Evaluating Factors Associated With Telehealth Appropriateness in Outpatient Rheumatoid Arthritis Encounters Using the Encounter Appropriateness Score for You (EASY)

METHODS

EASY score collection. We conducted a prospective cohort study of outpatient rheumatology encounters at Duke Health from January 1, 2021, to August 31, 2021. During the study period, all Doctor of Medicine (MD)/Doctor of Osteopathic Medicine (DO) faculty, advanced practice providers (APPs), and fellows in the rheumatology division were invited to document EASY scores after each rheumatology outpatient encounter with an established patient (ie, after each follow-up visit). The development of the EASY score and evidence of its validity have been previously published.^10,11

The EASY scoring system is intended to reflect providers’ perceptions of telehealth appropriateness for each outpatient encounter after the visit is concluded.¹⁰ In response to the prompt, “Which of the following encounter types would have been most appropriate for today’s visit (irrespective of the pandemic)?” providers rate their own encounters as follows: in-person or telehealth acceptable (EASY = 1), in-person preferred (EASY = 2), or telehealth preferred (EASY = 3; Supplementary Figure S1, available with the online version of this article). For the purposes of this study, telehealth appropriateness was evaluated as a binomial variable based on whether telehealth was acceptable or preferred for the encounter (EASY = 1 or 3) or in-person evaluation was preferred (EASY = 2).

Both in-person and telehealth (video or telephone) encounters were eligible to receive an EASY score. Only follow-up visits that received an EASY score were included in the study because new patient visits are generally conducted in-person. Encounters with patients who were not seen by a rheumatologist in our health system prior to the beginning of the study period were excluded in order to allow for comparisons with previous encounters, satisfy our RA definition, and avoid unintentionally including new patient visits in our analysis. The study did permit the use of multiple EASY scores per individual patient, provided each EASY score reflected a unique outpatient rheumatology encounter. EASY scores from providers who documented < 20 EASY scores during the study period were excluded to avoid including scores recorded by providers with minimal experience using the EASY scoring system.

Defining RA encounters. This study was limited to outpatient rheumatology encounters with adult patients with RA. To satisfy this requirement, an algorithm was employed to identify patients with RA from the electronic health record (EHR). This algorithm required both documented RA International Classification of Diseases, 10th revision (ICD-10) codes from ≥ 2 discrete outpatient encounters and a disease-modifying antirheumatic drug (DMARD) prescription during the study period.¹² An additional targeted search for patients with RA who were not receiving DMARD therapy was performed among outpatient rheumatology encounters using only ICD-10 codes for RA, and manual chart review was performed to confirm the RA diagnosis for these patients.

The medical record of each patient with RA was then screened against a comprehensive list of ICD-10 codes for other autoimmune inflammatory conditions, including systemic lupus erythematosus, Sjögren syndrome, systemic sclerosis, myositis, mixed connective tissue disease, undifferentiated connective tissue disease, juvenile idiopathic arthritis, seronegative spondyloarthropathies, systemic vasculitis, and sarcoidosis, among others. If an ICD-10 code for one of these conditions was present, manual chart review was performed to verify the diagnosis, and all patients with a confirmed additional or alternative autoimmune inflammatory disease were excluded.

Variables of interest. Variables evaluated as potential factors associated with telehealth appropriateness were extracted from the EHR for each outpatient RA encounter that received an EASY score. These variables were evaluated at the encounter level rather than the patient level to better reflect real-time decision making by providers in response to changing clinical scenarios over time.

Demographic variables included age, sex, self-reported race and ethnicity, insurance type, and service area (ie, distance traveled to clinic by zip code). RA clinical characteristics included seropositivity status, number and type of DMARDs used, patient-reported disease activity measured by the Routine Assessment of Patient Index Data 3 (RAPID-3) score,¹³ and comorbid noninflammatory causes of joint pain (ie, noninflammatory arthritis composite, osteoarthritis, fibromyalgia).

Data on DMARD use were extracted from the medication list attached to each encounter and were divided into conventional synthetic DMARDs, biologic DMARDs, and Janus kinase inhibitors. If no DMARDs were identified using this method, manual chart review of the encounter was performed to capture potential missing medication data. RA disease activity was assessed categorically with RAPID-3 scores: < 1 = remission, 1.1-2 = LDA, 2.1-4 = MDA, and > 4 = HDA.¹³ The presence of comorbid noninflammatory arthritis, osteoarthritis, or fibromyalgia during the encounter was evaluated using ICD-10 codes attached to the encounter, and the noninflammatory arthritis composite included unspecified forms of arthritis/arthropathy, bursopathies, enthesopathies, hypermobility, myalgias, and soft-tissue disorders related to overuse and pressure.

Previous and current encounter characteristics were also assessed, including previous encounter modality (in-person vs telehealth), whether the previous encounter was conducted by a different provider, current encounter modality (in-person vs telehealth), and time since last visit.

Finally, we assessed variables describing provider characteristics. These provider characteristics included average provider preference for telehealth from prior encounters and average provider modality (in-person vs telehealth) for prior encounters. Average provider preference for telehealth was assessed on a scale of 0-1, where a value of 0 implied that all of a provider’s previous EASY scores indicated in-person preferred (EASY = 2), and a value of 1 implied that all of a provider’s previous EASY scores indicated telehealth was appropriate (EASY = 1 or 3). Average provider modality on prior encounters was also assessed on a scale of 0-1, where a value of 0 implied that all of a provider’s previous encounters were conducted in-person, and a value of 1 implied that all a provider’s previous encounters were conducted by telehealth.

Statistical analysis. Robust Poisson regression with log link was used to evaluate the association of the variables of interest with telehealth appropriateness (telehealth appropriate [EASY = 1 or 3] vs in-person preferred [EASY = 2]) of outpatient RA encounters. Associations were evaluated at the encounter level rather than the patient level, as previously described. As a result, individual patients could have multiple encounters within the health system during the study period. To account for repeated measures, the model was fit using generalized estimating equations with an unstructured correlation matrix.¹⁴ All predesignated variables of interest were included in the model to assess the relative strength of each variable’s association with telehealth appropriateness. Each variable of interest was also modeled individually in a univariate fashion. Predetermined reference variables were chosen for categorical variables that generally reflected the largest subgroup of each variable (eg, female sex, White race, distance traveled to clinic ≤ 30 miles, previous encounter in-person). Relative risks (RR) with associated 95% CIs were used to assess the significance of all associations, and all statistical analyses were conducted using R version 4.2.2 (R Foundation for Statistical Computing). This study was approved by the Duke Health Institutional Review Board (no. Pro00105997). The datasets generated during the study are available from the corresponding author upon reasonable request.