Functional Independence Measure versus Short Form-36: relative responsiveness and validity

Introduction In recent years there has been an increased emphasis on the use of outcome measures in health care. This emphasis has involved an expansion from traditional measures of functional independence (Bohannon et al., 1990; Emhoff et al., 1991; Dodd et al., 1993; Long et al., 1993) to measures focused on health-related quality of life (HRQOL) (Bayley et al., 1995; Bombardier et al., 1995; McGuigan et al., 1995; DiFabio et al., 1998). Since HRQOL emphasizes patient perspectives, it has the potential to contribute valuable information about the outcomes of rehabilitation programmes. In the United States much of the rehabilitation that once occurred in traditional rehabilitation settings now takes place in subacute facilities. Little is known about the relative usefulness of functional and HRQOL measures in such settings. If either type of measure is to be advocated in such settings it must be shown to be responsive and predictive of other important variables. The purpose of this brief report, therefore, is to describe for a small sample of orthopaedic patients treated in a subacute setting: (1) the responsiveness of the Functional Independence Measure (FIM) and the SF-36 HRQOL measure, and (2) the correlation of the FIM and SF-36 with five other rehabilitation variables. Methods Twenty-eight patients with a primary orthopaedic diagnosis of the back, pelvis, or lower extremity, who were admitted to a subacute setting for rehabilitation, served as subjects. The majority of the subjects had joint arthroplasties (11 hip, 9 knee), but 4 had fractures of the pelvis or lower extremity, 3 had spinal stenosis, and 1 had a spinal fracture. Twenty (71.4%) were women. The age of the subjects was a mean 69.1 (SD = 11.9, range = 43–87) years. All subjects passed a mental screen and provided written informed consent prior to participation. Within 72 hours of admission and prior to discharge subjects had their functional status and health-related quality of life measured. Functional status was quantified using the Functional Independence Measure (FIM). The FIM is a well-established measure for which reliability and validity is well established (Dodd et al., 1993; Hamilton et al., 1994; Pollak et al., 1996). For this study only the subscale, domain (motor and cognitive), and total score were employed. Patients’ HRQOL was quantified using the acute version of the Short Form (SF)-36. The reliability and validity of this tool has been documented extensively (Ware, 1993; McHorney et al., 1994; Anderson et al., 1996;Kopjar, 1996). For this study, scale scores and component scores were employed (Ware, 1993). All statistical analysis was performed using the SPSS and Systat statistical packages. Responsiveness was examined using effect size calculations and t-tests. Validity was determined by the Pearson correlations of the FIM and SF-36 measures with five rehabilitation variables: physical therapy visits and units, occupational therapy visits and units, and length of stay. Results Table 1 summarizes data relevant to the responsiveness of FIM and SF-36 scores. On the basis of effect size conventions and the t-tests, all FIM motor domain subscale (self-care, sphincter, mobility, and locomotion) scores, the FIM motor domain score, and the total FIM score were responsive. Specifically, the effect size of the aforementioned scores was either moderate (>0.50) or large (>0.80) and the discharge scores were significantly higher than the admission scores (P < 0.001). This was not true of the FIM cognitive domain subscale (communication, social cognition) scores or the FIM cognitive domain score. On the basis of effect size conventions, only one SF-36 score demonstrated a moderate effect size (physical functioning = 0.55); no score had a large effect size. Only three SF-36 scores were significantly better at discharge than at admission (physical functioning, P = 0.005; general health, P = 0.004, and mental health, P = 0.033).Table 1: . Statistics summarizing Functional Independence Measure (FIM) and Short Form (SF)-36 Scores for admission and dischargeTable 2 presents descriptive statistics for five rehabilitation variables. Table 3 shows the correlations between the FIM and SF-36 measures and these variables. Four FIM measures (self-care and locomotion subscales, motor domain, and total) were correlated (r = – 0.403 to – 0.692) significantly with every one of these variables. No other FIM measure or any SF-36 measure was correlated significantly with any rehabilitation variable.Table 2: . Statistics summarizing rehabilitation variablesTable 3: . Correlations of FIM and SF-36 measures with five rehabilitation variablesDiscussion Responsiveness contributes to the usefulness of measures in settings in which they are employed. Most of the FIM, for which the responsiveness is well established in other settings (Bohannon et al., 1990; Fiedler et al., 2000) was responsive when applied to patients with an orthopaedic diagnosis treated in a subacute setting. The only clear-cut exception in this study involved the communication subscale, which reflects an area not much affected in the sample of orthopaedic patients in this study. In contrast, most of the SF-36 was not responsive. Clear exceptions were the physical functioning and general health scales. Although the SF-36 has been shown previously to be responsive in patients with an orthopaedic diagnosis, that responsiveness was demonstrated over a considerably longer period of time (Kriebich et al., 1996; Arslanian and Bond, 1999; Benroth and Gawande, 1999). The need for samples much larger than those of this study has been documented for the SF-36 when differences in the order of those in this study exist (Ward, 1993). Thus, these results were not terribly surprising. Prediction of outcomes also contributes to the usefulness of measurements. Only FIM scores were useful in this regard. Specifically, only FIM scores containing motor items were correlated significantly with outcomes. As the patients had orthopaedic diagnoses, FIM scores containing motor items might be expected to be more highly correlated with outcomes than FIM scores (e.g. cognitive domain) not containing motor items. Adding support to the validity of FIM scores found to be significant predictors of outcome was the consistency with which they were predictive. That is, any FIM score that was correlated significantly with one outcome measure was correlated significantly with all. Although other studies have supported the ability of FIM scores to predict outcomes (Bohannon et al., 1990; Emhoff et al., 1991; Dodd et al., 1993), we believe this is the first to show the scores to be predictive of the five outcomes selected (in this study) in a subacute setting. On the basis of our findings, we can recommend the FIM but not the SF-36 for use as an outcome measure in a subacute setting. While the measurement of HRQOL is important, the use of a generic measure such as the SF-36 cannot be advocated in a subacute setting, either in lieu of or in addition to the FIM.