Latent tuberculosis screening among new users of a biologic or targeted synthetic DMARD: gaps in screening overall and among Janus kinase inhibitors

Abstract

Objective:

We combined claims and EHR data to provide contemporary and accurate estimates of latent TB screening among new users of a biologic or targeted synthetic DMARD and assess potential gaps in testing by drug type, patient characteristics, and practice.

Methods:

Our denominator population was patients in the RISE registry and Medicare using a b/tsDMARD in 2018 without a claim or prescription in the year prior. TB screening was assessed in both Medicare and RISE 1 and 3 years before the medication start date. We calculated the proportion screened overall, by medication class, and by practice. We tested for demographic differences in screening using logistic regression.

Results:

In the year prior to drug starts, 65.6% of patients had any TB screening; in a 3-year window, 72.9% had any TB screening. Rates of screening within 1 year by drug type were greater or equal to the overall screening rate for most drugs except for JAKi (46%) and IL-17i (11.5%). A lower proportion of Hispanic and Asian patients were screened compared to White patients. Practice screening rates ranged from 20.0–92.9% of patients within 1 year.

Conclusion:

We report higher screening rates than have previously been published due to combining claims and EHR data. However, important safety gaps remain, namely, reduced screening among new users of a JAKi or IL-17i, among Asian and Hispanic patients, and low performing practices. Educational initiatives, team-based care delivery, task shifting and technological interventions to address observed gaps in patient safety procedures are needed.

Introduction

Many biologic or targeted synthetic DMARDs (b/tsDMARDs) increase the risk of reactivation of latent tuberculosis infection (LTBI), putting patients at risk for adverse events including death.^1–3 Accordingly, professional society guidelines recommend screening prior to starting most b/tsDMARDs.^{4, 5} While the increased risk of LTBI reactivation was first noted in connection with tumor necrosis factor inhibitors (TNFis), there is also an increased risk of reactivation among multiple other classes of b/tsDMARDs.⁶ With a growing number of these drugs available in rheumatology, it is important to investigate whether TB screening practices have kept pace with recommendations to minimize preventable patient safety events. Evaluation of TB screening is particularly important for newer classes of drugs, for which patient safety data are sparse.

Previous studies have reported rates of TB screening in select clinical populations^7–9 and in a national registry of US rheumatology practices.^{10, 11} Prior analyses of the Rheumatology Informatics System for Effectiveness (RISE) registry data showed that 29.7% (in the prior year) and 63.4% (at any time before or after the index date) of eligible patients received TB screening before starting a b/tsDMARD, which represents a significant gap in patient safety.¹¹ However, the magnitude of this gap may be smaller than previously reported, as this figure was derived from electronic health record (EHR) data in the RISE registry and thus may not have fully captured all screening due to lack of data capture from other practice settings. One way to address this is to use a combined data source of EHR and insurance claims, which together can provide a more comprehensive view of all laboratory testing completed in a given time window and can therefore enhance the validity of findings.^{12, 13}

To provide contemporary and accurate estimates of TB screening, we reported the proportion of new users of a b/tsDMARD who were screened for TB and evaluated potential gaps in testing by drug type, patient characteristics, and practice, using a combined national sample of patients with both EHR and claims data. This analysis aligns with several of the Centers for Medicare and Medicaid Services’ (CMS) National Quality Strategy Goals.¹⁴ In particular, patient safety with an ambitious but worthy goal of achieving zero preventable harms to patients.

Methods

Data Source

Data were taken from the American College of Rheumatology’s RISE registry with linkage to Medicare Parts A, B, and D claims. RISE is a national EHR enabled registry. As of 2018, RISE included validated data from 1113 providers in 226 practices, representing approximately 32% of the U.S. clinical rheumatology workforce.¹⁵ The Western IRB and UCSF Committee on Human Research approved this study.

