In this retrospective cohort study of patients with newly diagnosed psoriatic arthritis, the authors examine the association of treatment selection and costs with physician specialty.
Objectives: To describe current psoriatic arthritis treatment and costs by provider specialty using real-world claims data.
Study Design: Observational, retrospective cohort study of patients in the IBM MarketScan Commercial and supplemental Medicare databases.
Methods: Eligible patients had newly diagnosed psoriatic arthritis with 12 months of continuous enrollment pre– and post index date for their initial claim. Patients were assigned to 1 of 5 provider specialty cohorts. During the 1-year follow-up period, we collected psoriatic arthritis treatment agent and regimen type and total annual medical and health care costs. We used multivariate regression models to determine the conditional associations of provider specialty with costs.
Results: A total of 2132 patients with incident psoriatic arthritis qualified. Most providers were rheumatologists (n = 1365; 64%). Rheumatologists commonly prescribed oral small molecules (methotrexate, 56.3% of prescriptions; sulfasalazine, 8.6%; apremilast, 7.0%) as the index therapy, whereas 23.8% of prescriptions were for tumor necrosis factor inhibitors (adalimumab, 14.2%; etanercept, 7.9%; and infliximab, 1.7%). Compared with other specialists, dermatologists prescribed biologics and other specialty drugs more frequently—adalimumab (32.7%), apremilast (14.3%), etanercept (11.6%), and ustekinumab (8.8%)—and methotrexate less frequently (30.6%). The greatest unadjusted median health care costs were observed among dermatologists ($45,548) compared with rheumatologists ($30,411), primary care physicians ($29,927), rheumatologists/dermatologists ($27,393), and other specialists ($27,774). However, after adjusting for patient-level factors, multivariate regression analyses found that provider specialty was not associated with higher health care costs.
Conclusions: In patients with newly diagnosed psoriatic arthritis, physician specialty was associated with different medication choices but not costs.
Am J Manag Care. 2021;27(7):e226-e233. https://doi.org/10.37765/ajmc.2021.88706
Psoriatic arthritis is a chronic inflammatory musculoskeletal disease prevalent in approximately 0.06% to 0.25% of the US population1 and in about 1 in 3 patients with psoriasis.2,3 The heterogeneity of disease in these patients often leads to the need for specialists to manage the disease manifestations, including polyarticular arthritis, psoriasis, enthesitis, dactylitis, and spondylitis. With estimated annual total medical plus pharmacy costs of $28,676 per person,4 psoriatic arthritis needs to be closely managed by the physician. Pharmacologic treatment options have expanded in recent years and include oral agents tofacitinib and apremilast, tumor necrosis factor inhibitors, and newer biologics.5 Despite the complexity of the disease and the medications used to treat it, a recent study found that 58% of a cohort of patients with newly diagnosed psoriatic arthritis treated in the United States were initially treated with methotrexate. Of these, only one-third (34%) of those who initiated methotrexate monotherapy persisted on their treatment at the end of the first year.4 Although methotrexate has dominated as a first treatment choice in psoriatic arthritis in the past, consistent with the 2015 Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA) guidelines,6 the 2018 American College of Rheumatology/National Psoriasis Foundation (ACR/NPF) guidelines for the treatment of psoriatic arthritis recommend the use of tumor necrosis factor inhibitors over other medications for patients who recently received a diagnosis.5
Although psoriatic arthritis is a predominantly a rheumatic disease, both joint and skin involvement are key to diagnosing it. Thus, a variety of provider specialists may be responsible for diagnosis and treatment decisions, including rheumatologists, dermatologists, primary care providers, and other specialists. Survey studies including DISCONNECT7 and the United States subset of the Multinational Assessment of Psoriasis and Psoriatic Arthritis (MAPP) survey3 have reported differences in attitudes and treatment choices for psoriatic arthritis across provider specialties in the United States. Among veterans with psoriatic arthritis, those treated by rheumatologists were more likely to receive a treatment (either oral small molecule or biologic) than those seen by a nonrheumatologist.8 Although these studies’ findings suggest that provider specialty may be associated with different management and treatment strategies of patients with psoriatic arthritis, there is little real-world evidence describing these differences and their result on health care expenditure. If differences exist in the cost of managing psoriatic arthritis by provider type, it would be important for payers to understand this and how it may affect outcomes. This observational, retrospective study of data from a commercial claims database was designed to describe current psoriatic arthritis treatment patterns and health care costs among patients with newly diagnosed psoriatic arthritis as a function of the specialty of the treating provider.
