We conducted the study within a single large hospital Trust in the United Kingdom, screening all patients presenting to the ED with TLoC or referred to first seizure or syncope clinics within the window of February 10, 2022, to January 9, 2023. All participants recruited to the quantitative study were also invited to interview; we approached those who consented to interview sequentially.

We had a provisional target of 30 participants, using empirical data showing 24 participants reliably achieves saturation, the narrow specification of subject matter, and the study team’s expertise in the subject matter [25].

All participants in the quantitative study completed the PESQ at or shortly after presentation to the ED or primary care with TLoC. The majority completed it using a simple online platform. The PESQ comprises a brief (52-item) questionnaire comprising 4 demographic questions, 13 regarding patient medical history, and 35 peri-ictal symptoms. The online platform delivered these with sequential presentation of each question, with the option for binary yes or no responses (except for the demographic questions of age and years of education, which allowed for any integer answer). Participants could skip questions or navigate back and forth between them. Multimedia Appendix 1 lists PESQ items. For this study, all PESQ data were stored in anonymized fashion on a university secure database.

Patients could complete the questionnaire at the time of assessment and presentation using a tablet computer provided by the study team or else at home after their initial presentation through a simple browser-based application.

Two researchers (AW and DH) conducted remote (video, via Microsoft Teams, or telephone) semistructured interviews following a predefined interview schedule (Multimedia Appendix 2). Interviews lasted between 30 minutes and 1 hour. Interviewers noted initial reflections in contemporaneous logs, to support reflexive engagement with later analysis. An independent, nonclinical, professional transcription service transcribed all interviews for subsequent analysis.

We undertook thematic analysis [26] of transcribed interviews [27]. One researcher (LB; an experienced qualitative researcher with no clinical background or experience of TLoC) imported transcribed interviews into NVivo (Lumivero), and 2 researchers (LB and AW; a specialty registrar [senior resident] in neurology, with prior experience as a core trainee [junior resident] in cardiology and emergency medicine) developed and refined themes through iterative coding.

We assessed for saturation through interim analyses, stopping recruitment when saturation was reached.

We preregistered the study protocol on ClinicalTrials.gov (NCT05367999). Ethical approval came from the NHS Health Research Authority Edgbaston Research Ethics Committee (IRAS: 304114). We report results in line with the SRQR (Standards for Reporting Qualitative Research; checklist provided in Checklist 1) [28].

Of 2811 potential participants screened for recruitment, 1181 were eligible. Of these, 186 responded to the invitation to participate, and 133 also consented to approach for an interview. We approached 40 participants for interviews, aiming for diversity of age and gender. After 20 interviews, we achieved data saturation.

Of interview participants, 14 (70%) received final diagnoses of syncope, and 6 (30%) epilepsy. In total, 12 participants (60%) were female. The median age was 69 (range 17-90) years. Interviews were held a median of 69 (range 35‐283) days from initial presentation. Table 1 displays full participant demographics.

The semistructured interview protocol included assessment of both participants’ impressions of using the PESQ and of their assessments of its utility and acceptability in clinical practice. We identified 6 themes addressing the former—3 concerning the content of the questions and 3 the design of the tool—and 4 themes addressing the latter. These are summarized in Multimedia Appendix 3.

The experiences of our respondents provide general guidance for the development of patient-completed CDAs in the emergency setting, as well as for subsequent refinement of our specific tool. In general, respondents found a simple online questionnaire tool acceptable and understandable, though they were divided on their assessment of likely clinical benefit. They highlighted some specific features of such tools that they found of particular value; other uses to which such a tool could be put to support the patient navigation of their emergency assessment pathway; and identified important considerations in their design to maximize acceptability and potential benefit.

Some respondents found that having a questionnaire as a prompt for ictal recall and description was itself of value. It could serve as an aide-memoire for particular ictal experiences (“it…highlighted…some of the things that applied to me”; TL109) or even could provide the conceptual resources for articulating experiences that are “hard…to put into words yourself, not understanding it” (TL169). Previous work comparing structured and open interviews to extract seizure histories has demonstrated the value of such systematic closed questioning to enhance the yield of ictal history-taking [30,31] and support patients’ self-understanding and interpretation of experience [32]; our respondents suggest similar benefits can be found when answers are elicited in a patient-administered, self-report format (more practical in resource-limited settings, with online capture easily facilitating computer-aided interpretation of results). This is consonant with previous work demonstrating that, in general, electronic data collection is more effective than paper-based administration of the same questionnaires [6,33].

