This article has Open Peer Review reports available.
Assessing the sensitivity and representativeness of the Belgian Sentinel Network of Laboratories using test reimbursement data
© The Author(s). 2016
Received: 24 February 2016
Accepted: 25 May 2016
Published: 8 August 2016
The Belgian Sentinel Network of Laboratories (SNL) was created in 1983 in order to monitor trends in infectious diseases. Given the evolution of the surveillance system, such as the waivers, fusions and adhesions of laboratories over time, it is important to evaluate whether the SNL is still fit for purpose. This study aims to evaluate aspects of the sensitivity and representativeness of the SNL by means of a test coverage analysis.
We estimated test coverage of the SNL using the ratio of reimbursed tests performed by participating laboratories to the total number of tests performed between 2007 and 2012, for 12 (groups of) pathogens. We further evaluated the geographical difference coverage of the SNL at regional and provincial levels.
We found that test coverage of the SNL was stable over time and close to, or greater than, 50 % for the 12 (groups of) pathogens studied. These results hold for the three regions of Belgium but not for all provinces. We showed that some provinces had a low test coverage for some pathogens and that test coverage was more variable over time at provincial level.
This sensitivity and representativeness study based on test coverage suggests that the SNL is capable to describe trend and to monitor changes in the 12 (groups of) pathogens studied both at national and regional levels. Therefore, the SNL is useful to contribute to estimate the burden of disease and to inform preventive measures. It should however be reinforced to allow to be used as an alert system at provincial level.
In spite of the rising contribution of noncommunicable diseases to the global burden of disease, communicable diseases continue to threaten population health in high income countries. Public health authorities seek to evaluate this threat. In Belgium, a surveillance system for communicable diseases has been established in 1983 on the basis of a voluntary Sentinel Network of Microbiological Laboratories (SNL) . The Scientific Institute of Public Health (WIV-ISP) coordinates the SNL and weekly gathers laboratory-diagnosed cases reported by participating laboratories for more than 30 infectious diseases. The surveillance system aims to monitor trends in communicable diseases in order to be able to inform and alert health authorities on the current epidemiological situation. Its objectives include the uncovering of geographical and seasonal variations, the detecting of emerging threats , as well as the appraisal of the effectiveness of preventive measures such as vaccination .
The SNL has evolved over time. Since its creation, the surveillance system has adapted to the structural reorganisations, fusions, waivers and adhesions of laboratories; it has accounted for the use of new diagnosis methods such as PCR amplification; it has integrated European case-definitions; and it has encouraged laboratories to modernise their reporting method, moving from paper-based reporting to Electronic Data Interchange. Whereas some of the changes have had a positive impact on the quality of the surveillance, others might have altered the ability of the SNL to fulfil its missions. It is therefore important to regularly evaluate whether the surveillance system is fit for purpose .
To be able to detect changes in the epidemiological situation, the surveillance system should cover a suitably large and representative proportion of the Belgian population, and ensure that the population under study remains constant over time [5, 6]. These characteristics are embraced by two closely related attributes of surveillance systems: sensitivity and representativeness. In the context of surveillance systems evaluation, sensitivity, in addition to being a synonym of true positive rate, refers to the ability of a surveillance system to monitor changes in the number of cases over time and to detect outbreaks . Representativeness indicates whether a surveillance system accurately describes the occurrence of events under surveillance over time, person and place . Of particular importance is the ability to equally capture all infectious diseases under surveillance and to reflect non-homogeneous distributions of diseases across place. Some aspects of sensitivity and representativeness can be evaluated using coverage measures which indicate the proportion of the target population included in the surveillance system. Ideally, the coverage of a surveillance system should be constant over place and across pathogens to allow to detect regional outbreaks or specificities, and stable over time to allow meaningful interpretation of the time series.
The coverage of the SNL was annually monitored using the ratio of participating laboratories to the total number of laboratories. Data indicate constant ratios over the last 10 years, close to 60 %. The ratios were constant across Flanders, Wallonia and Brussels, which gives some indication of the geographical representativeness of the SNL. However, these figures are poor proxies of the coverage of the SNL because the laboratories have different catchment populations and greatly vary in terms of size and activity. A better measure of coverage was presented by Vandenberghe who analysed laboratory test reimbursement data to estimate test coverage of 15 pathogens. The study revealed that participating laboratories were performing more than 50 % of pathogen-specific tests, and that the test coverage was constant between 1999 and 2002 . The study also indicated that participating laboratories were more often connected to a hospital and tended to conduct more tests as compared to non-participating laboratories, but were similar with respect to other characteristics. It is unknown whether these differences could lead to systematic bias.
