Discordance of Global Estimates by Patients and Their Physicians in Usual Care of Many Rheumatic Diseases: Association With 5 Scores on a Multidimensional Health Assessment Questionnaire (MDHAQ) That Are Not Found on the Health Assessment Questionnaire (HAQ)

Authors


Abstract

Objective

To analyze discordance between global estimates by patients (PATGL) and their physicians (DOCGL) according to demographic and self-report variables on a Multidimensional Health Assessment Questionnaire (MDHAQ) in patients with many rheumatic diseases seen in usual care.

Methods

Each patient completed an MDHAQ at each visit, which includes scores for physical function, pain, and PATGL, each found on the traditional Health Assessment Questionnaire (HAQ), and scores for sleep quality, anxiety, depression, self-report joint count, and fatigue, which are not found on the HAQ. A random visit of 980 patients with any rheumatic diagnosis was analyzed in 3 categories: PATGL=DOCGL (within 2 of 10 units), PATGL>DOCGL (by ≥2 of 10 units), and DOCGL>PATGL (by ≥2 of 10 units), using descriptive statistics and multinomial logistic regression models.

Results

Patients included 145 with rheumatoid arthritis, 57 with systemic lupus erythematosus, 173 with osteoarthritis, 348 with other inflammatory diseases, and 257 with other noninflammatory diseases. Overall, PATGL=DOCGL in 509 (52%), PATGL>DOCGL in 371 (38%), and DOCGL>PATGL in 100 (10%). PATGL>DOCGL was associated significantly with older age, female sex, low formal education, Hispanic ethnicity, not working, high MDHAQ physical function and pain scores, and high scores for fatigue, poor sleep, anxiety, depression, and self-report joint count, which are not available on the HAQ. Pain and fatigue were significant in a final multinomial logistic regression; the other variables may raise awareness of discordance to clinicians.

Conclusion

Global estimates of patients indicated significantly poorer status than estimates of their physicians in 38% of 980 patients with rheumatic conditions, and were associated with demographic and MDHAQ scores, 5 of which are not available on the HAQ.

INTRODUCTION

A “gold standard” quantitative biomarker measure such as blood pressure or serum glucose that can be applied to all individual patients is not available in rheumatic diseases, and quantitative assessment is approached with pooled indices of multiple measures ([1]). Many indices include quantitative estimates by patients and/or physicians, in addition to laboratory tests and ancillary imaging data ([2-9]). Ideally, estimates of clinical severity by physicians and patients should be concordant. However, many reports indicate discordance in estimates of severity by physicians and many patients with rheumatoid arthritis (RA) ([10-20]), systemic lupus erythematosus (SLE) ([21-25]), juvenile idiopathic arthritis ([26-28]), ankylosing spondylitis ([11, 29]), systemic sclerosis ([30]), and fibromyalgia ([11]).

The patient history and physical examination are more prominent in clinical decisions in rheumatic diseases than in many chronic diseases ([31]). A patient global estimate (PATGL) may be viewed as a summary of the patient history, while a physician global estimate (DOCGL) may be viewed as a summary of the physical examination, also incorporating other information from a patient history, laboratory tests, and imaging studies. In reports that compared PATGL and DOCGL on 0–10 visual analog scales (VAS), concordance of the 2 estimates was seen in approximately half of patients, while PATGL was substantially higher than DOCGL in approximately one-third of patients and DOCGL was higher than PATGL in other patients ([16, 18, 24]).

In this report, we analyzed concordance and discordance of PATGL and DOCGL in patients with various rheumatic diseases, beyond RA or any specific disease. We identified demographic data and scores on a Multidimensional Health Assessment Questionnaire (MDHAQ) that are associated with discordance, to facilitate possible recognition of likely sources of discordance by rheumatologists in their clinical care.