However, our participants also highlighted the limitations of this format. They described various ways in which a finite symptom list, with binary yes or no answers, limited their ability to articulate their experience—whether from a surfeit of irrelevant questions, a lack of relevant ones, or feeling constrained by the insistence on the dichotomous presence or absence of a given experience, regarding which they might feel ambivalent or uncertain. Part of this issue may stem from a failure of the study team to articulate fully the purpose of the questionnaire—not to capture a complete description of the person’s TLoC experience, but rather to use certain previously identified highly discriminating features to predict likely etiology of the TLoC—but in part their concerns reflect those found more widely in the experiences of those completing symptom questionnaires. Exploring conditions as diverse as breathlessness and depression, patients completing symptom questionnaires describe ambivalent reactions to the depictions of their experience thus captured [34,35]. They may seek to reformulate questions, recontextualize them (as with TL101’s avoiding of the present or absent dichotomy, or TL106’s looking for follow-up exploration of symptoms), or reject their framing (as with TL184 concluding that the research was not directed toward their presentation) [35].

Beyond just highlighting this issue, our data suggest means to address it. Respondents were clear on the need for such a CDA to complement, not replace, in-person clinical assessment; and they suggested that a more nuanced description of their experience could be captured by complementing the structured questions with free-text symptom reporting. The combination of questionnaires and AI-supported analysis of free speech seizure descriptions may address this potential deficiency of an approach based on closed questions alone [36,37].

While much of the research on patient-completed CDAs has focused on their role in the management of chronic conditions [4,38], there is increasing recognition of the opportunities they offer in acute and emergency care. In particular, several studies have explored the role of patient-completed symptom checkers in supporting triage at the ED front door [8,9,39]. While self-triage CDAs for unselected presentations have been found to perform inadequately [39], quantitative assessments of acceptability and utility in this setting have generally shown patients are willing and able to interact with such tools [8,9]. Our study enriches this finding with qualitative data exploring patient preferences for interacting with such tools. As well as having a simple, navigable interface, they reported a preference for patient-facing outcomes to demonstrate benefit to them (explored more below).

In order to develop CDAs fit for supporting emergency and primary care practice, it is important to understand not just how the CDA performs in ideal conditions, but also how it will perform in the “real world” and how clinicians will use it in practice [40,41]; in particular, what are the barriers to and facilitators of physician uptake of a new CDA. Not all factors affecting uptake of CDAs depend on the design of the CDA itself; from a thematic analysis of semistructured interviews with emergency clinicians in the UK, Hayes identified that, along with the design of the CDA itself, factors concerning the care providers, the environment of their practice, and the institutions within which they worked might enable or impede widespread use of CDAs [42]. The Ottawa Acceptability of Decision Rules Instrument [43], a validated scale for measuring the likely clinical acceptability of new candidate CDAs for emergency care use, similarly considers factors both related to the CDA itself and the context in which it is to be used. The tool-related factors that affect uptake of CDAs can broadly be divided into 3 main areas of concern: how useful the CDA is; how accessible it is to apply in the working context; and how justifiable its decisions are—whether to patients, other clinicians, or in legal contexts. The 5 core principles of CDA design dictate that they must provide the right information to the right person in the right format through the right channel at the right time [44,45]. The majority of CDAs are designed to be completed and used by clinicians, but attending to the core principles of CDA design shows that the right information, right person, right format, and right time for information provision could also come through patients. Patient completion of CDAs prior to clinician assessment could empower patients to participate in their care and support shared decision-making [44], while reducing the time burden on clinicians and so improving the utility and accessibility of the CDA. Computerized CDAs already exist for the management of chronic conditions such as hypertension and multiple sclerosis [46,47]; patient-facing checklists have also been used successfully in ED settings to support patient-centered care [2]. Before real-world implementation, the PESQ would require further research addressing these implementation challenges, as well as multicenter validation of its performance.

It is important that such tools be able to be implemented in the ED setting, not only to allow swift and accurate triage but also because it maximizes capture of reliable information. The richness of subjective symptom descriptions decreases with time from the event for many TLoC presentations, reducing the utility of information captured [48]. Our respondents demonstrated this, with several unable to recall elements of their TLoC they reported in the CDA or even using the CDA itself. Allowing patients to return to their answers to support recall for other clinical purposes would be one means of enhancing the patient utility of such tools. Allowing delayed access to review answers through an online platform would also provide patients with the opportunity to reflect on their responses, an option valued in other work on patient-completed CDAs [38].

