SUMMARYThe Quality of Interactions Schedule (QUIS) was developed as part of the prospective evaluation of two residential domus units for elderly people with severe mental illnesses. QUIS is an observational strategy in which social interactions between residents and care staff are coded as positive social, positive care, neutral, negative protective or negative restrictive. The interrater reliabilities of both the observational technique and the category codings were good. In the domus evaluation, QUIS demonstrated that the number and quality of interactions were significantly improved in both domuses compared to a baseline hospital ward at 3 months, 6 months and 12 months follow-up. In particular, the number of negative interactions observed fell to almost zero in the domuses at 12 months.KEY woms-Domus, nursing home environments, quality of care, dementia, schizophrenia, QUIS In recent years, observational methods have found a useful place in the evaluation of continuing care facilities for elderly people with mental and physical impairments. In these settings the use of other means to explore residents' experiences, such as interviewing and self-rating, is seriously limited because of problems with respondent bias, reluctance to disclose sensitive or critical information, and unreliability due to cognitive impairment on the part of the residents (Clark and Bowling, 1990). As an alternative, observational strategies derived from other areas of sociological and anthropological research have been developed in these settings to obtain more objective data on the process and outcome of residential care.What may be coded and analysed through observation of the care process is potentially unlimited, and researchers are obliged to identify and develop a number of specific categories of observable * Address for correspondence behaviour and interaction that relate to the ideas that they and service providers have about the elderly, their needs and the purposes of continuing care. In one early study of four types of care setting for the elderly using observational methods, Godlove et al. (1982) developed an event-sampling strategy in which 23 activities and contacts with seven classes of people were coded and the frequency of their occurrence recorded by observers at 10-second intervals over a 6-hour period. This rather detailed and time-consuming procedure has been simplified and validated as the Short Observation Method (SOM) by Macdonald et al. (1985), in which the coded activities are restricted to 10 and the observation period restricted to a representative 1 -hour period before lunch. Similar approaches to the coding of observational data have been developed by others interested in the activities and behaviour of elderly people in residential settings (McCormack and Whitehead, 1981;Simpson et al., 1981; Ward et af., 1992).Quantitative approaches to observational data-
