Purpose:
Falls are among the main disabling events for elderly adults and the identification of old people prone to falls enables the development of preventive and rehabilitative strategies. This study aimed to develop a simple tool, based on easily obtained variables (anthropometric measurements, motor performance tests and sociodemographic characteristics), to early identify community-dwelling old people prone to falls.

Methods:
The population-based household study was conducted among 316 elders (60 years old) of both sexes, living in the urban area of Lafaiete Coutinho in Brazil. History of falls in the previous 12 months (dependent variable), sociodemographic characteristics, anthropometric measurements and motor performance tests results (explanatory variables) were recorded, and a multivariate logistic regression was applied to identify the association between the explanatory variables and the history of falls. Fall probability for each elderly adult was calculated from the logistic regression parameters, and the predictive power of the final model and the cutoff for higher propensity to fall were evaluated on the basis of the receiver operating characteristic curve.

Results:
The prevalence of falls was 25.8% and the final model was influenced by the variables of sex (female) and poor performance in the balance test. The estimated probability model predicted approximately 66.5% (95% CI, 61–72%) of the falls. The sensitivity and specificity were 58 and 70%, respectively.

Conclusions:
We conclude that there is a high prevalence of falls among the studied elderly individuals, and the proposed method allowed to construct a simple tool for screening old adults prone to fall.