As global life expectancy increases, the disease burden of age-associated neurodegenerative diseases, such as Parkinson disease (PD), is rising [1]. PD is the most common age-related motor neurodegenerative disorder [2], characterized primarily by motor symptoms [3]. However, PD also leads to nonmotor symptoms, such as a decline in cognition [4]. This decline often remains covert and goes unnoticed by clinicians and people with PD [4]. Despite the importance of cognitive evaluations for people with PD, current in-person assessment tools present several challenges.

Traditional cognitive screening tools for PD have long faced challenges related to accessibility, scalability, and PD specificity [5]. Accessibility is a persistent challenge, particularly for people with PD who face barriers to in-person cognitive assessments [5-7]. It is often challenging for people with PD to attend in-person cognitive assessments, especially those with severe motor and mobility impairments [8]. Motor impairments, including bradykinesia and postural instability, can make travel particularly challenging. In addition, PD-related fatigue may reduce their capacity to participate in testing [9,10]. Cognitive deficits may further complicate participation, as people with PD may struggle with task switching [11], working memory [12], visuospatial abilities [13], and impulsivity [14]. Thus, people with PD often find it challenging to navigate and commute to laboratories and clinics.

Beyond accessibility concerns, scalability is limited by the fact that most cognitive assessments must be administered by trained professionals, such as neuropsychologists or neurologists, who are often in high demand [15]. This issue is especially pronounced for people with PD living in remote areas with limited specialist access [5]. Although telehealth approaches can extend clinician-led cognitive assessments to remote settings, they remain constrained by specialist availability, scheduling burdens, and clinician time, limiting their scalability for large-scale screening [16]. Furthermore, recruitment challenges often result in small sample sizes (frequently fewer than 30 participants), drawn from single institutions and limited geographic regions [5], raising concerns about sampling bias and the generalizability of findings [17].

Regarding specificity, commonly used cognitive screening tools such as the Montreal Cognitive Assessment (MoCA) [18] and the Mini-Mental State Examination(S) [19] are generic and not tailored to the cognitive profile of PD. Although tasks such as clock drawing and cube copying are intended to measure visuospatial and executive functions—both relevant to cognitive decline in PD—their reliance on motor execution may confound interpretation, since motor impairments can limit performance independently of cognitive status.

We propose using a web-based service to overcome these barriers. Recent work has explored the feasibility of web-based cognitive assessments to enhance accessibility and scalability. Behavioral researchers in different domains have been relying on the internet to reach larger [20] and more diverse populations [21]. Although self-administered cognitive tasks have been evaluated in neurotypical controls (NT) [22-25] and individuals with mild cognitive impairment (MCI) [26], relatively few studies have examined their usability and validity in PD [27-31]. Given the cognitive profile of PD [32] even before PD-dementia, it is important to use sensitive tools capable of detecting cognitive differences between PD and NT, as well as between participants with PD without MCI (PD-non-MCI) and participants with PD with MCI (PD-MCI). Thus, aiming to further address these accessibility, scalability, and PD-specificity challenges, we developed the Online Rapid Cognitive Assessment for Parkinson’s Disease (ORCA-PD), a fully self-administered, web-based cognitive screening tool. ORCA-PD was tailored to capture early- to mid-stage PD-specific cognitive impairments without increasing motor demands for in-person visits.

Here, we aimed to evaluate the feasibility and preliminary validity of ORCA-PD as a remote, self-administered cognitive screening tool tailored for PD. First, we evaluated whether ORCA-PD can be feasibly completed outside the clinic in diverse geographical locations and whether users find it usable. Second, to situate ORCA-PD within existing assessment practices, we examined its convergent validity with the MoCA. Finally, we tested its ability to distinguish neurotypical individuals from people with PD, as well as PD-non-MCI from PD-MCI cohorts.

On average, participants completed the ORCA-PD in 15 (SD 4.5; range 12‐25) minutes. This reflects task completion time only and does not include presession setup (eg, audio or video checks) or any additional interaction time. Even with this distinction, this reflects minimal participant burden and supports the ORCA-PD’s usability for remote assessment. Participant feedback further supported usability: median ratings indicated that instructions were very clear (5, IQR=1), the display size was comfortable (5, IQR=0), the overall experience was enjoyable (4, IQR=1), and the completion time was good (5, IQR=1) (Figure 2).

