2007 HSR&D National Meeting Abstract
1058 — Tele-Health Care Management of High-Utilizing Veterans
Rupper RW (Salt Lake IDEAS Center and GRECC) , Sauer BC
(Salt Lake IDEAS Center), Shen S
(Salt Lake IDEAS Center), South B
(Salt Lake IDEAS Center), Bair B
(Salt Lake GRECC)
Most tele-health programs have focused on disease specific management. We developed a care-management program with patient selection and dialogue content focusing on a cross-section of the highest utilizing patients at the VA Salt Lake station.
The intervention involved daily monitoring of these patients by nurse managers using a HealthBuddy®. This study measures the veteran’s utilization and costs prior to, during, and after participation.
For 126 patients enrolled in tele-health management between November 2003 and September 2005, we measured rates of emergency room visits, inpatient hospitalizations, prescription fills, and DSS pharmacy and total costs for the two years prior to enrollment and for the enrollment period. Random effects Poisson regression was used to model rates, and random effects linear regression was used for log costs. Rates were also tracked for up to an additional two years for a group of 40 patients who disenrolled from the intervention.
Rates are expressed as average per quarter (SD). Emergency visits fell from 1.0(.80) prior to enrollment to 0.6(0.7) during enrollment. Similarly, hospitalization rates fell from 0.17(.20) to 0.10(0.19), and prescription fills fell from 17(10) to 14(9). Mean pharmacy costs fell from $507(494) to $479(568), and total costs fell from $4785(4151) to $4119(4212). The decreases from pre-enrollment to enrollment period are all significant when modeled using random effects regression. For the sub-group of veterans who disenrolled from care-management, reductions were inconsistent across utilization categories during both intervention and post-intervention status.
High-utilizing veterans enrolled in a tele-health care management program lowered their prior rates of utilization across all measured categories during the intervention. This consistency in gains was not achieved by veterans who disenrolled.
The long baseline period of pre-intervention rates and the apparent benefit of adherence lend promise to the suggestion that these results are attributable to the intervention. However, using propensity scoring and a difference in differences approach, we are currently comparing intervention patients to a cohort of comparable veterans receiving usual care to more rigorously address selection bias and regression toward the mean.