Fragmented data and high utilizers

JAMA**  had a short letter on the problems of identifying high utilizers based solely on hospital owned data:

Of 45 861 individuals with more than 5 emergency department visits to Maryland hospitals in 2014, 8438 (18.4%) visited more than 3 hospitals, 8905 (19.4%) visited 3 hospitals, 14 627 (31.9%) visited 2 hospitals, and 13 891 (30.3%) visited only 1 hospital. Only 27 251 (59.4%) of the 45 861 made more than 5 emergency department visits to 1 hospital, meaning that the rest would fail to be identified by individual hospital analyses.

As a result of new payment incentives that favor reductions in avoidable illnesses, more hospitals are working to identify and assist high utilizers. For more than two-thirds of frequent visitors to emergency departments inMaryland, however, no single hospital had a full picture of their care. Moreover, hospital-specific analyses would have failed to identify approximately 2 in 5 high utilizers. These results do not account for the fact that some facilities within the same hospital system may be able to share data….

This would be the argument for hospitals that are seeking to consolidate and merge with other hospitals.  They can share data better and see fairly quickly that Mr. Smith has shown up at East Nowhere General’s ER  on Monday and is now at St. Maria of Mercy’s ER on Wednesday morning with similar symptoms.  Those symptoms could be indicative of a long term problem or drug seeking behavior.  That would be the hospital argument.

It is also the case example of the argument that integrated payer provider delivery network make, especially those systems that do not have exclusive walled gardens.  An Integrated Delivery Network (IDN) owns the hospitals, owns the docs, and owns an insurance company.  A non-exclusive walled garden IDN would have both owned and non-owned hospitals submitting claims for the same individual.  The insurer would see the data stream and be able to create more comprehensive situational health and expense awareness than any single point in the provider chain.  This is especially true as the insurer side of the IDN is also seeing pharmacy /prescription claims.

Ideally an IDN would be able to integrate all of the data that they receive from a variety of streams, identify members who are likely to be high utilizers, split them into drug seeking and non-drug seeking streams and then apply evidence based interventions to either divert high cost utilization to lower cost utilization or actually solve the problem the person has so there is less repeat utilization.  That is the best case scenario.

One of the big problems with using claims data for decision making is lag.  ER claims are not always submitted the day of service.  At Mayhew Insurance, we figure that we’ll see roughly 40% of the ER claims come into our claim system within thirty days of the date of service.  By sixty days, we’re up in the high 90%’s, but billing is slow.  Furthermore, claims information are not clinical notes.  They will contain the information needed to pay not the information needed to treat.  Diagnosis information will be haphazard and idiosyncratic, rule-outs won’t be on the claim, the initial doctor’s thought process can’t be discovered, notes about previous activity are non-existent.  Finally, due to timing, the picture would be incomplete as one visit might be submitted, another visit at a second hospital is sitting in Accounting’s work queue, and the third hospital has a patient walking up to the triage desk.

If we are going to engage in population health management then we either need a comprehensive national medical records data repository with a common interface for a wide variety of legacy Electronic Medical Records (EMR) systems to dump into and pull out of.  That would be a security nightmare.  Or we need to be able to give people universally readable and secure medical record cards that can hold six months or a years’ worth of data at a time while also being able to be easily read and easily written upon.  I have no idea how to do that.  Counting on hospitals to manage population health with severely compromised data on high utilizers is asinine.

** Horrocks, D., Kinzer, D., Afzal, S., Alpern, J., & Sharfstein, J. M. (2016). The Adequacy of Individual Hospital Data to Identify High Utilizers and Assess Community Health. JAMA Internal Medicine JAMA Intern Med, 176(6), 856. doi:10.1001/jamainternmed.2016.1248