Agreement of Medication Information Derived From EHR Data Compared to Medicare Insurance Claims: An Analysis of Biologic Disease-Modifying Antirheumatic Drugs.

PURPOSE: Few studies have reported the agreement between medication information derived from ambulatory EHR data compared to administrative claims for high-cost specialty drugs. We used data from a national EHR-enabled registry, the Rheumatology Informatics System for Effectiveness (RISE), with linked Medicare claims in a population of patients with rheumatoid arthritis (RA) to investigate variations in agreement for different biologic disease-modifying agents (bDMARDs) between two data sources (RISE EHR data vs. Medicare claims), categorized by drug, route of administration, and patient insurance factors (dual eligibility). METHODS: Patients ≥ 65 years old, with ≥ 2 visits in RISE with RA ICD codes ≥ 30 days apart, and continuous enrollment in Medicare Parts B and D in 2017-2018 were included. We classified patients as bDMARD users or nonusers in Medicare claims or EHR data in 2018, and we calculated sensitivity, specificity, positive predicted value (PPV), and negative predicted value (NPV) of EHR data for identifying bDMARD users, using Medicare as the reference standard. We also calculated these metrics after stratifying by clinic-administered (Part B) versus. pharmacy-dispensed (Part D) bDMARDs and by patient dual-eligibility. RESULTS: A total of 26 097 patients were included in the study. Using Medicare claims as the reference standard, EHR data had a sensitivity of 75.0%-90.8% for identifying patients with the same medication and route. PPV for Part B bDMARDs was higher compared with Part D bDMARDs (range 94.3%-97.3% vs. 51.0%-69.6%). We observed higher PPVs for Part D bDMARDs among patients who were dual-eligible (range 82.4%-95.1%). CONCLUSION: The risk of misclassification of drug exposure based on EHR data sources alone is small for Medicare Part B bDMARDs but could be as high as 50% for Part D bDMARDs, in particular for patients who are not dually eligible for Medicare and Medicaid.

Association of Cardiovascular Outcomes With Low-Dose Glucocorticoid Prescription in Patients With Rheumatoid Arthritis.

OBJECTIVE: Many guidelines recommend limiting glucocorticoids in patients with rheumatoid arthritis (RA), but 40% of patients remain on glucocorticoids long term. We evaluated the cardiovascular risk of long-term glucocorticoid prescription by studying patients on stable disease-modifying antirheumatic drugs (DMARDs). METHODS: Using two claims databases, we identified patients with RA on stable DMARD therapy for >180 days. Proportional hazards models with inverse-probability weights and clustering to account for multiple observations were used to estimate the effect of glucocorticoid dose on composite cardiovascular outcomes (stroke or myocardial infarction [MI]). RESULTS: There were 135,583 patients in Medicare and 39,272 in Optum's de-identified Clinformatics Data Mart (CDM) database. Medicare and CDM patients had an incidence of 1.3 and 0.8 composite cardiovascular outcomes per 100 person-years, respectively. In the older, comorbid Medicare cohort, glucocorticoids were associated with a dose-dependent increase in composite cardiovascular outcomes in adjusted models with predicted one-year incidence of 1.4% (95% confidence interval [CI] 1.2%-1.6%) for ≤5 mg, 1.6% (95% CI 1.4%-1.9%) for >5 to 10 mg, and 1.8% (95% CI 1.2%-2.5%) for >10 mg versus 1.1% (95% CI 1.1%-1.2%) among patients not receiving glucocorticoids. There was no significant association among the CDM cohort. However, in the subgroup of younger patients with RA and higher cardiovascular risk, glucocorticoids were associated with a dose-dependent increase in composite cardiovascular outcomes. CONCLUSION: Among older patients with more comorbidities and younger patients with higher cardiovascular risk with RA on stable DMARD therapy, glucocorticoids were associated with a dose-dependent increased risk of MI and stroke, even at doses ≤5 mg/day. By contrast, no association was noted among younger, healthier patients with RA.

Incidence of checkpoint inhibitor-associated inflammatory arthritis, immunomodulation and mortality in cancer patients on immunotherapy: a retrospective cohort study.