This was an observational, retrospective cohort study of patients in the United States using claims data spanning January 2013 to March 2018.
The IBM MarketScan Commercial and supplemental Medicare databases provide access to medical and prescription drug claims for privately insured individuals, including individuals 65 years and older with Medicare supplemental coverage. The databases provide detailed cost, use, and outcomes data for health care services delivered in both inpatient and outpatient settings. The medical claims are linked to outpatient prescription drug claims and person-level enrollment data through the use of unique enrollee identifiers. All database records are deidentified and fully compliant with US patient confidentiality requirements, including the Health Insurance Portability and Accountability Act of 1996. Because the study collected only deidentified patient records and did not involve the collection, use, or transmittal of individually identifiable data, institutional review board approval to conduct this study was not necessary.
Patient Eligibility Criteria
Eligible study participants were 18 years or older and had at least 1 inpatient or at least 2 outpatient nondiagnostic medical claims for psoriatic arthritis treatment within 90 days of one another with a diagnosis of psoriatic arthritis between January 1, 2014, and March 31, 2017. The date of the earliest observed claim for psoriatic arthritis was the index diagnosis date; no diagnoses for psoriatic arthritis were permitted in the preindex period. Patients were excluded from this study if they had a claim from baseline through follow-up with a diagnosis of rheumatoid arthritis, ankylosing spondylitis, Crohn disease, ulcerative colitis, juvenile idiopathic arthritis, tuberculosis, hepatitis B, hepatitis C, HIV/AIDS, or any malignancy. Participants had at least 12 months of preindex continuous enrollment and at least 12 months of postindex continuous enrollment. The index treatment was an approved claim between 30 days prior to and 90 days after the index diagnosis date; patients must not have had a pharmacologic treatment for psoriatic arthritis during the first 335 days of the preindex period.
Assignment to Physician Specialty Cohort
Patients meeting eligibility criteria were assigned to 1 of 5 specialty cohorts based on the provider specialty listed on psoriatic arthritis claims appearing in the time preceding the index prescription date. The preceding 14 days were examined for oral small molecule medications and the preceding 30 days for biologic/newer medications. The following hierarchy was used in cases in which multiple physician specialties were observed: rheumatology and dermatology (ie, those who were double certified), rheumatology, dermatology, and primary care. Provider specialists who did not fit in these categories were assigned to a fifth cohort labeled “other physician specialty.” This hierarchy was based on the specialist types most relevant for patients with psoriatic arthritis. Because psoriatic arthritis is a rheumatologic disease, rheumatologists with or without dermatology certification were considered the most appropriate providers for these patients. Because psoriatic arthritis often develops in those already being treated by dermatologists for psoriasis, dermatologists were deemed the second most appropriate provider type. Some patients may be seen and treated by their primary care providers without being referred to a specialist; therefore, these providers were deemed the third most appropriate. Finally, those falling outside these categories were grouped together as other.
For pharmacologic treatments, study outcomes included the index treatment drug and regimen type (monotherapy or combination). We analyzed utilization and costs for all medical and health care services during the 12 months post medication index. Annual total medical costs included inpatient and outpatient costs (emergency department, outpatient hospital, office visits, infusion, and other outpatient services, not including outpatient pharmacy costs). Annual total health care costs included medical costs and all outpatient pharmacy costs. Health care costs were calculated based on paid amounts of adjudicated claims, including insurer and health plan payments and patient cost sharing in the form of co-payment, deductible, and coinsurance. All costs were adjusted for inflation using the medical care component of the Consumer Price Index and were standardized to 2018 US$.
