Latent Tuberculosis Infection Testing Practices in a Large US Integrated Healthcare System

Abstract

Background

Tuberculosis (TB) is a public health threat, with >80% of active TB in the United States occurring due to reactivation of latent TB infection (LTBI). We may be underscreening those with high risk for LTBI and overtesting those at lower risk. A better understanding of gaps in current LTBI testing practices in relation to LTBI test positivity is needed.

Methods

This study, conducted between 1 January 2008 and 31 December 2019 at Kaiser Permanente Southern California, included individuals aged ≥18 years without a history of active TB. We examined factors associated with LTBI testing and LTBI positivity.

Results

Among 3 816 884 adults (52% female, 37% White, 37% Hispanic, mean age 43.5 years [standard deviation, 16.1]), 706 367 (19%) were tested for LTBI, among whom 60 393 (9%) had ≥1 positive result. Among 1 211 971 individuals who met ≥1 screening criteria for LTBI, 210 025 (17%) were tested for LTBI. Factors associated with higher adjusted odds of testing positive included male sex (1.32; 95% confidence interval, 1.30–1.35), Asian/Pacific Islander (2.78, 2.68–2.88), current smoking (1.24, 1.20–1.28), diabetes (1.13, 1.09–1.16), hepatitis B (1.45, 1.34–1.57), hepatitis C (1.54, 1.44–1.66), and birth in a country with an elevated TB rate (3.40, 3.31–3.49). Despite being risk factors for testing positive for LTBI, none of these factors were associated with higher odds of LTBI testing.

Conclusions

Current LTBI testing practices may be missing individuals at high risk of LTBI. Additional work is needed to refine and implement screening guidelines that appropriately target testing for those at highest risk for LTBI.

Tuberculosis (TB) is a global public health threat, with an estimated 9.9 million TB cases (127 per 100 000) occurring worldwide in 2020 [1]. In the United States, 8904 new TB cases occurred in 2019 (2.7 per 100 000), with California accounting for one-quarter of cases [2]. Further, 2 million Californians (6% of the population) have latent TB infection (LTBI), which can progress to active TB without treatment [3].

In the United States, >80% of active TB cases occur due to reactivation of LTBI [4, 5]. Thus, domestic TB control is focused on preventing progression to active TB. While the duration of latency can be a lifetime for healthy individuals, reactivation often occurs within the first 2–5 years following infection in 5%–15% of infected individuals [6, 7]. The California Department of Public Health (CADPH) recommends screening asymptomatic adults based on birth, travel, or residence in a country with elevated TB for >1 month; immunosuppression; and close contact with active TB [8, 9]. However, guideline implementation has been suboptimal [8–10] as these risk factors are broadly defined, often not disclosed by patients, not documented, and can be confusing to providers. Recent statewide data in California demonstrated major gaps in both LTBI screening and treatment among patients with active TB [11]. On the other hand, outside of the screening guidelines, another indication for routine LTBI screening includes annual screening as a requirement for employment, school-based policies, or congregate settings such as nursing homes.

LTBI testing efforts may be disproportionally reaching populations with low risk of LTBI, while populations with the highest risk may be undertested [12, 13]. A better understanding of the current LTBI testing practices is needed to more appropriately direct testing efforts to the populations at highest risk. Therefore, we used 12 years of electronic health record (EHR) data from Kaiser Permanente Southern California (KPSC) to identify factors associated with LTBI testing and positivity among variables selected a prior based on expert knowledge and CADPH screening guidelines.

METHODS

Study Period and Setting

This study was conducted from 1 January 2008 through 31 December 2019 at KPSC, a large, integrated healthcare organization that serves >4.8 million residents of Southern California [14]. KPSC's EHR captures comprehensive health information of members, including demographic characteristics, diagnoses, laboratory tests, and medications from all care settings. Documentation of care received outside of KPSC is integrated into the EHR through reimbursement claims.

