Assessing proof of concept
Our results show that HIV providers can undertake these tasks as well as TB providers. During the study period nearly 30% of the patients who were screened for TB in the intervention clinics were screened by HIV service providers, showing that a relevant proportion of patients to be worked-up for TB, entered the system via HIV service providers. Until now, HIV service providers typically sent patients with presumed TB, i.e. patients who have been found to have symptoms of TB, to the TB service for TB testing and, if applicable, for treatment. Our study shows that HIV service providers can test, and if indicated also treat, patients with TB as well as the TB providers.
The percentage of unsuccessful tests can be an indication for the providers’ technical capacity to make use of GeneXpert testing. HIV and TB providers had a similar proportion of unsuccessful tests (2.1% and 1.9% respectively), at the same level as the national average (1.9%) [
28].
It is a well-known phenomenon in TB care in South Africa that not all patients with presumed TB (presence of TB symptoms) progress to testing and treatment, due to challenges of patient adherence, as well as shortcomings on the healthcare system [
19]. We found similar percentages of patients ‘lost-to-follow-up’ in the pre-treatment care cascade, from screening to treatment initiation, of patients who were cared for by HIV and TB service providers. It is possible that these patients eventually tested, and if applicable initiated TB treatment, at another facility [
29]. As we could see from our own results, patient moving between different facilities is a common practice – nearly half of the patients who started TB treatment at our study sites were transferred in, having been diagnosed with TB at another facility (clinic, hospital or a private general practitioner). Such movements, however, especially in patients who move out without giving or being able to give notice to the provider in advance, are unfortunately difficult to trace [
19,
29,
30].
An interesting finding is that HIV providers in our study had more than 50% higher TB detection rate among the patients they tested for TB than the TB service providers. While TB service providers had a detection rate of 10.4%, which is equivalent to the national average [
29], the percentage of positive TB tests of HIV service providers was 16.1%. Although the overall TB detection rate varied between clinics, HIV providers had markedly higher detection rates compared to TB providers in all three study sites.
Differences between facilities in the overall detection rate may be explained by differences in the local TB epidemic and the characteristics of living conditions in the catchment areas of the three facilities. Clinic 1, where TB detection rate was highest, is situated in an area where most residential structures are shacks. The area is very crowded, and so TB prevalence can be expected to be higher. In contrast, Clinic 2 and Clinic 3 are located in rural and semi-urban areas, respectively. As most structures are permanent and neighbourhoods less crowded, TB infection rates are expected to be lower. Note, that we reviewed GeneXpert test results from the 6 months prior to the intervention and saw similar proportions of positivity to those we found during the study period.
The different detection rates among people tested for TB between HIV and TB service providers within the clinic may be due to differences in the characteristics of the patients typically seen in the two service areas. HIV service providers, for example, tested predominantly symptomatic patients, and only patients who are HIV positive, hence patients with a high pre-test probability of having TB. In contrast, the persons TB providers sent for testing included (i) asymptomatic patients (e.g. patients who had contact with a person with TB (‘household contacts’)); (ii) HIV-negative individuals (individuals known to be HIV positive are predominantly cared for by the HIV service providers); and (iii) patients who are symptomatic because of diseases other than TB (e.g. chronic respiratory disease). Overall, a group of patients in whom the pre-test probability of having TB is very likely lower than in symptomatic patients living with HIV.
Other critical aspects to consider when evaluating the integration of TB and HIV services, especially the integration of full TB services into the scope of work of HIV service providers, are quality of care and treatment outcome. We can assess quality of care by examining the proportion of patients who received microscopic TB tests (Auramine stains) at the end of the intensive phase (‘On-treatment test’) to assess sputum conversion as a prerequisite to scale down to continuation phase treatment; and at the end of the continuation phase (‘End-of-treatment test’) in order to assess cure. Our results show that the overall percentages of on-treatment tests across both service providers were similarly low. HIV service providers had only 44.7% of their patients tested, while TB service providers had 55.3% tested (difference not significant, p = 0.2549). Even if considering that not all patients were able to produce a suitable sputum sample anymore, because bronchial secretion had subsided as a result of the treatment, the figures are clearly too low. The situation looks much better with respect to end-of-treatment testing: While 75.4% of the patients treated by TB service providers had an end-of-treatment test, HIV service providers had 81.3% of their patients tested (difference not significant; p = 0.6169). These findings indicate that HIV service providers are performing similarly to the TB service providers, regarding monitoring TB treatment. However, improvements, especially regarding on-treatment testing, are urgently needed for all providers.