Given the evolution of the SNL since 2002, little is known on whether the coverage of the network remained constant over the last years. In addition no studies have evaluated whether the SNL was representative across place, beyond the three regions of Belgium.
This study aims to evaluate aspects of the sensitivity and representativeness of the SNL by analysing test coverage, its geographical distribution, its consistency across pathogens and its stability over time. We estimated the coverage of the SNL using microbiology tests reimbursement data between 2007 and 2012, for 12 pathogens or groups of pathogens.
Diagnosis tests by pathogen in the Belgian National Institute for Health and Disability Insurance nomenclature (INAMI-RIZIV)
Search for anti-Borrelia IgG antibodies
Search for anti-Borrelia IgM antibodiesa
Search for anti-Borrelia IgG antibodies, in CSFa
Search for anti-Borrelia IgM antibodies, in CSFa
Search for anti-Borrelia IgG antibodies, immunoblot confirmation testa
Search for anti-Borrelia IgM antibodies, immunoblot confirmation testa
Search for anti-Borrelia IgG antibodies in CSF, immunoblot confirmation testa
Search anti-Borrelia IgM antibodies in CSF, immunoblot confirmation testa
Chlamydia molecular amplification
Search for cryptosporidium in stool, after enrichment
Gonorrhoeae molecular amplification
Search for anti-Hepatitis A IgM antibodies, non-isotopic method
Search for anti-Hepatitis A IgG antibodies
Search for anti-Mycoplasma pneumoniae IgM antibodies
Search for anti-Mycoplasma pneumoniae IgG antibodies
Rotavirus diagnosis, for children younger than 2 years old
Syphilis (treponema pallidum)
Serodiagnosis of treponema
Reimbursement data are used to assess aspects of the sensitivity and representativeness of the SNL by looking at different measures of test coverage. In the absence of information on the true number of (diagnosed) cases in the Belgian population, we used reimbursement data as a way of proxying case coverage . Test coverage is the ratio between the number of tests reimbursed by laboratories of the surveillance system and the total number of tests reimbursed. A high test coverage value indicates that most of the tests performed are captured by the surveillance system, and therefore, that cases are more likely to be reported to the surveillance system. Conversely, a low test coverage value indicates that most of the tests are performed by laboratories that do not take part to the surveillance system and therefore, that cases are less likely to be reported. We examine the test coverage of the SNL by pathogen at national level, its variations by region and province and the stability of the coverage over time for the period 2007–2012. These levels of analysis allow to inform on representativeness by time and place and to indicate whether pathogens are equally monitored. The extent and stability of test coverage values by place and time also informs on sensitivity, i.e. the ability to monitor change over time at different geographical levels. It is important to note that this study is not based on a sample but includes the total number of tests reimbursed in Belgium for the selected pathogens. Therefore, we report descriptive statistics and do not attempt to make statistical inference. Coverage values are graphically represented to allow to visually identify variations over pathogen, place or time.
Number of tests reimbursed by pathogen
Data also reveal pathogen variations in the number of tests. Fewer tests are reimbursed for Rotavirus, Chlamydia and Cryptosporidium (less than 100,000 annually) as compared to M. pneumoniae, Hepatitis A, Gonorrheae and enteric infections (more than 300,000 annually).
National test coverage
Regional test coverage
Provincial test coverage
We found that test coverage of the SNL was stable over time and close to, or greater than, 50 % for the 12 pathogens or groups of pathogens studied. These results hold for the three regions of Belgium. For the first time, this study also allowed to investigate provincial variations in test coverage. We showed that test coverage was constant over time in most provinces, despites a greater variability in the values.
Results indicate that the SNL is sensitive and representative for the surveillance of the selected infectious diseases. Effectiveness of vaccination strategies, seasonal variations, outbreaks and all sorts of variations in the occurrence of the pathogens are very likely to be captured by the surveillance system if they occur at a national or regional level. The SNL is nevertheless less likely to capture local changes over time in some provinces such as Namur, Walloon Brabant, Liege and Limburg.