Study Population and Statistical Analyses

LTBI Testing–Eligible Cohort

The LTBI testing–eligible cohort was assembled to examine factors associated with testing for LTBI among KPSC members aged ≥18 years (Figure 1). Date of cohort entry was 1 January 2008 or the first date that met the 24 months’ continuous enrollment requirement during the study period, whichever came later. We described demographic/clinical characteristics and LTBI risk factors (Supplementary Table 1) among individuals tested and never tested for LTBI. LTBI testing was defined as tuberculin skin test (TST) or interferon gamma release assay (IGRA). We also examined the proportion tested for LTBI among those who met the CADPH's LTBI screening criteria (birth in or travel to a TB-endemic country, immunosuppression, and exposure to active TB) [8].

LTBI-Tested Cohort

From the LTBI testing–eligible cohort, we restricted our study to individuals with ≥1 LTBI test (“LTBI-tested cohort”) to examine factors associated with testing positive among those tested for LTBI. A positive LTBI test was defined as a positive TST (≥5 mm) or IGRA (≥0.35 IU/mL). We described characteristics and LTBI risk factors of individuals with ≥1 positive test and those without any positive test. We also examined LTBI positivity among those who met the CADPH screening criteria.

For both cohorts, individuals with a history of active TB prior to cohort entry were excluded. Because a LTBI diagnosis code without an accompanying LTBI test could have been carried over from a previously treated infection, those with a diagnosis code without a test were excluded; a sensitivity analysis was performed that included these individuals. To identify factors associated with LTBI testing and positivity, we identified factors associated with LTBI testing and testing positive among variables chosen a priori based on expert knowledge informed by the CADPH screening guidelines, previous research, and clinical practice. Variables were assessed across each patient's follow-up rather than being anchored to a particular test to understand the odds of each outcome across the entire follow-up period. We used multivariable logistic regression to estimate adjusted odds ratios (aORs) for each factor, adjusting for all other factors. Multicollinearity was assessed by variance inflation factors, with values >5 indicating possible multicollinearity [15]. Basic model fit was assessed using the area under the receiver operating curve (AUROC), with values >0.7 considered acceptable [16]. Due to missing data on country of birth (55%), foreign-born status was imputed based on a previously published prediction algorithm [17] that used preferred language, race/ethnicity, and percent of population born outside the United States at the census level in logistic regression. The KPSC Institutional Review Board approved the study with a waiver of informed consent.

RESULTS

We identified 4 016 699 individuals aged ≥18 years with ≥2 years of KPSC membership and no history of active TB prior to cohort entry between 1 January 2008 and 31 December 2019 (Figure 1).

Demographic and Clinical Characteristics

LTBI Testing–Eligible Cohort

We identified 3 816 884 eligible individuals for the LTBI testing cohort (1 976 478 [52%] female, 1 403 448 [37%] White, 1 402 366 [37%] Hispanic, mean age 43.5 years [standard deviation, 16.1]). A total of 706 367 (18%) were tested for LTBI during the study period (11.4% IGRA only, 82.6% TST only, and 6.0% both IGRA and TST), of whom 469 766 (67%) were female and 317 501 (45%) were aged 18–35 years. Among the 3 110 517 never tested for LTBI, 1 506 712 (48%) were female and 1 095 942 (35%) were aged 18–35 years (Table 1). Distribution of race/ethnicity was similar for non-Hispanic White (37%), Hispanic (37%), and non-Hispanic Asian/Pacific Islander (11%) for tested and never-tested individuals, while 11% of the tested and 7.5% of the never-tested individuals were non-Hispanic Black. Of those tested for LTBI, 40 212 (5.7%) received an immunosuppressant during follow-up and 157 405 (22%) were born in a country with an elevated TB rate (with imputation [17]). Among those never tested for LTBI, 44 566 (1.4%) used an immunosuppressant and 929 832 (30%) were born in a country with an elevated TB rate.

A total of 1 211 971 eligible individuals had ≥1 risk factor for LTBI based on the current screening guidelines [8] (Table 2; breakdown of LTBI testing by positivity among individuals with ≥1 risk factor available in Supplementary Table 2). Among 183 741 immunosuppressed any time during follow-up and among 1 087 237 born in a country with an elevated TB rate (with imputation), 88 088 (48%) and 157 405 (14%) tested for LTBI, respectively. Among 24 803 with a history of travel >30 days to a country with an elevated TB rate and 5406 with a history of close contact with TB, 6704 (27%) and 4297 (79%) tested for LTBI, respectively.

