The problem of missing clinical data for research in psychopathology: some solution guidelines

There are no guidelines to help psychiatric researchers statistically adjust for missing data. We discuss the problems resulting from missing values, and illustrate some of them with examples from our work. Using structured instruments, we obtained clinical information from 241 patients. Some instrument items were not rated, and these did not occur randomly: hallucinations and delusions were most frequently unrated, especially in chronic schizophrenics, and patients with high scores for other psychopathology. Systematically assigning an intermediate value between present and absent to nonrated items was a satisfactory solution, unaffected by nonrandom missing values. This simple solution was equivalent to a complicated one (vectoring) in discriminating patients. When relationships between variables are linear, we recommend the intermediate value method as a practical solution to missing values. We stress that missing values do not mean missing information, and the most common response to missing values (dropping subjects) is least informative.