Demographic characteristics, including age, sex, insurance type, health plan type, and geographic region, were collected at the diagnosis index date. Clinical characteristics, including Deyo-Charlson Comorbidity Index9 score, the frequency of chronic comorbidities (psoriasis, diabetes [prediabetes, type 1, type 2], depression, anxiety, cardiovascular disease), the frequency of the most common medication classes (such as analgesics, anti-inflammatory agents, or psychotherapeutics), and health care utilization and costs, were collected during a 1-year pre- and postindex period. During the 1-year follow-up period, additional data were collected for psoriatic arthritis treatments and regimen type, as well as their associated costs.
Descriptive analyses summarized demographic characteristics, baseline characteristics, and outcomes by provider specialty and summarized medication prescribed using mean and SD, median, or frequency and percentage as appropriate. We assessed the association between provider specialties and outcomes using separate multivariable linear regression models with log-transformed costs adjusting for provider specialty, provider switching, baseline age, gender, Deyo-Charlson Comorbidity Index score, insurance type, geographic region, index year, diagnosing physician specialty, baseline health care expenditures, and index medication. Sensitivity analyses evaluated separate 2-level (patient and provider) multilevel linear regression models with log-transformed total health care or medical costs as a function of index provider specialty and confounding factors.
Role of the Funder
Amgen Inc, the funder of the study, designed this study and interpreted the data in collaboration with the non-Amgen authors. The authors at IBM analyzed the data in the IBM MarketScan Commercial and supplemental Medicare databases. The Amgen and non-Amgen authors provided joint approval to submit this manuscript for publication.
A total of 2132 patients qualified for the study and were included in the analysis (Table 1). Patients with psoriatic arthritis were most commonly prescribed their first treatment by a rheumatologist (n = 1365; 64%), followed by a primary care provider (n = 198; 9%) and a dermatologist (n = 147; 7%). Only 16 patients (< 1%) were treated by a double board-certified rheumatology/dermatology specialist; reporting on these data will be minimal. The remaining 406 (19%) patients were treated by other physician types and were assigned to the “other provider” category. This category consisted of 20 different specialties, with the most common specialty being internal medicine.
The demographics and baseline characteristics of patients were generally similar across provider specialties (Table 2). The mean age of patients at the index treatment date was 48.6 years, and just more than half (53.5%) were female. During the 12-month preindex period, the 3 most common comorbid clinical conditions were psoriasis (47.4%), cardiovascular disease (22.3%), and diabetes (18.0%). Overall, the most common medications during the 12-month preindex period were nonsteroidal anti-inflammatory agents (50.8%). The mean baseline total all-cause health care costs for patients treated for psoriatic arthritis were $11,485; these costs were similar across provider specialties.
Nearly all patients received monotherapy (98.5%) as a first treatment for psoriatic arthritis. Index treatment by provider specialty is shown in Figure 1. Overall, for all specialties combined, most patients received oral small molecules, of which the most widely prescribed was methotrexate (53.6% of prescriptions), followed by sulfasalazine (8.2%) and apremilast (7.5%). Approximately one-fourth of patients were prescribed tumor necrosis factor inhibitors (adalimumab [15.2%], etanercept [8.5%], and infliximab [1.6%]). When comparing the prescription patterns by provider specialty, rheumatologists most commonly prescribed oral small molecules (methotrexate, 56.3%; sulfasalazine, 8.6%; apremilast, 7.0%), whereas 23.8% of prescriptions were for tumor necrosis factor inhibitors (adalimumab, 14.2%; etanercept, 7.9%; and infliximab, 1.7%). Dermatologists prescribed biologics and other specialty drugs more frequently—adalimumab (32.7%), apremilast (14.3%), etanercept (11.6%), and ustekinumab (8.8%)—and methotrexate less frequently (30.6%) than the other specialists.
Index medication persistence rates were similar among patients treated by dermatologists (37.4%), rheumatologists (38.2%), and primary care providers (37.9%), but the rate of medication switching was lower among patients treated by dermatologists (28.6%) compared with those treated by rheumatologists and primary care providers (40.9% and 39.4%, respectively).
