Health IT—Helpful or Not?

The promise of health information technology (health IT) has long been touted as a major cost saving investment in the US healthcare system. President Obama is quoted as saying that updating and computerizing the healthcare system could save “billions of dollars per year.” In terms of patient mortality or length of stay, however, there have not been meaningful results for the majority of patients. A recent working paper by Jeffery S. McCullough, Stephen Parente, and Robert Town, entitled “Health Information Technology and Patient Outcomes:  The Role of Organizational and Informational Complementarities” attempts to quantify improved health outcomes across a variety of diagnoses and IT adoption rates with a nationwide study of Medicare patients.

The theory behind health IT‘s effect on quality is simple. It suggests that the technologies will be used to automate rule-based treatment guidelines and protocols along with patient data to reduce errors and suggest treatments, especially for patients with fairly simple diagnoses.  Additionally, for patients with more complex conditions, information management and communications across different specialists will be improved, leading to better case management. The former was the basis for receiving subsidies under the 2009 HITECH act.

In the paper, the authors employed a difference in difference (DID) regression to control for unobservable effects across hospitals. They used adoption information from the Health Information Management System Society (HIMSS) to document timing of adoption of various types of health IT technology, such as electronic medical records and order entry systems along with discharge records from all Medicare patients at fee-for-service hospitals who had been admitted for acute myocardial infarctions (AMI), congestive heart failure (CHF), coronary atherosclerosis (CA), and pneumonia (PN). The data contain over 6.6 million observations, along with rich demographic and patient history information. They used the data to construct patient severity measures, indicating the degree of risk for mortality. Importantly, the rich dataset allowed for estimation of heterogeneous effects across diagnoses. The researchers utilized AMI as a control, as important decisions on treatment for heart attacks are standardized and do not utilize health IT as much.

When analyzing the data, the authors found no statistically significant improvement for the marginal patient. That is, those cases that were projected to benefit most from health IT expansion saw no effect on outcomes. However, for patients with more complex and severe cases, there was a statistically significant improvement in outcomes. For patients in the 60^th decile or higher of severity, health IT adoption was associated with a reduction in mortality in about 200 per every 100,000 admissions.  With specific diseases, IT adoption prevented deaths in 500 out of every 100,000 admissions of PN patients and 100 deaths for every 100,000 admissions for CA and CHF patients. Notably, IT adoption had no effect on the control diagnosis, AMI.

The authors also looked for positive network externalities, based on the idea that more widespread adoption of health IT would bring more effective use of the technology. While they found no statistically significant correlation, the indicator they used is the estimated impact of neighboring hospitals’ adoption rates. This may not be a good indicator, as these hospitals are not necessarily sharing information. A better indicator would utilize information about the amount of information shared about patients with primary care physicians and other specialists through health IT. A report by the Kaiser Family Foundation states that the lack of health IT system uniformity has limited its efficacy and gains to its adoption. Access to the primary care physicians’ information or information on how integrated the IT network is with local physicians and specialists would provide a wider range of the patients’ history and be more indicative of the benefits to sharing information.

While this study indicates little to no gains from health IT adoption, it focused on the Medicare population, who are older, more likely to have negative health outcomes, and are not necessarily representative of the effects on the general population. Another limitation is that the study focused only on four diagnoses and looked only at admitted patients at hospitals. As most health care is provided by general practitioners and family doctors, one would expect benefits to accumulate in these practices at a higher rate than at hospitals. Still, we may need to temper our expectations of the gains to health IT adoption. Widely accepted diagnoses and procedures are likely to be practiced by competent physicians regardless of the assistance provided by health IT systems. This study suggests that this is the case, although health IT provides some benefits to those patients with the most complicated diagnoses.

Photo cc: Positive Force Production...