Accuracy of Veterans Administration databases for a diagnosis of rheumatoid arthritis

Racial and Ethnic Distribution of Rheumatic Diseases in Health Systems of the National Patient-Centered Clinical Research Network

OBJECTIVE

To evaluate the relative prevalence of 8 rheumatic and musculoskeletal diseases (RMDs) across racial and ethnic groups within the National Patient-Centered Clinical Research Network (PCORnet).

METHODS

Electronic health records from participating PCORnet institutions and systems from January 1, 2013, to December 31, 2018, were used to identify adult patients with ≥ 2 diagnosis codes for rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), osteoporosis (OP), granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), eosinophilic granulomatosis with polyangiitis (EGPA), giant cell arteritis (GCA), and Takayasu arteritis (TAK). Among those with race and ethnicity data available, we compared prevalence of RMDs by race and ethnicity.

RESULTS

Data from 28,059,546 patients were available for analysis. RA was more common in patients who were American Indian or Alaska Native vs White, with a prevalence of 11.57 vs 10.11/1000 (odds ratio [OR] 1.15, 95% CI 1.09-1.22). SLE was more common in patients who were Black or African American (6.73/1000), American Indian or Alaska Native (3.82/1000), and Asian (3.39/1000) vs White (2.80/1000; OR 2.43, 95% CI 2.39-2.46; OR 1.39, 95% CI 1.25-1.53; OR 1.26, 95% CI 1.21-1.31, respectively). SLE was more common in patients who were Hispanic vs non-Hispanic (prevalence 3.93 vs 3.45/1000, OR 1.14, 95% CI 1.12-1.16). TAK was more common in patients who were Asian vs White (prevalence 0.05 vs 0.04/1000, OR 1.43, 95% CI 1.00-2.03). OP, RA, and the vasculitides were all more common in patients who were White vs Black or African American.

CONCLUSION

These data provide important information on the prevalence of RMDs by race and ethnicity in the United States. PCORnet can be used as a reliable data source to study RMDs within a large representative population.

Changes in Patterns of Use of Advanced Therapies Following Emerging Data About Adverse Events in Patients With Rheumatoid Arthritis From the Veterans Affairs Health System

OBJECTIVE

To determine whether prescribing practices for Janus kinase inhibitors (JAKi), tumor necrosis factor inhibitors (TNFi), and non-TNFi biologic agents changed after the results of the Oral Rheumatoid Arthritis Trial (ORAL) Surveillance trial were released in January 2021.

METHODS

This is a retrospective study in adult patients with rheumatoid arthritis (RA) receiving advanced therapies within the Veterans Affairs Health System from January 2012 through September 2022. Eligible patients were required to have at least one diagnosis code for RA and to have received a biologic disease-modifying antirheumatic drug or JAKi. Treatment courses were defined from pharmacy dispensing data and the number of new courses of each advanced therapy was quantified over time. We assessed changes in the use of each therapy before and after the release of safety data (January 2021).

RESULTS

A total of 88,253 individual drug courses (in 34,656 unique patients) were included in the study. There was a consistent increase in the number and proportion of new courses of JAKi leading up to January 2021, which was followed by a significant net decrease in JAKi use through September 2022. There was significantly less tofacitinib use after the release of safety data, with a significant difference in the slope of change in use with time. In contrast, whereas TNFi use declined leading up to 2021, its use significantly increased after January 2021.

CONCLUSION

Changes in prescribing in response to new evidence emphasize the impact that safety trials have on prescribing practices. Ongoing study in this area, with attention to specific patient characteristics and risk profiles, will help characterize these changes in practice.

Clinical and Economic Burden of Herpes Zoster in Patients with Rheumatoid Arthritis: A Retrospective Cohort Study Using Administrative Claims

OBJECTIVE

To estimate the incremental healthcare resource utilization (HRU) and cost burden posed by herpes zoster (HZ) in adult patients with rheumatoid arthritis (RA) in the United States.

METHODS

A retrospective cohort study was conducted using an administrative claims database containing commercial and Medicare Advantage with Part D data, between October 2015 and February 2020. Patients with RA and HZ (RA+/HZ+) or RA without HZ (RA+/HZ-) were identified based on diagnosis codes and relevant medications. Outcomes measured included HRU and medical, pharmacy, and total costs at month 1, quarter 1, and year 1 after the index date (HZ diagnosis for RA+/HZ+ cohort, randomly assigned for RA+/HZ- cohort). Generalized linear models incorporating propensity scores and other covariates were used to estimate differences in outcomes between cohorts.

RESULTS

A total of 1866 patients from the RA+/HZ+ cohort and 38,846 patients from the RA+/HZ- cohort were included. Hospitalizations and emergency department visits occurred more frequently in the RA+/HZ+ than the RA+/HZ- cohort, especially in the month after HZ diagnosis (adjusted incidence rate ratio [95% confidence interval (CI)] for hospitalizations: 3.4 [2.8; 4.2]; emergency department visits: 3.7 [3.0; 4.4]). Total costs were also higher in the month after HZ diagnosis (mean adjusted cost difference [95% CI]: $3404 [$2089; $4779]), with cost differences driven by increased medical costs ($2677 [$1692; $3670]).

CONCLUSIONS

These findings highlight the high economic burden of HZ among individuals with RA in the United States. Strategies to reduce the risk of HZ in patients with RA (such as vaccination) may serve to reduce this burden. Video abstract.

Development and Validation of an Emergency Department Electronic Medical Record Gout Flare Alert

OBJECTIVE

Patients with acute gout are frequently treated in the emergency department (ED) and represent a typically underresourced and understudied population. A key limitation for gout research in the ED is the timely ability to identify acute gout patients. Our goal was to refine a multicriteria, electronic medical record alert for gout flares and to determine its diagnostic characteristics in the ED.

METHODS

The gout flare alert used electronic medical record data from ED nursing notes and was triggered by the term 'gout' preceding past medical history in the chief complaint, the term 'gout' and a musculoskeletal problem in the chief complaint, or the term 'gout' in the problem list and a musculoskeletal chief complaint. We validated its diagnostic properties to assess presence/absence of gout through manual medical record review using adjudicated expert consensus as the gold standard.

RESULTS

In January 2020, we analyzed 202 patient records from 2 university-based EDs; from these records, 57 patients were identified by our gout flare alert, and 145 were identified by other means as potentially having an acute gout flare. The gout flare alert's positive predictive value was 47% (95% confidence interval [95% CI] 34-60%), negative predictive value was 94% (95% CI 90-98%), sensitivity was 75% (95% CI 61-89%), and specificity was 82% (95% CI 76-88%). The diagnostic properties were similar at both institutions.

CONCLUSION

Our multicomponent gout flare alert had reasonable sensitivity and specificity, albeit a modest positive predictive value. An electronic gout flare alert may help enable the conduct of gout research in the ED setting.

Validation of algorithms to select patients with multiple myeloma and patients initiating myeloma treatment in the national Veterans Affairs Healthcare System

BACKGROUND

We aimed to evaluate and compare the performance of multiple myeloma (MM) selection algorithms for use in Veterans Affairs (VA) research.

METHODS

Using the VA Corporate Data Warehouse (CDW), the VA Cancer Registry (VACR), and VA pharmacy data, we randomly selected 500 patients from 01/01/1999 to 06/01/2021 who had (1) either one MM diagnostic code OR were listed in the VACR as having MM AND (2) at least one MM treatment code. A team reviewed oncology notes for each veteran to annotate details regarding MM diagnosis and initial treatment within VA. We evaluated inter-annotator agreement and compared the performance of four published algorithms (two developed and validated external to VA data and two used in VA data).

RESULTS

A total of 859 patients were reviewed to obtain 500 patients who were annotated as having MM and initiating MM treatment in VA. Agreement was high among annotators for all variables: MM diagnosis (98.3% agreement, Kappa = 0.93); initial treatment in VA (91.8% agreement; Kappa = 0.77); and initial treatment classification (87.6% agreement; Kappa = 0.86). VA Algorithms were more specific and had higher PPVs than non-VA algorithms for both MM diagnosis and initial treatment in VA. We developed the "VA Recommended Algorithm," which had the highest PPV among all algorithms in identifying patients diagnosed with MM (PPV = 0.98, 95% CI = 0.95-0.99) and in identifying patients who initiated their MM treatment in VA (PPV = 0.93, 95% CI = 0.90-0.96).

CONCLUSION

Our VA Recommended Algorithm optimizes sensitivity and PPV for cohort selection and treatment classification.

Risk of Prostate Cancer in US Veterans With Rheumatoid Arthritis

OBJECTIVE

Patients with rheumatoid arthritis (RA) have an increased risk of select cancers, including lymphoma and lung cancer. Whether RA influences prostate cancer risk is uncertain. We aimed to determine the risk of prostate cancer in patients with RA compared to patients without RA in the Veterans Health Administration (VA).

METHODS

We performed a matched (up to 1:5) cohort study of male patients with and without RA in the VA from 2000 to 2018. RA status, as well as covariates, were obtained from national VA databases. Prostate cancer was identified through linked VA cancer databases and the National Death Index. Multivariable Cox models compared prostate cancer risk between patients with RA and patients without RA, including models that accounted for retention in the VA system.

RESULTS

We included 56,514 veterans with RA and 227,284 veterans without RA. During 2,337,104 patient-years of follow-up, 6,550 prostate cancers occurred. Prostate cancer incidence (per 1,000 patient-years) was 3.50 (95% confidence interval [95% CI] 3.32-3.69) in patients with RA and 2.66 (95% CI 2.58-2.73) in patients without RA. After accounting for confounders and censoring for attrition of VA health care, RA was modestly associated with a higher prostate cancer risk (adjusted HR [HRadj ] 1.12 [95% CI 1.04-1.20]). There was no association between RA and prostate cancer mortality (HRadj 0.92 [95% CI 0.73-1.16]).

CONCLUSION

RA was associated with a modestly increased risk of prostate cancer, but not prostate cancer mortality, after accounting for relevant confounders and several potential sources of bias. However, even minimal unmeasured confounding could explain these findings.

