Linkage to Care and Treatment for TB and HIV among People Newly Diagnosed with TB or HIV-Associated TB at a Large, Inner City South African Hospital

Objective To assess the outcomes of linkage to TB and HIV care and identify risk factors for poor referral outcomes. Design Cohort study of TB patients diagnosed at an urban hospital. Methods Linkage to care was determined by review of clinic files, national death register, and telephone contact, and classified as linked to care, delayed linkage to care (>7 days for TB treatment, >30 days for HIV care), or failed linkage to care. We performed log-binomial regression to identify patient and referral characteristics associated with poor referral outcomes. Results Among 593 TB patients, 23% failed linkage to TB treatment and 30.3% of the 77.0% who linked to care arrived late. Among 486 (86.9%) HIV-infected TB patients, 38.3% failed linkage to HIV care, and 32% of the 61.7% who linked to care presented late. One in six HIV-infected patients failed linkage to both TB and HIV care. Only 20.2% of HIV-infected patients were referred to a single clinic for integrated care. A referral letter was present in 90.3%, but only 23.7% included HIV status and 18.8% CD4 cell count. Lack of education (RR 1.85) and low CD4 count (CD4≤50 vs. >250cells/mm3; RR 1.66) were associated with failed linkage to TB care. Risk factors for failed linkage to HIV care were antiretroviral-naïve status (RR 1.29), and absence of referral letter with HIV or CD4 cell count (RR1.23). Conclusions Linkage to TB/HIV care should be strengthened by communication of HIV and CD4 results, ART initiation during hospitalization and TB/HIV integration at primary care.


Introduction
South Africa is one of the 22 countries with a high tuberculosis (TB) burden [1]. In 2009, almost half a million new cases of tuberculosis were notified in South Africa, [2] corresponding to a rate of 971 cases per 100,000 population. The burden of HIVassociated tuberculosis in South Africa is enormous, with an estimated 73% of TB cases being HIV-infected, and one in four of all global HIV-associated TB cases occurring in South Africa [1].
Adherence to treatment for TB and HIV is an important determinant of a favourable treatment outcome, and those who default treatment, especially if soon after treatment initiation, continue to be a source of infection and transmission within the community [3,4]. Patients with drug resistant TB that default therapy pose an even greater risk to the public [5,6]. It is estimated that a person with smear-positive TB will infect 10-14 people per year [7]. The risk of transmitting HIV is mainly related to the HIV viral load and level of risk behaviour [8]. While treatment interruption during the course of TB treatment and antiretroviral treatment (ART) have been extensively studied [9][10][11][12][13], little is known about the linkage to care after the initial diagnosis [14], particularly among people diagnosed with HIV-associated TB at a hospital facility [15,16]. Because the cost-effectiveness of diagnostic interventions is in part determined by successful linkage to care [17,18], the rate of linkage to care should be monitored and evaluated along with other programmatic targets [18,19]. Currently, rates of and reasons for failure to link TB and HIV diagnosis with care are not collected as part of routine monitoring and evaluation in TB and ART programs, making evidence-based decisions on how to improve linkage to care difficult [20].
We aimed to determine the proportion of people who fail to link to care for TB and HIV treatment following a TB diagnosis at an inner city hospital in Johannesburg, South Africa, and assess risk factors for delayed or failed linkage to TB and HIV care.

Ethics statement
Study approval was obtained from the Human Research Ethic Committee of the University of Witwatersrand, Institutional Review Board of the University of North Carolina, the City of Johannesburg (CoJ) municipality and the Helen Joseph hospital. A waiver of written informed consent was granted by the ethics committees (retrospective review of existing medical records). Verbal informed consent for use of information provided by individuals contacted by phone was obtained from all individuals and documented in writing on the informed consent form used. All data was de-identified prior to analysis and analysed anonymously.

