Gaps in the care cascade for screening and treatment of refugees with tuberculosis infection in Middle Tennessee: a retrospective cohort study

We conducted a retrospective cohort study of refugees resettled in Middle Tennessee between January 1, 2012, and December 31, 2016, to characterize the TBI care cascade from 1) Initial TBI screening, 2) Follow-up evaluation for TBI positives, 3) Offer of TBI treatment, 4) Treatment initiation, to 5) Treatment completion. During the study period, all refugees relocated to Middle Tennessee were scheduled for an initial US health assessment at Siloam Family Health Center (Siloam) in Nashville, Tennessee (Supplementary Fig. 1). To assess factors associated with treatment initiation, we analyzed a subset of refugees who screened positive for TBI during their initial health assessment using a tuberculin skin test (TST) or interferon gamma release assay (IGRA) and were age 5 years or older at the time of resettlement. A positive screen was defined as a TST reaction of ≥10 mm of induration (≥5 mm if HIV positive) or a positive QuantiFERON (QFT) test [19]. Refugees were excluded from the analysis dataset if they 1) had TB disease at the time of evaluation at Metro Public Health Department (MPHD) in Nashville, TN; 2) had previously completed TBI treatment; 3) lived outside Davidson County; 4) were determined to have a false-positive screening test (i.e., a negative result on a follow-up TST or IGRA test); or 5) were advised by a provider not to start treatment. To assess factors associated with treatment completion, we excluded any refugees who did not start TBI treatment. Siloam and the Institutional Review Boards of Vanderbilt University and MPHD approved this study.

We used data from the Electronic Disease Notification System (EDN), a centralized electronic disease surveillance system used by the US Centers for Disease Control and Prevention (CDC) to identify all refugees who arrived in Middle Tennessee within the study period [20]. To identify refugees who screened positive for TBI, we used data from Siloam collected on the Tennessee Initial Refugee Health Assessment (TIRHA) form during initial screening visits. To determine treatment initiation, we abstracted data from the TB Risk Assessment Tool and the Patient Tracking Billing Management Information System (PTBMIS) at MPHD. To determine treatment completion, we abstracted TBI medication dosing data from PTBMIS at MPHD and used data from medical record review and the National Notifiable Disease Surveillance System (NNDSS) Base System (NBS), a data information system used to manage reportable disease data in Tennessee [21]. Because there was no unique identifier across databases, we used approximate string-matching techniques to match patients (Supplementary Tables 1A and 1B).

Eligible persons were treated for TBI with one of three regimens according to patient and provider decision: 1) 9 months of once-daily isoniazid (9H); 2) 4 months of once-daily rifampin (4R); or 3) starting in fall 2015, 3 months of observed once-weekly isoniazid and rifapentine (3HP). We defined treatment initiation as observed ingestion of the first dose of 3HP, or patient pick up of the first month’s supply of medication of 9H or 4R. We assessed treatment details using procedure codes in PTBMIS and by manual medical record review for details unable to be determined in PTBMIS. We assessed whether individuals met CDC guidelines for starting treatment within 60 days of screening positive for TBI and each individual’s treatment completion status 1 year (365 days) after the treatment start date. Treatment completion was defined as 1) having a “treatment complete” procedure code in the medical record, or 2) documentation of “treatment complete” by a healthcare professional in the medical record within 1 year from the start date.

We evaluated comorbidities in two ways. First, we estimated the overall comorbidity burden using the Elixhauser Comorbidity Index [22]. We assigned Elixhauser scores based on screening test results and ICD-9-CM codes recorded during the initial screening visit [23]. The study population was dichotomized into two groups: Elixhauser Score = 0 (No Comorbidity) or Elixhauser Score ≥ 1 (Any Comorbidity). Second, we evaluated four individual comorbidities including HIV, diabetes, liver disease and hypertension where the first three were associated with increased risk of TB disease [19, 24, 25] and the last one was the most commonly occurring comorbidity in the population. We also assessed age, sex, smoking history (any tobacco use: yes or no), alcohol abuse (excessive drinking of alcohol as determined by screening physician), type of screening test (TST or IGRA), distance travelled to TB clinic (straight line distance from home address to clinic), region of origin (created by categorizing countries of origin into regions based on the UN Standard Country or Area Codes for Statistical Use [26]), region of routing (created by categorizing countries where overseas medical exams were performed [26]), flu shot received at initial health assessment (yes or no), days to screening appointment from date of arrival in the US, year of arrival in the US, family size at time of entry (single, 2–3 persons, or > 3 individuals), and regimen (9H, 3HP, 4R or a combination). Geo-coded home addresses were used to calculate the straight line distance from home to the TB clinic by BatchGeo [27] to visualize the distribution and density for the study population.

We matched cases of TB disease from PTBMIS and NBS to the list of refugees screened for TBI at Siloam to determine whether any of them developed TB disease. We reviewed the medical record of all cases of TB disease that matched our screening dataset.

All data analyses were conducted using SAS Studio 3.6 (SAS Institute Inc., Cary, NC) and Stata 14.2 (Stata Corporation, College Station, TX, USA).

