Tuberculosis has been declared a global public health emergency. An estimated one-third of the world’s population is infected with tuberculosis; 10 million people developed tuberculosis disease in 2017 alone,¹ a number that may be underestimated due to under-reporting and lack of reliable data. The widespread implementation of the Xpert® MTB/RIF assay (Cepheid, Sunnyvale, California, United States) for detection of Mycobacterium tuberculosis and rifampicin resistance as a first-line tuberculosis diagnostic, has been hailed as the most significant advancement in decades, becoming the first molecular assay to provide a tuberculosis and first-line drug resistance diagnosis in just 2 hours. Following widespread adoption of this technology, the World Health Organization’s Agenda for Action on Digital Health for the End Tuberculosis Strategy called for 100% of all sites using rapid tuberculosis diagnostic instruments to be connected by 2020,² becoming the first to recognise the role of digital health in the fight against tuberculosis. Over the past 2 to 3 years, numerous countries have begun adopting connectivity platforms to help monitor and manage their GeneXpert fleet by collecting the vast amounts of rich clinical diagnostic and operational data produced by the instrument. Rarely before has such a rich data resource been both produced by a diagnostic instrument and been made available via connectivity platforms, at scale, for analysis. As yet, it remains largely untapped.³ If these data can be analysed, interpreted and translated into appropriate recommendations and actions, they have the potential to provide significant transformative impact on the management and effectiveness of infectious disease programmes worldwide.

GxAlert® (SystemOne, LLC, Northampton, Massachusetts, United States) is currently collecting data from GeneXpert platforms in 43 countries running Cepheid’s Xpert MTB/RIF assay. GxAlert is a connectivity platform that integrates directly with diagnostic instruments to collect and send a digital copy of test results and associated instrument metadata to an in-country or GxAlert server. From there, results can be sent and accessed via short message service and email alerts, Microsoft Excel (Microsoft Corp, Redmond, Washington, United States) reports and web dashboards. The type of data being collected includes not only the diagnostic result, but also information on when and where the test was run and by whom, demographic information about the patient (through an application called GxConnect), reagent lot numbers, probe data, cycle thresholds as well as instrument operational data such as instrument failures, inventory consumption and instrument downtime. From these data, critical insights can be gained or inferred about the tuberculosis programme and can help shed light on both clinical and operational return on investment. Data can also provide useful information on testing coverage, disease status and trends, circulating strains and drug resistance profiles, instrument utilisation rates, training needs, supply chain, inventory, and quality of the testing programme.⁴

But there is a challenge: even though countries now collect this type of data in large volumes, it is a new arena for them. Most high-disease burden countries lack the tools, resources and expertise required to analyse, understand and translate these data into improved programme and patient outcomes. A recent study by the Foundation for Innovative Diagnostics (FIND), found that despite large investments by donors to implement electronic data management systems, there is limited usage of the data to improve service delivery, mainly due to a lack of understanding and awareness of what data means.⁵ As a result, tuberculosis programmes are accumulating but not using the data being collected to drive decision making. This becomes apparent when one considers the various challenges still hindering tuberculosis programmes today, including gross under-utilisation of instruments,^3,6,7,8 high unsuccessful test and error rates (loss of tests),^9,10 cartridge stock-outs, instrument breakdowns, and lack of adequate module replacements and maintenance of instruments.¹¹

There is a dire need to build capacity in health data analytical skills amongst staff within national tuberculosis programmes (NTPs) in order to bolster the usage of data. To address this need, we designed a novel training programme to develop the expertise and skills required for the analysis and understanding of connected diagnostic data.

The TB Data Fellowship (TDF) programme was initiated in 2018 through a joint collaboration between SystemOne and Management Sciences for Health, with the support of the Tableau Foundation. Designed to build the foundation for sustainable in-country capacity, the programme aimed to enhance the understanding of tuberculosis data and its translation into actionable outputs. Achieving these goals required a new cadre of healthcare worker to be trained, one with the ability to understand and interpret the vast amounts of diagnostic and operational data being collected through connected diagnostic systems.

For the pilot programme, staff from the NTPs, national tuberculosis reference laboratories and ministries of health from several countries using the GxAlert connectivity platform were invited to apply. Countries invited included Bangladesh, Ethiopia, Ghana, Malawi, Mozambique and Nigeria. The selection criteria for the programme included a minimum of 2 years’ work experience in the field of tuberculosis, and more specifically the GeneXpert tuberculosis programme, and at least 6 months of work experience with GxAlert software. Participants also had to have experience working in Excel.

Two pilot training programmes were undertaken: the first accepted participants from five countries supported by the global Challenge TB project, funded by the United States Agency for International Development. The first countries were Bangladesh (n = 2), Ethiopia (n = 2), Mozambique (n = 2), Ghana (n = 1) and Malawi (n = 1). The second programme accepted participants from Nigeria only (n = 12). The training lasted 12 months and was structured in two parts: a 1-week in-person classroom training followed by an 11-month remote training and mentorship.

