Abstract Background People with Parkinson’s disease who meet criteria for mild cognitive impairment are at increased risk of dementia. It is not known which tests are more effective than others for identifying the risk of dementia. Methods At baseline, we assessed performance on 21 neuropsychological test measures spanning five cognitive domains in a prospective longitudinal study of 196 non-demented people with Parkinson’s. Elastic net logistic regression was used to identify a pair of tests from each cognitive domain that best predicted conversion to dementia over a four year period. The optimal tests most predictive of dementia were also determined when mild cognitive impairment was derived from a logistic-regression classifier that used all 21 measures simultaneously. Results With two tests per domain, the resulting mild cognitive impairment group (N=87/196) captured 44 of 51 individuals who converted to PDD; the out-of-sample relative risk of PDD was 8.0 (95% CI [4.3, 24]), similar to that achieved with the full battery (N=102/196, capturing 45/51, relative risk = 6.9). When selecting tests regardless of domain, there was strong evidence for three tests: Trail Making part B (Executive), Map Search (Attention), and CVLT-II word list acquisition (Episodic Memory). The logistic-regression classifier achieved an out-of-sample AUC of 0.90 [0.84, 0.96] and a relative risk of 12 [6, 39]. Conclusions An abbreviated selection of neuropsychological tests can identify non-demented patients who have a high relative risk of progression to PDD.

Cognitive tests that identify high risk of conversion to dementia in Parkinson’s disease