This study is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) checklist for cross-sectional studies [12] (see Checklist 1) and the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) [13].

The survey was conducted within the scope of the EU-funded AI-PROGNOSIS project, which aims to advance PD diagnosis and care through novel predictive models combined with digital biomarkers from everyday devices [11]. The survey was exploratory in nature.

The online survey consisted of 21 to 23 questions, with the exact number depending on the respondent’s group, that is, people with PD, persons at risk of PD, caregivers, and HCPs. Some questions asked respondents to rate statements. Including the latter, the survey consisted of 48 items. The survey included 15 to 17 mandatory (the number of questions depended on the stakeholder group) closed single-choice and multiple-choice questions, 48 mandatory 5-point Likert-scale items (ranging from 1=Strongly disagree to 3=Neutral to 5=Strongly agree), and 6 optional free-text comments. The questions were divided into 5 different sections: general information (16‐18 questions), perceived potential (6 questions), perceptions and attitudes (5 questions), intention to use (2 questions), and additional feedback (1 question). The number of general information questions comprised 7 universal questions and 9 to 11 stakeholder-specific questions, depending on the respondent group. The survey was designed by adopting a polytheoretical framework as a fusion of theories, frameworks, and models [14] to better understand the acceptance of AI-based tools in PD, namely the Theory of Planned Behavior [15], Social Cognitive Theory [16], the Health Belief Model [17], and the Technology Acceptance Model [18]. In particular, the Theory of Planned Behavior contributes insights into how attitudes, perceived norms, and perceived control shape stakeholders’ intentions to adopt AI-based tools. The Social Cognitive Theory emphasizes the role of self-efficacy and observational learning, highlighting how confidence in using digital technologies and exposure to peer practices may influence uptake. The Health Belief Model adds a focus on perceived susceptibility, severity, benefits, and barriers, which is particularly relevant for people with PD and caregivers evaluating preventive or supportive AI applications. Finally, the Technology Acceptance Model provides a lens on perceived usefulness and ease of use, central to understanding the willingness of HCPs and people with PDs to integrate AI into daily management. Together, this polytheoretical approach enables a more nuanced exploration of digital practices and desires, ensuring that the study captures both behavioral drivers and technological determinants across diverse stakeholder groups.

Using the EUSurvey [19] platform, the survey was available in English, French, German, Spanish, and Swedish. Translation was done in accordance with the World Health Organization [20], where a native speaker translated it and another native speaker back-translated the translation for cross-reference. Before dissemination, the survey design was pilot-tested by JLL through cognitive interviews with 2 persons with PD and 1 caregiver, which resulted in an expanded information text for each part.

In this paper, only Part 1: General information (41 items between all stakeholder versions) was analyzed [21] (Multimedia Appendix 1), which was intended to provide an exploratory snapshot of digital health practices across stakeholders. Other survey sections address more detailed user preferences and require a larger sample size for robust analysis; these data are planned to be presented in a subsequent publication.

Survey dissemination was done through the clinical and patient networks of the AI-PROGNOSIS project partners, focusing on European countries where the research team’s available languages are spoken (the United Kingdom, France, Germany, Spain, and Sweden). However, no geographical restrictions were imposed, allowing responses from all countries. Data were collected between December 2024 and October 2025.

A nonprobability sampling method was used as an exploratory sample with no random selection. Inclusion criteria for respondents were being 18 years or older and self-identifying as one of the following: (1) having a Parkinson diagnosis, (2) being at risk of PD, (3) being related to someone with PD, (4) being a caregiver to a person with PD, or (5) a HCP currently working with persons with PD. Individuals who indicated that they are at risk of developing PD and individuals who indicated that they are related to someone with PD were combined to form the “at risk” group in the analysis.