With respect to TB treatment outcomes, our study revealed that TB service providers performed better than HIV service providers (p = 0.0232). The proportion of patients with a successful treatment, i.e. patients who completed treatment without an end-of-treatment test, but with clinical improvement, or patients who had a negative end-of-treatment test and thus could be regarded as cured, was 68.3% for TB service providers and 53.3% for HIV service providers. Irrespective of any differences, completion/cure rates below 80%, are clearly not acceptable, as set by national policy. Several factors account for the overall low treatment success rates in the three study sites. First, high defaulter rates— the defaulter rate among patients treated by HIV service providers was 33.3%, more than 3-fold higher than among patients treated by TB service providers (9.9%, p = 0.0024). Second, high mortality rates. Although not statistically significant (p = 0.3904), mortality of 11.9% among patients treated by TB service providers was almost double the mortality of 6.7% among patients treated by HIV providers. Third, high percentages of patients transferred out, which affected both services equally (10%).
Our results, therefore, show that HIV providers are as capable as TB providers of screening their HIV-positive patients for TB symptoms, testing the ones with presumed TB, initiating TB treatment in those who are identified as having TB, and monitoring TB treatment. These results demonstrate that HIV service providers are technically and functionally capable of integrating TB services into the spectrum of services they provide.
The fact that the overall treatment success rates of both types of providers in our study are poor, and even worse where TB services are provided by HIV service providers, should not be used as an argument to oppose efforts of service integration. Rather, it reveals the limitations of service integration in improving treatment outcomes. The literature shows that service integration definitively contributes to achieving optimal treatment outcomes [
18]. However, treatment outcome is not determined only by the quality of services a patient receives at a facility. Factors related to the characteristics of the individual patient, and the population the patient is part of, may contribute as well. For example, the mobility of a patients, which is also partly linked to the responsiveness of the health system to deal with mobile populations and thus to the accessibility of health services in general, or health literacy which may impact on health-seeking behaviour and treatment adherence. Patients who are stably settled residents with good access to a local clinic, and who understand the need to seek healthcare early and adhere to any treatment for the entire course, even though the signs and symptoms of disease have vanished, would certainly benefit from an intervention that increases service quality at the local clinic. This is likely different for patients who only stay temporarily in the area, have difficulty accessing the clinic or services, or present themselves already in more advanced stages of TB. Further studies are required to assess factors that predict treatment outcome at the level of the individual patients and the populations they come from. Our study was not designed to determine this. When results of this study were shared with stakeholders at the provincial and district governments, attendees speculated that the high defaulter rates, especially among HIV patients, may be attributable to the fact that HIV patients in general have high default rates, because of the tiring lifelong treatment which may make them prone to stop taking their medications once they feel better.
Provider attitudes
The secondary objective of the study was to understand what the integration of TB into HIV services means for the staff and for clinic operations. At the start of the study, some TB providers were opposed to the intervention, and some even tried to sabotage the intervention. For example, we found from the records of the study coordinator, that providers in one facility purposefully misplaced TB forms from the HIV service areas. Such challenges improved over time, as observed when we interviewed providers, supervisors and program managers toward the end of the study period. When asked whether they agreed or disagreed with positive statements, responses were overwhelmingly positive; when asked about negative statements, the responses were mixed, and showed wider variations than the responses to the two positive statements. In other words, responses to the negative statements reveal that although the majority of respondents believed that service integration might not negatively affect the quality of HIV services, some concerns remain. Similarly, while the majority said that concerns regarding infection control are not an argument against the intervention, a number of respondents had reservations towards the intervention because of infection-control concerns.
These interviews showed almost uniform approval of the intervention, and a general opinion that service integration can improve health outcomes. These sentiments were echoed with stakeholders of all levels (facility, district, province, national) when results of the study were shared with them in dissemination meetings. Yet several critical issues emerged that must be addressed for successful integration of HIV and TB services, including scope of duties, administration, and infection control.