Setting and patient population
The TB focal point (TBFP) at the Helen Joseph public hospital in Johannesburg, South Africa, provides routine TB diagnosis for inpatients and some outpatients, TB education, and HIV counselling, testing and staging for TB suspects and TB patients. All activities are performed according to National TB guidelines [21,22]. During the study period, the South African guidelines recommended ART initiation as soon as the patient is stable and completed at least two weeks of TB treatment if CD4 count is less than 50 cells/mm or in the presence of other serious HIV ''illnesses''; at the end of the two-month intensive phase of TB therapy if CD4 count is between 50 cells/mm and 200 cells/mm; and after completion of TB therapy if CD4 count was above 200 cells/mm and there were no other HIV-related symptoms.
Patients are referred for continuation of TB treatment to the most appropriate local primary health care clinic based on the patient's home or work address, and provided with TB medication for 7 days. Patients on ART are provided with antiretroviral drugs for 30 days and ART-naïve HIV-infected patients are instructed to present for HIV care and treatment within 30 days of referral. National guidelines state that if possible, the patient should either be transported to the clinic by health services, be accompanied by a DOT supporter or social worker, or the local PHC clinic should collect the patient. If this is not possible the hospital should follow up directly with the clinic to confirm that the patient has arrived [21,22]. Because the TBFP does not have the capacity to accompany patients or to confirm arrival of all patients, a standard South African TB Programme referral letter (''pink'') form is given to the patient to deliver to the referral clinic. A copy of this referral letter is sent directly to the clinic where the patient is referred to, and a third copy is kept at the TBFP. The guidelines further instruct that the pink referral form must be completed in detail with all relevant information. [21,22] While the form includes a field ''other information'', there is no dedicated space for HIV status or CD4 count.
All adults (age 18 years or older) registered with active TB at the TBFP with TB between October 1 and December 31, 2009 and referred to a clinic within the CoJ health district were included in the analysis.

Data collection
The TBFP routinely collects socio-demographic data, clinical information, laboratory results related to TB and HIV care, patient contact information and the name of the clinic(s) patients are referred to for TB and HIV treatment.
For the study, data on the outcome of the referral process was collected 3 months (median 192 days, range 161-231) after referral from TBFP to ensure capture of late presenters (Figures 1  and 2). Referral clinics were visited to confirm patient arrival by interviewing the TB nurse and reviewing the clinic TB register. The patient TB file was reviewed to determine the exact date of patient arrival, standard referral form for documentation of HIV and CD4 cell count information, locate the pink form or any other referral documents, and review documentation on HIV status, care and treatment. For patients with HIV-associated TB, TherapyEdge TM an electronic database, used by several HIV clinics in Johannesburg, was reviewed to establish linkage to HIV care. CoJ clinics that did not use Therapy Edge-HIV TM were visited. In total, study staff visited 98 clinics within CoJ.
When documentation of patient arrival at the referral clinic for TB and HIV treatment was not found, the electronic South Africa's National Vital Registration system was consulted to verify the alive or deceased status of the patient. Patients that were alive or with unknown survival status were contacted telephonically using a standardized script to enquire about TB and, if applicable, HIV treatment. If patients reported they had presented for TB and/or HIV treatment, the name of the clinic was recorded and the clinic was visited to confirm attendance.

Data analysis
The primary outcome was time to arrival at the referral clinic, categorized as (i) linked to care, (ii) delayed linkage to care (iii) failed of linkage to care. Delayed linkage to care was defined as presentation for TB treatment more than 7 days or presentation for HIV care and treatment more than 30 days after hospital discharge, corresponding to the 7 day supply of TB treatment, 30 day supply of antiretroviral drugs, or a 30 days period during which the ART-naive patients were asked to present for HIV care. Failed linkage to care included patients who self-reported not being in care as well as patients who had no record of care and could not be traced by phone, assuming they had not presented for care. Initiation of care in people who could not be traced could not be excluded. We therefore explored the potential of misclassification in a sensitivity analysis in which all patients who could not be traced were classified as having linked to care. Patients who died within 7 days or 30 days of referral for TB or HIV care, respectively, were classified as deceased during referral process.
Multiple log-binomial regression analysis was performed to identify patient and referral characteristics independently associated with delayed and failed linkage to TB and HIV care, respectively. Measures of association were expressed as relative risks (RR) and their 95% confidence intervals (CI). Variables associated with the outcome in bivariate analysis at p#0.20 were considered in regression models. Data was analyzed using SAS 9.1 statistical software package (SAS institute, Inc., North Carolina, USA.