The validity of rheumatoid arthritis diagnoses in Finnish biobanks

OBJECTIVE

The aim of this study was to determine the validity of rheumatoid arthritis (RA) diagnoses in patients participating in Finnish biobanks.

METHOD

We reviewed the electronic medical records of 500 Finnish biobank participants: 125 patients with at least one visit with a diagnosis of seropositive RA, 125 patients with at least one visit with a diagnosis of seronegative RA, and 250 age- and gender-matched controls. The patients were chosen from five different biobank hospitals in Finland. A rheumatologist reviewed the medical records to assess whether each patients' diagnosis was correct. The diagnosis was compared with the diagnostic codes in the Finnish Care Register for Health Care (CRHC) and special reimbursement data of the Social Insurance Institution of Finland.

RESULTS

The positive predictive value (PPV) of CRHC diagnosis of RA (for seropositive and seronegative RA combined) was 0.82. For patients with a special reimbursement for anti-rheumatic medications for RA, the PPV was 0.89. The PPV was higher in patients with more than one visit. For one, two, five, and 10 visits, the PPV was 0.82, 0.85, 0.89, and 0.90, respectively, and for patients who also had the special reimbursement, the PPV was 0.89, 0.91, 0.93, and 0.94 for one, two, five, and 10 visits, respectively. In patients positive for anti-citrullinated protein antibodies, the PPV was 0.98.

CONCLUSION

These results demonstrate that the validity of RA diagnoses in Finnish biobanks was good and can be further improved by including data on special reimbursement for medication, number of visits, and serological data.

Agreement between self-report and administrative health data on occurrence of non-cancer chronic disease among participants of the BC generations project

Population-based studies of non-cancer chronic disease often rely on self-reported data for disease diagnosis, which may be incomplete, unreliable and suffer from bias. Recently, the British Columbia Generations Project (BCGP; n = 29,736) linked self-reported chronic disease history data to a Chronic Disease Registry (CDR) that applied algorithms to administrative health data to ascertain diagnoses of multiple chronic diseases in the Province of British Columbia. For the 10 diseases captured by both self-report and the CDR, including asthma, chronic obstructive pulmonary disease (COPD), diabetes, hypertension, multiple sclerosis, myocardial infarction, osteoarthritis, osteoporosis, rheumatoid arthritis, and stroke, we calculated Cohen's kappa coefficient to examine concordance of chronic disease status (i.e., ever/never diagnosed) between the data sources. Using CDR data as the gold standard, we also calculated sensitivity, specificity, and positive-predictive value (PPV) for self-reported chronic disease occurrence. The prevalence of each chronic disease was similar across both data sources. Substantial levels of concordance (0.66-0.73) and moderate to high sensitivities (0.64-0.92), specificities (0.98-0.99) and PPVs (0.55-0.84) were observed for diabetes, hypertension, multiple sclerosis, and myocardial infarction. We did observe degree of concordance to vary by age, sex, body mass index (BMI), health perception, and ethnicity across most of the chronic diseases that were evaluated. While administrative health data are imperfect, they are less likely to suffer from bias, making them a reasonable gold standard. Our results demonstrate that for at least some chronic diseases, self-report may be a reasonable method for case ascertainment. However, characteristics of the study population will likely have impacts on the quality of the data.

Validation and Improvement of a Convolutional Neural Network to Predict the Involved Pathology in a Head and Neck Surgery Cohort

The selection of patients for the constitution of a cohort is a major issue for clinical research (prospective studies and retrospective studies in real life). Our objective was to validate in real life conditions the use of a Deep Learning process based on a neural network, for the classification of patients according to the pathology involved in a head and neck surgery department. 24,434 Electronic Health Records (EHR) from the first visit between 2000 and 2020 were extracted. More than 6000 EHR were manually classified in ten groups of interest according to the reason for consultation with a clinical relevance. A convolutional neural network (TensorFlow, previously reported by Hsu et al.) was then used to predict the group of patients based on their pathology, using two levels of classification based on clinically relevant criteria. On the first and second level of classification, macro-average performances were: 0.95, 0.83, 0.85, 0.97, 0.84 and 0.93, 0.76, 0.83, 0.96, 0.79 for accuracy, recall, precision, specificity and F1-score versus accuracy, recall and precision of 0.580, 580 and 0.582 for Hsu et al., respectively. We validated this model to predict the pathology involved and to constitute clinically relevant cohorts in a tertiary hospital. This model did not require a preprocessing stage, was used in French and showed equivalent or better performances than other already published techniques.

Hydroxychloroquine and Risk of Long QT Syndrome in Rheumatoid Arthritis: A Veterans Cohort Study With Nineteen-Year Follow-up

OBJECTIVE

Recent evidence suggests that hydroxychloroquine use is not associated with higher 1-year risk of long QT syndrome (LQTS) in patients with rheumatoid arthritis (RA). Less is known about its long-term risk, the examination of which was the objective of this study.

METHODS

We conducted a propensity score-matched active-comparator safety study of hydroxychloroquine in 8,852 veterans (mean age 64 ± 12 years, 14% women, 28% Black) with newly diagnosed RA. A total of 4,426 patients started on hydroxychloroquine and 4,426 started on another nonbiologic disease-modifying antirheumatic drug (DMARD) and were balanced on 87 baseline characteristics. The primary outcome was LQTS during 19-year follow-up through December 31, 2019.

RESULTS

Incident LQTS occurred in 4 (0.09%) and 5 (0.11%) patients in the hydroxychloroquine and other DMARD groups, respectively, during the first 2 years. Respective 5-year incidences were 17 (0.38%) and 6 (0.14%), representing 11 additional LQTS events in the hydroxychloroquine group (number needed to harm 403; [95% confidence interval (95% CI)], 217-1,740) and a 181% greater relative risk (95% CI 11%-613%; P = 0.030). Although overall 10-year risk remained significant (hazard ratio 2.17; 95% CI 1.13-4.18), only 5 extra LQTS occurred in hydroxychloroquine group over the next 5 years (years 6-10) and 1 over the next 9 years (years 11-19). There was no association with arrhythmia-related hospitalization or all-cause mortality.

CONCLUSIONS

Hydroxychloroquine use had no association with LQTS during the first 2 years after initiation of therapy. There was a higher risk thereafter that became significant after 5 years of therapy. However, the 5-year absolute risk was very low, and the absolute risk difference was even lower. Both risks attenuated during longer follow-up. These findings provide evidence for long-term safety of hydroxychloroquine in patients with RA.

Potential long-term effect of tumor necrosis factor inhibitors on dementia risk: A propensity score matched retrospective cohort study in US veterans

INTRODUCTION

Tumor necrosis factor (TNF) inhibitors are widely used to treat rheumatoid arthritis (RA) and their potential to retard Alzheimer's disease (AD) progression has been reported. However, their long-term effects on the dementia/AD risk remain unknown.

METHODS

A propensity scored matched retrospective cohort study was conducted among 40,207 patients with RA within the US Veterans Affairs health-care system from 2000 to 2020.

RESULTS

A total of 2510 patients with RA prescribed TNF inhibitors were 1:2 matched to control patients. TNF inhibitor use was associated with reduced dementia risk (hazard ratio [HR]: 0.64, 95% confidence interval [CI]: 0.52-0.80), which was consistent as the study period increased from 5 to 20 years after RA diagnosis. TNF inhibitor use also showed a long-term effect in reducing the risk of AD (HR: 0.57, 95% CI: 0.39-0.83) during the 20 years of follow-up.

CONCLUSION

TNF inhibitor use is associated with lower long-term risk of dementia/AD among US veterans with RA.

Phenotypic Presentations of Heart Failure Among Patients With Chronic Inflammatory Diseases

Objective

Characterize incident heart failure (HF) phenotypes among patients with various chronic inflammatory diseases (CIDs).

Background

Several CIDs are associated with increased HF risk, but differences in HF phenotypes across CIDs are incompletely understood. No prior studies to our knowledge have manually adjudicated HF phenotypes across a CID spectrum.

Methods

We screened for patients with—and controls without—CIDs who had possible HF, then hand-adjudicated HF endpoints. Possible HF resulted from a single HF administrative code; HF was deemed definite/probable vs. absent using standardized, validated criteria. We queried adjudicated HF patients' charts to define specific HF phenotypes, then compared clinical, demographic, and HF phenotypic characteristics for HF patients with specific CIDs vs. non-CID controls using Fisher's exact test.

Results

Out of 415 possible HF patients, 192 had definite/probable HF. Significant differences in HF phenotypes existed across CIDs. Isolated right-sided HF was present in 27.8% of patients with SSc and adjudicated HF, which is more than twice as common as it was in any other CID. Left ventricular systolic dysfunction was most common in patients with HIV and lupus (SLE); mean LVEF was 45.0% ± 18.6% for HIV and 41.3% ± 17.1% for SLE, but was 57.7% ± 10.7% for SSc. Those with HIV and multiple CIDs were most likely to have coronary artery disease.

Conclusions

Different CIDs present with different phenotypes of physician-adjudicated HF, potentially reflecting different underlying inflammatory pathophysiologies. Larger studies are needed to confirm these findings, as are mechanistic studies focused on understanding specific immunoregulatory contributors to HF.

Depression readmission risk is elevated in multiple sclerosis compared to other chronic illnesses

OBJECTIVE

Assess readmissions for depression or suicide attempt (SA) after MS admission versus other chronic inflammatory illnesses.

METHODS

This retrospective cohort study identified MS, asthma, rheumatoid arthritis (RA), depression, and SA in the 2013 National Readmissions Database by International Classification of Diseases codes. Index admissions (MS, n = 7698; asthma, n = 93,590; RA, n = 3685) and depression or SA readmission rates were analyzed. Hazard ratios (HRs) estimated 1-year depression/SA readmission hazard, comparing MS to asthma or RA, adjusting for age, sex, psychiatric comorbidity, substance abuse, tobacco use, income, and index hospitalization characteristics.