OBJECTIVES: Immune checkpoint inhibitor (ICI) associated inflammatory arthritis (ICI-IA) occurs in 4-6% of ICI-treated patients based on one observational study. We identified cases of ICI-IA using administrative claims to study its incidence and characteristics at the population level. METHODS: We used the Medicare 5% sample to identify patients initiating ICIs. Cancer patients were identified by having ≥ 2 ICD-9/10-CM diagnosis codes from an oncologist for lung cancer, melanoma, or renal/urothelial cancer. ICI-IA was defined as having two Medicare claims ≥ 30 days apart with combinations of ICD-9/10-CM diagnosis codes that favored specificity. ICI-IA was identified in patients with a musculoskeletal diagnosis after ICI initiation, who had i.) no inflammatory arthritis or inflammatory rheumatic disease before ICI initiation ever, and ii) no musculoskeletal complaint in the one year prior to ICI. We examined DMARD utilization and visits to rheumatology in patients with ICI-IA. Landmark analysis and a time varying Cox proportional hazards model for overall survival was constructed. RESULTS: The incidence of ICI-IA was 7.2 (6.1-8.4) per 100 patient years. Patients with ICI-IA were mean (SD) age 73.5(7.0) years, 48% women, 91% white. Median(IQR) time from ICI initiation to first ICI-IA diagnosis was 124(56, 252) days. Only 24(16%) received care from a rheumatologist, and 24(16%) were prescribed a DMARD (46% by a rheumatologist). The HR for mortality in patients with ICI-IA was 0.86 (95% CI 0.59-1.26, p= 0.45). CONCLUSIONS: The incidence of ICI-IA identified in claims data is similar to that reported in observational studies, however, few patients are treated with a DMARD or see a rheumatologist. There was no difference in overall survival between ICI-treated patients with and without ICI-IA.

Real-World Effectiveness of Pegloticase Associated With Use of Concomitant Immunomodulatory Therapy.

OBJECTIVE: The objective of this study was to ascertain pegloticase persistence and adverse events associated with concomitant immunomodulatory drug treatment in patients with gout. METHODS: We conducted a retrospective analysis of patients with gout using the American College of Rheumatology's Rheumatology Informatics System for Effectiveness registry from January 2016 through June 2020. The first pegloticase infusion defined the index date. Based on concomitant immunomodulatory drug treatment, we identified three exposure groups: (1) immunomodulatory drug initiators (patients initiating an immunomodulatory prescription ±60 days from the index date), (2) prevalent immunomodulatory drug recipients (patients receiving their first immunomodulatory drug prescription >60 days before the index date with at least one prescription within ±60 days of the index date), and (3) immunomodulatory nonrecipients (patients receiving pegloticase without concomitant immunomodulatory drugs). We calculated the proportion of patients who achieved serum urate levels ≤6 mg/dL and who had laboratory abnormalities (white blood cell count <3.4 x 109/L, platelet count <135,000, hematocrit level <30%, alanine aminotransferase or aspartate aminotransferase level ≥1.5 times the upper limit normal value) within 180 days after the index date. Cox regression analyzed time to pegloticase discontinuation, controlling for potential confounders. RESULTS: We identified 700 pegloticase recipients (91 immunomodulatory drug initiators, 33 prevalent immunomodulatory drug recipients, and 576 nonrecipients), with a median follow-up of 14 months. Immunomodulatory drug recipients were less likely to discontinue pegloticase. The adjusted hazard ratios of pegloticase discontinuation associated with concomitant immunomodulatory drug initiation and prevalent treatment were 0.52 (95% confidence interval [CI] 0.37-0.75) and 0.69 (95% CI 0.42-1.16), respectively. Laboratory abnormalities were uncommon (<5%) and were not higher in concomitant immunomodulatory drug treatment. CONCLUSION: Consistent with clinical trials, results from this large observational registry suggest that concomitant immunomodulatory drug treatment improves pegloticase persistence.

Urate-lowering therapy, serum urate, inflammatory biomarkers, and renal function in patients with gout following pegloticase discontinuation.

