LAUSR

HoNOS is one of the most widely used clinician rated outcome measures in mental health services. A commonly encountered problem is that one or more of the 12 individual HoNOS items is left uncompleted (missing data rates of up to 25% have been reported), which affects the degree to which organisations can rely on the accuracy of historical HoNOS data. In this brief paper we outline a simple statistical method of predicting missing item scores for HoNOS, both in general adult and 65+ populations. The method accounts for the average pattern of responding being non-uniform across items: i.e., some HoNOS items consistently elicit higher scores than others. By calculating individual item weights based on a very large sample of fully completed HoNOS assessments, we were able to accurately predict the value of missing items in a new sample. We contrast the accuracy of this approach with two other simple statistical procedures, and show that the weighted means model returns a much lower error rate. Although this is not the only method of predicting missing items, it carries the advantages of being: (i) free of charge, (ii) easily applicable to large datasets using a spreadsheet and (iii) unreliant on the availability of previous assessment data for the same patients. We hope this method will be of use to other organisations that are processing large volumes of HoNOS data.