study population: 52,067,596 births (Jan 2015–Jul 2020); 3,115,628 during the 4-month lockdown windows. Population-based datasets had high data quality (typically <1% missing gestational age; <5% deviation from expected births during lockdown).
• Baseline rates (population-based): PTB 5.8% (Finland) to 11.8% (Brazil); very PTB 0.8% (Finland/Peru) to 2.0% (Brazil); spontaneous PTB 2.8% (NSW, Australia) to 9.2% (Brazil); stillbirth 2.5/1,000 (Finland) to 10.4/1,000 (Brazil).
• Preterm birth (population-based meta-analysis): Small reductions in PTB in months 1–3 of lockdown, not month 4: • Month 1: OR 0.96 (95% CI 0.95–0.98), P<0.001; I²=0%. • Month 2: OR 0.96 (0.92–0.99), P=0.03; I²=64%. • Month 3: OR 0.97 (0.94–1.00), P=0.09; I²=53%. • Month 4: OR 0.99 (0.96–1.01), P=0.34; I²=34%. Reductions were similar in HICs and upper-middle-income countries, with greater heterogeneity among upper-middle-income countries.
• Preterm birth (non-population-based): No clear association overall (e.g., Month 1 pooled OR 1.01, 95% CI 0.92–1.11); wide variability between facilities.
• Very preterm birth: No evidence of impact over months 1–4 in population-based data (pooled ORs ~1.00–1.02 with CIs spanning 1.00); similar null findings in non-population-based data.
• Spontaneous preterm birth: In HICs with data, small relative decreases (~3–4%) in months 1–3 post-lockdown (e.g., pooled HIC Month 1 OR 0.96, 95% CI 0.93–0.99); not observed in Brazil. In non-population-based data, Month 4 showed a decrease (OR 0.88, 95% CI 0.78–0.99, P=0.04).
• Stillbirth (population-based meta-analysis): • Month 1: Overall no clear increase (OR 1.04, 0.99–1.09, P=0.10; I²=0%), but HICs showed an increase (OR 1.14, 1.02–1.29, P=0.02), driven by Canada (OR 1.26, 1.04–1.51). • Months 2–3: Increased odds overall (Month 2 OR 1.07, 1.02–1.12, P=0.001; Month 3 OR 1.08, 1.02–1.13, P=0.004), largely driven by Brazil; restricting to HICs, no association (Month 2 OR 1.00, 0.88–1.12; Month 3 OR 0.99, 0.88–1.12). • Month 4: Possible increase overall (OR 1.07, 1.00–1.15, P=0.07); no association in HICs (OR 1.01, 0.87–1.18). In Brazil, increased stillbirth observed in months 2–4 (ORs 1.09, 1.10, 1.12).
• Sensitivity analyses: Restricting to live births, ≥28 weeks, or excluding Brazil and the USA yielded negligible changes in estimates.
• Interpretation: Modest, population-level PTB reductions (3–4%) during the first 3 lockdown months could translate to large absolute numbers globally (estimated ~50,000 PTBs averted in the first month assuming a 10.6% global pre-pandemic PTB rate). In HICs, decreased PTB did not appear to be offset by increased stillbirths; increases in stillbirths were most consistent in Brazil.

Applying standardized interrupted time series methods to large-scale, population-based data across 26 countries, the study demonstrates small but consistent reductions in overall preterm birth during the first three months of initial COVID-19 lockdowns, with no concurrent increase in very preterm birth and no consistent rise in stillbirths across HICs. These findings address methodological limitations of prior studies by accounting for pre-lockdown trends, using population-representative data, and evaluating PTB alongside stillbirth. Potential mechanisms for reduced spontaneous PTB include marked declines in circulating non-COVID infections due to reduced social contact and improved hygiene, and reductions in air pollution, both known PTB risk factors. A reduction in indicated (medically initiated) PTB might also have contributed in some settings, though this could not be directly examined. Increased stillbirths observed in Brazil and signals in some HICs early in lockdown may reflect disruptions in access to timely, quality antenatal and intrapartum care. Averaged at the population level, opposing effects in subgroups (e.g., reduced PTB risk via environmental changes versus increased risk via care disruptions) could balance out, yielding small net changes. Because only a small fraction of pregnancies were infected with SARS-CoV-2 during this early period, infection likely had limited influence on population-level estimates compared to the broader impact of restrictions. The study underscores the importance of robust, population-level surveillance and standardized definitions, as non-population-based data showed substantial variability and potential biases due to changes in care-seeking and facility case mix.

Across 18 population-based datasets from HICs and upper-middle-income countries, initial COVID-19 lockdowns were associated with modest (3–4%) reductions in overall preterm birth during the first three months, without evidence of compensatory increases in stillbirth in most HICs. Consistent increases in stillbirth were observed in Brazil, warranting further investigation. Although small in relative terms, these changes imply meaningful population impacts (e.g., an estimated ~50,000 PTBs averted in the first month globally). Elucidating causal pathways—such as reduced infections, improved air quality, and changes in obstetric practices—could inform future clinical and public health strategies to prevent PTB without compromising perinatal safety. The work also highlights the urgent need to enhance standardized, high-quality perinatal data systems, particularly in LMICs, and demonstrates the value of the iPOP platform for rapid, harmonized perinatal research during public health crises and beyond.

• Lockdown exposure was approximated using the Oxford Stringency Index threshold >50 at the national level, which may not capture subnational variability, enforcement, or individual experiences, especially in large countries (e.g., Brazil, USA).
• ITS analyses can be confounded by concurrent unmeasured events affecting outcomes.
• Aggregate data precluded subgroup analyses (e.g., by socioeconomic status, region, age) and disentangling spontaneous versus indicated PTB in all settings.
• Focus on the first four months post-lockdown primarily reflects pregnancies in the third trimester at lockdown onset; impacts on earlier gestations were not assessed.
• For rare outcomes (very preterm birth, stillbirth), estimates were imprecise with wide confidence intervals; absence of evidence is not evidence of no effect.
• Some countries had challenges in data capture, coverage, and quality (notably in non-population-based datasets), limiting generalizability; thus, primary inferences emphasize high-quality population-based data.