Can the modified functional independence measure be reliably obtained from the patient medical record by different raters?

BACKGROUND: The National Trauma Databank minimum dataset includes three items of the Functional Independence Measure (FIM) collected at hospital discharge. Reliable collection of the modified FIM through medical record review could enable trauma registries to obtain a functional outcome measure without patient interview. This study investigated the interrater reliability of the medical record modified FIM, and the level of agreement with the modified FIM scored through patient assessment.
METHODS: Four raters scored the modified FIM from the medical records of 44 major trauma patients postdischarge. The medical record modified FIM scores were compared with the modified FIM score obtained through patient assessment at discharge. Weighted (Kw) and unweighted [kappa] statistics were used to describe absolute and near agreement between pairs of raters.
RESULTS: The average level of agreement between raters was slight for the feeding (mean Kw, 0.13) item, fair for the expression (mean Kw, 0.28) item, and moderate for the locomotion (mean Kw, 0.47) item. Compared with the patient assessment modified FIM, the average level of agreement was moderate (mean Kw, 0.55) for the locomotion item, and fair for the expression and feeding items, with mean Kw of 0.38 and 0.20, respectively.
CONCLUSION: The interrater reliability of the medical record modified FIM, and agreement with the modified FIM obtained through patient assessment, was low. Meaningful comparisons of this outcome measure across registries and hospitals could be limited, and comparisons within a registry or hospital over time may be similarly affected by changes in staff.