Variability over time observed in some provinces could be accounted by changes in the participation status of laboratories and/or non-concordance between available accreditation numbers of laboratories and those used in the INAMI-RIZIV reimbursement database. The latter is likely to explain the sharp decrease in Namur in 2010 for example.
Regional figures are comparable to those described by the French Epibac Network in 2010, although the French network had a higher global coverage, close to 75 % . Comparison with a former study  seems to indicate that the coverage slightly decreased for a few pathogens (e.g. for Borrelia, Cryptosporidium and Hepatitis A) since the period 1999–2002. Differences in the results might be explained by factors such as changes in the reimbursement nomenclature (e.g. Borrelia), changes in the testing behaviours and changes in the types of laboratories participating to the surveillance system (although the proportion of laboratories participating has increased, some laboratories did quit the network since 2002 and fusions have been observed). In addition, unlike Vandenberghe , we defined laboratories not sending data as inactive, even though they officially belonged to the SNL. Our coverage values are therefore expected to be slightly lower and closer to reality. Unfortunately, data do not allow to indicate whether the slight decrease is an artefact or is real and might therefore have an impact on the trends of notification rates monitored by the WIV-ISP.
This study has several limitations. First, it relies on the quality and exhaustively of the reimbursement data, without being able to evaluate it. It is indeed likely that some tests performed by laboratories to identified pathogens are not included in the INAMI-RIZIV database. Second, our approach identified laboratories using certification number for reimbursement. As laboratory identification codes evolve over time and between sources, some tests could not be identified in the list of laboratories codes provided to the INAMI-RIZIV to obtain reimbursement data. In addition, multiple laboratories associated or belonging to the same group may share a same laboratory code (e.g. after a fusion) while not all laboratories from the group participate to the surveillance network. In this study we assumed that if at least one laboratory participated to the SNL, all laboratories from the same group and with the same identification code were also participating. This assumption does not always hold which means that coverage values could be slightly biased in that respect. Third, data were only available for a short period of time and do not allow to assess how changes in test coverage of the SNL since the 1980s might influence the meaning and interpretation of the reported number of cases over time. Data were furthermore only available for 12 pathogens which are fairly well distributed over place and widely tested. It is unknown whether results hold for pathogens which are less often tested (such as Hantavirus) or which are more clustered or subject to outbreaks (such as Hepatitis A, Legionella, Neisseria meningitis, or Listeria). Testing behaviours are also likely to differ by pathogen. Some tests, such as Syphilis, are systematically prescribed for some subgroups (e.g. pregnant women). These will tend to be well reported to the SNL as they are mainly performed by laboratory hospitals, which are overrepresented in the SNL. In certain circumstances, a high test coverage, might therefore not translate into representativeness or sensitivity. Alternative surveillance systems are therefore available in order to adequately monitor infectious diseases which are not evenly distributed in the general population, such as sexually transmitted diseases.
In other instances, individuals might perform multiple tests to diagnose the same disease. This might lead to an overestimation of the representativeness of the SNL if these repeated tests are not randomly distributed between SNL and other laboratories. If, for example, an individual is diagnosed at multiple university hospitals, which are known to be more likely to participate to the SNL, the measure of coverage might slightly be overestimated. The SNL would however be able to identify duplicates amongst the reported cases.
This study proved to be useful in showing potential limitations of the SNL to monitor infectious diseases with uneven geographical distributions. Results indicate that better geographical representativeness and sensitivity could be obtained by recruiting new participating laboratories in the four provinces where lower test coverage was detected: Namur, Walloon Brabant, Liege and Limburg.
Another lesson learned from this study is that one should be cautious when interpreting data from the SNL. Analysing provincial notification rates or changes over time might for example not be relevant for some pathogens. Automatic representation of provincial data should therefore be avoided. Rigorous analysis should preferably rely on multiple sources of data, as recently illustrated by Sabbe et al.  who analysed the impact of vaccination on rotavirus activity, using a diversity of data sources such as SNL data, test reimbursement data and hospitalisation data.
In this study, we evaluated restricted aspects of the surveillance system and indicated that the SNL, by design, has sufficient precision and reflects the Belgian situation without important systematic bias. Many other characteristics of the SNL have an influence on the capacity of the surveillance system to effectively monitor infectious diseases. A wider evaluation of the surveillance system is therefore strongly recommended in order to have a better understanding of the actual quality of the data reported by the SNL . For example, there might be systematic bias arising from the fact that some participating laboratories do not report all diagnosed cases, or only report cases for certain pathogens. Given the evolving nature of the SNL, the impact of changes in the network characteristics (type of data transfer, fusion of laboratories, evolution of pathogen detection techniques, cases definition) should be systematically documented and continuously assessed. Obtaining reimbursement data by laboratory, and not by province, is necessary for that purpose. National Reference Centre and Obligatory Notification data should also be used to evaluate the surveillance system.
