The Impact of Health Information Technology on Demand for Hospital Inpatient Services.

摘要

ealth Information Technology (IT) research has been focused on health IT adoption and the supply-side effects such as the quality and efficiency of health care. This demand analysis complements the existing supply-side analyses, allowing for a more complete understanding of the impact of health IT on health care markets. The impact of health IT on demand for hospital inpatient services is estimated using Medicare beneficiary inpatient hospital admissions as a measure of patient choices. Two complementary discrete choice models are used to model patients' choices with the underlying assumption that patients are making a utility maximizing decision. Berry's specification of a linear market share model provides mean effects of health IT on hospital market share at a national level. A patient-level conditional logit model which includes interactions of patient characteristics and health IT is also estimated for a subset of hospitals and diagnoses. Hospital inpatient admission data from 1999-2006 was obtained from the MedPAR file. The data for this study includes 100% of Medicare fee-for-service (FFS) beneficiaries over age 65. Hospital characteristics were obtained from American Hospital Association annual hospital survey. Hospital health IT system information is from the HIMSS/Dorenfest Integrated HEALTH CARE DELIVERY SYSTEM PLUS (IHDS+) DATABASE(TM). The impact of three technologies is evaluated: (1) Picture Archive and Communication System (PACS), (2) Computerized Physician Order Entry (CPOE) and (3) Electronic Medical Records (EMR). Combinations of these technologies are also studies. A panel data structure including hospital fixed effects is used to identify the impact of health IT on demand. The hospital fixed effects are included to control for endogeneity in hospitals' adoption of health IT and patient choices. The health IT variable and interaction terms were jointly significant in market level and individual choice models for CPOE but did not result in significant impacts on hospital demand. Patient-level conditional logit model results are used to calculate consumer surplus welfare measures for hospitals with both EMR and CPOE systems. In 2006 approximately 10% of the analysis sample of hospitals had adopted EMR and CPOE. The change from no adoption to the 2006 adoption level produces a