Evaluating the Quality of Health Care
5. Outcome Measures
Another problem is that the outcome of interest may be too rare to give us a reliable picture of quality for the entity of interest. For example, counting breast cancer deaths might seem like a good way to assess the effectiveness of health plans’ screening and treatment programs for breast cancer, but only about one of every 1,000 women over the age of fifty die in any given year. Thus, screening (versus no screening) might change a group or clinic’s death rate by only three per 10,000 patients a year. It becomes even harder to detect differences when it is not an either/or situation. If rates of screening varied by ten percentage points between clinics or medical groups, the difference in death rates might be only three per 10,000 patients per year (Eddy, 1998). To detect such differences, one would need a very large sample, much larger than might be available for individual providers or even group practices or small health plans. If one wanted to assess the quality of diabetes care using available quality measures, some have estimated that a physician would need to have more than 100 patients with diabetes for the quality measures to have adequate reliability (Hofer, et al., 1999), again a limitation if one wanted to assess the quality of care provided by individual physicians.