Is the Tinetti Performance Oriented Mobility Assessment (POMA) a feasible and valid predictor of short-term fall risk in nursing home residents with dementia?

Background: The feasibility and predictive validity of balance and gait measures in more severe stages of dementia have been understudied. We evaluated the clinimetric properties of the Tinetti Performance Oriented Mobility Assessment (POMA) in nursing home residents with dementia with a specific objective of predicting falls in the short term. Methods: Seventy-five ambulatory nursing home residents with dementia, mean age 81 8 years, participated in a prospective cohort study. All participants underwent the full POMA-test. Fall statistics were retrieved from incident reports during a three-months follow-up period. The predictive validity was expressed in terms of sensitivity and specificity. Loglinear regression analysis was used to examine the relationship between POMA scores and the occurrence of a fall. Results: The POMA showed several feasibility problems, with 41%, of patients having problems in understanding one or more instructions. The inter-rater reliability of the instrument was good. The predictive validity was acceptable, with a sensitivity of 70-85% and a specificity of 51-61%, for the POMA and its subtests, and an area under the curve (AUC) of 0.70 for POMA-Total (95% Cl: 0.53-0.81), 0.67 for POMA-Balance (95% CI: 0.52-0.81), and 0.67 for POMA-Gait (95% CI: 0.53-0.81). After loglinear regression analysis, only POMA-T was significant in predicting a fall (adjusted HR = 1.08 per point lower; 95%) Cl 1.00-1.17). Conclusions: Application of the POMA in populations with moderate to severe dementia is hampered by feasibility problems. Its implementation in clinical practice cannot therefore be recommended, despite an acceptable predictive validity. To refine our findings, large prospective studies on the predictive validity of the POMA in populations with mild, moderate and severe dementia are needed. In addition, the performance of mobility assessment methods that are less dependent on cognition should be evaluated