Abstract

Objective Patient-initiated follow-up (PIFU) for rheumatoid arthritis (RA) is a model of care delivery wherein patients contact the clinic when needed instead of having regularly scheduled follow-up. Our objective was to investigate the influence of different patient eligibility characteristics on the number of potentially deferred visits to inform future implementation of a PIFU strategy.

Methods We conducted a retrospective chart review of 7 rheumatologists’ practices at 2 university-based clinics between March 1, 2021, and February 28, 2022. Data extracted included the type and frequency of visits, disease management, comorbidities, and care complexities. Stable disease was defined as remission or low disease activity with no medication changes at all visits. The influence of patient characteristics on the number of deferrable visits in patients with stable disease was explored in 4 criteria sets that were based on early disease duration, medication prescribed, presence of care complexity elements, and comorbidity burden.

Results Records from 770 visits were reviewed from 365 patients with RA (71.5% female, 70% seropositive). Among all criteria sets, the proportion of visits that could be redirected varied between 2.5% and 20.9%. The highest proportion of deferrable visits was achieved when eligibility criteria included only stable disease activity and patients with RA on conventional synthetic disease-modifying antirheumatic drugs or no medications (n = 161, 20.9%).

Conclusion PIFU may result in a more efficient use of specialist healthcare resources. However, the applicability of such models of care and the number of deferred visits is highly dependent on patient characteristics used to establish eligibility criteria for that model. These findings should be considered when planning implementation trials.

Key Indexing Terms:

Rheumatoid arthritis (RA) is a debilitating autoimmune condition that, when left untreated, can result in progressive disability, systemic complications, and premature death.1,2 Although curative treatment for RA does not exist, early therapeutic interventions with disease-modifying antirheumatic drugs (DMARDs) can improve quality of life and reduce disability.3-5 The current standard of care for RA involves a treat-to-target (T2T) paradigm,6,7 including regular assessments of disease activity, with treatment modifications made if disease targets are not met. The interval for follow-up is typically determined by the rheumatologist based on an estimation of when the patient may require reassessment.

Although T2T has improved RA health outcomes, this approach results in lifelong regular specialist care, which is challenging given current and projected rheumatology workforce shortages.8-10 This approach results in a substantial proportion of visits for patients with otherwise stable disease. For example, in 1 center, 49% of RA follow-up appointments required no medication modifications and 38% were deemed unnecessary in the opinion of the attending rheumatologist.11 High volumes of follow-up visits for individuals with stable disease may affect access to speciality care for both new patients and existing patients with flares and other urgent health concerns. Given strained healthcare systems,12 new models of care are urgently needed to improve access to care when patients need it most.

To address these challenges, various care delivery models have been implemented, often using alternate care providers for follow-up of patients, to offload specialist care. One commonly investigated model includes nurse-led care.13 Although nurse-led care has been demonstrated to be effective and acceptable to patients,13-15 such models can exceed follow-up capacity and result in limited access over time.16 Nurse-led models also typically require rheumatologist supervision for DMARD prescription, and funding for interdisciplinary care models is challenging in many jurisdictions. Alternatively, primary care providers (PCPs) may support shared care models for patients with chronic diseases, including RA17; however, shortages of PCPs18 may make a shared care model unfeasible. Further, RA is currently not managed in primary care settings in many regions,11 leading to further implementation challenges, including the need for additional PCP rheumatology training.

Patient-initiated follow-up (PIFU) care models have been proposed as a strategy to better match patient needs to follow-up care.19 Patients are instructed to contact the rheumatology clinic in the case of flaring disease, new health concerns, or other comorbid conditions that require specialist follow-up instead of routinely scheduled follow-up appointments, to better meet patient care needs in a timely manner. Patients who are experiencing flares or other concerns are normally seen by a rheumatologist within 10 to 14 working days.19 Such models of care have been largely associated with reductions in rheumatology clinic visits with patient outcomes similar to those of usual care.19 PIFU models of care have been implemented in various ways, with examples in the United Kingdom20 as well as other jurisdictions.19