LTBI-Tested Cohort

Of the 706 367 tested for LTBI, 60 393 (9%) tested positive during the study period (36 335 [60%] female, 19 766 [33%] aged 18–35 years, 29 883 [49%] Hispanic, and 15 458 (26%) Asian/Pacific Islander (Table 1). Characteristics of this cohort were similar overall to the LTBI testing–eligible cohort. However, the proportion born in a country with an elevated TB rate (imputed or actual, 22% vs 28%) and the proportion with preference to speak a language spoken in a country with an elevated TB rate (7% vs 12%) were lower in the LTBI-tested cohort, while the proportion immunosuppressed during the study period (6% vs 2%) was higher. Among the 645 974 who never tested positive for LTBI, 433 431 (67%) were female, 253 579 (39%) were aged 18–35 years, 228 063 (35%) were Hispanic, and 64 043 (10%) were Asian/Pacific Islander. Among those with ≥1 positive test, 35 335 (59%) were born in a country with an elevated TB rate and 1174 (1.9%) had a history of travel >30 days to a country with an elevated TB rate. Among those who never tested positive, 122 070 (19%) were born in and 5530 (0.4%) had a history of travel >30 days to a country with an elevated TB rate.

Among the 210 025 individuals with ≥1 LTBI risk factor based on the CADPH guidelines who were tested for LTBI, LTBI positivity was 8% among individuals immunosuppressed any time prior to the first LTBI test, 22% among those born in a country with an elevated TB rate (with imputation), 18% among those with a history of travel >30 days to a country with an elevated TB rate, and 34% in those with a history of close contact with TB (Table 2).

Factors Associated With LTBI Testing Among the LTBI Testing–Eligible Cohort

Males had lower odds of testing for LTBI compared with females (aOR, 0.50; 95% confidence interval: .50–.51). Younger age groups had higher odds of testing for LTBI compared with older groups (Figure 2, Supplementary Table 3). The odds of testing were higher in non-Hispanic Blacks compared with their White counterparts (1.41, 1.40–1.43) and for those living in an area with median household income >$80 000 compared with <$40 000 (1.16, 1.14–1.17). Medicaid beneficiaries had higher odds of testing for LTBI (1.39, 1.37–1.41) compared with commercial insurance plans. Comorbidities associated with higher odds of LTBI testing included use of tumor necrosis factor-α inhibitors (34.2, 32.2–36.4), human immunodeficiency virus (HIV; 13.6, 12.8–14.5), end-stage renal disease (4.52, 4.39–4.65), and use of high-dose steroid (1.89, 1.85–1.93). Birth in a country with an elevated TB rate was associated with lower odds of testing (0.88, .87–.89) compared with those born in the United States. However, history of travel (>30 days) to a country with an elevated TB rate (1.55, 1.50–1.60) and prior exposure to TB (17.7, 16.6–19.0) were associated with higher odds of LTBI testing. Variance inflation factors for all variables (<5) and the AUROC (0.70) were acceptable.

Factors Associated With LTBI Positivity Among the LTBI-Tested Cohort

Males had higher odds of testing positive for LTBI (1.32, 1.30–1.35) compared with females (Figure 2, Supplementary Table 2). Compared with younger age groups (18–35 years), older individuals had higher odds of testing positive. The odds of testing positive were markedly higher among Asians/Pacific Islanders (2.78, 2.68– 2.88), Hispanics (2.45, 2.38–2.52), and Blacks (2.28, 2.20–2.37) compared with their White counterparts. Higher neighborhood deprivation was associated with higher odds of LTBI positivity (index quintile 5 vs 1; 1.17, 1.12–1.23). Diabetes was associated with higher odds of LTBI positivity, while odds of testing positive were lower for those with a history of organ transplant, use of high-dose steroids, immunosuppressants, dialysis, leukemia, and HIV. The odds of testing positive for those with a history of hepatitis B (HBV) or hepatitis C (HCV) infection were increased (1.45, 1.34–1.57 and 1.54, 1.44–1.66, respectively). Birth in a country with an elevated TB rate (3.40, 3.31–3.49), >30-day travel to a TB endemic country (1.74, 1.62–1.86), and prior exposure to TB (4.19, 3.91–4.50) were associated with increased odds of testing positive. Variance inflation factors for all variables (<5) and the AUROC (0.76) were acceptable.

