All records reviewed as part of this study are routinely collected and documented by RNTCP and NACP in India. Participation in interviews was entirely voluntary and we obtained written informed consent from all adult participants or from a parent of participants aged less than 15 years. Interviewers were blinded to the HIV testing status of participants. A total of three interview attempts were made for patients who could not be located initially. The entire study protocol was reviewed and approved by Institutional Ethics Committee of National TB Institute, Bangalore, India and Ethics Advisory Group of the International Union against TB and Lung Disease. Approval to conduct the study was obtained from Central TB Division and the State Health department.

This was a cross sectional study and the study procedures encompassed a mix of quantitative and qualitative data collection methods: 1) record review (from both TB and HIV programs) for all patients registered for TB treatment in South Parganas 24 District, 2) review of records documenting TB/HIV training for individual health personnel of the district, 3) review of HIV test-kit stock registers, 4) interview of a subset of patients with TB and 5) interview and focus group discussion (FGD) with TB care providers and HIV care providers.

South 24 Parganas District (population 7.7 million) is primarily a rural district of West Bengal with a HIV sero-prevalence of more than 1% among antenatal women [6]. Prevalence of HIV among people with TB is not known. TB control program services are available through a decentralized network of 102 general health care facilities. In addition, patients also frequently utilize health facilities in the city of Kolkata, located in an adjacent district, and subsequently complete treatment at one of the health care facility in South 24 Parganas districts. In accordance with RNTCP guidelines [7], the 102 South 24 Parganas facilities are grouped into 15 sub-district level TB program management units (TU) so each TU has 4 to 12 health care facilities. Patients with TB are registered within one month of treatment initiation, at one of the 15 TUs and are treated with standardized treatment regimens administered under direct observation. TB treatment consists of intensive phase (IP) for 2 to 3 months and continuation phase for 4 to 5 months. Each TU has a TB register maintained by a TB treatment supervisor to document details of patients started on anti-TB treatment. Register data also include date and place of diagnosis, date and place of treatment for TB, HIV status, and outcome of treatment. Each TB treatment supervisor works under the supervision and guidance of a Medical Officer of TB Control (MOTC). Of 102 health institutions that provide TB treatment services, 14 have a co-located HIV testing center called the Integrated Counseling and Testing Centre (ICTC). Each of the remaining 88 health institutions is affiliated with one of these 14 ICTCs in order to provide access to HIV testing services for their patients. Each ICTC is staffed by a counsellor and a lab-technician and is monitored by the NACP. HIV testing consists of one rapid test which if found positive is followed by two confirmatory rapid tests [8]. Details of each client (name, age, sex and address) attending the ICTC are documented in an identification register referred to as the Person Identification Digit (PID) register. The HIV test results are documented in a separate ICTC register using the unique PID number for each patient from the PID register.

The study was conducted during the period July 2010– March 2011.

The primary outcome variable was failure to be tested for HIV by the end of IP of anti-TB treatment. National guidelines recommend referral for HIV testing as soon as possible after diagnosis of TB; hence, in the present study, the time limit for defining the outcome variables was the end of the IP of treatment (i.e. 2 months for new cases and 3 months for retreatment cases). Multiple methods of data collection were used including the following quantitative and qualitative components:

Quantitative data from the data collection sheet and patient interviews were double-entered into Epi-Info software version 3.5.3 (Center of Disease Control and Prevention, Atlanta, USA), by independent data entry operators. Databases were compared and inconsistencies resolved through referral to the original data collection sheet. Statistical analysis was performed using SPSS software version 13.0. Relative risks with 95% confidence intervals (CI) were calculated for association between each exposure variable and the outcome of interest (failure to be tested for HIV within IP). Chi square (x²) tests were used for comparing proportions and p values <0.05 were considered statistically significant.

We used inductive analysis for qualitative data obtained from provider responses and FGDs (i.e., categories of analysis were not imposed a priori on the data but were identified through the analysis process) [9]. Provider perceptions about why patients with TB were not being tested for HIV during the IP were noted on spreadsheets. Similar answers were coded manually by the principal researcher. These codes were then grouped into emerging themes [10].

Of 1,651 patients registered for TB treatment, 18 (1%) had documentation of HIV test performed within six months prior to the time of TB diagnosis (i.e. had known HIV status) and were excluded from the study. Of 1,633 patients with TB included in the study, 1,555 (96%) were adults (≥15 years), 1167 (72%) were males, 1,375 (84%) lived in rural areas, and 1,356 (82%) did not have history of previous anti-TB treatment. A total of 1195 (73%) patients were diagnosed with TB in primary health centers, 325 (20%) in sub-divisional hospitals and 113 (7%) in medical college hospitals (Table 1). In addition to the health facilities in South 24 Parganas District, patients were also diagnosed with TB at twenty five other health facilities in the neighboring districts. Of 1,633 patients with unknown HIV status at the time of TB diagnosis, 425 (26%) were tested for HIV by the end of IP. Of these, 362/425 (85%) were tested within the first 15 days of initiation of treatment. Of all patients tested for HIV, 8 (2%) had a positive result.

