Human resources and health outcomes: cross-country econometric study
Sudhir Anand, Till Bärnighausen
The Lancet · 2004 · DOI: 10.1016/S0140-6736(04)17313-3
Countries: Cross-national (117-country sample; 83-country subsample with poverty data), predominantly low- and lower-middle-income countries
What Was Studied
Anand and Bärnighausen used the 2004 WHO dataset on health personnel (doctors, nurses, and midwives), combined with World Bank and UNDP data on income, poverty, and female literacy, to test — in a cross-sectional multiple-regression framework — whether a country's density of human resources for health predicts its maternal mortality rate, infant mortality rate, and under-five mortality rate. All variables were log-transformed so regression coefficients are interpretable as elasticities. There were 118 countries with complete data across all variables except income poverty; after excluding one outlier (Niger), the core sample comprised 117 countries (85 low- and lower-middle-income, 10 high-income). A second set of regressions added income poverty (share of the population below PPP$1-a-day) as a control, reducing the sample to 83 countries because poverty data were unavailable for many wealthy countries. The analysis was run twice — once with aggregate human-resources-for-health density, and again with doctor density and combined nurse-and-midwife density entered separately.
What They Found
Across 12 regression equations, aggregate density of human resources for health (doctors + nurses + midwives) significantly predicted all three mortality outcomes, with elasticities ranging from -0.474 (maternal mortality, the largest effect) to -0.212 (infant mortality, with poverty controls), all p≤0.0036 — meaning a 1% increase in health-worker density was associated with up to a 0.474% fall in maternal mortality, holding income, poverty, and female literacy constant. When disaggregated, doctor density alone remained significant for all three outcomes (elasticities -0.174 to -0.386, all p≤0.0029), contradicting three earlier studies that had reported no doctor effect (a 'doctor anomaly'). Nurse density was significant in only one of six regressions (maternal mortality without poverty controls, p=0.0443); the authors attribute this to measurement error in nurse counts biasing coefficients toward zero, not to nurses being unimportant. The models explained 79-84% of cross-country variance in mortality (R² 0.787-0.835) across samples of 83-117 countries.
What This Means for Nepal
This study's core finding — that human resources for health density predicts maternal and child mortality independently of income, poverty, and female literacy — reinforces the evidence base behind Nepal's own workforce-retention reform priority, since Nepal's persistently high maternal mortality ratio (151 per 100,000 live births, NDHS 2022) coexists with severe workforce shortages and maldistribution: only 37.9% of sanctioned physician/GP posts are filled nationally (down from 56.5% in 2015), Karnali province has 0% of consultant and physician posts filled, and Kathmandu Valley has roughly 180 times the doctor density of remote districts (1:850 versus 1:150,000). With roughly 2,000-2,400 doctors a year leaving via Nepal Medical Council 'Good Standing Certificates,' this cross-country evidence argues for the Ministry of Health and Food Hygiene (MoHP) to prioritise filling vacant rural physician posts — starting in Karnali, where fill rates are 0% — through binding rural-service requirements tied to KAHS/PAHS scholarships, rather than relying on FCHV-level community health workers alone to close the maternal-mortality gap.
Contextualisation
Nepal's own health workforce data illustrate the mortality link this cross-country study documents: only 37.9% of sanctioned physician and GP posts are filled nationally (down from 56.5% in 2015), Karnali — Nepal's poorest province — has 0% of its consultant and physician posts filled, and Kathmandu Valley has roughly 180 times the doctor density of remote districts (1 per 850 people versus 1 per 150,000). With Nepal's maternal mortality ratio still at 151 per 100,000 live births (NDHS 2022) and roughly 2,000-2,400 doctors a year obtaining Nepal Medical Council 'Good Standing Certificates' to emigrate, this study's finding that human-resources-for-health density predicts maternal and child mortality independently of income, poverty, and female literacy reinforces the evidence case behind MoHP's current health-workforce-retention reform priority.