Box 1. Significance & Innovations

  • Approximately 38% of patients had patient global estimates ≥2 units higher (on a 0–10 scale) than their physicians' global estimates, with similar patterns in rheumatoid arthritis, osteoarthritis, systemic lupus erythematosus, other inflammatory rheumatic diseases, and other noninflammatory rheumatic diseases.
  • Patients with higher global estimates than their doctors were more likely to be older, female, of Hispanic ethnicity, and not working full time, and have lower formal education and poor status for 9 variables on a Multidimensional Health Assessment Questionnaire (MDHAQ), including 4 found on the HAQ, i.e., physical function, pain, patient global estimate (by definition), and Routine Assessment of Patient Index Data 3 (comprised of scores for function, pain, and patient global), and 5 variables not found on the HAQ: fatigue, sleep quality, anxiety, depression, and self-report painful joint count.
  • In a final multinomial logistic regression model, the only independent explanatory variables for associations with discordance are pain and fatigue, although the other variables that are significant in bivariate analyses may be helpful to clinicians to recognize a likelihood of discordance between their global estimates and those of individual patients.
  • The MDHAQ provides a simple questionnaire, on a single sheet of paper, that can be completed easily by most patients in 5–10 minutes in the waiting area and reviewed in ∼10 seconds by a doctor, to facilitate awareness of many variables that may be associated with discordance in global estimates of physicians and patients.

MATERIALS AND METHODS

Data source and study patients

A database of patients seen by 2 rheumatologists (YY, JS) between July 13, 2005 and April 19, 2011 at the Seligman Center for Advanced Therapeutics of the New York University Hospital for Joint Diseases was analyzed. Approval by the Institutional Review Board of New York University School of Medicine was obtained for creation of an Access (Microsoft) database from retrospective review of medical records and data collected at each visit. The database included MDHAQ scores (with PATGL; see below), as well as demographic information, DOCGL, and a primary rheumatic diagnosis assigned by the rheumatologist. Data were extracted from the Access database for this study to include a random visit of each of the 980 patients for whom a DOCGL and a PATGL were recorded. The database was transferred to STATA 12.0 for Windows (StataCorp) for all analyses.

MDHAQ patient self-report questionnaire measures

Each patient in this setting (with any diagnosis) completed a 2-page (single sheet of paper) MDHAQ at each visit while waiting to see the physician, in the infrastructure of clinical care ([32, 33]). The MDHAQ includes 0–10 scales for the 3 patient-reported RA core data set ([2]) measures: physical function, pain, and PATGL. A higher score indicates a poorer status for all MDHAQ scales. Physical function is quantitated according to 10 activities, including 8 that are found on the HAQ ([34]), and 2 complex activities: “walk 2 miles or 3 kilometers” and “participate in recreation and sports as you would like” ([33]). Scoring is as on the HAQ (range 0–3: 0 = with no difficulty, 1 = with some difficulty, 2 = with much difficulty, and 3 = unable to do). The total 0–30 is divided by 3 for a 0–10 score, using a template on the MDHAQ. Pain and PATGL each are scored 0–10 on a 21-circle VAS. The three 0–10 scores for function, pain, and PATGL are totaled for a Routine Assessment of Patient Index Data 3 (RAPID3) composite index (total score 0–30) ([35]).

The MDHAQ ([33]) also includes 5 scales not found on the HAQ: a 21-circle VAS (0–10 scale) for fatigue; 3 psychological items concerning sleep quality, anxiety, and depression, queried in the patient-friendly HAQ format, but not scored formally in usual care; and a self-report joint count based on the Rheumatoid Arthritis Disease Activity Index (RADAI) ([36]). The RADAI assesses 8 specific joint groups (fingers, wrists, elbows, shoulders, hips, knees, ankles, and toes) bilaterally, with scores for pain in each joint group ranging from 0–3 (0 = “no pain” and 3 = “severe pain”), for a total of 0–48. The MDHAQ version is designed for patients with all rheumatic diseases and adds scores for “neck” and “back.” The RADAI is reviewed but not scored formally in usual care, although it may be scored for research studies ([36]). The MDHAQ also includes demographic data: sex, race/ethnicity, years of formal education, employment status, and marital status ([33]).

