FAM Properties

Validity: Marosszeky (1992) investigated the relative contribution of the FIM and FAM items within each of seven areas of function. He found that the FAM items contributed significantly to the overall proportion of disability in all categories and in the total score for the ratings 2 to 6. The FAM items did not contribute to extreme ratings of 1 or 7. He speculated that the FAM may become an even greater contributor to a valid measure of disability during the follow up period, as many of the FAM items relate to community functioning (i.e., car transfers, employability, community mobility, adjustment to limitations).

Rasch Analyses were completed on the FAM items at rehabilitation admission and discharge for data collected in the TBI National Database (Hall et al., 1993). FAM items rated at rehabilitation admission correlated significantly with indices of injury severity in a very similar pattern as FIM items.

The FAM does not appear to contribute beyond the FIM in predicting length of inpatient stay or costs. The FAM does appear to have increased sensitivity beyond the FIM alone for inpatient rehabilitation discharge and post-acute rehabilitation functional assessment. There is evidence that the FAM at rehabilitation discharge has less "ceiling effect" than the FIM and is more strongly related to rehabilitation charges than the FIM (TBI Model System National Database, Hall et al., 1996).

In a prospective study of individuals with TBI at 6 months and 24 months post-injury, FAM scores were predictive of return to work and community integration (Gurka et al, 1999). Overall, the FAM did not provide any additional information over the FIM except at 24 months where the FAM cognitive scores slightly improved the prediction of return to work.

Inter-rater Reliability: Preliminary data suggest that FAM items involving abstract concepts such as "attention" tend to be less reliable than directly observable behaviors (Hall, 1992). As a test of the FAM items interrater reliability, FAM training and testing vignettes (which also included the FIM items) were completed by trained raters, including a PT, OT, COTA, Data Analyst and researcher. Each item's three vignettes were written to cover low, intermediate, and high functioning levels within the 1 to 7 scale. The percentage of rating agreement was 89%. The Kappa score for the FIM was .87, and for the FAM, .85, both within the "very good" Kappa range.

Individual raters ranged in agreement with the "correct answer," from a high of 94% to a low of 81% of composite percentages across items. Keep in mind that these are ideal conditions, i.e., each rater is receiving the same information and sufficient information on the patient to do the rating, and each rater is trained and is using the definitions of terms in the Decision Tree in making ratings. What the findings do show is that FAM items can be rated with close to the same accuracy as FIM items, given the conditions stated above. This is encouraging because FAM items are generally more abstract (cognitive, psychological, social) than FIM items.

In another inter-rater reliability training endeavor, the four TBI Model Systems were given the vignettes to train and test several raters from different disciplines at each center. Each of the centers passed with an overall accuracy hovering just above 80% for the FIM+FAM.

In a recent study by Donaghy et al. (1998), all items of the FIM+FAM showed excellent or good intraclass correlation coefficients (ICC), with the exception of a FIM item, "Social Interaction" (.36). The next poorest correlation was for the FAM item "Emotional Status" (.62), considerably stronger than "Social Interaction." Consequently, the addition of the nine FAM items to the cognitive/psychosocial domain actually increased the overall reliability of this subscale to .74 from .69 for the FIM items alone. The total average ICC values for the 18 FIM items (.85) and for the 30 FIM+FAM items (.83) were comparable, both falling in the excellent range.