Towards national systems for continuous surveillance of antimicrobial resistance: Lessons from tuberculosis

In a Perspective on the research article from Jacobson and colleagues, Amitabh Suthar and colleagues from the Centers for Disease Control and Prevention discuss the importance of and considerations for developing real-time and large-scale reporting systems for tracking and controlling antimicrobial resistance.

diagnostics are being developed and deployed and may affect DST technologies that are registered, centrally procured, and nationally distributed across laboratories and/or health facilities. Linking national diagnostic algorithms to essential diagnostic lists can facilitate national access to DST [7]. In the absence of this linkage, there could be discordance between national policy and access, which would introduce geographical biases in continuously generated data. Third, DST coverage within national health insurance benefit packages, as part of universal health coverage, may affect uptake. Specifically, if DST is not covered by benefit packages and requires out-of-pocket payments, vulnerable and poor populations could decline this service and introduce selection bias into continuously generated data.
WHO currently recommends universal DST in its End TB Strategy [8]. Unfortunately, in 2016, only 26 of 48 surveyed countries reported a national policy and algorithms indicating universal access to DST, and globally, only 41% of bacteriologically confirmed new and previously treated cases were tested for at least rifampicin resistance [9]. Upon identifying resistance to rifampicin, recommended diagnostic algorithms assess susceptibility to additional drugs to determine whether the TB strain is multidrug resistant (i.e., MDR TB) or extensively drug resistant (i.e., XDR TB) [6,10,11]. Since 2011, the South African TB diagnostic algorithm begins with smear and/or culture to confirm Mycobacterium tuberculosis complex; if positive, screening for rifampicin resistance is evaluated with Xpert MTB/RIF and is confirmed using line probe assay [5]. Information on out-of-pocket expenditures required for DSTs as part of this algorithm was not specified by McIntosh and colleagues. Nonetheless, overall, McIntosh and colleagues reported that 49% of confirmed TB cases were tested for rifampicin resistance [5]. Further expansion toward universal eligibility, access, and uptake of DST is key to maximizing patient coverage.

Information completeness
National health information systems facilitate DST data collection, transmission, storage, and analysis. Communication between laboratory information systems and electronic health records helps ensure DST results are linked to an individual patient record. Unfortunately, recent studies have called into question the completeness and reliability of the laboratory and electronic surveillance systems. It is possible to learn from TB in South Africa [12][13][14]. McIntosh and colleagues leveraged the National Health Laboratory Services database in Western Cape Province [5]. Since this database does not contain unique patient identifiers, a deduplication algorithm utilizing name, surname, age, location, and sex was used to link specimen results to individual patients. In this study, 2.1% of specimens were removed from the South African analysis because an identifier or location was missing, and a further 15% of specimens were removed from geospatial analyses because the specimens were submitted from nonclinic locations [5]. Countries may need to determine the minimum data elements required to meet their pathogen's surveillance objectives.

Laboratory performance
Robust laboratory quality management systems are the cornerstone of accurate and reliable DST results. Such systems encompass the complete spectrum of diagnostic services from collection of specimens, storage in proper conditions, transportation to laboratories, laboratory testing, and result reporting. Recognizing the critical role of laboratory quality for appropriate patient care and disease surveillance efforts, WHO recommends that by 2020, all laboratories conducting TB testing are enrolled in external quality assessment (EQA) programs for all tests they conduct, all National Reference Laboratories are accredited to ISO 15189 standards, and sites performing DST are deemed proficient by EQA panel testing [15]. Developing national competencies, training the workforce to instill them, and monitoring the performance of laboratory systems are important means to achieve this target for TB and to achieve similar standards for other pathogens [15]. In recent years, a number of tools and resources have been developed to guide TB laboratory quality monitoring and improvement [16]. Enhanced global support and commitment, however, are required to sustain achievements to date and address noted weaknesses in DST laboratory performance to achieve WHO targets and ensure provision of accurate test results [17]. McIntosh and colleagues reported that 0.1% of specimens were excluded because they were indeterminate, suggesting that a high level of laboratory performance is possible.

Surveillance approaches
Given the potential for selection, information, and measurement bias in continuous surveillance data systems, disease programs and countries will be left wondering whether they can make policy decisions on the basis of continuously generated data or whether surveys are also needed. Surveys can be robust and detailed and provide nationally representative data but can be logistically complex and expensive and require external technical support [6]. Many countries conduct nationally representative drug resistance surveys for TB [6]. South Africa is one country that is transitioning away from TB drug resistance surveys after national scale-up of DST [9,18]. The continuous TB drug resistance surveillance system in Western Cape produced a rifampicin resistance estimate similar to that generated in a recent national survey (4.6% versus 4.3%, respectively) [19]. However, only approximately 47% of diagnosed TB cases in the continuous system met the minimum requirements for analysis (i.e., specimen collected for DST, minimum data collected, and valid laboratory result available). If there are biases in the data from other provinces, ongoing surveillance in a sample of nationally representative sites (i.e., continuous sentinel surveillance) may be an interim solution to generate data while continuing to build a national system for continuous surveillance (Table 1).

Future outlook
Standardized AMR tools can help low-and middle-income countries ascertain biases in different surveillance approaches and guide them to an appropriate selection based on their pathogen(s) of interest, objectives, and context. This guidance is available for TB [6]. On the other hand, not all surveillance approaches have been fully described for HIV, malaria, and priority bacterial pathogens [20][21][22]. DST in clinical care also plays a key role in system development, as routinely generated genotyping data can be leveraged for national systems for continuous surveillance when DST is clinically indicated. In addition to potentially informing national decisions, continuous surveillance has increased geographical granularity. McIntosh and colleagues were able to detect a resistance outbreak in a district and characterize districts with higher resistance due to poor health system access or higher migration of drug-resistant bacteria [5]. This novel use of spatial analysis helps local programs "see" where AMR resistance is accumulating and monitor trends over time, providing a potentially powerful tool to guide control and prevention efforts. This use of AMR surveillance data may need further encouragement and promotion. Moreover, routinely generated genotyping data have been used to characterize sources of transmission and guide prevention efforts [23]. Having these data available as an early warning could allow swift action and mitigate perpetuation of AMR strains. Therefore, regardless of whether a final policy decision is taken based on continuously generated data, more countries may find it worthwhile to explore the use of continuous AMR surveillance as part of national strategies to contain AMR.