One of the challenges of introducing new models of care with alternative follow-up strategies is ensuring appropriate patient selection. PIFU care models have typically been used for patients with more stable disease.19 However, “stable disease” may be variably defined. Whereas disease activity scores and treatment modalities are commonly employed to ascertain patients’ suitability for alternative follow-up care models, the influence of social determinants of health (such as socioeconomic status and access to primary care) and additional complexities associated with patient care (including comorbidities, frailty, and mental health issues) have been inadequately considered in model development.19

To improve prompt access to rheumatological care at our site, we plan to pilot a PIFU strategy. We envision that deferral of stable visits will enable rheumatologists to provide more timely care for patients with acute care needs and to see new referrals, resulting in optimized population outcomes. To support planning for this new model of care, the objective of the present study was to investigate how different patient eligibility characteristics would influence the potential number of deferred visits.

METHODS

Study design. We conducted a retrospective chart review of 7 rheumatologists’ practices within 2 university-based clinics in Alberta, Canada. We included charts of individuals with RA and reviewed all rheumatology visits (in-person and virtual) between March 1, 2021, and February 28, 2022, to obtain a recent sample.

Patient selection. Recruitment of patients occurred with the assistance of Alberta Health Service Health System Access. Participating healthcare providers (HCPs) supplied their practice identifier, and a random subset of individuals from their clinic rosters with at least 1 billing code for RA in the last year who were aged ≥ 18 years were selected for chart review. We estimated, based on prior work,11 that approximately 50% of individuals could be eligible for PIFU care based on the stability of the disease course and a disease activity status of remission/low disease activity (LDA). For our study, we estimated that we would require 385 individuals to achieve a similar proportion of individuals with stable disease (95% CI 45-55%).21 We oversampled by 10% to account for any loss in eligible cases due to errors in identification of RA, for a target sample of 424 patients.

Data collection. We conducted chart reviews of electronic medical records to obtain accurate information on patient characteristics and rheumatology visit processes of care. A standardized abstraction tool was developed in REDCap22,23 (hosted by the Clinical Research Unit, Cumming School of Medicine, University of Calgary) and pilot-tested prior to use. We also developed a data abstraction guide that outlined instructions, definitions, and decision rules for each variable. Four independent abstractors (SS, KD, NMSH, KW) were trained and subsequently conducted the chart reviews. Twenty charts were reviewed in duplicate for training purposes; charts were reviewed by a single abstractor.

Patient and visit characteristics. We reviewed patients’ charts to determine duration of rheumatology care, frequency of rheumatology visits, modality of rheumatology visits (in-person, virtual telephone, virtual video conference), and rheumatology HCPs at each visit.

The Rheumatic Disease Comorbidity Index (RDCI) is a validated index that was used to assess the comorbidity burden in all patients, chosen for its excellent performance in accounting for comorbid illness, particularly when studying death and physical functioning outcomes using administrative data.24 We reviewed all rheumatology consult notes within the patient charts in our specified time frame to identify 11 comorbid conditions included in the RDCI. Each comorbidity was scored using the RDCI’s predetermined scale, which is based on its potential effect on an individual’s health. These scores were then summated to calculate the total score for the RDCI, which represents the combined burden of all comorbidities.24 RDCI scores range from 0 to 9, with 0 indicating the absence of any listed comorbidity.

Eleven elements associated with care complexity that could affect PIFU were considered during chart reviews based on a patient-reported measure of care complexity called the INTERMED Self-Assessment.25 The elements were recorded as present/absent: housing vulnerability, language barriers, health literacy concerns, frailty, financial/insurance concerns, limited social/familial support, therapeutic challenges (including concordance with treatment plan), mental health concerns, distance from care, and negative experience with HCPs. Current smoking, alcohol, and other substance use patterns were also abstracted. Complexities were considered present if there was documented evidence of the above elements from either the physician or patient perspective.

Disease management at visits. We also examined disease activity and management at each rheumatology visit. Stability of disease was assessed based on data availability in the following priority order: Disease Activity Score in 28 joints (DAS28) based on C-reactive protein (DAS28-CRP), DAS28 based on erythrocyte sedimentation rate (DAS28-ESR), and Clinical Disease Activity Index (CDAI). Patients were classified as having stable disease if (1) their DAS28 or CDAI score indicated they were in remission (DAS-28 < 2.6 or CDAI < 2.8) or LDA (DAS28 between 2.6 and 3.2 or CDAI 2.8-10)26 and (2) no medication changes were made at the visit (including no change, addition, stoppage, or dosage increase or decrease). Where disease activity scores were undocumented or consult notes lacked required information for posthoc DAS28 and/or CDAI calculation, absence of a change in medication was used as a proxy for stable disease. For individuals to be classified as having stable disease, these criteria needed to be met across all visits within the study period.

