Given the increasing number of patients requiring palliative care and the need for more professionals who are able to provide care for the dying comfortably, assessment of medical attitudes toward end-of-life care is becoming a key aspect of medical education. The present study aimed to establish whether the Frommelt Attitude Toward the Care Of the Dying, Form B (FATCOD–B) meets current psychometric standards of validity for an assessment tool in medical education.

The participants were 200 undergraduate medical students. Since in a previous study the FATCOD–B was found to have a weak structure due to poor item validity, a refined version was proposed and tested in the present study. Confirmatory factor analysis and the Rasch model were employed to assess its dimensionality and psychometric properties.

The construct measured by the FATCOD–B continues to be misspecified. The tool has a two-dimensional structure. The first is well-structured and demonstrates appreciable measurement and discriminant capabilities. The second has low validity because its measurement capabilities are based on weakly correlated items.

Our results suggest that the FATCOD–B measures a two-dimensional construct and that only its first dimension is a robust measurement tool for use in medical education to evaluate undergraduates' attitudes about caring for the dying.

A steady increase in the number of patients requiring end-of-life care has been observed during the last decades. The assessment of healthcare students' attitudes toward end-of-life care is an important step in their curriculum, as it provides information about their disposition to practice palliative medicine. The Frommelt Attitude Toward Care of the Dying Scale (FATCOD–B) was developed to detect such a disposition, but its psychometric properties are yet to be clearly defined.

A convenience sample of 608 second-year medical students participated in our study in the 2012/2013 and 2013/2014 academic years. All participants completed the FATCOD–B. The sample was randomly divided in two subsamples. In the item analysis, reliability (Cronbach's α), internal consistency (item–total correlations), and an exploratory factor analysis (EFA) were conducted using the first subsample (n = 300). Using the second subsample (n = 308), confirmatory factor analysis (CFA) was performed using the robust ML method in the Lisrel program.

Reliability for all items was 0.699. Item–total correlations, ranging from 0.03 to 0.39, were weak. EFA identified a two-dimensional orthogonal solution, explaining 20% of total variance. CFA upheld the two-dimensional model, but the loadings on the dimensions and their respective indicators were weak and equal to zero for certain items.

The findings of the present study suggest that the FATCOD–B measures a two-dimensional construct and that several items seem in need of revision. Future research oriented toward building a revised version of the scale should pay attention to item ambiguity and take particular care to distinguish among items that concern emotions and beliefs related to end-of-life care, as well as their subjects (e.g., the healthcare provider, the patient, his family).