Provider specialties had a wide variation in unadjusted annual total medical and health care costs for patients treated for psoriatic arthritis, as indicated by the medians and interquartile ranges (Figure 2). The highest median health care cost was observed among dermatologists ($45,548) compared with rheumatologists ($30,411), primary care physicians ($29,927), and other specialists ($27,774). Mean (SD) health care costs were $46,616 ($30,438) for dermatologists, $34,505 ($32,027) for rheumatologists, $34,028 ($30,365) for primary care physicians, and $32,562 ($30,225) for other specialists. Total median medical costs for patients treated for psoriatic arthritis were more similar and had less variation within each provider specialty (Figure 2). Mean (SD) medical costs were $11,513 ($22,693) for rheumatologists, $9102 ($15,042) for dermatologists, $10,677 ($21,979) for primary care physicians, and $11,639 ($21,398) for other specialists.
Multivariate Regression Analyses
Results of multivariate analyses examining the associations between physician specialty and both medical and health care costs after adjusting for demographic and baseline characteristics and index medications are presented in Table 3. Provider specialty was not associated with differences in either medical or health care costs. For medical costs during the postindex period, several factors were statistically significant with increasing costs. These factors include older age, female sex, baseline health care costs, diabetes, depression, cardiovascular disease, geographical regions, and index provider switch. Provider specialty compared with a rheumatologist and the choice of medication (etanercept, adalimumab, and apremilast) compared with methotrexate were not associated with increased medical costs.
For total health care costs during the postindex period, the treatment choices of etanercept, adalimumab, and apremilast compared with methotrexate were all associated with higher costs (all P ≤ .001). Other significant covariates included rural designation (compared with urban), some comorbid conditions (depression, psoriasis, diabetes), and baseline health care costs. Importantly, provider specialty was not associated with higher health care costs at follow-up after adjusting for observable patient characteristics. However, switching index provider was associated with higher health care costs.
To evaluate the robustness of our multivariate results, we tested several assumptions about the data. Clusters of patients may have been treated by the same health care provider (eg, one rheumatologist treating a cluster of patients with psoriatic arthritis), and this could have affected the results. To evaluate this, we generated a multilevel model clustering on provider specialties and accounting for intraclass correlation among patients nested within providers (eAppendix Table 1A [eAppendix available at ajmc.com]). Most providers (85% of 524 providers) saw only 1 patient in the sample, and the intraclass correlation was near zero for all linear multilevel models, suggesting that among the approximately 15% of providers who saw multiple patients, patient outcomes were not highly correlated with each other. Additionally, analyses using a combination of zip code and provider specialty as a proxy for provider identifier (ID) (eAppendix Table 1B) produced nearly identical results to the models that ignore intraprovider correlation entirely (ie, main regression models). Further, it appears that studying only those patients with explicit IDs (ie, provider IDs) may induce biased results on some outcomes (eAppendix Table 2). For instance, in the analysis of total medical costs, the results from the entire study population produced different results from analogous models of the subset with valid provider IDs, suggesting that provider ID may not be missing at random. Specifically, some coefficients differ in magnitude and/or direction, and some significant trends in the subset models are no longer significant when evaluating the entire study population. In other models, however, including total health care costs, all variations of the multilevel models are relatively consistent as it concerns the primary exposure variable (provider specialty). Given these 2 characteristics of the sample, low intraclass correlation and the bias introduced by including only those patients with explicit IDs, we concluded that the results from the main model specifications are robust regardless of clustering or patient ID assumptions.
In this retrospective cohort study of patients with newly diagnosed psoriatic arthritis within a commercial claims database in the United States, providers prescribed methotrexate most often as the first medication for the treatment of psoriatic arthritis. Methotrexate represented just more than half of all prescriptions overall (53.6%), whereas tumor necrosis factor inhibitors represented approximately one-fourth (25.3%) of first treatments prescribed for psoriatic arthritis. Dermatologists most often prescribed adalimumab, followed by methotrexate, then etanercept. Compared with other specialists, dermatologists were more likely to prescribe adalimumab, etanercept, apremilast, and ustekinumab. Dermatologists, who represented just 7% of the provider specialists, made treatment decisions more in line with the 2018 ACR/NPF guidelines.5 This finding regarding patients who obtained their first treatment for psoriatic arthritis from a dermatologist may reflect that these patients had severe skin disease or other unobservable characteristics not captured in claims data.