RESULTS

MS had more baseline depression (24.7%) versus asthma (15.6%) and RA (14.6%). Ninety-day depression readmission rate was higher in MS (0.5%) than asthma (0.3%) and RA (0.03%). Depression readmission HR was higher after MS admission versus asthma (HR = 1.37, 95% confidence interval (CI) = 1.00-1.86, p = 0.0485) and RA (HR = 4.68, 95% CI = 1.60-13.62, p = 0.0047). HR was not different for SA readmission across groups. Depression readmission HR was more than double in MS patients with psychiatric disease or substance abuse versus RA or asthma patients with either comorbidity.

CONCLUSION

Depression readmission risk after MS hospitalization was elevated versus asthma/RA. Substance use and baseline psychiatric comorbidity were more strongly associated with depression readmission in MS patients.

Comparative Risk of Incident Coronary Heart Disease Across Chronic Inflammatory Diseases

Background: Chronic inflammatory diseases (CIDs) are considered risk enhancing factors for coronary heart disease (CHD). However, sparse data exist regarding relative CHD risks across CIDs. Objective: Determine relative differences in CHD risk across multiple CIDs: psoriasis, rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), human immunodeficiency virus (HIV), systemic sclerosis (SSc), and inflammatory bowel disease (IBD). Methods: The cohort included patients with CIDs and controls without CID in an urban medical system from 2000 to 2019. Patients with CIDs were frequency-matched with non-CID controls on demographics, hypertension, and diabetes. CHD was defined as myocardial infarction (MI), ischemic heart disease, and/or coronary revascularization based on validated administrative codes. Multivariable-adjusted Cox models were used to determine the risk of incident CHD and MI for each CID relative to non-CID controls. In secondary analyses, we compared CHD risk by disease severity within each CID. Results: Of 17,049 patients included for analysis, 619 had incident CHD (202 MI) over an average of 4.4 years of follow-up. The multivariable-adjusted risk of CHD was significantly higher for SLE [hazard ratio (HR) 1.9, 95% confidence interval (CI) 1.2, 3.2] and SSc (HR 2.1, 95% CI 1.2, 3.9). Patients with SLE also had a significantly higher risk of MI (HR 3.6, 95% CI 1.9, 6.8). When CIDs were categorized by markers of disease severity (C-reactive protein for all CIDs except HIV, for which CD4 T cell count was used), greater disease severity was associated with higher CHD risk across CIDs. Conclusions: Patients with SLE and SSc have a higher risk of CHD. CHD risk with HIV, RA, psoriasis, and IBD may only be elevated in those with greater disease severity. Clinicians should personalize CHD risk and treatment based on type and severity of CID.

Validation of algorithms for selecting rheumatoid arthritis patients in the Tuscan healthcare administrative databases

Validation of algorithms for selecting patients from healthcare administrative databases (HAD) is recommended. This PATHFINDER study section is aimed at testing algorithms to select rheumatoid arthritis (RA) patients from Tuscan HAD (THAD) and assessing RA diagnosis time interval between the medical chart date and that of THAD. A population was extracted from THAD. The information of the medical charts at the Rheumatology Unit of Pisa University Hospital represented the reference. We included first ever users of biologic disease modifying anti-rheumatic drugs (bDMARDs) between 2014 and 2016 (index date) with at least a specialist visit at the Rheumatology Unit of the Pisa University Hospital recorded from 2013 to the index date. Out of these, we tested four index tests (algorithms): (1) RA according to hospital discharge records or emergency department admissions (ICD-9 code, 714*); (2) RA according to exemption code from co-payment (006); (3) RA according to hospital discharge records or emergency department admissions AND RA according to exemption code from co-payment; (4) RA according to hospital discharge records or emergency department admissions OR RA according to exemption code from co-payment. We estimated sensitivity, specificity, positive and negative predicted values (PPV and NPV) with 95% confidence interval (95% CI) and the RA diagnosis median time interval (interquartile range, IQR). Two sensitivity analyses were performed. Among 277 reference patients, 103 had RA. The fourth algorithm identified 96 true RA patients, PPV 0.78 (95% CI 0.70-0.85), sensitivity 0.93 (95% CI 0.86-0.97), specificity 0.84 (95% CI 0.78-0.90), and NPV 0.95 (95% CI 0.91-0.98). The sensitivity analyses confirmed performance. The time measured between the actual RA diagnosis date recorded in medical charts and that assumed in THAD was 2.2 years (IQR 0.5-8.4). In conclusion, this validation showed the fourth algorithm as the best. The time interval elapsed between the actual RA diagnosis date in medical charts and that extrapolated from THAD has to be considered in the design of future studies.

Accuracy of US Administrative Claims Codes for the Diagnosis of Autoinflammatory Syndromes

OBJECTIVE

To determine the accuracy of case definitions for autoinflammatory syndromes (AISs) based on administrative claims codes compared with rheumatology records in the electronic medical record (EMR).

METHODS

An AIS screening filter of administrative codes was applied to a large tertiary care EMR database to extract all possible AIS cases. We manually chart reviewed all patients who were evaluated by a rheumatologist to determine their reference standard diagnosis of adult onset Still's disease (AOSD), Behçet's disease (BD), and familial Mediterranean fever (FMF). We calculated sensitivity, specificity, positive predictive values, negative predictive values, and area under the receiver operating characteristic curve of specific codes for diagnosing AIS subtypes.

RESULTS

We identified 273 individuals with possible AIS, of which 72 (26.4%) had a true AIS diagnosis, including 24 with AOSD, 32 with BD, and 9 with FMF. For all 3 AIS subtypes, the estimates of specificities and negative predictive values for specific administrative codes were excellent (>95%). Sensitivity estimates were excellent (>89%) for BD and FMF codes and lower for AOSD (46%-50%). Positive predictive values were excellent for BD (>99%) and AOSD (>86%) and lower for FMF (>53%). Area under the receiver operating characteristic curve estimates were excellent for BD (97%-98%) and FMF (93%) and very good for AOSD (75%).

CONCLUSIONS

This is the first study to characterize the accuracy of specific administrative codes for the diagnosis of AOSD, BD, and FMF in a large tertiary care EMR. Validation in external EMRs and linked EMR-administrative databases is needed to enable future clinical outcomes research of AIS.

Trends of lymphoma incidence in US veterans with rheumatoid arthritis, 2002-2017

OBJECTIVE

Past epidemiological studies have consistently demonstrated a link between rheumatoid arthritis (RA) and the incidence of lymphoma and it has been posited that high systemic inflammatory activity is a major risk determinant of lymphomagenesis. Given advances in the therapeutic armamentarium for RA management in recent years, the resulting lower level of disease activity could have led to a decline in lymphoma incidence in patients with RA. This study examined recent trends in lymphoma incidence in US veterans with RA.

METHODS

Patients with RA were identified in the Veterans Affairs (VA) Corporate Data Warehouse. Lymphoma incidence was identified through the end of 2018 from the VA Central Cancer Registry and compared among patients diagnosed during 2003-2005, 2006-2008, 2009-2011 and 2012-2014.

RESULTS

Among persons diagnosed with RA during 2003-2005, the incidence of lymphoma in the next 6 years was 2.0 per 1000 person-years. There was a steady decline in lymphoma incidence during the corresponding 6 years following diagnosis in the subsequent three cohorts, with a rate of 1.5 per 1000 person-years in the 2012-2014 cohort (incidence relative to that in the 2003-2005 cohort=0.79 (95% CI 0.58 to 1.1)). There was no similar decline in lymphoma incidence in VA patients diagnosed with osteoarthritis.

CONCLUSION

We observed a decline in lymphoma incidence in recent years among American veterans with RA. Further studies are needed to evaluate the specific factors driving this decline.

Association Between Tumor Necrosis Factor Inhibitor Exposure and Inflammatory Central Nervous System Events

Question

Is exposure to tumor necrosis factor inhibitors associated with risk of inflammatory demyelinating and nondemyelinating central nervous system events in patients with an autoimmune disease?

Findings

In this case-control study of 212 patients with or without inflammatory CNS events, exposure to tumor necrosis factor inhibitors was associated with an increased risk of inflammatory central nervous system events. The association was similar for both inflammatory demyelinating and nondemyelinating central nervous system events.

Meaning

The association observed between exposure to tumor necrosis factor inhibitor and increased risk of inflammatory demyelinating and nondemyelinating central nervous system events warrants future research to ascertain whether the association may indicate de novo inflammation or exacerbation of already aberrant inflammatory pathways.

Importance

Tumor necrosis factor (TNF) inhibitors are common therapies for certain autoimmune diseases, such as rheumatoid arthritis. An association between TNF inhibitor exposure and inflammatory central nervous system (CNS) events has been postulated but is poorly understood.

Objective

To evaluate whether TNF inhibitor exposure is associated with inflammatory demyelinating and nondemyelinating CNS events in patients with an indication for TNF inhibitor use and to describe the spectrum of those CNS events.

Design, Setting, and Participants

A nested case-control study was conducted using the medical records of patients with autoimmune diseases treated at 3 Mayo Clinic locations (Rochester, Minnesota; Scottsdale, Arizona; and Jacksonville, Florida) between January 1, 2003, and February 20, 2019. Patients were included if their records reported International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, diagnostic codes for US Food and Drug Administration–approved autoimmune disease indication for TNF inhibitor use (ie, rheumatoid arthritis, ankylosing spondylitis, psoriasis and psoriatic arthritis, Crohn disease, and ulcerative colitis) and diagnostic codes for inflammatory CNS events of interest. Patients were matched 1:1 with control participants by year of birth, type of autoimmune disease, and sex.

Exposures

TNF inhibitor exposure data were derived from the medical records along with type of TNF inhibitor, cumulative duration of exposure, and time of exposure.

Main Outcomes and Measures

The main outcome was either inflammatory demyelinating (multiple sclerosis and other diseases such as optic neuritis) or nondemyelinating (meningitis, meningoencephalitis, encephalitis, neurosarcoidosis, and CNS vasculitis) CNS event. Association with TNF inhibitor was evaluated with conditional logistic regression and adjusted for disease duration to determine the odds ratios (ORs) and 95% CIs. Secondary analyses included stratification of outcome by inflammatory demyelinating and nondemyelinating CNS events and by autoimmune disease (rheumatoid arthritis and non–rheumatoid arthritis).