BACKGROUND/PURPOSE: Little is known about long-term clinical outcomes or urate-lowering (ULT) therapy use following pegloticase discontinuation. We examined ULT use, serum urate (SU), inflammatory biomarkers, and renal function following pegloticase discontinuation. METHODS: We conducted a retrospective analysis of gout patients who discontinued pegloticase using the Rheumatology Informatics System for Effectiveness (RISE) registry from 1/2016 to 6/2022. We defined discontinuation as a gap ≥ 12 weeks after last infusion. We examined outcomes beginning two weeks after last dose and identified ULT therapy following pegloticase discontinuation. We evaluated changes in lab values (SU, eGFR, CRP and ESR), comparing on- treatment (≤ 15 days of the second pegloticase dose) to post-treatment. RESULTS: Of the 375 gout patients discontinuing pegloticase, median (IQR) laboratory changes following discontinuation were: SU: +2.4 mg/dL (0.0,6.3); eGFR: -1.9 mL/min (- 8.7,3.7); CRP: -0.8 mg/L (-12.8,0.0); and ESR: -4.0 mm/hr (-13.0,0.0). Therapy post-discontinuation included oral ULTs (86.0%), restarting pegloticase (4.5%), and no documentation of ULT (9.5%), excluding patients with multiple same-day prescriptions (n = 17). Oral ULTs following pegloticase were: 62.7% allopurinol, 34.1% febuxostat. The median (IQR) time to starting/restarting ULT was 92.0 days (55.0,173.0). Following ULT prescribing (≥ 30 days), only 51.0% of patients had SU < 6 mg/dL. Patients restarting pegloticase achieved a median SU of 0.9 mg/dL (IQR:0.2,9.7) and 58.3% had an SU < 6 mg/dL. CONCLUSION: Pegloticase treats uncontrolled gout in patients with failed response to xanthine oxidase inhibitors, but among patients who discontinue, optimal treatment is unclear. Based on this analysis, only half of those starting another ULT achieved target SU. Close follow-up is needed to optimize outcomes after pegloticase discontinuation.

Wearable activity tracker study exploring rheumatoid arthritis patients' disease activity using patient-reported outcome measures, clinical measures, and biometric sensor data (the wear study).

Background: Digital health studies using electronic patient reported outcomes (ePROs), wearables, and clinical data to provide a more comprehensive picture of patient health. Methods: Newly initiated patients on upadacitinib or adalimumab for RA will be recruited from community settings in the Excellence NEtwork in RheumatoloGY (ENRGY) practice-based research network. Over the period of three to six months, three streams of data will be collected (1) linkable physician-derived data; (2) self-reported daily and weekly ePROs through the ArthritisPower registry app; and (3) biometric sensor data passively collected via wearable. These data will be analyzed to evaluate correlations among the three types of data and patient improvement on the newly initiated medication. Conclusions: Results from this study will provide valuable information regarding the relationships between physician data, wearable data, and ePROs in patients newly initiating an RA treatment, and demonstrate the feasibility of digital data capture for Remote Patient Monitoring of patients with rheumatic disease.

Changes in the Workforce Characteristics of Providers Who Care for Adult Patients With Rheumatologic and Musculoskeletal Disease in the United States.

OBJECTIVE: The aim of this study was to describe the adult rheumatology workforce in the United States, assess change in rheumatology providers over time, and identify variation in rheumatology practice characteristics. METHODS: Using national Medicare claims data from 2006 to 2020, clinically active rheumatology physicians and advanced practice providers (APPs) were identified. Each calendar year was used for inclusion, exclusion, and analysis, and providers were determined to be entering, exiting, or stable based upon presence or absence in the prior or subsequent years of data. Characteristics (age, gender, practice type, rural, and region) of rheumatologists were determined for 2019 and in mutually exclusive study periods from 2009 to 2011, 2012 to 2015, and 2016 to 2019. The location of rheumatology practice was determined by billing tax identification and mapped. Demographics of physicians exiting or entering the rheumatology workforce were compared separately to those stable by logistic regression. RESULTS: The clinically active adult rheumatology workforce identified in US Medicare in 2019 was 5,667 rheumatologists and 379 APPs. From 2009 to 2020, the number of rheumatologists increased 23% and the number of APPs increased 141%. There was an increase in female rheumatologists over time, rising to 43% in 2019. Women and those employed by a health care system were more likely to exit, and those in a small practice or in the South were less likely to exit. CONCLUSION: The overall number of clinically active rheumatology providers grew more than 20% over the last decade to a high of 6,036 in 2020, although this rate of growth appears to be flattening off in later years.

COVID-19 Vaccine Uptake, Hesitancy, and Flare in a Large Rheumatology Practice Network.