Study Population

Our sample comprised all individuals in RISE with continuous Medicare enrollment in 2017 and 2018 who were new users of a qualifying b/tsDMARD in 2018 in Medicare. New users were defined as those with (1) a claim in Medicare for a b/tsDMARD in 2018 but no claim at least 365 days before their index date (index date is defined as the date of the first claim for a qualifying b/tsDMARD in Medicare in 2018), and (2) no prescription for any b/tsDMARD in RISE between 365 and 30 days before their index date. The 30-day grace period was included to allow a small gap between when data enters the EHR and subsequently registered in Medicare claims databases. Patients could only be included once.

Qualifying b/tsDMARDs included the following drugs used in rheumatology and approved for use by the FDA before 2019: TNFis (adalimumab, certolizumab, etanercept, golimumab and infliximab), cytotoxic T-lymphocyte associated protein-4) CTLA-4 inhibitors (abatacept), interleukin (IL)-1 inhibitors (anakinra, canakinumab, rilonacept), IL-6 inhibitors (sarilumab, siltuximab, tocilizumab), IL-23 inhibitors (guselkumab, tildrakizumab), IL-12/23 inhibitor (ustekinumab), IL-17 inhibitors (brodalumab, ixekizumab, secukinumab), and janus kinase (JAK) inhibitors (baricitinib, tofacitinib).

TB screening tests and screening window

Latent TB screening was identified in both Medicare claims and RISE registry EHR data to obtain a comprehensive picture of completed screening. TB screening in Medicare was identified using Healthcare Common Procedure Coding System (HCPCS) codes for TB laboratory tests: interferon gamma release assays (IGRA) tests, TB skin tests (e.g., purified protein derivative (PPD)), a documented medical exception (e.g., a patient who has been treated for TB in the past), and codes for the Quality Payment Program measure 176 for a TB screen or a medical exception. TB screening in RISE was identified by searching for the same HCPCS codes as in Medicare and identification and categorization of terms identifying TB blood tests (e.g., Quantiferon TB Gold), TB skin tests (e.g., Mantoux), or documented medical exceptions (e.g., “history of tuberculosis”). Evidence of latent or active TB treatment (i.e., prescriptions for isoniazid, rifapentine or rifampin) at any time prior to the index date in either Medicare or RISE also fulfilled the TB testing measure. We combined these data sources and identified whether patients had received any screening, as well as the type of screening test. In the primary analysis, we calculated the proportion of patients screened in the year prior to receipt of the new b/tsDMARD, consistent with the Quality Payment Program measure. In a sensitivity analysis, we permitted TB screening in the 3 years before and up to 30 days after the patient’s index date. The testing windows are depicted in Figure 1.

Figure 1.

Testing windows for TB. The primary analysis included evidence of TB screening in either Medicare (light gray bar) or RISE (dark gray bar) within 12 months of the index date (vertical dotted line) and evidence of TB treatment at any previous time. In our sensitivity analysis we widened the screening window to the beginning of 2016 for Medicare (due to data availability) and to 3 years before the index date in RISE. Both data sources also had a 30-day grace period after the index date.

Covariates

We extracted the following variables for each patient from RISE: age, sex, self-reported race and ethnicity, zip code and socioeconomic status (SES; as measured by the Area Deprivation Index (ADI), a measure of socioeconomic disadvantage at the patient residential Census Block Group level, which is calculated using 17 indicators from the American Community Survey that encompass income, education, employment, and housing conditions).¹⁶ The ADI was used as a proxy of individual SES. We matched patient zip codes to USDA rural-urban commuting area codes categorized into urban; large rural city/town; small and isolated small rural town.¹⁷ Medicare/Medicaid dual eligible status was extracted from the Medicare beneficiary file.

Statistical Analyses

We present the demographic characteristics of our sample overall and assess differences between patients that did and did not receive TB screening using independent samples t tests and chi-square tests of independence, as appropriate. We computed the proportion of new b/tsDMARD users screened overall, by type of TB test, and by each b/tsDMARD medication class. Using logistic regression, we tested for differences in screening by age, sex, race and ethnicity, insurance, urban-rural category and SES in a fully adjusted model. We used generalized estimating equations to account for clustering by practice. We calculated the proportion of patients screened among practices with at least 20 new b/tsDMARD users during the study period to investigate practice variation in TB screening. The threshold of 20 patients was chosen following standard reporting requirements for CMS to ensure the reliability of estimates.¹⁸