Patients characteristics
During the 3-month study period, 678 adults were registered with active TB at the TBFP, of which 593 were referred to a CoJ clinic and included in the analysis. Demographic and clinical characteristics of patients referred to CoJ clinics did not differ from the overall cohort (p.0.2). Median age was 35.9 years (IQR 30.6-43.5), 50% were female, 82.6% were of South African nationality, 80% had an education level of grade 8 or higher, and 54.7% were unemployed ( Table 1

Linkage to TB care
Overall, 137 (23%) patients failed to link to TB care and 456 (77%) presented to a clinic for continuation of TB care and treatment ( Table 2). Among the 456 who had linked to care, 126 (27.6%) had presented to a clinic other than the one they had been referred to by TBFP staff. The median time to presentation was 5 days (IQR 2 -10) but 138 (30.3%) arrived late, after their 7 day supply of TB drugs expired. Among the 415 patients who arrived for TB care at a CoJ clinic, 375 (90.3%) had a clinic file that contained a referral letter but this letter contained the HIV status in only 23.7% of patients, and the CD4 cell count was communicated in as few as 18.8% of HIV-infected TB patients. Among the 137 (23%) patients classified as failure to link to TB care, 6 (4.4%) had died within the first 7 days after TBFP referral, 38 (27.7%) self-reported they had never presented at a TB clinic and 93 (67.9%) could not be traced.

Linkage to HIV care and treatment
Among the 515 patients with HIV-associated TB, 486 were referred to a CoJ clinic and included in the analysis. Of the 486 patients, 186 (38.3%) failed to link to HIV care and 300 (61.7%) presented for HIV care and treatment ( Table 2). Among those that linked to HIV care, 66 (22.0%) had presented to a clinic different than the one they had been referred to by the TBFP staff. The median time to presentation was 17 days (IQR 5-48), but 32% presented late, more than 30 days after referral by the TBFP. Among the 186 patients classified as failed linkage to care, 10 (5.4%) had died within the first 30 days of referral, 82 (44.1%) admitted they had never presented at a HIV clinic, and 94 (50.5%) could not be traced.
The majority of patients (79.8%) had been referred to two separate clinics, one for TB treatment and one for HIV care and treatment. Among ART-naïve patients, 12% were inappropriately referred to a clinic not accredited for ART initiation.  Correlation of linkage to TB and HIV care and treatment (in patients with recorded arrival times)

Sensitivity analysis
Despite intensive investigations, 15.7% of patients could not be traced regarding linkage to TB care and 19.3% could not be traced to obtain information on HIV care. In the primary analysis, these patients were assumed not to have linked to care, which may have overestimated the rate of failed linkage to care. A sensitivity analysis in which all patients who could not be traced were classified as having linked to care, gave similar results compared to the main analysis. Being a non-South African (RR 1.30 95% CI 0.64-2.64), lack of formal education (RR 1.32 95% CI 0.61-2.88) and low CD4 count (#50 vs. .250cells/mm 3 ; RR 3.29 95% CI 0.42-25.97) resulted in a higher risk of failed linkage to TB care, but the association was no longer statistically significant, possibly due to the lower numbers of patients who failed to link to care in this analysis. Regarding linkage to HIV care, being ART-naïve at referral (RR 5.47 95% CI 2.06-14.49) and referral to an inappropriate HIV clinic (RR 1.93 95% CI 1.25-2.98) were associated with failure to link to HIV care in the sensitivity analysis; being a non-South African (RR 1.08 95% CI 0.69-1.70) and lack of HIV status and CD4 count on the referral letter received at the TB clinic (RR 0.94 95% CI 0.64-1.38) were no longer associated with failing linkage to HIV care in the sensitivity analysis.