OBJECTIVE: The goal of this study was to ascertain COVID-19 vaccine uptake, reasons for hesitancy, and self-reported flare in a large rheumatology practice-based network. METHODS: A tablet-based survey was deployed by 108 rheumatology practices from December 2021 to December 2022. Patients were asked about COVID-19 vaccine status and why they might not receive a vaccine or booster. We used descriptive statistics to explore the differences between vaccination status and vaccine and booster hesitancy, comparing patients with and without autoimmune and inflammatory rheumatic diseases (AIIRDs). We used multivariable logistic regression to examine the association between vaccine uptake and AIIRD status and self-reported flare and AIIRD status. We reported adjusted odds ratios (aORs). RESULTS: Of the 61,158 patients, 89% reported at least one dose of vaccine; of the vaccinated, 68% reported at least one booster. Vaccinated patients were less likely to have AIIRDs (44% vs 56%). A greater proportion of patients with AIIRDs were vaccine hesitant (14% vs 10%) and booster hesitant (21% vs 16%) compared to patients without AIIRDs. Safety concerns (28%) and side effects (23%) were the main reasons for vaccine hesitancy, whereas a lack of recommendation from the physician was the primary factor for booster hesitancy (23%). Patients with AIIRD did not have increased odds of self-reported flare or worsening disease compared to patients without with AIIRD (aOR 0.99, 95% confidence interval [CI] 0.94-1.05). Among the patients who were vaccine hesitant and booster hesitant, 12% and 39% later reported receiving a respective dose. Patients with AIIRD were 32% less likely to receive a vaccine (aOR 0.68, 95% CI 0.65-0.72) versus patients without AIIRD. CONCLUSION: Some patients who are vaccine and booster hesitant eventually receive a vaccine dose, and future interventions tailored to patients with AIIRD may be fruitful.

Treatment of Polymyalgia Rheumatica by Rheumatology Providers: Analysis from the American College of Rheumatology RISE Registry.

OBJECTIVE: This study describes the demographics, comorbidities, and treatment patterns in a national cohort of patients with polymyalgia rheumatica (PMR) who received care from rheumatology providers. METHODS: Patients with PMR were identified in the American College of Rheumatology Rheumatology Informatics System for Effectiveness registry from 2016 to 2022. Use of glucocorticoids and immunomodulatory antirheumatic medications used as steroid-sparing agents were examined overall and in a subgroup of patients new to rheumatology practices, the majority with presumed new-onset PMR. In these new patients, multivariate logistic regressions were performed to identify factors associated with persistent glucocorticoid and steroid-sparing agent use at 12 to 24 months. RESULTS: A total of 26,102 patients with PMR were identified, of which 16,703 new patients were included in the main analysis. Patients were predominantly female (55.8%) and White (46.7%), with a mean age of 72.0 years. Hypertension (81.2%), congestive heart failure (52.4%), hyperlipidemia (41.3%), and ischemic heart disease (36.0%) were the most prevalent comorbidities. At baseline, 92.3% of patients were on glucocorticoids, and only 13.1% were on a steroid-sparing agent. At 12 to 24 months, most patients remained on glucocorticoids (63.8%). Although there was an increase in use through follow-up, antirheumatic medications were prescribed only to a minority (39.0%) of patients with PMR. CONCLUSION: In this large US-based study of patients with PMR receiving rheumatology care, only a minority of patients were prescribed steroid-sparing agents during the first 24 months of follow-up; most patients remained on glucocorticoids past one year. Further identification of patients who would benefit from steroid-sparing agents and the timing of steroid-sparing agent initiation is needed.

Using Electronic Health Records and Linked Claims Data to Assess New Medication Use and Primary Nonadherence in Rheumatology Patients.

OBJECTIVE: The objective of this study was to determine the proportion of new medication prescriptions observed in electronic health records (EHR) that represent true incident medication use, accounting for undocumented previous prescriptions (prevalent medication use) and failure to initiate treatment (primary nonadherence) with linked administrative claims data as the reference standard. METHODS: Using single-specialty rheumatology EHR data from more than 700 community practices in the United States linked to administrative claims data, we identified first (index) EHR prescriptions and assessed the positive predictive value (PPV) of different EHR-derived new user definitions to identify true incident use (no prior claims). We then assessed how often index EHR prescriptions that met a definition of new use resulted in primary nonadherence (no subsequent claims). RESULTS: Overall, 12,405 index EHR prescriptions were identified with PPVs of 0.59 to 0.67 for true incident use. PPVs increased to 0.76 to 0.85 by excluding medications listed during the EHR medication reconciliation process and further increased to 0.87 to 0.93 by requiring ≥12 elapsed months since the first rheumatology office visit. Primary nonadherence at three months was observed in 33% to 38% overall and varied substantially by medication class, ranging from 15% to 23% for conventional synthetic disease-modifying antirheumatic drugs (DMARDs) to 54% to 64% for targeted synthetic DMARDs. CONCLUSION: New DMARD use was accurately distinguished from prevalent use with EHR prescriptions and simple new user definitions that include current medications collected during medication reconciliation. Primary nonadherence was frequent and varied by DMARD class. This has important implications for epidemiologic studies using EHR data and for optimal delivery of clinical care.