Implementation in an ED setting would therefore benefit both from an online application to which users could return after their initial attendance and from hardware available in the ED so that patients could interact with the CDA as early as possible in their pathway. In this study we achieved this through making available to participants in the acute hospital setting a tablet PC, through which they could complete the CDA in its browser-based form; but making it online in this fashion also allowed home completion at a later date. However, further implementation research would be necessary to see how successfully this approach could be used for all patients, including, for example, those with language barriers or who were transiently disoriented at the time of presentation.

Respondents drew a distinction between clinical benefit—that which achieves the assessing clinician’s aims (in particular, diagnosis)—and patient benefit—addressing the patient’s immediate needs (prominently, information and management of uncertainty) [24]. They proposed ways in which a CDA could be used for patient benefit. Access to answers after the time of completion (eg, through a print-out, or option to log in and return to questionnaire answers) could help them as a reference point in later assessment and care. They also indicated that the time of questionnaire completion would be an ideal opportunity to address patients’ informational needs. An online interface could easily provide patients as well as clinicians information about the results of the CDA and its implications for ongoing management or links to explanations of assessment pathways and interim management while awaiting definitive diagnosis. As reported elsewhere [24], these were predominant concerns of respondents after their first assessment for TLoC, and this represents a significant opportunity to add value to patient-completed CDAs.

While it was not a theme explored extensively by our participants, perhaps because of the study literature clarifying the protocol for research data storage that would differ from widescale health care implementation, data governance will always be important for online CDAs. As with any digital health tool, future iterations of PESQ should consider patient data privacy and security, ensuring compliance with standards such as General Data Protection Regulations (GDPR) or the US Health Insurance Portability and Accountability Act (HIPAA).

Existing tools for supporting CDA development and implementation (such as the Theoretical Domains Framework [38,49] or GUIDES checklist [13]) provide high-level overviews of barriers and enablers of implementation; our data allow for more granular practical guidance. On the basis of our data, we offer a set of good practice points for the development of patient-completed online CDAs, summarized in Table 2.

We note some important limitations. The inclusion criteria for the quantitative study necessarily left some groups underrepresented, particularly those with insufficient English language proficiency to complete the questionnaire or learning disabilities that would similarly make self-completion difficult. To an extent this is by design—since cross-language and cross-cultural validation would be necessary anyway after development of the English-language version of such a CDA prior to its employment in different languages. We did include within our sample a diverse age range, including older populations who may be less familiar with online tools—it is a strength of this study to note that even those who self-described as “not a hundred percent au fait” with computer use were able to use the tool without difficulty. Previous research has suggested that older adults are less likely to engage with tablet-based mobile health interventions [4], but that when they do are no less likely to complete them than younger patients, suggesting that perceptions of digital literacy may be lower amongst older patients than actual inability to engage with such tools. To remove such barriers, CDA design can emphasize aspects more likely to be acceptable in this demographic [50], such as simple user interfaces and minimizing numbers of clicks and scrolls, without intrusive features like pop-ups.

We also recruited only a small proportion of those eligible to participate. Anecdotally, the study team noted that recruitment rates appeared higher when potential participants were approached directly and contemporaneously with their assessment, rather than retrospectively by letter. Future research could assess differential recruitment rates and effects on sample representativeness of different recruitment pathways.

Additionally, as a measure of acceptability of such a CDA, we only describe qualitative findings; further work could complement this with quantitative data, both survey responses identifying users’ evaluations of acceptability (along dimensions guided by the data presented in this study) and objective measures, eg, noncompletion of the questionnaire.

In reporting the main quantitative outcomes of this study, we discuss the further validation and implementation research necessary to determine the practical value of the use of such a CDA in routine practice [23]. The results from this study highlight the need to combine this with further research refining the acceptability and utility of the CDA to patients and clinicians. Furthermore, before application in different cultural contexts and different languages, cross-language and cross-cultural validation would be required. Our present research provides guidance—summarized in Table 2—for how such a program of research could proceed.

In this study we find that an online, patient-completed questionnaire tool to support a CDA for the differential diagnosis of TLoC is generally acceptable to users, though opinions diverge on the likely utility of such a tool. Respondents highlighted the distinction between clinician and patient benefit—noting that while interests overlap, they are not coextensive—and offered means to improve the tool such that, in addition to supporting a CDA for clinical benefit, it could address the informational needs of patients at the time of their first assessment for TLoC.