FALLS - AN EXAMINATION OF 3 REPORTING METHODS IN NURSING-HOMES

Objective: To examine the concordance of various fall reporting methods and to use the results to recommend a preferred method of ascertaining fall frequency for residents of nursing homes, both for research and in the collection of federally mandated nursing home data. Design: A cohort study followed for 858 patient months, with a mean individual follow-up of 6.6 months. Measurements: Falls were independently ascertained monthly by three methods: review of administrative incident reports, nursing home chart abstraction, and structured interview of subjects. Concordance of events was assessed using measures of simple agreement and Kendall's Tau-b. Simple correlation and multiple regression were used to evaluate the relation of age, sex, gender, depression, mental status, and functional status with degree of concordance between self-reported falls and chart-recorded falls. Setting: One academic and six community nursing homes in San Antonio, Texas. Participants: 131 long-stay nursing home residents, greater than 60 years of age, dependent in at least two activities of daily living, and mildly cognitively impaired. Results: Falls were ascertained in 74 of the 131 individuals; 53 subjects fell 124 times by incident report, 58 had 140 falls according to chart review, and 66 subjects self-reported 232 falls. Greatest agreement between reporting methods was shown for incident report and chart review, with a Kendall's Tau-b of 0.88; self-report and chart-review agreement was 0.56; and self-report and incident agreement was 0.53. Estimated total fall events were more often (P = 0.001) identified by chart review (92%) than incident report (82%). Although concordance was higher for non-fallers, no significant relationships were observed between concordance and age, sex, race, depression, mental status, and functional status. Also, there was no systematic relationship between length of follow-up and degree of concordance. Conclusions: Fall frequency varies by ascertainment method, with chart review reflecting a greater number of fall events than the traditionally counted incident reports.