PURPOSE: This study compared measurement accuracy between the converted FIM score generated from varied 4- and 8-item short forms (SFs) and the original FIM–13-item score. The comparisons were tested at (1) individual patient level using point differences and (2) group classification level using functional related group (FRG).

RATIONALE: The national prospective payment system (PPS) for inpatient medical rehabilitation reimbursement incorporated FIM to predict the cost of patient care resources since 2002. The FRG was part of case-mix groups for predicting the resources needed to provide patient care. The Centers for Medicare and Medicaid Services used the FRG to calculate PPS payment in inpatient rehabilitation facility (IRF). Patients in IRF were classified into different FRG groups based on their specific diagnosis.

DESIGN: A retrospective, secondary data analysis using a national veterans sample

PARTICIPANTS: A total of 2,499 veterans was included in this study. The majority of the sample was male (96.2%), White (65.6%), and married (42.5%) and lived in an acute unit at the same rehabilitation facility (44.5%) or at home (20.1%) prior to transit to the another facility. We included only participants who completed all the test items and had at least one of the following diagnoses: stroke (42.6%), lower-extremity amputation (18.9%), knee replacement (22.7%), and hip replacement (15.8%).

METHOD: This study used a retrospective data of 2,500 veterans from the Veterans Austin Information Technology Center. Each participant completed the full FIM October 2008–September 2010 in the settings of inpatient rehabilitation facility and community living center in the veterans health care system.

ANALYSIS: We utilized the short forms created in the previously developed FIM Minimum Data Set (MDS) item bank. Although FIM 13-item was treated as a standard (an actual score; FIMa), the FIM 8-item, FIM 4-item, MDS 13-item, MDS 8-item, and MDS 4-item were converted to the FIM scores (converted FIM; FIMc). At the individual level, we compared point difference of FIMa and FIMc using the Wilcoxon signed-rank test.

We also calculated percentages of point differences below 5 points and 10 points. At the group level, we compared FRG classification results using κ statistics. We also used paired-samples t test to compare FRG results across four diagnostic groups.

RESULTS: The mean converted FIM (FIMc) scores generated from FIM 8-item, FIM 4-item, MDS 13-item, MDS 8-item, and MDS 4-item SFs were within 1.07 and 0.05 points of the mean actual FIM 13-item (FIMa) scores. Wilcoxon signed-rank test showed a significant difference between the FIM 13-item and MDS 8-item (p < .05) and FIM 4-item SFs (p < .05). All forms had >55% of point differences within 10 points (range from 56.14%–99.04%) but above 30% within 5 points (range from 31.05%– 92.20%).

Kappa statistics overall demonstrated a fair to substantial (.34–.90) strength of agreement between functional-related group classifications generated from the FIMa and FIMc scores of each diagnosis, but the knee replacement FRG had the averaged least κ values. Paired-samples t test showed the amputation FRG had significant difference between the FIM 13-item and MDS 13-item, MDS 8-item, and MDS 4-item SFs (p < .05).

DISCUSSION: The results showed mixed findings. Individual score comparisons of FIMa and FIMc had low percentage of point agreement, whereas κ statistics showed different levels of agreement across diagnostic groups.

IMPACT STATEMENT: We implied that using short form for group classification might obtain better measurement accuracy than using individual converted scores.