We analyzed 41 questions (Part 1) from the AI-PROGNOSIS survey on a predictive AI-based tool for PD care [21] (Multimedia Appendix 1). These multiple-choice questions were analyzed descriptively and summarized through frequency tables, figures, and in-text descriptions. Similar responses (eg, “good” and “excellent”) were dichotomized where applicable. The analysis was carried out by JLL in JASP v0.19.1.0. Where the same question was asked of multiple stakeholder groups, these responses are presented side-by-side in figures or text for comparison. χ² tests were performed to examine stakeholder differences in desired AI features. Because all comparisons involved 2×2 contingency tables, continuity correction by Yates was applied [22]. Open-ended qualitative questions were not included due to low response rates.

The survey responses were anonymous. The project was submitted for ethical review and was deemed exempt from formal approval by the Swedish Ethical Review Authority, Dnr 2024-05505-01.

This paper presents findings from a survey on perspectives of AI and digital tools for Parkinson management across 4 stakeholder groups: people with PD, people at risk of developing PD, caregivers, and HCPs. Where relevant, findings are presented by the respondent group to reflect differing roles, experiences, and needs. As the respondent groups differ in sample size, the respondents (n) are presented in a table or in text (xx/xx, %).

A total of 255 responses were received from 18 different countries, with Spain (n=94, 37%) and France (n=66, 26%) contributing the most (Multimedia Appendix 2). Half (132/255, 51.8%) of the respondents were persons with PD (Table 1). Most of the persons with PD (67/132, 50.8%) and caregivers (8/21, 38.1%) were retired, while the majority of at-risk respondents (20/29, 69%) and HCPs (61/73, 83.6%) were employed full-time. Additionally, 79.2% (202/255) of all respondents had completed higher education (bachelor’s degree or above).

Most people with PD (53/132, 40.2%) were aged 65 to 74 years, and half were retired (67/132, 50.8%). People with PD reported a complex symptom profile, commonly citing bradykinesia (85/132, 64.4%) and rigidity (85/132, 64.4%) as the symptoms they experience most frequently and managing their PD with medication (123/132, 93.2%) and exercise (104/132, 78.8%). Nearly one-third were diagnosed 1 to 3 years ago (42/132, 31.8%), followed by 4 to 6 years ago (37/132, 28%). Individuals at risk consisted of 2 out of 29 (6.9%) participants who self-reported risk, and a majority of individuals were related to people with PD (classified as at-risk in this study). This group was generally younger than people with PD (10/29, 34.5% were 55‐64 years old) and employed (20/29, 69% full-time).

Caregivers were primarily spouses (12/21, 57.1%) or adult children (8/21, 36.1%), of which the majority had been providing care for over a decade (9/21, 42.9%) or 6 to 10 years (5/21, 23.8%). Their perceived support levels were notably low, with most rating the support they received as “very poor/poor” (14/21, 66.7%). The most positive support level reported was “fair” (6/21, 28.6%).

HCPs were typically younger, with most aged 25 to 44 (49/73, 67.1%) years and working full-time (61/73, 83.6%) in a hospital (45/73, 61.6%). The sample was professionally diverse, including neurologists, nurses, and physiotherapists.

The most reported health goals, according to people with PD, were to “minimize symptoms with lifestyle factors (eg, diet, physical activity, sleep habits)” (123/132, 93.2%) and “slow Parkinson disease progression through lifestyle factors (eg, diet, physical activity)” (112/132, 84.6%). The most commonly reported challenge among persons with PD was managing motor symptoms (70/132, 53%; Table 2).

Caregivers most often cited providing emotional support (14/21, 66.7%), other (14/21, 66.7%), balancing caregiving with other responsibilities (13/21, 61.9%), and helping with mobility and physical activities (12/21, 57.1%) as their primary caregiving challenges (Table 3). When asked about their knowledge of PD, nearly half of them rated it as “moderate” (10/21, 47.6%), and 43.9% (9/21) reported it as “high/very high.” The most frequently reported information sources were health care professionals (11/21, 52.4%) and educational websites (10/21, 47.6%).