Of 6776 individuals who completed an initial health assessment in the 5-year study period, 1681 (25%) screened positive for TBI. Of these, we excluded 388 individuals from the analysis, and at the MPHD an additional 85 refugees were found to be ineligible for TBI treatment (Fig. 1).

The study population of 1208 refugees offered TBI treatment was made up predominantly of African refugees from Somalia, D.R. Congo, and Sudan as well as Asian refugees from Burma, Bhutan, and Iraq. Nepali, Somali and Arabic were the most commonly spoken languages and less than 3% reported English as their primary language. The median age was 31 years. About 34% of the study population were in the Any Comorbidity Group with the most common comorbidities being uncomplicated hypertension, obesity, liver disease, and uncomplicated diabetes mellitus (Supplementary Table 2). Individuals in the Any Comorbidity Group were significantly older than those in the No Comorbidity Group, were more likely to be smokers, to not have received influenza vaccination, and to have entered the country in a family of 2 or 3 (Table 1).

Of the 1208 refugees eligible for treatment, 690 (54%) started treatment (Fig. 2). Of these, 294 (43%) started TBI treatment within 60 days of screening positive. The most common recorded reason for non-initiation of TBI treatment was declination of treatment (55%). Other reasons included loss to follow-up (8%) and relocation (5%) (Table 2).

In the univariate analysis, male sex, having an IGRA screening test rather than a TST and having an initial health assessment in 2013 were significantly associated with increased treatment initiation. Older age, living further away from the TB clinic and having entered the country in a family of 2 or 3 were significantly associated with decreased treatment initiation (Table 3). We did not find a significant association between Any Comorbidity and No Comorbidity groups, but when we assessed individual comorbidities, hypertension was associated with treatment initiation.

In the adjusted model, male sex and an IGRA screening test were significantly associated with increased odds of treatment initiation (Adjusted Odds Ratio [aOR] _Male: 1.44; 95% Confidence Interval [CI]: 1.08, 1.92; aOR_IGRA: 2.79; 95% CI: 1.57, 4.93). Living farther away from the TB clinic was significantly associated with decreased odds of treatment initiation (aOR: 0.91; 95% CI: 0.83, 0.99). A higher percentage of non-initiators (50%) lived more than the median straight-line distance from the TB clinic than initiators (47%), however this difference was not statistically significant. We did not find a significant interaction between sex and family size or sex and region of origin.

Of the 690 refugees who started treatment, 432 (63%) completed treatment; 281 (41%) did so within a year of treatment initiation. Non-completion was due to loss to follow-up, declination of further treatment, relocation, and provider decisions to terminate treatment early (Table 3). In the univariate analysis, receipt of influenza vaccination during the screening appointment and having a region of origin in Southern or South-eastern Asia were significantly associated with increased treatment completion (Table 4). We did not find a significant association between Any Comorbidity and No Comorbidity groups, but when we assessed individual comorbidities, diabetes was associated with increased treatment completion. Additionally, we found a statistically significant nonlinear relationship between the duration from US arrival to screening appointment and treatment completion.

In the adjusted model, the associations remained significant (aOR _Diabetes: 7.27; 95% CI: 1.93, 27.30; aOR _Flu: 1.65; 95% CI: 1.14, 2.40; OR_SEAsia: 2.30; 95% CI: 1.43, 3.70; OR_SAsia: 1.64; 95% CI: 1.02, 2.64). We modeled the nonlinear association between days to screening appointment and treatment completion using restricted cubic splines (Supplementary Fig. 2). Adjusting for other covariates, the odds of completing treatment decreased for individuals who were screened earlier (e.g., 11 or 14 days after arrival into the US) compared to those who were screened later (28 days after arrival into the US) (aOR _{Screen (11 vs 28)}: 0.47; 95% CI: 0.26, 0.86; aOR _{Screen (14 vs 28)}: 0.48; 95% CI: 0.27, 0.86). Given this finding, we also included a non-linear term for days to screening appointment in the multivariable treatment initiation model, however this was not significant.

Of the 6776 individuals who were screened for TBI between 2012 and 2016, 21 (0.3%) developed TB disease during the study period (Supplementary Fig. 3). All 21 individuals who developed TB disease screened positive for TBI at their initial health visit Siloam. Of the 21 cases, 15 (71%) were diagnosed with TB disease at the follow-up clinical evaluation at MPHD, with a median time from US entry to TB disease diagnosis at MPHD (time to diagnosis) of 64 days. Of the remaining 6 cases, 4 individuals declined TBI treatment due to pregnancy (n = 3) or issues with transportation (n = 1) with a median time to diagnosis of 478 days. The other 2 cases were reported in individuals who had partially completed treatment on the 9H regimen (33 and 66% completed) with 602 days and 591 days to diagnosis respectively. One additional refugee developed TB disease 15 days after US entry and was referred directly to MPHD for treatment before having an initial health screen. There were no reported cases of TB disease among refugees who completed TBI treatment in our dataset.