Deciphering Perspectives: A European Survey on Clinical Decision Support Tools for Dementia and Alzheimer’s Disease

The burden of neurocognitive diseases like Alzheimer’s disease (AD), pose significant challenges to public health and healthcare systems. And an increase in life expectancy has contributed to the shrinking workforce [1‒5]. One of the existing challenges in managing this problem is obtaining timely and accurate diagnosis. This is in addition to the unavailability of personalized treatment. Hence, there is a need for targeted interventions and healthcare strategies to address the growing burden of these conditions.

Scientific advancements in the areas of diagnosis and treatment have been made for AD management, which spread across various domains of clinical management. For diagnosis, PredictND integrates biomarkers to support differential diagnosis of neurocognitive diseases [6], while ADappt tool facilitates risk calculation of AD [7]. Also, a combination of machine learning with neuroimaging has been demonstrated to enhance the diagnostic accuracy of neurodegenerative disorders [8‒11]. Following diagnosis, another essential yet challenging domain in clinical management is communication amongst patients and care partners, which can influence patient outcomes. A personalized diagnostic report has been designed to solve this challenge and promote clear communication [12]. For the treatment of AD, majority of available therapies have been found to temporarily alleviate symptoms; and slow down AD progression by targeting brain amyloid deposits [13]. This initiated the development of several disease modifying therapies (DMTs) [13‒15], and Donanemab has recently been approved by the US Food and Drug Administration for the treatment of early AD [16]. There is a push for earlier diagnosis because of new DMT that require initiation in early disease stages.

Despite the advances mentioned, challenges like effective diagnosis persist, and AI can bridge that gap. A proposed solution is use of clinical decision support tools [17‒19] to assist disease management. The potential of decision support tools in streamlining clinical processes and improving overall efficiency has been highlighted [20, 21]. Aside from the core functionalities, tool implementation in clinical practice should also be considered.

Different factors have been reported to influence the acceptance and adoption of digital clinical decision support systems (CDSSs) in healthcare facilities. The human, organization, and technology (HOT) fit framework [22] aids in the assessment of factors that affect adoption and implementation of healthcare tools like CDSS. Perceived usefulness and trust in the knowledge base of the tools [23] have been purported as contributory reasons for utilization. Clinicians’ views of the system align with their acceptance [24, 25]. Consequently, they influence the redesign of the system to fit their workflows. Few studies [26‒28] have assessed healthcare providers’ opinions on the decision support tools within this field. However, functions regarding DMT were not explored.

In a new Innovative Health Initiative project, PROMINENT, we are developing a platform for precision medicine specifically targeting the diagnosis and treatment of neurocognitive diseases and their comorbidities [29]. To create a relevant and practical platform, inputs from the intended users will inform the development. Hence, this study aimed to determine clinician’s perspective of CDSS in the management of dementia and AD across Europe. The survey was designed to answer two research questions: (1) What are the required functionalities in a CDSS for the management of dementia and AD from a clinician’s perspective? With a specific focus on DMT, (2) What factors impact the uptake and use of a CDSS?

This was a cross-sectional survey conducted from June 2023 to September 2023. The study was carried out within various healthcare facilities across Europe.

Participants were invited from the European Alzheimer’s Disease Consortium, which comprises 65 member centers. We targeted clinicians with field expertise, including neurologists, psychiatrists, and geriatricians. The selection rationale for these participants was to ensure that survey captured practical insights into real-world clinical needs and reflect the diversity in practices across Europe. Invitations were sent via institutional email lists containing the link to the survey. All participants were informed about the study’s purpose and confidentiality, and they provided digital informed consents.

The questionnaire used in our survey was adapted from a previous study [26], which evaluated the views of medical professionals on the use of computer tools in memory clinics. We expanded on the previous study questionnaire by including questions about experience with current CDSS and adding a section for DMT and treatment functionalities required from the new CDSS. The questionnaire was created using the Research Electronic Data Capture (REDCap) web application and designed in English. There were four main sections. The first section captured basic demographic information of the study participants including age, country of practice, profession, and type of healthcare practice. The second section explored knowledge and experience with CDSS, using a sliding scale format for this exploration. The third section explored the potential usefulness of CDSS functions across the domains of diagnosis, treatment decisions – DMT, prognosis and communication with patients or care partners. Some examples of questions included were the potential usefulness of CDSS in predicting the likelihood of the patient having AD, assessing eligibility for DMT. A Likert-scale format was employed for this section. Participants rated their responses on a scale from 1 “not useful” to 5 “very useful.” The final section had a set of 12 questions to gather opinions on HOT factors that impact CDSS implementation in a healthcare setting. These questions helped us understand the barriers and facilitators related to CDSS adoption. The domains captured were system quality, information quality, system use, and environment. These were also asked using a five-level Likert scale ranging from “not important” to “very important.” The entire questionnaire form can be found in supplementary material A (for all online suppl. material, see https://doi.org/10.1159/000544801).