Physician measures

Each of the 2 rheumatologists (YY, JS) scored a 0–10, 21-circle VAS for DOCGL. The physicians were not blinded to the PATGL or other information included on the MDHAQ, and reviewed the MDHAQ with the patient after the patient entered the examination room. Therefore, they were aware of the MDHAQ/RAPID3 (including PATGL) while taking a medical history, performing a physical examination, discussing treatment options, and writing prescriptions. DOCGL was assigned at the conclusion of the encounter. The physicians did not specifically review the PATGL before assigning DOCGL, although they had seen the PATGL earlier in the encounter.

Comparisons of PATGL and DOCGL

A random visit of each patient age >25 years seen between 2005 and 2011 by the 2 rheumatologists that included both PATGL and DOCGL was selected for the analyses presented in this report. A random visit was chosen, rather than the first visit, to include as many patients as possible, since some patients had their first visit prior to the practice of recording DOCGL at each visit, or DOCGL was missing at the first visit. Similar findings were seen in analyses of the first visit, but with fewer numbers (data not shown).

Patients were classified into 5 diagnosis categories: RA, SLE, osteoarthritis (OA), other inflammatory diseases, and other noninflammatory diseases. Patients with inflammatory diseases included those with connective tissue diseases, Sjögren's syndrome, vasculitis, gout, psoriatic arthritis, spondyloarthritides, “unspecified arthritis,” sarcoidosis, familial Mediterranean fever, Still's disease, reactive arthritis, and antiphospholipid syndrome. Patients with noninflammatory diseases included those with back pain, osteoporosis, fibromyalgia, “musculoskeletal disorders,” and “unspecified arthralgia.” In order for the report to describe a series of patients seen in a typical rheumatology setting, 1,772 visits of 646 patients with Behçet's syndrome (a particular interest of YY &lsqbr;[37, 38]&rsqbr;) were excluded from this study.

Discordance was defined as a difference of ≥2 units between 0–10, 21-circle VAS for DOCGL and PATGL, based on previous reports ([19]). Analyses also were performed using cut points of 1.5 or 2.5 units to identify discordance, with generally similar results (which therefore are not presented in this report). All patients were classified into 3 categories: patients with similar PATGL and DOCGL (PATGL=DOCGL) within 2 of 10 units, those with PATGL ≥2 units higher than DOCGL (PATGL>DOCGL), and those with DOCGL ≥2 units higher than PATGL (DOCGL>PATGL).

Statistical analysis

Demographic and MDHAQ variables were compared for the 5 diagnosis categories in 3 groups, i.e., PATGL>DOCGL, PATGL=DOCGL, and DOCGL>PATGL, using analysis of variance or the Kruskal-Wallis test if needed because of non-normal distributions. Spearman's rank order correlation coefficients were computed to estimate possible relationships between DOCGL and PATGL. Correlations were interpreted according to guidelines ([39]), with 0–0.09 indicating no correlation, 0.10–0.29 indicating low correlation, 0.30–0.49 indicating moderate correlation, and ≥0.50 indicating high correlation.

The Bland-Altman method was used to assess agreement between DOCGL and PATGL ([40]). The Bland-Altman method calculates the mean difference between 2 methods of measurement (the “bias”), and 95% limits of agreement as the mean difference ±2 SDs. It is expected that the 95% limits of agreement include 95% of differences between the 2 measurement methods. A Bland-Altman plot is included as a visual presentation of the 95% limits of agreement. The smaller the range, the better the agreement between these 2 methods of measurement.

Multinomial logistic regression model analyses were performed to identify independent explanatory variables for positive (PATGL>DOCGL) and negative (DOCGL>PATGL) discordance. Initially, bivariate analyses were performed. Demographic variables were dichotomized into 2 groups for age (<45 versus ≥45 years), sex (male versus female), formal education (≤14 versus >14 years), race/ethnicity (white versus each other group), marital status (married versus not), and work status (full time versus not). The RADAI self-report joint count also was dichotomized (≤4 versus >4). Physical function, pain, and fatigue were categorized using tertiles, with the lower tertile as the reference group in the models. The final multinomial logistic regression model was constructed, starting with those variables that were significant in bivariate analyses at the 0.05 level, and then using backward elimination to select the final model.

