Home  >  Analytics   >   Haunted by the Past: The Legacy of Claims Data Continues

Haunted by the Past: The Legacy of Claims Data Continues

by corasharma | August 22, 2013

Assume for the moment that we are indeed looking towards a future of widespread value-based reimbursement and provider-led population health management (PHM). Claims data should come to be seen as increasingly irrelevant, right?

This is what I initially assumed before starting research for our upcoming report: Clinical Analytics for Population Health Management (CAPH) Market Trends Report. (Editor’s Note: Report is scheduled for release next week.) Now, I am quite certain that claims data is here to stay, at least for the near future.

Let me first make the distinction between billing data and adjudicated claims data.

  • Billing data (837) is provider-owned. This data is sourced from PMS or hospital billing systems, and healthcare organizations (HCOs) have traditionally run analytics on this data for the purpose of increasing fee-for-service (FFS) reimbursement.
  • Adjudicated Claims data (medical claims, PBM Data, eligibility data) is payer-owned. HCOs entering into risk contracts with commercial payers should ALWAYS get access to this data from their payers; Medicare shared savings plan (MSSP) ACOs are automatically granted access to adjudicated claims data from CMS.

Both billing data and adjudicated claims data continue to be leveraged within provider-led PHM efforts. Billing data still plays a role in pay for performance (P4P) quality adjudication, as well as in supplying ICD/CPT to the PHM system when no EHR is to be found (albeit inaccurately).

While billing data has its limited uses, there are several benefits to adjudicated claims data that should not be discounted in provider-led PHM implementations going forward:

Claims-based analytics systems are mature. Payers have been running analytics on claims data for more than a decade. As a result claims data has become highly scrubbed and claims-centric vendors have developed a set of robust analytics offerings.

Truven, Verisk, Optum, ActiveHealth Management, et. al., have developed highly capable offerings for statistics-based predictive analytics, as well as for the traditional analytics technology stack. These include groupers (by disease, by episode), provider benchmarking, disease registry attribution, utilization vocabularies, risk scoring and risk adjustment, and rules for identifying care gaps.

Claims can help in creating the integrated patient record. The clinical-based patient record is richer than claims-based patient records but only includes data from one HCO and does not go far back in time. In contrast the claims-based patient record is sparse but more longitudinal across providers and deeper historically.

A much more complete integrated patient record can be created by combining the two. This patient record enriches a claims-based data skeleton with clinical data from the EHR and labs (yes, this record is a bit odd in that the richness of data falls off a cliff at some timestamp in the past)

Claims data can be used to link cost data to population/clinical data. HCOs taking on patient health risk absolutely need to tie cost data to population/clinical data. For example, an HCO might want to know the yearly costs for a specific high-risk diabetic patient cohort.

Adjudicated claims data reveals the charges paid, and currently HCOs are using these charges to approximate a notion of “cost”. Disclaimer: the usefulness of such approximation is up for debate, and this is not a post that discusses the enigma that is health care cost data or activity based costing.

Claims data enables clinical network management. Unless an HCO is a vertically integrated one like Kaiser Permanente, there is always the risk that care will be delivered outside of that HCO’s network. With post-adjudicated claims data, an HCO that is trying to manage affiliated providers in order to keep patients sticky to its network can:

  • Detect leakage. Which patients are visiting outside providers? Which physicians within the network are referring outside?
  • Manage inter-network referrals. Steer patients towards the most efficient (low cost, high clinical KPIs) providers for their condition.
  • Benchmark providers: Which physicians should be kicked out of the clinical network due to poor referrals management, excessive cost?

Looking forward. An HCO looking to leverage clinical and claims data analytics for risk contracts has many questions left to answer, for example:

  • Predictive modeling: Do we feed clinical data into claims-based predictive models, segregate claims and clinical data, or build one-off machine-learning clinical-based predictive models?
  • Integrated patient record: Should we build claims+clinical integrated patient records, or ignore claims data completely?
  • Cost Approximation: Should we use adjudicated claims charges to approximate cost, or take a more complex approach?
  • Exposing Claims Data to the a RBAC User Type  Do care managers want to know risk scores and cohort cost rollups? Do risk managing executives want claims-data risk-based functionality integrated with clinical data as well?
  • How to leverage payer-sourced eligibility data? Combine this with EHR demographics data?

The bottom line is that the touch-points between clinical and claims data are still fuzzy and there is no standard approach to integrating the two. In the coming year we hope to see further efforts in standardizing these touch-points among vendors that are lucky enough to be versed in both clinical and claims datasets.


Leave a Reply

Your email address will not be published. Required fields are marked *

Stay up to the minute.