Like people with PD, HCPs also identified symptom management as a primary challenge in PD care. However, there was no difference in ranking of nonmotor symptoms (49/73, 67.1%) and motor symptoms (47/73, 64.4%), which was followed by “providing patient and caregiver education” (39/73, 53.4%; Table 4). When asked about the overall effectiveness of current treatments, half (36/73, 49.4%) of HCPs rated them as “good” or “excellent”, while 45.2% (33/73) rated them as “fair.” Regarding methods of caregiver involvement, HCPs most commonly reported providing education and resources (51/73, 69.9%), maintaining regular communication (46/73, 63%), and involving caregivers in treatment planning (43/73, 58.9%). Fewer reported offering support groups or counseling (35/73, 48%).

In terms of managing PD risk, few had consulted a HCP regarding their perceived risk (5/29, 17.2%) or taken steps to manage it (eg, specific lifestyle recommendations, regular check-ups, self-tracking; 8/29, 27.6%). Perceived risk factors were not only primarily genetic predisposition (23/29, 79.3%) but also included age (11/29, 37.9%), environmental exposures (11/29, 37.9%), head injuries (5/29, 17.2%), and others (6/29, 20.7%). The majority (18/29, 62.1%) had not experienced any symptoms associated with prodromal PD, and levels of concern were evenly split between moderate-to-high concern (14/29, 48.3%) and little-to-no concern (15/29, 51.7%).

Across all groups, the majority self-rated their digital literacy level as intermediate (Figure 1). The lowest reported levels were among informal caregivers, although they were also the smallest sample.

Participants across all groups showed interest in using digital tools and AI to support PD (risk) management. The majority of at-risk respondents reported being “possibly” or “definitely” interested in using apps or wearable devices for monitoring risk or managing early symptoms (23/29, 79.3%). Similarly, most caregivers expressed interest in digital tools to support their caregiving responsibilities (14/21, 66.7%), and most HCPs reported interest in integrating AI into PD care (56/73, 76.7%).

Despite this interest, the current use of such tools remains limited. The majority of persons with PD reported not using a smartwatch or mobile tracking app (108/132, 81.8%). However, half expressed an intention to begin tracking (70/132, 53%), and nearly a quarter already did (31/132, 23.5%). No caregivers reported using mobile apps for PD care, and very few used wearables (3/21, 14.3%); most of them relied on observation and memory (18/21, 85.7%).

Respondents were presented with a stakeholder-specific list of potential AI features. The following paragraphs describe which features were prioritized within and across respondent groups (Figure 2). Symptom tracking emerged as the most frequently selected AI feature across all respondent groups, though secondary priorities varied between groups. HCPs also placed strong emphasis on tools that enhance patient and caregiver engagement (47/73, 64.4%), a priority less commonly noted by people with PD (30/132, 22.7%).

To further compare priorities for AI features between people with PD and HCPs, 5 χ² tests of independence were conducted to determine whether these stakeholders select specific features at different rates. Group differences were not statistically significant for early diagnosis and risk assessment, χ²₁ (n=205)=0.21, P=.64, Cramer V=0.03; monitoring and tracking symptoms, χ²₁ (n=205)=1.75, P=.19, Cramer V=0.09; personalized treatment recommendations, χ²₁ (n=205)=2.74, P=.10, Cramer V=0.12; or predicting disease progression, χ²₁ (n=2o5)=0.03, P=.87, Cramer V=0.01. These effects were all small in magnitude.

A statistically significant and meaningful stakeholder difference was, however, found for enhancing patient and caregiver engagement, χ²₁ (n=205)=34.78, P<.001, Cramer V=0.41. HCPs prioritized this feature significantly more frequently than people with PD, indicating a moderate-to-large association (Multimedia Appendix 3).