Participant Engagement and Adherence to Providing Smartwatch and Patient-Reported Outcome Data: Digital Tracking of Rheumatoid Arthritis Longitudinally (DIGITAL) Real-World Study.

BACKGROUND: Digital health studies using electronic patient-reported outcomes (ePROs) and wearables bring new challenges, including the need for participants to consistently provide trial data. OBJECTIVE: This study aims to characterize the engagement, protocol adherence, and data completeness among participants with rheumatoid arthritis enrolled in the Digital Tracking of Arthritis Longitudinally (DIGITAL) study. METHODS: Participants were invited to participate in this app-based study, which included a 14-day run-in and an 84-day main study. In the run-in period, data were collected via the ArthritisPower mobile app to increase app familiarity and identify the individuals who were motivated to participate. Successful completers of the run-in period were mailed a wearable smartwatch, and automated and manual prompts were sent to participants, reminding them to complete app input or regularly wear and synchronize devices, respectively, during the main study. Study coordinators monitored participant data and contacted participants via email, SMS text messaging, and phone to resolve adherence issues per a priori rules, in which consecutive spans of missing data triggered participant contact. Adherence to data collection during the main study period was defined as providing requested data for >70% of 84 days (daily ePRO, ≥80% daily smartwatch data) or at least 9 of 12 weeks (weekly ePRO). RESULTS: Of the 470 participants expressing initial interest, 278 (59.1%) completed the run-in period and qualified for the main study. Over the 12-week main study period, 87.4% (243/278) of participants met the definition of adherence to protocol-specified data collection for weekly ePRO, and 57.2% (159/278) did so for daily ePRO. For smartwatch data, 81.7% (227/278) of the participants adhered to the protocol-specified data collection. In total, 52.9% (147/278) of the participants met composite adherence. CONCLUSIONS: Compared with other digital health rheumatoid arthritis studies, a short run-in period appears useful for identifying participants likely to engage in a study that collects data via a mobile app and wearables and gives participants time to acclimate to study requirements. Automated or manual prompts (ie, "It's time to sync your smartwatch") may be necessary to optimize adherence. Adherence varies by data collection type (eg, ePRO vs smartwatch data). INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.2196/14665.

Assessment of Real-World Patient-Reported Outcomes in Patients Initiating Biologic Agents for the Treatment of Autoimmune Diseases: An Observational Study in Four Patient-Powered Research Networks.

Background: The most reliable and meaningful approach for inclusion of patient-reported outcomes (PROs) in the evaluation of real-world clinical effectiveness of biologics in the treatment of autoimmune diseases is u ncertain. This study aimed to assess and compare the proportions of patients who had abnormalities in PROs measuring important general health domains at the initiation of treatment with biologics, as well as the effects of baseline abnormalities on subsequent improvement. Methods: PROs were collected for patient participants with inflammatory arthritis, inflammatory bowel disease, and vasculitis using Patient-Reported Outcomes Measurement Information System instruments. Scores were reported as T-scores normalized to the general population in the United States. Baseline PROs scores were collected near the time of biologic initiation, and follow-up scores were collected 3 to 8 months later. In addition to summary statistics, the proportion of patients with PROs abnormalities (scores ≥5 units worse than the population norm) was determined. Baseline and follow-up scores were compared, and an improvement of ≥5 units was considered significant. Results: There was wide variation across autoimmune diseases in baseline PROs scores for all domains. For example, the proportion of participants with abnormal baseline pain interference scores ranged from 52% to 93%. When restricted to participants with baseline PROs abnormalities, the proportion of participants experiencing an improvement of ≥5 units was substantially higher. Conclusion: As expected, many patients experienced improvement in PROs following initiation of treatment with biologics for autoimmune diseases. Nevertheless, a substantial proportion of participants did not exhibit abnormalities in all PROs domains at baseline, and these participants appear less likely to experience improvement. For PROs to be reliably and meaningfully included in the evaluation of real-world medication effectiveness, more knowledge and careful consideration are needed to select the most appropriate patient populations and subgroups for inclusion and evaluation in studies measuring change in PROs.

Identifying inpatient mortality in MarketScan claims data using machine learning.