Sensitivity Analysis

As a sensitivity analysis we assessed TB screening rates among those with incident vs prevalent disease overall as well as stratified by drug type. To complete this analysis we restricted our sample to those with continuous Medicare enrollment from 2016–2018 and searched for ICD codes for common rheumatic diseases that are typically treated with b/tsDMARDs (in alphabetical order): ankylosing spondylitis, anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis, antiphospholipid syndrome, axial spondylarthritis, Crohn’s disease, giant cell arteritis, inflammatory myopathies, mixed connective tissue disease, other arthritis, polymyalgia rheumatica, psoriasis, psoriatic arthritis, rheumatoid arthritis, sarcoidosis, Sjogren’s disease, systemic lupus erythematosus, systemic sclerosis, undifferentiated connective tissue disease, ulcerative colitis, other vasculitis. A diagnosis required two codes at least 30 days apart. Incident cases must have had at least one year of observation without an ICD code and have no prevalent case of any of the searched for diseases. Patients with multiple diagnoses were assigned to one according to the following hierarchy: SLE > PSA > AS > RA > axSpA/other spondylarthritis > CTD/vasculitis > psoriasis > Crohn’s/ulcerative colitis and the proportion screened by diagnosis was assessed.

Results

Among 2853 new b/tsDMARD users, mean age was 73±8.6 years, 72% were female, and 73% were non-Hispanic White (Table 1). In the year prior to drug starts, 65.6% of patients had any latent TB screening documented, including 61.3% with an IGRA, 4.7% with a TST, 3.8% with a documented medical exception; 2.2% had prior TB treatment. In our sensitivity analysis using a 3-year window, 72.9% had any TB screening (Table 2). In bivariable analyses comparing differences in TB screening by demographic characteristics, men, and high-SES individuals were statistically significantly less likely to be screened (Table 1). A lower proportion of Hispanic and Asian patients were screened compared to White patients.

Table 1.

Demographic characteristics of new b/tsDMARD users in RISE/Medicare sample in 2018 overall and among those screened within 1 year and within 3 years with a 30-day grace period.

	Total (N=2853)		Any screening within 1 year (N=1870)			Any screening within 3 years with 30-day grace period (N=2081)
Characteristic	N	%	N	% screened	p-value	N	% screened	p-value
Age (mean, sd)	73.54	8.58	73.50	8.25	0.74	73.41	8.24	0.21
Sex
Women	2059	72.17	1370	66.54	0.07	1526	74.11	0.02
Men	794	27.83	500	62.97		555	69.90
Race/ethnicity
White	2071	72.59	1362	65.77	0.13	1520	73.39	0.21
African American	197	6.91	138	70.05		153	77.66
Hispanic	132	4.63	74	56.06		93	70.45
Asian	35	1.23	22	62.86		22	62.86
Other/mixed	15	0.53	12	80.00		12	80.00
Unknown	403	14.13	262	65.01		281	69.73
Insurance
Medicare only	2419	84.79	1581	65.36	0.62	1746	72.18	0.03
Medicaid (dual eligible)	434	15.21	289	66.59		335	77.19
Rural-Urban designation
Urban	2329	81.63	1508	64.75	0.22	1679	72.09	0.15
Large rural city/town	275	9.64	187	68.00		215	78.18
Small/isolated rural	209	7.33	149	71.29		157	75.12
Unknown	40	1.40	26	65.00		30	75.00
National ADI
First quartile (highest SES)	683	23.94	397	58.13	<0.01	450	65.89	<0.01
Second quartile	656	22.99	437	66.62		490	74.70
Third quartile	640	22.43	440	68.75		484	75.63
Fourth quartile (lowest SES)	647	22.68	453	70.02		499	77.13
Unknown	227	7.96	143	63.00		158	69.60

Table 2.

Proportion of new b/tsDMARD users screened for TB within 1 year and 3 years with a 30-day grace period overall and by TB screening method.

