Cognitive structure and progression in Parkinson’s Disease: Insights from a tablet-based assessment

Cognitive structure and progression in Parkinson’s Disease: Insights from a tablet-based assessment | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Cognitive structure and progression in Parkinson’s Disease: Insights from a tablet-based assessment Tim Feige, Anika Frank, Jonas Bendig, Andrea Epler, Charlotte Harbarth, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7585530/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Cognitive impairment represents an important burden for patients with Parkinson’s disease (PwPD). Digital tools may improve accessibility and provide richer assessments than traditional paper–pencil tests; open-source generation can support the availability of comparable assessments in different cohorts. We therefore implemented a digital, tablet-based cognitive assessment (DiCo) comprising 13 commonly used tests as an open-source tool. 97 PwPD without overt cognitive impairment (43% women) completed the entire DiCo. Clustering of feature correlations and conditional dependencies indicated a predominantly mutual organization of cognitive performance in PwPD. Exploratory factor analysis identified five interrelated latent factors, most of which were derived from single tests. Factors correlated moderately with traditional neuropsychological tests and questionnaires. Machine learning identified working memory as the most predictive features of the MoCA. Latent profile analysis revealed four cognitive subgroups, mainly reflecting severity, with one group characterized by selective reflection impulsivity. Exploratory longitudinal analyses suggest partly independent trajectories of cognitive and motor progression, with a data-driven composite score detecting changes not captured by the MDS UPDRS III. Taken together, the DiCo demonstrated good feasibility in PD, and individual tests might be sufficient to substitute for MoCA or FAB in a research context. Health sciences/Health care Health sciences/Neurology Biological sciences/Neuroscience Biological sciences/Psychology Social science/Psychology Parkinson’s disease cognition digital open source Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction Parkinson’s Disease (PD) is the second most common neurodegenerative disease and the fastest growing neurological disorder in the world with regard to prevalence, disability and deaths 1 , 2 . PD is defined by its cardinal motor symptoms resting tremor, rigidity and bradykinesia. In addition, PD encompasses a multitude of non-motor symptoms (NMS). People with PD (PwPD) have a 2.8 to 6-fold higher risk of developing dementia than the general population 3 . Several longitudinal studies have demonstrated that almost all PwPD develop dementia if they live with PD for more than ten years 4 . Hence, cognitive impairment is one of the most common NMS and arguably the NMS that PwPD fear most 5 . Cognitive impairment in PD can be classified based on severity and based on the affected functional systems. The following severity levels are currently delineated: normal cognition (NC), Subjective Cognitive Decline (SCD), Mild Cognitive Impairment (MCI), and Parkinson's Disease Dementia (PDD) 6 – 8 . In addition, the following functional domains are distinguished by the Movement Disorder Society: attention/working memory, executive functions, language, memory, and visuospatial functions 9 . The timing, profile, and rate of cognitive decline can be highly variable in PD 10 . The observation of distinct functional impairment profiles in MCI has led to the hypothesis of the "dual syndrome," suggesting two cognitive subtypes of PD-MCI. These are based on the impairment of distinct anatomical regions: the frontal syndrome, associated with executive and attentional deficits, and the posterior cortical syndrome, associated with visual-spatial and memory deficits 11 . Digital assessments build on a long tradition of psychophysical research in PD and potentially offer the opportunity to collect features of cognitive function beyond what is captured in classical paper-pencil-based neuropsychological testing (NPT). Hence, digital assessments can potentially reveal novel patterns and relationships 12 . In addition, digital testing methods have been suggested to allow for low-threshold longitudinal measurements with high detail, applicable with minimal effort in a large population 13 . Digital assessments should therefore provide at least the information obtained by commonly used screening instruments such as the Montreal Cognitive Assessment (MoCA) 14 and the Frontal Assessment Battery (FAB) 15 . Among digital methods, open-source tools support the use of comparable tests in different cohorts as a basis for the generation of poolable datasets for data-driven investigation into cognitive profiles associated with neurodegenerative disorders and the modeling of disease progression trajectories. We therefore intended to design a digital, tablet-based cognitive assessment (DiCo) that is easily applicable in clinical practice, expanding on readily available, open-source systems and frameworks such as the Experiment Factory (EF) 16 and jsPsych 17 . The Selection of tests was based on a systematic review of the literature. Only established assessments or adaptations thereof were included in the implementation. Extensive usability testing with PwPD and PD experts was performed to ensure applicability in research and practice. In order to determine the type of information acquired in the DiCo, conditional dependencies were visualized and an exploratory factor analysis was conducted. To determine to what extent data captured by the DiCo can reproduce the information contained in MoCA and FAB, Random Forest Regressors were trained. Finally, to address the question of heterogeneity, a Latent Profile Analysis was performed, finding four clusters of participants. Three clusters mainly differed in severity, whereas the remaining cluster was characterized by predominant reflection impulsivity. Results Study participants After usability testing, the tablet-based digital assessment was administered to a cohort of n = 100 participants (43% women) of whom 97 completed the entire test battery. The three dropouts had different reasons: Two patients had to terminate the study earlier because of scheduled follow-up medical appointments and one felt too exhausted to continue the testing. Demographic and clinical data of all participants are displayed in Table 1 . With an average disease duration of 6.4 years and a mean LEDD of 672 mg, our sample is beyond the typical ‘honeymoon period’ of Parkinson’s disease, which is usually defined as the first 2–5 years of stable response to levodopa 18 . The average score in the Montreal Cognitive Assessment (MoCA) was 26 points. Completing the entire assessment took 67 minutes on average, the interquartile range was 59-71 minutes. From each test, several different features were extracted, including mean and standard deviation of the reaction time, percentage of correct answers and others. In total, 54 features were extracted from the 13 tests. A subsample of n = 50 participants additionally took part in a paper-pencil neuropsychological assessment (NPT) that was mainly comprised of the CERAD+ 19 . Questionnaire (Q-Data) were available for n = 76 participants. An overview of all NPT and Q-Data variables is provided in Table 3 . Another subset of n = 28 participants was available for a follow-up assessment. The mean interval between the two assessments was 25 months. Demographics of this subset are also included in Table 1 . Table 1: Demographic and Clinical Characteristics Feature (at baseline) Entire cohort (n=97) Follow-up (n=28) Gender, % male 1) 55 (57%) 13 (46 %) Age (years) 2) 63 (10.1) 62 (9.8) Disease Duration (years) 2) 6 (4.7) 6.4 (3.9) Age at onset (years) 2) 56 (10.5) 56 (10.6) LEDD Total (mg) 2) 672 (338.8) 665 (355.6) LEDD LDopa (mg) 2) 418 (197.7) 365 (189.3) LEDD DA (mg) 2) 244 (141.6) 269 (126.8) Neuroleptics 1) 6 (6%) 1 (3.5 %) Hallucinations 1) 5 (5%) 0 Vascular Risk present 1) 34 (35%) 7 (25 %) UPDRS III 2) 19 (9.0) 17.4 (7.5) MoCA 2) 26 (2.8) 27.8 (1.8) FAB 2) 16.5 (1.9) 17.4 (.8) SCD 1) 53 (54.7%) 17 (60 %) 1) N and %; 2) Mean and SD Abbreviations: LEDD - Levodopa Equivalent Daily Dose; DA – Dopamine Agonist; MoCA - Montreal Cognitive Assessment, FAB - Frontal Assessment Battery, SCD - Subjective Cognitive Decline Usability testing confirms the feasibility of the DiCo in an elderly population A tablet-based cognitive assessment (DiCo) was developed by implementing 13 commonly used test paradigms that cover response inhibition, cognitive flexibility, attention, and working memory ( Table 3 ). To make sure that the DiCo can be used efficiently in an elderly population, we placed particular emphasis on a user-friendly experience and on the clarity of test instructions. Training sessions were implemented to acquaint participants with the use of the tablet computer, in particular with the functionality of the touch input, and practice trials were incorporated into the assessments to provide participants with an opportunity to familiarize themselves with the testing procedures before the actual assessments. Comprehensive usability testing was conducted with a mixed-methods approach that incorporated quantitative measures and heuristic evaluations, which are described in more detail in the supplement. Quantitative measures were obtained through the System Usability Scale (SUS) 20 and the User Experience Questionnaire (UEQ) 21 . The SUS is a widely used questionnaire that assesses the overall usability of a system, generating a numeric score that reflects user satisfaction. SUS ratings revealed a high overall satisfaction with the system, while some participants expressed the need for support and some time to get familiar with the use ( Fig. 1c ). The UEQ, on the other hand, provides an evaluation of the user experience across multiple dimensions, including attractiveness, perspicuity, efficiency, and stimulation. The cognitive testing battery achieved high scores in the UEQ domains of attractiveness (1.88 ± 0.81), stimulation (2.00 ± 0.84), and novelty (1.80 ± 0.98), indicative of a high “hedonic quality” ( Fig. 1b ). The rating for the pragmatic quality was above average (indicated by perspicuity, efficiency and dependability, mean 1.36). DiCo captures performance-based dimensions that are stable across analytic approaches To determine whether the extracted DiCo features reflect cognitive domains, task-specific variance, or broader performance dimensions, we conducted three complementary analyses: hierarchical cluster analysis, network modeling and exploratory factor analysis (EFA). The hierarchical cluster map ( Fig. 2 ) was based on pairwise feature correlations and revealed three interpretable clusters (C) of features: time-based measures such as response time (C1), accuracy-related measures (C2), and a residual cluster mainly comprised of features from BART, Information Sampling and Reversal Learning (C3). Correlation Coefficients were stronger within C1 and C2 than within C3, indicating less internal coherence within the third cluster. Because good performance is associated with short response times and high accuracy, we inverted the time-based measures in C1 to align higher values with better performance. After this transformation, the cluster structure changed and yielded two main clusters ( supplemental Fig. 1 ): one combining both response time and accuracy measures across cognitive tests, and another residual cluster similar to the original C3. For the network modeling, we calculated the sparse inverse covariance matrix (precision matrix) using the graphical Lasso algorithm. The precision matrix identifies conditional dependencies between individual variables while controlling for the influence of all others. The resulting network structure showed strong within-test connectivity and sparse connections between tests ( Fig. 3a ). Among the connections within the same test, 66.7% (n=14 of 21) were realized, whereas only 10.1% (n=36 of 357) of connections across tests were realized. In Fig. 3b , we color-coded the nodes in the network graph according to the three variable clusters derived from the hierarchical cluster map ( Fig. 2 ). The cluster assignments aligned remarkably well with the structure of the network that was derived from the precision matrix. Particularly, time-based features (C1) formed a densely interconnected core, whereas impulsivity-related measures (C3) appeared largely isolated from the rest of the network. This convergence supports the interpretability and structural validity of the initial cluster solution and suggests that the DiCo captures performance-based dimensions that are stable across analytic approaches. Hierarchical clustering captures bivariate associations based on overall similarity, while network analysis estimates conditional dependencies between variables. Both approaches describe the observed variable structure at the surface level — in terms of correlation or partial correlation — without modeling the underlying sources of shared variance. To examine whether the DiCo features coalesce into interpretable latent dimensions that could be analogous to the concept of cognitive domains, we conducted an exploratory factor analysis (EFA). EFA explicitly aims to identify latent dimensions that account for the covariance structure, thereby allowing for a theoretically meaningful interpretation of shared cognitive processes. After excluding features with low sampling adequacy (Kaiser-Meyer-Olkin Test, KMO < .50), low communalities (< .40) and eigenvalues < 1.0 22 , the EFA yielded five factors ( Fig. 5a ). These factors explained 64.6% of the total variance, exceeding the commonly used threshold of 60% in social sciences 23 . The first factor consisted of features from StopSignal, Go-Nogo and the Flanker task, specifically those derived from the “Go” conditions that did not require inhibition or discrimination. Since those trials primarily capture aspects of attention, we chose to name this factor “visual attention”. The remaining four factors mostly consisted of variables from the same test and were thus named after the cognitive domain these tests have been associated with: Stroop – Interference control , Tower of London – Problem solving , Information Sampling – Decision-Making , and N-Back – Working Memory . We refer to them as DiCo factors from here on. We then tested for correlations between DiCo factors. The correlation of Decision Making with the other factors was small ( Fig. 4b ), but the remaining four factors were moderately to strongly interrelated with correlation coefficients ranging from r = 0.51 to 0.62. This high degree of interrelatedness suggests a considerable degree of shared variance across cognitive domains. Construct validity of DiCo factors To evaluate the construct validity of the identified DiCo factors, we calculated Pearson correlation coefficients with summary scores obtained from neuropsychological tests (NPT) and validated questionnaires (Q) ( Fig. 5 ). We only observed moderate associations; the strongest correlation coefficients with a DiCo-derived factor ranged between r = 0.30 and r = 0.64. Due to the coding schemes of the individual variables ( Table 3) , NPT measures tended to correlate positively with DiCo factors whereas several questionnaire-derived variables were negatively associated. The DiCo factor Visual Attention was moderately associated with the following variables obtained by NPT ( Fig. 5a ): finding similarities (WIE; r = .36), executive functions (CERAD TMT-B, r = -.34) and semantic verbal fluency (CERAD ANIM, r =.33). Interference Control was most strongly correlated with executive functioning (CERAD TMT-B, r = -.64). Problem-Solving was related to semantic verbal fluency (CERAD ANIM, r = .48) and executive functioning (CERAD TMT-B, r = -.52). Working Memory was most strongly related to semantic verbal fluency (CERAD ANIM, r = .43). Decision Making showed only weak associations with the NPT variables, except for a moderate correlation with vocabulary knowledge (MWT-B, r = .37). This latter association may reflect a response style where participants prematurely accepted options as valid without careful evaluation potentially mirroring increased reflection impulsivity. For the questionnaire data ( Fig. 5b ), the DiCo factor Visual Attention showed moderate negative relationships with conscientiousness (BFI Conscientiousness, r = -.47), functional abilities in everyday life (FAQ total, r = -.45) and lack of perseverance (UPPS Lack Persev., r = -.44). Interference Control was weakly to moderately related to patients’ global impression of condition (PGI Severity, r = -.36), fatique (Fatigue FSS, r = -.34) and gambling (ICDRD Gambling, r = -.33). Problem-Solving was most strongly related to the participants subjective global impression of condition (PGI Severity, r = -.46) and moderately related to lack of perseverance (UPPS Lack Persev., r = -.38). Working Memory overall also showed weak correlations to Q-Data variables with urgency (UPPS Urgency, r = -.34) and lack of perseverance (UPPS Lack Persev., r = -.28) being the most prominent. Decision-Making again showed only weak correlations to questionnaire variables with urgency (UPPS Urgency, r = -.28) being the strongest. Table 2 Comparison of clinical variables between clusters Feature Cluster 1 (N=13) Cluster 2 (N=53) Cluster 3 (N=21) Cluster 4 (N=10) Kruskal-Wallis Mann-Whitney U Mann-Whitney U (Holm-corrected) Age (years) 60.7 (9.0) 60.7 (10.1) 66.2 (9.6) 69.5 (9.4) p= .013 1,2 < 3,4 n.s. Disease Duration (years) 6.7 (3.3) 5.8 (4.6) 6.7 (5.3) 8.1 (5.1) p= .403 n.s. n.s. Age at onset (years) 54 (8.9) 54.6 (10.1) 59.5 (10.6) 61.4 (12.33) p = .039 2 < 3 n.s LEDD Total (mg) 685.6 (283.5) 627.3 (355.4) 721.8 (345.6) 789.6 (298.5) p = .272 n.s. n.s. LEDD LDopa (mg) 420.5 (168.8) 355.0 (177.0) 496.0 (223.3) 603.6 (105.5) p < .001 2 < 3; 1,2 < 4 2 3 2 > 3 Vascular Risk present (%) 23.1 37.7 28.6 50 p =.449 n.s. n.s. UPDRS III 18.8 (9.1) 17.4 (8.7) 21.3 (7.6) 25.9 (10.8) p = .047 2 4 2 > 4 FAB 16.8 (1.4) 17.0 (1.3) 16.6 (1.5) 13.5 (3.4) p = .004 1 ,2,3 > 4 1,2,3 > 4 SCD (%) 38.5 47.2 71.4 80.0 p = .058 n.s. n.s. Note. Kruskal–Wallis test was used to assess group differences across the four clusters, as the data were not normally distributed and group sizes were unequal. Given the small sample sizes in some clusters, we chose pairwise Mann–Whitney U tests for post-hoc comparisons, as this approach is less conservative and more robust than Dunn’s test in small samples. For each comparison, both the uncorrected and Holm-corrected differences are reported. Abbreviations : n.s. – not significant, LEDD - Levodopa Equivalent Daily Dose; DA – Dopamine Agonist; MoCA - Montreal Cognitive Assessment, FAB - Frontal Assessment Battery, SCD - Subjective Cognitive Decline Longitudinal changes primarily affect composite motor score, Working Memory, Problem-Solving and Interference Control Longitudinal assessments were available for a subset of our cohort. The mean delay between the two assessments was 25 months. To compare cognitive changes across domains and with respect to motor changes, we computed Standardized Response Means (SRMs), defined as the mean change divided by the standard deviation of the change scores ( Fig. 7a ). Next to SRMs for each DiCo factor, we used the MoCA score as a commonly used cognitive assessment and the motor score UPDRS III. Because the UPDRS III can be affected by dopaminergic medication, we determined, in addition, a compound motor score using Uniform Manifold Approximation and Projection (UMAP). The UMAP-derived motor score exhibited a moderate effect size for change (SRM = .57). The MoCA (SRM = -.36) and Working Memory (SRM = -.32) showed a small to moderate effect size whereas Problem-Solving (SRM = -.19), UPDRS III (SRM = .16) and Interference Control (SRM = -.11) yielded only very small effect sizes. The observation that Problem Solving showed the most pronounced changes with age ( Fig. 6d ) but relatively small longitudinal changes ( Fig. 7e ) is consistent with the notion that longitudinal trajectories in patients are not only driven by age. Accordingly, the compound motor score showed only a weak correlation with age (r = 0.27; supplemental Fig. 2 ), indicating that it reflects disease duration and progression more strongly than chronological age. Accordingly, individual trajectories showed stronger changes for the compound motor score ( Fig. 7h ) than for MoCA ( Fig. 7c ) and Working Memory ( Fig. 7d ). Paired t-tests showed significant differences between baseline and follow-up values for the compound motor score (t = -2.857, p= .008), but not for UPDRS III (t = -.844, p = .405) or MoCA (t=1.424, p=.165). Associations of digital cognitive assessments with MoCA and FAB To investigate the relationship of factors obtained from the DiCo with overall cognition and executive function, we trained two separate Random Forest Regressors to predict the scores of the Montreal Cognitive Assessment (MoCA, Model 1) and the Frontal Assessment Battery (FAB, Model 2). Factor scores, sociodemographic and clinical variables were used as predictor variables. Hyperparameter tuning was achieved via grid search and 5-fold cross-validation. Model performance and feature importances were evaluated across multiple random seeds using repeated cross-validation (5 folds, 2 repeats), resulting in a robust estimation across 150 model runs. Mean absolute Error (MAE) and Root Mean Squared Error (RMSE) were computed to evaluate