The data in this analysis provide insights into clinical practice patterns across physician specialties, showing prescribing patterns consistent with the 2015 GRAPPA guidelines for most specialists, except for dermatologists. These 2015 guidelines recommend oral small molecules and tumor necrosis factor inhibitors as first-line treatments, with the suggestion that methotrexate could be tried first.6 In contrast, the more recent 2018 ACR/NPF guidelines5 for the treatment of psoriatic arthritis conditionally recommend the use of tumor necrosis factor inhibitors over oral small molecules as a first treatment for psoriatic arthritis. As these guidelines were published after the time frame for our study, physician prescribing behavior may be expected to change. Future research will be required to determine the adherence of providers to these new recommendations. Although the DISCONNECT7 and United States MAPP survey3 studies had suggested differences in treatment choices for psoriatic arthritis based on physician specialty, the current study has extended these findings by identifying provider specialty differences through real-world claims data and calculating their association with costs.
We observed substantial variation in pharmacologic treatment and total health care costs within and across physician specialties, with dermatologists prescribing a higher frequency of newer medications relative to methotrexate. Although the unadjusted cost outcomes show higher spending by patients treated by dermatologists, our multivariate analysis shows that this higher spending is explained by differences in patient characteristics, not by the type of provider. Although the current study found no evidence of a statistically significant relationship between provider specialty and costs, we believe that there may be unobserved variable bias in the specific patients treated by different specialties. That is, there may be unobserved patient or health system characteristics that cause some providers to see a specific kind of patient. This idea may be supported by our finding that provider switching, regardless of other baseline characteristics, was associated with higher costs and may reflect the switch to a specialist. Future real-world studies are needed to better understand how first therapy affects both patient outcomes and health care costs.
This study was limited to individuals with commercial health coverage or private Medicare supplemental coverage in the United States; results may not be generalizable to patients with psoriatic arthritis with other types of insurance or without health insurance coverage. Misclassification of psoriatic arthritis is possible because patients were identified through administrative claims data instead of medical records; these data are subject to data coding limitations and data entry error. However, we tried to minimize this risk by employing methodology recommended by Lee et al10 and used 2 codes for psoriatic arthritis and a therapy. This strategy leads to a higher positive predictive value for a true diagnosis of psoriatic arthritis. Although not flawless, our patient classification algorithm minimized the likelihood of misclassification of patients in our sample. The rheumatologist/dermatologist group had a low sample size; therefore, observations for this group are likely not generalizable. There may be unobserved variable selection bias by provider specialty, as patient characteristics and disease features may differ across provider specialties. For example, patients with less severe disease may be more likely to be treated by a primary care provider, whereas those with more severe disease may be referred to a specialist or may seek out a specialist themselves. As claims data do not capture disease severity, we could not address its influence on our results.
Because psoriatic arthritis can be a progressive disease, treatment selection may have important implications for patient outcomes. Further research is necessary to determine the basis for these differences in treatment choice and costs and to evaluate the implications for outcomes in patients over the long term. Future analyses may examine the relationships between patients and physician specialty and their association with disease activity outcomes of interest.
In this retrospective cohort study of patients prescribed treatment for newly diagnosed psoriatic arthritis, we found that various physician specialties prescribed different initial treatments for their patients. In our study, psoriatic arthritis was treated by a variety of specialists, most of whom were rheumatologists.
Dermatologists were more likely than other specialists to initially prescribe adalimumab, etanercept, apremilast, and ustekinumab in the era before (2014 to 2018), but consistent with, the 2018 ACR/NPF guidelines. Other provider types, including the predominant physician specialist type in the study, rheumatologists, were more likely to prescribe methotrexate as the initial treatment for psoriatic arthritis. We found variation in prescription patterns and costs within each specialty, as well as variation in health care costs between specialties. Payers concerned about differential health care costs across specialist types, particularly for rheumatologists and dermatologists, may want to consider that different patients may require different treatments depending on their specific disease manifestations.
Wanda Krall, PhD (on behalf of Amgen Inc), and Linda Runft, PhD (Amgen), provided medical writing support in the development of this manuscript.
Author Affiliations: Amgen Inc (EJMH, EK, BS), Thousand Oaks, CA; IBM Watson Health (JT, NMZ, PC, BL), Bethesda, MD; University of Pennsylvania (AO), Philadelphia, PA.
Source of Funding: Amgen Inc.
