Population Health vendors shouldn’t take patient matching for granted

If you visit any one of the thousands of population health vendor websites and sift through the promises of achieving the triple aim, you’ll most likely see a three-step process to filling the gaps of care:

  1. Aggregate patient data across silos.
  2. Analyze the aggregated data with predictive analytics and sophisticated risk stratification models.
  3. Generate configurable, out-of-the box clinical quality measures.

Some even offer the next logical step – addressing the newly discovered gaps of care with care coordination/management applications. Regardless of how far down the path of continued care they go, all population health tools must first aggregate data across silos and create the coveted longitudinal patient record. They have no choice. Pulling together patient data across disparate systems is fundamental to delivering clinical quality measures.

Because of its necessity, very few vendors point to patient matching as a differentiator. You don’t see Ford or Honda spend their advertising dollars on the fact that their cars have brakes. Similarly, population health vendors simply check the box and take for granted that patient matching is a function that just works, instead focusing their message on their proprietary components: predictive models, stratification logic, configurability, stack completeness, total cost of ownership (TCO), use of social determinants of health, a fancy user interface, ease of use, and a rich set of clinical quality measures. The problematic reality is that most of these vendors are simply passing the buck and relying on their clients’ abilities to uniquely identify patients and match patient records across data silos. For example, population health vendors are often depending on medical record numbers (MRNs) and existing enterprise IDs (EIDs) to link patient data before they begin the structural data aggregation process.

It is easy to see how this can create problems in the end deliverable if the client doesn’t have high quality patient matching. Spoiler alert: the clients rarely do. AHIMA estimates that healthcare providers have 8% - 12% duplicate medical record rates within their own systems, and then when exchanging data with other organizations (which should be done in a population health setting), that duplicate rate jumps to over 50%. A 50%+ duplicate rate is very difficult to ignore, so some population health vendors started including patient matching services in their offerings. However, these services are utilizing the same matching technologies that produced the 50% duplication rate in the first place. Population Health vendors are experts in predictive data models, risk stratification, data warehousing, reporting and a long list of other sophisticated technologies; patient matching isn’t one of them.

Ask your population health vendor today how they are matching your records. If they say they built or bought probabilistic or deterministic algorithms, ask them about Referential Matching. They are likely, and unknowingly, leaving many patient records out of your quality measurements.