Preferences among individuals at risk for PD were examined separately using a tailored set of potential AI features. Among at-risk individuals (Table 5), AI-generated lifestyle recommendations aimed at prevention were more popular than progress reports and health insights (23/29, 79.3% and 15/29, 51.7%, respectively). Communication with health care professionals was the second most cited feature for both persons at risk (18/29, 62.1%) and caregivers (13/21, 61.9%; Table 6).

This study explored how different stakeholder groups—people with PD, people at risk, informal caregivers, and health care professionals—currently engage with digital tools, what they expect from AI-based solutions, and where their priorities converge or diverge in managing PD. The results highlight the complexity of managing PD and suggest both universal needs and distinct priorities across these 4 groups regarding the potential application of predictive AI. Interest in predictive AI was high across all stakeholder groups, with symptom tracking being the most desired feature; however, secondary priorities varied by group. Additionally, despite this interest, reported technological use was limited. Most of the people with PD did not use symptom-tracking apps or wearables, nor did they currently track their condition, though many expressed intentions to start. This pattern aligns with broader digital health literature demonstrating that interest in technology does not necessarily translate into adoption or sustained use [22]. Prior studies have found cost [23], discomfort (of wearables) [23], and fear of replacing in-person care [24] to constrain adoption for older adults with chronic illness. Addressing these barriers will be essential if AI-driven solutions are to move from intention to sustained practice.

Based on this survey, the most critical insight for the development of new AI applications is the universal demand for symptom tracking; it was the most frequently selected desired AI feature across all respondent groups (ie, persons with PD, at-risk individuals, caregivers, and HCPs). This demand directly correlates with the perceived primary challenges in PD care, namely, for people with PD, the greatest challenge reported was managing motor symptoms; for HCPs, symptom management remains a primary challenge, with nearly equal emphasis on nonmotor symptoms and motor symptoms. Also, for caregivers, a symptom-tracking aid might be beneficial, given that they reported relying heavily on memory and observation for tracking. This consensus on symptom tracking supports the initiative to use digital biomarkers from everyday devices for PD monitoring, a direction in which many research projects are heading. Effective self-management, which is increasingly emphasized in PD care, depends heavily on using both internal resources (like self-tracking) and external support (like digital tools).

While symptom tracking was shared, the secondary priorities for AI tools reveal stakeholder-specific challenges that must be addressed to ensure user-centered design (Figure 2). In particular, at-risk respondents expressed strong interest in preventive lifestyle recommendations but reported limited engagement with HCPs regarding their risk. In contrast, the HCPs emphasized caregiver engagement and education above what people with PD did. These contrasts suggest the need for differentiated design approaches: preventive guidance for at-risk individuals, caregiver support features for HCPs and caregivers, and self-tracking tools for persons with PD.

There is extensive evidence that many patients desire to take a more active role in their care and that digital health tools can aid in doing so. Additionally, research shows that people with PD and health care professionals desire to use digital health tools to facilitate co-care between them [25], and older adults have shown higher motivation to engage with technology when there is a clear value in terms of health outcomes [25,26].

This active role is tied closely to the primary health goals reported by people with PD in our study, which were centered on minimizing symptoms and slowing PD progression through lifestyle factors (eg, diet, physical activity, and sleep habits) or, in other words, factors that one can influence outside of health care. Similarly, individuals at risk, whose perceived risk factors include genetic predisposition and age, prioritized AI-generated lifestyle recommendations aimed at prevention. This shared focus confirms the importance of integrating nonpharmacological, preventative, and lifestyle-based recommendations into AI tools for both diagnosed individuals and those at risk.

A qualitative study within the AI-PROGNOSIS project also emphasized the demand for actionable insights, though it also demonstrated how individual and personal this question is. The degree to which participants felt they could influence their disease varied, as did their desire to know predictions about their disease [27]. This emphasis on actionability is consistent with prior work by Schaeffer et al [28], who found that expert clinicians and researchers in prodromal PD held divergent views on whether and when to disclose PD risk, with actionability emerging as a key factor in their reasoning. Together, these findings highlight the importance of individualized prognostic counseling for those at risk and with PD and the role of the clinician in determining the patient’s “wish to know” [28,29].