Discussion
We observed inadequate referral outcomes despite the implementation of a TB focal point infrastructure with a strong emphasis on integrated TB/HIV diagnosis. Almost one in four (23%) of patients failing to link to TB treatment at a primary care clinic, and 30% of those linking to care arrived after their 7 day supply of drugs expired. Linkage to HIV care was even more inadequate, with 38.3% of HIV-positive patients failing to present themselves to HIV care. Among HIV-infected TB patients, only 55.5% achieved linkage to both TB and HIV care, and one in six failed linkage to both TB and HIV care. Our findings are in contrast to the observations by Edginton et al [16] who observed that the introduction of a TB care centre at another large hospital in Johannesburg, South Africa, improved patient referral with 93% of TB patients achieving successful referral to a clinic for continuation of TB treatment and nurses reporting that patients arrived ''well-informed and motivated to continue treatment with referral documentation that were accurate and complete'' [16].
Most socio-demographic characteristics, including age, gender and employment status, were not associated with poor outcomes of the referral process. Lack of formal education resulted in a twofold risk of delayed and failed linkage to TB care, and being non-South African increased the risk by about 40%. As observed elsewhere, staff may not spend enough time on patient education and may not be able to communicate in the patient's native language [16]. In a Ugandan study, not being aware that TB is curable and not knowing the duration of anti-TB treatment was associated with poor linkage to TB treatment in patients with HIV-associated TB [23]. Foreign-born individuals and those who lack formal education may find it especially challenging to interpret information provided at referral and may be less empowered to seek continuation of care.
High rates of losses to care between HIV diagnosis and HIV care have been observed in many settings, highlighting the need for robust referral systems and a better understanding of risk factors [10][11][12][24][25][26][27]. In our cohort of patients with HIVassociated TB, being ART-naive at time of referral was associated with a 30% increased risk of failed linkage to HIV care, and around a 40-50% increased risk for delayed linkage to HIV and TB care, although the latter was not statistically significant. In contrast to Basset et al [20], who found that patients with a CD4 cell count #100 cell/mm 3 were twice as likely to be lost to care before ART initiation, CD4 count was not a risk factor for failure to link to HIV care in our population of HIV-infected TB patients. The increased risk of failure to link to HIV care among TB patients who are ART-naïve at time of referral highlights a lost opportunity for reducing early mortality among those with advanced immunosuppression [28][29][30] and suggest that initiation of ART during hospitalization, may not only improve survival but also reduce the likelihood of failure of linkage to care at primary care level. The efforts towards an integrated TB/HIV approach at the hospital TBFP resulted in a 95.8% HIV testing rate and 84% of HIV-positive patients being evaluated by CD4 cell count. Unfortunately, the information on HIV status reached the primary care providers in only 23.7% of patients who linked to TB care, and the results of CD4 cell count were available at primary care level for only 18.8% of HIV-infected TB patients who linked to TB care. Failure to communicate HIV-related information to the health care worker, often because of confidentiality concerns, negatively impacts on patient care, as demonstrated by a 23% increased risk of failure to link to HIV care, and results in great inefficiency and high costs of duplicating efforts at different levels of the health care system. Our findings highlights the need to communicate HIV status and CD4 count at time of referral and suggest that national programs should consider revising the standard referral document if this does not contain designed entries for HIV-related information.
Integration of TB and HIV care at primary care level was poor, with 80% of patients with HIV-associated TB being referred to two separate clinics, one for TB and one for HIV care. The Pro-TEST study in urban Zambia demonstrated that delivery of integrated services, including a strong referral system reduces patient costs and is financially feasible and inexpensive [17].
Our study had several strengths, including the assessment of linkage to both TB and HIV care in a single cohort of patients, and extending data collection from the conventional review of registers or files at the referral clinic to review of an electronic HIV database, national death register, and telephone contact with patients who could not otherwise be traced [31][32][33][34][35]. The observation that 28% of patients who linked to care choose to go to a clinic other than the one they were referred to for continuation of TB treatment demonstrates the importance of this approach. Our study also had several limitations. First, even though the sample size was large, some estimates were imprecise due to small cell numbers in covariate strata. Second, despite intensive investigations, exact dates of arrival were missing in up to 9% of patients, 15.7% of patients could not be traced regarding linkage to TB care and 19.3% could not be traced to obtain information on HIV care. In the analysis, these patients were assumed not to have linked to care, which may have overestimated the rate of failed linkage to care. The results of a sensitivity analysis, in which these patients were classified as having linked to care, however suggested that potential misclassification did not substantially influence our conclusions. In conclusion, the proportion of patients that failed to link to TB and HIV care was unacceptably high despite efforts to integrate TB and HIV care and streamline referral for patients diagnosed with TB at the hospital. The high rates of HIV co-infection and the high degree of immunosuppression in the population highlight the urgent need for strengthened linkages to and integration of care. Factors associated with linkage to care were predominantly health system related and could be improved through appropriate interventions. Efforts to improve linkage to care could include ART initiation during hospitalization to reduce the proportion of TB patients who are ART-naïve at time of referral; improving the quality and content of patient education and adapting referral information to the patient's level of education to overcome failure of linkage to care among those who lack formal education; discussing referral clinic options with the patient instead of simply assuming that the preferred clinic from the patient perspective is the one closest to their home; referring patients to a single clinic for integrated TB and HIV care; ensuring that ART-naïve patients are referred to a clinic that is accredited to initiate ART; and changing the standard reference referral letter to ensure that information on TB, HIV status and CD4 cell count is shared.