Predicting anxiety and depression in elderly patients using machine learning technology
| Autor: | Arkaprabha Sau, Ishita Bhakta |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2017 |
| Předmět: |
patient diagnosis
geriatrics learning (artificial intelligence) medical diagnostic computing feature selection medical disorders false positive rate random forest classifier health-related factors socio-demographic factors predictive model automated diagnosis patient load time constraint morbidities geriatric population mental health problems machine learning technology elderly patients depression prediction anxiety prediction Medical technology R855-855.5 |
| Zdroj: | Healthcare Technology Letters (2017) |
| Druh dokumentu: | article |
| ISSN: | 2053-3713 |
| DOI: | 10.1049/htl.2016.0096 |
| Popis: | Anxiety and depression are two important mental health problems among the geriatric population. They are often undiagnosed and directly or indirectly responsible for various morbidities. Early and timely diagnosis has immense effect on appropriate management of anxiety and depression along with its co-morbidities. Owing to time constraint and enormous patient load, especially in developing county such as India it is hardly possible for a physician or surgeon to identify a geriatric patient suffering from anxiety and depression using any psychometric analysis tool. So, it is of utmost importance to develop a predictive model for automated diagnosis of anxiety and depression among them. This Letter aims to develop an appropriate predictive model, to diagnose anxiety and depression among older patient from socio-demographic and health-related factors, using machine learning technology. Ten classifiers were evaluated with a data set of 510 geriatric patients and tested with ten-fold cross-validation method. Highest prediction accuracy of 89% was obtained with random forest (RF) classifier. This RF model was tested with another data set from separate 110 older patients for its external validity. Its predictive accuracy was found to be 91% and false positive (FP) rate was 10%, compared with gold standard tool. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|
Plný text