Deriving a Model for Predicting Hospital Falls

Background: In the United States 700,000 to 1,000,000 people fall in the hospital annually, 1/3 result in injury. One single fall averages $14,000, resulting in an increase in hospital length of stay and burden on hospital budget. In St. Joseph Hospital of Orange, from calendar year 2019 to 2020, there was an increase in falls from 178 to 185 falls, despite the use of a telesitter. At time of data collection, 12 telesitter cameras were initiated after a fall. An investigation was deemed necessary to determine the cause of the increase and the factors related to patient falls. Purpose: The purpose is to derive and validate predictors of falls by identifying criteria responsible for falls in a population of in-patients in an acute care setting. Compare research findings responsible for falls with current fall scales. Lastly, increase awareness with bedside nurses of patients most at risk for falls. Methods: The study utilized a retrospective cross-sectional design with a review of the electronic health records from calendar years of 2018 and 2019. Patients included are over the age of 18 and who were admitted to inpatient units in the hospital. A comprehensive literature review and comparison of current fall scales provided for identification of similarities, differences, and gaps among fall scales and identified common fall factors. Findings from the literature review were used to select variables for this study. The statistical methods and modeling used were descriptive statistics, continuous variables, categorical variables and bivariate analysis. Results: A total of 1,247 patient records, 929 records were randomized, while the other 318 records represented patients who fell during the hospital stay. Patient demographics shown to be statistically significant were age, gender, length of stay, and diagnosis. Identified patient behavior at most risk for falls are withdrawn, restless, anxious, and agitated. Lastly, if patient takes sedatives, anti-convulsants, anti-psychotics, and anticoagulants put a patient at risk for falls. Statistical analysis identified the factors posing the greatest risk. The strongest individual predictor was dizziness and vertigo; individuals were 7.2 times more likely to fall than those without dizziness/vertigo. Results also demonstrated a two-level “high” Morse Fall Risk with those with a 65 or greater score having double the risk of falling than those scoring 45-64. The fall predictor model derived from this study predicted 82% of the falls. This was especially significant when compared to the Morse Fall Scale which only predicted 62% of the falls. Conclusions: Results of the study will contribute to changes in policy and procedure on fall interventions for low, moderate, and high fall risk patients. Learning which variables are most likely to be present in a patient who could fall, can increase a bedside nurse awareness, and improve patient safety. Implications for practice: For future research, we would like to utilize the data and create a new model for predicting patient falls. Partner with other ministries to replicate study to see if results are similar. Incorporate the developed model to classify patient\u27s at risk for falls or early visual camera implementation