Sensitivity Analyses

Results from sensitivity analysis that included individuals with an LTBI diagnosis code only (without an accompanying LTBI test) identified the same factors associated with LTBI testing and LTBI positivity (Supplementary Table 3).

DISCUSSION

Overall, our findings suggest that the current LTBI testing strategies may not be reaching the populations with the highest risk while disproportionately reaching lower-risk groups. Individuals without risk factors are routinely tested due to regulatory requirements for employment, schools, and congregate settings. Factors associated with higher odds of testing (Black race, higher income, Medicaid, immunosuppression, HIV, travel to a country with an elevated TB rate, and exposure to active TB) did not overlap with factors associated with higher odds of testing positive (male sex, 18- to 35-year age group, Asian/Pacific Islanders, Hispanics, lower income, smoking, diabetes, HBV/HCV infection, and birth in a country with an elevated TB rate). More importantly, for risk factors recognized by the current LTBI screening criteria, such as birth in or >30 day travel to a country with an elevated TB rate, testing rates were low, despite their high positivity rates.

We observed differences in the demographic and socioeconomic characteristics associated with LTBI testing and test positivity. Although the 35- to 50-year age group had the highest odds of testing positive for LTBI, younger adults (18–35 years) had the highest odds of testing; the oldest age group (≥75 years) was least likely to be tested. LTBI testing in younger groups as a requirement for schools and participation in extracurricular activities as well as preemployment screening may provide some explanation. While there are no specific age-related guidelines for LTBI testing, older adults may be perceived as a lower TB risk group and thus tested less frequently [18]. These observations are consistent with US meta-analysis on LTBI diagnosis and treatment, suggesting that older age is associated with lower screening rates and LTBI treatment referral/initiation [19]. Medicaid beneficiaries were more likely to test compared with those covered by Medicare, commercial insurance, or uninsured; these patterns are consistent with previous US reports [18]. At KPSC and elsewhere, programs for Medicaid beneficiaries with disabilities and/or special needs such as adult day care require LTBI testing. We also observed that males had lower odds of LTBI testing but higher odds of testing positive compared with females, which may be related to differences in health-seeking behaviors. LTBI testing as a condition of employment for occupations such as educators and healthcare professionals such as nurses with female predominance may provide additional explanation. Hispanics and Asian/Pacific Islanders were generally more likely to test positive but not necessarily more likely to test for LTBI compared with their non-Hispanic White counterparts. Disparities in TB burden in racial/ethnic minority groups are well documented [20]. Social determinants of health that lead to disparities have drawn attention in many disease areas including cardiovascular diseases and cancer, but attention to disparities in TB prevention is limited [12].

Comorbidities that are known risk factors for TB, including organ transplant, immunosuppression, dialysis, leukemia, and HIV, were associated with higher odds of LTBI testing but lower odds of testing positive, which is encouraging. These conditions are associated with a high risk of progression to active TB [21], and LTBI testing is warranted. Individuals with such conditions are more often tested for LTBI than the general population, and the higher testing volume may contribute to the lower test positivity. Higher testing rates in individuals receiving immunosuppressants may also be driven by LTBI testing recommendations prior to therapy initiation. However, individuals with diabetes were less likely to be tested for LTBI, and those with HBV or HCV infection were not necessarily more likely to be tested despite their higher testing positivity. Higher prevalence of LTBI and TB has been reported in the setting of HBV/HCV infection compared with the general population [22, 23]. Recent data from California have shown high test positivity in individuals with diabetes and HBV infection [18], and LTBI/HBV coinfection risk was highest among persons born in high-incidence countries and among Asians and Blacks [24]. These patients were also less likely to be prescribed LTBI treatment after testing positive compared with patients without HBV infection or diabetes [18]. Although HBV and HCV have been commonly reported as TB risk factors, they are not recognized by the current guidelines. Efforts to reduce morbidity and mortality associated with coinfection with TB and HBV/HCV will require healthcare systems to screen and treat individuals at risk.