prediction performance. To identify the most relevant predictors for MoCA and FAB, our main interest was in the averaged feature importances and their standard deviation across all folds and random seeds. For Model 1, grid search revealed the following best hyperparameters: n_estimators = 50, min_samples_split=2,min_sample_leaf=4, max_depth=10. Model 1 predicted the MoCA scores with a MAE of 1.87 (SD = 0.31) and a RMSE of 2.21 (SD = 0.39). These values indicate that, on average, the model's predictions are reasonably close to the actual scores. Working Memory was, by far, the most important factor for predicting the MoCA score ( Fig. 8a ). Grid search yielded n_estimators = 100, min_samples_split = 5, min_samples_leaf = 2 and max_depth = 5 as optimal hyperparameters for predicting the FAB. Model 2 predicted the FAB ( Fig. 8b ) scores with a MAE of 1.05 (SD = 0.31) and a RMSE of 1.58 (SD = 0.56). In contrast to Model 1, no single feature stood out for predicting the FAB. Total levodopa equivalent dose and Interference Control were almost equally important for predicting the FAB, followed by Problem-Solving and Disease Duration . The more evenly distributed feature importance could be explained, for instance, by collinearity among predictors. We then trained a second set of random forest regressors to predict the change in MoCA and the change in the compound motor score. The MoCA and compound motor scores at baseline were the most important features to predict the change in MoCA, followed by the time between the two assessments and UPDRS III at baseline ( Fig. 9a ). To predict the change in Motor Score, days between the assessments, motor score at baseline and Decision Making were the most important features ( Fig. 9b ). Interestingly, baseline Decision-Making was associated negatively with change in Motor Score (r = -.31, p = .11; Fig. 9d), indicating that higher reflection impulsivity may predict stronger motor decline over time. Given the small number of participants (n = 28) in the longitudinal cohort, these results need to be interpreted with caution. After controlling for baseline values, the residualized MoCA and motor change scores only exhibited a weak correlation (r = -.28, p = .15; Fig. 9c ), suggesting that the progression of motor and cognitive performance over time may develop independently from one another. Discussion In this study, we developed and validated an open-source tablet-based cognitive assessment battery for PwPD. 54 features were obtained in 13 test paradigms. Three structural analyses were conducted, which converged on the notion that cognitive performance of PwPD without manifest cognitive impairment, as captured by the DiCo, is characterized by strong intercorrelations across domains. Specifically, the main difference between clusters C1 and C2 obtained by correlation-based clustering ( Fig. 2 ) consisted in the direction of changes, and inverting the time-based measures resulted in one cluster combining both response time and accuracy measures across cognitive tests, and another residual cluster similar to the original C3 ( supplemental Figure 1 ). In the network analysis, features clustered within the same test, but not along other domains ( Fig. 3a ). The EFA revealed five latent factors, four of which were strongly interrelated ( Fig. 4b ). These factors also correlated strongly with neuropsychological test results ( Fig. 5a ) and questionnaire data ( Fig. 5b ). Clustering of patients showed homogenous differences between 3 of 4 clusters ( Fig. 6a ), which can be classified as severity-based using the nomenclature by Pourzinal et al. 24 and which are consistent with a gradual decline across all cognitive functions captured by the DiCo. These findings are consistent with a mutualistic, network-like structure of cognitive abilities as suggested by recent theoretical developments in cognitive science 25 . The age-dependent decline of the DiCo factors Attention, Working Memory, Interference Control and Problem-Solving ( Fig. 6 , x-axis) is consistent with the hypothesis that cognitive abilities are interdependent and decline jointly. Correlations between cognitive tasks arise from reciprocal interactions between individual cognitive processes, and cognitive abilities reinforce one another throughout development and aging. The quantitative differences between patient clusters 1, 3 and 4 ( Fig. 6 , color coding) indicate that, in PwPD, cognitive performance declines along the same axes. These results are in line with analyses of cognitive test scores obtained in 698 PwPD from the DEMPARK and LANDSCAPE cohorts 26 . This rather uni-dimensional structure of cognitive test results does not support traditional neuropsychological models, which have conceptualized cognitive abilities as separable domains, including attention, working memory, executive functions, language, memory and visuospatial abilities. This modular view of cognition underlies diagnostic frameworks like the MDS criteria for MCI in PwPD, which require deficits in at least two of these distinct domains 9 . The DiCo factors correlated moderately with established paper-pencil neuropsychological tests ( Fig. 5a ). Given the substantial methodological differences, this correlation strength is not unexpected. Previous studies comparing paper-pencil versions of a neuropsychological test with its exact digital implementation found correlations of 0.63 for TMT-A, 0.77 for TMT-B, and 0.68 for the Color-Word Interference Test 27 . Assessing the relationship between the processing speed subtests Coding and Symbol Search from the Wechsler Intelligence Scale for Children, Ferriola 28 found correlations of 0.67 and 0.61 respectively. In addition, the lower correlation coefficients we observed can be explained by the fact that the DiCo tests are not exact implementations of the neuropsychological tests we related them to. Finally, the DiCo factor Visual Attention represents a composite of features from three different tests. The apparent lack of domain structure within the DiCo features and the surprisingly poor association with paper-pencil tests and questionnaires is reminiscent of a comparison of digital assessments of self-regulation described by Eisenberg et al 29 . The associations we did observe between DiCo factors and paper-pencil tests are theoretically plausible. For example, Interference Control and the TMT-B have both been linked to executive functioning 30,31 . Similarly, Problem-Solving and verbal fluency have both been linked to microstructural white matter changes in newly diagnosed PwPD, suggesting a shared neuroanatomical basis 32 . For questionnaires, most associations to DiCo factors were even more moderate ( Fig. 5b ). Interestingly, the correlation between Problem-Solving and participants’ subjective impression of disease severity was rather strong, which underlines previous findings that cognitive symptoms play an important role in PwPDs’ quality of life 33–35 . The DiCo factor Decision-Making consistently emerged as rather independent, being less correlated with the other factors ( Fig. 4b ), less correlated with neuropsychological test results and questionnaire data ( Fig. 5 ). Decision-Making did not decline with age ( Fig. 6e ) and separated the two largest clusters of patients ( Fig. 6a ): cluster 2 showed similar performance as cluster 1 in most DiCo factors but was significantly impaired in Decision-Making . The DiCo factor Decision-Making contributed little to the prediction of MoCA and FAB ( Fig. 7 ). Decision-Making was mainly driven by features of the information sampling test ( Fig. 4a ). The information sampling test responds to reflection impulsivity, i.e., the tendency to make rapid decisions without sufficient information gathering or consideration of alternatives. Dopamine agonists increase reflection impulsivity 36 and all dopaminergic treatment can cause impulse-control disorders 37–39 . This raises the possibility that the Decision-Making feature results not from aging or PD, but from dopaminergic medication. Interestingly, Decision-Making was associated with changes in the composite motor score ( Fig. 10b ), which includes medication information next to the UPDRS III ( supplemental Figure 2 ). In addition, we observed a trend where a higher proportion of dopamine agonists in the medication profile was associated with poorer performance on the information sampling task – particularly among participants who performed well on other cognitive tests (cluster 2 vs. cluster 1, Table 2 ). In this study, cluster assignment based on DiCo-derived factor scores revealed four subgroups with distinct performance profiles ( Fig. 6a ). These four clusters were poorly explained by age, disease duration, age at disease onset, presence of cardiovascular risk factors, and motor impairment ( Table 2 ). Also, the classical tests MoCA and FAB did not differ significantly between clusters 1, 2 and 3 ( Table 2 ). The digital assessment implemented in this study can therefore identify differences that are not observed with these measurements. This might result from the fact that the DiCo measures slightly different properties than paper-pencil tests ( Fig. 5 ). In addition, digital assessments might provide finer granularity. In this context, the more or less uni-dimensional structure of the DiCo results implies that the DiCo composition could be re-evaluated to reduce redundancies. In the subset of patients where longitudinal data was available, we observed changes between baseline and follow-up in three DiCo factors. Working Memory exhibited a standardized response mean (SRM) comparable to the MoCA, suggesting that the N-Back task, which mainly loaded on Working Memory ( Fig. 4a ) might serve as a simple research tool for longitudinal research. The very low correlation between residual changes in MoCA and composite motor score ( Fig. 10c ) underscores the importance of examining cognitive and motor progression as partly independent trajectories in PwPD. These effects were not observed when using the UPDRS III score to predict motor progression. In fact, we did not observe a significant difference in UPDRS III scores between baseline and follow-up ( Fig. 8d ). Similarly, Holden et al. found UPDRS III to increase 4 points per year in unmedicated PD Patients whereas patients on medication progressed only 1.2 points per year. Dopaminergic medication can therefore mask motor progression and reduce the sensitivity of UPDRS III to subtle longitudinal changes. The significant differences observed between baseline and follow-up in the UMAP-based composite score suggest that this approach may offer greater sensitivity to progressive motor changes in PwPD than clinical ratings alone. Our sample of 97 patients primarily included participants in the transition phase of PD without overt cognitive impairment. As for other cohorts in early PD 40 , this potentially limited the detection of dementia-related patterns, but allowed for an analysis of subtle cognitive differences that may already be present in early stages of the disease. For longitudinal assessments in particular, the number of patients included in the study was relatively small and analyses should be interpreted with caution. Because participants were recruited in an outpatient clinic, PwPD with impaired mobility were underrepresented in our study, limiting the generalizability of the feasibility assessments. Moreover, the cohort is quite homogenous with respect to ethnic disparities. Multicenter validation studies are therefore required, and the availability of the app on github along with the test implementations will facilitate such confirmation. In summary, this study describes a comprehensive digital cognitive assessment that has been extensively usability-tested and can be readily applied for research on cognitive impairment in PwPD. Researchers with specific interests may choose specific tasks identified to correlate with specific cognitive domains based on the associations described here. In addition, our analyses suggest that features extracted from digital assessments can contain information not included in paper-pencil based test. The prognostic value of this data needs to be confirmed in longitudinal studies. Methods Study population and design The study was approved by the institutional review board of the Technische Universität Dresden (BO-EK-494112020) and written informed consent was obtained from all participants. Participants were recruited at the Department of Neurology of the Dresden University Hospital. In this observational exploratory study, a sample size of 100 cases was targeted. To ensure a diverse study population, the inclusion criteria for our study were deliberately broad, encompassing individuals aged 18 and above who met the diagnostic criteria for PD as stipulated by the Movement Disorder Society 41 . Individuals with severe cognitive or psychiatric disorders that could impede the successful completion of the cognitive test battery were excluded. Clinical and demographic data were collected, including motor and cognitive status (Hoehn & Yahr stage, UPDRS part III, Montreal cognitive assessment) as well as the presence of disease-related complications. Selection and implementation of tests The selection of digital cognitive tests for our study was informed by a literature review centered on cognition in PD. We systematically examined research findings and consensus-based recommendations and identified specific cognitive domains implicated in PD pathology, with a particular focus on response inhibition, cognitive flexibility, attention, and working memory 42 . A data-driven identification of cognitive subtypes identified two types of clustering models: severity-based and domain-based. The majority of severity-based models revealed three distinct clusters of PwPD. The domain-based models identified two to six clusters with mnestic, visuospatial, executive and attentional impairment being the most common subtypes 24 . The 13 selected tests were implemented using the open-source JavaScript framework jsPsych, optimized for touch-input and deployed via an Experiment Factory Container. To ensure a distraction-free and uniform testing environment, an iOS client application was developed. The app as well as the test implementations are available on github: https://github.com/tfeige91/DiCo-iOS-. All participants conducted the test on an 8th-generation Apple iPad at maximum display brightness. The app was designed to have participants undergo a brief instructional session to familiarize them with the touch controls prior to the actual experiments. During the instructional session, participants were required to accurately tap a circular area displayed on the screen. Each successful tap was met with immediate feedback, designed to reinforce correct interactions and improve proficiency of the touch control system. This preliminary exercise ensured that all patients were comfortable and adept at using the touch interface, allowing for a consistent starting point in the subsequent experimental tasks. All participants conducted the tests in a strategically designed order (Table 1) optimized for participant engagement and balanced for cognitive strain. Demanding tests that required high levels of concentration and attention were interleaved by shorter, less strenuous tasks that in the usability testing were perceived as more enjoyable or attractive. This approach was intended to mitigate fatigue and maintain participant motivation throughout the testing session. Additionally, simpler tasks where intense attention and quick reaction times were not the primary focus (e.g. the Information Sampling Task) were included in the second half of the test series. Usability Testing Usability testing employed a mixed-methods approach that incorporated heuristic evaluations with experts and quantitative, as well as qualitative measures, in PwPD. The Systems Usability Scale (SUS) provides a subjective assessment of usability by the patient and involves a set of ten Likert scale questions 20 . Participants provide ratings on a scale from 1 to 5 (strongly disagree to strongly agree) for odd-numbered questions and from 5 to 1 (strongly agree to strongly disagree) for even-numbered questions. The SUS-score is calculated by subtracting 1 from the user's response for odd-numbered questions and subtracting the user's response from 5 for even-numbered questions, summing these adjusted scores, and then multiplying the total by 2.5 to yield a score ranging from 0 to 100. Higher scores indicate better perceived usability, with scores above 68 considered above average and scores above 80 considered excellent 43 . The UEQ (User Experience Questionnaire) 21 is a standardized questionnaire used to assess the user experience of products, systems, or services. A total of 26 items measure six key dimensions: Attractiveness, Perspicuity (Clarity), Efficiency, Dependability, Stimulation, and Novelty. Participants rate each item on a seven-point scale, ranging from -3 (extremely bad) to +3 (extremely good). Mean values for each domain are analyzed more frequently than generating an overall score. Values > 0.8 are considered a positive evaluation 44 . In addition, we used a customized version of the UEQ for each test to ensure more precise feedback concerning the difficulty and speed level of each test as well as the perceived informative value of the test. The testing process followed an iterative design, allowing for step-wise refinement based on feedback from each testing iteration ensuring that it met the specific needs and usability requirements of PD patients and enhanced its overall effectiveness in cognitive assessment. Analysis and Feature Extraction Data analysis was conducted using Python V.3.10 and R V4.2.3. The Jupyter notebook is available on Github: https://github.com/tfeige91/DiCo-iOS-. 54 features were extracted from the 13 tests, including mean and standard deviation of the reaction time, percentage of correct answers and others. A full list of features can be found in the supplemental text. To mitigate the risk of multicollinearity, we identified pairs of correlations with a threshold of .85 and randomly excluded one variable. A total of four variables from four different tests were thereby excluded from the analysis: flanker_rt_mean_congruent, bart_correctedTotal, kirby_expValue, and StopSignal_correct_go. Furthermore, we clipped features at +/- 3 * interquartile range to minimize the influence of extreme outliers. Next, we calculated Variance Inflation Factors (VIFs) separately for each test and excluded variables with a VIF > 10. This test-wise approach ensured that at least one representative variable from each task remained for subsequent structural data analyses, including cluster mapping, network analysis, and exploratory factor analysis. Correlations and Networks To detect groups of variables that cohere based on shared variance, we performed a hierarchical agglomerative cluster analysis on the pairwise Pearson correlation matrix of all DiCo-derived variables using average linkage as the linkage criterion and 1 – |r| as the distance metric. The resulting dendrogram and clustered heatmap were generated using the clustermap function from the Seaborn package (version 0.13.2) in Python. This approach groups variables with similar correlation profiles, allowing an intuitive identification of variable clusters that reflect shared cognitive domains or common measurement characteristics. To identify robust relationships between the test variables, we calculated the sparse inverse covariance matrix (precision matrix) using the GraphicalLassoCV function from scikit-learn in Python using default parameters. The sparse inverse covariance matrix reveals information about the conditional dependency of two variables controlling for the effects of all other variables in the dataset. We used the non-diagonal values of the precision matrix to draw a network graph in Gephi, utilizing the force-directed layout algorithm ForceAtlas2 46 . The following tuning parameters were used in the depiction of the final graph: Scaling = 8.0, Gravity = 0.015, Stronger gravity = active. Factor Analysis To identify the underlying latent cognitive domains that can be measured by the DiCo we performed Exploratory Factor Analysis (EFA). Calculating the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (MSA) yielded an unsatisfactory value of < .60. We therefore followed the method proposed by Kaiser and Rice 47 and calculated the MSA for each variable separately, excluding those having an MSA < .50. This procedure led to the exclusion of 24 features, leaving 30 features. No variable from Reversal Learning , Kirby and BART remained in the EFA dataset. We then performed EFA with a maximum likelihood estimation and oblique (Promax) rotation. The decision on f , the number of factors to estimate, was based on the factors’ eigenvalues, keeping those with an eigenvalue > 1.0, which is the commonly used Kaiser’s criterion 22 in factor analysis. We performed EFA and calculated factor scores using the open-source Python module factor_analyzer 48 version 0.5.0. Following common best practices 49 we then excluded variables with communalities < .40 and repeated the EFA using the same extraction procedure described above. Clustering We performed a latent profile analysis (LPA) on the factor scores to identify possible subgroups. LPA aims to preserve information about participants on an individual level and classifies individuals into distinct groups based on differing combinations of personal and environmental traits. Compared to traditional, non-latent clustering methods (e.g., k-means clustering, hierarchical clustering), LPA treats profile membership as an unobserved categorical variable, where its value indicates which profile an individual belongs to with a certain degree of probability. It thus assigns individuals to clusters based on probabilities estimated directly from the model 50 . That means the researcher does not have to specify a certain number of clusters upfront, but the optimal solution can be based on different goodness of fit indices such as the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) as well as other measures like entropy, i.e. the goodness of cluster separation (1.0 being the best) and Bootstrap Likelihood Ratio Test (BLRT) that - if significant - suggests the superiority of the k-cluster solution over the k-1-cluster solution. We performed LPA with the R-package tidyLPA 51 . To determine if there are differences between the clusters, we performed a Kruskal-Wallis test followed by Dunn’s test. Predicting screening assessments and feature importance We trained a Random Forest regressor using scikit-learn to predict both the MoCA and the FAB using the factor scores derived from the EFA (as described above) as well as age, age at initial diagnosis, disease duration and LEDD. RF is a robust method known for its efficacy in handling complex, non-linear relationships among variables. 