	New users of Biologics or tsDMARDs (n=2853)
Screening type	Within 1 year	Within 3 years with 30-day grace period
Any screening	65.55	72.94
Screened with IGRA	61.27	68.98
Screened with TST	4.73	7.05
Documented medical exception	3.82	5.89
Prior TB treatment*	2.21	2.21

IGRA = Interferon-Gamma Release Assay; TST = Tuberculin Skin Test

^*Prior TB treatment at any time.

Of 2853 new b/tsDMARD users, 52.2% (n=1490) started a TNFi, 16.7% (n=476) started a CTLA-4i, 13.8% (n=393) an IL-17i, 7.3% (n=207) an IL-6i, 6.1% (n=174) a JAKi, 3.6% (n=103) an IL-12/IL-23i, and 0.4% (n=10) an IL-1i or IL-23i. Due to small sample sizes, IL-1i and IL-23i were not considered in further analyses. Rates of screening within 1 year by drug type were greater or equal to the overall screening rate (65.6%) for most drugs except for JAKi (46%) and IL-17i (11.5%). Similar results were seen in the 3-year and 30-day grace period time windows (Figure 2).

Figure 2.

Proportion of new b/tsDMARD users screened for TB by medication class.

In a fully adjusted logistic regression, patients in the highest quartile of SES had statistically significantly lower odds of being screened within 1 year (odds ratio (OR)=0.61; 95%CI 0.40–0.94) compared to the lowest quartile, and Hispanic patients were statistically significantly less likely to be screened compared to White patients within 1 year (OR=0.64; 95%CI 0.46–0.90).

51 practices had at least 20 eligible patients totaling 1900 patients; 953 patients in 129 practices were not included in this analysis. Screening rates ranged from 20.0–92.9% of patients within 1 year (Figure 3). The median screening rate within practices was 70.0% (IQR: 58.3% - 78.7%).

Figure 3.

Proportion of new b/tsDMARD users screened for TB, by practice among practices with at least 20 patients (n=51). Each vertical column represents one practice in the RISE registry.

Within 12 months of initiation 79.2% (81.6% within 3-years) of patients with incident disease were screened (supplemental table 1). There was a significant difference in the proportion screened for TB within 1 year by disease (in descending order): psoriasis 80.7%, AS 75.6%, RA 74%, other spondylarthritis 72.9%, PSA 68.1%, GI 64.4%, SLE 59.6%, CTD 59.5% (data not shown).

Discussion

Using a comprehensive data source that includes both EHR data and Medicare claims, we found that just over 1 in 3 new users of a b/tsDMARD did not receive TB screening within the recommended time window of one year prior to new b/tsDMARD initiation. When the screening window was expanded to 3 years, performance only slightly improved with approximately 1 in 4 new users not receiving screening. We found important variation in the proportion screened between drug classes, with more than half of new users of a JAKi and nearly 90% of new users of IL-17i not receiving screening. Demographically, we found Hispanic, Asian, and higher SES users were less likely to be screened. While the mean proportion screened among practices was around the national average, the range was large and differences between high- and low-performing practices could be a fruitful area for future research and quality improvement.

The overall estimates of TB screening presented here are higher than we have previously reported using the RISE registry alone.¹¹ This is the result of combining Medicare claims with EHR data in RISE to improve the validity of our outcome assessment and denominator population.¹² The biggest innovation of this study is the integration of Medicare claims which uniquely identified a meaningful proportion of additional TB tests, suggesting these were performed in healthcare settings outside of the rheumatology practice and may therefore have not entered the participating rheumatologist’s EHR in an accessible manner. Importantly, there are very few situations in which a Medicare recipient would receive a TB screen and not have it appear as a claim. Therefore, we are confident our study has captured nearly all screens ordered during the study period. The EHR data added complementary information as well. For example, we found that EHR data were more likely to capture medical exceptions, allowing us to better capture when TB screening was not indicated. Additionally, we excluded some patients that met the new user definition in Medicare claims but had a record of use of a b/tsDMARD in the EHR, which can happen when a physician provides a free sample to a patient.¹⁹ Taken together, these results suggest previously published screening rates likely underestimate the true rate. Future work is planned to quantify the magnitude of these biases in each data source.

