ISQUA16-1327ESTIMATING THE EFFECTS OF FALL PREVENTION INTERVENTIONS BY ADJUSTING FOR THE RISK OF FALLING WITH A PROPENSITY SCORE

Abstract

Objectives: The hospital fall prevention programs that reduced fall rate included fall-risk assessments, multifactorial interventions, post fall follow-up, and integration with electronic records (Spelstra, 2012). However, patients who have received multifactor-ial interventions and patients who have not have different levels of fall-risk. This hampers comparison of patients who have received interventions and patients who have not and it precludes a determination of the effectiveness of individual interventions. The aim of this study was to examine the effects of fall prevention interventions by adjusting for the fall-risk with a propensity score. Methods: This study was conducted at an acute-care hospital with 716 beds. Electronic medical records at this hospital were linked to the Fall Risk Assessments Tool (Morita, 2010) and a standard plan to prevent falls (Oki, 2012). Subjects were admitted to wards or units other than Pediatrics, the NICU, and the ICU from April 2009 to March 2013. The Falls Risk Assessments Tool has a sensitivity of 74.5%, and specificity of 79.6%, and an area under the ROC curve of 0.822. Patients were assessed for 19 risk factors for falls. Standard interventions to prevent falls were listed along with 49 specific interventions to prevent falls on a screen displaying the patient's medical records. These interventions were based on meta-analyses, guidelines, and a root cause analysis at the hospital conducted by interdisciplinary team. Logistic regression models for risk of fall was applied to gain odds ratio (OR) and 95% confidence interval (CI) for 49 specific interventions to prevent fall, separately. To adjust background difference among those with and without a given intervention, a propensity score was entered to the model. This score was calculated individually using logistic regression model for each intervention and intervention-specific patients' background. Analysis was performed using the Stata ver. 13. Results: Subjects had a mean age of 66.5 years (SD: 17.4). Of the 61,949 total patients, 1,071 (1.72%) had fallen after their current admission. The odds ratio for 19 risk of falling were higher in "had previously fallen after admission" (OR 8.5, CI 7.0 to 10.4), "impaired balance" (OR 3.7, CI 3.3 to 4.2), "being restless"(OR 3.6, CI 2.9 to 4.4).The intervention that was most effective at preventing falls was to provide patients taking a narcotic, an antidepressant, a sleep aid, or an antihypertensive diuretic with "toileting assistance after taking a laxative or a diuretic (OR 0.43, CI 0.3 to 0.7 p<0.00)." This intervention was implemented for 1,198 of 13,582 patients, and its rate of implementation was low at 8.8%. Other interventions were "informing a physician of the efficacy of a patient's medication and adjusting that medication" (OR 0.56, rate of implementation 8.1%), "informing family members of the condition of agitated patients and asking that they accompany those patients" (OR 0.57, rate of implementation 35.8%), and "paying closer attention to patients who have transferred departments, wards, or units while performing nightly rounds" (OR 0.57, rate of implementation 38.6%). Conclusion: The results showed that three interventions to prevent falls in accordance with the patients' risks improved the fall ratio by 0.43-0.56 times. But results also revealed that such interventions are implemented at a low rate. Thus, interventions to prevent falls should be implemented at a higher rate.