Stakeholder desires for AI tools extend beyond symptom monitoring to actionable, personalized insights [30,31]. In our study, people with PD emphasized lifestyle-based recommendations to minimize symptoms and slow progression, while at-risk individuals prioritized preventive guidance tailored to genetic and age-related risk. Caregivers highlighted the need for tools that reduce burden and provide structured support. The emphasis on actionable, personalized insights aligns with recent studies showing that the adoption of digital tools in PD is driven less by data collection itself and more by perceived clinical or lifestyle benefits [30,31]. These findings suggest that AI solutions should integrate predictive analytics with practical, user-centered recommendations that address both medical and nonmedical aspects of PD care.

HCPs desired tools that would enhance patient and caregiver engagement, and the literature demonstrated that HCPs recognize caregiver involvement as vital for the effective use of digital tools and clinical decision-making [32]. This priority directly addresses the third greatest challenge HCP respondents identified: providing patient and caregiver education. Respondents also reported involving caregivers by providing education and resources and maintaining regular communication. An AI tool could potentially aid in these tasks to overcome educational and communication gaps. The literature supports education as a challenge for HCPs, where a provider must balance how much to discuss and when. For example, discussing advanced PD services with someone having mild symptom presentation can cause unnecessary worry [33].

Notably, this desire for AI-facilitated engagement was less commonly cited by people with PD, who seemed to prioritize taking a more active role in self-managing their PD. However, family-centered care approaches in PD have been found to improve patient safety and self-efficacy [33]. A possible explanation for why people with PD did not prioritize patient and caregiver engagement to the extent HCPs did might be because patients who are able to participate in research may already be quite independent and less reliant on a caregiver and thus prioritize their own experience and features they find useful for self-care. Meanwhile, HCPs might focus on caregiver engagement due to the unique and clinically valuable insights a caregiver holds [28], especially among people with PD and cognitive symptoms. Research suggests that caregivers who utilize mHealth apps have substantially higher odds of engaging in health communication and making informed health decisions in collaboration with providers [33]. Consequently, HCPs are advised to encourage and support mHealth use among caregivers, which may be reflected in our results.

Communication with health care professionals was the second most cited desired feature for both caregivers and persons at risk. For caregivers, who listed emotional support and balancing responsibilities as main challenges, improving communication suggests a need for reliable channels to share information and seek guidance. For at-risk individuals, who rarely consult HCPs about their risk, AI-assisted monitoring may provide insights into when to connect with HCPs and provide data to help structure the visit.

Early diagnosis was not a feature prioritized by people with PD or HCP respondents. Literature shows that diagnosing PD can profoundly affect mental health and identity, which may explain why some respondents may not welcome earlier detection in the absence of disease-modifying therapies and robust counseling [34]. Early-detection tools raise a prospective double-edged sword, as they could accelerate recruitment into trials and enable earlier supportive care, but without careful specificity, follow-up protocols, and shared decision-making, they may also widen harms from overdiagnosis. PD is commonly overdiagnosed, especially in its early stages and among nursing-home residents or patients assessed by clinicians without specialized training [35], and longitudinal evidence further indicates that initial PD diagnoses are frequently revised over time [36].

On the other hand, people with PD have typically already lost over half of their dopamine-producing neurons, so earlier diagnosis, in situations where such severe loss can be prevented, could be extremely valuable. For highly sedentary individuals, implementing regular high-intensity exercise before such extensive loss could potentially delay onset and symptom presentation. In a proof-of-concept study of patients with mild and early PD, 6 months of high-intensity exercise was found to induce beneficial neural changes, reversing the expected decline in key biomarkers of the dopaminergic system [37].