RESULTS

Demographic characteristics of patients

The study included 980 patients: 145 with RA, 57 with SLE, 173 with OA, 348 with other inflammatory rheumatic diseases, and 257 with other noninflammatory rheumatic diseases. The mean age was 50.2 years; patients with OA were the oldest, while SLE patients were the youngest (Table 1). Sixty-eight percent of the patients were women; female predominance was seen in all diagnosis groups, ranging from 54% in patients with other inflammatory rheumatic diagnoses to 93% in SLE patients. Sixty-eight percent of patients were white, 9% were African American, 13% were Hispanic, 6% were Asian, and 3% were another race/ethnicity. Full-time work was being performed by 46% of all patients, with the fewest in OA patients (32%; consistent with older age). The median formal education level was 16 years for all patients, an unusually high level among patients seen in rheumatology settings ([41]) (Table 1).

Table 1. Demographic and clinical characteristics of 980 patients according to diagnosis*
 All patients (n = 980)Diagnostic groupsPa
RA (n = 145)SLE (n = 57)OA (n = 173)ID (n = 348)NID (n = 257)
  1. Values are the mean ± SD for variables with a normal distribution and the median (interquartile range [IQR]) for variables with a skewed distribution. Qualitative variables are shown as the percentage. RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; OA = osteoarthritis; ID = other inflammatory rheumatic disease; NID = other noninflammatory rheumatic disease; MDHAQ = Multidimensional Health Assessment Questionnaire; RAPID3 = Routine Assessment of Patient Index Data 3; RADAI = Rheumatoid Arthritis Disease Activity Index ([36]) self-report joint count.
  2. aP value is from analysis of variance for quantitative variables with a normal distribution, the Kruskal-Wallis test for skewed variables, and the chi-square test for qualitative variables; not adjusted for multiple comparisons.
Demographic variables       
Age, mean ± SD years50.2 ± 16.549.3 ± 15.837.8 ± 12.662.6 ± 12.446.9 ± 15.849.6 ± 16.5< 0.001
Female sex, %687493785469< 0.001
Race/ethnicity, %      < 0.001
White685550677275 
African American91520107.06.9 
Hispanic13181317127.8 
Asian610152.34.95.9 
Other32.32.53.13.93.9 
Education, median (IQR) years16 (14–18)16 (13–18)16 (15–18)16 (13–18)16 (14–18)16 (14–18)0.56
Working full time, %464753325854< 0.001
Clinical variables (possible score range), median (IQR)       
MDHAQ function (0–10)1.1 (0.3–3)1.7 (0.3–3.7)0.3 (0–1.7)1.7 (0.7–3.3)1.1 (0–3)1 (0–2.3)< 0.001
MDHAQ pain (0–10)4 (2–7)4.7 (2–7)2 (0.5–5)5 (3–7.5)3.5 (1–7)4.5 (2–7)< 0.001
RAPID3 (0–30)10 (4.5–15.5)11 (4–16.7)5.3 (2.5–10.8)11.7 (6.7–16.7)9.3 (3.2–15.3)10 (5–14)< 0.001
Get a good sleep (0–3)1 (0–1)1 (0–1)1 (0–1)1 (0–1)1 (0–1)1 (0–1)0.89
Anxiety (0–3)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0.88
Depression (0–3)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0 (0–1)0.63
RADAI for pain (0–48)4 (1–10)5 (2–17.5)2 (0–4.5)6 (4–12)4 (0–9)4 (1–8)< 0.001
MDHAQ fatigue (0–10)3.7 (0–7)5 (0.5–8)5 (1.7–8)3.2 (1–7)3 (0–7)3.7 (0–7)0.08

Global estimates of patients and physicians

Among all patients, the median PATGL was 4 versus a median DOCGL of 2.5, 1.5 units lower than their patients (Figure 1 and Table 2). In the 5 diagnosis categories, median discordance ranged from 1.1 units in SLE to 1.8 units in RA on a 0–10 scale (Table 2). The median PATGL ranged from 2.5 in SLE to 5.0 in OA and RA, while the median DOCGL ranged from 2.0 in SLE to 2.5 in other diagnoses (Figure 1), a narrower range. PATGL was correlated significantly with DOCGL, albeit relatively modestly (ρ = 0.42, P < 0.001; ρ = 0.50 and ρ = 0.42 for rheumatologists 1 and 2, respectively) (Table 2). Again, results were relatively similar in the 5 diagnosis groups, as correlations of PATGL with DOCGL ranged from ρ = 0.29 in OA to ρ = 0.57 in SLE (Table 2).

