Monitoring diseases based on register data: Methods and application in the Danish swine production

Ana Carolina Lopes Antunes

    Research output: Book/ReportPh.D. thesisResearch

    381 Downloads (Pure)


    The spread of diseases is one of the most important threats to animal production and public health. Disease spread causes considerable economic losses for the agricultural sector and constitutes trade-limiting factors, as transmission to countries free from disease should beavoided. Monitoring and surveillance systems are critical for the timely and effective control of infectious diseases. The ability of a system to detect changes in the disease burden depends on the choice of data source. Many factors can lead to inconsistent data collection among populations and it is therefore important to assess the quality of data before use in disease monitoring and surveillance. Over the past decade, several studies have focused on using statistical control methods to detect outbreaks of (re-)emerging diseases in the context of syndromic surveillance – both inhuman and veterinary medicine – in an attempt to supplement traditional sentinel surveillance. However, it may not be possible to generalize the performance of these methods to the context of other countries (where data have different characteristics), or to the context of endemic diseases. Lower incidence rates are normally expected for endemic diseases compared to highlyinfectious (re-emerging) diseases, due to control measures such as vaccination or health management programs. Furthermore, the data collected differ from those obtained from traditional surveillance (generally related to incidence monitoring), due to its focus on the endemic scenario, with less frequently sampled data. This reflects the added complexity of monitoring endemic diseases, as disease burden is affected not only by the incidence, but also by the duration and recovery rate. The aim of this thesis was to evaluate existing register data related to veterinary health, asa tool for monitoring swine diseases in Denmark. This included: i) describing and evaluating the quality of data (regarding the potential for disease monitoring and surveillance) in Danish databases related to swine health; ii) assessing the feasibility of studying changes in datarecords over time to detect changes that might indicate disease spread between swine herds;iii) evaluating the performance of different time-series methods for the monitoring and surveillance of endemic diseases, as well as assessing the impact of noise in the data on the results when using these methods. Some of the work presented was focused on endemic diseases, using Porcine Reproductive and Respiratory Syndrome (PRRS) as example. Interviews were conducted with relevant stakeholders in order to assess the data quality of seven databases: the Central Husbandry Register (CHR), the swine movement databasexii (SMD), the national Danish database of drugs for veterinary use (VetStat), laboratorydiagnostic data from the National Veterinary Institute – Technical University of Denmark (DTU-Vet lab) and the Pig Research Centre - SEGES (VSP-SEGES lab), the Specific Pathogen Free System (SPF System) and the Meat Inspection database. The guidelines from the European Centre for Disease Prevention and Control (ECDC) for monitoring data quality and surveillance systems were used. The findings showed that limitations included delayed transfer of data to databases and incomplete representation of Danish swine herds. Laboratory submission data for testing PRRS were used to study temporal changes in datarecords, due to the large amount of diagnostic data available. The laboratory data proved to be useful for monitoring temporal patterns of disease occurrence. The fact that some Danish swine herds are tested monthly allows for changes in disease prevalence and incidence to be monitored, which is an example of sentinel surveillance. However, for other herds, the frequency of testing (i.e. the representativeness of the data) depends on factors such as the herd status, farmer compliance, the value of the animal, commercial purposes and ongoing control and eradication programs. This limitation did not apply to the mortality data, which is available for all Danish swine herds on a monthly basis. However, observed changes might be due to disease occurrence, or as a result of changes in herd management or a lack of accuracy in the calculation of mortality. Several scenarios representative of changes in endemic disease sero-prevalence programs were simulated to test the performance of different monitoring methods. These included univariate process control algorithms applied directly to the simulated data, as well as using the forecast errors and trend-based methods. The performance of these methods was evaluated based on the sensitivity and time taken to detect changes, which showed that some methods were more efficient than others for specific patterns. Therefore, choosing a single temporal monitoring method is challenging, and the objectives of the monitoring program and the differing performance of the methods in detecting a specific pattern should be taken into account. Changes in the noise of the data had an impact on the univariate process controlalgorithms, while the trend-based methods provided a consistent approach to monitoring changes in disease or sero-prevalence. The findings of this thesis may serve as a basis for the improvement of monitoring swinediseases in Denmark. Although the available databases have the potential for use in disease monitoring and surveillance of swine herds in Denmark, improvements are needed for accurate and real-time implementation. Further research relating to the improvement of data quality, as well as combining different data sources for monitoring endemic diseases in Denmark is needed.
    Original languageEnglish
    Place of PublicationFrederiksberg C
    Publisher National Veterinary Institute, Technical University of Denmark
    Number of pages44
    Publication statusPublished - 2016


    Dive into the research topics of 'Monitoring diseases based on register data: Methods and application in the Danish swine production'. Together they form a unique fingerprint.

    Cite this