Routine Blood Indicators Combined With Sleep Questionnaires for the Identification of Cognitive Impairment in Parkinson's Disease: A Cross-Sectional Study With a Leakage-Free Nested Modeling Framework.

backgroundCognitive impairment (CI) is a prevalent non-motor complication of Parkinson's disease (PD), yet accessible and reliable screening aids remain limited. This cross-sectional study aimed to quantify the strength of association and the cross-sectional classification value of routine blood indicators combined with standardized sleep questionnaires for prevalent CI in PD, using a strictly leakage-free model-development pipeline.

methodsA total of 347 PD patients were enrolled and classified into PD-CI (n = 143) and PD-NCI (n = 204) groups based on the Montreal Cognitive Assessment (MoCA < 26). The cohort was first divided into training (70%, n = 242) and test (30%, n = 105) sets by stratified sampling; all subsequent feature selection and tuning were confined to the training data. Within the training set, 20 candidate predictors were screened using L1-penalized logistic regression with the lambda.1se criterion, yielding 10 variables. Four nested logistic regression models (baseline clinical, baseline plus blood, baseline plus sleep, and full model) were constructed alongside random forest (RF) and gradient boosting machine (GBM) classifiers. Discrimination was evaluated by the area under the receiver operating characteristic curve (AUC) with DeLong confidence intervals and between-model DeLong tests, calibration (slope, intercept, and Brier score with bootstrap intervals) and decision curve analysis; the untouched test set was scored once, and generalization was estimated by repeated 5 × 2 cross-validation. Robustness of the outcome definition was examined with education-adjusted and alternative MoCA cut-offs.

resultsThe training-set LASSO retained 10 predictors: education, disease duration, Hoehn-Yahr stage, platelet-to-lymphocyte ratio (PLR), monocyte-to-lymphocyte ratio (MLR), homocysteine, C-reactive protein (CRP), Pittsburgh Sleep Quality Index (PSQI), Epworth Sleepiness Scale (ESS), and REM Sleep Behavior Disorder Screening Questionnaire (RBDSQ). In the untouched test set, the full logistic regression model achieved an AUC of 0.943 (95% CI: 0.896-0.990) in the test set, higher than baseline (0.844), and comparable to the baseline-plus-blood (0.937) and baseline-plus-sleep (0.889) models; the incremental gain over baseline was statistically significant (DeLong p = 0.009). Repeated cross-validation gave a more conservative AUC of 0.921 (SD = 0.029). RF (0.956) and gradient boosting (0.944) did not outperform logistic regression. Calibration was acceptable (slope 1.18, intercept 0.03; test Brier 0.086, 95% CI 0.057-0.120) and findings were stable across alternative MoCA cut-offs (AUC 0.909-0.931).

conclusionCombining routine blood indicators and sleep questionnaires is cross-sectionally associated with prevalent CI in PD and adds information beyond clinical variables alone. Because the design is cross-sectional, single-center, and only internally validated, the model should be regarded as a candidate screening aid that requires external, preferably multicenter and prospective, validation before any clinical use.