Although LTBI testing is recommended for asymptomatic adults who were born or resided in countries with increased TB prevalence [9, 10], testing rates in individuals who were born in or traveled (>30 days) to such countries were low (14%–27%), despite their high LTBI positivity rates (18%–25%). The high TB burden in non-US born persons compared with US-born persons has been consistently reported [25]. Information on factors such as place of birth, travel to a country with an elevated TB rate, and prior exposure to TB being associated with LTBI positivity is not routinely collected or documented in primary care settings where initial TB screening often occurs.

Our work has some limitations. First, some TB risk factors were identified at any time during the study period, regardless of the timing relative to LTBI testing. This was to avoid heavy loss of information by defining risk factors only at the beginning of the study period and to maximize identification of risk factors across the study period where multiple LTBI tests could be ordered. Second, we excluded those with a LTBI diagnosis code without an accompanying LTBI test during the study period. Some individuals with LTBI could have been missed if they tested positive outside of KPSC prior to KPSC enrollment. However, sensitivity analysis including those with a diagnosis code alone was consistent with the main analysis. Third, data on country of birth were missing for more than half of the study population, and we used imputed values using previously published methods [17]; the distribution of country of birth with and without imputed data was balanced. Fourth, among those tested for LTBI, 82.6% were tested only with TST and 17.4% with either IGRA or TST, with fewer IGRAs performed earlier in the study. Due to the lower specificity (eg, affected by prior Bacille Calmette-Guérin vaccine) and sensitivity of TST, misclassification of LTBI is possible, especially in the earlier years when TST was used more frequently. Last, odds ratios are relative measures of effect, which may be statistically significant due to large sample size and thus should be interpreted jointly with the actual percentages presented in Tables 1 and 2 to better understand clinical relevance.

Overall, we observed that although some groups at high risk of LTBI are being appropriately tested, the current LTBI testing practices are missing important high-risk groups (eg, non-US born). Most TB cases result from reactivation of longstanding LTBI [26], representing a missed opportunity for prevention of active TB by appropriately testing and treating those at the highest risk. Routine testing of low-risk populations is not recommended and may result in unnecessary evaluations and treatment [27]. Nevertheless, many low-risk individuals will continue to be tested as a requirement for employment or congregate settings. Thus, increasing testing in high-risk individuals who are missed by the current testing practices is a priority. Further, routine assessment for TB risk factors, some of which are poorly documented in current clinical practice such as place of birth and travel history, is needed to more appropriately test high-risk populations. Further, other important risk factors such as HBV/HCV infection and race/ethnicity are readily available yet not considered in current screening guidelines. Additional work is needed to refine guidelines that are feasible to operationalize and more appropriately target populations with the highest LTBI risk as well as to identify strategies to support guideline implementation.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author Contributions. Concept and design: J. H. K., H. F., L. X. Q., J. S., K. J. B., S. Y. T. Acquisition, analysis, or interpretation of data: J. H. K., H. F., L. X. Q., K. L., J. S., K. J. B., B. J. L., S. Y. T. Drafting of the manuscript: J. H. K. Critical revision of the manuscript for important intellectual content: J. H. K., H. F., L. X. Q., K. L., J. S., S. S., K. J. B., B. J. L., S. Y. T. Statistical analysis: H. F., L. X. Q., K. L. Obtained funding: S. Y. T. Administrative, technical, or material support: S. S., B. C. S. Supervision: S. Y. T. All authors approved the final version of the manuscript. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. All authors take responsibility for the overall content and conduct of the study and the decision to publish.

Acknowledgments. The authors thank the members of Kaiser Permanente Southern California for helping to improve care through the use of information collected through our electronic health record systems.

Disclaimer. The findings and conclusions presented here are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC).

Financial support. This study was funded by the National Institutes of Health (NIH; grant 5R01AI151072). The travel encounter data development was funded through the Vaccine Safety Datalink from the Centers for Disease Control and Prevention.

References

Author notes