Through the 12-month programme, participants were taught how to understand GeneXpert tuberculosis diagnostic and operational data being collected by the GxAlert platform and to translate this data into insights about their respective tuberculosis programmes (Table 1). They identified programme gaps and appropriate intervention needs in different programme areas (Table 1), while learning how to lower operating costs (by reducing supervision frequency and troubleshooting services), improve the quality of the programme and manage the programme more effectively.

The training has yielded new country-level insights and programme improvements due to improved data use and decision making as demonstrated by numerous technical reports, conference abstracts and presentations. For example, in Bangladesh, participants have used data about instrument utilisation rates to influence the NTP to improve referral mechanisms for underperforming sites and further optimise GeneXpert placement within their NTP to better meet testing demand.¹² These insights have also helped the NTP plan for future placement of additional GeneXpert machines. In Ethiopia, analysis of instrument utilisation and subsequent proactive monitoring have led to an improvement in utilisation, from 28% to 75%.¹³ Such dramatic improvements can translate immediately into programme return on investment – whereby tuberculosis programmes can make existing resources go much farther than expected and deploy resources more effectively when receiving future grants or allocating domestic budgets.

By teaching participants how to monitor unreportable test rates or the number of tests resulting in errors, no results and invalid results, programme efficiency and response speed can be improved. Unreportable or unsuccessful tests do not provide a clinically valid result to the patient and thus need to be repeated. Besides the cost in ‘lost’ cartridges, when one considers that the actual cost per test performed has been estimated at $23.00 (United States dollors [USD]) and the cost per diagnosis at $99.00 USD,¹⁴ unsuccessful tests represent a significant cost to the health system. The majority of unsuccessful tests are due to error results and can, to a large extent, be corrected. Unfortunately, countries seldom know how to interpret error codes to inform appropriate corrective actions. The TDF helped participants categorise error codes according to their suspected sources and, through doing this, identify the most frequent types of errors to troubleshoot while pinpointing the sites needing supervision and follow-up. This real-time support is less costly compared to conventional monitoring, which requires a person to visit sites to troubleshoot issues, without any understanding of which issues pertain to which sites. Across all participating countries, the majority of errors were user or technical errors. These errors are associated with incorrect specimen processing or volumes added to the cartridge.¹⁵ For example, in Ghana, up to 67% of error results were identified as user related, and this insight has led to the introduction of refresher training and targeted supervision for sites.¹⁶ The same issue was identified in Nigeria and Bangladesh, where both programmes have managed to reduce their national error rates due to targeted supervision, refresher training for laboratory staff and regular feedback to laboratories aimed at addressing the high incidence of these user related errors.^12,17

Another focus area of the TDF training that has led to significant programmatic improvement is the monitoring of testing fleet and instrument downtime. A challenge faced by many tuberculosis programmes is that GeneXpert instruments are often located at remote facilities, leading to delays in maintenance and replacement of broken modules. By monitoring trends in how instruments report in real-time through connectivity tools, the Bangladesh NTP are now identifying directly when modules are down or instruments require calibration. Through this real-time monitoring of instrument performance, Bangladesh has managed to reduce instrument maintenance turn-around time from anywhere between 5 and 14 months to just 2 weeks, and is now also maintaining 90% module functionality.¹²

Various connectivity solutions exist to collect diagnostic data, some of which have already been adopted widely. Connectivity tools can play a major role in addressing many of the challenges that tuberculosis programmes face by facilitating the central collection and aggregation of diagnostic instrument data so that it can be analysed. However, the introduction of connectivity tools is not sufficient to ensure improved programme management. Well-functioning health systems need to utilise this data at all levels in order to drive evidence-based decisions and interventions to improve the quality of care provided.¹⁸

To our knowledge, the TDF programme is the first of its kind to build capacity and resources in low- and middle-income countries for the analysis of tuberculosis data collected from connected diagnostics. To address sustainability, the programme provided participants with the much-needed tools required to drive data analytics, counting on participants to lead the ongoing analysis of tuberculosis-related health data from the national GeneXpert programme in their respective countries. We chose Tableau software as a data analytics and visualisation tool, because it enables users to explore, manipulate and create visual representations of large amounts of data in order to produce insights as well as to communicate those insights to a broader audience. We leveraged an existing connectivity footprint, namely the GxAlert system (SystemOne, LLC, Northampton, Massachusetts, United States), to gain access to GeneXpert tuberculosis data within each respective country, but the programme is translatable to any connectivity platform collecting tuberculosis GeneXpert data. While the initial pilot programme focused on tuberculosis, the intent is to expand into related disease streams and diagnostics within the ministry of health, such as HIV. By providing an online tool (Tableau Online) where participants could post their visualisations and insights, the programme also provides a platform whereby countries can share best practices and help to create value.

The TDF programme has already seen rapid development and analysis of country key performance indicators leading to immediate publications, programme engagements and strengthening.^12,13,16,17 By using the data to recognise programme gaps and identify needs, issues and priorities, participants have been equipped to help develop their national strategies, address challenges and inform data-driven decision making.