We administered the questionnaire digitally using the REDCap platform, which facilitated secure data collection and management. Participants accessed the survey via personalized links, and reminders were sent periodically to enhance response rates. Responses were collected anonymously, and data were stored in a secure environment at Karolinska Institutet.

There were no mandatory questions in the questionnaire; this was done to encourage willing and honest responses. Therefore, there were item missing responses. This was addressed during the data cleaning process, by excluding incomplete responses that had no influence on study outcomes. We also excluded respondents with no experience in the field of dementia. Following the exclusion, 84 questionnaires were eligible for analysis. Participant characteristics and survey outcomes were analyzed on Microsoft Excel using descriptive summaries.

At least one participant was anticipated from each of the 65 member centers, and we received 117 responses. A filtration criterion was applied during data cleaning. These filter criteria excluded respondents that had no clinical experience in the field of dementia or AD, missing or no response about CDSS and drop-outs.

Figure 1 depicts the geographical distribution of survey respondents. A total of 15% did not state their country of practice. Respondents were distributed across 18 countries in Europe. Amongst those that indicated their country of practice, the top three most represented countries were the Netherlands, Norway, and Spain with a share of 20%, 18%, and 8%, respectively. The United Kingdom, Belgium, France, Finland, Sweden, Portugal had similar representation ranging from 2 to 4%.

A total of 77% (N = 63) survey respondents were specialist physicians including neurologists, geriatricians, and psychiatrists. Over half of the specialist physicians had more than 10 years of clinical experience. A total of 13% respondents were psychologists, and nearly half of them had 3–5 years of relevant clinical experience. The remaining 10% comprised nurses, research physicians, and trainee physicians. More than half of the respondent category had no previous knowledge or experience with CDSS.

Table 1 contains descriptive information on the baseline characteristics. Participants were mostly in the 30–59-year age group. With regard to their primary workplace, 63% worked at tertiary referral centers or university clinics, 31% at specialist memory clinics, and the remaining distributed across community care centers and nursing homes.

Of the 84 respondents who answered this survey, 63 respondents answered the section on future diagnostic functionalities for the system. Majority (90%) indicated that predicting the likelihood of the patient having AD was the most vital function. This was followed by predicting differential diagnosis and identifying patients for screening, reported by about 70%. The least required function as claimed by a little over half of them was predicting amyloid positivity. Figure 2 shows more details on the responses.

Over three-quarter (79%) of the respondents indicated ability to assess treatment eligibility as most useful. While identifying contraindications to DMT (72%) and predicting response to treatment (70%) had a great degree of usefulness. The least required functions in this category were predicting effects on quality of life and predicting effects on institutionalization. Figure 3 displays the responses of this section.

This section received 64 responses from the total number of participants. Two functions – predicting future cognitive decline and future functional decline were deemed the most useful by 69% of respondents, followed by predicting the ability to drive a car by 61% of respondents. Predicting mortality was considered the least useful by 47% of respondents in this domain. Figure 4 illustrates the responses of this section.

We received 58 responses in this section. Using HOT fit framework, information quality dimension had the highest frequency of importance, with 90% of the respondent’s indicating accuracy/validity of output as the most important facilitator/barrier. A total of 79% stated that documentation and evidence of the CDSS were important. User friendliness was a significant facilitator, highlighted by 88% of respondents, closely followed by integration with electronic health records system with 87% of respondents. The least important factors were system customizability (64%) and colleagues experience with system (51%). Figure 5 contains all the descriptions of this section.

This survey explored the perspectives of clinicians on required functionalities for a CDSS. Most significant findings for diagnostic functions were the ability to predict the likelihood of AD and differential diagnosis. Predicting amyloid positivity as the least required function was a surprising finding, considering the significance of amyloid positivity in management of AD. Plausible explanations for this finding may be the reliance on established diagnostic tools to assess amyloid positivity, or some participants like psychologists, who are not directly involved with diagnosis may not be familiar with the relevance of this marker. This could have resulted in underestimating its utility as an essential functionality. For treatment domain, assessing eligibility for treatment and identifying contraindications to DMT had the most responses, while predicting future cognitive and functional decline were the most useful prognostic functions.

Our study findings add to the expectation of a CDSS in the context of DMT. The ability to identify eligible patients for this therapy drives clinical efficiency, since not all patients with AD may be suitable for this intervention [15]. In addition, predicting how patients are likely to respond to this therapy can help tailor treatment plans that could lead to improved patient outcomes. Adverse events like infusion-related reactions, cerebral microhemorrhages [13, 15] have been associated with DMT, justifying the predictive function of experiencing these events. Amongst the least prioritized functionalities were effects on cost of care and institutionalization. One possible explanation for this finding is that the clinicians are not directly responsible for them. Or perhaps the cost of care is difficult to forecast due to other contributory factors.