Figure 1.

Box plot of patient and physician global estimates of status by diagnostic category: rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), osteoarthritis (OA), other inflammatory rheumatic diseases (ID), and other noninflammatory rheumatic diseases (NID).

Table 2. Physician global estimate (DOCGL), patient global estimate (PATGL), differences between DOCGL versus PATGL, and levels of concordance and discordance in 980 patients according to diagnosis*
 All patients (n = 980)Diagnostic groups
RA (n = 145)SLE (n = 57)OA (n = 173)ID (n = 348)NID (n = 257)Pa
  1. Values are the median (interquartile range [IQR]) for quantitative variables and the number (percentage) for qualitative variables unless indicated otherwise. RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; OA = osteoarthritis; ID = other inflammatory rheumatic disease; NID = other noninflammatory rheumatic disease.
  2. aP value from the Kruskal-Wallis test for quantitative variables and chi-square test for qualitative variables; not adjusted for multiple comparisons.
DOCGL (0–10), median (IQR)2.5 (1.5–3.5)2.5 (1.5–3.5)2 (1.5–3.5)2.5 (2–3.5)2.5 (1.5–3.5)2.5 (1.5–3.5)0.036
PATGL (0–10), median (IQR)4 (1.5–6.5)5 (1.5–7)2.5 (1–5)5 (2–6.5)3.5 (1.5–6.5)4 (1.5–6.5)0.044
Difference between PATGL and DOCGL, mean ± SD1.5 ± 2.71.8 ± 2.91.1 ± 2.51.7 ± 2.71.3 ± 2.71.5 ± 2.60.219
Spearman's correlation of PATGL with DOCGL0.420.450.570.290.470.35< 0.001 for all
PATGL and DOCGL discordance groups, no. (%)      0.295
PATGL>DOCGL371 (38)63 (43)16 (28)75 (43)119 (34)98 (38) 
PATGL=DOCGL509 (52)69 (48)36 (63)80 (46)193 (55)131 (51) 
DOCGL>PATGL100 (10)13 (9)5 (9)18 (10)36 (10)28 (11) 

Among all 980 patients, PATGL=DOCGL in 509 (52%), PATGL>DOCGL in 371 (38%), and DOCGL>PATGL in 100 (10%) (Table 2). As seen for median values, the proportions of patients in the 3 categories varied over a relatively narrow range in the 5 diagnostic groups: PATGL=DOCGL from 46% (OA) to 63% (SLE), PATGL>DOCGL from 28% (SLE) to 43% (RA and OA), and DOCGL>PATGL from 9% (RA and SLE) to 11% (other noninflammatory rheumatic diseases).

A Bland-Altman analysis indicated that the limits of agreement ranged from −6.82 to 3.85, with a mean difference of −1.48 (Figure 2). The average discordance between DOCGL and PATGL may be considered large enough to be clinically important, and the 2 measures cannot be considered equivalent. The scatter is highest for average values around 6, and appears lower for high and low values of the average (Figure 2).

Figure 2.

Bland-Altman plot comparing the average of global estimates of status by patient (PATGL) and physician (DOCGL) against the difference between global estimates (PATGL − DOCGL). The bubble size is determined by the number of subjects with that combination of average and difference in PATGL and DOCGL.

