Predictors of latent tuberculosis treatment initiation and completion at a U.S. public health clinic: a prospective cohort study

All adults (>17 years) who met CDC guidelines for LTBI treatment between January 11, 2008 and May 6, 2009 at Wake County Tuberculosis Clinic in Raleigh, North Carolina were included in the study. Written informed consent was waived by the Duke Institutional Review Board, as this was considered an evaluation of a public benefit program, and thus eligible for such a waiver under the United States Code of Federal Regulations part 45 46.116(c)(1). The study was also approved by the Wake County Human Rights-Consumer Affairs and Human Research Committee.

On entry to the TB clinic, a routine medical interview was conducted, followed by a self-administered multiple choice questionnaire addressing social factors and attitudes toward LTBI treatment (Table 1). The questionnaire was developed based on the results of a focus group study assessing TB-related knowledge, attitudes, and behaviors among key populations in the region [18]. The questionnaire was validated by administering it to a pilot population of individuals taking isoniazid therapy at baseline and one month later; in this group test-retest response concordance was >0.75.

All LTBI medications were provided free of charge by the health department from an on-site pharmacy, which recorded all medication pickups in a computerized system.

Primary endpoints were 1) LTBI therapy initiation, defined as picking up at least one month of medication, and 2) therapy completion, defined as picking up nine months of INH therapy within a 12-month period or four months of rifampin within a six-month period. Univariate analysis was performed with survey responses, demographics, and medical risk data, using chi square test or Fischer’s exact test for categorical variables as appropriate. For multi-level categorical variables, pairwise testing was performed with chi square Univariate analysis was also performed to examine the relationship between the risk that an individual would transmit TB in the future and the likelihood of a) initiating and b) completing LTBI treatment. Subjects were placed in three risk categories, which were based on a combination of medical risk (risk of progression to active TB) and social risk (transmission risk due to suboptimal healthcare utilization). Subjects with history of transplant, HIV, silicosis, intravenous drug use, close contact to an infectious TB case, or with evidence of old tuberculosis on chest radiograph were placed in the “High risk” category. Persons with diabetes, end-stage renal disease, history of gastrectomy, homelessness, or being underweight were placed in the “Moderate risk” category, and all others were considered “Low risk.” Backward elimination was used to arrive at a final log binomial model consisting of independent variables significantly associated with completion of LTBI therapy at p < 0.10 by univariate analysis plus any significant confounders. Significant confounding was defined as a change in the risk ratio for a second variable of interest of >10% when the confounder was included (vs. not included) in the model.

Residences of study participants were geocoded and mapped using ArcMap 9.3 GIS software. The point density analyst tool was used to calculate distances between each residence address and the Wake County Health Department. Mean distance from the health department was compared between those who initiated and completed therapy and those who did not, using the student’s t-test. Residence zip codes were linked to percentage of area residents living below the poverty line in each zip code, based on Census 2000 data. Zip codes with less than 5% of the population living below the poverty line were categorized as “Low poverty,” those with 5-10% below the poverty line “Moderate poverty” and those with greater than 10% below the poverty line “High poverty.” Univariate relationships between these three categories and LTBI therapy initiation and completion were assessed.

The 496 participants were predominantly foreign-born (65%) and racial/ethnic minorities (87%), with mean age 39.1 + 12.3 years (Table 2). Of the 496, 130 (26%) persons initiated LTBI therapy and 70 (54% of those initiating) completed therapy. Of those who were foreign-born, the most frequent regions of origin were Africa (19%) and Latin America (20%). 61% of persons included in the study were referred after a tuberculin skin test (TST) was performed as part of employment screening, and 19% received a TST as part of a contact investigation. The overall cohort included a number of former or current smokers (32%), drug users (9%), and persons with history of incarceration (14%) or homelessness (11%).

In the univariate analysis, factors significantly associated with treatment initiation (P < 0.10) included older age, African birthplace, close contact with an infectious TB case, non-employment reason for TST, end-stage renal disease (ESRD), immunosuppressed status, HIV infection, previous incarceration, residence in current home for greater than one year, plans to remain at same home greater than one year, lower educational level, having a regular physician, and fear of getting sick with tuberculosis (Tables 1 and 2). In multivariable analysis, close contact with a TB case, non-employment reason for screening, lower educational level, having a regular physician, fear of getting sick with TB, and prior incarceration were independently associated with treatment initiation (Table 3). Since there were few (n < 20) persons with specific medical risk factors (ESRD, immunosuppressed status, HIV), medical and social risk were collapsed into three risk strata as described in the Methods section. More persons at “high risk” of progressing to and/or transmitting active TB initiated LTBI therapy (54/120, 45%) than those at “moderate risk” (24/69, 35%) or “low risk” (52/307, 17%)(P < 0.01).

Among those initiating treatment, factors associated with treatment completion (P < 0.10) included African or Asian birthplace, absence of tobacco history, prior incarceration, and plan to tell friends/family about positive TST. In multivariable analysis, planning to tell friends/family about positive TST was independently and significantly associated with treatment completion (RR 2.0, 95% CI 1.0-3.9). The proportion of persons at high risk of progressing to and/or transmitting active TB completing therapy was not significantly different from those at moderate or low risk (p = 0.6).

Of those persons initiating therapy, 31 were prescribed rifampin, with a 61% completion rate, and 99 persons were prescribed INH, with a 52% nine-month completion rate (p = 0.3). At least six months of INH was completed by 62/99 (63%) persons. Seven persons were switched from isoniazid to rifampin, and four of these completed a full course of rifampin. One person was switched from rifampin to isoniazid and that individual completed a full course of isoniazid. Compliance with both therapies declined over time (Figure 1).

In geographic analysis, 391 residential addresses were mapped with ArcGIS software. Of the 130 who initiated therapy, 99 participants had addresses that could be mapped to a physical location in North Carolina. Distance from the health department was not significantly different between addresses of persons who initiated LTBI therapy and those who did not initiate therapy (p = 0.4). Similarly, treatment completion was not significantly associated with distance from the health department (p = 0.9). Of all participants, 378 participants had zip codes with poverty data available from the 2000 U.S. census. Treatment initiation did not vary significantly by whether a participant lived in a high, moderate, or low-poverty level area (p = 0.6). Of those who initiated therapy, there was a trend toward greater completion of therapy by persons living in a low-poverty area (81% of those initiating completed therapy), compared to those living in a moderate (49% completed) or high poverty area (42%), at p = 0.08.