This was a qualitative study relating to the proposed implementation of the PROMise model. The study was conducted in the outpatient rheumatology clinic in Singapore General Hospital. Singapore General Hospital is the largest tertiary hospital in Singapore and serves patients from across the country. In-depth interviews were conducted from November 2024 to May 2025 to elicit both the HCPs’ and patients’ views on the potential implementation of the PROMise model.

The proposed PROMise model remotely monitors patients with axSpA through an automated review of laboratory tests and PROM results. Stable patients will be identified during their routine visits based on a BASDAI score of lower than 4, which indicates that disease activity is inactive in the patients [21]. The patients will then be given a PROM questionnaire, in this case BASDAI, to fill out prior to their upcoming automated review. The BASDAI questionnaire can be completed remotely in the patients’ homes. The schedule of their next appointment will then be determined algorithmically based on the outcome of their automated review (see Figure 1). Patients will have a minimum of one in-person visit to the rheumatologist per year. The PROMise model is under development at the time of this writing, and once refined, is expected to be integrated as part of routine assessment in all future follow-up visits.

Patients were eligible if they were aged 21 years and older and had been diagnosed with axSpA by a rheumatologist. All patients fulfilled the 2009 Assessment of SpondyloArthritis International Society criteria with a BASDAI score of <4 [2]. HCPs were eligible if they were caring for a patient with axSpA at the time of this writing. HCPs had a diverse variety of roles in the routine care of patients with axSpA; those that could potentially be affected by the implementation of the PROMise model were of particular interest. This included rheumatologists, nurses, allied health professionals, and clinic operations staff. Participants were purposively sampled from a list of patients, who were screened based on eligibility criteria. Purposive sampling was used to ensure representation across key demographic and clinical characteristics. Both patients and HCPs were purposively recruited based on sex, age, and ethnicity. Patients were additionally recruited based on the number of years since diagnosed with axSpA, while HCPs were recruited based on seniority and their role in the care of patients with axSpA. A sampling matrix that included gender, age, ethnicity, and years since diagnosis for patients and years of clinical experience and professional role for HCPs was used to monitor key participant characteristics throughout recruitment. Recruitment was iteratively adjusted to ensure diversity across these characteristics, with targeted efforts made to include underrepresented groups. The eligible patients were approached in person at the outpatient rheumatology clinic by a research staff who explained the purpose of the study.

The CFIR framework was used as a guide to design a standardized, semistructured interview guide (Multimedia Appendix 1) [22]. Individual in-depth interviews were conducted and audio-recorded for both HCPs and patients by 2 trained facilitators (HZ and XRC). Field notes were taken when needed. The interviewers were not part of the care team for the participants. The interviews lasted around 30 minutes on average and were conducted remotely via online video conferencing. Data saturation was defined as the point at which no new codes or subthemes emerged within the CFIR domains represented in this study. Saturation was achieved with at least one code or subtheme identified per domain and confirmed through 3 additional interviews for both patients and HCPs, which did not yield new codes or subthemes within these domains. We also collected basic demographics such as age, sex, education, and ethnicity, as well as the number of years since axSpA was first diagnosed for patients and the number of years in their position at the time of this writing for HCPs.

The interviews were transcribed verbatim, and the transcripts were checked against the initial audio recording by the coders (HZ, MH, ZX, and XRC) to ensure accuracy and enhance familiarity with the dataset. Coding was performed by independent coders who worked in pairs. Transcripts were thematically analyzed based on Braun and Clarke [23-25], adopting a structured, codebook-based approach and ensuring clarity and transparency in analytic procedures. Each coder independently reviewed transcripts and generated initial codes. The coders then met to compare coding, discuss differences, and organize codes into potential categories and subthemes. A codebook was developed and iteratively refined to document coding definitions, decisions, and subtheme development. Coding discrepancies were discussed and resolved through consensus, with input from a third researcher (YHK) where required. Regular team discussions were held to review and refine the coding framework and thematic structure to enhance analytic rigor. To enhance trustworthiness, several strategies were used. Credibility was supported through investigator triangulation and repeated engagement with the data. Dependability was strengthened by maintaining an audit trail of coding decisions and codebook development. Confirmability was addressed through team-based discussions to minimize individual bias. Transferability was facilitated by providing contextual details of the study setting and participants to enable assessment of applicability to other contexts. NVivo software (version 15; Lumivero) was used for data management. The reporting of qualitative methods followed the Consolidated Criteria for Reporting Qualitative Research (COREQ) checklist (Checklist 1) [26].