Though we are confident we have captured nearly all TB screens ordered or excepted during our study period, it is possible that a small proportion of the population received more remote TB screening that was not captured in our analyses. While we were restricted to a 3-year look back given the available claims data, previous analyses that had access to screening at any time report gaps.^{11, 20} As a sensitivity analysis, we assessed screening rates among those with incident disease in our study population. Within 12 months of initiation 79.2% (81.6% within 3-years) of patients with incident disease were screened (supplemental table 1) suggesting that at a minimum, screening rates need to be increased among first-time users as directed by the quality measure. It is also reasonable to argue that screening three or more years prior to initiating a new biologic is too remote and warrants renewed TB screening when (re)initiating a biologic or JAKi. This is particularly relevant for our study cohort as we excluded patients with use of any qualifying b/tsDMARD in the year before the index date.

We identified LTBI screening rates below the overall average for patients initiating JAKi and IL-17i. Neither JAKi nor IL-17i are first-line therapeutics suggesting remote testing could explain these gaps. However, our sensitivity analysis showed reduced screening rates among patients initiating JAKi or IL-17i with both incident and prevalent disease in both the 1- and 3-year windows (with the exception of the 3-year screening rate for patients with prevalent disease initiating a JAKi being 70.3%). These sample sizes are reduced, particularly among incident user cohorts, and our findings require replication in a larger cohort. However, these findings are particularly concerning in the case of JAKi as emerging evidence suggests that the risk of LTBI reactivation with JAKi is similar to that of TNFi.⁶ While some research suggests there appears to be a low risk of LTBI reactivation associated with the IL-17i secukinumab,⁶ there remains uncertainty around the risk associated with exposure to other IL-17i. Most of the data pertaining to other IL-17i are derived from randomized controlled trials which have (appropriately) excluded patients with a positive TB screen or included only patients who have received prophylactic latent TB treatment and have a relatively short duration of follow-up.⁶ Long-term extension studies and real-world data from large registries is better suited to address the risk of LTBI reactivation associated with the use of IL-17i. RISE is well poised to address this question once more follow-up time accumulates and linkage to more recent Medicare claims is completed.

Overall, we see relatively small differences in screening by most patient characteristics suggesting that patient characteristics are not major drivers of quality of care for TB screening. In our multivariable model there were no statistically significant differences in the odds of screening by age, sex, dual eligible status, or rural-urban commuting designation. We reported reduced odds of screening among high SES participants, and Hispanic or Asian compared to White patients. The difference in screening by SES may be reflective of the physician judgement of patient risk as TB risk is patterned by SES in the United States.²¹ Regarding race-ethnicity, this result was not always statistically significant, but this was likely due to the relatively small number of Hispanic and Asian patients. This potential disparity is troubling as rates of TB in the United States are higher among these racial and ethnic groups²² and TB is endemic in many parts of South America and Asia.²³ While universal screening in the general population is not cost-effective compared to risk-based screening in low-prevalence settings like the United States²⁴ these findings are in line with prior work that has found physicians have poor recognition of risk factors before prescribing immunosuppressive therapy.²⁵

The main strength of this study is the use of a combined EHR and claims dataset, incorporating data from over 200 rheumatology practices in the United States. To date, this is the most comprehensive picture of TB screening among new users of a b/tsDMARD. However, our results are still limited in important ways. We did not have access to scanned documents or clinical notes, so there may be a small number of tests that were performed but for which there was no Medicare billing. We were restricted to a 3-year look back given the available claims data; it is possible that a small proportion of the population received more remote TB screening that was not captured in our analyses. However, it is reasonable to also argue that screening three or more years ago is too remote and warrants renewed TB screening when (re)initiating a biologic or JAKi. Finally, while the RISE registry represents a large portion of US rheumatology practices, it is unclear how these results may generalize to practices that are not included, particularly academic medical center practices. Further, given the enrollment criteria for Medicare, it is unclear how these results may generalize to younger patients or those who are not dually eligible for Medicare and Medicaid.