Through this study, we illustrated the importance of evaluating (some) attributes of surveillance systems and showed that available reimbursement data could serve as a tool for improving the representativeness of a surveillance system in the absence of other data sources. Reimbursement data have been underused so far. In the future, one can think of using reimbursement data together with the number of reported cases in order to infer the incidence of a disease. Such a procedure would assume similar test positivity for participating and non-participating laboratories, an assumption which deserves to be further investigated. Having regular reimbursement data by laboratory would therefore both allow to monitor the quality of surveillance and open up new opportunities.
This test coverage study suggests that the SNL is capable of describing the epidemiological situation and monitoring changes in the 12 pathogens or groups of pathogens studied both at national and regional levels. It should however be reinforced in four out of the 11 provinces to allow to be used as a sensitive alert system at provincial level.
INAMI-RIZIV, Belgian National Institute for Health and Disability Insurance; SNL, sentinel network of laboratory; WIV-ISP, Belgian Scientific Institute of Public Health.
The INAMI-RIZIV kindly provided test reimbursement data for the selected pathogens.
The funding agencies are the Flemish Community and French Speaking Community in Belgium. The Scientific Institute of Public Health is in its work independent of the funding agencies in the design, data collection, analysis, interpretation and writing of the manuscript.
Availability of data and materials
Reimbursement data used in this manuscript are the property of the INAMI-RIZIV. Any request to obtain access to the data should therefore directly be addressed to the INAMI-RIZIV.
SQ and NB designed the overall study. NB conducted the analyses and drafted the initial manuscript. SQ, GM and YD commented on and developed subsequent drafts with NB. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Consent for publication
Ethics approval and consent to participate
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Walckiers D, Stroobant A, Yourassowsky E, Lion J, Cornelis R. A sentinel network of microbiological laboratories as a tool for surveillance of infectious diseases in Belgium. Epidemiol Infect. 1991;106(2):297–303.View ArticlePubMedPubMed CentralGoogle Scholar
- Bleyenheuft C, Lernout T, Berger N, Rebolledo J, Leroy M, Robert A, Quoilin S. Epidemiological situation of Lyme borreliosis in Belgium, 2003 to 2012. Arch Public Health. 2015;73(1):33.View ArticlePubMedPubMed CentralGoogle Scholar
- Sabbe M, Berger N, Blommaert A, Ogunjimi B, Grammens T, Callens M, Van Herck K, Beutels P, Van Damme P, Bilcke J: Sustained low rotavirus activity and hospitalisation rates in the post-vaccination era in Belgium. Eurosurveillance (in press).Google Scholar
- Romaguera RA, German RR, Klaucke DN, Teutsch SM, Churchill RE. Evaluating public health surveillance. In: Teutsch SM, Churchill RE, editors. Principles and practice of public health surveillance. 2nd ed. Oxford: Oxford University Press; 2000. p. 176–93.Google Scholar
- European Centre for Disease Prevention and Control. Data quality monitoring and surveillance system evaluation – A handbook of methods and applications. Stockholm: ECDC; 2014.Google Scholar
- German RR, Lee LM, Horan JM, Milstein R, Pertowski C, Waller M. Updated guidelines for evaluating public health surveillance systems. MMWR Recomm Rep. 2001;50:1–35.PubMedGoogle Scholar
- Vandenberghe H. Schatting van de landelijke representativiteit van het netwerk van peillaboratoria. In: Van Oyen H, editor. Diagnostic et surveillance des maladies infectieuses. Brussels: WIV-ISP; 2004. p. 19–27.Google Scholar
- INAMI. Nomensoft: Nomenclature des prestations de santé. 2014. https://www.riziv.fgov.be/webprd/appl/pnomen/search.aspx?lg=F Accessed 22 Feb 2016.
- Institut de veille sanitaire. Le réseau Epibac. 2009. http://www.invs.sante.fr/publications/epibac/epibac.html#mat Accessed 22 Feb 2016.