52 Hyperparameters were optimized via grid search with 5-fold cross-validation, minimizing the mean absolute error (MAE). Model performance was then evaluated using repeated 5×2-fold cross-validation, with three different random seeds per split to account for variance introduced by both data partitioning and the stochastic nature of the algorithm. For each model, the MAE and root mean squared error (RMSE) were recorded. Feature importances were averaged across all runs, and we additionally quantified their standard deviation due to random forest construction and cross-validation splits. Longitudinal analysis A total of 28 patients completed follow-up assessments using a subset of 6 DiCo tests: Simple Reaction Time Test, Stroop Test, N-Back, Tower of London, Flanker, and the Judgement of Line Orientation Task . The average interval between baseline and follow-up was 770 days (SD = 236). To evaluate cognitive change over time, we calculated follow-up factor scores for Working Memory , Interference Control, and Problem-Solving . These were derived using the factor loadings established from the baseline sample and restricted to features from the subset of tasks available at both timepoints. Correlations between these follow-up scores and the full-model factor scores (using the complete feature set) exceeded r = .95, confirming the robustness of the reduced-score estimation. We calculated Pearson correlation coefficients to assess the relationship between the change in the MoCA score and the time between baseline and follow-up. Features responsive to longitudinal change To quantify the magnitude of longitudinal change, we calculated the Standardized Response Mean (SRM) for each cognitive domain. The SRM is defined as the mean change score between baseline and follow-up divided by the standard deviation of the change scores. It represents an effect size metric commonly used to evaluate sensitivity to change, with values around 0.2, 0.5, and 0.8 typically interpreted as small, moderate, and large effects, respectively. Calculation of a continuous motor score To derive a continuous motor score that integrates multiple medication-related and clinical variables, we applied Uniform Manifold Approximation and Projection (UMAP) 53,54 , a non-linear dimensionality reduction technique that preserves the local and global structure of high-dimensional data in a low-dimensional embedding. We selected four variables that reflect motor symptom severity and medication status — levodopa equivalent daily dose from dopamine agonists (LEDD_DA), from levodopa (LEDD_L), total LEDD, and the MDS-UPDRS III motor score — and concatenated data from both baseline and follow-up assessments. The UMAP algorithm was applied to the z-standardized variables with the following parameters: n_components=1, n_neighbors=30, and random_state=24. The resulting one-dimensional embedding was then min-max scaled to a range between 0 and 1, with higher values indicating more severe motor impairment. Since UMAPs do not yield explicit feature weights, we calculated pairwise Pearson correlation coefficients between the input features and the UMAP-derived motor score to assess their individual contributions ( supplementary Fig. 2 ). We conducted paired t-tests on both the UPDRS III and the UMAP-derived Motor Score to assess whether significant changes occurred between baseline and follow-up assessments. Identifying baseline features to predict cognitive and motor progression To exploratorily identify relevant features for predicting the change in both MoCA and the motor score we calculated feature importances using the same Random Forest Regressor algorithm described above. Relation of cognitive and motor changes To examine the relationship between cognitive and motor changes over time, we calculated the change scores (Δ) for the MoCA and the UMAP-derived motor score by subtracting baseline values from follow-up values. To control for baseline effects, we performed linear regression analyses by regressing follow-up scores on baseline scores for each measure. The residuals from these regressions reflect change independent of baseline performance. We then computed the Pearson correlation between these residuals to assess the association between motor and cognitive change. To further investigate the role of reflection impulsivity and medication in motor progression, we conducted two linear regression analyses. First, we tested whether baseline Decision-Making performance predicted changes in motor functioning (Δ Motor Score). Second, we examined the association between the proportion of dopamine agonists in the total levodopa equivalent dose at baseline (DA/LEDD Total) and motor progression. In both models, we estimated a linear regression line and computed Pearson correlation coefficients to quantify the strength of the relationship. Results were visualized using regression plots with 95 % confidence intervals. Abbreviations LEDD Levodopa Equivalent Daily Dose DA Dopamine Agonist MoCA Montreal Cognitive Assessment, FAB-Frontal Assessment Battery, SCD-Subjective Cognitive Decline Declarations Author contributions: Tim Feige and Anika Frank contributed equally to this work. TF, AF, JB, BF conceived the study. TF, AF, JB, AE, CH, JK, JJ, NS conducted the research. TF, AF, JB, BF analysed the data. TF, AF, JB, BF wrote the initial draft. HR, BF supervised the research. All authors reviewed the manuscript. Competing interests: The authors have no competing interests as defined by Nature Portfolio, or other interests that might be perceived to influence the results and/or discussion reported in this paper. Data availability: Data is provided within the manuscript or supplementary information files. References Aarsland, D. et al. Parkinson disease-associated cognitive impairment. Nat Rev Dis Primers 7 , 47 (2021). Dekker, M. C. J. et al. Parkinson’s Disease Research on the African Continent: Obstacles and Opportunities. Front. Neurol. 11 , 512 (2020). De Lau, L. M. L., Schipper, C. M. A., Hofman, A., Koudstaal, P. J. & Breteler, M. M. B. Prognosis of Parkinson Disease: Risk of Dementia and Mortality: The Rotterdam Study. Arch Neurol 62 , 1265 (2005). Aarsland, D. & Kurz, M. W. The epidemiology of dementia associated with Parkinson disease. Journal of the Neurological Sciences 289 , 18–22 (2010). Mantri, S. et al. In Their Own Words: Fears Expressed by People with Parkinson’s Disease in an Online Symptom Database. J Parkinsons Dis 14 , 865–872 (2024). Kjeldsen, P. L. & Damholdt, M. F. Subjective cognitive complaints in patients with Parkinson’s disease. Acta Neurol Scand 140 , 375–389 (2019). Cammisuli, D. M., Cammisuli, S. M., Fusi, J., Franzoni, F. & Pruneti, C. Parkinson’s Disease–Mild Cognitive Impairment (PD-MCI): A Useful Summary of Update Knowledge. Front. Aging Neurosci. 11 , 303 (2019). Dubois, B. et al. Diagnostic procedures for Parkinson’s disease dementia: Recommendations from the movement disorder society task force. Movement Disorders 22 , 2314–2324 (2007). Litvan, I. et al. Diagnostic criteria for mild cognitive impairment in Parkinson’s disease: Movement Disorder Society Task Force guidelines. Movement Disorders 27 , 349–356 (2012). Roheger, M., Kalbe, E. & Liepelt-Scarfone, I. Progression of Cognitive Decline in Parkinson’s Disease. JPD 8 , 183–193 (2018). Kehagia, A. A., Barker, R. A. & Robbins, T. W. Cognitive Impairment in Parkinson’s Disease: The Dual Syndrome Hypothesis. Neurodegener Dis 11 , 79–92 (2013). Jain, A. K., Murty, M. N. & Flynn, P. J. Data clustering: a review. ACM Comput. Surv. 31 , 264–323 (1999). Düzel, E., Thyrian, J. R. & Berron, D. Innovation in der Diagnostik – mobile Technologien. Nervenarzt 90 , 914–920 (2019). Nasreddine, Z. S. et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. Journal of the American Geriatrics Society 53 , 695–699 (2005). Dubois, B., Slachevsky, A., Litvan, I. & Pillon, B. The FAB: A frontal assessment battery at bedside. Neurology 55 , 1621–1626 (2000). Sochat, V. The Experiment Factory: Reproducible Experiment Containers. JOSS 3 , 521 (2018). De Leeuw, J. R., Gilbert, R. A. & Luchterhandt, B. jsPsych: Enabling an Open-Source CollaborativeEcosystem of Behavioral Experiments. JOSS 8 , 5351 (2023). Alonso-Canovas, A. et al. The Early Treatment Phase in Parkinson’s Disease: Not a Honeymoon for All, Not a Honeymoon at All? Journal of Parkinson’s Disease 13 , 323–328 (2023). Fillenbaum, G. G. & Mohs, R. CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) Neuropsychology Assessment Battery: 35 Years and Counting. JAD 93 , 1–27 (2023). Brooke, J. System usability scale (SUS): a quick-and-dirty method of system evaluation user information. Reading, UK: Digital equipment co ltd 43 , 1–7 (1986). Laugwitz, B., Held, T. & Schrepp, M. Construction and Evaluation of a User Experience Questionnaire. in HCI and Usability for Education and Work (ed. Holzinger, A.) 63–76 (Springer Berlin Heidelberg, Berlin, Heidelberg, 2008). Kaiser, H. F. The Application of Electronic Computers to Factor Analysis. Educational and Psychological Measurement 20 , 141–151 (1960). Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E. & Tatham, R. L. Multivariate data analysis. Uppersaddle River. Multivariate Data Analysis (5th ed) Upper Saddle River 5 , 207–219 (1998). Pourzinal, D. et al. Systematic review of data‐driven cognitive subtypes in Parkinson disease. Euro J of Neurology 29 , 3395–3417 (2022). Van Der Maas, H. L. J. et al. A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological Review 113 , 842–861 (2006). Balzer-Geldsetzer, M. et al. Parkinson’s Disease and Dementia: A Longitudinal Study (DEMPARK). Neuroepidemiology 37 , 168–176 (2011). Latendorf, A., Runde, L. M., Salminen, T. & Steinert, A. Digitization of neuropsychological diagnostics: a pilot study to compare three paper-based and digitized cognitive assessments. Aging Clin Exp Res 33 , 1585–1597 (2021). Ferriola, A. M. A COMPARISON OF THE WISC-V PROCESSING SPEED SUBTESTS: PAPER-PENCIL AND DIGITAL FORMATS. (2019). Eisenberg, I. W. et al. Uncovering the structure of self-regulation through data-driven ontology discovery. Nat Commun 10 , 2319 (2019). Scarpina, F. & Tagini, S. The Stroop Color and Word Test. Front. Psychol. 8 , (2017). Arbuthnott, K. & Frank, J. Trail Making Test, Part B as a Measure of Executive Control: Validation Using a Set-Switching Paradigm. Journal of Clinical and Experimental Neuropsychology 22 , 518–528 (2000). Duncan, G. W. et al. Gray and white matter imaging: A biomarker for cognitive impairment in early P arkinson’s disease? Movement Disorders 31 , 103–110 (2016). Henry, K. et al. Cognitive Predictors of Quality of Life in Parkinson’s Disease: A Three-year Longitudinal Study (P4-3.018). Neurology 102 , 2615 (2024). Tang, Y. et al. Cognitive Function and Quality of Life in Parkinson’s Disease: A Cross-Sectional Study. JPD 10 , 1209–1216 (2020). Lawson, R. A. et al. Cognitive decline and quality of life in incident Parkinson’s disease: The role of attention. Parkinsonism & Related Disorders 27 , 47–53 (2016). Djamshidian, A. et al. Dopamine Agonists Rather than Deep Brain Stimulation Cause Reflection Impulsivity in Parkinson’s Disease. Journal of Parkinson’s Disease 3 , 139–144 (2013). Weintraub, D. et al. Association of Dopamine Agonist Use With Impulse Control Disorders in Parkinson Disease. Arch Neurol 63 , 969 (2006). Grosset, K. A. et al. Problematic gambling on dopamine agonists: Not such a rarity. Movement Disorders 21 , 2206–2208 (2006). Antonelli, F. et al. Dopamine‐agonists and impulsivity in Parkinson’s disease: Impulsive choices vs. impulsive actions. Human Brain Mapping 35 , 2499–2506 (2014). Foltynie, T., Brayne, C. E. G., Robbins, T. W. & Barker, R. A. The cognitive ability of an incident cohort of Parkinson’s patients in the UK. The CamPaIGN study. Brain 127 , 550–560 (2004). Postuma, R. B. et al. MDS clinical diagnostic criteria for Parkinson’s disease: MDS-PD Clinical Diagnostic Criteria. Mov Disord. 30 , 1591–1601 (2015). Litvan, I. et al. MDS task force on mild cognitive impairment in Parkinson’s disease: Critical review of PD‐MCI. Movement Disorders 26 , 1814–1824 (2011). Sauro, J. A Practical Guide to the System Usability Scale: Background, Benchmarks & Best Practices . (Measuring Usability LLC, 2011). Schrepp, M. User experience questionnaire handbook: All you need to know to apply the ueq successfully in your projects (2015). URL http://www. ueq. online. org (2022). Nielsen, J. Heuristic evaluation. in Usability inspection methods 25–62 (John Wiley & Sons, Inc., 1994). Jacomy, M., Venturini, T., Heymann, S. & Bastian, M. ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software. PLoS ONE 9 , e98679 (2014). Kaiser, H. F. & Rice, J. Little Jiffy, Mark Iv. Educational and Psychological Measurement 34 , 111–117 (1974). Biggs, J. factor_analyzer Documentation. Costello, A. B. & Osborne, J. W. Best Practices in Exploratory Factor Analysis: Four Recommendations for Getting the Most From Your Analysis. Exploratory Factor Analysis . Spurk, D., Hirschi, A., Wang, M., Valero, D. & Kauffeld, S. Latent profile analysis: A review and “how to” guide of its application within vocational behavior research. Journal of Vocational Behavior 120 , 103445 (2020). Rosenberg, J., Beymer, P., Anderson, D., Van Lissa, C. j. & Schmidt, J. tidyLPA: An R Package to Easily Carry Out Latent Profile Analysis (LPA) Using Open-Source or Commercial Software. JOSS 3 , 978 (2018). Fife, D. A. & D’Onofrio, J. Common, uncommon, and novel applications of random forest in psychological research. Behav Res 55 , 2447–2466 (2022). McInnes, L., Healy, J. & Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. Preprint at https://doi.org/10.48550/ARXIV.1802.03426 (2018). Healy, J. & McInnes, L. Uniform manifold approximation and projection. Nat Rev Methods Primers 4 , 82 (2024). Lejuez, C. W. et al. Evaluation of a behavioral measure of risk taking: The Balloon Analogue Risk Task (BART). Journal of Experimental Psychology: Applied 8 , 75–84 (2002). Cools, R., Clark, L., Owen, A. M. & Robbins, T. W. Defining the Neural Mechanisms of Probabilistic Reversal Learning Using Event-Related Functional Magnetic Resonance Imaging. J. Neurosci. 22 , 4563–4567 (2002). Benton, A. L. Visuospatial Judgment: A Clinical Test. Arch Neurol 35 , 364 (1978). Wechsler Memory Scale - Fourth Edition: Deutsche Übersetzung und Adaptation der WMS-IV von David Wechsler. (2015). Schretlen, D., Bobholz, J. H. & Brandt, J. Development and psychometric properties of the brief test of attention. The Clinical Neuropsychologist 10 , 80–89 (1996). Warrington, E. K. & James, M. Testbatterie Für Visuelle Objekt- Und Raum- Wahrnehmung (Deutsche Übersetzung von Beckers K, Canavan A) . (Thames Valley Test Company, Bury St. Edmunds, 1992). Wechsler Adult Intelligence Scale - Fourth Edition. Deutschsprachige Adaptation Der WAIS-IV von D. Wechsler . (Pearson Assessment & Information GmbH, Frankfurt/M, 2012). Storch, A. et al. Non-motor Symptoms Questionnaire und Scale für das idiopathische Parkinson-Syndrom: Interkulturell adaptierte Versionen in deutscher Sprache. Nervenarzt 81 , 980–985 (2010). Trenkwalder, C. et al. Parkinson’s disease sleep scale—validation of the revised version PDSS‐2. Movement Disorders 26 , 644–652 (2011). Bloch, K. E., Schoch, O. D., Zhang, J. N. & Russi, E. W. German Version of the Epworth Sleepiness Scale. Respiration 66 , 440–447 (1999). Starkstein, S. E. et al. Reliability, validity, and clinical correlates of apathy in Parkinson’s disease. J Neuropsychiatry Clin Neurosci 4 , 134–139 (1992). Krupp, L. B., LaRocca, N. G., Muir-Nash, J. & Steinberg, A. D. The fatigue severity scale. Application to patients with multiple sclerosis and systemic lupus erythemato-sus. Arch. Neurol 46 , 1121–1123 (1989). Beck, A. T., Steer, R. Ab. & K, G. BDI-II, Beck Depression Inven-Tory :Manual San Antonio, Tex . (Psychological Corp, 1996). Pfeffer, R. I., Kurosaki, T. T., Harrah, C. H., Chance, J. M. & Filos, S. Measurement of Functional Activities in Older Adults in the Community. Journal of Gerontology 37 , 323–329 (1982). Rabin, R. & Charro, F. D. EQ-SD: a measure of health status from the EuroQol Group. Annals of Medicine 33 , 337–343 (2001). Jenkinson, C., Fitzpatrick, R., Peto, V., Greenhall, R. & Hyman, N. The PDQ-8: Development and validation of a short-form parkinson’s disease questionnaire. Psychology & Health 12 , 805–814 (1997). Rammstedt, B., Kemper, C. J., Klein, M. C., Beierlein, C. & Kovaleva, A. A Short Scale for Assessing the Big Five Dimensions of Personality: 10 Item Big Five Inventory (BFI-10). methods, data, analyses 7 , 17 Pages (2013). Bayard, S. et al. A multidimensional approach to impulsivity in Parkinson’s disease: measurement and structural invariance of the UPPS Impulsive Behaviour Scale. Psychol. Med. 46 , 2931–2941 (2016). Schmidt, R. E., Gay, P., d’Acremont, M. & Van Der Linden, M. A German Adaptation of the UPPS Impulsive Behavior Scale: Psychometric Properties and Factor Structure. Swiss Journal of Psychology 67 , 107–112 (2008). Weintraub, D. et al. Questionnaire for impulsive‐compulsive disorders in Parkinson’s Disease–Rating Scale. Movement Disorders 27 , 242–247 (2012). Guy, W. ECDEU Assessment Manual for Psychopharmacology . (US Department of Health, Education, and Welfare, Public Health Service …, 1976). Patton, J. H., Stanford, M. S. & Barratt, E. S. Factor structure of the barratt impulsiveness scale. J. Clin. Psychol. 51 , 768–774 (1995). Foa, E. B. et al. The Obsessive-Compulsive Inventory: Development and validation of a short version. Psychological Assessment 14 , 485–496 (2002). Baumann‐Vogel, H., Valko, P. O., Eisele, G. & Baumann, C. R. Impulse control disorders in Parkinson’s disease: don’t set your mind at rest by self‐assessments. Euro J of Neurology 22 , 603–609 (2015). Table Table 3 is available in the Supplementary Files section Additional Declarations No competing interests reported. Supplementary Files FeigeFrankDiCosupplement.pdf Table3.