Inpatient Rehabilitation Facilities: Alternative Definitions of Hospital Specialization Presented by Roberta Constantine, RN, MBA, PhD RTI International Presented at AcademyHealth 2007 Annual Research Meeting June 4, 2007 Orlando, Florida 1440 Main Street ■ Suite 310 ■ Waltham, MA 02451-1623 Phone 781-434-1711 Fax 781-434-1701 e-mail RConstantine@rti.org RTI International is a trade name of Research Triangle Institute Introduction About 8 percent of all hospital users are admitted to IRFs for PAC services (over 470 thousand admissions between June 30th of 2004 and 2005.) These specialized hospitals provide acute-level treatment in physical medicine and rehabilitation. IRF patients typically have a primary diagnosis related to musculoskeletal or neurological disorders or injuries and need acute-level physical rehabilitation treatments. 2 Medicare Policy In 2002, IRFs moved from TEFRA cost-based to case-mix, adjusted PPS. Hospitals must meet IRF certification requirements to be paid under the IRF PPS, else IPPS with much lower rates. To qualify for IRF certification, a hospital or unit must have admitted during its last cost reporting period, a certain percentage of cases within 13 diagnostic groups (75% Rule). 3 IRF Certification Groups Stroke Spinal cord injury, Neurological disorders Congenital deformity, Burns Amputation, Systemic vasculidities Fracture of femur (hip Knee/Hip replacement fracture), Active rheumatoid arthritis or Brain injury, severe/ advanced osteoarthritis 4 Study Purpose Are there alternative ways of defining IRF specialization that would allow for changes in medical practice? 5 Research Questions Do individual inpatient rehabilitation facilities (IRFs) specialize in certain populations? If so, what types of admissions, who are the patient populations and what types of IRFs are specializing? If specialization does occur does it affect patient outcomes? 6 Methods We analyzed data from: Inpatient Rehabilitation Facility Patient Assessment Instrument (IRFPAI), Patients’ prior acute, inpatient claims, and Provider certification data to examine whether other factors might be useful in classifying IRFs. Focus on the relationship between patient diagnosis, IRF characteristics, and functional outcomes. 7 IRF Characteristics Rehab Unit (%) Freestanding(%) Total(%) No. of Facilities 970 (82) 210 (18) 1,180 (100) No. of Admits 302,524 (68) 168,171 (36) 470,695 (100) Mean Admits 312 801 399 Median Admits 260 766 290 Nonprofit 668 (69) 71 (34) 739 (63) For Profit 159 (16) 131 (62) 290 (25) 8 Rehabilitation Impairment Categories (RICs) The Medicare Program classifies IRF patients by RICs. There are a total of 21 RICs, this analysis examined five in depth. Percent of Total Frequency Admissions Replace of Lower Extremity Joint 108,339 23% Stroke 84,114 18% Fracture of the Lower Extremity 64,847 14% Neurological 26,106 5.5% Burns 328 0.07% 9 Top IRFs with Highest Number of RIC Admits RIC Admits as % Total Admits RIC Admits of Total IRF Admits ReplacLE 5,343 1,612 30% Stroke 5,343 863 16% FracLE 5,343 550 10% Neuro 2,032 703 35% Burns 1,046 58 6% 10 FIM Scores by IRF Affiliation Units Freestanding RIC ADM CHG ADM DC ReplaceLE 77 26 72 31 49 25 Stroke 58 22 57 28 FracLE 65 24 58 26 Neuro 66 20 57 29 Burns 62 19 11 Creation of New Variables to Characterize Specialization Examined both the number of admissions for a RIC (RIC admits) and the number of admissions for a particular RIC as a percent of the facility’s total admissions (RIC_PCNT) simultaneously. RIC admits RIC_PCNT Quantiles 0-50 51-100 0-50 51-100 Low, Low X X Low, High X X High, Low X X High, High X X 12 Frequency Distribution and Percentage ( ) of Specialization Variables LL LH HL HH ReplacLE 527 (46) 340 (30) 74 (7) 209 (18) Stroke 417 (35) 480 (41) 127 (11) 155 (13) FracLE 428 (37) 449 (39) 99 (9) 182 (16) Neuro 548 (48) 262 (23) 