To assess the ORCA-PD’s sensitivity to expected patterns of task difficulty, we evaluated whether participant performance declined with increasing task difficulty. Such a pattern would support the expected performance pattern (ie, sanity check) of the online tasks in measuring the intended cognitive constructs (Figure 3). As expected, in 3 versions of the digit span task, participants performed significantly better on the easy level (level 1) than on the hard level (level 2) trials (forward: t₁₁₁=8.99, P<.001; Cohen d=0.85; backward: t₁₁₀=10.20, P<.001; Cohen d=0.97; order: t₁₁₁=11.1, P<.001; Cohen d=1.05), supporting the expected performance pattern of these online tasks. In the LOM task, participants performed significantly better on the easy trials (t₁₁₁=12.18, P<.001; Cohen d=1.15). In addition, easy trials showed high accuracy (98%), whereas the hard trials showed lower mean accuracy (76.6%). In the ATS, participants performed significantly better on repeat (easy) trials than on switch (hard) trials (t₁₁₁=6.23, P<.001; Cohen d=0.59).

Regarding questionnaires, for impulsivity, internal consistency was medium (Cronbach α=0.68; McDonald ω=0.80). For the self-reported SCC questionnaire, internal consistency was good (Cronbach α=0.83; McDonald ω=0.91). For the informant-reported SCC questionnaire, internal consistency was strong (Cronbach α=0.90; McDonald ω=0.92).

To examine relationships among the different ORCA-PD measures, we computed a correlation matrix of all ORCA-PD components across participants (Figure 4). The heatmap below displays interfeature correlations (range –0.04 to 0.70), which is consistent with the multidimensional structure of the ORCA-PD.

Additionally, to evaluate the discriminant validity of ORCA-PD and its ability to dissociate cognitive from motor functions, correlations were examined with a motor measure, the MDS-UPDRS III. As expected, no significant correlations were found between any ORCA-PD measures and the MDS-UPDRS III scores, supporting the notion that ORCA-PD primarily assesses cognitive abilities (DSF: ρ=–0.25, P=.11; DSB: ρ=–0.19, P=.25; DSO: ρ=–0.24, P=.14; ATS: ρ=–0.07, P=.67; and LOM: ρ=–0.09, P=.57).

Importantly, to assess convergent validity, we examined the correlation between the ORCA-PD naive score and MoCA scores (Figure 5, n=56). As expected, across both groups, a significant positive correlation was observed (ρ=0.45, 95% CI 0.21-0.64; P<.001), indicating a moderate association. When examined separately, significant associations were observed in both the PD (ρ=0.41, 95% CI 0.12-0.64; P=.004) and NT (ρ=0.62, 95% CI 0.15-0.86; P=.007) groups, supporting the preliminary validity of ORCA-PD.

To examine the ability of ORCA-PD to distinguish between people with or without MCI, we classified participants according to established criteria for MCI (ie, MoCA between 19 and 25) [54]. In a linear regression analysis adjusting for education and age, neither education (β=0.37; P=.34) nor age (β=–0.27; P=.18) was a significant predictor of ORCA-PD naive scores. As expected, the non-MCI group demonstrated higher ORCA-PD naive scores compared to the MCI group (β=6.07; P=.046). When restricting the analysis to PD only, neither education (β=0.37; P=.40) nor age (β=–0.23; P=.40) was a significant predictor of ORCA-PD naive scores. As expected, the non-MCI group again demonstrated higher ORCA-PD naive scores compared to the MCI group (β=7.22; P=.046). This analysis was not conducted on NT alone, as there were fewer than 5 participants in this group with MCI. These results support the discriminative ability of the ORCA-PD, demonstrating that it can capture cognitive impairments.

Importantly, the ORCA-PD’s discriminative ability was also assessed by comparing ORCA-PD scores between PD and NT (Figure 6). In the regression model, neither education (β=0.27; P=.36) nor age (β=–0.19; P=.16) was a significant predictor of ORCA-PD naive scores. As expected, the group effect was significant (β=–4.18; P=.048), with PD scoring lower than NT.

To explore the specific cognitive domains contributing to the group differences, the NT and PD groups on all individual measures were compared. NT participants performed significantly better than PD on DSO accuracy (t₁₀₇=2.67, P=.004). In addition, NT participants scored significantly higher in impulsivity than participants with PD (t₁₁₀=2.15, P=.02). No significant group differences were found on DSF (t₁₁₀=0.16, P=.44), DSB (t₁₀₉=0.87, P=.19), ATS (t₁₁₀=0.27, P=.39), LOM (t₁₁₀=–0.06, P=.52), or SCC measures (self: t₁₁₀=1.28, P=.10 and other: t₁₁₀=0.44, P=.33).