As the use of biologic and tsDMARDs increases in rheumatology and in other medical specialties, building both local and national systems to ensure patient safety is paramount, given the persistent gaps in screening procedures as well as significant practice variation identified in this study. At the national level, audit and feedback mechanisms such as the RISE registry that use standardized quality measures are important, as are efforts to work with national EHR vendors to ease the documentation and retrieval burden of TB testing information. At the local health system and clinic levels, developing standardized workflows suited to the resources available to clinicians and the patient population being served is important. Using quality improvement tools such as clinician feedback, pharmacy checks prior to dispensing, and EHR order sets have been used successfully in prior studies. As one example, Baker et al. describe the successful implementation of a Best Practice Advisory which significantly increased TB screening rates.²⁶ Finally, it would be fruitful to interview high-performing practices to understand local workflows and disseminate findings that might be useful to other rheumatology practices in the country.

The incidence of active TB in low-TB incidence settings is approximately 4–6 cases/100,000 person/years and data from post-marketing studies and clinical registries suggest that individuals using anti-TNF biologics have a 2-to-6 fold increased risk and the risk is higher in patients also using glucocorticoids or conventional synthetic DMARDs (e.g. methotrexate).^27–29 While the United States is a low incidence country, CMS’s National Quality Strategy Goal of achieving zero preventable harm¹⁴ suggest any gap in care should be mitigated.

In conclusion, we found significant gaps in patient safety among new users of a b/tsDMARD with concerning differences by type of b/tsDMARD, demographic characteristics and practice. Educational initiatives, team-based care delivery, task shifting and technological interventions to address observed gaps in patient safety procedures are needed.

Table 3.

Odds Ratio for screening for TB within 1 year and 3 years with a 30-day grace period by demographics among new b/tsDMARD users from fully adjusted logistic regression models.

	Within 1 year		Within 3 years with 30-day grace period
Characteristic	OR	95% CI	OR	95% CI
Age	1	(0.99 – 1.01)	1	(0.99 – 1.01)
Sex
Women	ref	-	ref	-
Men	0.88	(0.73 – 1.05)	0.83	(0.68 – 1.02)
Race/ethnicity
White	ref	-	ref	-
African American	1.1	(0.74 – 1.63)	1.09	(0.70 – 1.69)
Hispanic	0.64	(0.46 – 0.90)	0.77	(0.55 – 1.08)
Asian	0.99	(0.36 – 2.74)	0.62	(0.21 – 1.82)
Other/mixed	2.14	(0.73 – 6.28)	1.42	(0.48 – 4.24)
Unknown	1.03	(0.80 – 1.33)	0.88	(0.69 – 1.12)
Insurance
Medicare only	ref	-	ref	-
Medicaid (dual eligible)	1.02	(0.77 – 1.34)	1.24	(0.95 – 1.63)
Rural-Urban designation
Urban	ref	-	ref	-
Large rural city/town	1.04	(0.78 – 1.39)	1.23	(0.92 – 1.65)
Small/isolated rural	1.16	(0.81 – 1.68)	1.01	(0.69 – 1.46)
Unknown	1.16	(0.36 – 3.71)	1.38	(0.47 – 4.08)
National ADI
First quartile (highest SES)	0.61	(0.40 – 0.94)	0.64	(0.39 – 1.06)
Second quartile	0.87	(0.66 – 1.15)	0.94	(0.69 – 1.28)
Third quartile	0.96	(0.76 – 1.21)	0.96	(0.74 – 1.25)
Fourth quartile (lowest SES)	ref	-	ref	-
Unknown	0.73	(0.52 – 1.03)	0.68	(0.48 – 0.98)

Significance and Innovation.

• Ours is the first study to assess screening for latent tuberculosis infection in patients initiating biologic or targeted synthetic (b/ts) DMARD therapy using a dataset which combined information from both claims and electronic health records.

• We find higher, but still suboptimal, rates of screening than previously published. Just over 1 in 3 new users of a b/tsDMARD did not receive TB screening within the recommended time window of one year prior to new b/tsDMARD initiation.

• We found important variation in the proportion screened between drug classes, with more than half of new users of a JAKi and nearly 90% of new users of IL-17i not receiving screening.

• Demographic disparities in screening persist with Hispanic, Asian, and higher socioeconomic status (SES) initiators being less likely to be screened.