Variables associated with discordant PATGL>DOCGL or DOCGL>PATGL

In bivariate analyses, patients in the PATGL>DOCGL group were more likely to be older and female, have a lower education level, and be Hispanic or African American rather than white or Asian, and less likely to be married and working full-time, compared to patients for whom PATGL=DOCGL (Table 3). Patients in the PATGL>DOCGL group had significantly higher scores on the 3 core data set measures found on the HAQ and MDHAQ, physical function, pain, and PATGL (as expected), as well as RAPID3, the composite index of these 3 scores. Patients in the PATGL>DOCGL group also had significantly higher scores on 5 MDHAQ measures not included on the HAQ: poor sleep, anxiety, depression, RADAI self-report joint count, and fatigue (Table 3). In the final multinomial logistic regression model, selecting variables through backward elimination, pain and fatigue were the only significant explanatory variables for PATGL>DOCGL (Table 4).

Table 3. Demographic and clinical characteristics of patients according to level of concordance or discordance between patient (PATGL) and physician global estimates (DOCGL)*
 All patients (n = 980)Groups by PATGL and DOCGL discordance/concordancePa
PATGL>DOCGL (n = 371)PATGL=DOCGL (n = 509)DOCGL>PATGL (n = 100)
  1. Values are the mean ± SD for normal distributions, the median (interquartile range [IQR]) for skewed distributions, and the percentage for qualitative variables. MDHAQ = Multidimensional Health Assessment Questionnaire; RAPID3 = Routine Assessment of Patient Index Data 3; RADAI = Rheumatoid Arthritis Disease Activity Index (36) patient self-report joint count.
  2. aP value from analysis of variance for quantitative variables with a normal distribution, the Kruskal-Wallis test for skewed variables, and chi-square test for qualitative variables; not adjusted for multiple comparisons.
Demographic measures     
Age ≥45 years, %59.964.955.4640.012
Female sex, %687464640.001
Education, median (IQR) years16 (14–18)16 (12–18)16 (14–18)17 (16–19)< 0.001
Race/ethnicity, %    0.129
White68637078 
African American91195 
Hispanic1317128 
Asian6568 
Other3431 
Married, %47.844.549.153.20.246
Working full time, %45.938.850.549.00.001
Physician measure (possible score range)     
DOCGL (0–10), mean ± SD2.6 ± 1.62.5 ± 1.02.5 ± 1.74 ± 1.9< 0.0001
MDHAQ measures (possible score range), median (IQR)     
MDHAQ function (0–10)1.1 (0.3–3)2.3 (1–4)0.7 (0–2)0.3 (0–1.3)< 0.0001
MDHAQ pain (0–10)4 (2–7)7 (5–8.5)3 (1–5)2.5 (0.5–5)< 0.0001
PATGL (0–10)4 (1.5–6.5)7 (5–8)2.5 (1–4)0.5 (0–2)< 0.0001
RAPID3 (0–30)10 (4.5–15.5)15.7 (12–19.4)6.5 (3–10.8)4 (1–7.8)< 0.0001
Get a good sleep (0–3)1 (0–1)1 (1–2)1 (0–1)0 (0–1)< 0.0001
Anxiety (0–3)0 (0–1)1 (0–1)0 (0–1)0 (0–1)< 0.0001
Depression (0–3)0 (0–1)1 (0–1)0 (0–1)0 (0–1)< 0.0001
RADAI for pain (0–48)4 (1–10)8 (4–16)3 (0–7)2 (0–7)< 0.0001
MDHAQ fatigue (0–10)3.5 (0.5–7)7 (4–8.5)2.5 (0–5)0 (0–3)< 0.0001
Table 4. Unadjusted and adjusted ORs from multinomial logistic regression models of discordance between patient (PATGL) and physician global estimates (DOCGL) of status*
PredictorPATGL>DOCGL (by ≥2 units)DOCGL>PATGL (by ≥2 units)
Unadjusted OR (95% CI)Adjusted OR (95% CI)Unadjusted OR (95% CI)Adjusted OR (95% CI)
  1. OR = odds ratio; 95% CI = 95% confidence interval; MDHAQ = Multidimensional Health Assessment Questionnaire; RADAI = Rheumatoid Arthritis Disease Activity Index ([34]) patient self-report joint count.
  2. aStatistically significant results at the 0.05 level. The adjusted results are from the final multinomial logistic regression model that selected only pain and fatigue through backward elimination.
Demographic variables    
Age, ≥45 vs. <45 years1.49 (1.13–1.96)a1.43 (0.92–2.23)
Sex, male vs. female0.61 (0.45–0.83)a1.01 (0.64–1.59)
Education, >14 vs. ≤14 years0.60 (0.44–0.81)a1.77 (0.99–3.17)
Race/ethnicity  
Asian vs. white0.89 (0.46–1.72) 1.11 (0.44–2.82) 
African American vs. white1.28 (0.77–2.12) 0.51 (0.18–1.50) 
Hispanic vs. white1.61 (1.04–2.49)a 0.61 (0.25–1.50) 
Other vs. white1.61 (0.73–3.54) 0.39 (0.05–3.00) 
Married, yes vs. no0.84 (0.63–1.10)1.18 (0.76–1.84)
Working full time, yes vs. no0.61 (0.47–0.80)a1.01 (0.65–1.56)
MDHAQ variables (possible score range)    
MDHAQ function (0–10)  
1 vs. 03.42 (2.32–5.02)a 0.65 (0.40–1.09) 
2 vs. 06.6 (4.53–9.75)a 0.59 (0.33–1.06) 
MDHAQ pain (0–10)    
1 vs. 04.03 (2.65–6.15)a3.09 (1.98–4.81)a0.80 (0.49–1.30)1.05 (0.63–1.74)
2 vs. 015.57 (10.10–23.99)a8.15 (5.04–13.16)a0.86 (0.47–1.59)1.38 (0.68–2.79)
Get a good sleep (0–3)  
1 vs. 01.53 (1.10–2.13)a 0.55 (0.35–0.89)a 
2 vs. 03.63 (2.48–5.32)a 0.46 (0.22–0.94)a 
Anxiety (0–3)  
1 vs. 01.85 (1.37–2.51)a 0.63 (0.37–1.05) 
2 vs. 02.69 (1.73–4.17)a 0.58 (0.24–1.42) 
Depression (0–3)  
1 vs. 02.05 (1.49–2.81)a 0.83 (0.48–1.41) 
2 vs. 03.46 (2.16–5.54)a 0.72 (0.27–1.90) 
RADAI for pain (0–48), >4 vs. ≤43.59 (2.48–5.21)a0.72 (0.39–1.34)
MDHAQ fatigue (0–10)    
1 vs. 02.54 (1.70–3.80)a1.81 (1.18–2.79)a0.38 (0.27–0.65)a0.38 (0.22–0.65)a
2 vs. 09.35 (6.30–13.87)a3.94 (2.52–6.18)a0.38 (0.19–0.73)a0.32 (0.15–0.67)a