As an exploratory analysis, which needs to be further validated in future studies, we compared scores between English- and Hebrew-speaking cohorts. At the level of overall ORCA-PD naive scores, differences between English- and Hebrew-speaking participants were not statistically significant in the NT group (t_27.49=2.02, P=.05). However, in the PD group, the English-speaking cohort had significantly higher scores (t_31.67=2.19, P=.04). At the task level, language-related differences were observed for digit span forward in the NT group, where performance was significantly higher in the English-speaking cohort (t_20.16=2.90, P=.009). In the ATS task in the PD group, performance was higher in the English-speaking cohort (t_36.49=2.19, P=.04). For all other measures, no significant language differences were observed in the NT group (DSB: P=.68; DSO: P=.19; LOM: P=.29; ATS: P=.17; impulsivity: P=.29; SCC-self: P=.79; SCC-other: P=.96) or in the PD group (DSF: P=.05; DSB: P=.92; DSO: P=.16; LOM: P=.72; impulsivity: P=.19; SCC-self: P=.06; SCC-other: P=.08).

The current in-person cognitive screening tools face challenges related to accessibility, scalability, and PD specificity. We aimed to develop and tailor a new sensitive tool capable of identifying cognitive impairments in early- to mid-stage PD in a manner that was both accessible and scalable. Accordingly, ORCA-PD was developed as a fully self-administered, web-based cognitive screening tool tailored to capture PD-specific cognitive impairments.

First, as reflected by high completion rates (93%) and participant ratings (4-5 out of 5), ORCA-PD showed high usability, further supporting its feasibility across diverse populations (more than 30 geographical locations). Second, ORCA-PD naive scores correlated with MoCA scores, supporting preliminary validity. While the NT estimate should be interpreted with caution due to the small sample size, the moderate correlation with MoCA suggests that ORCA-PD captures overlapping but not identical constructs, offering information beyond a generic cognitive screener. As with all screening tools designed to enable accessible and fast identification of cognitive impairment, ORCA-PD was not meant to replace comprehensive assessment tools but to prompt early diagnosis of PD with MCI. The moderate correlation with the MoCA supports ORCA-PD’s role as a complementary tool, particularly when subtle, domain-specific changes are expected earlier in the PD course [52]. Third, ORCA-PD differentiated between groups, with performance following expected trends: NT>PD and PD-non-MCI>PD-MCI, confirming its discriminative capacity.

When examining previous literature, recent studies have begun to explore the potential of remote cognitive testing in PD. One study [28] demonstrated the feasibility and test-retest reliability of an unsupervised online cognitive battery in PD. However, the study did not assess validity, nor did it evaluate MCI using an additional screening measure. Another study [31] examined smartphone-based cognitive testing in daily contexts, but the sample size was small (n=27) and required a supervised laboratory session. A third study [27] demonstrated the feasibility of large-scale remote data collection in participants with PD using self-report questionnaires, but it lacked performance-based cognitive measures. A fourth pilot study [29] consisted of virtual neuropsychological testing using traditional batteries with a small sample size (n=35) and reliance on supervised administration. A fifth study [30] evaluated a lengthy 19-task online cognitive battery in individuals with PD and individuals with REM (rapid eye movement) sleep behavior disorder, identifying a post hoc 20-minute subset sensitive to group differences. However, the tasks relied heavily on reaction time, including motor-dependent tasks (ie, mouse use), which can confound cognitive measurement in PD. Additionally, the tasks were not chosen based on previous empirical independent findings and were not prospectively tested. In contrast, ORCA-PD was prospectively evaluated. The task selection was guided by prior meta-analyses and reviews that provided independent evidence for their sensitivity to PD. Additionally, the ORCA-PD tasks were specifically designed to reduce reliance on motor skills, thereby reducing the risk of motor-related confounds. This was also demonstrated by the lack of correlation between the ORCA-PD measures and the MDS-UPDRS III scores.

We developed and tailored ORCA-PD as a cognitive assessment for early- to mid-stage PD. In our sample, although inclusion criteria allowed participants with H&Y scores of 4 or lower, 98% had an H&Y score of 3 or lower (none had a score of 5; mean MDS-UPDRS III=21.9, mean disease duration=7.4 y). These findings are consistent with mild-to-moderate motor severity [55].

Of note, ORCA-PD was designed for remote, unsupervised administration, which likely contributed to the high rates of completion and usability observed, supporting its potential clinical applicability. Our sample did not include participants with dementia, as ORCA-PD was designed for early- to mid-stage PD, and self-administered testing may be less feasible in advanced stages. Taken together, validating ORCA-PD in this disease stage is advantageous because it targets the period when mild cognitive changes emerge.