The CFIR-ERIC matching tool was used to develop and outline the expert-endorsed Level 1 and 2 strategies. The identified CFIR barriers were input into the CFIR–ERIC matching tool. The tool’s algorithm is derived from an expert consensus panel that ranked how well different implementation strategies address specific CFIR barriers. It then generates a prioritized list of strategies by aggregating the importance ratings assigned by experts across the selected domains [17]. In the development of the CFIR-ERIC matching tool, experts were invited to endorse their top 7 strategies for overcoming a specific CFIR barrier. A total of 18 Level 1 and Level 2 strategies were endorsed by more than 50% of experts and between 20% and 50% of experts, respectively. The CFIR-ERIC output was then analyzed to determine suitable implementation strategies to overcome the main barriers identified.

This study was approved by the SingHealth Centralized Institutional Review Board (Ref No. 2024‐4044). Informed written consent was obtained from all participants before the interviews. All data collected from participants were deidentified. As a token of appreciation, each participant received an SGD $50 (US $39.34) coupon upon completion of the interview. The study was conducted in compliance with the principles of the 1964 Declaration of Helsinki.

A total of 32 interviews were conducted. The participant characteristics are summarized in Table 1. The mean (SD) age for patients was 39.4 (11.7) years and for HCPs was 37.9 (7.2) years, respectively. The HCPs were employed in a diverse range of roles in the clinical setting, from clinic operation staff to rheumatologists; all were involved in the care of patients with axSpA. The HCPs had between 1 year and 23 years of service (Table 1).

Using CFIR as the a priori framework for identifying the barriers and facilitators to implementation of the novel PROMise, all 5 domains (innovation, outer setting, inner setting, individuals, and implementation process) were elicited. We identified 4 main facilitators to the implementation of the PROMise model, which are presented in Table 2.

We identified 4 main barriers to the implementation of the PROMise model (Table 3).

Overall, the CFIR-ERIC matching tool produced a total of 35 Level 1 and Level 2 strategies (Table 4). There were 8 Level 1 strategies, which had expert agreement of more than 50%. Of the 8 Level 1 ERIC strategies, 7 strategies cut across all 3 barriers of innovation cost, individuals (need), and planning. The strategy of accessing new funding cut across only the barriers of innovation costs and planning, as it did not address the needs of patients. The patients required additional reassurance that they would still receive timely medical attention under this model of care, which could not be resolved by accessing new funding.

The cumulative percentages exceeded 100% because experts were invited to endorse the same implementation strategy for multiple CFIR barriers independently.

This study sought to identify and understand the facilitators and barriers to implementing a novel model of care for patients with axSpA (PROMise) from both the perspectives of patients and HCPs using the CFIR framework. Facilitators included (1) reduced inconvenience and costs for patients and reduced patient load in the clinic, (2) need for an alternative model of care, (3) similarity to these workflows, and (4) suitable patient selection for PROMise. Barriers included concerns for the (1) financial sustainability of PROMise, (2) cultural conditions, (3) patient safety, and (4) increased workload for HCPs. To the best of our knowledge, this is the first study to break down the facilitators and barriers to the implementation of a novel model of care and suggest implementation strategies to address barriers identified prior to the formal implementation of the novel model of care. This study provides insights for future studies that aim to implement remote monitoring strategies in axSpA or other chronic diseases.

