Does Medicare Managed Care Affect Diabetic Patient Cost, Use and Quality Associated with Different Medical Labor Inputs?

The Medicare market has changed dramatically since the passage of the Medicare Modernization Act (MMA) in 2003. The managed care share of the Medicare population has nearly tripled from 13% in 2004 to 30% in 2014 (Kaiser Family Foundation, 2015).  At the same time, the production of medical care for patients with chronic conditions has altered significantly as the use of non-physician medical labor has increased.  In this paper we use claims data for the diabetic population from 2011 to 2013 to compare how health care demand, cost and quality differ between fee-for-service (FFS) and Medicare Advantage (MA) plans.  Furthermore, we examine whether there are annual differences in these measures for nurse practitioners (NPs) and primary care physicians (PCPs).

The data for this analysis come from two sources.  For the traditional FFS Medicare program, we use the 5% Standard Analytic File of claims and beneficiary data representing roughly 1.7 M covered lives. For the MA population, we use claims and enrollment data furnished by the Health Care Cost Institute (HCCI), which represents claims from several large insurers operating across multiple states.  We focus on a subset of the data, namely seven states with large MA populations and with varying levels of scope of practice for nurse practitioners. We stratified the study population to compare the healthiest versus least healthy diabetics using a medical productivity index (MPI).  A key design feature of the MPI is the use of common variables found in insurance claims data used by all public and private insurers.  Given the high variability in the ratio between provider charges and payment, the use of a simple set of non-cost variables permits robust comparison of clinical input and output at the patient level without relying on negotiated provider prices for services as a potential biased metric of resource consumption. 

We found significant differences in health care demand, cost and quality between the diabetic population seen by NP versus PCP in both the MA and FFS populations.  For example, in the FFS population, average inpatient costs for the sickest strata of patients were higher for PCPs ($38,937) than NPs ($9,727) in 2012. Average use of HbA1c diagnostic tests was lower for PCPs (0.69) than NPs (0.72).  In the MA population, inpatient costs for patients managed by PCPs were lower ($23,270) than NPs ($36,188), but HbA1c testing was higher in NP-managed patients (0.30) than the PCP-managed patients (0.28).  A similar pattern held for the strata of healthiest patients.  Preliminary Generalized Linear Model (GLM) estimates across all three years, accounting for time and location, show the MA population with lower cost and a neutral impact on quality compared to the FFS population.  The impact of NP vs. PCP varies but suggests a pattern of lower non-institutional use and cost for NP-MA versus PCP-MA that is consistent across all three years.  Although very early, these findings could inform current labor market policy development as well as providing evidence of the impact of explicitly managed health Medicare insurance.