Our study has several limitations. First, some measures (eg, SCC) showed limited sensitivity to between-group differences. Future studies should refine these tasks or replace them with more PD-sensitive alternatives, such as language-based measures. Second, although we assessed validity, we did not evaluate test-retest reliability. Third, the scoring of some tasks (eg, digit span) is relatively coarse, where a small number of trials per difficulty level may limit granularity. Yet, the ORCA-PD design is typical of screening instruments, such as the MoCA and MMSE. The ORCA-PD, in fact, includes more trials than these widely used measures (4 vs 2 in digit span), which may provide greater sensitivity for detecting impairments. Our primary goal at this stage was to implement a simple approach that allows direct comparison across predefined difficulty levels, serving as an assessment of the expected performance pattern (ie, sanity check) of the online task design. In addition, we prioritized accuracy-based scoring over reaction time to minimize the influence of motor slowing, which may disproportionately affect reaction time measures in PD. While reaction time may provide additional sensitivity, it may also introduce motor confounds. Fourth, MoCA data were available for only a subset of participants, resulting in a smaller sample. Although this subsample exceeded the minimum sample size required by the power analysis, the small sample size may limit the generalizability of these findings. Fifth, the MoCA was used both as a measure for convergent validity and to operationalize MCI classification, introducing a degree of circularity that limits the independence of the validation and may inflate observed associations. Sixth, medication status at the time of testing was not controlled. A majority of participants (<70%) were treated with dopaminergic medication, and the median time since the last medication intake was 120 minutes. Although prior work [56,57] suggests that dopaminergic medication has limited effects on cognitive screening measures such as the MoCA, with more modest and domain-specific effects, variability in medication state may have introduced additional heterogeneity in performance. This study was not designed to examine ON versus OFF medication effects, and future studies should systematically evaluate the influence of medication state and timing of testing on ORCA-PD performance.

Seventh, ORCA-PD scores were not compared against a comprehensive neuropsychological battery, and the use of a single-instrument cutoff does not meet MDS level II criteria. Future research should validate ORCA-PD against independent, gold-standard neuropsychological assessments and include parallel forms to estimate test-retest reliability, while recognizing that ORCA-PD is designed primarily as a screening tool. Eighth, online-based participation may favor individuals who are more technologically adept or less cognitively impaired; therefore, future studies should directly compare online and in-person administrations to assess potential selection and usability biases.

Ninth, some NT participants were recruited through relatives or caregivers of individuals with PD, which may introduce selection bias. Individuals in close proximity to people with a PD diagnosis may differ from the general population in terms of health care engagement and motivation to participate in research. This may limit the external validity of the findings, as the control group may not be fully representative of the general population.

In addition, race, quality of education, area deprivation index, and related factors were not collected. This represents a substantive limitation, particularly given the goal of developing an accessible tool for diverse populations, and it restricts our ability to evaluate generalizability across demographic and socioeconomic contexts. This is an important goal for future studies.

Language-related effects were examined as a secondary exploratory analysis. While some differences were observed (eg, DSF), these effects were not consistent across tasks or groups. However, our study was not designed or powered to evaluate cross-language differences or establish language-specific norms. Additionally, we did not aim to validate ORCA-PD in different languages. The relatively small sample, along with important potential demographic and cultural differences between language groups, further limits interpretability. Future studies should be designed to establish language-specific norms and examine between-language differences. Finally, the cross-sectional design precluded longitudinal inferences, and future longitudinal studies are needed to examine ORCA-PD’s sensitivity to cognitive change over time. Addressing these limitations will help maximize the tool’s clinical and research utility.

Although we developed ORCA-PD for people with PD, ORCA’s framework and principles can be expanded. Additional studies could focus on different populations, with an emphasis on varying levels of education and both rural and urban regions. Given the relatively high education level of our current sample, future validation in less-educated and lower-resource populations will be critical to ensure accessibility and generalizability. Socioeconomic information was not collected in this study, which limits our ability to assess the representativeness of our sample across social classes; future work should incorporate these measures to evaluate equity of access and usability for all. Furthermore, this platform need not be limited to PD. By selecting tasks matched to PD cognitive signatures and adjusting task difficulty and response modalities, our ORCA approach can be adapted to other conditions (eg, cerebellar ataxia). Such adaptations will require independent validation and condition-specific calibrations.

In summary, ORCA-PD is a brief, self-administered cognitive screening test for people with PD that demonstrates feasibility, utility, and preliminary validity that can be used in research and clinical settings upon further validation. By enabling unsupervised home testing, it has the potential to complement clinic-based screening and support between-visit cognitive monitoring. ORCA-PD could provide an accessible, scalable, and specific tool for cognitive screening in PD.