The findings of cost and increased workload being barriers are common in the implementation process [27,28]. In our study, even though there were concerns with the direct costs of the new model of care, the greater concern of the HCPs was the lack of clarity regarding the financial structure of the proposed model of care. HCPs also mentioned a lack of clarity on how to set aside dedicated time to review patients’ laboratory tests and PROMs results who require follow-up in the PROMise model. Although both ERIC strategies of accessing new funding and developing an implementation blueprint were recommended for the mitigation of the abovementioned barriers, developing an implementation blueprint would be more relevant in this particular context. The key features of the implementation blueprint should include cost evaluation, patient billing, and detailed workflow for not just clinicians but all HCPs involved in the PROMise model. Hence, an implementation blueprint would shed light on the financial sustainability of the model of care and increase workload transparency for the HCPs [29].

Both patients and HCPs brought up considerations regarding technological and language literacy. Poor technological literacy has been described in local publications [30]. Even though the vast majority of Singapore residents are literate in English (97.1%), there is still a small minority who will face challenges completing the PROM questionnaire in English [31]. For patients who either have challenges in technological or language literacy, more efforts will be needed to determine how support can be provided. Furthermore, these patients are likely to be older and have an increased likelihood of co-morbidities, hence requiring closer monitoring from HCPs [32,33]. However, patients with axSpA are likely to be younger, as pointed out by the HCPs. Hence, the majority of the patients with axSpA will still benefit from the PROMise model. For the small minority of patients with lower technological and language literacy and/or more severe illness, it is likely a longer transition period will be required to ease these patients into the PROMise model as suggested, or to determine if they are suitable at all.

Patient safety was a concern for a minority of patients. Patients desired a direct avenue to seek help in between the scheduled reviews by the algorithm. This was in contrast to findings by other studies where other patients felt a sense of security under a remote monitoring model of care [34]. However, patients in this study did mention that their concerns would be greatly alleviated if they were able to contact the clinic staff directly regarding the scheduling of appointments if they experience worsening symptoms. Hence, the future implementation of the PROMise model should include and inform patients of the access to a direct helpline to address their safety concerns.

With regard to facilitators from the patients’ perspectives, the benefits of remote monitoring are well documented. Our study found similar facilitators such as patient convenience and reduced traveling costs [35,36].

Based on the recommended ERIC strategies of conducting a local needs assessment, assessing readiness, and identifying barriers and facilitators; and involving, obtaining, and using patient and family feedback, continuous feedback from patients and HCPs will be collected and used to improve the PROMise model. The continued feedback from patients and HCPs is integral to the sustainability of the PROMise model after implementation [37].

The strengths of this study include purposive sampling of the HCPs involved in the care of patients with axSpA in Singapore General Hospital. We ensured the inclusion of a wide range of ages, levels of clinical experience, and diverse clinical and nonclinical roles that would be affected by the implementation of the PROMise model. Therefore, our data captured a broad range of views and was representative of the views of HCPs that would be affected by this model of care.

There are a few limitations to our study. First, our study focused on interviews with patients and HCPs involved in the care of patients with axSpA. We lack representation of other stakeholders such as the health care leadership. These stakeholders may have provided a more comprehensive understanding of the barriers and facilitators of the implementation process. However, we tried to mitigate this by including staff involved in operations to understand the barriers and facilitators to the implementation of PROMs. Moving forward, we will solicit feedback during our meetings with the health care leadership. Second, as the proposed PROMise model of care has not been piloted yet, the patients and HCPs interviewed have not had first-hand experience with the model of care. Hence, the perspectives that they provided were only based on their perception of what a PROMise is. They may have a different perspective after they have experienced the model of care personally, especially after the implementation process has been tailored according to recommendations. Hence, we will consistently solicit feedback from patients and HCPs after the model of care has been piloted to better understand and improve upon the implementation process.

Our study found a wide range of facilitators and barriers to the implementation of our novel PROMise model of care, where patients with axSpA will be monitored remotely through laboratory tests and PROM results. We also presented the ERIC strategies that would possibly mitigate the identified CFIR barriers. These findings will guide the implementation of the PROMise model care in patients with axSpA. Future works should evaluate the implementation success of the improved PROMise model of care.