Patients in the DOCGL>PATGL group had higher levels of formal education and lower fatigue scores. In bivariate analyses, these patients were more likely to have low scores for sleep quality and fatigue, but no significant differences in the other MDHAQ scores, compared to the PATGL=DOCGL group (Table 4). In the final multinomial logistic regression model, fatigue was the only significant explanatory variable.

DISCUSSION

This study is consistent with previous reports ([10-30]) that 52% of global estimates of patients and physicians were concordant, while 38% were characterized by higher PATGL compared to physicians (PATGL>DOCGL) and 10% by higher DOCGL compared to patients (DOCGL>PATGL). Higher PATGL>DOCGL was associated with older age, female sex, lower formal education level, and higher MDHAQ scores for physical function and pain (as well as PATGL, by definition), as reported previously ([16, 18, 19, 24, 42]). This report provides new information that PATGL>DOCGL also is associated significantly with not working, Hispanic ethnicity, and high scores for 5 items on the MDHAQ not found on the HAQ: fatigue, poor sleep quality, anxiety, depression, and RADAI self-report joint count. By contrast, DOCGL>PATGL was associated with high education, good sleep quality, and low fatigue scores. Furthermore, similar levels of discordance were seen in patients with many rheumatic diseases, including RA, OA, SLE, and composite groups of patients with inflammatory diseases or noninflammatory diseases.