61 (5) 267 (24) Burns 85 (59) 14 (10) 21 (15) 23 (16) 13 Functional Improvement Measure (FIM) The FIMTM instrument is “a measure of disability not impairment” (IRF-PAI Training Manual, 2004). It is a basic indicator of the severity of disability intended to measure what a patient actually does, regardless of the impairment, not what they should be able to do. There are 18 FIM items in all, scored from 1 total assist to 7 independent. There are six FIM categories; self-care, sphincter control, transfers, locomotion, communication, and social cognition 14 IRFs with Highest Percentage of RIC Admits as Percentage of Total Admits RIC Admits as % Total Admits RIC Admits of Total IRF Admits ReplacLE 630 439 71% Stroke 78 73 96% FracLE 69 35 51% Neuro 385 190 49% Burns 97 39 40% 15 Patient Severity of Illness (SOI) Patient SOI was estimated during the index acute care, inpatient hospitalization using the All Patient Refined DRG (APR-DRG) grouping software developed by 3M Health Information Systems. The SOI subclass has values of 0-4 corresponding to: no relevant comorbidities (0), minor severity (1), moderate severity (2), major severity (3), or extreme severity (4). A match was performed on IRF-PAI admissions to MEDPAR acute care admissions and a total of 370,308 claims matched the 470,695 IRFPAI records (78.7%). 16 OLS Models for FIM Outcomes for Selected RICs Dependent Variables: Total FIM Change Score, Total Motor Change Score (self-care, sphincter control, transfers, and locomotion), Total Cognitive Change Score (communication and social cognition) Independent Variables: Patient Characteristics – age, gender, white, severity of illness, LOS, Total FIM admission score IRF Characteristics – Region, Total Admissions, Rehab Unit Indicator, Specialization variables 17 Model Results – Adjusted R-Squared Total FIM Total Motor FIM Total Cognitive FIM Change Score Change Score Change Score ReplacLE 0.35 0.14 0.25 Stroke 0.12 0.11 0.11 FracLE 0.13 0.11 0.14 Neuro 0.09 0.06 0.10 Burns 0.15 0.16 0.09 18 Summary of OLS Regressions, Dependent Variable, Total FIM Change Score *P= or < 0.05. PTSOI1 is the reference category in the analysis; LL is the reference category in the analysis; New England is the reference category in the analysis. 19 Model Results IRF Characteristics Overall, specialization tended to be statistically significant (0.01 or less) but varied by direction, RIC, change score being measured, and specialization category. When the specialization variables were significant, they generally increased the FIM change score. Volume had the greatest influence: HL 1.6 points (Reple); 1.36 (Stroke); 1.7 (Neuro) High percentages in small admission groups were next most influential (LH) in general and especially in rarer cases 5.2 (burns); 0.67 (Reple); 0.69 (FracLE); 0.48 (Neuro) High volume, High percent had mixed results -0.37 (Stroke); -2.3 (Burns); 1.2 (Neuro) 20 Model Results - Patient Characteristics Patient’s severity of illness for indexed acute care hospitalization was the most significant patient variable. Varied across RICs and models – ranged less than one to six points. Other patient characteristics: Female tended to have a positive affect. Age was significant and negative for most models but less than one point effect. Ethnicity – being white was generally significant and positive LOS tended to be significant and positive but less than one point across the models. 21 Policy Implications Using definitions of functional and medical severity to determine appropriate IRF admissions would be consistent with the consideration that PAC is provided on a continuum. Volume and Percent of Admissions have significant effects on rehabilitation outcomes. These types of measures allow for variation within conditions to identify the acute-level rehabilitation case. 22
"Inpatient Rehabilitation Facilities"