According to the survey, user friendliness, accuracy of output, and integration with existing health records significantly influence implementation. Limited familiarity with CDSS may raise concerns about disruptions to established workflows and confidence in its utility. Addressing this issue may require organizational changes like tailored training programs to improve user proficiency. Also, clinical leadership could assist with the transition, by communicating the tool benefits in relation to patient outcomes and other organizational goals. This may increase confidence and tool acceptance. Additionally, integrating CDSS within clinical settings could depend on its interoperability with existing health records and compatibility with other health information technology. Technical barriers such as lack of standardized data formats, difference in protocols and policies, and varying level of system security are some of the potential issues that could impact integration. Customizing the CDSS to align with the data and governance standards of the settings could be a potential solution.

Various studies have demonstrated the effectiveness of decision support systems and machine learning techniques in the management of dementia [17, 30, 31]. A previous European survey on functionalities and implementation of a computer tool in clinics reported the preference of these tools for diagnostic, prognostic, and communication [26]. Within the same study, technical and user ability were cited as essential implementation factors, although setting was limited to memory clinics. Our study adds on to this survey by expanding on CDSS functionalities to incorporate DMT, and providing specific requirements of each functionality. A systematic review [32] on implementation success factors found information and system quality factors like ease of use and compatibility with existing health records as high contributory factors, which was similar to our study findings. Furthermore, technological factors have also been demonstrated to affect implementation in primary care setting [33]. In hospitals, practical usability, trust in output, and fit with workflows influenced utilization of CDSS [34].

This study exhibited several strengths. First, the participant heterogeneity in terms of their specialization, healthcare setting, and countries broadened the perspective of the findings. Diversity in the respondent pool ensured that survey results captured several viewpoints, which enhanced result transferability to other settings. Second, knowledge from this survey would be translated to real-world settings, as the survey was directly linked to the design and development of a future CDSS.

Conversely, there were a couple of limitations. First of all, the questionnaire mostly comprised of technological implementation factors, with reduced spotlight on human and organizational factors. This restricted the consideration of potential insights that could hinder tool adoption and integration. Furthermore, selection bias may have been introduced during data analysis, while mitigating the issue of incomplete data, as responses could have skewed towards more informed clinicians. Finally, participants were selected purposively from a clinical group, which limited the sample size and representation of some countries. Therefore, findings may not fully capture the perspectives of all potential users of the tool, which could affect the generalizability of the findings. However, there was at least one survey respondent from each cardinal geographic regions, which provided sufficient diversity to achieve the survey objectives.

Future studies could improve response rate and inclusion of underrepresented countries by increasing their sample area and sampling approach to ensure a more balanced representation. Also, broader participation could be promoted by utilizing multi-lingual questionnaires.

The study provides guidance for the development and implementation of future clinical decision support tools in the management of dementia and AD. Although experiences with CDSS are still limited, considering the needs and preferences of the clinicians, can help facilitate adoption and integration into workflows. This approach, coupled with the insights gained from the study, can assist the development of clinical support tools that are practical and useful in a real-world setting.

Participants signed a digital informed consent prior to data collection. This consent was embedded in the digital questionnaire, and those who withheld consent did not complete the questionnaire. All data were collected anonymously. Personal data such as your name, email address or date of birth were not obtained. Data were kept confidential and results only presented at group level. Additionally, data will be retained for 5 years and stored in a secure IT environment at the Karolinska Institutet. As no sensitive data were collected or processed, the study did not require ethics review according to Swedish research ethics legislation.

This study was funded by the European Union, the private members, and those contributing partners of the IHI JU. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the aforementioned parties. Neither of the aforementioned parties can be held responsible for them. Bengt Winblad was a member of the journal’s Editorial Board at the time of submission.

This project is supported by the Innovative Health Initiative Joint Undertaking (IHI JU) under grant agreement No 101112145. The JU receives support from the European Union’s Horizon Europe research and innovation programme, Combinostics Oy and Bioarctic AB. “The funder had no role in the design, data collection, data analysis, and reporting of this study.” H.R. is recipient of the Memorabel Dementia Fellowship 2021 (ZonMw project no. 10510022110004) and Alzheimer Nederland InterACT grant (project no. WE.08-2022-06).

L.J.: conceptualization, validation, and writing – review; I.U.: data analysis, writing – original draft, review, and editing; X.X., B.W., S.A., H.F.M.R.-M., and E.A.: writing – review. All authors read and approved the final manuscript.

The dataset is not publicly available because individual privacy could be compromised; however, it can be made available from the corresponding author on reasonable request.