Assessing influence of patient characteristics on number of deferrable visits. Our prior review of PIFU models of care19 identified that inclusion in these models varied in 4 important ways: disease duration, disease activity, comorbidities, and use of specific medications. We used these in addition to care complexity considerations to define criteria sets (Table 1), which varied based on patient characteristics, including differing definitions of established vs early RA (≥ 1 or ≥ 2 years following diagnosis, or not considered), treatment (conventional synthetic DMARDs [csDMARD] or no DMARD treatment, or biologic DMARD [bDMARD] or targeted synthetic DMARD [tsDMARD]), comorbidity burden (RDCI score 0-2), and care complexity (absent based on criteria described above, or not considered in the eligibility criteria).

Table 1.

Outline of 4 new proposed patient eligibility criteria sets to be considered for patient-initiated follow-up models of care.

Data analysis. Descriptive analysis was conducted using R27 to analyze patient characteristics and estimate the number of deferred visits based on the criteria sets described above (Table 1). Means (SD) and medians (IQR) along with other descriptive statistics (counts and proportions) were calculated.

Ethics. This study was reviewed and approved by the Conjoint Health Research Ethics Board at the University of Calgary (REB22-0487). Informed consent was waived due to the observational nature of the study.

RESULTS

From the clinic rosters of the 7 rheumatologists involved, 569 patients were eligible for inclusion, and 424 were randomly selected for chart reviews, meeting our target sample size. Further screening excluded 59 patients who either lacked an RA diagnosis, had no rheumatology visits within the study time frame, or were not under the care of 1 of the 7 rheumatologists (Supplementary Figure S1, available with the online version of this article). This led to the inclusion of 365 patients and 770 rheumatologist visits for review. Most visits were in person (n = 528, 68.6%), with 242 conducted by telephone (31.4%). The median age of patients was 65.0 years, with 71.5% being female. Approximately 70% of patients were seropositive (rheumatoid factor or anticyclic citrullinated peptide antibody positive) and the median duration of rheumatological care was 7 years. In relation to the study start date, 15.9% of patients received their RA diagnosis within 2 years, and 8.8% within the last year.

At study baseline, 54 (14.8%) patients were in remission, 15 (4.1%) had LDA, 30 (8.2%) had moderate disease activity, 12 (3.3%) had high disease activity, and 255 (69.9%) had an unknown status due to missing disease activity scores. Mean (SD) tender joint counts were 2.7 (5.0) and mean (SD) swollen joint counts 1.9 (3.8). Sixty-eight (18.6%) patients had evidence of joint erosion, 42 (11.5%) had no erosive damage, 20 (5.5%) had pending diagnostic imaging, and 235 (64.4%) had an unknown erosion status within the study period.

Patient treatment characteristics are shown in Table 2. The most common treatment regimen included only csDMARDs in 156 patients (42.7%), followed by 89 patients (24.4%) treated with combination therapy of bDMARD + csDMARD. Other regimens included tsDMARD monotherapy (4.4%), bDMARD monotherapy (14.8%), and tsDMARD + csDMARD combination therapy (8.5%).

Table 2.

Baseline patient characteristics (n = 365).a

Within the 1-year study period, 770 visits were reviewed, with a median of 2 visits per patient. Approximately 69% of patients had 1 or 2 rheumatologist visits during the year, and only 3.3% had ≥ 5. Median (IQR) number of visits in person and by telephone were 1 (1-2) and 0 (0-1) per patient, respectively. The most common care complexities were mental health concerns (10.1%), diagnostic and therapeutic challenges (8.8%), and financial/insurance concerns (8.5%). Table 3 describes the types and frequency of all the care complexities identified. The range of scores on the RDCI was between 0 and 7, with a median of 1. The most prevalent comorbidities included hypertension (27.1%), lung disease (20%), and cardiovascular diseases (14%).

