A push towards more accurate evaluation of health care quality has become a major public policy goal in recent years. The Department of Health and Human Services now has a website ranking hospital quality and California has a Healthcare Quality Report Card. One problem with simple quality measures such as mortality or morbidity rates are that physicians who have a patient base with a lower initial health level are penalized—according to the scoring—for treating the patients who need care the most. For this reason, health care rankings have moved from outcome-based rankings to process-based rankings. For instance, the California Report Card ranks insurance plans by such factors as whether or not individuals with diabetes had an eye exam and the percentage of pregnant women who had a check-up visit 21-56 days after delivery.
One puzzling finding Kahn et al. (2007) encounters in their data is that ambulatory care centers with better process of care scores have patient with a worse health outcomes. Using a simple ordinary least squares (OLS) procedure would lead to the erroneous conclusion that more care leads to worse health. Of course, individuals who are sicker are specifically the patients who need to undergo the most procedures.
To control for this problem, Kahn and co-authors use an instrument variables method. The independent variable—the process of care variables—is instrumented with a structure of care variable. The structure or care is defined in the data by a set of medical organization dummy variables. This IV estimation technique will provide unbiased, precise estimates if the medical organization dummies are correlated with the process of care, but the structure of care only affects health outcomes as mediated by the process of care. I believe that this is a reasonable assumption.
Using this methodology, the authors do in fact find that better processes lead to better health outcome. Moving a patient from a process of care score in the lower quartile to a process of care score in the highest quartile will lead to approximately a 10% increase in the PCS health score.
While the conclusion of this paper seems obvious—better process of care scores lead to better health outcome—the authors have done a yeoman’s job of proving this empirically. Further, the paper casts doubt upon the conclusions from other studies which use OLS regressions to compare health interventions and health outcomes.
- Kahn, Katherine; Diana M. Tisnado; John L. Adams; Honghu Liu; Wen-Pin Chen; Fang Ashlee Hu; Carol M. Mangione; Ronald D. Hays; Cheryl L. Damberg (2007) “Does Ambulatory Process of Care Predict Health-Related Quality of Life Outcomes for Patients with Chronic Disease?” Health Services Research 42 (1p1), 63–83.