A risk model for the prediction of recurrent falls in community-dwelling elderly: a prospective cohort study

Abstract

The object of this article was to determine the predictive value of risk factors for recurrent falls and the construction of a fall risk model as a contribution to a mobility assessment for the identification of community-dwelling elderly at risk for recurrent falling in general practice. The design was a prospective cohort study (n = 311). There were four primary health care centers. A sample stratified on previous falls, age, and gender of community-dwelling elderly persons aged 70 years or over (n = 311) was taken from the respondents to a mail questionnaire (n = 1660). They were visited at home to assess physical and mental health, balance and gait, mobility and strength. A 36-week follow-up with telephone calls every 6 weeks was conducted. Falls and fall injuries were measured. During follow-up 197 falls were reported by 33% of the participants: one fall by 17% and two or more falls by 16%. Injury due to a fall was reported by 45% of the fallers: 2% hip fractures, 4% other fractures, and 39% minor injuries. A fall risk model for the prediction of recurrent falls with an area under the curve (AUC) of 0.79, based on logistic regression analysis, showed that the main determinants for recurrent falls were: an abnormal postural sway (OR 3.9; 95% Cl 1.3-12.1), two or more falls in the previous year (OR 3.1; 95% Cl 1.5-6.7), low scores for hand grip strength (OR 3.1; 95% Cl 1.5-6.6), and a depressive state of mind (OR 2.2; 95% CI 1.1-4.5). To facilitate the use of the model for clinical practice, the model was converted to a "desk model" with three risk categories: low risk (0-1 predictor), moderate risk (two predictors), and high risk (> or =3 predictors). A fall risk model converted to a "desk model," consisting of the predictors postural sway, fall history, hand dynamometry, and depression, provides added value in the identification of community-dwelling elderly at risk for recurrent falling and facilitates the prediction of recurrent falls.