Sunday, September 30, 2012
Online dating as a model for health care outcomes research
I heard an interview with Dan Ariely while listening to CNBC today. He is one of my favorite authors but I was not aware of his work done focusing on online dating services. What struck me about his work and comments was the nature of the critique; online dating services may be of limited value because they collect the wrong data. They collect data that is easily defined and relatively easy to collect such as income, education, height, and weight. Based upon the information presented, users of these services appear to be remarkably superficial in terms of what they value. However, they may be superficial because of the information presented to them.
I was struck by the parallels to current state of outcomes research efforts in health care. We tend to focus on measures which often are process measures or outcomes measures which are at best proxies to something more important. We use administrative databases because they are the only sources of structured data we might have access to. We do so because in the absence of robust things to measure, we end up collecting data on what we can collect and measuring what is measurable.
Ariely talks about collecting match-making questionnaires on a variety of people, some of whom he liked and others from those he did not care for. He then reviewed the blinded questionnaires and attempted to sort out from these data whether he could discern whether he could tell those which came from the people he had a personal affinity for vs. those he did not click with. He could not. Thus the information collected was not helpful in predicting personal chemistry. In the same sense I would predict that the current state of data collection and analysis used for assessing outcomes and quality will suffer from similar issues. By focusing on what we can measure, we are likely to focus on things that are not particularly important in the delivery of health care.
I think the lesson is not that we should not strive to collect data to drive practice change. We just need to be VERY circumspect about what we strive to collect and how we use it. Our processes focusing on performance improvement also need to be a focus of performance improvement.