Author Disclosures: Dr Maksabedian Hernandez, Dr Stolshek, and Dr Karis are employees of and stockholders in Amgen Inc, which owns and manufactures 2 of the medications used to treat psoriatic arthritis. At the time of writing of this paper, Mr Tkacz, Ms Zimmerman, Dr Chan, and Dr Limone were all employees of IBM, which was contracted by Amgen to conduct this work. Dr Ogdie reports consultancies or paid advisory boards for AbbVie, Amgen, Bristol Myers Squibb, Celgene, Corrona, Janssen, Lilly, Novartis, and Pfizer; grants received from Amgen, Novartis, and Pfizer; and royalties to her husband from Novartis.
Authorship Information: Concept and design (EJMH, JT, PC, BS); acquisition of data (JT, BL); analysis and interpretation of data (EJMH, JT, NMZ, PC, BL, AO, EK, BS); drafting of the manuscript (EJMH, JT, AO, EK); critical revision of the manuscript for important intellectual content (EJMH, JT, NMZ, BL, AO, EK, BS); statistical analysis (EJMH, NMZ, BL); obtaining funding (EJMH); administrative, technical, or logistic support (EJMH, JT, PC, BL); and supervision (EJMH, JT).
Address Correspondence to: Ervant J. Maksabedian Hernandez, PhD, MPhil, Amgen Inc, 1120 Veterans Blvd, Bldg ASF3, Rm 3060P, South San Francisco, CA 94080. Email: email@example.com.
1. Ogdie A, Weiss P. The epidemiology of psoriatic arthritis. Rheum Dis Clin North Am. 2015;41(4):545-568. doi:10.1016/j.rdc.2015.07.001
2. Jönsson B, Ström O, Eisman JA, et al. Cost-effectiveness of denosumab for the treatment of postmenopausal osteoporosis. Osteoporos Int. 2011;22(3):967-982. doi:10.1007/s00198-010-1424-x
3. Lebwohl MG, Kavanaugh A, Armstrong AW, Van Voorhees AS. US perspectives in the management of psoriasis and psoriatic arthritis: patient and physician results from the population-based Multinational Assessment of Psoriasis and Psoriatic Arthritis (MAPP) survey. Am J Clin Dermatol. 2016;17(1):87-97. doi:10.1007/s40257-015-0169-x
4. Maksabedian EJ, Tkacz J, Lopez-Gonzalez L, Higgins K, Ogdie A, Stolshek B. Psoriatic arthritis treatment patterns and costs among pharmacologic treatment–naïve patients. Am J Manag Care. 2020;26(8):e252-e257. doi:10.37765/ajmc.2020.44075
5. Singh JA, Guyatt G, Ogdie A, et al. Special article: 2018 American College of Rheumatology/National Psoriasis Foundation guideline for the treatment of psoriatic arthritis. Arthritis Care Res (Hoboken). 2019;71(1):2-29. doi:10.1002/acr.23789
6. Coates LC, Kavanaugh A, Mease PJ, et al. Group for Research and Assessment of Psoriasis and Psoriatic Arthritis 2015 treatment recommendations for psoriatic arthritis. Arthritis Rheumatol. 2016;68(5):1060-1071. doi:10.1002/art.39573
7. Husni ME, Fernandez A, Hauber B, et al. Comparison of US patient, rheumatologist, and dermatologist perceptions of psoriatic disease symptoms: results from the DISCONNECT study. Arthritis Res Ther. 2018;20(1):102. doi:10.1186/s13075-018-1601-4
8. Walsh JA, Pei S, Penmetsa GK, et al. Treatment patterns with disease-modifying antirheumatic drugs in U.S. veterans with newly diagnosed rheumatoid arthritis, psoriatic arthritis, or ankylosing spondylitis. J Manag Care Spec Pharm. 2019;25(11):1218-1228. doi:10.18553/jmcp.2019.25.11.1218
9. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613-619. doi:10.1016/0895-4356(92)90133-8
10. Lee H, Ford JA, Jin Y, et al. Validation of claims-based algorithms for psoriatic arthritis. Pharmacoepidemiol Drug Saf. 2020;29(4):404-408. doi:10.1002/pds.4950