Table 3.

Rheumatology visits and care complexity at patient level (n = 365).a

Table 4 estimates the number of visits that could potentially be deferred during a PIFU model of care for criteria set 1 (defined in Table 1), where patients were either on a csDMARD or had no medication. Patients were not eligible if they had active disease status, advanced therapy (tsDMARDs or bDMARDs), medication changes, and the presence of an element of care complexity. Results were stratified by different definitions of established RA status and different RDCI cut-off values. When criteria included an RDCI value of ≤ 2 and an RA diagnosis of ≥ 1 year, the greatest number of potentially deferred visits (n = 86, 11.2%) was observed due to the decrease in stringency of the eligibility criteria. Contrastingly, the least number of potentially deferred visits (n = 49, 6.4%) was determined when patients had an RDCI value of 0, indicating no comorbidities, and an RA diagnosis of ≥ 2 years (Table 4).

Table 4.

Estimated number of deferred rheumatologist visits if patients on conventional synthetic DMARD or no medication were enrolled in a PIFU care model based on different cut-offs for established RA and comorbidities using the RDCI.

Table 5 estimates the number and proportion of potentially deferred visits if criteria set 2 was used for PIFU inclusion (defined in Table 1) for patients on advanced therapy (tsDMARD or bDMARD). Exclusion criteria were active disease status, medication changes, and presence of an element of care complexity. The greatest proportions of potentially deferred visits were observed when the inclusion criteria consisted of RDCI ≤ 2 and a diagnosis of RA for ≥ 1 year (n = 79, 10.3%), followed by an RDCI ≤ 2 and a diagnosis of RA of ≥ 2 years (n = 75, 9.7%). After accounting for a mandatory 1-year advance therapy renewal appointment in these groups, 30 (3.9%) and 28 (3.6%) visits could be potentially deferred, respectively.

Table 5.

Estimated number of deferred rheumatologist visits if patients on targeted synthetic or biologic DMARDs were enrolled in a PIFU strategy based in different cut-offs for established RA and comorbidities using the RDCI.

Our inclusion criteria in criteria sets 3 and 4 (Table 1) focused solely on treatment modality and disease status, excluding other variables such as care complexities, comorbidities, and disease duration. Criteria set 3 included only patients on csDMARD or no medications who had stable disease status. This criteria set included 99 patients and 161 visits, which resulted in 20.9% of potentially deferrable visits, the highest proportion among all criteria sets (Table 6). For patients taking tsDMARDs or bDMARDs with stable disease status (criteria set 4), there were 43 patients, accounting for 120 visits. After taking into consideration the mandatory 1-year advance therapy renewal appointment, this led to 43 visits or approximately 5.6% of potentially deferrable visits.

Table 6.

Summary comparison of criteria sets and estimated percentages of deferred visits.

DISCUSSION

Given current workforce shortages in rheumatology,8 the focus of our study was to identify the proportion of visits for individuals with RA who had stable disease and could potentially be offered a PIFU model of care. In doing so, rheumatologists could prioritize caring for patients in greater need of specialist management. Our study estimates that within academic rheumatology practices, the proportion of visits identified that could be deferred to this model of care varied between 2.5% and 20.9%, depending on patient eligibility characteristics.

We previously conducted a scoping review of RA PIFU strategies and found that the implementation varied between studies.19 One important area of variation included patient eligibility characteristics. Some studies included a predefined disease duration for inclusion, whereas others included treatment or disease activity and/or stability requirements. A minority considered comorbidities as exclusion criteria, and none explicitly considered elements of care complexity. Although the effectiveness of PIFU was evaluated in most studies, many were underpowered.19 Further, the health equity considerations of the model were rarely investigated or discussed.19 The present study demonstrates that there is variability in the number of potentially deferrable visits depending on patient eligibility criteria. This information should be used in planning future PIFU studies to ensure they are sufficiently powered for the outcomes of interest.