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7585530","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":529105377,"identity":"8c12130b-7aac-40e3-ab60-58cabbfa2c6c","order_by":0,"name":"Tim Feige","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Tim","middleName":"","lastName":"Feige","suffix":""},{"id":529105380,"identity":"0a3744dc-dc53-4d6e-b11e-7579cfa66f84","order_by":1,"name":"Anika Frank","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Anika","middleName":"","lastName":"Frank","suffix":""},{"id":529105382,"identity":"aa74f51d-be29-40c6-a510-bd9d999c508a","order_by":2,"name":"Jonas Bendig","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Jonas","middleName":"","lastName":"Bendig","suffix":""},{"id":529105385,"identity":"ec9771a7-69b8-43f1-ab03-e82b6a6336f4","order_by":3,"name":"Andrea Epler","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Andrea","middleName":"","lastName":"Epler","suffix":""},{"id":529105388,"identity":"77ca952e-a1ac-4ef5-a0ba-e34d99eaa086","order_by":4,"name":"Charlotte Harbarth","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Charlotte","middleName":"","lastName":"Harbarth","suffix":""},{"id":529105390,"identity":"8ad2fd82-d96f-4d19-b8f4-912607c0b2c5","order_by":5,"name":"Julia Kunze","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Julia","middleName":"","lastName":"Kunze","suffix":""},{"id":529105392,"identity":"c9ded051-d7ba-46fb-8008-b0c36a65547d","order_by":6,"name":"Johanna Jungk","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Johanna","middleName":"","lastName":"Jungk","suffix":""},{"id":529105396,"identity":"bababd18-d6e2-4376-8ae0-1753dad2891e","order_by":7,"name":"Nils Schnalke","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Nils","middleName":"","lastName":"Schnalke","suffix":""},{"id":529105398,"identity":"2e5a77a1-094c-4e96-838e-faa22e642b18","order_by":8,"name":"Heinz Reichmann","email":"","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":false,"prefix":"","firstName":"Heinz","middleName":"","lastName":"Reichmann","suffix":""},{"id":529105401,"identity":"a3193549-f921-4085-b7d5-7e614455f040","order_by":9,"name":"Björn H. Falkenburger","email":"data:image/png;base64,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","orcid":"","institution":"University Hospital Carl Gustav Carus, Technische Universität Dresden","correspondingAuthor":true,"prefix":"","firstName":"Björn","middleName":"H.","lastName":"Falkenburger","suffix":""}],"badges":[],"createdAt":"2025-09-10 18:38:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7585530/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7585530/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93618112,"identity":"eb73e8c9-10be-4fe7-b5cb-95954b726e93","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1232112,"visible":true,"origin":"","legend":"","description":"","filename":"FeigeFrankDiCoMajorRevisionClean.docx","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/e12f3759e8e491d7d0ad9ab3.docx"},{"id":93618894,"identity":"d4a58944-3a61-4fa1-82d6-d2905daf47bf","added_by":"auto","created_at":"2025-10-15 17:26:20","extension":"json","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10270,"visible":true,"origin":"","legend":"","description":"","filename":"10ed0d5540f0457289b6624ec3115fd5.json","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/e4f56148cf1fdb76b836510a.json"},{"id":93616782,"identity":"dabf0e41-f7a2-468f-9d35-db1a363b996a","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"pdf","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1115969,"visible":true,"origin":"","legend":"","description":"","filename":"FeigeFrankDiCosupplement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/a8ae9b4d9abc06d0047a4c2d.pdf"},{"id":93616778,"identity":"985dd685-d7fd-480c-babe-65a7ecb751f1","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"xml","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":244842,"visible":true,"origin":"","legend":"","description":"","filename":"10ed0d5540f0457289b6624ec3115fd51enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/2bab3bd376c505cda9e1d8a0.xml"},{"id":93618113,"identity":"08a2f1c2-e1ca-40e4-be34-93559e550d40","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"pdf","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":109314,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/cd9630f33bdf8af4b3043acc.pdf"},{"id":93617672,"identity":"cc4bcb7a-bbb3-4d07-b337-4977e85c52e2","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63530,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/0bc4b352b9748eaae5d133c5.pdf"},{"id":93616771,"identity":"2c72a22d-cb11-4d38-907c-93e8e7fc0edf","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"pdf","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52695,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.3.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/dadb3ba22d5dadae2cf0531a.pdf"},{"id":93619186,"identity":"76b07cc0-1b64-4903-9a7a-c9788f8140b0","added_by":"auto","created_at":"2025-10-15 17:34:20","extension":"pdf","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28331,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/5cfec7f50f8b4e0bc450523b.pdf"},{"id":93617677,"identity":"642d1ba3-94c3-42ba-8eb5-9895c9bf0ca6","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"pdf","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39843,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/38e807aa80506354f25be595.pdf"},{"id":93616781,"identity":"682152fa-f43a-491e-bb79-3d0ec0372bc3","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"pdf","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54938,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/a399b9237d06c9268ee818a0.pdf"},{"id":93616772,"identity":"2c43f523-fca6-4b67-8ea8-a676350ae742","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"pdf","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34148,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.7.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/e54ebf9417c9af8c4748194c.pdf"},{"id":93618892,"identity":"366529bd-f0cf-465a-85ee-81c436934899","added_by":"auto","created_at":"2025-10-15 17:26:20","extension":"pdf","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24248,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.8.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/1178fc7573bb7bcd27d36abd.pdf"},{"id":93618117,"identity":"acc00be0-13fe-4f32-9652-4ff460491a56","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"pdf","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35945,"visible":true,"origin":"","legend":"","description":"","filename":"Fig.9.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/8c7118143bb92bfe995631dd.pdf"},{"id":93616794,"identity":"e1c8d2f3-56b4-41db-9510-47c276016fe8","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"emf","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1330392,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/bbb95a6f12c8ba5bd786f9e4.emf"},{"id":93616791,"identity":"b90941e3-2811-4ab2-a2e0-ab15a8540b73","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"emf","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":685512,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/a2878736d0ca0b86c7bae8c4.emf"},{"id":93616796,"identity":"2702b626-25f5-47d7-8a1f-95bf3c633a90","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"emf","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":195500,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/4916da96c76a2f668e314cc1.emf"},{"id":93616792,"identity":"04102379-f167-470e-9f22-af2207f1e87f","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"emf","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":423520,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/98e51906281e617c6565aa72.emf"},{"id":93618118,"identity":"59c9587b-5e64-49cb-b5dd-b0b5c32527b9","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"emf","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":859400,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/c3741fadbe9792ac5b68cc9d.emf"},{"id":93616788,"identity":"12dbda41-9e30-487e-9c41-d797301014aa","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"emf","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":138248,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/3a893f60de956b4d144824c7.emf"},{"id":93617685,"identity":"fdfb7bd5-93db-4460-87b5-c9b47a7f16ce","added_by":"auto","created_at":"2025-10-15 17:10:21","extension":"emf","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":188420,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/ac52f4cb844264bed2f0243e.emf"},{"id":93616795,"identity":"73c97f00-8cdb-4c8e-9192-5ed61534410a","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"emf","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":110452,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/d5f273995f8ddb3e3f5c72d6.emf"},{"id":93616803,"identity":"cfaea913-e54a-44f9-8c40-38f6ae4deec4","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"emf","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":155152,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.emf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/adf9f3cd8a03b74feb31b42b.emf"},{"id":93617679,"identity":"fc832739-2361-45ad-b97c-d1e8f2b256ac","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10226,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/b7aab2a983e0c3fd949c840d.png"},{"id":93616793,"identity":"af7dc387-35db-4e0f-b48b-7f3652ced7bf","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24817,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/e61c6a895c78e3435b2ceda4.png"},{"id":93617683,"identity":"4cdd0946-18e1-486c-bc4d-875bd49f22dc","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11259,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/ae9024a427ea703b34722088.png"},{"id":93618896,"identity":"9b242540-8843-4073-a353-5e34d9b85476","added_by":"auto","created_at":"2025-10-15 17:26:20","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19604,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/d9af7599391605f23376c17a.png"},{"id":93618895,"identity":"c835ce96-26ec-4f1a-82f9-898f69c28f9d","added_by":"auto","created_at":"2025-10-15 17:26:20","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39138,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/0a8cb3d4a79d1afd68256bc1.png"},{"id":93618120,"identity":"a3a5c8f3-4342-47e6-8d29-f1ab29ecb2d7","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21718,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/63eb30bc89353ecef51bbf9d.png"},{"id":93616797,"identity":"c93e3ae1-931b-4e14-84e0-b768fe4a72bf","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35177,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/cb17b48ea7f8e86ed3e8d64c.png"},{"id":93616799,"identity":"3a6af8af-d7fe-4055-b1a9-8b098ae94e19","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14039,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/84dd9fdceb49a0aeb7ab45e6.png"},{"id":93616800,"identity":"63347a70-b09c-40e4-ad75-9e0b986e4e66","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"png","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20657,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/334560bf72f9bb94efabd621.png"},{"id":93617686,"identity":"748b26e0-85b9-4818-adfc-5c7bc93f85ae","added_by":"auto","created_at":"2025-10-15 17:10:21","extension":"xml","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":241470,"visible":true,"origin":"","legend":"","description":"","filename":"10ed0d5540f0457289b6624ec3115fd51structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/93a4b53f08a601926b898592.xml"},{"id":93616802,"identity":"24b09466-5d58-4224-871d-ce9ffb1140d9","added_by":"auto","created_at":"2025-10-15 17:02:21","extension":"html","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":261271,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/a197af5c186a27f1bbf2c9da.html"},{"id":93617666,"identity":"9351f43e-8373-4201-87b9-2415d88583d1","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":145659,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eUsability testing reveals high overall satisfaction in established usability measures. (a) \u003c/strong\u003e\u003cem\u003eOverall UEQ ratings (a) for the DiCo (black diamond and whiskers) in the six established domains against the established benchmarks (colored bars). The measured scale means (±standard deviation) are set in relation to existing values from a benchmark data set.\u003c/em\u003e\u003cem\u003e\u003cstrong\u003e (b)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e SUS ratings from n=15 PwPD participants of the usability testing. For each question, patients responded on a 5-item Likert scale (strongly disagree, disagree, neutral, agree, strongly agree).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/b7be68a29f3f4043cd3af80d.png"},{"id":93616759,"identity":"c3dab54c-74d9-4ad7-9381-2eefc04d82a5","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":133501,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCluster map of DiCo feature correlations revealing three clusters. \u003c/strong\u003e\u003cem\u003eC1 includes reaction time and timing-related measures. C2 comprises accuracy-based variables that reflect actual task performance in terms of correctness. C3 represents a residual group, primarily consisting of features from impulsivity-related tasks. Supplemental Figure 1 displays the same analysis with measures in cluster 1 inverted. \u003c/em\u003eAbbreviations used in this figure are defined in \u003cstrong\u003eTable 3.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/0a341b1f8ab93dd449483024.png"},{"id":93616762,"identity":"1b9bdfe2-d19d-4659-9886-3cce1e0fd22b","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":63389,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConditional dependencies predominantly occur within the same cognitive test. \u003c/strong\u003e\u003cem\u003eA graphical lasso was used to estimate a sparse graph of undirected relationships between all digital variables (a and b). Variables were colored according to the parent test \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(a)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eor according to their association with the clusters from Fig. 2 \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(b).\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eEdges in the graph reflect absolute conditional dependency greater than 0.10. The size of the nodes increases with the number of connections, the thickness of the edges increases with stronger absolute conditional dependency.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/f6bdb14c41b807d4370926ef.png"},{"id":93616765,"identity":"caba2f22-b184-483c-b1e3-b67fcfd7ddb8","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":84505,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExploratory Factor Analysis reveals four intercorrelated latent factors. \u003c/strong\u003e(a) Five extracted factors and factor loadings of DiCo variables. (b) Correlations between the five extracted factors. SS = StopSignal, GN = Go-Nogo, FL = Flanker, ST = Stroop, LO = Judgement of Line Orientation, TOL = Tower of London, IS = Information Sampling, NB = NBack-Task; Further abbreviations used in this figure are defined in \u003cstrong\u003eTable 3.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/12be8e04547d786b57592a20.png"},{"id":93618111,"identity":"ab9074d3-e5ed-426d-aac4-112057dc1d6b","added_by":"auto","created_at":"2025-10-15 17:18:20","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":168813,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConstruct validity of DiCo factors. \u003c/strong\u003eDepicted are Pearson correlations between DiCo-derived factors and external neuropsychological measures (a) as well as self-report questionnaire data (b). Neuropsychological data were available for n = 50 and questionnaire data for n = 77 participants. Abbreviations used in this figure are defined in \u003cstrong\u003eTable 3\u003c/strong\u003e.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/0b7cd1d9fb915e92aebfa4c4.png"},{"id":93617668,"identity":"4010be0f-452c-4246-bcab-2bf28ec9f4ad","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":73318,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLatent profile analysis on DiCo factor scores revealed four clusters. \u003c/strong\u003e(a) \u003cem\u003eParticipants were clustered by their factor scores using Latent Profile Analysis. Clusters 1 (n = 13), 3 (n = 21) and 4 (n =10) differed consistently across all factors, whereas cluster 2 (n = 53) performed similarly to cluster 1 in four of the five factors but significantly worse in Decision-Making. (b-f) Associations between each factor and age with data points color-coded according to the cluster membership. All factors except Decision-Making showed significant age-related decline.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/bc92b4c7a79cbdf4dc01a250.png"},{"id":93616763,"identity":"0c518c53-3c96-4a75-b9ce-a66c21b93c17","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":44915,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLongitudinal analyses of cognitive and motor outcomes. \u003c/strong\u003e\u003cem\u003eThe UMAP-derived motor score yielded the highest SRM with MoCA and Working Memory being almost equally sensitive to change (a). MoCA scores showed slight improvements with increasing time between measurements (b). Individual trajectories for MoCA as well as the three DiCO-derived factors Working Memory, Problem-Solving and Interference Control (c-f). While UPDRS III remained relatively stable from baseline to follow-up (g), the UMAP-derived motor score showed progression in almost every participant (h).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/05033c29301b21d2cfbce460.png"},{"id":93617667,"identity":"ddbae2e2-c079-4665-8898-5401a3f45d82","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":26212,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAssociations between DiCo factors and cognitive screening tests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWe trained Random Forest models to predict MoCA \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(a) \u003c/strong\u003e\u003c/em\u003e\u003cem\u003eand FAB \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(b)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003escores using DiCo factor scores alongside sociodemographic and clinical variables. Prediction accuracy was evaluated across 150 repeated cross-validation runs. For MoCA, Working Memory and Visual Attention were the most important predictors. For FAB, feature importance was more evenly distributed, with Interference Control and total LEDD being most influential.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/d0e857d2a62e118e4d2651ee.png"},{"id":93617670,"identity":"d286ec68-5618-4c95-941e-4ba05ec79ed5","added_by":"auto","created_at":"2025-10-15 17:10:20","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":41956,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePredictors of Cognitive and Motor Changes. \u003c/strong\u003e\u003cem\u003eRelative feature importances from Random Forest regressors predicting changes in MoCA \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(a)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e and UMAP-derived Motor Score \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(b).\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eFor the prediction of motor decline, baseline (BL) DiCO-derived Decision-Making was among the most important predictors, following time between assessments and baseline motor performance. Baseline-corrected changes in MoCA and Motor Score \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(c)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003ewere only weakly negatively correlated, suggesting that cognitive and motor progression may occur independently. Additionally, lower Decision-Making performance at baseline \u003c/em\u003e\u003cem\u003e\u003cstrong\u003e(d)\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e were associated with greater motor deterioration.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/533cfb87b6cedda947990b78.png"},{"id":105860248,"identity":"be40a993-b2a7-4deb-96fc-83a721871b89","added_by":"auto","created_at":"2026-04-01 01:10:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2302562,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/cc4c2bd0-a400-4eb0-acfb-cef714763f46.pdf"},{"id":93616768,"identity":"c84d063a-c5ec-4f6c-aaf8-77888534d3d1","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1115969,"visible":true,"origin":"","legend":"","description":"","filename":"FeigeFrankDiCosupplement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/5b63bcfb7fe5c6c633724013.pdf"},{"id":93616760,"identity":"0fa3d103-71fc-4ccb-b2b4-36cfd17abebe","added_by":"auto","created_at":"2025-10-15 17:02:20","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":48472,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-7585530/v1/663e8d89561c0f35b23bb467.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Cognitive structure and progression in Parkinson’s Disease: Insights from a tablet-based assessment","fulltext":[{"header":"Introduction","content":"\u003cp\u003eParkinson\u0026rsquo;s Disease (PD) is the second most common neurodegenerative disease and the fastest growing neurological disorder in the world with regard to prevalence, disability and deaths\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. PD is defined by its cardinal motor symptoms resting tremor, rigidity and bradykinesia. In addition, PD encompasses a multitude of non-motor symptoms (NMS). People with PD (PwPD) have a 2.8 to 6-fold higher risk of developing dementia than the general population\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. Several longitudinal studies have demonstrated that almost all PwPD develop dementia if they live with PD for more than ten years\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Hence, cognitive impairment is one of the most common NMS and arguably the NMS that PwPD fear most\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eCognitive impairment in PD can be classified based on severity and based on the affected functional systems. The following severity levels are currently delineated: normal cognition (NC), Subjective Cognitive Decline (SCD), Mild Cognitive Impairment (MCI), and Parkinson's Disease Dementia (PDD)\u003csup\u003e\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. In addition, the following functional domains are distinguished by the Movement Disorder Society: attention/working memory, executive functions, language, memory, and visuospatial functions\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. The timing, profile, and rate of cognitive decline can be highly variable in PD\u003csup\u003e10\u003c/sup\u003e. The observation of distinct functional impairment profiles in MCI has led to the hypothesis of the \"dual syndrome,\" suggesting two cognitive subtypes of PD-MCI. These are based on the impairment of distinct anatomical regions: the frontal syndrome, associated with executive and attentional deficits, and the posterior cortical syndrome, associated with visual-spatial and memory deficits\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eDigital assessments build on a long tradition of psychophysical research in PD and potentially offer the opportunity to collect features of cognitive function beyond what is captured in classical paper-pencil-based neuropsychological testing (NPT). Hence, digital assessments can potentially reveal novel patterns and relationships\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. In addition, digital testing methods have been suggested to allow for low-threshold longitudinal measurements with high detail, applicable with minimal effort in a large population\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Digital assessments should therefore provide at least the information obtained by commonly used screening instruments such as the Montreal Cognitive Assessment (MoCA)\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e and the Frontal Assessment Battery (FAB)\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. Among digital methods, open-source tools support the use of comparable tests in different cohorts as a basis for the generation of poolable datasets for data-driven investigation into cognitive profiles associated with neurodegenerative disorders and the modeling of disease progression trajectories.