Study design
Data were drawn and anonymised from the Adelphi Real-World RA Disease Specific Programme (DSP), a cross-sectional, geographically diverse, real-world survey of rheumatologists and their patients with RA [21].
Participants of the DSP comprised US rheumatologists meeting the DSP inclusion criteria (described below) who completed a survey about their attitudes and stated behaviors regarding the treatment of their RA patients. Following the survey, rheumatologists then completed patient record forms for at least the eight consecutive RA patients who attended an appointment at their clinic. The patient forms collected information on patient demographics, symptomology (including marginal bone erosion and synovium inflammation), disease severity, treatment history and concomitant conditions. While symptomology, including osteoporosis and non-RA-related bone/joint inflammation was recorded, the rheumatologists were not asked to indicate how the symptoms were assessed. In addition, rheumatologists were not required to calculate or consult any composite scores (e.g. DAS28(3)-ESR, RAPID-3, etc.) specifically for the DSP research. Test results and disease activity scores may have been separately obtained as part of the routine clinical work-up. Rheumatologists also indicated their own satisfaction with the patient’s RA control. Data collection was conducted between January and March 2014. All participating rheumatologists were compensated according to fair market research rates, reflecting the time needed to complete all the forms.
Inclusion criteria
Rheumatologists were considered eligible to participate in the DSP if they met the following self-reported criteria: consultations with and medical management of ≥ 8 patients with RA per month, and graduation from medical school between 1975 and 2010.
Patients were considered eligible for inclusion of their data in the DSP if the following criteria were met: the patient aged ≥ 18 years, had rheumatologist confirmed and documented diagnosis of RA, and was not currently enrolled in a clinical trial. Finally, only patients for whom a DAS28(3)-ESR and CDAI score could be calculated, for purposes of the primary and sensitivity analyses, respectively, were included in the analysis population.
Patient demographics and baseline clinical characteristics
Data on patient demographics and baseline clinical characteristics were obtained. These included the following variables: Age, sex, bodymass index (BMI), race, time since diagnosis of RA, worst ever pain experienced, current level of pain, mean, current DAS28(3) score, presence of marginal bone erosion, synovium inflammation, osteoporosis present, RA-related bone/joint inflammation or damage present, whether on biologic biologic disease modifying antirheumatic drug (bDMARD) treatment, managed by physician based in hospital or mixed (hospital + office) practice, rheumatologist has an agreed T2T measure for patient, and the length of time managed by current rheumatologist.
Outcome measures
To evaluate rheumatologist-reported use of standardized disease activity measures, each rheumatologist was asked how he or she determined RA remission and which standard measure (if any) was typically used in assessing RA disease activity. In each patient record, physicians also stated whether the DAS28, ACR/EULAR, RAPID-3, and/or Health Assessment Questionnaire-Disability Index (HAQ-DI) assessment(s) were completed for the patient during the reference consultation.
RA disease remission for each patient was assessed via a direct question “Is this patient currently in remission? Yes or no?” (i.e. rheumatologist-reported assessment), via a calculation of RA disease activity using standardized measures, and based on information provided by rheumatologists on the record forms.
DAS28(3)-ESR (primary analysis)
Primary analysis was conducted using DAS28(3)-ESR as this maximized the number of patients and is one of a selection of standardized measures advocated by the ACR [7, 8].
The most recent Tender Joint Count (TJC), Swollen Joint Count (SJC), and ESR values were used to calculate the DAS28(3)-ESR based on the published scoring equations [22]. Two outcome categories were defined: remission (DAS28(3)-ESR < 2.6) and no remission (DAS28(3)-ESR ≥ 2.6) [23].
CDAI (sensitivity analysis)
CDAI was included in the sensitivity analysis as an alternative disease activity measure that does not require measurement of an acute phase reactant [24], therefore affording it a greater feasibility for implementing it in clinical practice. CDAI was calculated for patients for whom data on TJC, SJC, Evaluator’s (rheumatologist) Global Assessment (EGA) of disease activity based on a visual analog scale (0–10 cm), and Patient Global Assessment (PGA) measures had been provided on the patient forms, and scored via published equations [25]. Two outcome categories were similarly defined on this measure: remission, (CDAI ≤ 2.8) and no remission (CDAI > 2.8) [26].
Remission discordance/concordance
The outcomes of the clinical assessment as based on physician judgment versus the standardized measures were then compared to create four groups for analysis purposes:
-
(1)
Concordant/in remission: patient in remission as per physician judgment, confirmed by standardized measure.
-
(2)
Concordant/not in remission: patient not in remission as per physician judgment, confirmed by standardized measure.
-
(3)
Rheumatologist-negative discordance: patient in remission as per physician judgment, but has active disease per standardized measure (i.e. rheumatologist underestimated disease activity versus standardized measure).
-
(4)
Rheumatologist-positive discordance: patient not in remission as per physician judgment, but standardized measure indicates no disease activity i.e. in remission (i.e. rheumatologist overestimated disease activity versus standardized measure).
Statistical analysis
All analyses were conducted with Stata® 14.0 or later (StataCorp LLC, College Station, TX, US). Descriptive analyses provided the frequency (n) and percentage (%) of rheumatologist self-reported use of standardized measures and of patients assigned to each of the four cohorts, including respective rheumatologist and patient characteristics. Bivariate and multivariate analyses were conducted to identify factors associated with Rheumatologist-negative discordance versus the Concordant/in remission cohort. Odds ratios (OR), 95% confidence interval (CI), and P values indicated the robustness and significance of the results. The differences between cohorts were examined across various patient characteristics, including patient demographics, disease status, RA symptoms, treatment, doctor-patient relationship, patient-reported data including health-related quality of life, and rheumatologist self-reported characteristics such as workload and practice setting. Fisher Exact test, Mann-Whitney test, or Pearson’s chi-squared test assessed significant differences between patient subgroups on binary, non-parametric, and categorical outcomes, respectively. Additionally, Kernel density estimations using the Gaussian Kernel function were calculated and plotted for DAS28(3)-ESR for the four concordance groups.
Multivariate analyses then identified patient and physician characteristics independently associated with Rheumatologist-negative discordance of remission. Variables hypothesized to be associated with negative discordance were selected for inclusion in a logistic regression model. Variable selection was guided by disease knowledge; variables included physician practice type, whether a T2T management approach was in place for the patient, time since patient diagnosis, change in patient pain (from worse ever to current painFootnote 1), and presence of RA-related bone or joint inflammation or damage.
Standard errors were adjusted in the regressions to model the intragroup correlation (or clustering) of patients within rheumatologist practice using the Huber and White sandwich estimator of variances [27]. A 95% significance level was used throughout.