Interactive visualization to assist fall-risk assessment of community-dwelling elderly people. (January 2019)

• Record Type: Journal Article
• Authors: Sun, Tien-Lung; Huang, Chien-Hua
• Is Part Of: Information visualization. Volume 18:Number 1(2019)
• Page Start: 33
• Page End: 44
• Publication Date: 2019-01
• Subjects: Interactive visualization -- visual exploration -- fall-risk assessment -- elderly people; Information visualization -- Periodicals; 006.605
• DOI: 10.1177/1473871617721243
• Languages: English
• ISSNs: 1473-8716

Abstract

In fall-risk assessment, clinical experts have to provide accurate assessment of high-risk individuals using vast amounts of collected data. In this article, we propose an interactive visualization approach for clinical experts to improve their interpretation of assessment scores and facilitate the decision-making process. Fall-risk assessment data on a total of 356 community-dwelling elders were collated. The Short-Form Berg Balance Scale and 3-Meter Timed Up and Go test were used to screen elderly people with high fall risks. A series of interactive visualization techniques were conducted. After grouping by the literature and a statistical 5% outlier method, some disputed elderly people were examined through interactive visualization. Finally, receiver operating characteristic analysis was conducted using previous fall experience (faller or non-faller) and the three methods. Receiver operating characteristic analysis revealed that the area under the curve was the highest (0.87, 95% confidence interval: 0.80–0.94) for the interactive visualization process compared to the other methods (literature, 0.81 (95% confidence interval: 0.71–0.90); statistical 5% outlier, 0.80 (95% confidence interval: 0.70–0.90)). Through the interactive visualization approach, the clinical experts were able to determine the screening results and their relationship with the decision boundary more rapidly and accurately, demonstrating that this approach is useful for risk assessment in the medical domain.