Gridlock from diagnosis to treatment of multidrug-resistant tuberculosis in Tanzania: low accessibility of molecular diagnostic services and lack of healthcare worker empowerment in 28 districts of 5 high burden TB regions with mixed methods evaluation

For the fourth consecutive year, tuberculosis (TB) is the leading killer from an infectious disease worldwide, and despite curative antibiotic treatment, remains the leading cause of death for people living with human immunodeficiency virus (PLWH) [1]. The confluence of HIV and poorly integrated health systems in Tanzania, not unlike other countries in sub Saharan Africa, has fueled a resurgence in TB including multidrug resistant (MDR)-TB [2]. MDR-TB is defined as in-vitro resistance of Mycobacterium tuberculosis to at least isoniazid and rifampicin and is associated with worse treatment outcomes, higher mortality and longer treatment durations than drug-susceptible TB [3]. The World Health Organization (WHO) declared MDR-TB as a public health crisis with need to scale up diagnosis of MDR-TB, increase the number of patients initiated on MDR-TB treatment and control the ongoing transmission [4]. Strategies for scale up include universal drug-susceptibility testing and delivery of this laboratory capacity closer to the point of care, with initial focus on those patients at higher risk for MDR-TB such as those co-infected with HIV or with a prior history of TB treatment, so-called “retreatment” cases. While critical to implement, these strategies place a significant burden of agency on the front-line healthcare worker to recognize the need for novel diagnostics, access and utilize novel diagnostics effectively, and support patients to bridge to proper treatment. In many situations, this burden represents an unfunded mandate.

The WHO estimates that for the Africa region only 48% of incident cases of MDR-TB were actually detected in 2014, and among those detected in the same year, only 68% were started treatment [1]. Prior to 2012, the health systems in Tanzania relied on smear microscopy for TB diagnosis, and only in rare cases was MDR-TB established by conventional culture-based phenotypic methods which took up to 3 months for results and which were rarely available for practical clinical care. In 2009, MDR-TB treatment was made available through a national referral process, and molecular diagnostics were simultaneously being commercially distributed and endorsed by the WHO and other stakeholders. Excluding assays in the research setting, current molecular diagnostics available for clinical care in Tanzania are the Xpert MTB/RIF (Cepheid, USA), and the Genotype MTBDRplus (Hain Lifescience, Germany). The Genotype MTBDRplus screens for both isoniazid (inhA and katG mutation) and rifampicin (rpoB mutation) resistance and requires at least 2 days for completion. The Xpert MTB/RIF assay can be completed in 2 h and requires debatably less technical expertise, but only detects rifampin resistance [5].

Although these molecular diagnostics have desired accuracy and capacity to produce rapid results for clinical action, in other settings outside of Tanzania the effect has neither been transformed into rapid initiation of treatment nor improved patients outcomes [6‐8]. Using 2016 Global Burden of Disease data investigators have recently attributed 5.0 million excess deaths in low and middle income countries were attributable to receipt of poor-quality care and 3.6 million were due to non-utilization of existing healthcare services [9]. We hypothesized that a lack of improvement in TB outcomes in Tanzania was similarly symptomatic of poor quality care and non-utilization of existing healthcare, and that understanding implementation barriers faced by front-line healthcare workers may produce remediable solutions [10, 11]. This mixed-methods study was part of a nationwide series of investigations supported by the WHO that we have termed “gridlock from MDR-TB diagnosis to treatment,” to inform strategies for optimal use of molecular diagnostic technology to control the TB epidemic and reshape the healthcare workforce.

Twenty-eight districts (51%) were randomly selected from the 55 districts to participate in the study. Only 11 (39% of selected districts) had laboratories with capacity for performing MDR-TB diagnosis with molecular diagnostics and were therefore included. Thus, 28 DTLCs, 28 DACs and 11 Laboratory Managers were eligible. Of those eligible front-line healthcare workers, 64 (96%) participated; including all 11 (100%) of the laboratory managers. A cohort of presumed MDR-TB patients that received services in the previous year either at a TB or HIV clinic was summarized.

This study examined three important settings within the healthcare system: TB clinics, HIV clinics and laboratories with MDR-TB diagnostics at the front-line of patient care in Tanzania. In so doing, we identified factors that potentially influenced both a lack of diagnosis and a delay in diagnosis of MDR-TB (Fig. 4). We uncovered widespread underuse of molecular diagnostics for the two specific cohorts at higher risk of MDR-TB and for whom international consensus recommends the use of molecular tests, those with retreatment TB and PLWH. The majority of front-line healthcare workers, as represented by DACs and DTLCs surveyed, demonstrated opportunity for knowledge improvement on the use of molecular diagnostics, and knowledge was more deficient for GenotypeMTBDRplus compared to XpertMTB/RIF, and for both technologies in DACs compared to DTLCs. While these knowledge gaps are not dissimilar from recent studies of front-line healthcare workers in other sub-Saharan African countries and elsewhere regarding molecular diagnostics for TB [17], the scale of effect was profound. Only 40% of previously treated TB cases were tested for MDR-TB, lower than the global estimate of 60% [1], and when accounting for the those tested but for whom results were not communicated to the clinic, only 30% of all cases of presumed MDR-TB received the opportunity for the most accurate and effective treatment, which is exceptionally low compared to other settings [18].

