Cost of Care for Veterans Receiving Primary Care in Patient Aligned Care Teams (PACT) vs. Geriatric PACT; Controlling for Selection Effects

Tuesday, June 14, 2016
Lobby (Annenberg Center)

Author(s): Ciaran S. Phibbs; Orna Intrator; Kenneth Shay; Bruce Kinosian; Winifred Scott; Sharon Dally; Thomas Edes; Richard Allman



Many systems have started using team-based primary care models and VA has invested heavily in its version, Patient Aligned Care Teams (PACTs).  The VA Office of Geriatrics & Extended Care [GEC] has establish geriatric-focused (GeriPACT) with lower caseloads (2/3 of PACT primary care, with additional disciplines such as pharmacy) to address the complex needs of patients with high demands for GEC services, especially those at high-risk for institutional placement or to be a high-cost patient due to multiple chronic conditions and/or cognitive or functional decline.  We examined if GeriPACT had an effect on total healthcare expenditures for these high-risk patients.


Among VA patients who received care in a PACT or GeriPACT in FY2013 at 84 VA facilities with verified GeriPACT, we identified those with a JEN Frailty Index >2 to target those who needed GEC services.  Patients who received care in both types of teams were excluded. Care provided by Medicare was included to insure capture all relevant diagnoses.  Total expenditures were adjusted for regional wage variation. NOSOS, VA’s refinement of the Medicare HCC model to include adjustments for service-connected conditions, psychiatric conditions and medication use, was used to create expected expenditures for each patient.  Regression models were used to examine the effect of GeriPACT on log expenditures.  As no valid instruments were available, propensity scores were used to establish a control sample that more closely matched GeriPACT patients.


Study cohort included 745,795 Veterans, including 26,296 in GeriPACT only.  The GeriPACT patients were much older, mean age of 82 vs. 64, and had higher NOSOS and JFI scores.  The differences in NOSOS increased with age, while the differences in JEN decreased with age.  Without propensity matching, management by GeriPACT was associated with significantly lower than predicted costs; 23% (95% CI 16-31%) lower for all ages and 11% (4-17%) for age>65.  These results persisted when controlling for other severity measures.  When the propensity-matched sample was used, there were no significant differences in the costs between the two groups; GeriPACT 4% (95% CI -0.04-12%) more expensive.  No age stratification was done in the propensity-matched sample as 95% of the GeriPACT patients were over age 65. 


Sample selection complicates the evaluation of GeriPACT as it targets patients at-risk for needing high-cost services.  While management by GeriPACT was associated with lower than expected costs for the select 3.5% of Veterans managed by GeriPACT in a simple analysis, there were no cost differences with propensity matching.  This is opposite what would be expected with adverse selection into GeriPACT.  Given that GeriPACT targets high-risk patients, it is likely that unobserved selection bias remains.  While it needs additional investigation, we believe that the direction of the change to using propensity matching is being driven by the low risk of many of the controls in the unmatched sample.  Additional study is needed to help establish criteria for which patients have the largest potential benefit from this more intensive care.  GeriPACT could also be affecting the quality of care in ways that didn’t affect costs.