Logo image
Back
The great equalizer?: hospital generative AI, robotics and population health
Doctoral - Dissertation   Open access

The great equalizer?: hospital generative AI, robotics and population health

Aaron Johnson
Doctor of Business Administration (D.B.A.), Drexel University
Mar 2026
DOI:
https://doi.org/10.17918/00011282
pdf
Johnson_Aaron_20265.23 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Causal inference Digital divide Generative AI Hospital technology adoption Population health outcomes Public Health Machine Learning
Hospitals are rapidly piloting generative AI (GenAI) and robotics, but their population-level effects and diffusion remain unclear. Drawing on Diffusion of Innovation (DOI) and Resource-Based View (RBV) theory, this dissertation links hospital technology capabilities to county health by integrating the 2024 AHA Annual Survey (hospital AI/robotics) with County Health Rankings (access, outcomes and context) for all U.S. counties (N = 3,143). Methods include index and component-level stepwise OLS with controls (policy, environment, behaviors, socioeconomic factors, log-population, Census Division fixed effects), doubly robust causal estimation via AIPW/IPTW for premature mortality (YPLL) and geospatial accessibility and LISA clustering. Three findings emerge. First (RQ1), technology adoption is associated with better access, outcomes, and efficiency. At the index level, robotics correlates with improved access ([beta] = -30.73) and outcomes ([beta] = -51.22), while GenAI is positively associated with hospital operational efficiency ([beta] = +1.21, p=0.0067). AIPW estimates suggest lower premature death rates where hospitals use AI for routine task automation (ATE = -933 YPLL; q=.06) and robotics (ATE = -470 YPLL; q=0.024). Second (RQ2), GenAI dampens the harm of unhealthy community behaviors: the IV3xGenAI interaction is negative and significant ([beta] = -11.84; on YPLL, AI task-automation interactions [beta]=-26.63 to -30.68), supporting a technology "buffer" or "Great Equalizer" effect; policy-tech interactions indicate diminishing marginal efficiency gains from Medicaid expansion where tech adoption is already high (IV1xGenAI [beta]=-2.46; IV1xRobotics [beta]=-0.92). Third (RQ3), diffusion is uneven: ~71% of the U.S. population lives within 30 minutes of an AI-enabled hospital (vs. ~84% for robotics), and the worst-served decile travels ~61 miles to reach AI-enabled care (vs. ~11 miles to any hospital), quantifying a technology access penalty that risks widening disparities absent incentives for targeted investment. Together, these results offer early national evidence that GenAI and robotics are associated with improved access, lower premature mortality, and greater efficiency, while GenAI uniquely moderates behavior-outcome links. The findings motivate policies that both accelerate diffusion to rural "cold spots" and prioritize high-impact, workflow-embedded AI to maximize public health ROI.

Metrics

165 File views/ downloads
69 Record Views

Details

Logo image