Development of a clinical prediction model for falls in individuals with COPD

Chronic obstructive pulmonary disease (COPD) is linked to an increased risk of falls, however, there is no accurate method for predicting falls in this population. This study aimed to develop and internally validate a clinical prediction model for falls in individuals with COPD.

A secondary analysis was conducted using data from a recent fall prevention trial. Participants with COPD who reported a 12-month history of falls, concerns with balance or recent near falls were tracked for falls over 12 months prospectively. Baseline data included demographics and measures of balance, mobility and health status. A predictive model was developed using backward-selected multivariate logistic regression with fall status (no falls versus ≥1 fall) as the dependent variable and 17 baseline candidate predictors as independent variables. Using the bootstrap resampling method for internal validation, model performance was assessed for discrimination by the concordance (c) statistic and calibration by the expected to observed (E:O) ratio, calibration in the large (CITL) and calibration slope. The final model was adjusted for optimism using the bootstrap shrinkage factor.

Of 178 participants (mean age 73±9 years; 83 females), 74 (42%) reported ≥1 fall over 12 months, totalling 188 falls. The predictive model identified three factors associated with 12-month future falls: reporting a 12-month history of ≥2 falls (OR=3.59, CI (1.65 to 7.82)), more chronic conditions (OR=1.14, CI (1.01 to 1.28)) and worse Timed Up and Go Dual-Task test scores (OR=1.04, CI (1.00 to 1.09)). The final prediction model achieved acceptable discrimination (c-statistic=0.69, CI (0.61 to 0.78)) and calibration (E:O ratio=1.01, CITL=-0.01 and calibration slope=0.93).

A history of ≥2 falls, having more chronic conditions and impaired mobility under cognitive demand predicts future falls in individuals with COPD. The prediction model showed acceptable internal validation. External validation is needed to confirm these findings.

NCT02995681; clinicaltrials.gov.