Mental health disorders in older adults often go undetected because access to geriatric services is limited and caregivers lack suitable screening tools. This study designs a Certainty Factor (CF) model for the SiLansia early detection module, covering four hypotheses: depression, cognitive impairment/dementia, anxiety, and delirium. The knowledge base contains 15 symptoms derived from the Geriatric Depression Scale, AD8/INA-AD8, DSM-5/ICD-11, and WHO ICOPE. Expert CF values were established through a three-stage strategy: literature proxies, internal expert judgment, and planned external validation. Eleven IF-THEN rules combine evidence in parallel (AND operator) within each rule and sequentially (MYCIN formula) across rules. A Python prototype was tested on 29 synthetic cases, without human subjects. At a CF threshold of 0.6, precision was 1.00 for all hypotheses and micro-average accuracy was 0.92. Delirium reached perfect recall, but recall for depression, dementia, and anxiety was low (0.43–0.50) because of comorbid cases, single moderate symptoms, and the structure of the rule base. Lowering the threshold to 0.4 raised micro-average recall from 0.59 to 0.95, at the cost of one false positive. These results are preliminary, based on simulation only, and do not represent clinical performance. External expert validation and ethically approved testing with real caregivers and elderly participants are required.
Copyrights © 2026