The Institutional Review Board of Boston University Medical Campus approved this study and the surveys employed. Participation in the study was voluntary and not linked to treatment.

The study setting was a Boston, MA academic medical center's TB clinic. TB physicians and nurses provided TB screening, verbal and written patient education, including treatment risks and benefits (using interpreters as needed), and treatment of LTBI (1). Baseline serum transaminases were determined in all patients. An initial visit for usual medical care, including time in the clinic, Radiology, Pharmacy, and the Clinical Laboratory, typically lasted approximately 3 to 4 h. At follow-up visits with a nurse, patients were clinically assessed for toxicity and given a 1-mo medication refill, as well as additional education. These visits usually lasted between 30 and 60 min. Serum transaminases were monitored periodically in those at increased risk for isoniazid-induced hepatitis and in those over 35 yr of age. Patients referred from specific neighborhood health centers returned there to nurse-staffed TB follow-up clinics by pre-arrangement. While primary care physicians may have provided initial referrals, they were generally not involved in treatment of LTBI within this system.

Individuals 18 or more years of age prescribed at least 6 mo of daily isoniazid only for LTBI, and who spoke English, Spanish, or Chinese, were included in the study. Patients were excluded if they were prescribed medication other than isoniazid for treatment of LTBI, had treatment delayed, or were suspected to have or had TB disease. From July 14, 2002 to September 28, 2003, 217 subjects were voluntarily enrolled, with a refusal rate of eligible patients of approximately 30%.

The survey was pilot-tested before the study, measured 4.0 on the Microsoft (Redmond, WA) Word Flesch-Kincaid English readability scale, which uses the formula (0.39) (total words/total sentences) + 11.8 (total syllables/total words) − 15.59 to yield a number corresponding to a United States educational grade level. The survey was also translated into Chinese and Spanish. During the enrollment period, eligible patients were approached after the initial TB clinic evaluation and patient education had been concluded. There was no time limit in which to complete the written survey, but typically respondents were finished within 10 min. Answers were kept confidential from clinic staff.

We used χ² analysis to determine significance of variables relative to the primary study endpoint, completion of 6 mo of isoniazid within 9 mo, using the EpiInfo, July 2002 (CDC, Atlanta, GA) statistical program. For multivariate regression analysis, the adjusted relative risks of completion were estimated using a log-binomial model with the GENMOD procedure in SAS version 9.1 (Cary, NC).

Surveys were completed by a total of 217 individuals: 148 in English, 22 in Spanish, and 47 in Chinese. The respondents were 90% foreign-born and included a substantial proportion of racial minorities (Table 1). There were more women among survey participants (54.4%, versus 44.2% men). A majority of participants were under the age of 35 (52%), with a median age of 34 yr. Participant demographics were similar to those of patients with LTBI who were offered treatment at this clinic (4). The most common TB risk factor was foreign-birth in a TB endemic country. Approximately one third of participants received their follow-up at neighborhood health centers.

A total of 29.0% completed at least 6 mo of isoniazid, with 19.4% completing the full 9-mo course and 9.2% completing between 6 and 9 mo of treatment. Nearly two-thirds (65.9%) of the participants self-discontinued treatment either by declaration or by not returning to the clinic within 3 mo. Another 5.3% had treatment discontinued by a TB provider due to adverse events, including itching, rash, fatigue, malaise, headache, dizziness, nausea, abdominal discomfort, and diarrhea.

Demographic and TB risk factors did not predict failure to complete treatment of LTBI. The completion rates by survey language were not significantly different.

Concern about venipuncture, which approximately half of respondents either did not want or were unsure if they wanted, was a significant predictor of treatment non-completion by univariate analysis (relative risk [RR], 0.63 [0.39–1], 95% confidence interval [CI], p = 0.048) and by multivariate analysis (odds ratio [OR], 0.43 [0.22–0.85], 95% CI, p = 0.015). This view alone accounted for 20% of non-completers and was a shared predictor among 18% of all non-completers (Table 2).

Univariate predictors of treatment non-completion included not having a primary care physician (RR, 0.48 [0.25–0.87], p = 0.013) and not valuing regular visits to a health care professional for health maintenance (RR, 0.27 [0.07–0.85], p = 0.019) (Table 3). Patients who perceived a low risk for developing active TB without LTBI treatment were significantly less likely to complete LTBI treatment (RR, 0.35 [0.18–0.67], p = 0.001). This was also true for those who completed only the English version of the survey (RR, 0.26 [0.10–0.62], 95% CI, p = 0.001). Multivariate analysis also identified low TB risk perception as a significant predictor of treatment non-completion (OR, 0.31 [0.13–0.72], 95% CI, p = 0.007). This view alone accounted for 37% of non-completers and was a shared predictor among 18% of all non-completers. The majority of respondents felt that they were not infected with TB, and believed that BCG vaccine would prevent the development of TB, but these views did not predict failure to complete treatment. Reluctance to have venipuncture, and perception of low TB risk, as both independent and shared predictors, accounted for 75% of non-completers.