In a multinomial logistic regression model, the 2 significant independent explanatory variables for PATGL>DOCGL, identified through backward elimination, were pain and fatigue. These findings might be expected on the basis of high correlations of PATGL with pain and fatigue scores, which also are associated with demographic variables and other MDHAQ scores. The demographic variables and other MDHAQ scores are presented to provide clues for a clinician to recognize individual patients who are more likely to show discordance of DOCGL and PATGL.

A recent report suggests that intensification of therapy is based primarily on outcomes reported by the patient rather than the doctor ([43]). In analyses of clinical trials results, both DOCGL and PATGL usually are more efficient to distinguish active from control treatments compared to joint counts and laboratory tests ([44, 45]).

The MDHAQ is a simple questionnaire that can be completed by the patient in the waiting area in ∼5–10 minutes ([46, 47]), and can raise awareness in any health professional of variables that may be associated with discordance in PATGL versus DOCGL. The MDHAQ facilitates calculation of RAPID3 in ∼5 seconds, with scores that are correlated at high levels of significance with the Clinical Disease Activity Index (CDAI) and Disease Activity Score in 28 joints (DAS28) ([48]). The MDHAQ presents additional potential advantages in busy clinical settings to help prepare the patient for the encounter, improve doctor–patient communication by providing an agenda or roadmap, and save time for the doctor ([47]).

It would appear optimal if PATGL and DOCGL were concordant in formulating clinical decisions. Discordance has been reported to be associated with poorer outcomes in several diseases ([19, 49, 50]). A patient who overestimates severity may push a doctor to initiate treatments that may have a poor risk/benefit ratio. By contrast, patients who underestimate severity may have poor adherence to treatment and/or may lead their physicians to underrate the impact of disease. An initial step to improve concordance is to record quantitatively both the PATGL and DOCGL at each visit. Unfortunately, this practice has not yet been introduced into the majority of usual rheumatology care settings ([51]).

These observations also extend the concept that various rheumatic diseases, which differ substantially in pathophysiology, treatment, and many other features, are similar from the viewpoint of the patient experiencing the disease. The MDHAQ is effective to monitor patients with all rheumatic diseases ([52]). RAPID3 documents similar change in status in OA, SLE, spondyloarthritis, and gout as in RA ([53]). The RADAI self-report joint count on the MDHAQ is informative in all rheumatic diseases ([54]). Therefore, diseases such as RA, SLE, OA, spondyloarthritis, and gout appear far more similar than different from the patients' perspectives, although they appear very different from the doctors' perspectives.

This study includes several limitations. All patients were seen at one site in New York with an atypically high level of formal education ([41]), and information from other settings would certainly be of value. Only 2 rheumatologists, both of whom were men in their late 30s, were involved; they were the only 2 rheumatologists in this setting who completed a DOCGL on most patients. Evidence of some variation between DOCGL estimated by the 2 rheumatologists would suggest that further analyses of physician variables that may influence DOCGL versus PATGL would be of interest, such as age, clinical setting, sex, etc. However, such studies would require far more than 2 rheumatologists to ascertain meaningful trends. Although 980 patients were studied, many more patients would be needed to recognize possible subsets of patients and physicians who are likely to overestimate or underestimate severity compared to one another. Finally, the data are cross-sectional, and it would appear of value to analyze data longitudinally, to recognize whether discordance is stable or changes over time and whether it is associated with differences in patient outcomes, as documented in one study in SLE ([22]).

Nonetheless, the data presented would appear to provide a strong rationale for a rheumatologist to score a DOCGL and ask a patient to score a PATGL at each visit in usual care. A PATGL is required to compile a Disease Activity Score (DAS28) or CDAI. If the patient is given a piece of paper or computer screen to indicate a PATGL, it is rather simple to include an entire MDHAQ, which occupies 2 sides of a single piece of paper. The MDHAQ also can save time for the doctor, particularly when used with a simple form available to record a DOCGL ([55]). It is suggested that this practice be incorporated into all usual rheumatology care.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Pincus had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Castrejón, Pincus.

Acquisition of data. Yazici, Samuels.

Analysis and interpretation of data. Castrejón, Yazici, Samuels, Luta, Pincus.

Ancillary