\u003c/p\u003e\u003cp\u003eWe therefore intended to design a digital, tablet-based cognitive assessment (DiCo) that is easily applicable in clinical practice, expanding on readily available, open-source systems and frameworks such as the \u003cem\u003eExperiment Factory\u003c/em\u003e (EF)\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e and jsPsych\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. The Selection of tests was based on a systematic review of the literature. Only established assessments or adaptations thereof were included in the implementation. Extensive usability testing with PwPD and PD experts was performed to ensure applicability in research and practice. In order to determine the type of information acquired in the DiCo, conditional dependencies were visualized and an exploratory factor analysis was conducted. To determine to what extent data captured by the DiCo can reproduce the information contained in MoCA and FAB, Random Forest Regressors were trained. Finally, to address the question of heterogeneity, a Latent Profile Analysis was performed, finding four clusters of participants. Three clusters mainly differed in severity, whereas the remaining cluster was characterized by predominant reflection impulsivity.\u003c/p\u003e"},{"header":"Results","content":"\u003ch2\u003eStudy participants\u003c/h2\u003e\n\u003cp\u003eAfter usability testing, the tablet-based digital assessment was administered to a cohort of \u003cem\u003en\u0026nbsp;\u003c/em\u003e= 100 participants (43% women) of whom 97 completed the entire test battery. The three dropouts had different reasons: Two patients had to terminate the study earlier because of scheduled follow-up medical appointments and one felt too exhausted to continue the testing. Demographic and clinical data of all participants are displayed in \u003cstrong\u003eTable 1\u003c/strong\u003e. With an average disease duration of 6.4 years and a mean LEDD of 672 mg, our sample is beyond the typical \u0026lsquo;honeymoon period\u0026rsquo; of Parkinson\u0026rsquo;s disease, which is usually defined as the first 2\u0026ndash;5 years of stable response to levodopa\u003csup\u003e18\u003c/sup\u003e. The average score in the Montreal Cognitive Assessment (MoCA) was 26 points.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCompleting the entire assessment took 67 minutes on average, the interquartile range was 59-71 minutes. From each test, several different features were extracted, including mean and standard deviation of the reaction time, percentage of correct answers and others. In total, 54 features were extracted from the 13 tests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA subsample of n = 50 participants additionally took part in a paper-pencil neuropsychological assessment (NPT) that was mainly comprised of the CERAD+\u003csup\u003e19\u003c/sup\u003e. Questionnaire (Q-Data) were available for n = 76 participants. An overview of all NPT and Q-Data variables is provided in \u003cstrong\u003eTable 3\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eAnother subset of n = 28 participants was available for a follow-up assessment. The mean interval between the two assessments was 25 months. Demographics of this subset are also included in \u003cstrong\u003eTable 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1: \u003cem\u003eDemographic and Clinical Characteristics\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"98%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eFeature (at baseline)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eEntire cohort (n=97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eFollow-up (n=28)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eGender, % male\u003csup\u003e1)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e55 (57%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e13 (46 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eAge (years)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e63 (10.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e62 (9.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eDisease Duration (years)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e6 (4.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e6.4 (3.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eAge at onset (years)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e56 (10.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e56 (10.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eLEDD Total (mg)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e672 (338.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e665 (355.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eLEDD LDopa (mg)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e418 (197.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e365 (189.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eLEDD DA (mg)\u003csup\u003e\u0026nbsp;2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e244 (141.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e269 (126.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eNeuroleptics\u003csup\u003e1)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e6 (6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e1 (3.5 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eHallucinations\u003csup\u003e1)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e5 (5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eVascular Risk present\u003csup\u003e1)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e34 (35%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e\u0026nbsp;7 (25 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eUPDRS III\u003csup\u003e2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e19 (9.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e17.4 (7.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eMoCA\u003csup\u003e2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e26 (2.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e27.8 (1.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eFAB\u003csup\u003e2)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e16.5 (1.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e17.4 (.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003eSCD\u003csup\u003e1)\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e53 (54.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 33px;\"\u003e\n \u003cp\u003e17 (60 %)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003e1) N and %; 2) Mean and SD\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAbbreviations:\u0026nbsp;\u003c/em\u003eLEDD - Levodopa Equivalent Daily Dose; DA \u0026ndash; Dopamine Agonist; MoCA - Montreal Cognitive Assessment, FAB - Frontal Assessment Battery, SCD - Subjective Cognitive Decline\u003c/p\u003e\n\u003ch2\u003eUsability testing confirms the feasibility of the DiCo in an elderly population\u003c/h2\u003e\n\u003cp\u003eA tablet-based cognitive assessment (DiCo) was developed by implementing 13 commonly used test paradigms that cover response inhibition, cognitive flexibility, attention, and working memory (\u003cstrong\u003eTable 3\u003c/strong\u003e). To make sure that the DiCo can be used efficiently in an elderly population, we placed particular emphasis on a user-friendly experience and on the clarity of test instructions. Training sessions were implemented to acquaint participants with the use of the tablet computer, in particular with the functionality of the touch input, and practice trials were incorporated into the assessments to provide participants with an opportunity to familiarize themselves with the testing procedures before the actual assessments. Comprehensive usability testing was conducted with a mixed-methods approach that incorporated quantitative measures and heuristic evaluations, which are described in more detail in the supplement.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eQuantitative measures were obtained through the System Usability Scale (SUS)\u003csup\u003e20\u003c/sup\u003e and the User Experience Questionnaire (UEQ)\u003csup\u003e21\u003c/sup\u003e. The SUS is a widely used questionnaire that assesses the overall usability of a system, generating a numeric score that reflects user satisfaction. SUS ratings revealed a high overall satisfaction with the system, while some participants expressed the need for support and some time to get familiar with the use (\u003cstrong\u003eFig. 1c\u003c/strong\u003e). The UEQ, on the other hand, provides an evaluation of the user experience across multiple dimensions, including attractiveness, perspicuity, efficiency, and stimulation. The cognitive testing battery achieved high scores in the UEQ domains of attractiveness (1.88 \u0026plusmn; 0.81), stimulation (2.00 \u0026plusmn; 0.84), and novelty (1.80 \u0026plusmn; 0.98), indicative of a high \u0026ldquo;hedonic quality\u0026rdquo; (\u003cstrong\u003eFig. 1b\u003c/strong\u003e). The rating for the pragmatic quality was above average (indicated by perspicuity, efficiency and dependability, mean 1.36).\u003c/p\u003e\n\u003ch2\u003eDiCo captures performance-based dimensions that are stable across analytic approaches\u003c/h2\u003e\n\u003cp\u003eTo determine whether the extracted DiCo features reflect cognitive domains, task-specific variance, or broader performance dimensions, we conducted three complementary analyses: hierarchical cluster analysis, network modeling and exploratory factor analysis (EFA).\u003c/p\u003e\n\u003cp\u003eThe hierarchical cluster map (\u003cstrong\u003eFig. 2\u003c/strong\u003e) was based on pairwise feature correlations and revealed three interpretable clusters (C) of features: time-based measures such as response time (C1), accuracy-related measures (C2), and a residual cluster mainly comprised of features from BART, Information Sampling and Reversal Learning (C3). Correlation Coefficients were stronger within C1 and C2 than within C3, indicating less internal coherence within the third cluster. Because good performance is associated with short response times and high accuracy, we inverted the time-based measures in C1 to align higher values with better performance. After this transformation, the cluster structure changed and yielded two main clusters (\u003cstrong\u003esupplemental Fig. 1\u003c/strong\u003e): one combining both response time and accuracy measures across cognitive tests, and another residual cluster similar to the original C3.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor the network modeling, we calculated the sparse inverse covariance matrix (precision matrix) using the graphical Lasso algorithm. The precision matrix identifies conditional dependencies between individual variables while controlling for the influence of all others. The resulting network structure showed strong within-test connectivity and sparse connections between tests (\u003cstrong\u003eFig. 3a\u003c/strong\u003e). Among the connections within the same test, 66.7% (n=14 of 21) were realized, whereas only 10.1% (n=36 of 357) of connections across tests were realized.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn \u003cstrong\u003eFig. 3b\u003c/strong\u003e, we color-coded the nodes in the network graph according to the three variable clusters derived from the hierarchical cluster map (\u003cstrong\u003eFig. 2\u003c/strong\u003e). The cluster assignments aligned remarkably well with the structure of the network that was derived from the precision matrix. Particularly, time-based features (C1) formed a densely interconnected core, whereas impulsivity-related measures (C3) appeared largely isolated from the rest of the network. This convergence supports the interpretability and structural validity of the initial cluster solution and suggests that the DiCo captures performance-based dimensions that are stable across analytic approaches.\u003c/p\u003e\n\u003cp\u003eHierarchical clustering captures bivariate associations based on overall similarity, while network analysis estimates conditional dependencies between variables. Both approaches describe the observed variable structure at the surface level \u0026mdash; in terms of correlation or partial correlation \u0026mdash; without modeling the underlying sources of shared variance. To examine whether the DiCo features coalesce into interpretable latent dimensions that could be analogous to the concept of cognitive domains, we conducted an exploratory factor analysis (EFA). EFA explicitly aims to identify latent dimensions that account for the covariance structure, thereby allowing for a theoretically meaningful interpretation of shared cognitive processes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAfter excluding features with low sampling adequacy (Kaiser-Meyer-Olkin Test, KMO \u0026lt; .50), low communalities (\u0026lt; .40) and eigenvalues \u0026lt; 1.0 \u003csup\u003e22\u003c/sup\u003e, the EFA yielded five factors (\u003cstrong\u003eFig. 5a\u003c/strong\u003e). These factors explained 64.6% of the total variance, exceeding the commonly used threshold of 60% in social sciences\u003csup\u003e23\u003c/sup\u003e. The first factor consisted of features from StopSignal, Go-Nogo and the Flanker task, specifically those derived from the \u0026ldquo;Go\u0026rdquo; conditions that did not require inhibition or discrimination. Since those trials primarily capture aspects of attention, we chose to name this factor \u0026ldquo;visual attention\u0026rdquo;. The remaining four factors mostly consisted of variables from the same test and were thus named after the cognitive domain these tests have been associated with: Stroop \u0026ndash; \u003cem\u003eInterference control\u003c/em\u003e, Tower of London \u0026ndash; \u003cem\u003eProblem solving\u003c/em\u003e, Information Sampling \u003cem\u003e\u0026ndash; Decision-Making\u003c/em\u003e, and N-Back \u0026ndash; \u003cem\u003eWorking Memory\u003c/em\u003e. We refer to them as DiCo factors from here on.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe then tested for correlations between DiCo factors. The correlation of \u003cem\u003eDecision Making\u003c/em\u003e with the other factors was small (\u003cstrong\u003eFig. 4b\u003c/strong\u003e), but the remaining four factors were moderately to strongly interrelated with correlation coefficients ranging from r = 0.51 to 0.62. This high degree of interrelatedness suggests a considerable degree of shared variance across cognitive domains.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eConstruct validity of DiCo factors\u003c/h2\u003e\n\u003cp\u003eTo evaluate the construct validity of the identified DiCo factors, we calculated Pearson correlation coefficients with summary scores obtained from neuropsychological tests (NPT) and validated questionnaires (Q) (\u003cstrong\u003eFig. 5\u003c/strong\u003e). We only observed moderate associations; the strongest correlation coefficients with a DiCo-derived factor ranged between r = 0.30 and r = 0.64. Due to the coding schemes of the individual variables (\u003cstrong\u003eTable 3)\u003c/strong\u003e, NPT measures tended to correlate positively with DiCo factors whereas several questionnaire-derived variables were negatively associated.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; The DiCo factor \u003cem\u003eVisual Attention\u003c/em\u003e was moderately associated with the following variables obtained by NPT (\u003cstrong\u003eFig. 5a\u003c/strong\u003e): finding similarities (WIE; r = .36), executive functions (CERAD TMT-B, r = -.34) and semantic verbal fluency (CERAD ANIM, r =.33). \u003cem\u003eInterference Control\u003c/em\u003e was most strongly correlated with executive functioning (CERAD TMT-B, r = -.64). \u003cem\u003eProblem-Solving\u003c/em\u003e was related to semantic verbal fluency (CERAD ANIM, r = .48) and executive functioning (CERAD TMT-B, r = -.52). \u003cem\u003eWorking Memory\u0026nbsp;\u003c/em\u003ewas most strongly related to semantic verbal fluency (CERAD ANIM, r = .43). \u003cem\u003eDecision Making\u003c/em\u003e showed only weak associations with the NPT variables, except for a moderate correlation with vocabulary knowledge (MWT-B, r = .37). This latter association may reflect a response style where participants prematurely accepted options as valid without careful evaluation potentially mirroring increased reflection impulsivity.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor the questionnaire data (\u003cstrong\u003eFig. 5b\u003c/strong\u003e), the DiCo factor \u003cem\u003eVisual Attention\u003c/em\u003e showed moderate negative relationships with conscientiousness (BFI Conscientiousness, r = -.47), functional abilities in everyday life (FAQ total, r = -.45) and lack of perseverance (UPPS Lack Persev., r = -.44). \u003cem\u003eInterference Control\u003c/em\u003e was weakly to moderately related to patients\u0026rsquo; global impression of condition (PGI Severity, r = -.36), fatique (Fatigue FSS, r = -.34) and gambling (ICDRD Gambling, r = -.33). \u003cem\u003eProblem-Solving\u003c/em\u003e was most strongly related to the participants subjective global impression of condition (PGI Severity, r = -.46) and moderately related to lack of perseverance (UPPS Lack Persev., r = -.38). \u003cem\u003eWorking Memory\u003c/em\u003e overall also showed weak correlations to Q-Data variables with urgency (UPPS Urgency, r = -.34) and lack of perseverance (UPPS Lack Persev., r = -.28) being the most prominent. \u003cem\u003eDecision-Making\u0026nbsp;\u003c/em\u003eagain showed only weak correlations to questionnaire variables with urgency (UPPS Urgency, r = -.28) being the strongest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2 \u003cem\u003eComparison of clinical variables between clusters\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eFeature\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eCluster 1\u003c/p\u003e\n \u003cp\u003e(N=13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eCluster 2\u003cbr\u003e\u0026nbsp;(N=53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eCluster 3\u003c/p\u003e\n \u003cp\u003e(N=21)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003eCluster 4\u003cbr\u003e\u0026nbsp;(N=10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003eKruskal-Wallis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003eMann-Whitney U\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003eMann-Whitney U\u003c/p\u003e\n \u003cp\u003e(Holm-corrected)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e60.7 (9.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e60.7 (10.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e66.2 (9.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e69.5 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep= .013\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e1,2 \u0026lt; 3,4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eDisease Duration (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e6.7 (3.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e5.8 (4.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e6.7 (5.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e8.1 (5.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep= .403\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eAge at onset (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e54 (8.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e54.6 (10.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e59.5 (10.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e61.4 (12.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .039\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026lt; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eLEDD Total (mg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e685.6 (283.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e627.3 (355.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e721.8 (345.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e789.6 (298.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .272\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eLEDD LDopa (mg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e420.5 (168.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e355.0 (177.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e496.0 (223.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e603.6 (105.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep \u0026lt; .001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026lt; 3; 1,2 \u0026lt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026lt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eLEDD DA (mg)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e204.4 (106.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e253.3 (109.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e206.4 (158.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e338.5 (300.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .159\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eDA/LEDD Total (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e33.6 (18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e44.6 (22.