The following analysis, although not a randomized controlled trial, tests a potential, and important, association between oral and systemic health. Data are derived from a convenience sample of insured persons with both diabetes and periodontal disease. Although the generalizability of these results to other populations (such as the uninsured) is not known, we believe that these findings could serve as a springboard for further research exploring this association. This compelling preliminary analysis may be of interest to researchers in many arenas, including dentistry, chronic disease, and healthcare costs.
Mining Insurance Data to Answer Clinical Questions
The exploratory studies needed to detect associations between clinical conditions and potential contributing factors (demographic, environmental, genetic, or medical) pose a special challenge in research. On one hand, large sample sizes are needed to detect subtle influences in the presence of strong known effects (eg, smoking) or confounders. On the other hand, the tentative nature of the hypothesis may not justify the effort and cost of large (usually multicenter) randomized controlled trials at an early stage of knowledge. When possible, researchers seek to "mine" historical records as an early step in determining the credibility of a hypothesis. We, as well as others, have found insurance records to be especially rich and reliable sources of health data (such as healthcare costs). What they lack in medical, dental, behavioral, and adherence detail is often offset by their uniformity and sheer size. The trick is to formulate the research question in a way that it can be answered from the available data.
This article discusses how data from a group of large private insurance plans were used to investigate whether periodontal health affects the cost of medical care in patients with type 2 diabetes. Although the findings are interesting, the process may also be relevant to other clinicians trying to interpret reports derived from these valuable data sources.