Taken together, the results reveal both convergence and divergence in stakeholder priorities. Convergence was evident in the universal demand for symptom tracking and interest in predictive AI. Divergence appeared in the emphasis on caregiver engagement among HCPs, preventive lifestyle guidance among at-risk individuals, and burden reduction among caregivers. Explicitly recognizing these shared and distinct needs is critical for participatory design, ensuring that future AI-driven solutions are broadly acceptable while remaining tailored to the unique contexts of each stakeholder group.

This study used a nonprobability sampling approach, with recruitment primarily conducted through project networks and online dissemination channels. This strategy enabled participation from multiple countries and stakeholder groups, enhancing the overall diversity and breadth of perspectives captured. However, the same approach may have introduced sampling biases, including uneven representation across stakeholder groups and an overrepresentation of highly educated respondents, with 80% holding a bachelor’s degree or higher. Combined with the relatively small sample size and the exploratory nature of the study, these factors limit the generalizability of subgroup comparisons. Accordingly, the findings should be interpreted as preliminary and are intended to inform future large-scale studies using more systematic and inclusive recruitment strategies.

While multicountry inclusion is a strength, it also introduces challenges in interpretation. Contextual factors such as health care systems, access to digital tools, and cultural attitudes toward technology vary widely across the countries included. These wide contextual differences may have influenced how participants perceived and responded to questions about symptom management and digital support tools. Additionally, although the AI tools were briefly explained (including functions, features, and aims), differing understandings of AI likely influenced responses and acceptability.

As AI-assisted tools become more of a reality than a fantasy, we must ensure robust user-centered design, which accounts for differences in stakeholder preferences. This entails participatory methods with efforts to include underrepresented populations, including those with lower education levels and participants outside of EU countries. As these AI-based tools develop, continued involvement of people with PD and stakeholders is crucial.

This survey identifies specific features stakeholders prioritize; future discussion should focus on how to overcome the practical obstacles in implementation and adoption to realize the potential of AI in PD care. Symptom tracking was a top desired feature, though there are already many tools available aimed at filling out this need. As suggested by Madanian et al [8], this is likely in part due to such tools being developed for HCPs rather than for patients or for collaboration between both parties. As HCPs emphasized engaging patients and caregivers, future research might examine how data collected and analyzed through AI tools (eg, detailed symptom tracking and prognosis) can be seamlessly integrated into clinical workflows in such a way as to facilitate shared decision-making.

Finally, building on the importance of participatory approaches in the design of AI-based solutions, the effective integration of AI into PD management depends on the development of tools that are trustworthy, inclusive, and transparent. Explainable and reliable AI systems are essential to safeguard patient safety and foster clinician confidence, while inclusive model design and validation help mitigate bias and promote equitable care across diverse populations. Transparency in AI-driven decision-making further supports ethical governance, facilitates clinical adoption, and enables informed use in real-world practice.

This study aimed to identify and compare digital practices and desires for AI-based tools for PD among people with PD, persons at risk of PD, caregivers, and HCPs. The survey revealed high interest in AI-based tools for PD management across all stakeholders, with symptom tracking identified as the most universally desired feature. Despite this interest, current use of digital tools remains limited, revealing a gap between interest and adoption. Barriers such as usability, cost, digital literacy, and trust in technology may be constraining adoption. Secondary priorities differed, with people with PD and at-risk individuals favoring actionable lifestyle recommendations, caregivers desiring improved communication, and HCPs prioritizing patient and caregiver engagement. There was a statistically significant difference between persons with PD and HCPs and the desire for an AI feature that enhances patient and caregiver engagement. These findings highlight the importance of user-centered design approaches that foreground stakeholder needs and priorities, thereby enhancing the relevance, usability, and uptake of digital tools. By translating predictive insights into actionable support for self-care and collaborative decision-making, such approaches may foster sustained engagement, adherence, and more effective integration of AI-based technologies into PD care.