Adding TB testing and treatment to the services rendered by HIV providers means that they learn a new task that may increase their workload, if no adjustments are made, such as allocating more staff. As no additional staff were appointed or shifted to the HIV service as part of the intervention, indeed some staff complained of increased workload. Therefore, HIV and TB service integration should be accompanied by adequate human resource shifts. This is especially true in larger facilities, where staff tend to be more ‘specialized’, such that individual staff members are expected to cover only a certain section of the spectrum of services. This vertical distribution of services also allows for the appointment of less-qualified staff for specific tasks, where they can work under the supervision of a more qualified staff member. This is a common practice in larger facilities, and was the case in Clinic 2, where the TB focal person was not a qualified nurse. Therefore, service integration seems to have a larger impact on the individual service providers in larger facilities. The ‘specialized’ nurse may fear for his/her ‘specialist’ status and need to become competent in other service areas; less qualified staff may fear for their employability. Unavoidably, any change in operational practices can have negative impact on some staff. Key to getting staff support for service integration of HIV and TB services is to clearly communicate plans, address foreseeable positive and negative implications of the intervention, take concerns of the staff seriously, and respond promptly to any unforeseeable developments.
The increase in paperwork, which was criticized by some providers when interviewed, was indeed a concern in this study. It was mainly due to the fact that we could not get permission from the provincial office to introduce a second TB register for the HIV providers. Their concern was that this could result in duplicate records if a patient is registered in one side of the facility, then is treated by a provider from the other side. As a result, HIV providers had to record ‘their’ TB patient information on a separate notepad and then transfer the cases at the end of the day into the TB register, which was kept in the TB service areas. These registers are a critical monitoring and evaluation tool of the National TB-control program. If TB services are rendered at more than one service point in a clinic, each of these service points must be reliably linked. While we could not satisfyingly resolve this issue for this study, we believe that it should be possible to run more than just one register per clinic without affecting the quality of reporting. The use of electronic registers may also resolve this issue.
Infection control was mentioned as a challenge by one provider and emerged as a serious concern during the dissemination meetings, particularly at the district level. Stakeholders were understandably concerned about having symptomatic TB patients treated in the same area as other patients, especially potentially immunocompromised patients with HIV infection. The discussions further revealed that infection control is generally not managed and practiced well at the facilities currently. Therefore, for successful integration of HIV and TB services, infection control must be reviewed and improved across all facilities. Relatively simple infection-control measures, such as open windows or outside or otherwise well-ventilated waiting areas, should be practiced facility wide. Depending on the physical structure of the facility, another infection control measure can be to screen patients as soon as they enter the facility, at the triage or vital signs station, and prioritize those with a cough, or have them wait in a separate area [
31].
Limitations of the study
The study provides important insights on the feasibility of HIV and TB service integration but has some limitations. First, at the start of intervention implementation not all staff members followed the protocol equally well. We believe this was because they did not see this as part of their normal duties, and some may have feared negative consequences for their job. Through the appointment of local study assistants, regular and ad hoc visits by the study coordinator, repeated training and establishing close and trustful communication lines with facility staff, we were able to minimize non-adherence to the study protocol and procedures and conducted the study as intended.
Second, in order to investigate how HIV providers managed TB testing and TB treatment, we utilized routinely captured data from TB screening and case registers, routine data from NHLS, as well as patient-records. Although we believe that data quality overall was good, some information which we intended to use for more detailed analyses was not reliably recorded in the files, such as the presence and duration of TB symptoms. We also identified discrepancies between NHLS data and the TB screening registers and some data quality issues within the NHLS data set (e.g., misspelling of names and incorrectly captured dates of birth). These quality issues occurred only in a fraction of cases, were managed through comprehensive data cleaning, and did not influence our findings.
Third, because the project which funded the study ended, we could assess treatment outcome only on approximately 70% of all the patients whose records we followed. However, as we considered in the final outcome analyses only cases that were long enough on treatment to allow for completion of treatment within the study period, we believe that the essence of our findings regarding outcomes remains unaffected by the shortening of the data collection period.
Finally, this proof of concept study tested the feasibility of the intervention in only three public primary health-care facilities. Therefore, generalizing our findings to other settings should only be done with caution. Nevertheless, we strongly believe that our findings are of great value to the health sector in South Africa, especially to guide the ‘integrated chronic disease management’ which is part of the ‘Ideal Clinic’ programme and a major part of the South African health sector reform. Our findings may also be relevant to other countries with similar settings and a high burden of TB, HIV, and TB/HIV co-infection.