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e25.6 (17.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e35.2 (32.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .012\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026gt; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026gt; 3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eVascular Risk present (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e23.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e37.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e28.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep =.449\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eUPDRS III\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e18.8 (9.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e17.4 (8.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e21.3 (7.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e25.9 (10.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .047\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026lt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eMoCA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e27.4 (1.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e26.9 (2.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e25.6 (3.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e23.6 (3.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .011\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e1,2 \u0026gt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e2 \u0026gt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eFAB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e16.8 (1.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e17.0 (1.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e16.6 (1.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e13.5 (3.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e1 ,2,3 \u0026gt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003e1,2,3 \u0026gt; 4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 151px;\"\u003e\n \u003cp\u003eSCD (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e38.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e47.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e71.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 109px;\"\u003e\n \u003cp\u003e80.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003ep = .058\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 115px;\"\u003e\n \u003cp\u003en.s.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNote. Kruskal\u0026ndash;Wallis test was used to assess group differences across the four clusters, as the data were not normally distributed and group sizes were unequal. Given the small sample sizes in some clusters, we chose pairwise Mann\u0026ndash;Whitney U tests for post-hoc comparisons, as this approach is less conservative and more robust than Dunn\u0026rsquo;s test in small samples. For each comparison, both the uncorrected and Holm-corrected differences are reported.\u0026nbsp;\u003c/em\u003eAbbreviations\u003cem\u003e:\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e n.s.\u003cem\u003e\u0026nbsp;\u0026ndash; not significant,\u0026nbsp;\u003c/em\u003eLEDD - Levodopa Equivalent Daily Dose; DA \u0026ndash; Dopamine Agonist; MoCA - Montreal Cognitive Assessment, FAB - Frontal Assessment Battery, SCD - Subjective Cognitive Decline\u003c/p\u003e\n\u003ch2\u003eLongitudinal changes primarily affect composite motor score, Working Memory, Problem-Solving and Interference Control\u0026nbsp;\u003c/h2\u003e\n\u003cp\u003eLongitudinal assessments were available for a subset of our cohort. The mean delay between the two assessments was 25 months. To compare cognitive changes across domains and with respect to motor changes, we computed Standardized Response Means (SRMs), defined as the mean change divided by the standard deviation of the change scores (\u003cstrong\u003eFig. 7a\u003c/strong\u003e). Next to SRMs for each DiCo factor, we used the MoCA score as a commonly used cognitive assessment and the motor score UPDRS III. Because the UPDRS III can be affected by dopaminergic medication, we determined, in addition, a compound motor score using Uniform Manifold Approximation and Projection (UMAP). The UMAP-derived motor score exhibited a moderate effect size for change (SRM = .57). The MoCA (SRM = -.36) and \u003cem\u003eWorking Memory (SRM = -.32)\u003c/em\u003e showed a small to moderate effect size whereas \u003cem\u003eProblem-Solving\u003c/em\u003e (SRM = -.19), UPDRS III (SRM = .16) and \u003cem\u003eInterference Control\u003c/em\u003e (SRM = -.11) yielded only very small effect sizes. The observation that \u003cem\u003eProblem Solving\u003c/em\u003e showed the most pronounced changes with age (\u003cstrong\u003eFig. 6d\u003c/strong\u003e) but relatively small longitudinal changes (\u003cstrong\u003eFig. 7e\u003c/strong\u003e) is consistent with the notion that longitudinal trajectories in patients are not only driven by age. Accordingly, the compound motor score showed only a weak correlation with age (r = 0.27; \u003cstrong\u003esupplemental Fig.\u003c/strong\u003e\u003cstrong\u003e \u003c/strong\u003e\u003cstrong\u003e2\u003c/strong\u003e), indicating that it reflects disease duration and progression more strongly than chronological age.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAccordingly, individual trajectories showed stronger changes for the compound motor score (\u003cstrong\u003eFig. 7h\u003c/strong\u003e) than for MoCA (\u003cstrong\u003eFig. 7c\u003c/strong\u003e) and \u003cem\u003eWorking Memory\u003c/em\u003e (\u003cstrong\u003eFig. 7d\u003c/strong\u003e). Paired t-tests showed significant differences between baseline and follow-up values for the compound motor score (t = -2.857, p= .008), but not for UPDRS III (t = -.844, p = .405) or MoCA (t=1.424, p=.165).\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eAssociations of digital cognitive assessments with MoCA and FAB\u003c/h2\u003e\n\u003cp\u003eTo investigate the relationship of factors obtained from the DiCo with overall cognition and executive function, we trained two separate Random Forest Regressors to predict the scores of the Montreal Cognitive Assessment (MoCA, Model 1) and the Frontal Assessment Battery (FAB, Model 2). Factor scores, sociodemographic and clinical variables were used as predictor variables. Hyperparameter tuning was achieved via grid search and 5-fold cross-validation. Model performance and feature importances were evaluated across multiple random seeds using repeated cross-validation (5 folds, 2 repeats), resulting in a robust estimation across 150 model runs. Mean absolute Error (MAE) and Root Mean Squared Error (RMSE) were computed to evaluate prediction performance. To identify the most relevant predictors for MoCA and FAB, our main interest was in the averaged feature importances and their standard deviation across all folds and random seeds.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor Model 1, grid search revealed the following best hyperparameters: n_estimators = 50, min_samples_split=2,min_sample_leaf=4, max_depth=10. Model 1 predicted the MoCA scores with a MAE of 1.87 (SD = 0.31) and a RMSE of 2.21 (SD = 0.39). These values indicate that, on average, the model\u0026apos;s predictions are reasonably close to the actual scores. \u003cem\u003eWorking Memory\u0026nbsp;\u003c/em\u003ewas, by far, the most important factor for predicting the MoCA score (\u003cstrong\u003eFig. 8a\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGrid search yielded n_estimators = 100, min_samples_split = 5, min_samples_leaf = 2 and max_depth = 5 as optimal hyperparameters for predicting the FAB. Model 2 predicted the FAB (\u003cstrong\u003eFig. 8b\u003c/strong\u003e) scores with a MAE of 1.05 (SD = 0.31) and a RMSE of 1.58 (SD = 0.56). In contrast to Model 1, no single feature stood out for predicting the FAB. Total levodopa equivalent dose and \u003cem\u003eInterference Control\u003c/em\u003e were almost equally important for predicting the FAB, followed by \u003cem\u003eProblem-Solving\u003c/em\u003e and \u003cem\u003eDisease Duration\u003c/em\u003e. The more evenly distributed feature importance could be explained, for instance, by collinearity among predictors.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe then trained a second set of random forest regressors\u003cem\u003e\u0026nbsp;\u003c/em\u003eto predict the change in MoCA and the change in the compound motor score. The MoCA and compound motor scores at baseline were the most important features to predict the change in MoCA, followed by the time between the two assessments and UPDRS III at baseline (\u003cstrong\u003eFig. 9a\u003c/strong\u003e). To predict the change in Motor Score, days between the assessments, motor score at baseline and \u003cem\u003eDecision Making\u003c/em\u003e were the most important features (\u003cstrong\u003eFig. 9b\u003c/strong\u003e). Interestingly, baseline \u003cem\u003eDecision-Making\u003c/em\u003e was associated negatively with change in Motor Score (r = -.31, p = .11; Fig. 9d), indicating that higher reflection impulsivity may predict stronger motor decline over time. Given the small number of participants (n = 28) in the longitudinal cohort, these results need to be interpreted with caution.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAfter controlling for baseline values, the residualized MoCA and motor change scores only exhibited a weak correlation (r = -.28, p = .15; \u003cstrong\u003eFig. 9c\u003c/strong\u003e), suggesting that the progression of motor and cognitive performance over time may develop independently from one another.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we developed and validated an open-source tablet-based cognitive assessment battery for PwPD. 54 features were obtained in 13 test paradigms. Three structural analyses were conducted, which converged on the notion that cognitive performance of PwPD without manifest cognitive impairment, as captured by the DiCo, is characterized by strong intercorrelations across domains. Specifically, the main difference between clusters C1 and C2 obtained by correlation-based clustering (\u003cstrong\u003eFig. 2\u003c/strong\u003e) consisted in the direction of changes, and inverting the time-based measures resulted in one cluster combining both response time and accuracy measures across cognitive tests, and another residual cluster similar to the original C3 (\u003cstrong\u003esupplemental Figure 1\u003c/strong\u003e). In the network analysis, features clustered within the same test, but not along other domains (\u003cstrong\u003eFig. 3a\u003c/strong\u003e). The EFA revealed five latent factors, four of which were strongly interrelated (\u003cstrong\u003eFig. 4b\u003c/strong\u003e). These factors also correlated strongly with neuropsychological test results (\u003cstrong\u003eFig. 5a\u003c/strong\u003e) and questionnaire data (\u003cstrong\u003eFig. 5b\u003c/strong\u003e). Clustering of patients showed homogenous differences between 3 of 4 clusters (\u003cstrong\u003eFig. 6a\u003c/strong\u003e), which can be classified as severity-based using the nomenclature by Pourzinal et al.\u003csup\u003e24\u003c/sup\u003e and which are consistent with a gradual decline across all cognitive functions captured by the DiCo.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;These findings are consistent with a mutualistic, network-like structure of cognitive abilities as suggested by recent theoretical developments in cognitive science\u003csup\u003e25\u003c/sup\u003e. The age-dependent decline of the DiCo factors \u003cem\u003eAttention, Working Memory, Interference Control\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;Problem-Solving\u003c/em\u003e (\u003cstrong\u003eFig. 6\u003c/strong\u003e, x-axis)\u0026nbsp;is consistent with the hypothesis that\u0026nbsp;cognitive abilities are interdependent and decline jointly. Correlations between cognitive tasks arise from reciprocal interactions between individual cognitive processes, and cognitive abilities reinforce one another throughout development and aging. The quantitative differences between patient clusters 1, 3 and 4\u0026nbsp;(\u003cstrong\u003eFig. 6\u003c/strong\u003e, color coding) indicate that, in PwPD, cognitive performance declines along the same axes.\u0026nbsp;These results are in line with analyses of cognitive test scores obtained in 698 PwPD from the DEMPARK and LANDSCAPE cohorts\u003csup\u003e26\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;This rather uni-dimensional structure of cognitive test results does not support traditional neuropsychological models, which have conceptualized cognitive abilities as separable domains, including attention, working memory, executive functions, language, memory and visuospatial abilities. This modular view of cognition underlies diagnostic frameworks like the MDS criteria for MCI in PwPD, which require deficits in at least two of these distinct domains\u003csup\u003e9\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThe DiCo factors correlated moderately with established paper-pencil neuropsychological tests (\u003cstrong\u003eFig. 5a\u003c/strong\u003e). Given the substantial methodological differences, this correlation strength is not unexpected. Previous studies comparing paper-pencil versions of a neuropsychological test with its exact digital implementation found correlations of 0.63 for TMT-A, 0.77 for TMT-B, and 0.68 for the Color-Word Interference Test\u003csup\u003e27\u003c/sup\u003e. Assessing the relationship between the processing speed subtests \u003cem\u003eCoding\u003c/em\u003e and \u003cem\u003eSymbol Search\u003c/em\u003e from the Wechsler Intelligence Scale for Children, Ferriola\u003csup\u003e28\u003c/sup\u003e found correlations of 0.67 and 0.61 respectively. In addition, the lower correlation coefficients we observed can be explained by the fact that the DiCo tests are not exact implementations of the neuropsychological tests we related them to. Finally, the DiCo factor \u003cem\u003eVisual Attention\u003c/em\u003e represents a composite of features from three different tests. The apparent lack of domain structure within the DiCo features and the surprisingly poor association with paper-pencil tests and questionnaires is reminiscent of a comparison of digital assessments of self-regulation described by Eisenberg et al\u003csup\u003e29\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe associations we did observe between DiCo factors and paper-pencil tests are theoretically plausible. For example, \u003cem\u003eInterference Control\u003c/em\u003e and the TMT-B have both been linked to executive functioning\u003csup\u003e30,31\u003c/sup\u003e. Similarly, \u003cem\u003eProblem-Solving\u0026nbsp;\u003c/em\u003eand verbal fluency have both been linked to microstructural white matter changes in newly diagnosed PwPD, suggesting a shared neuroanatomical basis\u003csup\u003e32\u003c/sup\u003e. For questionnaires, most associations to DiCo factors were even more moderate (\u003cstrong\u003eFig. 5b\u003c/strong\u003e). Interestingly, the correlation between \u003cem\u003eProblem-Solving\u003c/em\u003e and participants’ subjective impression of disease severity was rather strong, which underlines previous findings that cognitive symptoms play an important role in PwPDs’ quality of life\u003csup\u003e33–35\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThe DiCo factor \u003cem\u003eDecision-Making\u003c/em\u003e consistently emerged as rather independent, being less correlated with the other factors (\u003cstrong\u003eFig. 4b\u003c/strong\u003e), less correlated with neuropsychological test results and questionnaire data (\u003cstrong\u003eFig. 5\u003c/strong\u003e). \u003cem\u003eDecision-Making\u003c/em\u003e did not decline with age (\u003cstrong\u003eFig. 6e\u003c/strong\u003e) and separated the two largest clusters of patients (\u003cstrong\u003eFig. 6a\u003c/strong\u003e): cluster 2 showed similar performance as cluster 1 in most DiCo factors but was significantly impaired in \u003cem\u003eDecision-Making\u003c/em\u003e. The DiCo factor \u003cem\u003eDecision-Making\u003c/em\u003e contributed little to the prediction of MoCA and FAB (\u003cstrong\u003eFig. 7\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDecision-Making\u003c/em\u003e was mainly driven by features of the information sampling test (\u003cstrong\u003eFig. 4a\u003c/strong\u003e). The information sampling test\u0026nbsp;responds to reflection impulsivity, i.e., the tendency to make rapid decisions without sufficient information gathering or consideration of alternatives. Dopamine agonists increase reflection impulsivity\u003csup\u003e36\u003c/sup\u003e and all dopaminergic treatment can cause impulse-control disorders\u003csup\u003e37–39\u003c/sup\u003e. This raises the possibility that the \u003cem\u003eDecision-Making\u0026nbsp;\u003c/em\u003efeature results not from aging or PD, but from dopaminergic medication.\u0026nbsp;Interestingly, \u003cem\u003eDecision-Making\u003c/em\u003e was associated with changes in the composite motor score (\u003cstrong\u003eFig. 10b\u003c/strong\u003e), which includes medication information next to the UPDRS III (\u003cstrong\u003esupplemental Figure 2\u003c/strong\u003e). In addition,\u0026nbsp;we observed a trend where a higher proportion of dopamine agonists in the medication profile was associated with poorer performance on the information sampling task – particularly among participants who performed well on other cognitive tests (cluster 2 vs. cluster 1, \u003cstrong\u003eTable 2\u003c/strong\u003e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn this study, cluster assignment based on DiCo-derived factor scores revealed four subgroups with distinct performance profiles (\u003cstrong\u003eFig. 6a\u003c/strong\u003e). These four clusters were poorly explained by age, disease duration, age at disease onset, presence of cardiovascular risk factors, and motor impairment (\u003cstrong\u003eTable 2\u003c/strong\u003e). Also, the classical tests MoCA and FAB did not differ significantly between clusters 1, 2 and 3 (\u003cstrong\u003eTable 2\u003c/strong\u003e). The digital assessment implemented in this study can therefore identify differences that are not observed with these measurements. This might result from the fact that the DiCo measures slightly different properties than paper-pencil tests (\u003cstrong\u003eFig. 5\u003c/strong\u003e). In addition, digital assessments might provide finer granularity. In this context, the more or less uni-dimensional structure of the DiCo results implies that the DiCo composition could be re-evaluated to reduce redundancies.\u003c/p\u003e\n\u003cp\u003eIn the subset of patients where longitudinal data was available, we observed changes between baseline and follow-up in three DiCo factors. \u003cem\u003eWorking Memory\u003c/em\u003e exhibited a standardized response mean (SRM) comparable to the MoCA, suggesting that the N-Back task, which mainly loaded on \u003cem\u003eWorking Memory\u003c/em\u003e (\u003cstrong\u003eFig. 4a\u003c/strong\u003e) might serve as a simple research tool for longitudinal research. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe very low correlation between residual changes in MoCA and composite motor score (\u003cstrong\u003eFig. 10c\u003c/strong\u003e) underscores the importance of examining cognitive and motor progression as partly independent trajectories in PwPD.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;These effects were not observed when using the UPDRS III score to predict motor progression. In fact, we did not observe a significant difference in UPDRS III scores between baseline and follow-up (\u003cstrong\u003eFig. 8d\u003c/strong\u003e). Similarly, Holden et al. found UPDRS III to increase 4 points per year in unmedicated PD Patients whereas patients on medication progressed only 1.2 points per year. Dopaminergic medication can therefore mask motor progression and reduce the sensitivity of UPDRS III to subtle longitudinal changes. The significant differences observed between baseline and follow-up in the UMAP-based composite score suggest that this approach may offer greater sensitivity to progressive motor changes in PwPD than clinical ratings alone.\u003c/p\u003e\n\u003cp\u003eOur sample of 97 patients primarily included participants in the transition phase of PD without overt cognitive impairment. As for other cohorts in early PD\u003csup\u003e40\u003c/sup\u003e, this potentially limited the detection of dementia-related patterns, but allowed for an analysis of subtle cognitive differences that may already be present in early stages of the disease. For longitudinal assessments in particular, the number of patients included in the study was relatively small and analyses should be interpreted with caution. Because participants were recruited in an outpatient clinic, PwPD with impaired mobility were underrepresented in our study, limiting the generalizability of the feasibility assessments. Moreover, the cohort is quite homogenous with respect to ethnic disparities. Multicenter validation studies are therefore required, and the availability of the app on github along with the test implementations will facilitate such confirmation.