PURPOSE: Inpatient mortality is an important variable in epidemiology studies using claims data. In 2016, MarketScan data began obscuring specific hospital discharge status types for patient privacy, including inpatient deaths, by setting the values to missing. We used a machine learning approach to correctly identify hospitalizations that resulted in inpatient death using data prior to 2016. METHODS: All hospitalizations from 2011 to 2015 with discharge status of missing, died, or one of the other subsequently obscured values were identified and divided into a training set and two test sets. Predictor variables included age, sex, elapsed time from hospital discharge until last observed claim and until healthcare plan disenrollment, and absence of any discharge diagnoses. Four machine learning methods were used to train statistical models and assess sensitivity and positive predictive value (PPV) for inpatient mortality. RESULTS: Overall 1 307 917 hospitalizations were included. All four machine learning approaches performed well in all datasets. Random forest performed best with 88% PPV and 93% sensitivity for the training set and both test sets. The two factors with the highest relative importance for identifying inpatient mortality were having no observed claims for the patient on days 2-91 following hospital discharge and patient disenrollment from the healthcare plan within 60 days following hospital discharge. CONCLUSION: We successfully developed machine learning algorithms to identify inpatient mortality. This approach can be applied to obscured data to accurately identify inpatient mortality among hospitalizations with missing discharge status.

Accuracy of administrative claims prescription fill data to estimate glucocorticoid use and dose in patients with rheumatoid arthritis.

PURPOSE: To assess accuracy of administrative claims prescription fill-based estimates of glucocorticoid use and dose, and approximate bias from glucocorticoid exposure misclassification. METHODS: We identified adults with rheumatoid arthritis with linked Medicare and CorEvitas registry data. An algorithm identifying glucocorticoid use and average dose over 90 days from Medicare prescription fills was compared to physician-reported measures from a CorEvitas visit during the same period, using weighted kappa to compare doses (none, ≤5 mg, 5-10 mg, >10 mg/day). A deterministic sensitivity analysis examined the effect of exposure misclassification on estimated glucocorticoid-associated infection risk from a prior study. RESULTS: We identified 621 observations among 494 patients. Prescription fills identified glucocorticoid use in 41.9% of observations versus 31.1% identified by CorEvitas physician-report. For glucocorticoid use (yes/no), prescription fills had sensitivity 88.1% (95% CI 82.7-92.3), specificity 79.0% (74.8-82.7), PPV 65.4% (59.3-71.2), NPV 93.6% (90.6-95.9), and 81.8% agreement with CorEvitas, with kappa 0.61 (moderate to substantial agreement). There was 89.5% agreement between prescription fills and physician-reported doses, with weighted kappa 0.56 (moderate agreement). Applying these results to a prior Medicare study evaluating glucocorticoid-associated infection risk [risk ratio 1.44 (95% CI 1.41-1.48)] led to an externally adjusted risk ratio of 1.74 when accounting for exposure misclassification, representing -17% bias in infection risk estimate. CONCLUSIONS: This study supports the use of claims data to estimate glucocorticoid use and dose, but investigators should account for exposure misclassification, which may lead to underestimates of glucocorticoid risks. Our results could be applied to adjust risk estimates in other studies that use prescription fills to estimate glucocorticoid use.

Machine Learning Approaches to Classify Self-Reported Rheumatoid Arthritis Health Scores Using Activity Tracker Data: Longitudinal Observational Study.

BACKGROUND: The increasing use of activity trackers in mobile health studies to passively collect physical data has shown promise in lessening participation burden to provide actively contributed patient-reported outcome (PRO) information. OBJECTIVE: The aim of this study was to develop machine learning models to classify and predict PRO scores using Fitbit data from a cohort of patients with rheumatoid arthritis. METHODS: Two different models were built to classify PRO scores: a random forest classifier model that treated each week of observations independently when making weekly predictions of PRO scores, and a hidden Markov model that additionally took correlations between successive weeks into account. Analyses compared model evaluation metrics for (1) a binary task of distinguishing a normal PRO score from a severe PRO score and (2) a multiclass task of classifying a PRO score state for a given week. RESULTS: For both the binary and multiclass tasks, the hidden Markov model significantly (P<.05) outperformed the random forest model for all PRO scores, and the highest area under the curve, Pearson correlation coefficient, and Cohen κ coefficient were 0.750, 0.479, and 0.471, respectively. CONCLUSIONS: While further validation of our results and evaluation in a real-world setting remains, this study demonstrates the ability of physical activity tracker data to classify health status over time in patients with rheumatoid arthritis and enables the possibility of scheduling preventive clinical interventions as needed. If patient outcomes can be monitored in real time, there is potential to improve clinical care for patients with other chronic conditions.

Using pooled electronic health records data to conduct pharmacoepidemiology safety studies: Challenges and lessons learned.