Results

A total of 212 individuals were included: 106 patients with inflammatory CNS events and 106 control participants without such events. Of this total, 136 were female (64%); the median (interquartile range) age at disease onset for patients was 52 (43-62) years. Exposure to TNF inhibitors occurred in 64 patients (60%) and 42 control participants (40%) and was associated with an increased risk of any inflammatory CNS event (adjusted OR, 3.01; 95% CI, 1.55-5.82; P = .001). These results were similar when the outcome was stratified by demyelinating and nondemyelinating CNS events. Secondary analyses found the association was predominantly observed in patients with rheumatoid arthritis (adjusted OR, 4.82; 95% CI, 1.62-14.36; P = .005).

Conclusions and Relevance

This study found that exposure to TNF inhibitors in patients with autoimmune diseases appeared to be associated with increased risk for inflammatory CNS events. Whether this association represents de novo or exacerbated inflammatory pathways requires further research.

Adverse drug event rates in pediatric pulmonary hypertension: a comparison of real-world data sources

Objective

Real-world data (RWD) are increasingly used for pharmacoepidemiology and regulatory innovation. Our objective was to compare adverse drug event (ADE) rates determined from two RWD sources, electronic health records and administrative claims data, among children treated with drugs for pulmonary hypertension.

Materials and Methods

Textual mentions of medications and signs/symptoms that may represent ADEs were identified in clinical notes using natural language processing. Diagnostic codes for the same signs/symptoms were identified in our electronic data warehouse for the patients with textual evidence of taking pulmonary hypertension-targeted drugs. We compared rates of ADEs identified in clinical notes to those identified from diagnostic code data. In addition, we compared putative ADE rates from clinical notes to those from a healthcare claims dataset from a large, national insurer.

Results

Analysis of clinical notes identified up to 7-fold higher ADE rates than those ascertained from diagnostic codes. However, certain ADEs (eg, hearing loss) were more often identified in diagnostic code data. Similar results were found when ADE rates ascertained from clinical notes and national claims data were compared.

Discussion

While administrative claims and clinical notes are both increasingly used for RWD-based pharmacovigilance, ADE rates substantially differ depending on data source.

Conclusion

Pharmacovigilance based on RWD may lead to discrepant results depending on the data source analyzed. Further work is needed to confirm the validity of identified ADEs, to distinguish them from disease effects, and to understand tradeoffs in sensitivity and specificity between data sources.

Epidemiology of cardiac or orthopedic procedures in gout versus rheumatoid arthritis: a national time-trends study

Aims

To examine the secular trends in the number and rates of in-hospital cardiac and orthopedic procedures in people with gout and rheumatoid arthritis (RA), and the United States (US) general population, from 1998 to 2014.

Methods

We examined the frequency of seven common cardiac and orthopedic procedures in hospitalized people with gout, RA, or the general population using the 1998-2014 US National Inpatient Sample (NIS). Poisson regression evaluated the differences in frequencies in 1998 versus 2014, between gout and RA, and within each cohort.

Results

Both in-hospital cardiac and orthopedic procedures increased in gout and RA with time, in contrast with declining cardiac procedures in the general US population. Cardiac procedures were significantly higher in gout versus RA in 1998 (59% higher) and 2014 (92% higher). The rate of cardiac procedures increased from 36.6 to 82.8 in gout and from 20.1 to 33.1 in RA per 100,000 NIS claims from 1998 to 2014. Orthopedic procedures became more common than cardiac procedures in gout and RA by 2014. In RA, the cardiac-orthopedic procedure volume difference was significant in 1998 and 2014. We noted no significant difference between cardiac versus orthopedic procedures in 1998 in gout, but the difference was significant in 2014. Orthopedic procedures in gout were significantly lower than RA in 1998 (33% lower), but were significantly higher than RA in 2014 (5% higher).

Conclusion

Increasing in-hospital cardiac procedures in gout and RA contrasting with declining general US population rates indicated that optimal management of systemic inflammation and an early diagnosis of gout and RA are needed. The rate of increase in orthopedic procedures exceeded that in cardiac procedures. A much greater volume and rate of increase in common in-hospital cardiac and orthopedic procedures in gout compared to RA indicates that an aggressive approach to treat-to-target in gout is needed to potentially reduce the associated healthcare burden and cost.

Plain language summary

Cardiac and orthopedic procedures rising faster for gout compared to rheumatoid arthritis in the United States We performed a national US study of the most common cardiac versus orthopedic procedures from 1998 to 2014. We found that over time, the number and the rate of cardiac procedures increased in people with gout (2.2-fold higher) or rheumatoid arthritis (1.6-fold higher). This was surprising, since during the same time, we noted a decrease in cardiac procedures in the general U.S. population. The rate of cardiac procedures in gout was 2.5-fold higher than that in rheumatoid arthritis, 82.8 vs. 33.1 per 100,000 NIS claims in 2014. Interestingly, orthopedic procedures were more common than cardiac procedures in both gout and RA in all periods. Also, the difference in the numbers of cardiac vs. orthopedic procedures increased over time in both gout and RA. Gout outpaced rheumatoid arthritis for both the total number and the rate of cardiac or orthopedic procedures over time. Therefore, our study provides an understanding of an increasing procedure burden in gout compared to rheumatoid arthritis, and to the general U.S. people with these conditions.

Impact of ICD10 and secular changes on electronic medical record rheumatoid arthritis algorithms

OBJECTIVE

The objective of this study was to compare the performance of an RA algorithm developed and trained in 2010 utilizing natural language processing and machine learning, using updated data containing ICD10, new RA treatments, and a new electronic medical records (EMR) system.

METHODS

We extracted data from subjects with ≥1 RA International Classification of Diseases (ICD) codes from the EMR of two large academic centres to create a data mart. Gold standard RA cases were identified from reviewing a random 200 subjects from the data mart, and a random 100 subjects who only have RA ICD10 codes. We compared the performance of the following algorithms using the original 2010 data with updated data: (i) a published 2010 RA algorithm; (ii) updated algorithm, incorporating ICD10 RA codes and new DMARDs; and (iii) published algorithm using ICD codes only, ICD RA code ≥3.

RESULTS

The gold standard RA cases had mean age 65.5 years, 78.7% female, 74.1% RF or antibodies to cyclic citrullinated peptide (anti-CCP) positive. The positive predictive value (PPV) for ≥3 RA ICD was 54%, compared with 56% in 2010. At a specificity of 95%, the PPV of the 2010 algorithm and the updated version were both 91%, compared with 94% (95% CI: 91, 96%) in 2010. In subjects with ICD10 data only, the PPV for the updated 2010 RA algorithm was 93%.

CONCLUSION

The 2010 RA algorithm validated with the updated data with similar performance characteristics as the 2010 data. While the 2010 algorithm continued to perform better than the rule-based approach, the PPV of the latter also remained stable over time.

Development and validation of lupus nephritis case definitions using United States veterans affairs electronic health records

OBJECTIVE

International Classification of Diseases (ICD) codes are commonly used to identify patients with rare diseases in electronic health records (EHRs). However, misclassification is common, impacting the validity of study results. In this study, we compared the accuracies of several ICD-based case definitions of lupus nephritis (LN) in identifying United States veterans with LN.

METHODS

Using the Department of Veterans Affairs (VA) EHR, we identified all veterans with ≥1 ICD-9 or 10 diagnostic codes for systemic lupus erythematosus (SLE) between October 1, 1999 and September 30, 2017. A cohort was randomly selected for diagnostic validation and 9 ICD-based LN case definitions were applied to this cohort. The diagnostic accuracy of each definition was assessed against gold standard criterion of biopsy-proven LN.

RESULTS

18,420 veterans had ≥1 ICD-9 or 10 diagnostic codes for SLE; 981 were randomly selected for diagnostic validation. 95 veterans (9.7%) had biopsy-proven LN. The case definitions had high specificity and NPV but variable sensitivity and PPV. The definition containing ≥2 ICD -9 codes for SLE and ≥2 nephritis indicators had the highest combination of sensitivity and specificity (87.4% and 94.6% respectively). ICD-10 code for LN had high specificity (99.8%) and PPV (93.9%).

CONCLUSION

ICD-based case definitions of LN in the VA population have high specificity and NPV but variable sensitivity and PPV. Our results may help guide the design of future LN studies in VA cohorts. The choice of specific case definitions depends on the relative importance of different accuracy measures to individual studies.

Posttraumatic Stress Disorder, Depression, Anxiety, and Persistence of Methotrexate and TNF Inhibitors in Patients with Rheumatoid Arthritis

OBJECTIVE

To examine the relationship of posttraumatic stress disorder (PTSD) with earlier treatment discontinuation and medication adherence in US veterans with rheumatoid arthritis (RA).

METHODS

Veterans Affairs (VA) administrative data (2005-2014) were used to define unique dispensing episodes of methotrexate (MTX) and tumor necrosis factor inhibitors (TNFi) for veterans with RA. Diagnosis codes were used to categorize patients into mutually exclusive groups: PTSD (with/without depression/anxiety), depression/anxiety without PTSD, and neither psychiatric diagnosis. Multivariable Cox proportional hazards models were used to evaluate associations between psychiatric diagnoses and time to disease-modifying antirheumatic drug discontinuation (lapse in refill >90 days). Multivariable logistic regression was used to examine associations of diagnoses with medication nonadherence (proportion of days covered <0.8).

RESULTS

There were 15 081 dispensing episodes of MTX and 8412 dispensing episodes of TNFi. PTSD was independently associated with a greater likelihood of earlier discontinuation of both MTX (hazard ratio [HR] 1.15 [1.10-1.21]) and TNFi (HR 1.20 [1.13-1.28]). Depression/anxiety had a comparable risk of discontinuation for both MTX (HR 1.14 [1.10-1.19]) and TNFi (HR 1.16 [1.10-1.22]). Depression/anxiety, but not PTSD, was associated with higher odds of MTX (odds ratio [OR] 1.12 [1.03-1.22]) and TNFi (OR 1.14 [1.02-1.27]) nonadherence.

