Is Acute Rehabilitation More Cost-Effective Than Subacute Rehabilitation? Bruce Vogel, PhD This research was funded by the Department of Veterans Affairs, Health Services Research and Development Service (HSR&D), Stroke Quality Enhancement Research Initiative (QUERI), RRP 06-184, through the VA Rehabilitation Outcomes Research Center (RORC), North FL/South GA Veterans Health System, Gainesville, FL. This presentation does not necessarily represent the opinions or beliefs of the Department of Veterans Affairs, VISN 8, or the RORC. Collaborators Tracey Barnett, Ph.D. Dean Reker, Ph.D. Xinping Wang, Ph.D. Today’s Objectives To describe important changes in the structure of VA inpatient rehabilitation over the past 15 years To estimate differences in VA costs and outcomes for VA stroke patients in acute versus subacute rehabilitation units To draw some preliminary conclusions about the cost-effectiveness of alternative rehabilitation sites from the VA perspective to inform policy Restructuring VA Rehab Between 1995 and 2003, the VA closed 28 of its 59 acute hospital-based inpatient rehabilitation units (ARBUs), or 47% of its original complement of such units Over the same time, the VA established 24 new subacute nursing home-based inpatient rehabilitation units (SRBUs). Central Question What has been the impact of this major restructuring on VA rehabilitation costs and outcomes? Data Sources Demographic and clinical data from chart abstractions in previous study of the Stroke Impact Scale 483 post-acute stroke patients from 27 acute and subacute units at 23 different VA Medical Centers in FY2002-FY2005 VA cost data for index rehab stay and 24 months post-stay from VA Decision Support System National Data Extracts (DSS NDE) Dependent Variables Cost data divided into index stay, shortterm (0-3 months post-stay), long-term (4-24 months post-stay) and total 24 month costs Outcomes measured by admission and discharge total, motor, and cognitive Functional Independence MeasureTM (FIMTM) scores Independent Variables Both cost and functional status models controlled for: ARBU vs. SRBU type of stroke time from onset to admission admission motor and cognitive FIMTM scores demographics (age, martial status, and race) VA medical center Methods Followed standard econometric practice in estimating cost models (Manning and Mullahy, 1998) by choosing between OLS and generalized linear models based on residual kurtosis and heteroskedasticity. Estimated a distributed lag model of functional outcomes where discharge FIMTM is modeled as a function of admission FIMTM Examined full and parsimonious models (p<.20) Descriptive Statistics Variable Frequency (%) Acute (N=315) Frequency (%) Subacute (N=168) Mean (SD) 96.34 (21.43) 22214 (13818) Mean (SD) 91.97 (25.01) 24861 (17819) N (%) N (%) 187 (64.71) 124 (65.26) 102 (35.29) 66 (34.74) 300 (64.66) 11 (73.33) 164 (35.34) 4 (26.67) 159 (66.81) 152 (63.07) 79 (33.19) 89 (36.93) 133 (65.84) 135 (65.85) 16 (80.0) 8 (61.54) 19 (48.72) 69 (34.16) 70 (34.15) 4 (20.0) 5 (38.46) 20 (51.28) Mean (SD) 68.45 (11.25) Mean (SD) 68.85 (10.89) 71.08 (21.33) 24.21 (7.57) 46.88 (16.27) 26.29 (103.83) 65.38 (24.47) 21.30 (8.36) 44.07 (18.96) 45.71 (308.83) 27.63 (5.97) 68.70 (17.14) 25.33 (7.21) 66.64 (20.22) Dependent Variables Discharge FIM total score* Total index cost* Independent Variables Race White Other Gender Male Female Marital Married Not Married Impairment Code 1.1 1.2 1.3 1.4 1.9 Age Admission Score FIM Total* FIM Cognitive FIM Motor* Onset to Admission Descriptive Discharge Score FIM Cognitive* FIM Motor* * p <.05 Cost and FIMTM Model Runs Variable Acute RBU Impairment Code 1.1 Impairment Code 1.2 Impairment Code 1.3 Impairment Code 1.4 Impairment Code 1.9 Onset to Admit FIM Motor at Admission FIM Cognitive at Admission Age Married White Site 516 523 528 546 549 552 553 578 583 598 600 605 618 630 635 642 648 671 672 673 678 688 691 *p < .05; †p < .10 Full Total Index Stay Costs 2 R = .41 -5979.90* 765.07 1866.22 -2693.24 -183.92 reference -.27 -451.23* -75.75 -204.07* 41.99 1390.77 -7382.23† 2029.84 8397.23† -5887.32 8993.35* -4486.92 -6194.42 4128.98 12709.93* 7304.03* -890.91 -6843.06 2855.75 6627.13 -7677.20† -4266.90 -2015.00 -1291.84 -790.60 -2017.43 -4072.49 2628.07 reference Parsimonious Total Index Stay Costs 2 R = .40 -5850.79* -475.41* -191.82* -6105.27 1252.22 9106.78* -5975.21 9848.49* -3496.23 -6406.22 4223.46 13071.03* 7487.36* -550.78 -6053.16 3695.13 6926.72 -7466.79† -4502.05 -896.68 -546.89 -1013.02 -1110.00 -3580.50 2907.72 Full Total FIM 2 R = .68 8.63* .38 3.68 -4.21 -1.68 reference 0.002 .71* 1.03* -.32* .65 -1.80 12.28* 2.16 -2.05 8.75* -.88 1.72 9.06 -2.48 3.93 -.74 -1.96 15.40* 10.41* 9.04† -4.66 -4.97 13.09* -.30 -3.27 .75 16.18* 8.59† reference Parsimonious Total FIM 2 R = .68 8.57* .29 3.63 -3.90 -1.84 reference .71* 1.03* -.31* 11.50* 1.33 -2.49 8.67* -1.08 1.50 9.60† -2.82 3.53 -1.21 -2.05 15.13* 9.54* 9.37† -5.05 -4.44 12.18* -.78 -2.27 .26 15.63* 9.11* reference Significant Descriptives ARBU patients had higher admission and discharge FIMTM scores than SRBU patients (71.1 vs. 65.4 at admission; 96.3 vs. 92.0 at discharge) No differences in total, motor, or cognitive FIMTM gains were significant 100 90 80 70 60 50 40 30 20 10 0 ARBU SRBU Admit FIM Dischg FIM Significant Descriptives Average index stay total cost was lower for ARBU patients ($22,214 vs. $24,861) 25000 24500 24000 23500 23000 22500 22000 21500 21000 20500 ARBU SRBU Index Stay Costs Moving On . . . What happens to these results when we control for the independent variables listed above? Significant Model Results We find a considerably larger statistically significant difference in total index costs of approximately $6,000 (~25% of the index stay cost) After regression adjustment, we find statistically significantly higher total discharge FIMTM scores in ARBUs than in SRBUs (~8.6 FIMTM points) Cost Effectiveness? When outcomes improve and costs either decrease or stay the same, cost-effectiveness is a given. Implications While moving patients out of the hospital can often be cost effective, this may not always be the case VA policymakers may want to reconsider the trend of replacing acute, hospital-based rehab units with subacute, nursing home-based units Caveats Failed to find significant differences for short-term, long-term, and total two-year costs – implications? Observational data – Could unobserved selection bias be driving these results? Only examined VA cost – “business case” Only examined stroke patients – Do these results carry over to all rehab patients?