PURPOSE: We assessed the suitability of pooled electronic health record (EHR) data from clinical research networks (CRNs) of the patient-centered outcomes research network to conduct studies of the association between tumor necrosis factor inhibitors (TNFi) and infections. METHODS: EHR data from patients with one of seven autoimmune diseases were obtained from three CRNs and pooled. Person-level linkage of CRN data and Centers for Medicare and Medicaid Services (CMS) fee-for-service claims data was performed where possible. Using filled prescriptions from CMS claims data as the gold standard, we assessed the misclassification of EHR-based new (incident) user definitions. Among new users of TNFi, we assessed subsequent rates of hospitalized infection in EHR and CMS data. RESULTS: The study included 45 483 new users of TNFi, of whom 1416 were successfully linked to their CMS claims. Overall, 44% of new EHR TNFi prescriptions were not associated with medication claims. Our most specific new user definition had a misclassification rate of 3.5%-16.4% for prevalent use, depending on the medication. Greater than 80% of CRN prescriptions had either zero refills or missing refill data. Compared to using EHR data alone, there was a 2- to 8-fold increase in hospitalized infection rates when CMS claims data were added to the analysis. CONCLUSIONS: EHR data substantially misclassified TNFi exposure and underestimated the incidence of hospitalized infections compared to claims data. EHR-based new user definitions were reasonably accurate. Overall, using CRN data for pharmacoepidemiology studies is challenging, especially for biologics, and would benefit from supplementation by other sources.

Geographic Variation in Disease Burden and Mismatch in Care of Patients With Rheumatoid Arthritis in the United States.

OBJECTIVE: Our objective was to evaluate the factors associated with regional variation of rheumatoid arthritis (RA) disease burden in the US. METHODS: In a retrospective cohort analysis of Rheumatology Informatics System for Effectiveness (RISE) registry data, seropositivity, RA disease activity (Clinical Disease Activity Index [CDAI], Routine Assessment of Patient Index Data-version 3 [RAPID3]), socioeconomic status (SES), geographic region, health insurance type, and comorbidity burden were recorded. An Area Deprivation Index score of more than 80 defined low SES. Median travel distance to practice sites' zip codes was calculated. Linear regression was used to analyze associations between RA disease activity and comorbidity adjusting for age, sex, geographic region, race, and insurance type. RESULTS: Enrollment data for 184,722 patients with RA from 182 RISE sites were analyzed. Disease activity was higher in African American patients, in those from Southern regions, and in those with Medicaid or Medicare coverage. Greater comorbidity was prevalent in patients in the South and those with Medicare or Medicaid coverage. There was moderate correlation between comorbidity and disease activity (Pearson coefficient: RAPID3 0.28, CDAI 0.15). High-deprivation areas were mainly in the South. Less than 10% of all participating practices cared for more than 50% of all Medicaid recipients. Patients living more than 200 miles away from specialist care were located mainly in Southern and Western regions. CONCLUSION: A disproportionately large portion of socially deprived, high comorbidity, and Medicaid-covered patients with RA were cared for by a minority of rheumatology practices. Studies are needed in high-deprivation areas to establish more equitable distribution of specialty care for patients with RA.

Identification of Multimorbidity Patterns in Rheumatoid Arthritis Through Machine Learning.

OBJECTIVE: Recognizing that the interrelationships between chronic conditions that complicate rheumatoid arthritis (RA) are poorly understood, we aimed to identify patterns of multimorbidity and to define their prevalence in RA through machine learning. METHODS: We constructed RA and age- and sex-matched (1:1) non-RA cohorts within a large commercial insurance database (MarketScan) and the Veterans Health Administration (VHA). Chronic conditions (n = 44) were identified from diagnosis codes from outpatient and inpatient encounters. Exploratory factor analysis was performed separately in both databases, stratified by RA diagnosis and sex, to identify multimorbidity patterns. The association of RA with different multimorbidity patterns was determined using conditional logistic regression. RESULTS: We studied 226,850 patients in MarketScan (76% female) and 120,780 patients in the VHA (89% male). The primary multimorbidity patterns identified were characterized by the presence of cardiopulmonary, cardiometabolic, and mental health and chronic pain disorders. Multimorbidity patterns were similar between RA and non-RA patients, female and male patients, and patients in MarketScan and the VHA. RA patients had higher odds of each multimorbidity pattern (odds ratios [ORs] 1.17-2.96), with mental health and chronic pain disorders being the multimorbidity pattern most strongly associated with RA (ORs 2.07-2.96). CONCLUSION: Cardiopulmonary, cardiometabolic, and mental health and chronic pain disorders represent predominant multimorbidity patterns, each of which is overrepresented in RA. The identification of multimorbidity patterns occurring more frequently in RA is an important first step in progressing toward a holistic approach to RA management and warrants assessment of their clinical and predictive utility.