\u003c/p\u003e\n\u003cp\u003eIn summary, this study describes a comprehensive digital cognitive assessment that has been extensively usability-tested and can be readily applied for research on cognitive impairment in PwPD. Researchers with specific interests may choose specific tasks identified to correlate with specific cognitive domains based on the associations described here. In addition, our analyses suggest that features extracted from digital assessments can contain information not included in paper-pencil based test. The prognostic value of this data needs to be confirmed in longitudinal studies.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eStudy population and design\u003c/h2\u003e\n\u003cp\u003eThe study was approved by the institutional review board of the Technische Universität Dresden (BO-EK-494112020) and written informed consent was obtained from all participants. Participants were recruited at the Department of Neurology of the Dresden University Hospital. \u0026nbsp; In this observational exploratory study, a sample size of 100 cases was targeted. To ensure a diverse study population, the inclusion criteria for our study were deliberately broad, encompassing individuals aged 18 and above who met the diagnostic criteria for PD as stipulated by the Movement Disorder Society\u003csup\u003e41\u003c/sup\u003e. Individuals with severe cognitive or psychiatric disorders that could impede the successful completion of the cognitive test battery were excluded. Clinical and demographic data were collected, including motor and cognitive status (Hoehn \u0026amp; Yahr stage, UPDRS part III, Montreal cognitive assessment) as well as the presence of disease-related complications.\u003c/p\u003e\n\u003ch2\u003eSelection and implementation of tests\u003c/h2\u003e\n\u003cp\u003eThe selection of digital cognitive tests for our study was informed by a literature review centered on cognition in PD. We systematically examined research findings and consensus-based recommendations and identified specific cognitive domains implicated in PD pathology, with a particular focus on response inhibition, cognitive flexibility, attention, and working memory\u003csup\u003e42\u003c/sup\u003e. A data-driven identification of cognitive subtypes identified two types of clustering models: severity-based and domain-based. The majority of severity-based models revealed three distinct clusters of PwPD. The domain-based models identified two to six clusters with mnestic, visuospatial, executive and attentional impairment being the most common subtypes\u003csup\u003e24\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThe 13 selected tests were implemented using the open-source JavaScript framework \u003cem\u003ejsPsych,\u0026nbsp;\u003c/em\u003eoptimized for touch-input and deployed via an \u003cem\u003eExperiment Factory\u003c/em\u003e Container. To ensure a distraction-free and uniform testing environment, an iOS client application was developed. The app as well as the test implementations are available on github: https://github.com/tfeige91/DiCo-iOS-. All participants conducted the test on an 8th-generation Apple iPad at maximum display brightness.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe app was designed to have participants undergo a brief instructional session to familiarize them with the touch controls prior to the actual experiments. During the instructional session, participants were required to accurately tap a circular area displayed on the screen. Each successful tap was met with immediate feedback, designed to reinforce correct interactions and improve proficiency of the touch control system. This preliminary exercise ensured that all patients were comfortable and adept at using the touch interface, allowing for a consistent starting point in the subsequent experimental tasks.\u003c/p\u003e\n\u003cp\u003eAll participants conducted the tests in a strategically designed order (Table 1) optimized for participant engagement and balanced for cognitive strain. Demanding tests that required high levels of concentration and attention were interleaved by shorter, less strenuous tasks that in the usability testing were perceived as more enjoyable or attractive. This approach was intended to mitigate fatigue and maintain participant motivation throughout the testing session. Additionally, simpler tasks where intense attention and quick reaction times were not the primary focus (e.g. the Information Sampling Task) were included in the second half of the test series.\u003c/p\u003e\n\u003ch2\u003eUsability Testing\u003c/h2\u003e\n\u003cp\u003eUsability testing employed a mixed-methods approach that incorporated heuristic evaluations with experts and quantitative, as well as qualitative measures, in PwPD. The Systems Usability Scale (SUS) provides a subjective assessment of usability by the patient and involves a set of ten Likert scale questions\u003csup\u003e20\u003c/sup\u003e. Participants provide ratings on a scale from 1 to 5 (strongly disagree to strongly agree) for odd-numbered questions and from 5 to 1 (strongly agree to strongly disagree) for even-numbered questions. The SUS-score is calculated by subtracting 1 from the user's response for odd-numbered questions and subtracting the user's response from 5 for even-numbered questions, summing these adjusted scores, and then multiplying the total by 2.5 to yield a score ranging from 0 to 100. Higher scores indicate better perceived usability, with scores above 68 considered above average and scores above 80 considered excellent\u003csup\u003e43\u003c/sup\u003e. The UEQ (User Experience Questionnaire)\u003csup\u003e21\u003c/sup\u003e is a standardized questionnaire used to assess the user experience of products, systems, or services. A total of 26 items measure six key dimensions: Attractiveness, Perspicuity (Clarity), Efficiency, Dependability, Stimulation, and Novelty. Participants rate each item on a seven-point scale, ranging from -3 (extremely bad) to +3 (extremely good). Mean values for each domain are analyzed more frequently than generating an overall score. Values \u0026gt; 0.8 are considered a positive evaluation\u003csup\u003e44\u003c/sup\u003e. In addition, we used a customized version of the UEQ for each test to ensure more precise feedback concerning the difficulty and speed level of each test as well as the perceived informative value of the test.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe testing process followed an iterative design, allowing for step-wise refinement based on feedback from each testing iteration ensuring that it met the specific needs and usability requirements of PD patients and enhanced its overall effectiveness in cognitive assessment.\u003c/p\u003e\n\u003ch2\u003eAnalysis and Feature Extraction\u003c/h2\u003e\n\u003cp\u003eData analysis was conducted using Python V.3.10 and R V4.2.3. The Jupyter notebook is available on Github: https://github.com/tfeige91/DiCo-iOS-.\u003c/p\u003e\n\u003cp\u003e54 features were extracted from the 13 tests, including mean and standard deviation of the reaction time, percentage of correct answers and others. A full list of features can be found in the supplemental text. To mitigate the risk of multicollinearity, we identified pairs of correlations with a threshold of .85 and randomly excluded one variable. A total of four variables from four different tests were thereby excluded from the analysis: flanker_rt_mean_congruent, bart_correctedTotal, kirby_expValue, and StopSignal_correct_go.\u003c/p\u003e\n\u003cp\u003eFurthermore, we clipped features at +/- 3 * interquartile range to minimize the influence of extreme outliers.\u003c/p\u003e\n\u003cp\u003eNext, we calculated Variance Inflation Factors (VIFs) separately for each test and excluded variables with a VIF \u0026gt; 10. This test-wise approach ensured that at least one representative variable from each task remained for subsequent structural data analyses, including cluster mapping, network analysis, and exploratory factor analysis.\u003c/p\u003e\n\u003ch2\u003eCorrelations and Networks\u003c/h2\u003e\n\u003cp\u003eTo detect groups of variables that cohere based on shared variance, we performed a hierarchical agglomerative cluster analysis on the pairwise Pearson correlation matrix of all DiCo-derived variables using average linkage as the linkage criterion and 1 – |r| as the distance metric. The resulting dendrogram and clustered heatmap were generated using the clustermap function from the Seaborn package (version 0.13.2) in Python. This approach groups variables with similar correlation profiles, allowing an intuitive identification of variable clusters that reflect shared cognitive domains or common measurement characteristics. To identify robust relationships between the test variables, we calculated the sparse inverse covariance matrix (precision matrix) using the GraphicalLassoCV function from scikit-learn in Python using default parameters. The sparse inverse covariance matrix reveals information about the conditional dependency of two variables controlling for the effects of all other variables in the dataset. We used the non-diagonal values of the precision matrix to draw a network graph in Gephi, utilizing the force-directed layout algorithm ForceAtlas2\u003csup\u003e46\u003c/sup\u003e. The following tuning parameters were used in the depiction of the final graph: Scaling = 8.0, Gravity = 0.015, Stronger gravity = active.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eFactor Analysis\u003c/h2\u003e\n\u003cp\u003eTo identify the underlying latent cognitive domains that can be measured by the DiCo we performed Exploratory Factor Analysis (EFA). Calculating the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (MSA) yielded an unsatisfactory value of \u0026lt; .60. We therefore followed the method proposed by Kaiser and Rice\u003csup\u003e47\u003c/sup\u003e and calculated the MSA for each variable separately, excluding those having an MSA \u0026lt; .50. This procedure led to the exclusion of 24 features, leaving 30 features. No variable from \u003cem\u003eReversal Learning\u003c/em\u003e, \u003cem\u003eKirby\u003c/em\u003e and \u003cem\u003eBART\u003c/em\u003e remained in the EFA dataset. We then performed EFA with a maximum likelihood estimation and oblique (Promax) rotation. The decision on \u003cem\u003ef\u003c/em\u003e, the number of factors to estimate, was based on the factors’ eigenvalues, keeping those with an eigenvalue \u0026gt; 1.0, which is the commonly used Kaiser’s criterion\u003csup\u003e22\u003c/sup\u003e in factor analysis. We performed EFA and calculated factor scores using the open-source Python module factor_analyzer\u003csup\u003e48\u003c/sup\u003e version 0.5.0.\u003c/p\u003e\n\u003cp\u003eFollowing common best practices\u003csup\u003e49\u003c/sup\u003e we then excluded variables with communalities \u0026lt; .40 and repeated the EFA using the same extraction procedure described above.\u003c/p\u003e\n\u003ch2\u003eClustering\u003c/h2\u003e\n\u003cp\u003eWe performed a latent profile analysis (LPA) on the factor scores to identify possible subgroups. LPA aims to preserve information about participants on an individual level and classifies individuals into distinct groups based on differing combinations of personal and environmental traits. Compared to traditional, non-latent clustering methods (e.g., k-means clustering, hierarchical clustering), LPA treats profile membership as an unobserved categorical variable, where its value indicates which profile an individual belongs to with a certain degree of probability. It thus assigns individuals to clusters based on probabilities estimated directly from the model\u003csup\u003e50\u003c/sup\u003e. That means the researcher does not have to specify a certain number of clusters upfront, but the optimal solution can be based on different goodness of fit indices such as the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC) as well as other measures like entropy, i.e. the goodness of cluster separation (1.0 being the best) and Bootstrap Likelihood Ratio Test (BLRT) that - if significant - suggests the superiority of the k-cluster solution over the k-1-cluster solution. We performed LPA with the R-package tidyLPA\u003csup\u003e51\u003c/sup\u003e. To determine if there are differences between the clusters, we performed a Kruskal-Wallis test followed by Dunn’s test.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003ePredicting screening assessments and feature importance\u003c/h2\u003e\n\u003cp\u003eWe trained a Random Forest regressor using scikit-learn to predict both the MoCA and the FAB using the factor scores derived from the EFA (as described above) as well as age, age at initial diagnosis, disease duration and LEDD. RF is a robust method known for its efficacy in handling complex, non-linear relationships among variables.\u003csup\u003e52\u003c/sup\u003e Hyperparameters were optimized via grid search with 5-fold cross-validation, minimizing the mean absolute error (MAE). Model performance was then evaluated using repeated 5×2-fold cross-validation, with three different random seeds per split to account for variance introduced by both data partitioning and the stochastic nature of the algorithm. For each model, the MAE and root mean squared error (RMSE) were recorded. Feature importances were averaged across all runs, and we additionally quantified their standard deviation due to random forest construction and cross-validation splits.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLongitudinal analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 28 patients completed follow-up assessments using a subset of 6 DiCo tests: \u003cem\u003eSimple Reaction Time Test, Stroop Test, N-Back, Tower of London, Flanker, and the Judgement of Line Orientation Task\u003c/em\u003e. The average interval between baseline and follow-up was 770 days (SD = 236). To evaluate cognitive change over time, we calculated follow-up factor scores for \u003cem\u003eWorking Memory\u003c/em\u003e, \u003cem\u003eInterference Control,\u003c/em\u003e and \u003cem\u003eProblem-Solving\u003c/em\u003e. These were derived using the factor loadings established from the baseline sample and restricted to features from the subset of tasks available at both timepoints. Correlations between these follow-up scores and the full-model factor scores (using the complete feature set) exceeded r = .95, confirming the robustness of the reduced-score estimation. We calculated Pearson correlation coefficients to assess the relationship between the change in the MoCA score and the time between baseline and follow-up.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeatures responsive to longitudinal change\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo quantify the magnitude of longitudinal change, we calculated the Standardized Response Mean (SRM) for each cognitive domain. The SRM is defined as the mean change score between baseline and follow-up divided by the standard deviation of the change scores. It represents an effect size metric commonly used to evaluate sensitivity to change, with values around 0.2, 0.5, and 0.8 typically interpreted as small, moderate, and large effects, respectively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCalculation of a continuous motor score\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo derive a continuous motor score that integrates multiple medication-related and clinical variables, we applied Uniform Manifold Approximation and Projection (UMAP)\u003csup\u003e53,54\u003c/sup\u003e, a non-linear dimensionality reduction technique that preserves the local and global structure of high-dimensional data in a low-dimensional embedding. We selected four variables that reflect motor symptom severity and medication status — levodopa equivalent daily dose from dopamine agonists (LEDD_DA), from levodopa (LEDD_L), total LEDD, and the MDS-UPDRS III motor score — and concatenated data from both baseline and follow-up assessments. The UMAP algorithm was applied to the z-standardized variables with the following parameters: n_components=1, n_neighbors=30, and random_state=24. The resulting one-dimensional embedding was then min-max scaled to a range between 0 and 1, with higher values indicating more severe motor impairment. Since UMAPs do not yield explicit feature weights, we calculated pairwise Pearson correlation coefficients between the input features and the UMAP-derived motor score to assess their individual contributions (\u003cstrong\u003esupplementary Fig. 2\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eWe conducted paired t-tests on both the UPDRS III and the UMAP-derived Motor Score to assess whether significant changes occurred between baseline and follow-up assessments.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentifying baseline features to predict cognitive and motor progression\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo exploratorily identify relevant features for predicting the change in both MoCA and the motor score we calculated feature importances using the same Random Forest Regressor algorithm described above.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRelation of cognitive and motor changes\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo examine the relationship between cognitive and motor changes over time, we calculated the change scores (Δ) for the MoCA and the UMAP-derived motor score by subtracting baseline values from follow-up values. To control for baseline effects, we performed linear regression analyses by regressing follow-up scores on baseline scores for each measure. The residuals from these regressions reflect change independent of baseline performance. We then computed the Pearson correlation between these residuals to assess the association between motor and cognitive change.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo further investigate the role of reflection impulsivity and medication in motor progression, we conducted two linear regression analyses. First, we tested whether baseline Decision-Making performance predicted changes in motor functioning (Δ Motor Score). Second, we examined the association between the proportion of dopamine agonists in the total levodopa equivalent dose at baseline (DA/LEDD Total) and motor progression. In both models, we estimated a linear regression line and computed Pearson correlation coefficients to quantify the strength of the relationship. Results were visualized using regression plots with 95 % confidence intervals.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eLEDD\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eLevodopa Equivalent Daily Dose\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eDA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eDopamine Agonist\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv class=\"DefinitionListEntry\"\u003e\u003cdiv class=\"Term\"\u003eMoCA\u003c/div\u003e\u003cdiv class=\"Description\"\u003e\u003cp\u003eMontreal Cognitive Assessment, FAB-Frontal Assessment Battery, SCD-Subjective Cognitive Decline\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eAuthor contributions: Tim Feige and Anika Frank contributed equally to this work.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTF, AF, JB, BF conceived the study. TF, AF, JB, AE, CH, JK, JJ, NS conducted the research. TF, AF, JB, BF analysed the data. TF, AF, JB, BF wrote the initial draft. HR, BF supervised the research. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting interests: The authors have no competing interests as defined by Nature Portfolio, or other interests that might be perceived to influence the results and/or discussion reported in this paper.\u003c/p\u003e\n\u003cp\u003eData availability: Data is provided within the manuscript or supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAarsland, D. \u003cem\u003eet al.\u003c/em\u003e Parkinson disease-associated cognitive impairment. \u003cem\u003eNat Rev Dis Primers\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 47 (2021).\u003c/li\u003e\n \u003cli\u003eDekker, M. C. J. \u003cem\u003eet al.\u003c/em\u003e Parkinson\u0026rsquo;s Disease Research on the African Continent: Obstacles and Opportunities. \u003cem\u003eFront. Neurol.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 512 (2020).\u003c/li\u003e\n \u003cli\u003eDe Lau, L. M. L., Schipper, C. M. A., Hofman, A., Koudstaal, P. J. \u0026amp; Breteler, M. M. B. Prognosis of Parkinson Disease: Risk of Dementia and Mortality: The Rotterdam Study. \u003cem\u003eArch Neurol\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 1265 (2005).\u003c/li\u003e\n \u003cli\u003eAarsland, D. \u0026amp; Kurz, M. W. The epidemiology of dementia associated with Parkinson disease. \u003cem\u003eJournal of the Neurological Sciences\u003c/em\u003e \u003cstrong\u003e289\u003c/strong\u003e, 18\u0026ndash;22 (2010).\u003c/li\u003e\n \u003cli\u003eMantri, S. \u003cem\u003eet al.\u003c/em\u003e In Their Own Words: Fears Expressed by People with Parkinson\u0026rsquo;s Disease in an Online Symptom Database. \u003cem\u003eJ Parkinsons Dis\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 865\u0026ndash;872 (2024).\u003c/li\u003e\n \u003cli\u003eKjeldsen, P. L. \u0026amp; Damholdt, M. F. Subjective cognitive complaints in patients with Parkinson\u0026rsquo;s disease. \u003cem\u003eActa Neurol Scand\u003c/em\u003e \u003cstrong\u003e140\u003c/strong\u003e, 375\u0026ndash;389 (2019).\u003c/li\u003e\n \u003cli\u003eCammisuli, D. M., Cammisuli, S. M., Fusi, J., Franzoni, F. \u0026amp; Pruneti, C. Parkinson\u0026rsquo;s Disease\u0026ndash;Mild Cognitive Impairment (PD-MCI): A Useful Summary of Update Knowledge. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 303 (2019).\u003c/li\u003e\n \u003cli\u003eDubois, B. \u003cem\u003eet al.