Although the effectiveness of this model of care was demonstrated by one of the first and also the longest randomized trials of patient-initated follow-up in rheumatology by Hewlett et al28 in consecutively enrolled patients, this model of follow-up care may not be appropriate for patients with early disease. Following diagnosis, patients may require time to adjust to the diagnosis, as well as build a trusting relationship with the care team. It also may take time to attain remission or LDA. For this reason, we selected patients with established RA, defined as disease duration of either ≥ 1 or ≥ 2 years following diagnosis. Further, visits were considered potentially deferrable if patients were in LDA or remission, with no treatment changes. We recognize that for some patients it may take longer to attain adequate disease control and some may require more than 1 to 2 years to gain comfort with their knowledge of their own disease to enable participation in such a model of care. For others, lifelong regular rheumatology follow-up may truly be necessary due to disease that is difficult to control. Although they were not part of our study due to the retrospective nature of the work, patient preferences, health literacy, and comfort levels with taking an active role in their healthcare should be considered prior to enrollment in PIFU.

Due to insurance requirements in Canada, patients on bDMARDs and tsDMARDs require annual medication renewal. Further, these medications can be prescribed only by specialists, which precludes returning patients to primary care. For people with well-controlled disease, annual visits could be nurse- and/or pharmacist-led to optimize healthcare resources.13,29 Renewals of medications for patients treated with csDMARDs and participating in a PIFU strategy could be done in either primary care or specialty care, depending on PIFU implementation. Further, in our study, we specified patients must be on stable treatment, meaning no new medications added or removed nor changes in dosing. However, changes in international rheumatology guidelines may need to be considered when developing eligibility criteria for future PIFU strategies. For example, in Australia and Canada,30,31 new guidance suggests it may be appropriate to discuss bDMARD and tsDMARD tapering with patients who have stable disease. Although PIFU may be highly appropriate for individuals participating in treatment tapering, as it could improve access to care for flares, these patients may require more frequent disease monitoring to ensure disease stability. Future studies should investigate the appropriateness and safety of including individuals participating in PIFU during treatment tapering.

Other important considerations for a PIFU model of care include physician and practice suitability and compensation. Mid to late career rheumatologists have higher numbers of patients with stable disease suitable for a PIFU pathway, in contrast to early career rheumatologists. Remuneration is also important to consider in a new care model and there may be different risks and benefits to PIFU in academic and community practices. Last, physician burnout should be considered as it may be affected by PIFU models, as physicians would see a higher volume of more urgent and complex cases in daily practice.

Many patients with RA have comorbid conditions,24 with pulmonary and cardiovascular disease associated with higher mortality.32,33 There are no clear guidelines about the appropriateness of including patients with specific comorbidities and/or a high overall comorbidity burden in PIFU models. Our study explores the effect of comorbidity burden, using the RDCI24 as a criterion for PIFU inclusion. We selected the RDCI given evidence that the total burden of comorbidities is associated with worse RA outcomes34-37 and increased mortality.38,39 Conversely, well-managed comorbidities, such as hypertension or diabetes, may not add significant specialist care complexity as these are managed in primary care. Therefore, it is possible our evaluation using RDCI thresholds could have excluded patients who may otherwise have been appropriate for PIFU. Conversely, there may be comorbidities not included in the RDCI that add significant rheumatology care complexity and could be considered as exclusion criteria for PIFU. Examples include mental health conditions other than depression (included in the RDCI) or dementia. Chronic pain may also complicate PIFU models as it is more prevalent in individuals with difficult-to-treat RA.40 It is also possible that through repeated scheduled follow-up in speciality care, patients may receive more screening for comorbidities. Future studies of PIFU should therefore investigate the consequences of this model of care on patient comorbidity identification and management alongside RA-specific outcomes.

There are many potential factors beyond comorbidity burden and RA disease status that can add to care complexity and affect patient outcomes. Examples include social determinants of health, such as lower health literacy, fewer social supports, and strained financial resources.41-43 Our study attempted to identify such factors that could potentially influence the appropriateness for patient inclusion in PIFU. Currently, in our center, there are no routine strategies for screening patients to identify social determinants of health or characteristics that could increase care complexity. Nevertheless, where documented, our study captured a substantial burden of care complexity for some individuals that should be considered in future PIFU implementation. Compounding these issues, Statistics Canada reported in 2019 that 15% of Canadians stated they did not have a regular PCP.18 This was even higher for individuals from rural, immigrant, and refugee communities, and this issue has been further exacerbated in many regions due to the coronavirus disease 2019 (COVID-19) pandemic. This is particularly relevant, as some PIFU strategies19 require a PCP to monitor blood work to assess for bone marrow suppression and hepatic or renal toxicity that may complicate RA treatment; PCPs may also require additional training to manage RA monitoring. These examples highlight how care complexity, and the lack of access to primary care, may hinder the safe transition of patients with stable disease to alternative models of speciality follow-up care.

