Adaptive cognitive testing in cerebrovascular disease and vascular dementia
Background/Aims: To examine whether brevity can be combined with precision in measuring global cognitive ability in patients with cerebrovascular disease (CVD) or vascular dementia (VaD). Longer tests (e.g. the CAMCOG) are precise but inefficient, whereas brief tests (e.g. the MMSE) are efficient but imprecise. Methods: A simulated computerized adaptive testing (CAT) algorithm using existing CAMCOG data from 284 patients with CVD of whom 55 were diagnosed with VaD. CAT was used to estimate each individual patient's total score on a large precise test (the CAMCOG). CAT repeatedly selected only items of appropriate difficulty, depending on whether the previous item was (in)correctly responded to. CAT estimates were compared with total scores on the whole CAMCOG. Results: Even though there was an average test reduction of more than 40%, CAT estimates were in very high agreement with the whole test results (intraclass correlation >0.97) and had similar accuracy for the diagnosis of dementia (area under the curve = 0.94). Conclusion: CAT combines efficiency with precision in the measurement of global cognitive ability in CVD patients.