I went to see my family physician for my annual visit. He mentioned that the chronic disease registry listed me as having hypertension and diabetes. That's not so, and my family physician recognized the registry assignment errors immediately.
I know where each of those mistakes arose. On one occasion 2 years ago, I had minimal elevation of my blood pressure, and many subsequent normal blood pressures. And in 2013, I attended an employee health screening (fingestick specimens) where I had a barely elevated fasting blood sugar of 108, with a normal upper limit of 100. The actual diagnostic criteria for diabetes are either a blood sugar 126 mg/dL or higher, or an A1C level 6.5 percent or higher on two separate occasions. I certainly don't qualify for the diagnosis of diabetes by either criteria.
What went wrong?
There must be a rule in my doctor's population health registry model that says "single abnormal value = disease", even though that is almost never the rule from the physician perspective.
For my healthcare organization to really manage population health, all the parties will need to trust the data and its analysis, and both must be trustworthy. The most convenient data (a single measure, misapplied in my case) may not be meaningful or trustworthy. The analysis can't just be trustworthy 70% of the time.
Good data + good models = good analysis
Good data + faulty models = faulty analysis
Understanding a population health registry takes hard work and clinical insight. When the data comes from multiple sources, such as billing diagnoses, clinical records, and from payment sources, there are many opportunities for error. Your team needs to understand the source of every data element.