CONCLUSION

Veterans with RA and comorbid PTSD, depression, or anxiety had poor persistence of MTX and TNFi therapies. These results suggest that earlier discontinuation and low adherence to therapy among patients with RA with these psychiatric comorbidities may contribute to worse disease outcomes. Mechanisms by which these comorbidities contribute to lower adherence deserve further investigation and may lead to targeted interventions to improve disease outcomes.

Mental health conditions and the risk of chronic opioid therapy among patients with rheumatoid arthritis: a retrospective veterans affairs cohort study

OBJECTIVE

Patients with rheumatoid arthritis (RA) often receive opioid analgesics for pain management. We examined the association between mental health conditions and the risk of chronic opioid therapy.

METHODS

A retrospective cohort of veterans with RA initiating opioid use was assembled using Veterans Health Administration databases (2001-2012). Mental health conditions included anxiety (N = 1108, 12.9%), depression (N = 1912, 22.2%), bipolar disease (N = 131, 1.5%), and post-traumatic stress disorder (N = 768, 8.9%) and were identified by ICD coded diagnoses and use of specific medications. Cohort members were followed from opioid initiation through chronic opioid therapy, defined as the continuous availability of opioids for at least 90 days. Multivariable Cox proportional hazard regression models assessed the association between mental health conditions and chronic opioid therapy accounting for relevant covariates. Subgroup analyses examined whether the strength of the observed association varied by the duration of the initial opioid prescription.

RESULTS

We identified 14,767 patients with RA with 22,452 episodes of opioid use initiation. Mental health conditions were identified in 8607 (38.3%) patients. Compared with patients without mental health conditions, patients with mental health conditions have a higher risk of developing chronic opioid therapy (469.3 vs 378.1 per 1000 person-years, adjusted hazard ratio [aHR] 1.18, 95% CI 1.09, 1.29). The increased risk was highest for those with a history of opioid use disorder (aHR 1.94, 95% CI 1.09, 3.46) and also elevated for those with other substance use disorders (aHR 1.35, 95% CI 1.05, 1.73). Duration of the initial opioid prescription was independently associated with chronic opioid therapy, regardless of the estimated opioid daily dose.

CONCLUSIONS

History of mental health conditions and duration of the initial opioid prescription were associated with an increased risk of chronic opioid therapy among patients with RA.Key Points• Approximately a third of patients with RA are exposed to opioid analgesics.• Patients with RA and history of mental health disease, especially substance use disorders, who initiate opioid use have an increased risk of chronic opioid therapy.• This study provides insight in an underrepresented population of mainly male patients with RA.

Using genetics to prioritize diagnoses for rheumatology outpatients with inflammatory arthritis

It is challenging to quickly diagnose slowly progressing diseases. To prioritize multiple related diagnoses, we developed G-PROB (Genetic Probability tool) to calculate the probability of different diseases for a patient using genetic risk scores. We tested G-PROB for inflammatory arthritis-causing diseases (rheumatoid arthritis, systemic lupus erythematosus, spondyloarthropathy, psoriatic arthritis, and gout). After validating on simulated data, we tested G-PROB in three cohorts: 1211 patients identified by International Classification of Diseases (ICD) codes within the eMERGE database, 245 patients identified through ICD codes and medical record review within the Partners Biobank, and 243 patients first presenting with unexplained inflammatory arthritis and with final diagnoses by record review within the Partners Biobank. Calibration of G-probabilities with disease status was high, with regression coefficients from 0.90 to 1.08 (1.00 is ideal). G-probabilities discriminated true diagnoses across the three cohorts with pooled areas under the curve (95% CI) of 0.69 (0.67 to 0.71), 0.81 (0.76 to 0.84), and 0.84 (0.81 to 0.86), respectively. For all patients, at least one disease could be ruled out, and in 45% of patients, a likely diagnosis was identified with a 64% positive predictive value. In 35% of cases, the clinician's initial diagnosis was incorrect. Initial clinical diagnosis explained 39% of the variance in final disease, which improved to 51% (P < 0.0001) after adding G-probabilities. Converting genotype information before a clinical visit into an interpretable probability value for five different inflammatory arthritides could potentially be used to improve the diagnostic efficiency of rheumatic diseases in clinical practice.

Screening of Hyperlipidemia Among Patients With Rheumatoid Arthritis in the United States

OBJECTIVE

To determine the proportion of primary lipid screening among patients with rheumatoid arthritis (RA) and compare it with those among patients with diabetes mellitus (DM) and patients with neither RA nor DM, and to assess whether primary lipid screening varied according to the health care provider (rheumatologist versus non-rheumatologist).

METHODS

We analyzed claims data from US private and public health plans from 2006-2010. Eligibility requirements included continuous medical and pharmacy coverage for ≥12 months (baseline period) and >2 physician diagnoses and relevant medications to define RA, DM, RA and DM, or neither condition. Among the 330,695 eligible participants, we calculated the proportion with a lipid profile ordered during the 2 years following baseline. Time-varying Cox proportional hazard models were used to determine the probability of hyperlipidemia screening in participants with RA according to provider specialty.

RESULTS

More than half of the patients were ages 41-71 years. Among patients with RA (n = 12,182), DM (n = 62,834), RA and DM (n = 1,082), and those who did not have either condition (n = 167,811), the proportion screened for hyperlipidemia was 37%, 60%, 55%, and 41%, respectively. Patients with RA who visited a rheumatologist and a non-rheumatology clinician during follow-up had a 55% (95% confidence interval 1.36-1.78) higher screening probability than those who only visited a rheumatologist.

CONCLUSION

Primary lipid screening was suboptimal among patients with RA. It was also lower for patients with DM and minimally different from the general population. Screening was higher for RA patients who received care from both a rheumatologist and a non-rheumatologist (e.g., primary care physician).

Differential Associations of Chronic Inflammatory Diseases With Incident Heart Failure

OBJECTIVES

The purpose of this study was to compare the risks of incident heart failure (HF) among a variety of chronic inflammatory diseases (CIDs) and to determine whether risks varied by severity of inflammation within each CID.

BACKGROUND

Individuals with CIDs are at elevated risk for cardiovascular diseases, but data are limited regarding risk for HF.

METHODS

An electronic health records database from a large urban medical system was examined, comparing individuals with CIDs with frequency-matched controls without CIDs, all of whom were receiving regular outpatient care. Rates of incident HF were determined by using the Kaplan-Meier method and subsequently used multivariate-adjusted proportional hazards models to compare HF risks for each CID. Exploratory analyses determined HF risks by proxy measurement of CID severity.

RESULTS

Of 37,636 patients (n = 18,278 patients with CIDs; and n = 19,358 controls without CIDs) there were 960 incident HF cases over a median of 3.6 years. Risks for incident HF were significantly or borderline significantly elevated for patients with systemic sclerosis (hazard ratio [HR]: 7.26; 95% confidence interval [CI]: 5.72 to 9.21; p < 0.01), systemic lupus erythematosus (HR: 3.15; 95% CI: 2.41 to 4.11; p < 0.01), rheumatoid arthritis (HR: 1.39; 95% CI: 1.13 to 1.71; p < 0.01), and human immunodeficiency virus (HR: 1.28; 95% CI: 0.99 to 1.66; p = 0.06). There was no association between psoriasis or inflammatory bowel disease and incident HF, although patients with those CIDs with higher levels of C-reactive protein had higher risks for HF than controls.

CONCLUSIONS

Systemic sclerosis and systemic lupus erythematosus were associated with the highest risks of HF, followed by rheumatoid arthritis and HIV. Measurements of inflammation were associated with HF risk across different CIDs.

Characterizing Health Outcomes in Idiopathic Pulmonary Fibrosis using US Health Claims Data

BACKGROUND

Idiopathic pulmonary fibrosis (IPF) is a life-threatening interstitial lung disease (ILD). Characterizing health outcomes of IPF patients is challenging due to disease rarity.

OBJECTIVE

This study aimed to identify the burden of disease in patients newly diagnosed with IPF.

METHODS

Patients with ≥1 claim with an IPF diagnosis were identified from a United States healthcare insurer's database (2000-2013). Patients with other known causes of ILD or aged <40 years were excluded. Subgroups were compared based on the 2011 change in International Classification of Diseases, 9th Revision (ICD-9) definition of IPF and occurrence of IPF testing. The prevalence and incidence of preselected health conditions of clinical interest were estimated.

RESULTS

Median age of newly diagnosed patients (n = 7,298) was 62 years (54.0% male). Restricting to patients with IPF diagnostic testing did not substantially affect cohort characteristics, nor did ICD-9 IPF coding change. Mean follow-up was 1.7 years; 16.8% of patients died; and a substantial proportion of patients were censored due to end of health plan enrollment (50.7%) and other causes of ILD (19.6%). The incidence of pulmonary hypertension, lung cancer, and claims-based algorithm proxy for acute respiratory worsening of unknown cause was 22.5, 17.6, and 12.6 per 1,000 person-years, respectively.

CONCLUSIONS

Patients with IPF had a high disease burden with a variety of health outcomes observed, including a high rate of mortality. Database censoring due to changes in enrollment or other ILD diagnoses limited follow-up. Altering cohort entry definitions, including IPF testing or ICD-9 IPF coding change, had little impact on cohort baseline characteristics.

Improving accuracy of self-reported diagnoses of rheumatoid arthritis in the French prospective E3N-EPIC cohort: a validation study

OBJECTIVES

The French E3N-EPIC (Etude Epidémiologique auprès des femmes de la Mutuelle générale de l'Education Nationale-European Prospective Investigation into Cancer and Nutrition) cohort enrolled 98 995 women aged 40 to 65 years at inclusion since 1990 to study the main risk factors for cancer and severe chronic conditions in women. They were prospectively followed with biennially self-administered questionnaires collecting self-reported medical, environmental and lifestyle data. Our objective was to assess the accuracy of self-reported diagnoses of rheumatoid arthritis (RA) and to devise algorithms to improve the ascertainment of RA cases in our cohort.

DESIGN

A validation study.

PARTICIPANTS

Women who self-reported an inflammatory rheumatic disease (IRD) were asked to provide access to their medical record, and to answer an IRD questionnaire. Medical records were independently reviewed.

