Health information technology (HIT) has great potential to improve health care; however, there is limited information about HIT effects on clinical care, especially with respect to commercially available technology in community-based ambulatory clinics. The dearth of quantitative information is particularly concerning given the significant resources needed for implementation and the limited adoption of HIT in the United States. We propose to evaluate the effects of ambulatory HIT on quality, safety, and resource use between 2004 and 2007. We will use a quasi-experimental pre-post design with concurrent controls, within a large, integrated health delivery system (IDS). In this natural experiment, the staggered HIT implementation across 110 primary care teams will occur over 38 months, starting in November 2004. This new commercial HIT will include an electronic medical record integrated with an order-entry system and decision-support; the current system is primarily based on the paper medical record. The IDS's current automated databases will permit consistent capture of our outcomes and patient-level covariates before and after the new HIT, but are notably distinct from the new HIT with respect to providing real-time clinical data. We will focus on a cohort of 780,000 IDS members with at least one of five chronic diseases (asthma, coronary artery disease, heart failure, diabetes mellitus, and hypertension) in January 2004; these patients may be particularly sensitive to HIT-related changes in ambulatory care. The presence of HIT in the primary care team will be the main predictor. The quality and safety measures include guideline-adherent drug dispensations and laboratory monitoring, drug adherence, and physiologic disease control (as measured by laboratory tests). The resource-use measures include emergency department visits, nonelective hospitalizations, and office visits. Using Poisson regression models with patient-level random effects, we will test the hypotheses that HIT is associated with improved quality and safety measures, and HIT is associated with lower visit rates. We will have the ability to detect small changes in our outcomes, e.g., 80% power to detect a 3% change in the relative rate of nonelective hospitalizations in the CHF group. We will make adjustments for patient, insurance, and organizational factors, including socioeconomic status, case mix, cost-sharing, and care-delivery structure, using automated data and annual surveys. In short, this natural experiment provides an ideal setting to understand the value of HIT in community-based, ambulatory care. ? ? ?
