Related Experiment Video
Updated: May 31, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Diagnosing delirium in elderly Thai patients: utilization of the CAM algorithm
Nahathai Wongpakaran1, Tinakon Wongpakaran, Putipong Bookamana
1Department of Psychiatry, Faculty of Medicine, Chiang Mai University, Thailand. nkuntawo@med.cmu.ac.th
Background:
Delirium is a common illness among elderly hospitalized patients. However, under-recognition of the condition by non-psychiatrically trained personnel is prevalent. This study investigated the performance of family physicians when detecting delirum in elderly hospitalized Thai patients using the Thai version of the Confusion Assessment Method (CAM) algorithm.
Methods:
A Thai version of the CAM algorithm was developed, and three experienced Thai family physicians were trained in its use. The diagnosis of delirium was also carried out by four fully qualified psychiatrists using DSM-IV TR criteria, which can be considered the gold standard. Sixty-six elderly patients were assessed with MMSE Thai 2002, in order to evaluate whether they had dementia upon admission. Within three days of admission, each patient was interviewed separately by a psychiatrist using DSM-IV TR, and a family physician using the Thai version of the CAM algorithm, with both sets of interviewers diagnosing for delirium.
Results:
The CAM algorithm tool, as used by family physicians, demonstrated a sensitivity of 91.9% and a specificity of 100.0%, with a PPV of 100.0% and an NPV of 90.6%. Interrater agreement between the family physicians and the psychiatrists was good (Cohen's Kappa = 0.91, p < 0.0001). The mean of the time the family physicians spent using CAM algorithm was significantly briefer than that of the psychiatrists using DSM-IV TR.
Conclusions:
Family physicians performed well when diagnosing delirium in elderly hospitalized Thai patients using the Thai version of the CAM algorithm, showing that this measurement tool is suitable for use by non-psychiatrically trained personnel, being short, quick, and easy to administer. However, proper training on use of the algorithm is required.