PRIMARY AND SECONDARY OUTCOME MEASURES

Positive predictive values (PPV) of self-reported RA alone, then coupled with the IRD questionnaire, and with a medication reimbursement database were assessed. These algorithms were then applied to the whole cohort to ascertain RA cases.

RESULTS

Of the 98 995 participants, 2692 self-reported RA. Medical records were available for a sample of 399 participants, including 305 who self-reported RA. Self-reported RA was accurate only for 42% participants. Combining self-reported diagnoses to answers to a specific IRD questionnaire or to the medication reimbursement database improved the PPV (75.6% and 90.1%, respectively). Using the devised algorithms, we could identify 964 RA cases in our cohort.

CONCLUSION

Accuracy of self-reported RA is poor but adding answers to a specific questionnaire or data from a medication reimbursement database performed satisfactorily to identify RA cases in our cohort. It will subsequently allow investigating many potential risk factors of RA in women.

Risk of Recurrent Staphylococcus aureus Prosthetic Joint Infection in Rheumatoid Arthritis Patients-A Nationwide Cohort Study

Background

Treatment of rheumatoid arthritis (RA) often involves immune-suppressive therapies. Concern for recurrent prosthetic joint infection (PJI) in RA patients might be high and could reduce use of joint implantation in these patients. We aimed to evaluate the risk of recurrence of PJI in RA patients compared with osteoarthritis (OA) patients by utilizing a large health care system.

Methods

We conducted a retrospective cohort study of all patients admitted for a Staphylococcus aureus PJI who underwent debridement, antibiotics, and implant retention (DAIR) or 2-stage exchange (2SE) between 2003 and 2010 at 86 Veterans Affairs Medical Centers. Both RA patients and the comparison group of osteoarthritis (OA) patients were identified using International Classification of Diseases, Ninth Revision, codes. All index PJI and recurrent positive cultures for S. aureus during 2 years of follow-up were validated by manual chart review. A Cox proportional hazards regression model was used to compare the time to recurrent PJI for RA vs OA.

Results

In our final cohort of 374 veterans who had either DAIR or 2SE surgery for their index S. aureus PJI, 11.2% had RA (n = 42). The majority of the cohort was male (97.3%), and 223 (59.6%) had a methicillin-susceptible S. aureus PJI. RA patients had a similar risk of failure compared with OA patients, after adjusting for covariates (hazard ratio, 0.81; 95% confidence interval, 0.48–1.37).

Conclusions

Prior diagnosis of RA does not increase the risk of recurrent S. aureus PJI. Further studies are needed to evaluate the effect of different RA therapies on outcomes of episodes of PJI.

Initiation of Disease-Modifying Therapies in Rheumatoid Arthritis Is Associated With Changes in Blood Pressure

PURPOSE

This study reports the effect of disease-modifying therapies for rheumatoid arthritis (RA) on systolic and diastolic blood pressure (SBP, DBP) over 6 months and incident hypertension over 3 years in a large administrative database.

METHODS

We used administrative Veterans Affairs databases to define unique dispensing episodes of methotrexate, leflunomide, sulfasalazine, hydroxychloroquine, tumor necrosis factor inhibitors, and prednisone among patients with RA. Changes in SBP and DBP in the 6 months before disease-modifying antirheumatic drug initiation were compared with changes observed in the 6 months after initiation. The risk of incident hypertension within 3 years (new diagnosis code for hypertension and prescription for antihypertensive) was also assessed. Multivariable models and propensity analyses assessed the impact of confounding by indication.

RESULTS

A total of 37,900 treatment courses in 21,216 unique patients contributed data. Overall, there were no changes in SBP or DBP in 6 months prior to disease-modifying antirheumatic drug initiation (all P > 0.62). In contrast, there was a decline in SBP (β = -1.08 [-1.32 to -0.85]; P < 0.0001) and DBP (β = -0.48 [-0.62 to -0.33]; P < 0.0001) over the 6 months following initiation. The greatest decline was observed among methotrexate and hydroxychloroquine users. Methotrexate users were 9% more likely to have optimal blood pressure (BP) after 6 months of treatment. Patients treated with leflunomide had increases in BP and a greater risk of incident hypertension compared with patients treated with methotrexate (hazard ratio, 1.53 [1.21-1.91]; P < 0.001).

CONCLUSIONS

Blood pressure may improve with treatment of RA, particularly with methotrexate or hydroxychloroquine. Leflunomide use, in contrast, is associated with increases in BP and a greater risk of incident hypertension.

Identifying Breast Cancer Distant Recurrences from Electronic Health Records Using Machine Learning

Accurately identifying distant recurrences in breast cancer from the Electronic Health Records (EHR) is important for both clinical care and secondary analysis. Although multiple applications have been developed for computational phenotyping in breast cancer, distant recurrence identification still relies heavily on manual chart review. In this study, we aim to develop a model that identifies distant recurrences in breast cancer using clinical narratives and structured data from EHR. We applied MetaMap to extract features from clinical narratives and also retrieved structured clinical data from EHR. Using these features, we trained a support vector machine model to identify distant recurrences in breast cancer patients. We trained the model using 1,396 double-annotated subjects and validated the model using 599 double-annotated subjects. In addition, we validated the model on a set of 4,904 single-annotated subjects as a generalization test. In the held-out test and generalization test, we obtained F-measure scores of 0.78 and 0.74, area under curve (AUC) scores of 0.95 and 0.93, respectively. To explore the representation learning utility of deep neural networks, we designed multiple convolutional neural networks and multilayer neural networks to identify distant recurrences. Using the same test set and generalizability test set, we obtained F-measure scores of 0.79 ± 0.02 and 0.74 ± 0.004, AUC scores of 0.95 ± 0.002 and 0.95 ± 0.01, respectively. Our model can accurately and efficiently identify distant recurrences in breast cancer by combining features extracted from unstructured clinical narratives and structured clinical data.

Natural Language Processing for EHR-Based Computational Phenotyping

This article reviews recent advances in applying natural language processing (NLP) to Electronic Health Records (EHRs) for computational phenotyping. NLP-based computational phenotyping has numerous applications including diagnosis categorization, novel phenotype discovery, clinical trial screening, pharmacogenomics, drug-drug interaction (DDI), and adverse drug event (ADE) detection, as well as genome-wide and phenome-wide association studies. Significant progress has been made in algorithm development and resource construction for computational phenotyping. Among the surveyed methods, well-designed keyword search and rule-based systems often achieve good performance. However, the construction of keyword and rule lists requires significant manual effort, which is difficult to scale. Supervised machine learning models have been favored because they are capable of acquiring both classification patterns and structures from data. Recently, deep learning and unsupervised learning have received growing attention, with the former favored for its performance and the latter for its ability to find novel phenotypes. Integrating heterogeneous data sources have become increasingly important and have shown promise in improving model performance. Often, better performance is achieved by combining multiple modalities of information. Despite these many advances, challenges and opportunities remain for NLP-based computational phenotyping, including better model interpretability and generalizability, and proper characterization of feature relations in clinical narratives.

eDRAM: Effective early disease risk assessment with matrix factorization on a large-scale medical database: A case study on rheumatoid arthritis

Recently, a number of analytical approaches for probing medical databases have been developed to assist in disease risk assessment and to determine the association of a clinical condition with others, so that better and intelligent healthcare can be provided. The early assessment of disease risk is an emerging topic in medical informatics. If diseases are detected at an early stage, prognosis can be improved and medical resources can be used more efficiently. For example, if rheumatoid arthritis (RA) is detected at an early stage, appropriate medications can be used to prevent bone deterioration. In early disease risk assessment, finding important risk factors from large-scale medical databases and performing individual disease risk assessment have been challenging tasks. A number of recent studies have considered risk factor analysis approaches, such as association rule mining, sequential rule mining, regression, and expert advice. In this study, to improve disease risk assessment, machine learning and matrix factorization techniques were integrated to discover important and implicit risk factors. A novel framework is proposed that can effectively assess early disease risks, and RA is used as a case study. This framework comprises three main stages: data preprocessing, risk factor optimization, and early disease risk assessment. This is the first study integrating matrix factorization and machine learning for disease risk assessment that is applied to a nation-wide and longitudinal medical diagnostic database. In the experimental evaluations, a cohort established from a large-scale medical database was used that included 1007 RA-diagnosed patients and 921,192 control patients examined over a nine-year follow-up period (2000-2008). The evaluation results demonstrate that the proposed approach is more efficient and stable for disease risk assessment than state-of-the-art methods.

Positive predictive value of first-time rheumatoid arthritis diagnoses and their serological subtypes in the Danish National Patient Registry

Purposes

To assess whether the positive predictive value (PPV) of first-time rheumatoid arthritis (RA) diagnosis registration in the Danish National Patient Registry increases when data are linked to the RA treatment codes and to assess the PPV of first-time RA diagnoses according to RA serological subtypes.

Methods

Participants from the Danish Diet, Cancer, and Health cohort with at least one RA diagnosis, registered at one of the Central Denmark Region hospitals in the Danish National Patient Registry during the period 1977–2016, were identified. Register-based RA diagnoses were verified by scrutinizing medical records against RA classification criteria or clinical case RA. PPVs for overall RA, seropositive RA, and other RA were calculated for two models: first-time RA diagnosis registration ever in the Danish National Patient Registry and first-time RA diagnosis registration ever where subsequently a prescription had been redeemed for a synthetic disease-modifying antirheumatic drug.

Results

Overall, 205 of 311 first-time register-based RA diagnoses were verified (PPV: 61.9%; 95% CI: 56.9–67.0). Regarding RA serological subtypes, 93 of 150 register-based seropositive RA (PPV: 62.0; 95% CI: 53.9–69.5) and 36 of 144 other RA (PPV: 25.0; 95% CI: 18.5–32.8) were confirmed. When register-based RA diagnosis codes were linked to RA treatment codes, the PPVs increased substantially: the PPV for overall RA was 87.7% (95% CI: 82.5–91.5), the PPV for seropositive RA was 80.2% (95% CI: 71.6–86.7), and the PPV for other RA was 41.1% (95% CI: 30.2–52.9).

