The Gift of a Lifetime: The Hospital, Modern Medicine, and Mortality

The paper investigates whether the rise of modern hospitals and medical technology contributed to the early- to mid-twentieth-century decline in mortality, challenging the view that modern medicine (aside from sulfa drugs) had little impact before the 1950s. The authors pose several questions: Did improved access to modern hospitals reduce mortality? Were impacts heterogeneous by race in the Jim Crow South? Through what mechanisms—such as attracting higher-quality physicians or enabling better use of new medical technologies—did hospitals affect health? And did benefits extend beyond infancy into later life? The study focuses on The Duke Endowment’s hospital modernization program (starting in 1926) in the Carolinas, which funded construction and expansion, equipment upgrades, management reforms, and explicitly sought to serve both Black and White patients and attract better physicians, especially in rural areas. Using this context, the paper aims to identify causal effects of hospital modernization on infant and later-life mortality and to illuminate the mechanisms and racial disparities involved.

The paper situates its contribution within several literatures: (1) the drivers of early- to mid-twentieth-century mortality decline—comparing roles of public health and sanitation (e.g., water filtration, sewerage) versus medical care—providing rare causal evidence on hospitals’ roles pre-1950; (2) short- and long-run consequences of prenatal/infant health interventions and the fetal origins literature, extending it by documenting both short-run and adult mortality effects from a large hospital funding program and racial gap implications; (3) health workforce quality and regulation, complementing work on physician and midwife licensing by showing non-legislative infrastructure investments can attract higher-quality physicians; (4) complementarities between infrastructure and medical innovation, demonstrating that improved hospital quality increased the effectiveness of sulfa drugs; (5) management, standardization, and organizational practices in health care, highlighting the Endowment’s bundled approach (funding plus managerial reforms); and (6) the effects of large-scale private philanthropy (e.g., Rockefeller, Rosenwald) on public health and welfare, documenting substantial, cost-effective health gains and informing later public programs (e.g., Hill-Burton Act).

Data: The authors compile a county-by-year-of-birth panel for North Carolina (NC), drawing on (a) The Duke Endowment’s Annual Reports on hospital capital appropriations and payments (1927–1942), including project purpose/type; (b) NC death certificates (1917–1963) to construct infant deaths by county and year of birth; (c) Vital Statistics for births by county-year-race (1922–1948); (d) Social Security Administration Numident public-use records (deaths verified through 2007) to construct later-life mortality at ages 56–65 for 1932–1942 birth cohorts by county-year of birth and follow-up year; (e) American Medical Directory (1914–1940) to measure physician counts, quality proxies (school admission standards and years since graduation), licensure year, and location; (f) baseline pneumonia mortality (1922–1926) to form shares for sulfa-drug shift-share analyses; and (g) county controls (illiteracy, race shares, urbanization, retail sales per capita, county health departments). Treatment: An intent-to-treat (ITT) indicator turns on in the first year a county receives a Duke capital appropriation (and stays on). The authors also analyze intensive margins using cumulative appropriations and payments (2017 dollars). Empirical design: Primary estimates use difference-in-differences (DiD) with staggered adoption, county and year fixed effects, Poisson pseudo-maximum likelihood (PPML) to handle zeros and heteroskedasticity, and birth-count weights. Event studies test for pre-trends. Robustness includes: stacked regressions (Cengiz et al. 2019) in PPML and OLS; alternative DiD estimators (de Chaisemartin–D’Haultfoeuille; Callaway–Sant’Anna; Sun–Abraham; Gardner); Goodman-Bacon decompositions; varied functional forms (levels, log, log(·+1)); alternate samples and timing windows; exclusion of smallest counties; alternate control groups using non-Carolina Southern counties; randomization inference by permuting treated counties and preserving timing; and IV strategies. IV strategy: Instrumented intensive-margin exposure using the interaction of (i) cumulative Endowment returns over time and (ii) county eligibility (presence of a not-for-profit hospital pre-program in NC), leveraging variation in funding capacity and initial eligibility. Mechanism analyses: (1) Physician labor market—effects on total and quality-stratified physician supply per 100,000; (2) Complementarity with medical innovation—triple-difference combining Duke rollout (DiD), sulfa drugs (post-1937 shift), and baseline pneumonia mortality (1922–1926 shares) to test whether hospital modernization amplified sulfa effectiveness, especially in high-baseline pneumonia counties. Cost-effectiveness: Using intensive-margin estimates, compute cost per infant death averted from appropriations and payments (2017 dollars).