Although our study has been informative for our long-term aim of implementing a PIFU model, there were limitations. Our sample was identified at random from 7 rheumatologists at 2 clinics from the same academic center, and findings may not be generalizable to community practices or different geographic regions. Patients with RA were identified with at least 1 billing code, which may have influenced sampling. Patient charts were reviewed by a single abstractor due to funding and time limitations, and it is possible that abstraction in duplicate could have resulted in different findings for some variables. All abstractors were encouraged to bring questions for review at regular team meetings, which likely reduced abstraction differences. Rheumatology consult notes also varied between structured and unstructured formats among rheumatologists, leading to incomplete disease activity ascertainment. As a result, absence of medication changes was used as a proxy for stable disease; future work is ongoing at our site to improve collection of disease activity scores. Serologic status was not considered as one of the PIFU criteria as it had not emerged as a predictor of outcomes in our prior review; however, this may be interesting to collect as a variable in future studies. Patients who were on no medications were not analyzed in terms of their need for follow-up; however, this could be a future direction for study. Also, as our study period coincided with the COVID-19 pandemic, it is possible that patient preference for virtual and/or telephone visits increased, posing further challenges in objectively assessing disease activity or other aspects of patient care. Similarly, the lack of systematic documentation of care complexity elements may have resulted in underreporting. We also did not examine the role of pregnancy and/or pregnancy planning as exclusion criteria for PIFU models of care.

Our results confirm that there are patients with stable disease who could be eligible for PIFU, although possibly a lower proportion than previously reported11; this needs to be considered when powering future implementation trials. Nevertheless, redirecting individuals with stable RA to a PIFU model may result in more efficient use of healthcare resources. Although we considered various patient characteristics in our criteria sets to evaluate the influence on the number of potentially redirectable visits, we recommend a more nuanced approach for implementation. We suggest using shared decision making between the rheumatologist and patient to review the risks and benefits of PIFU. Further, there remain many elements of the PIFU model beyond patient inclusion criteria that require development. For example, PCPs involved in a PIFU model need a defined care pathway, including timely return to speciality care without re-referral. Additionally, physician and practice suitability, along with compensation models, need to be considered. In our center, work is ongoing to draw upon our findings, as well as upon models developed in other jurisdictions,19,20 to define and pilot a PIFU care pathway.

ACKNOWLEDGMENT

Data were extracted from the Alberta Health Services Enterprise Data Warehouse with in-kind support provided by the Alberta SPOR SUPPORT Unit Data and Research Services team.

Footnotes

This study was funded by a grant from the Canadian Initiative for Outcomes in Rheumatology cARE (CIORA). CEHB is supported by funding from Arthritis Society Canada (ASC) Stars Career Development Award funded by the Canadian Institutes of Health Research (CIHR) Institute of Musculoskeletal Health and ASC (STAR-19-0611/CIHR SI2-169745). SS is supported by the Canadian Rheumatology Association Foundation. SLM is supported by funding from the ASC Stars Career Development Award (STAR-18-0189). EL is supported by funding from the Health System Impact Fellowship Award funding by the CIHR and Alberta Health Services. DL is supported by the Mary Pack Chair in Rheumatology from the University of British Columbia and ASC.

MRWB has received consulting fees from AbbVie, AstraZeneca, GSK, Janssen, and Sanofi-Genzyme, unrelated to the current work. The remaining authors declare no conflicts of interest relevant to this article.

Accepted for publication January 2, 2024.Copyright © 2024 by the Journal of Rheumatology

This is an Open Access article, which permits use, distribution, and reproduction, without modification, provided the original article is correctly cited and is not used for commercial purposes.