The study team estimated policy impacts on annual child mortality and maternal mortality using a staggered 2x2 Difference-in-Differences framework over 1990–2023. The authors reported that the Ghana Essential Health Interventions Program (GEHIP, 2012) produced a reduction in child mortality of approximately 9.5 percent relative to the global trend. The National Newborn Health Strategy (NNHS, 2019) was associated with a larger point estimate for child survival — a 16.4 percent reduction relative to the global comparator — but the investigators flagged mild pre-existing drift in the NNHS event-study window and therefore considered that estimate to be an upper bound.
On maternal outcomes, the GEHIP window again delivered the most credible effect: authors reported a 7.2 percent reduction in maternal mortality below the global comparator, and they noted event-study diagnostics showed a near-flat pre-period average treatment effect on the treated (ATT) series in that window. By contrast, the Free Maternal Care policy (2008) produced no detectable improvement in maternal mortality relative to either the global average or the lower-middle-income comparator. The Community-Based Health Planning and Services initiative (CHPS, 1999) generated a large apparent maternal mortality effect in naive comparisons, but investigators invalidated this effect after observing a strong pre-existing downward trend in the event-study window.
The authors summarized these patterns as evidence that interventions focused on supply and quality of care produced the most credible mortality improvements in Ghana, while financial-access expansions showed stronger detectable effects on child survival than on maternal survival.
Investigators conducted event-study plots to assess the parallel trends assumption for each policy window. Where the pre-period ATT series was near-flat, the authors treated post-treatment estimates as more credible. GEHIP’s maternal mortality result benefited from such an event-study pattern. Where pre-period drift or strong pre-existing trends were present, the team downgraded causal interpretation. NNHS’s child mortality estimate was explicitly described as an upper bound because mild pre-period drift was detectable. CHPS’s large maternal effect was rejected as causal because the event-study window contained a marked downward mortality trend preceding the intervention.
The paper therefore relied not only on point estimates from the staggered 2x2 Difference-in-Differences model but also on graphical and diagnostic evidence from event-study checks to determine which policy windows produced credible causal inferences.
Data indicated differential performance by intervention type. Supply-side and quality-focused initiatives — the authors point to GEHIP and the National Newborn Health Strategy as examples — were associated with the most credible reductions in both child and, to a lesser extent, maternal mortality. Financial-access reforms, such as the Free Maternal Care policy and broader insurance reforms, did not consistently translate into maternal mortality declines in these analyses. The authors interpreted the null maternal finding for Free Maternal Care as consistent with the view that increased facility use requires concurrent facility capacity and quality improvements before mortality declines will appear.
The team observed stronger signals for child survival following access-expanding policies than for maternal survival, and they stressed that effect credibility depended on the absence of pre-treatment trends in the event-study windows. These distinctions influenced which estimates the authors presented as robust versus suggestive or upper-bound effects.
The study used only aggregated, publicly available mortality indicators from the World Bank World Development Indicators database. Specifically, investigators used the under-5 mortality rate (per 1,000 live births) and the modeled maternal mortality ratio (per 100,000 live births) drawn from the World Bank series referenced in the manuscript. The analytic period covered 1990 through 2023. The Difference-in-Differences approach was a staggered 2x2 design; the study team compared Ghana's trajectories against two comparator groups: the global average and the lower-middle-income country group. Event-study plots were used to evaluate parallel trends for each policy window.
The authors reported that no individual-level or identifiable data were used, and therefore IRB approval was not required. They also declared no competing interests. The manuscript is a preprint posted to medRxiv and has not been peer reviewed.
The authors acknowledged several limitations in their approach and interpretation. Their causal claims depend on the parallel trends assumption; where event-study checks show pre-period drift, the authors treated estimates as upper bounds or invalidated them. The analysis used aggregated World Bank indicators; the authors did not report subnational or facility-level data, nor did they report clinical process or capacity metrics that might explain why access reforms failed to reduce maternal mortality. The preprint noted that prior evaluations relied on descriptive trends and cross-sectional comparisons; this study attempted a stronger counterfactual but remains constrained by the available aggregated series.
The manuscript did not report specific implementation fidelity, dosing, or mechanism-level data for each policy window. The authors did not provide randomized or quasi-random assignment at subnational levels in the text; if such details exist they were not reported in the preprint. Open questions thus include how facility capacity, quality of obstetric care, and subnational heterogeneity mediated the observed effects, and whether longer follow-up after some policies (for example NNHS) would alter the credibility of estimated impacts once pre-treatment drift concerns are resolvable.
Readers should also note that the article is a preprint and has not been peer reviewed; the authors caution against using these findings to guide clinical practice without further validation and independent review.