Previous proportions of retreatment patients without drug-susceptibility testing in Tanzania were reported as high as 90% from a 9-year audit in 2013 [19], so it can be inferred that availability of molecular diagnostics has slightly improved this testing pattern. Yet we believe the rigor of our standardized data collection methods among randomly selected districts from the 5 highest burdened regions is most reflective of current practice patterns and workforce capacity. As such, the lack of diagnostics utilization was even more concerning when restricted to HIV/AIDS clinics where the vast majority of presumed TB cases were screened using conventional smear microscopy [20], despite well-defined WHO recommendations to use XpertMTB/RIF for improved sensitivity of TB detection and MDR-TB diagnosis [21]. Front-line healthcare workers were further compromised in their ability to deliver the recommended diagnostics secondary to unreliable laboratory services as machines were frequently out of service or cartridges or other consumables out of supply, as well as a lack of standardization and support for the specimen referral mechanisms. Our findings suggest that the current health systems have been slow to implement international standards of TB care surrounding the use of molecular diagnostics and goals to diagnose and treat MDR-TB, which have likely led to ongoing transmission, and without considerable structural change in the healthcare workforce and massive resource influx will never see the benefit of these technological advances [22].

We also noted that even in the cases of successful MDR-TB diagnosis, 20% were lost to follow-up or died prior to treatment initiation. While data for this observation were representative of only one of the regions studied, the median time from results communicated to the clinic to referral for treatment was 5 days (IQR 1–15). This time to referral should not result in early death, and we suspect rather that death was influenced by extensive disease from delay in presentation and delay in ultimate performance of the diagnostic test. We previously reported that extensive lung destruction was associated with a higher early mortality in the first cohort of MDR-TB treated cases [23].

Unfortunately, the majority of DTLCs and DACs were unable to stipulate why molecular diagnostics were not used more frequently, and we interpret this to indicate a generalized lack of ownership over this aspect of the health system. From our local findings, we postulate that initiatives to deliver continuing medical education, such as in the form of modules available on the web and accessed through mobile devices, and linking this training to professional recertification, would vastly increase uptake of new technology. Currently there are no formal training or demonstrations of knowledge required for professional recertification in Tanzania. We also propose transfer of molecular diagnostics ownership and maintenance to public-private partnered management as responsibilities for machine maintenance and consumables provisions are borne by the end-users. When this strategy has been used in other TB endemic settings, diagnostic uptake has been increased and sustained [24].

In addition to proper use of existing molecular diagnostics machines, reflective of the quality of services, we observed, as others have, a considerable need to scale up access, reflective of the quantity of care, as the availability of diagnostics was present in only 39% of the surveyed districts [25]. While the original target was for one XpertMTB/RIF machine in every district in Tanzania, this quantity of machine is still below the number of machines needed to efficiently screen the number of patients with retreatment TB or PLWH and presenting with TB symptoms [26]. Given the current lack of availability and proper use of molecular diagnostics in Tanzania, it remains very difficult to study the benefit of rollout of these technologies to-date. Nevertheless, this type of implementation study may be more important than ever given our observations regarding equipment maintenance and servicing and stock out of consumables, which in at least one case interrupted operation for up to 5 months.

While the retrospective review of programmatic records has inherent limits such as missing information and recall bias, the former was infrequent as exemplified by consistent reporting of demographics and HIV status among retreatment TB cases. We attempted data extraction from records from TB and HIV/AIDS clinics, as well as laboratory records, which also minimized incomplete data. However, some clinical information, such as the number of prior episodes of TB treatment or the number of times a patient had accessed care or diagnostics prior to the time period studied, were not available but may have informed the egregiousness of lack of testing for MDR-TB or the further systems errors in repeating microscopy despite negative or non-diagnostic prior testing. We followed the STROBE guidelines for reporting of results with the intent to provide a policy guide for stakeholders in Tanzania [27]. However, our sampling strategy failed to capture any laboratories or districts utilizing GenotypeMTBDRplus which limited lack of interpretation of its use and may explain the lack of knowledge by the DACs and DTLCs with this assay.

The majority of cases at higher risk for MDR-TB missed the opportunity of early screening, diagnosis and treatment. Identified bottlenecks in this process include a lower level of knowledge of molecular diagnostics for TB among front-line healthcare providers, poor access to molecular diagnostics based on the estimated number of retreatment TB cases and PLWH in the high burden regions surveyed, unreliable equipment maintenance, and inconsistent availability of supplies and consumables. These elements point toward health systems barriers that further remove the patient from the center of TB care decision making, as is advocated in the End TB Strategy [28]. Add to the major epidemic of TB/HIV, the emerging non-communicable disease interaction of TB and diabetes mellitus, and the cycling infectious epidemics such as dengue and cholera, we forecast that health systems in Tanzania responsible for TB care will be ill-equipped to realize the benefit of molecular diagnostics without large-scale changes [29, 30].