Data from patient records will increasingly be used for care planning, quality assessment, research, health planning and allocation of resources. Knowledge about the accuracy of such secondary data, however, is limited and only a few studies have been conducted on the accuracy of nursing recording. The aim of this study was to analyse the concordance between the nursing documentation in nursing homes and descriptions of some specific problems of nurses and patients. Comparisons were made between wards where nurses had received training in structured recording based on the nursing process (study group) and wards where no intervention had taken place (reference group). Data were collected from the patient records of randomly selected nursing home residents (n=85). The methods used were audits of patient records and structured interviews with residents and nurses. The study revealed considerable deficiencies in the accuracy of the patient records when the records were compared with the reports from nurses and residents. The overall agreement between the interview data from nurses and from the patient records was low. Concordance was better in the study group as compared with the reference group in which the recorded data were structured only following chronological order. The study unequivocally demonstrates that there are major limitations in using records as a data source for the evaluation, planning and development of care.