Â鶹´«Ã½Ó³»­

Â鶹´«Ã½Ó³»­

Explore the latest content from across our publications
Join The Â鶹´«Ã½Ó³»­

Log In

Forgot Password?
Create New Account

Loading... please wait

Abstract Details

Data Mining applied to Poststroke Depression in Colombian Elderly Patients
Cerebrovascular Disease and Interventional Neurology
P15 - Poster Session 15 (12:00 PM-1:00 PM)
4-005

To determine the prevalence of depressive symptoms in elderly patients with a medical history of stroke and associated variables using a data mining technique.
Neurovascular disease is one of the leading causes of death and disability among adults worldwide. Particularly, 75% of strokes occur in patients older than 65 years and around one-third develop depressive symptoms after the event. Functional outcomes are compromised by poststroke depression. 
This is a secondary analysis of the Colombian SABE study which included 23.694 adults older than 60 years. Prevalence of depressive symptoms was calculated based on the Geriatric Depression Scale (GDS) developed by Yesavage and colleagues. Cognitive function and independence were assessed using the Abbreviated Mini-Mental Examination and Barthel Index. A data mining approach was implemented to create a decision tree highlighting those attributes that provide the highest information gain for the relative entropy of the system. 
In this sample, 1.080 subjects (4.57%) had a history of stroke. After applying the selection criteria, depressive symptoms were present in 37.68%. In these patients, higher frequencies of positive answers were found for items evaluating low mood (p=0.007), thoughts of death (p=0.044) and fatigue (p=0.000). The data mining analysis of 829 instances converged in the construction of an 8-leaf and 4-level tree defining presence of depressive symptoms as the class. 80.69% of instances were correctly classified and the highest information gain was obtained by cognitive functioning, communicative interest and independence in activities of daily living. Combination of these three variables increased the association with poststroke depressive symptoms (OR 14.92, CI95% 10.31-21.85, p<0.001).
Prevalence of depressive symptoms in Colombian elderly patients is similar to that previously reported in the literature. Decision trees obtained through data mining are useful representations to understand hierarchy and combination of associated variables with poststroke depression.
Authors/Disclosures
Gabriel D. Pinilla Monsalve, MD, MSc (Montreal University Geriatrics Institute)
PRESENTER 		Dr. Pinilla Monsalve has nothing to disclose.
No disclosure on file
No disclosure on file