Conclusion

The first-time RA diagnoses in the Danish National Patient Registry should be used with caution in epidemiology research. However, linking registry-based RA diagnoses to the subsequent RA treatment codes increases the probability of identifying true RA diagnoses, especially overall RA and seropositive RA.

xCT (SLC7A11)-mediated metabolic reprogramming promotes non-small cell lung cancer progression

Many tumors increase uptake and dependence on glucose, cystine or glutamine. These basic observations on cancer cell metabolism have opened multiple new diagnostic and therapeutic avenues in cancer research. Recent studies demonstrated that smoking could induce the expression of xCT (SLC7A11) in oral cancer cells, suggesting that overexpression of xCT may support lung tumor progression. We hypothesized that overexpression of xCT occurs in lung cancer cells to satisfy the metabolic requirements for growth and survival. Our results demonstrated that 1) xCT was highly expressed at the cytoplasmic membrane in non-small cell lung cancer (NSCLC), 2) the expression of xCT was correlated with advanced stage and predicted a worse 5-year survival, 3) targeting xCT transport activity in xCT overexpressing NSCLC cells with sulfasalazine decreased cell proliferation and invasion in vitro and in vivo and 4) increased dependence on glutamine was observed in xCT overexpressed normal airway epithelial cells. These results suggested that xCT regulate metabolic requirements during lung cancer progression and be a potential therapeutic target in NSCLC.

Treatment of rheumatoid arthritis with biologic agents lowers the risk of incident chronic kidney disease

Rheumatoid arthritis is associated with reduced kidney function, possibly due to chronic inflammation or the use of nephrotoxic therapies. However, little is known about the effects of using the newer novel non-nephrotoxic biologic agents on the risk of incident chronic kidney disease (CKD). To study this we used a cohort of 20,757 United States veterans diagnosed with rheumatoid arthritis with an estimated glomerular filtration rate (eGFR) of 60 mL/min/1.73m² or more, recruited between October 2004 and September 2006, and followed through 2013. The associations of biologic use with incident CKD (eGFR under 60 with a decrease of at least 25% from baseline, and eGFR under 45 mL/min/1.73m²) and change in eGFR (<-3, -3 to <0 [reference], and ≥0 mL/min/1.73m²/year) were examined in propensity-matched patients based on their likelihood to initiate biologic treatment, using Cox models and multinomial logistic regression models, respectively. Among 20,757 patients, 4,617 started biologic therapy. In the propensity-matched cohort, patients treated (versus not treated) with biologic agents had a lower risk of incident CKD (hazard ratios 0.95, 95% confidence interval [0.82-1.10] and 0.71 [0.53-0.94] for decrease in eGFR under 60 and under 45 mL/min/1.73m², respectively) and progressive eGFR decline (multinomial odds ratios [95% CI] for eGFR slopes <-3 and ≥0 [versus -3 to <0] mL/min/1.73m²/year, 0.67 [0.58-0.79] and 0.76 [0.69-0.83], respectively). A significant deceleration of eGFR decline was also observed after biologic administration in patients treated with biologics (-1.0 versus -0.4 [mL/min/1.73m²/year] before and after biologic use). Thus, biologic agent administration was independently associated with lower risk of incident CKD and progressive eGFR decline.

DPP-4 Inhibitor-Induced Rheumatoid Arthritis Among Diabetics: A Nested Case-Control Study

INTRODUCTION

The risk of rheumatoid arthritis (RA) associated with dipeptidyl peptidase-4 inhibitor (DPP-4i) use is unclear. This study assesses the RA risk associated with DPP-4i use among a diabetic cohort initiating second-line therapy.

METHODS

This was a nested case-control study, using the adult diabetic population starting second-line antidiabetic therapy from IMS LifeLink Plus^® database (2006-2015). Cases were those with two or more RA diagnosis, at least one prescription, and 180 days enrollment prior to the event date (earliest of the two: first RA diagnosis, first RA prescription). Controls were drawn from the nest after matching (1:15) with cases on index date (± 90 days), age (± 5 years), sex, and event date (imputed to have the same time difference between cohort entry and event date as the matched case). Exposure and covariate information was gathered from the 180-day period prior to event date. Conditional logistic regression was used to assess exposure among cases and controls. Adjusted analysis was carried out after controlling for important medications and comorbidities.

RESULTS

The final sample consists of 790 cases and 11,850 controls; of these, 151 cases (19.11%) and 2177 controls (18.37%) had DPP-4i claims during the exposure assessment period. DPP-4i therapy was not significantly associated with the development of RA after adjusting for covariates (OR = 1.156, 95% CI 0.936-1.429). Changing the exposure definition or exposure window to 1 year and subgroup analyses yielded similar results except for the non-insulin-using subgroup (OR = 1.299, 95% CI 1.001-1.985) which showed a significant positive association.

CONCLUSION

DPP-4i were not significantly associated with the risk of RA compared with other second-line antidiabetic therapies.

Predictors of Use of Monitored Anesthesia Care for Outpatient Gastrointestinal Endoscopy in a Capitated Payment System

BACKGROUND & AIMS

Use of monitored anesthesia care (MAC) for gastrointestinal endoscopy has increased in the Veterans Health Administration (VHA) as in fee-for-service environments, despite the absence of financial incentives. We investigated factors associated with use of MAC in an integrated health care delivery system with a capitated payment model.

METHODS

We performed a retrospective cohort study using multilevel logistic regression, with MAC use modeled as a function of procedure year, patient- and provider-level factors, and facility effects. We collected data from 2,091,590 veterans who underwent outpatient esophagogastroduodenoscopy and/or colonoscopy during fiscal years 2000-2013 at 133 facilities.

RESULTS

The adjusted rate of MAC use in the VHA increased 17% per year (odds ratio for increase, 1.17; 95% confidence interval, 1.09-1.27) from fiscal year 2000 through 2013. The most rapid increase occurred starting in 2011. VHA use of MAC was associated with patient-level factors that included obesity, obstructive sleep apnea, higher comorbidity, and use of prescription opioids and/or benzodiazepines, although the magnitude of these effects was small. Provider-level and facility factors were also associated with use of MAC, although again the magnitude of these associations was small. Unmeasured facility-level effects had the greatest effect on the trend of MAC use.

CONCLUSIONS

In a retrospective study of veterans who underwent outpatient esophagogastroduodenoscopy and/or colonoscopy from fiscal year 2000 through 2013, we found that even in a capitated system, patient factors are only weakly associated with use of MAC. Facility-level effects are the most prominent factor influencing increasing use of MAC. Future studies should focus on better defining the role of MAC and facility and organizational factors that affect choice of endoscopic sedation. It will also be important to align resources and incentives to promote appropriate allocation of MAC based on clinically meaningful patient factors.

The Impact of Diagnostic Code Misclassification on Optimizing the Experimental Design of Genetic Association Studies

Diagnostic codes within electronic health record systems can vary widely in accuracy. It has been noted that the number of instances of a particular diagnostic code monotonically increases with the accuracy of disease phenotype classification. As a growing number of health system databases become linked with genomic data, it is critically important to understand the effect of this misclassification on the power of genetic association studies. Here, I investigate the impact of this diagnostic code misclassification on the power of genetic association studies with the aim to better inform experimental designs using health informatics data. The trade-off between (i) reduced misclassification rates from utilizing additional instances of a diagnostic code per individual and (ii) the resulting smaller sample size is explored, and general rules are presented to improve experimental designs.

Incidence and Prevalence of Cardiovascular Risk Factors Among Patients With Rheumatoid Arthritis, Psoriasis, or Psoriatic Arthritis

OBJECTIVE

To estimate prevalence and incidence of cardiovascular (CV) risk factors of hypertension, diabetes mellitus, hyperlipidemia, and obesity in patients with rheumatoid arthritis (RA), psoriasis, or psoriatic arthritis (PsA).

METHODS

Patients with RA, psoriasis, or PsA were identified based on medical and pharmacy claims from the MarketScan claims databases from January 1, 2002 through December 31, 2014. Primary outcomes included age- and sex-standardized prevalence of CV risk factors during the 12 months preceding diagnosis date and incidence rates per 1,000 patient-years, with 95% confidence intervals (95% CIs) during followup.

RESULTS

Prevalence for RA, psoriasis, and PsA cohorts for hypertension was 18.6% (95% CI 18.3-18.8), 16.6% (95% CI 16.3-17.0), and 19.9% (95% CI 19.4-20.4), respectively; for diabetes mellitus 6.2% (95% CI 6.1-6.4), 6.3% (95% CI 6.0-6.5), and 7.8% (95% CI 7.4-8.2); for hyperlipidemia 9.9% (95% CI 9.7-10.1), 10.4% (95% CI 10.2-10.7), and 11.6% (95% CI 11.2-12.0); and for obesity 4.4% (95% CI 4.2-4.6), 3.8% (95% CI 3.5-4.0), and 6.0% (95% CI 5.6-6.5). Incidence rates per 1,000 patient-years during followup for RA, psoriasis, and PsA cohorts, respectively, for hypertension were 74.0 (95% CI 72.5-75.5), 68.2 (95% CI 65.9-70.4), and 79.8 (95% CI 76.3-83.3); for diabetes mellitus 10.6 (95% CI 10.1-11.1), 13.0 (95% CI 12.1-13.8), and 14.7 (95% CI 13.5-16.0); for hyperlipidemia 40.3 (95% CI 39.4-41.3), 47.1 (95% CI 45.4-48.7), and 52.0 (95% CI 49.6-54.3); and for obesity 24.4 (95% CI 23.4-25.4), 26.4 (95% CI 25.0-27.8), and 32.9 (95% CI 30.6-35.2).

CONCLUSION

Patients with RA, psoriasis, and PsA have high prevalence and incidence of CV risk factors, suggesting the need for risk factor monitoring of these patients.

Changes in Body Mass Related to the Initiation of Disease-Modifying Therapies in Rheumatoid Arthritis

OBJECTIVE

Unintentional weight loss is important and can be predictive of long-term outcomes in patients with rheumatoid arthritis (RA). This study was undertaken to assess how primary therapies for RA may influence changes in body mass index (BMI) in RA patients from a large administrative database.