Machine Learning Applied to Patient-Reported Outcomes to Classify Physician-Derived Measures of Rheumatoid Arthritis Disease Activity.

OBJECTIVE: Patient-reported outcome (PRO) data have assumed increasing importance in the care of patients with rheumatoid arthritis (RA), yet physician-derived disease activity measures, such as Clinical Disease Activity Index (CDAI), remain the most accepted metrics to assess disease activity. The possibility that newer longitudinal PRO data might be used as a proxy for the CDAI has not been evaluated. METHODS: Using data from a large pragmatic trial, we evaluated patients with RA initiating golimumab intravenous or infliximab. The classification target was low disease activity (LDA) (CDAI ≤10) at the first visit between months 3 and 12. Data were randomly partitioned into training (80%) and test (20%) data sets. Multiple machine learning (ML) methods (eg, random forests, gradient boosting, support vector machines) were used to classify CDAI disease activity category, conduct feature selection, and assess feature importance. Model performance evaluated cross-validated error, comparing different ML approaches using both training and test data. RESULTS: A total of 494 patients were analyzed, and 36.4% achieved LDA. The most important classification features included several Patient-Reported Outcomes Measurement Information System measures (social participation, pain interference, pain intensity, and physical function), patient global, and baseline CDAI. Among all ML methods, random forests performed best. Overall model accuracy and positive predictive values for all ML methods were approximately 80%. CONCLUSION: ML methods coupled with longitudinal PRO data appear useful and can achieve reasonable accuracy in classifying LDA among patients starting a new biologic. This approach has promise for real-world evidence generation in the common circumstance when physician-derived disease activity data are not available yet PRO measures are.

Incidence of dementia in patients with rheumatoid arthritis and association with disease modifying anti-rheumatic drugs - Analysis of a national claims database.

OBJECTIVE: To evaluate the risk of incident dementia associated with the use of biologics or targeted synthetic DMARDs (b/tsDMARD) compared to conventional synthetic (cs) DMARDS only in patients with rheumatoid arthritis (RA). METHODS: We analyzed claims data from the Center for Medicare & Medicare Services (CMS) from 2006-2017. Patients with RA were identified as adults ≥40 years old and two RA diagnoses by a rheumatologist > 7 and < 365 days apart. Patients with a prior diagnosis of dementia were excluded. Use of cs/b/tsDMARDs was the exposure of interest. Person-time was classified as either: 1) b/tsDMARD exposed, which included tumor necrosis factor alpha inhibitors (TNFi)-bDMARDs, non-TNFi-bDMARDs or tsDMARDs with or without csDMARDs; 2) csDMARD-exposed: any csDMARD without b/tsDMARD. Patients could contribute time to different exposure groups if they changed medications. Incident dementia was defined as: 1 inpatient OR 2 outpatients ICD-9-CM or ICD-10 claims for dementia, OR prescription of a dementia-specific medication (rivastigmine, galantamine, memantine, donepezil, tacrine). Age-adjusted incident rates (IR) were calculated, and univariate and multivariate Cox proportional hazard models were used to calculate Hazard Ratios (HR) and 95% confidence intervals (CI). RESULTS: We identified 141,326 eligible RA patients; 80% female and 75.3% white, median age 67 years and mean (SD) exposure time of 1.1 (1.5) years. There were 233,271 initiations of c/b/tsDMARDS and 3,794 cases of incident dementia during follow up. The crude IR of dementia was 2.0 (95% CI 1.9-2.1) per 100 person-years for patients on csDMARDs and 1.3 (95% CI 1.2-1.4) for patients on any b/tsDMARD. Patients on b/tsDMARDs had an adjusted 19% lower risk for dementia than patients on csDMARDs [HR 0.81 (95% CI 0.76-0.87)]. Subgroup analysis found comparable risk reductions between TNFi, non-TNFi, and tsDMARDs. on the risk of dementia. CONCLUSIONS AND RELEVANCE: The incidence of dementia in patients with RA was lower in patients receiving b/tsDMARDs when compared to patients on csDMARD only. No differences were observed between different classes of b/tsDMARDs, suggesting that decreased risk is possibly explained by the overall decrease in inflammation rather than a specific mechanism of action of these drugs.