Relative to patients diagnosed with or treated for TB at a facility with collocated ICTC, risk of failure to be tested for HIV were higher for patients diagnosed with or treated for TB in facilities without a co-located ICTC (RR 1.27, 95% CI 1.20–1.35). Age, sex, positive sputum test result and history of prior TB treatment were not associated with failure to receive testing for HIV (Table 2).

The RNTCP district program manager, NACP district program manager, all MOTCs, all TB supervisors at the 15 TUs, and all counselors of 14 ICTCs were found to have been formally trained using the standard TB/HIV training modules. Other medical officers and healthcare workers were not trained formally but were informed of the national policy of ‘routine offer of HIV testing for all patients with TB’ in their regular monthly meetings. From July through September 2010, there were no test kits available at 6 of 14 ICTCs for periods ranging from 10 to 20 days. However, at the district level, there were more test kits distributed to ICTCs than were used.

Of the 204 patients randomly selected, 169 (83%) were interviewed, 25 were unavailable even after three attempts, and 10 died before they could be interviewed. Of 169 patients interviewed, six were children under 15 years so we interviewed the parent who accompanied the children during the TB diagnosis process. There was no statistically significant difference between patients who were interviewed compared to all patients enrolled by age group, sex, completion of HIV testing during IP, residence type or history of previous treatment (data not shown). Of 169 patients interviewed, only 67 (40%) reported that they were referred for HIV testing and of these, 53 (79%) went to an ICTC. Of the 53 patients who went to an ICTC, 47 (89%) were tested for HIV and 6 (11%) were not. Of the 47 patients tested for HIV, 25 (47%) reported that they were counseled prior to being tested. Of 67 patients who reported that they were referred for HIV testing, 41 (61%) reported they were not informed of the purpose of the test by the TB care provider.

Of the six patients who reported that they went to ICTC after being referred for testing but did not get tested, two (33%) reported that they did not get tested because the ICTC was closed and four (66%) reported that there were no counselors to conduct HIV counseling or testing. Of the 14 persons who were referred but did not go to ICTC, eight (57%) reported they thought that the test was not necessary, four (29%) reported that the ICTC was too far and two (14%) reported that they were told by the TB program staff to go for HIV testing at another time because testing kits were not available but did not go for testing at a later date. Of these 14 persons referred to ICTC but who did not go, 3 were referred from a facility that was co-located with an ICTC. In total, 122/167 (72%) of interviewed patients were not tested for HIV during the IP of anti-TB treatment (Table 3).

Among patients who reported that they were referred for HIV testing, patients who reported being referred by a DOT worker had higher risk of not undergoing HIV testing relative to patients who reported being referred by a Medical Officer (RR 3.53, 95% CI 1.62–9.57). We found that among patients referred by DOT workers 12/29 (41%) were exposed to co-located facilities compared to 22/38 (58%) of patients referred by Medical. Patients referred for HIV testing during anti-TB treatment were less likely to get tested relative to patients referred at the time of diagnosis (RR 2.91, 95% CI 1.19–7.10), (Table 4). Patients’ marital status, literacy, occupation and income were not significantly associated with failure to get HIV tested. Among patients diagnosed with TB, 43% (34/79) of patients diagnosed in a facility with a co-located ICTC were referred for HIV testing compared to 36% (33/90) of patients diagnosed at a facility without a co-located ICTC.

Key informant interviews were conducted with 16 MOs and four DOT providers. More than half (13/20) of providers reported awareness of the policy to refer all patients with TB for HIV testing and 35% (7/20) reported that program managers of both the national TB control and AIDS control programs provided minimal supervisory support. Almost all providers (17/20) believed that patients would not undergo HIV testing even when recommended by a provider. Most providers (15/20) reported that they had not been formally trained using standard-HIV training modules, and that they do not regularly refer all patients with TB for HIV testing. We learned that providers do not refer for HIV testing because they believe that ICTCs are not co-located with TB facilities, that they are too far for patients to travel especially in view of poor public transport facilities, and that ICTCs provided irregular service and frequently suffered from HIV testing kit shortages (Table 5). Providers felt that very few patients were likely to be HIV positive and therefore questioned the rationale for referring patients for testing.

In FGD, TB treatment supervisors indicated that they had received training on TB/HIV guidelines but were not yet conversant with correct reporting procedures. They also shared concerns about the shortage of HIV testing kits and felt that physicians were apathetic towards referral of HIV testing for patients with TB. ICTC counselors indicated that they receive few referrals for HIV testing from TB treatment providers, minimal supportive supervision from both TB and HIV programs, and intermittent stock outs of kits.