FALLS AMONG MIDDLE-AGED VETERANS: STEPS TOWARDS PREVENTION

点击翻译按钮获取中文摘要

Background. Among middle-aged individuals (45-65 years), falls that occur in the community (community falls) are a leading cause of non-fatal injuries treated in hospital emergency departments and are responsible annually for the loss of 422,000 disability-adjusted life-years (DALYs). Intrinsic risk factors (risk factors inherent to the individual) likely contribute significantly to falls risk in this age group, but a consistently effective approach to outpatient fall prevention has not been realized within the VA. Objectives. The proposed project will explore community falls among middle-aged Veterans by characterizing prevalence incidence, sequelae, and risk factors for medically significant community falls among middle-aged Veterans (SA1). We will then develop a risk prediction tool to calculate the one year probability of a community fall (SA2). Long-term, we will develop a tool that will provide useful information to clinicians (RNs, APRNs, MDs, PAs) regarding falls risk and that will be easy to use. To this end, we will explore barriers and facilitators that clinicians experience when using clinical decision support tools, highlighting input from RNs and APRNs in the context of a multidisciplinary team (SA3). This project challenges the assumption held by most healthcare providers that community falls related to intrinsic risk factors are only a problem in older adults. We suggest that this is an important problem among middle-aged adults as well but that risk factors differ by age group, suggesting that interventions appropriate to older adults may not be effective among middle-aged. This project will provide the information necessary to develop falls prevention interventions for middle-aged Veterans. This project also uses an innovative approach to identify falls in the EHR: the use of machine learning to identify falls in radiology reports. Methods. We will use data obtained from the electronic health record (EHR) of Veterans ages 45-65 in the VA Birth Cohort. We have developed a machine learning algorithm that identifies community falls in radiology reports and will validate this algorithm in the VA Birth Cohort. We will develop a reference standard from a randomly selected subset of the radiology reports in this cohort that have been reviewed by a clinician and identified as addressing a fall or not. These results will be compared with those from the algorithm. We will first calculate rates of occurrence of community falls, rates of related injury, hospitalization and death, and the prevalence of related risk factors among middle-aged Veterans. Descriptive statistics (means, medians, frequencies, and standard deviations) will be used to characterize the distribution of risk factors and outcomes among the study participants. We will then develop a prediction tool for community falls in middle-aged Veterans. We will apply Bayesian Model Averaging which will identify a small group of risk factor models within a given range of the minimal value of the Bayesian Information Criterion. The final model will be an average of this small set of models. We will also assess facilitators and barriers to the successful implementation and use of clinical decision support tools by clinician-members of patient aligned care teams (RNs, APRNs, MDs, PAs).To maximize the utility of our falls prediction tool, we will interview all types of clinicians, with a particular focus on RNs and APRNs, to assess barriers and facilitators to clinical decision support implementation and use by potential end- users. The information from these interviews will inform future studies that address the development and implementation of the falls prediction tool as an important element of clinical care. Expected results. We anticipate that the machine learning algorithm will detect falls with a sensitivity >90%. We anticipate that falls risk factors identified in middle-aged Veterans will be different from those identified in older age groups, suggesting that falls prevention interventions will also differ. Screening efforts will need to take clinician preferences into account.