METHODS

Unique dispensing episodes of methotrexate, prednisone, leflunomide, and tumor necrosis factor inhibitors (TNFi) administered to RA patients were identified from the US Department of Veterans Affairs pharmacy databases. Values for C-reactive protein (CRP) level and BMI closest to the time point within 30 days of the treatment course start date and at follow-up time points were linked. Missing laboratory values were imputed. Weight loss was defined as a decrease in BMI of >1 kg/m(2) . Regression models were used to evaluate changes in BMI during each drug treatment as compared to treatment with methotrexate. To assess the impact of confounding by indication, propensity scores for use of each drug were incorporated in analyses using matched-weighting techniques.

RESULTS

In total, 52,662 treatment courses in 32,859 RA patients were identified. At 6 months from the date of prescription fill, weight gain was seen among patients taking methotrexate, those taking prednisone, and those taking TNFi. On average, compared to methotrexate-treated patients, prednisone-treated patients had significantly more weight gain, while leflunomide-treated patients demonstrated weight loss. In multivariable models, more weight loss (β = -0.41 kg/m(2) , 95% confidence interval [95% CI] -0.46, -0.36; P < 0.001) and a greater risk of weight loss (odds ratio 1.73, 95% CI 1.55, 1.79; P < 0.001) were evident among those receiving leflunomide compared to those receiving methotrexate. Treatment with prednisone was associated with greater weight gain (β = 0.072 kg/m(2) , 95% CI 0.042, 0.10; P < 0.001). These associations persisted in analyses adjusted for propensity scores and in sensitivity analyses.

CONCLUSION

Leflunomide is associated with significantly more, but modest, weight loss, while prednisone is associated with greater weight gain compared to other therapies for RA.

Risk of Herpes Zoster in Autoimmune and Inflammatory Diseases: Implications for Vaccination

OBJECTIVE

The herpes zoster (HZ) vaccine is recommended for adults in the US ages ≥60 years who do not have weakened immune systems. It is unclear how the risk of HZ varies according to age and disease conditions in younger patients with autoimmune or inflammatory (AI) diseases. This study was undertaken to evaluate the age-stratified incidence of HZ in patients with AI diseases as compared to older adults for whom the HZ vaccine is currently recommended by the US Centers for Disease Control and Prevention.

METHODS

Using linked data obtained from patients who were insured by US commercial and government health care plans during the period 2007-2010, 7 cohorts of patients with AI diseases were assembled: systemic lupus erythematosus (SLE), inflammatory bowel disease (IBD), rheumatoid arthritis (RA), psoriatic arthritis (PsA), psoriasis (PsO), ankylosing spondylitis (AS), and gout. Two comparator cohorts were also assembled as controls: adult patients with diabetes and adult subjects without AI diseases or diabetic conditions. HZ was identified using diagnostic codes. Age-specific incidence rates (IRs) of HZ were calculated and compared to the IRs of HZ in control subjects ages 61-70 years who were without AI diseases or diabetic conditions.

RESULTS

After review of the linked data, the following number of enrollment periods were identified: 8,395 for patients with SLE, 7,916 for patients with IBD, 50,646 for patients with RA, 2,629 for patients with PsA, 4,299 for patients with PsO, 1,019 for patients with AS, 58,934 for patients with gout, 214,631 for control patients with diabetes, and 330,727 for control subjects without AI diseases and diabetic conditions. The respective highest and lowest IRs of HZ during the study were 19.9 per 1,000 person-years in the SLE cohort and 6.8 per 1,000 person-years in the gout cohort, as compared to an IR of 5.3 per 1,000 person-years in control subjects without AI diseases or diabetic conditions. The age-specific IRs of HZ in patients with RA and those with SLE ages ≥40 years were 1.5-2 times greater than those observed in older healthy adults (IR 8.5 per 1,000 person-years), for whom the vaccine is currently recommended.

CONCLUSION

SLE, IBD, and RA are AI diseases associated with a higher risk of HZ compared to that in older adults for whom vaccination is currently recommended, suggesting that individuals with these conditions who are as young as age 40 years could potentially benefit from the HZ vaccine.

Defining Disease Phenotypes in Primary Care Electronic Health Records by a Machine Learning Approach: A Case Study in Identifying Rheumatoid Arthritis

OBJECTIVES

1) To use data-driven method to examine clinical codes (risk factors) of a medical condition in primary care electronic health records (EHRs) that can accurately predict a diagnosis of the condition in secondary care EHRs. 2) To develop and validate a disease phenotyping algorithm for rheumatoid arthritis using primary care EHRs.

METHODS

This study linked routine primary and secondary care EHRs in Wales, UK. A machine learning based scheme was used to identify patients with rheumatoid arthritis from primary care EHRs via the following steps: i) selection of variables by comparing relative frequencies of Read codes in the primary care dataset associated with disease case compared to non-disease control (disease/non-disease based on the secondary care diagnosis); ii) reduction of predictors/associated variables using a Random Forest method, iii) induction of decision rules from decision tree model. The proposed method was then extensively validated on an independent dataset, and compared for performance with two existing deterministic algorithms for RA which had been developed using expert clinical knowledge.

RESULTS

Primary care EHRs were available for 2,238,360 patients over the age of 16 and of these 20,667 were also linked in the secondary care rheumatology clinical system. In the linked dataset, 900 predictors (out of a total of 43,100 variables) in the primary care record were discovered more frequently in those with versus those without RA. These variables were reduced to 37 groups of related clinical codes, which were used to develop a decision tree model. The final algorithm identified 8 predictors related to diagnostic codes for RA, medication codes, such as those for disease modifying anti-rheumatic drugs, and absence of alternative diagnoses such as psoriatic arthritis. The proposed data-driven method performed as well as the expert clinical knowledge based methods.

CONCLUSION

Data-driven scheme, such as ensemble machine learning methods, has the potential of identifying the most informative predictors in a cost-effective and rapid way to accurately and reliably classify rheumatoid arthritis or other complex medical conditions in primary care EHRs.

Validation of administrative codes for calcium pyrophosphate deposition: a Veterans Administration study

BACKGROUND

Despite high prevalence, progress in calcium pyrophosphate deposition (CPPD) has been limited by poor awareness and absence of validated approaches to study it in large data sets.

OBJECTIVES

We aimed to determine the accuracy of administrative codes for the diagnosis of CPPD as a foundational step for future studies.

METHODS

We identified all patients with an International Classification of Diseases, Ninth Revision, Clinical Modification code for chondrocalcinosis (712.1-712.39) or pseudogout/other disorders of mineral metabolism (275.49), and convenience sample selected a comparison group with gout (274.00-03 or 274.8-9), or rheumatoid arthritis (714.0) from 2009 to 2011 at a Veterans Affairs medical center. Each patient was categorized as having definite, probable, or possible CPPD or absence of CPPD based on the McCarty and Ryan criteria using chart abstracted data including crystal analysis, radiographs, and arthritis history.

RESULTS

Two hundred forty-nine patients met the clinical gold standard criteria for CPPD based on medical records, whereas 48 patients met definite criteria, 183 probable, and 18 met possible criteria. The accuracy of administrative claims with a code of 712 or 275.49 for definite or probable CPPD was as follows: 98% sensitivity (95% confidence interval, 96%-99%), 78% specificity (74%-83%), 91% positive predictive value, and 94% negative predictive value.

CONCLUSIONS

At this center, single administrative code 275.49 or 712 accurately identifies patients with CPPD with a positive predictive value of 91%. These findings suggest that administrative codes can have strong clinical accuracy and merit further validation to allow adoption in future epidemiologic studies of CPPD.

The Biobank Portal for Partners Personalized Medicine: A Query Tool for Working with Consented Biobank Samples, Genotypes, and Phenotypes Using i2b2

We have designed a Biobank Portal that lets researchers request Biobank samples and genotypic data, query associated electronic health records, and design and download datasets containing de-identified attributes about consented Biobank subjects. This do-it-yourself functionality puts a wide variety and volume of data at the fingertips of investigators, allowing them to create custom datasets for their clinical and genomic research from complex phenotypic data and quickly obtain corresponding samples and genomic data. The Biobank Portal is built upon the i2b2 infrastructure [1] and uses an open-source web client that is available to faculty members and other investigators behind an institutional firewall. Built-in privacy measures [2] ensure that the data in the Portal are utilized only according to the processes to which the patients have given consent.

Tumor Necrosis Factor Inhibition and Head and Neck Cancer Recurrence and Death in Rheumatoid Arthritis

The objective of this retrospective cohort study was to determine the effect of tumor necrosis factor inhibitor (TNFi) therapy on the risk of head and neck cancer (HNC) recurrence or HNC-attributable death in patients with rheumatoid arthritis (RA). RA patients with HNC were assembled from the US national Veterans' Affairs (VA) administrative databases, and diagnoses confirmed and data collected by electronic medical record review. The cohort was divided into those treated with non-biologic disease-modifying anti-rheumatic drugs (nbDMARDs) versus TNF inhibitors (TNFi) after a diagnosis of HNC. Likelihood of a composite endpoint of recurrence or HNC-attributable death was determined by Cox proportional hazards regression. Of 180 patients with RA and HNC, 31 were treated with TNFi and 149 with nbDMARDs after the diagnosis of HNC. Recurrence or HNC-attributable death occurred in 5/31 (16.1%) patients in the TNFi group and 44/149 (29.5%) patients in the nbDMARD group (p = 0.17); it occurred in 2/16 (13%) patients who received TNFi in the year prior to HNC diagnosis but not after. Overall stage at diagnosis (p = 0.03) and stage 4 HNC (HR 2.49 [CI 1.06-5.89]; p = 0.04) were risk factors for recurrence or HNC-attributable death; treatment with radiation or surgery was associated with a lower risk (HR 0.35 [CI 0.17-0.74]; p = 0.01 and HR 0.39 [CI 0.20-0.76]; p = 0.01 respectively). Treatment with TNFi was not a risk factor for recurrence or HNC-attributable death (HR 0.75; CI 0.31-1.85; p = 0.54). We conclude that treatment with TNFi may be safe in patients with RA and HNC, especially as the time interval between HNC treatment and non-recurrence increases. In this study, TNF inhibition was not associated with an increase in recurrence or HNC-attributable death.