• Infant mortality: Duke support reduced infant mortality by roughly 10% (PPML ITT). Race-specific effects: −16.4% for Black infants and −7.1% for White infants, narrowing the Black-White infant mortality gap by about one-third. Event studies show no pre-trends and declines emerging with expected lags consistent with appropriation-to-implementation timing. - Contribution to statewide decline: From 1922–26 to 1938–42, infant mortality fell ~36% overall. Back-of-envelope calculations attribute about 28% of this decline to Duke support (43% for Black infants, 20% for White infants). - Later-life mortality: Exposure at birth to Duke support reduced mortality at ages 56–65 by about 9% (conditional on survival to age 56), with similar magnitudes for Black and White cohorts. - Intensive margin and cost-effectiveness: Per $1 million (2017$) in payments, infant mortality fell by about 6.6% (White) to 9.1% (Black). Conversion rate from appropriation to payment ≈ 0.45–0.48. Estimated cost per infant life saved ≈ $20,000 based on payments (≈$39,000 per life saved per appropriation), substantially more cost-effective than many later insurance-based interventions. - Mechanisms: • Physician workforce: Duke-supported counties saw an ≈11% increase in high-quality physicians and a ≈9% decline in low-quality physicians per 100,000, shifting average physician graduation year forward by ~12 years. • Complementarity with innovation: Hospital modernization and sulfa drugs were complementary; in counties with higher baseline pneumonia mortality, the mortality reduction from Duke support increased significantly after sulfa’s introduction, roughly doubling gains in high-need places. - Heterogeneity by project type: Additions, new plants, equipment, and conversions to nonprofit all show comparable effects on infant mortality; conversions show weaker payment-based intensive-margin effects but retain extensive-margin benefits. - Robustness: Results are stable across PPML/OLS/log specifications, stacked designs, modern DiD estimators, placebo randomization, alternative control groups (non-Carolina Southern counties), varied sample windows, and IV using Endowment returns × eligibility. No evidence of pre-trends; balance tests show no significant changes in economic covariates (e.g., retail sales per capita). - Ancillary outcomes: No meaningful changes in fertility; maternal mortality reductions comparable in magnitude to infant mortality but imprecisely estimated. - Equity: Absolute reductions were larger for Black infants (about −12 per 1,000) than for White infants (about −4 per 1,000), reducing racial gaps by one-third.

The findings show that modernizing hospital infrastructure and management significantly lowered infant mortality and produced lasting reductions in adult mortality, contradicting the notion that modern medicine had negligible effects before the 1950s. The larger benefits for Black infants indicate that improved access to higher-quality facilities and staff can reduce racial health disparities, especially where prior access was limited. Mechanism evidence supports two pathways: (1) hospital upgrades altered local physician labor markets by attracting better-trained doctors and reducing the share of lower-quality practitioners, raising the local frontier of medical knowledge; and (2) improved hospital capacity heightened the effectiveness of medical innovations like sulfa drugs, especially where infectious disease burden was high, demonstrating complementarities between infrastructure and technology. Robustness checks and IV estimates reinforce causal interpretation and mitigate concerns over selection, staggered adoption bias, and spillovers. Policy relevance includes the historical role of private philanthropy as a catalyst for later public programs (e.g., Hill-Burton), implications for current rural hospital closures and physician attraction strategies, and lessons for developing countries: bundling infrastructure investment with management support and oversight can generate large, equitable, and persistent health gains.

The study provides causal evidence that early twentieth-century hospital modernization substantially reduced infant mortality and improved survival into late middle age, with especially large gains for Black infants that narrowed racial gaps. Complementarities between infrastructure and innovation amplified benefits following the advent of sulfa drugs. The program was highly cost-effective, saving an infant life for roughly $20,000 (2017$). Beyond quantifying impacts, the paper documents mechanisms—upgrading the physician workforce and enhancing the productivity of new technologies—enabled by bundled capital and managerial reforms. Policy implications include the value of philanthropic-public models for scaling health infrastructure (as seen in the Hill-Burton Act), the potential costs of closing rural hospitals today, and guidance for health investments in low-resource settings where complementary management and oversight can magnify returns. Future research directions include: evaluating parallel impacts in South Carolina; assessing longer-term and broader outcomes (education, earnings); quantifying cumulative exposure effects across the life course; and studying analogous large-scale hospital construction programs (e.g., Hill-Burton) for their contributions to mid-century mortality declines.

• Geographic scope: Primary analyses focus on North Carolina due to data availability; results may not fully generalize to South Carolina or other regions. - Later-life mortality data: Public-use Numident is incomplete before 1988 and right-censored at 2007, restricting cohorts to 1932–1942 and ages 56–65 and precluding full lifespan analyses. - Measurement: Infant deaths rely on death certificates by county and year of birth; potential measurement issues (e.g., stillbirth classification) are examined with robustness checks that show stable results. - Treatment definition: Preferred estimates are intent-to-treat based on appropriations; actual payments and implementation timing may lag; intensive-margin analyses and IV mitigate this but cannot capture all nuances of utilization. - Potential migration and exposure dynamics: Long-run effects condition on survival and do not track migration; individuals may have received subsequent exposures as hospital capacity evolved. - External validity: The historical, philanthropic, and segregated context differs from contemporary settings; however, mechanisms and complementarities likely retain relevance. - Unobserved outcomes: The study focuses on mortality; potential gains in human capital or income are not directly measured.