Health IT Diffusion and Physician Density

• In Event: Health IT and Outcomes

Abstract

Digital transformation in health care promises to improve the practice of medicine. While extensive literature examines how health information technology (HIT) impacts the quality of care, its effect on the density of working physicians, the key players in care provision and primary users of the new technology, remains underexplored. We fill this gap by examining how the diffusion of hospital HIT affects the density of hospital-based (HB) physicians at the county level. We exploit the sharp increase in HIT adoption rates across US counties during 2005–2017, using an event-study framework following East et al. (AER 2023). To minimize potential bias from TWFE estimators, we employ the De Chaisemartin and d’Haultfoeuille (REStat 2024) estimator.  

We find that counties with a rapid HIT diffusion experienced an average 11% increase in HB physician density, corresponding to 2 physicians per 100k population, relative to comparison counties. This growth is mainly driven by early-career physicians and is concentrated in primary care shortage areas. These results suggest that HIT adoption enhances efficiency and creates new opportunities for young physicians (under age 45) in less competitive areas. HB physicians in treated counties are found to receive higher Medicare payments and perform more Medicare services, likely due to administrative efficiencies and improved care coordination that increases patient throughput. Hospitals in treated counties experienced relatively higher profits following HIT diffusion, driven by substantially increased revenue compared to modestly increased cost. Our results have important policy implications. They suggest that enhancing HIT infrastructure could attract more (especially young) physicians to areas facing shortages and expand care capacity, thereby reducing geographic disparities in health care access. 

The rapid diffusion of HIT was largely driven by the HITECH Act and various indirect incentives related to HIT in the ACA, both of which add (withhold) Medicare payments to hospitals for adoption (non-adoption) of HIT. These large regulatory drivers provide strong exogenous incentives that reduce concerns of endogeneity in HIT adoption. The lack of differential pre-trends in HB physician density between treated and control counties in event studies lends support to our identifying assumption. 

Nonetheless, we apply additional approaches to address potential threats to identification. First, to account for unobserved heterogeneity across counties, we include an extensive set of controls and fixed effects. We also examine how typical county characteristics evolved before and after treatment and find no evidence of correlation with the timing of rapid adoption. Second, to address concerns about potential spillovers of local HIT diffusion, we apply a matching-based DiD method, restricting control counties to be geographically distant but otherwise observably comparable, and find similar estimates. Third, we drop counties experiencing substantial mergers and acquisitions or vertical integration and find our results are not driven by healthcare consolidation. Furthermore, placebo tests confirm no significant effects on physician density for those in practices or specialties that are less affected by HIT adoption (e.g., physicians engaging in research/teaching activities, physicians in less HIT-relevant specialties such as dermatology and occupational medicine, or physicians employed by federal agencies).

Author

• Jianjing Lin, University of Massachusetts at Amherst
  Presenting Author