“My patients are different.” How many times have heath system and health plan leaders heard this argument during alignment discussions with physicians?

There are so many factors that may influence a physician’s care decisions. An essential challenge in provider benchmarking is fairly and accurately comparing the provider performance of one provider to another in a way that is comparing apples-to-apples.

Therefore, any solution that aspires to create fair provider performance benchmarks will need to first address the physician’s inherent (and valid) argument, “my patients are different”. Provider benchmarking can only spark changes in behavior if clinicians trust that the case-mix adjustment methodologies underpinning the healthcare analytics account for their patients’ demographic, clinical, and social determinants of health (SDoH).

Analytics companies, such as 3M, Milliman, MCG, and Optum, have attempted to make case-mix / severity adjustments in their assessments of provider performance, oftentimes without the granular SDoH insights needed to paint the full picture. Based on our analysis and collaborations with industry partners, we’ve uncovered several drawbacks and inconsistencies in the provider benchmarking methodologies most commonly seen in the market.

To address the challenges of incumbent models, Clarify created a differentiated approach to provider performance benchmarking that takes into account case mix, which include SDoH insights on what is impacting patients on each physician’s panel. The differentiators can be summarized as follows:

  • Inclusion of comprehensive case-mix / severity-adjustment factors. These include demographics, clinical diagnoses and SDoH insights drawn from the largest patient-level dataset on the market.
  • Precise modeling using machine learning. Instead of cohort-based approaches, we use generalized linear models which look at each individual provider’s panel and the values associated with their unique patients. This is basically the difference between using calculus to calculate (our approach) vs. summing rectangles to approximate (cohort) area under the curve.
  • Comparison of the benchmark value to actual value. This means that you do not require any additional “severity score” when comparing outcomes.
  • Not a black box. Clarify can highlight how each variable contributes to the case-mix / severity-adjustment, not just provide a black box score.
  • Dynamic benchmarking and custom care groupings. Because of the speed and precision of our machine learning platform, Clarify allows for flexible interrogation and drill-down within any category, including the ability to change the comparator dataset on the fly

Realizing the promise of big data in healthcare rests on the ability to effectively distill meaning from traditionally fractured, unconsumable data sources. It is only with fair and trusted provider performance benchmarks that healthcare organizations will truly move the needle in achieving their goals for cost and quality optimization.

Read our latest white paper The Future of Benchmarking.