\u003c/em\u003e Diagnostic procedures for Parkinson\u0026rsquo;s disease dementia: Recommendations from the movement disorder society task force. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 2314\u0026ndash;2324 (2007).\u003c/li\u003e\n \u003cli\u003eLitvan, I. \u003cem\u003eet al.\u003c/em\u003e Diagnostic criteria for mild cognitive impairment in Parkinson\u0026rsquo;s disease: \u003cem\u003eMovement\u003c/em\u003e Disorder Society Task Force guidelines. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 349\u0026ndash;356 (2012).\u003c/li\u003e\n \u003cli\u003eRoheger, M., Kalbe, E. \u0026amp; Liepelt-Scarfone, I. Progression of Cognitive Decline in Parkinson\u0026rsquo;s Disease. \u003cem\u003eJPD\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 183\u0026ndash;193 (2018).\u003c/li\u003e\n \u003cli\u003eKehagia, A. A., Barker, R. A. \u0026amp; Robbins, T. W. Cognitive Impairment in Parkinson\u0026rsquo;s Disease: The Dual Syndrome Hypothesis. \u003cem\u003eNeurodegener Dis\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 79\u0026ndash;92 (2013).\u003c/li\u003e\n \u003cli\u003eJain, A. K., Murty, M. N. \u0026amp; Flynn, P. J. Data clustering: a review. \u003cem\u003eACM Comput. Surv.\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 264\u0026ndash;323 (1999).\u003c/li\u003e\n \u003cli\u003eD\u0026uuml;zel, E., Thyrian, J. R. \u0026amp; Berron, D. Innovation in der Diagnostik \u0026ndash; mobile Technologien. \u003cem\u003eNervenarzt\u003c/em\u003e \u003cstrong\u003e90\u003c/strong\u003e, 914\u0026ndash;920 (2019).\u003c/li\u003e\n \u003cli\u003eNasreddine, Z. S. \u003cem\u003eet al.\u003c/em\u003e The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. \u003cem\u003eJournal of the American Geriatrics Society\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 695\u0026ndash;699 (2005).\u003c/li\u003e\n \u003cli\u003eDubois, B., Slachevsky, A., Litvan, I. \u0026amp; Pillon, B. The FAB: A frontal assessment battery at bedside. \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 1621\u0026ndash;1626 (2000).\u003c/li\u003e\n \u003cli\u003eSochat, V. The Experiment Factory: Reproducible Experiment Containers. \u003cem\u003eJOSS\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 521 (2018).\u003c/li\u003e\n \u003cli\u003eDe Leeuw, J. R., Gilbert, R. A. \u0026amp; Luchterhandt, B. jsPsych: Enabling an Open-Source CollaborativeEcosystem of Behavioral Experiments. \u003cem\u003eJOSS\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 5351 (2023).\u003c/li\u003e\n \u003cli\u003eAlonso-Canovas, A. \u003cem\u003eet al.\u003c/em\u003e The Early Treatment Phase in Parkinson\u0026rsquo;s Disease: Not a Honeymoon for All, Not a Honeymoon at All? \u003cem\u003eJournal of Parkinson\u0026rsquo;s Disease\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 323\u0026ndash;328 (2023).\u003c/li\u003e\n \u003cli\u003eFillenbaum, G. G. \u0026amp; Mohs, R. CERAD (Consortium to Establish a Registry for Alzheimer\u0026rsquo;s Disease) Neuropsychology Assessment Battery: 35 Years and Counting. \u003cem\u003eJAD\u003c/em\u003e \u003cstrong\u003e93\u003c/strong\u003e, 1\u0026ndash;27 (2023).\u003c/li\u003e\n \u003cli\u003eBrooke, J. System usability scale (SUS): a quick-and-dirty method of system evaluation user information. \u003cem\u003eReading, UK: Digital equipment co ltd\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 1\u0026ndash;7 (1986).\u003c/li\u003e\n \u003cli\u003eLaugwitz, B., Held, T. \u0026amp; Schrepp, M. Construction and Evaluation of a User Experience Questionnaire. in \u003cem\u003eHCI and Usability for Education and Work\u003c/em\u003e (ed. Holzinger, A.) 63\u0026ndash;76 (Springer Berlin Heidelberg, Berlin, Heidelberg, 2008).\u003c/li\u003e\n \u003cli\u003eKaiser, H. F. The Application of Electronic Computers to Factor Analysis. \u003cem\u003eEducational and Psychological Measurement\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 141\u0026ndash;151 (1960).\u003c/li\u003e\n \u003cli\u003eHair, J. F., Black, W. C., Babin, B. J., Anderson, R. E. \u0026amp; Tatham, R. L. Multivariate data analysis. Uppersaddle River. \u003cem\u003eMultivariate Data Analysis (5th ed) Upper Saddle River\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 207\u0026ndash;219 (1998).\u003c/li\u003e\n \u003cli\u003ePourzinal, D. \u003cem\u003eet al.\u003c/em\u003e Systematic review of data‐driven cognitive subtypes in Parkinson disease. \u003cem\u003eEuro J of Neurology\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 3395\u0026ndash;3417 (2022).\u003c/li\u003e\n \u003cli\u003eVan Der Maas, H. L. J. \u003cem\u003eet al.\u003c/em\u003e A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. \u003cem\u003ePsychological Review\u003c/em\u003e \u003cstrong\u003e113\u003c/strong\u003e, 842\u0026ndash;861 (2006).\u003c/li\u003e\n \u003cli\u003eBalzer-Geldsetzer, M. \u003cem\u003eet al.\u003c/em\u003e Parkinson\u0026rsquo;s Disease and Dementia: A Longitudinal Study (DEMPARK). \u003cem\u003eNeuroepidemiology\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 168\u0026ndash;176 (2011).\u003c/li\u003e\n \u003cli\u003eLatendorf, A., Runde, L. M., Salminen, T. \u0026amp; Steinert, A. Digitization of neuropsychological diagnostics: a pilot study to compare three paper-based and digitized cognitive assessments. \u003cem\u003eAging Clin Exp Res\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1585\u0026ndash;1597 (2021).\u003c/li\u003e\n \u003cli\u003eFerriola, A. M. A COMPARISON OF THE WISC-V PROCESSING SPEED SUBTESTS: PAPER-PENCIL AND DIGITAL FORMATS. (2019).\u003c/li\u003e\n \u003cli\u003eEisenberg, I. W. \u003cem\u003eet al.\u003c/em\u003e Uncovering the structure of self-regulation through data-driven ontology discovery. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 2319 (2019).\u003c/li\u003e\n \u003cli\u003eScarpina, F. \u0026amp; Tagini, S. The Stroop Color and Word Test. \u003cem\u003eFront. Psychol.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2017).\u003c/li\u003e\n \u003cli\u003eArbuthnott, K. \u0026amp; Frank, J. Trail Making Test, Part B as a Measure of Executive Control: Validation Using a Set-Switching Paradigm. \u003cem\u003eJournal of Clinical and Experimental Neuropsychology\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 518\u0026ndash;528 (2000).\u003c/li\u003e\n \u003cli\u003eDuncan, G. W. \u003cem\u003eet al.\u003c/em\u003e Gray and white matter imaging: A biomarker for cognitive impairment in early P arkinson\u0026rsquo;s disease? \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 103\u0026ndash;110 (2016).\u003c/li\u003e\n \u003cli\u003eHenry, K. \u003cem\u003eet al.\u003c/em\u003e Cognitive Predictors of Quality of Life in Parkinson\u0026rsquo;s Disease: A Three-year Longitudinal Study (P4-3.018). \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, 2615 (2024).\u003c/li\u003e\n \u003cli\u003eTang, Y. \u003cem\u003eet al.\u003c/em\u003e Cognitive Function and Quality of Life in Parkinson\u0026rsquo;s Disease: A Cross-Sectional Study. \u003cem\u003eJPD\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1209\u0026ndash;1216 (2020).\u003c/li\u003e\n \u003cli\u003eLawson, R. A. \u003cem\u003eet al.\u003c/em\u003e Cognitive decline and quality of life in incident Parkinson\u0026rsquo;s disease: The role of attention. \u003cem\u003eParkinsonism \u0026amp; Related Disorders\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 47\u0026ndash;53 (2016).\u003c/li\u003e\n \u003cli\u003eDjamshidian, A. \u003cem\u003eet al.\u003c/em\u003e Dopamine Agonists Rather than Deep Brain Stimulation Cause Reflection Impulsivity in Parkinson\u0026rsquo;s Disease. \u003cem\u003eJournal of Parkinson\u0026rsquo;s Disease\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 139\u0026ndash;144 (2013).\u003c/li\u003e\n \u003cli\u003eWeintraub, D. \u003cem\u003eet al.\u003c/em\u003e Association of Dopamine Agonist Use With Impulse Control Disorders in Parkinson Disease. \u003cem\u003eArch Neurol\u003c/em\u003e \u003cstrong\u003e63\u003c/strong\u003e, 969 (2006).\u003c/li\u003e\n \u003cli\u003eGrosset, K. A. \u003cem\u003eet al.\u003c/em\u003e Problematic gambling on dopamine agonists: Not such a rarity. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 2206\u0026ndash;2208 (2006).\u003c/li\u003e\n \u003cli\u003eAntonelli, F. \u003cem\u003eet al.\u003c/em\u003e Dopamine‐agonists and impulsivity in Parkinson\u0026rsquo;s disease: Impulsive choices vs. impulsive actions. \u003cem\u003eHuman Brain Mapping\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 2499\u0026ndash;2506 (2014).\u003c/li\u003e\n \u003cli\u003eFoltynie, T., Brayne, C. E. G., Robbins, T. W. \u0026amp; Barker, R. A. The cognitive ability of an incident cohort of Parkinson\u0026rsquo;s patients in the UK. The CamPaIGN study. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 550\u0026ndash;560 (2004).\u003c/li\u003e\n \u003cli\u003ePostuma, R. B. \u003cem\u003eet al.\u003c/em\u003e MDS clinical diagnostic criteria for Parkinson\u0026rsquo;s disease: MDS-PD Clinical Diagnostic Criteria. \u003cem\u003eMov Disord.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1591\u0026ndash;1601 (2015).\u003c/li\u003e\n \u003cli\u003eLitvan, I. \u003cem\u003eet al.\u003c/em\u003e MDS task force on mild cognitive impairment in Parkinson\u0026rsquo;s disease: Critical review of PD‐MCI. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1814\u0026ndash;1824 (2011).\u003c/li\u003e\n \u003cli\u003eSauro, J. \u003cem\u003eA Practical Guide to the System Usability Scale: Background, Benchmarks \u0026amp; Best Practices\u003c/em\u003e. (Measuring Usability LLC, 2011).\u003c/li\u003e\n \u003cli\u003eSchrepp, M. User experience questionnaire handbook: All you need to know to apply the ueq successfully in your projects (2015). \u003cem\u003eURL http://www. ueq. online. org\u003c/em\u003e (2022).\u003c/li\u003e\n \u003cli\u003eNielsen, J. Heuristic evaluation. in \u003cem\u003eUsability inspection methods\u003c/em\u003e 25\u0026ndash;62 (John Wiley \u0026amp; Sons, Inc., 1994).\u003c/li\u003e\n \u003cli\u003eJacomy, M., Venturini, T., Heymann, S. \u0026amp; Bastian, M. ForceAtlas2, a Continuous Graph Layout Algorithm for Handy Network Visualization Designed for the Gephi Software. \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e98679 (2014).\u003c/li\u003e\n \u003cli\u003eKaiser, H. F. \u0026amp; Rice, J. Little Jiffy, Mark Iv. \u003cem\u003eEducational and Psychological Measurement\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 111\u0026ndash;117 (1974).\u003c/li\u003e\n \u003cli\u003eBiggs, J. factor_analyzer Documentation.\u003c/li\u003e\n \u003cli\u003eCostello, A. B. \u0026amp; Osborne, J. W. Best Practices in Exploratory Factor Analysis: Four Recommendations for Getting the Most From Your Analysis. \u003cem\u003eExploratory Factor Analysis\u003c/em\u003e.\u003c/li\u003e\n \u003cli\u003eSpurk, D., Hirschi, A., Wang, M., Valero, D. \u0026amp; Kauffeld, S. Latent profile analysis: A review and \u0026ldquo;how to\u0026rdquo; guide of its application within vocational behavior research. \u003cem\u003eJournal of Vocational Behavior\u003c/em\u003e \u003cstrong\u003e120\u003c/strong\u003e, 103445 (2020).\u003c/li\u003e\n \u003cli\u003eRosenberg, J., Beymer, P., Anderson, D., Van Lissa, C. j. \u0026amp; Schmidt, J. tidyLPA: An R Package to Easily Carry Out Latent Profile Analysis (LPA) Using Open-Source or Commercial Software. \u003cem\u003eJOSS\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 978 (2018).\u003c/li\u003e\n \u003cli\u003eFife, D. A. \u0026amp; D\u0026rsquo;Onofrio, J. Common, uncommon, and novel applications of random forest in psychological research. \u003cem\u003eBehav Res\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 2447\u0026ndash;2466 (2022).\u003c/li\u003e\n \u003cli\u003eMcInnes, L., Healy, J. \u0026amp; Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. Preprint at https://doi.org/10.48550/ARXIV.1802.03426 (2018).\u003c/li\u003e\n \u003cli\u003eHealy, J. \u0026amp; McInnes, L. Uniform manifold approximation and projection. \u003cem\u003eNat Rev Methods Primers\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 82 (2024).\u003c/li\u003e\n \u003cli\u003eLejuez, C. W. \u003cem\u003eet al.\u003c/em\u003e Evaluation of a behavioral measure of risk taking: The Balloon Analogue Risk Task (BART). \u003cem\u003eJournal of Experimental Psychology: Applied\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 75\u0026ndash;84 (2002).\u003c/li\u003e\n \u003cli\u003eCools, R., Clark, L., Owen, A. M. \u0026amp; Robbins, T. W. Defining the Neural Mechanisms of Probabilistic Reversal Learning Using Event-Related Functional Magnetic Resonance Imaging. \u003cem\u003eJ. Neurosci.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 4563\u0026ndash;4567 (2002).\u003c/li\u003e\n \u003cli\u003eBenton, A. L. Visuospatial Judgment: A Clinical Test. \u003cem\u003eArch Neurol\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 364 (1978).\u003c/li\u003e\n \u003cli\u003eWechsler Memory Scale - Fourth Edition: Deutsche \u0026Uuml;bersetzung und Adaptation der WMS-IV von David Wechsler. (2015).\u003c/li\u003e\n \u003cli\u003eSchretlen, D., Bobholz, J. H. \u0026amp; Brandt, J. Development and psychometric properties of the brief test of attention. \u003cem\u003eThe Clinical Neuropsychologist\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 80\u0026ndash;89 (1996).\u003c/li\u003e\n \u003cli\u003eWarrington, E. K. \u0026amp; James, M. \u003cem\u003eTestbatterie F\u0026uuml;r Visuelle Objekt- Und Raum- Wahrnehmung (Deutsche \u0026Uuml;bersetzung von Beckers K, Canavan A)\u003c/em\u003e. (Thames Valley Test Company, Bury St. Edmunds, 1992).\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWechsler Adult Intelligence Scale - Fourth Edition.\u0026nbsp;\u003c/em\u003e\u003cem\u003eDeutschsprachige Adaptation Der WAIS-IV von D. Wechsler\u003c/em\u003e. (Pearson Assessment \u0026amp; Information GmbH, Frankfurt/M, 2012).\u003c/li\u003e\n \u003cli\u003eStorch, A. \u003cem\u003eet al.\u003c/em\u003e Non-motor Symptoms Questionnaire und Scale f\u0026uuml;r das idiopathische Parkinson-Syndrom: Interkulturell adaptierte Versionen in deutscher Sprache. \u003cem\u003eNervenarzt\u003c/em\u003e \u003cstrong\u003e81\u003c/strong\u003e, 980\u0026ndash;985 (2010).\u003c/li\u003e\n \u003cli\u003eTrenkwalder, C. \u003cem\u003eet al.\u003c/em\u003e Parkinson\u0026rsquo;s disease sleep scale\u0026mdash;validation of the revised version PDSS‐2. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 644\u0026ndash;652 (2011).\u003c/li\u003e\n \u003cli\u003eBloch, K. E., Schoch, O. D., Zhang, J. N. \u0026amp; Russi, E. W. German Version of the Epworth Sleepiness Scale. \u003cem\u003eRespiration\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 440\u0026ndash;447 (1999).\u003c/li\u003e\n \u003cli\u003eStarkstein, S. E. \u003cem\u003eet al.\u003c/em\u003e Reliability, validity, and clinical correlates of apathy in Parkinson\u0026rsquo;s disease. \u003cem\u003eJ Neuropsychiatry Clin Neurosci\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 134\u0026ndash;139 (1992).\u003c/li\u003e\n \u003cli\u003eKrupp, L. B., LaRocca, N. G., Muir-Nash, J. \u0026amp; Steinberg, A. D. The fatigue severity scale. Application to patients with multiple sclerosis and systemic lupus erythemato-sus. \u003cem\u003eArch. Neurol\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 1121\u0026ndash;1123 (1989).\u003c/li\u003e\n \u003cli\u003eBeck, A. T., Steer, R. Ab. \u0026amp; K, G. \u003cem\u003eBDI-II, Beck Depression Inven-Tory\u003c/em\u003e\u003cem\u003e:Manual San Antonio, Tex\u003c/em\u003e. (Psychological Corp, 1996).\u003c/li\u003e\n \u003cli\u003ePfeffer, R. I., Kurosaki, T. T., Harrah, C. H., Chance, J. M. \u0026amp; Filos, S. Measurement of Functional Activities in Older Adults in the Community. \u003cem\u003eJournal of Gerontology\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 323\u0026ndash;329 (1982).\u003c/li\u003e\n \u003cli\u003eRabin, R. \u0026amp; Charro, F. D. EQ-SD: a measure of health status from the EuroQol Group. \u003cem\u003eAnnals of Medicine\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 337\u0026ndash;343 (2001).\u003c/li\u003e\n \u003cli\u003eJenkinson, C., Fitzpatrick, R., Peto, V., Greenhall, R. \u0026amp; Hyman, N. The PDQ-8: Development and validation of a short-form parkinson\u0026rsquo;s disease questionnaire. \u003cem\u003ePsychology \u0026amp; Health\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 805\u0026ndash;814 (1997).\u003c/li\u003e\n \u003cli\u003eRammstedt, B., Kemper, C. J., Klein, M. C., Beierlein, C. \u0026amp; Kovaleva, A. A Short Scale for Assessing the Big Five Dimensions of Personality: 10 Item Big Five Inventory (BFI-10). \u003cem\u003emethods, data, analyses\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 17 Pages (2013).\u003c/li\u003e\n \u003cli\u003eBayard, S. \u003cem\u003eet al.\u003c/em\u003e A multidimensional approach to impulsivity in Parkinson\u0026rsquo;s disease: measurement and structural invariance of the UPPS Impulsive Behaviour Scale. \u003cem\u003ePsychol. Med.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 2931\u0026ndash;2941 (2016).\u003c/li\u003e\n \u003cli\u003eSchmidt, R. E., Gay, P., d\u0026rsquo;Acremont, M. \u0026amp; Van Der Linden, M. A German Adaptation of the UPPS Impulsive Behavior Scale: Psychometric Properties and Factor Structure. \u003cem\u003eSwiss Journal of Psychology\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 107\u0026ndash;112 (2008).\u003c/li\u003e\n \u003cli\u003eWeintraub, D. \u003cem\u003eet al.\u003c/em\u003e Questionnaire for impulsive‐compulsive disorders in Parkinson\u0026rsquo;s Disease\u0026ndash;Rating Scale. \u003cem\u003eMovement Disorders\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 242\u0026ndash;247 (2012).\u003c/li\u003e\n \u003cli\u003eGuy, W. \u003cem\u003eECDEU Assessment Manual for Psychopharmacology\u003c/em\u003e. (US Department of Health, Education, and Welfare, Public Health Service \u0026hellip;, 1976).\u003c/li\u003e\n \u003cli\u003ePatton, J. H., Stanford, M. S. \u0026amp; Barratt, E. S. Factor structure of the barratt impulsiveness scale. \u003cem\u003eJ. Clin. Psychol.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 768\u0026ndash;774 (1995).\u003c/li\u003e\n \u003cli\u003eFoa, E. B. \u003cem\u003eet al.\u003c/em\u003e The Obsessive-Compulsive Inventory: Development and validation of a short version. \u003cem\u003ePsychological Assessment\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 485\u0026ndash;496 (2002).\u003c/li\u003e\n \u003cli\u003eBaumann‐Vogel, H., Valko, P. O., Eisele, G. \u0026amp; Baumann, C. R. Impulse control disorders in Parkinson\u0026rsquo;s disease: don\u0026rsquo;t set your mind at rest by self‐assessments. \u003cem\u003eEuro J of Neurology\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 603\u0026ndash;609 (2015).\u003cu\u003e\u003cbr\u003e\u003c/u\u003e\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 3 is available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Parkinson’s disease, cognition, digital, open source","lastPublishedDoi":"10.21203/rs.3.rs-7585530/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7585530/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCognitive impairment represents an important burden for patients with Parkinson\u0026rsquo;s disease (PwPD). Digital tools may improve accessibility and provide richer assessments than traditional paper\u0026ndash;pencil tests; open-source generation can support the availability of comparable assessments in different cohorts. We therefore implemented a digital, tablet-based cognitive assessment (DiCo) comprising 13 commonly used tests as an open-source tool. 97 PwPD without overt cognitive impairment (43% women) completed the entire DiCo. Clustering of feature correlations and conditional dependencies indicated a predominantly mutual organization of cognitive performance in PwPD. Exploratory factor analysis identified five interrelated latent factors, most of which were derived from single tests. Factors correlated moderately with traditional neuropsychological tests and questionnaires. Machine learning identified working memory as the most predictive features of the MoCA. Latent profile analysis revealed four cognitive subgroups, mainly reflecting severity, with one group characterized by selective reflection impulsivity. Exploratory longitudinal analyses suggest partly independent trajectories of cognitive and motor progression, with a data-driven composite score detecting changes not captured by the MDS UPDRS III. Taken together, the DiCo demonstrated good feasibility in PD, and individual tests might be sufficient to substitute for MoCA or FAB in a research context.\u003c/p\u003e","manuscriptTitle":"Cognitive structure and progression in Parkinson’s Disease: Insights from a tablet-based assessment","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-15 17:02:15","doi":"10.21203/rs.3.rs-7585530/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"019092b7-c90e-4755-80c6-7e6d6afee1b0","owner":[],"postedDate":"October 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":56237062,"name":"Health sciences/Health care"},{"id":56237063,"name":"Health sciences/Neurology"},{"id":56237064,"name":"Biological sciences/Neuroscience"},{"id":56237065,"name":"Biological sciences/Psychology"},{"id":56237066,"name":"Social science/Psychology"}],"tags":[],"updatedAt":"2026-05-03T14:55:27+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-15 17:02:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7585530","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7585530","identity":"rs-7585530","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}