On the Use of Count Model to Predict Falls in Community-Dwelling Elderly: Using Klosa (Korean Longitudinal Study of Ageing) Data

Purpose: Falls has been caused of increasing rates of morbidity or mortality in elderly population (Centers for Disease Control and Prevention, 2013). In community-dwelling adults aged 65 years or older, one in three in the US (Kannus et al., 2007) and 17.2% of South Korean (Korea Ministry of Health and Welfare, 2009) suffers a fall each year. It’s been led to spend of much of budgets from medical finance. And, the severity of complications of fall accidents has increased the length of stay of the old in hospital or care facility (Donaldson et al, 2005). Even though many of studies have been proposed the predictors or solutions for fall prevention, falls are still methodological issues in geriatric nursing field. This study suggests predicting the causes of increasing fall accidents in elderly by analyzing the national data with count model. The objectives of the study were: (1) to ascertain the risk factors for falls in community-dwelling elderly; (2) to determine whether risk factor profiles differ between first time fallers and recurrent fallers; and (3) to build decision tree map of fall down risks in elderly and to suggest effective interventions for first time fallers and recurrent fallers each.

Methods: Secondary data analysis was conducted on information collected Korean Longitudinal Study of Aging (KLoSA). KLoSA is a national panel data set that is publicly available. Data collected from 4,163 community-dwelling elderly in 2006 and 2008. Each subject was assessed by individual records of history and physical performance tests. Falls were recorded in frequency of fall for past 2 years. Data were analyzed by Chi-square, t-test, and zero-inflated negative binominal regression. Count models were estimated using STATA version 10.0 and regression tree with R program.

Results: The incidence of falls among community-dwelling elderly in Korea was 6.5%. Significant predictors of being a non-faller or a faller were vision, place of residence, pain and depression (p < .05). And significant predictors of being a recurrent faller were place of residence, alcohol and fear of falls (p < .05). For diagnosis and predicting regression trees of recurrent fallers were as follows (characteristics of high risk group): fear of falls (always worried), height (<175.5cm), age (<78.5year), vision (very bad), Quality of life (lower), and then height (<159.5cm).

Conclusion: These results provide new points of view of nursing implication for fall prevention of elderly in Korea. First, this study provides additional methodological option to study for prevention of falls and decrease the numbers of recurrent falls in elderly. This paper summarizes information to help guide the health care providers in choosing the high risk group of falls and the most appropriate preventing intervention. Second, this secondary-analysis provides comprehensive evidence-based assessment of risk factors for falls and recurrent falls in older people, confirming their multifactor etiology. Thus, the findings of this study will be the basis for effective intervention program to prevent falls and repeat falls.