Predicting 12-Month Future Falls in People with Chronic Obstructive Pulmonary Disease

Abstract

Chronic Obstructive Pulmonary Disease (COPD) has been previously linked to falls. However, there are no established tools to accurately predict falls in this population. Predicting whether an individual with COPD is at an increased likelihood of falling can help clinicians initiate early fall prevention interventions and mitigate the risk of falling. The main objective of this dissertation was to identify predictors of 12-month future falls in people with COPD that are potentially feasible for use in clinical settings. The first manuscript was a prospective cohort study among community-dwelling older adults with COPD, evaluating the reliability and validity of four balance measures: the Brief Balance Evaluation Systems Test (BESTest), Single Leg Stance (SLS) test, Timed Up and Go (TUG) test, and TUG Dual-Task (TUG-DT) test. The results demonstrated that the four balance measures had excellent inter-rater and test-retest reliability but did not have evidence for predictive validity for predicting 12-month future falls in this population. Although these balance measures could not predict falls with sufficient accuracy, they may have a role in multifactorial fall risk assessment. The Brief BESTest and the SLS test may be better suited for identifying balance impairment, while the TUG and TUG-DT tests may be used to screen for mobility limitations. Overall, the results suggested that additional factors beyond balance should be considered to predict future falls in COPD. The second manuscript was a secondary analysis of prospectively collected data among individuals with COPD to develop and internally validate a clinical prediction model for 12-month future falls. A clinical prediction model with acceptable discrimination and calibration was identified. The model demonstrated that a higher likelihood of future falls can be predicted by a reported 12-month history of two or more falls, higher number of total chronic conditions, and worse TUG-DT test scores. The results highlight the need for external validation of the model to inform future application in different groups of people with COPD. The third manuscript was a secondary preliminary analysis of prospectively collected data among community-dwelling older adults with COPD to externally validate the clinical prediction model developed in the second manuscript. The clinical prediction model was externally validated and recalibrated, achieving acceptable calibration and discrimination. It was also demonstrated that the prediction model had superior clinical net benefit (i.e., more true positives and fewer false positives) when compared to screening for fall history alone, with decision thresholds set to 30-50% (the point at which the probability of an event is deemed actionable by a clinician). The results suggested that the clinical prediction model may be promising for detecting future falls in community-dwelling older adults and is superior to screening for fall history alone. However, the impact of the prediction model in clinical practice must be evaluated before it can be recommended outside of research settings. In conclusion, future falls in people with COPD are predicted by a 12-month history of two or more falls, a higher number of total chronic conditions, and worse mobility under cognitive demand as measured by the TUG-DT test. This work informs the future development of fall prevention guidelines for people with COPD.