4075 — De-intensification opportunities and patterns across VA facilities.
Lead/Presenter: Rob Holleman ,
COIN - Ann Arbor
All Authors: Hofer TP (Ann Arbor Center of Innovation, University of Michigan Medical School), Klamerus ML (Ann Arbor Center of Innovation) Holleman R (Ann Arbor Center of Innovation) Bernstein S (Ann Arbor Center of Innovation, University of Michigan Medical School) Kerr EA (Ann Arbor Center of Innovation, University of Michigan Medical School)
To quantify the amount and variation in opportunities for deintensification of care, to understand if we could meaningfully aggregate these overuse measures, and if so whether we could identify patterns in the variation of overuse that would suggest facility or system level approaches to support deintensification
Two expert panels assessed the validity of 44 actionable and measurable deintensification recommendations. Thirty-seven recommendations were considered valid by the panel, as assessed by the RAND/UCLA Appropriateness Method. Our funding and timeline allowed us to operationalize 15 recommendations as measures that could be applied to national VA data available from the CDW. We included Veterans who were active users of VA health care based on visits to the VA and primary care within a one year period preceding the observation window. Algorithms based on the recommendations used data available from both the VA CDW and CMS data and included procedure codes (i.e., CPT codes), diagnosis codes (i.e., ICD-9 codes) as well as medication prescriptions. Data from the VA CDW only include medication fill dates (in addition to prescriptions), laboratory values, and patient demographic information. A shrunken estimate of the mean opportunity for deintensification and 95% CI was calculated from a 2-level hierarchical generalized linear binomial mode. We calculated predicted deintensification opportunity rates for each measure with site of care as the cluster identifier.
The opportunities for deintensification, varied widely across the 15 measures from 1% to over 80%. The medication deintensification measures had higher than 50% rates of deintensification opportunities but smaller numbers of eligible patients than the cancer screening measures. However by virtue of much larger denominators (100,000 to over 4 million), the absolute number of people to whom the testing and screening measures applied was larger, with over half of the measures each suggesting overuse in over 100,000 people. There were relatively low correlations (0.30 or less) between measures across facilities, other than for closely related deintensficiation measures for the same disease (three diabetes management measures and two opioid and benzodiazepine overuse measure).
There are significant opportunities for deintensifcation of tests, treatments and preventive services within the VA with wide variation in deintensification opportunities that exist across VA clinical sites. But our results do not suggest actionable patterns of variation, given the low correlation between measures, and thus offer little evidence to support creating aggregate measures to assess deintensification, even within primary care
Our findings suggest that at least within VHA, a measurable â€œculture of efficiencyâ€ that promotes deintensification does not exist, and that missed deintensification opportunities seem to occur idiosyncratically within sites of care. To address such missed opportunities, systems will need to have multifactorial approaches that address both cultural changes broadly and individual opportunities specifically, including monitoring both underuse and overuse through a portfolio of balanced measures. Our study contributes a set of measures that can be used to motivate actions to stop or scale back unnecessary services within a broader effort focused on improving the value of health care.