Evaluation of underreporting of salmonellosis and shigellosis hospitalised cases in Greece, 2011: results of a capture-recapture study and a hospital registry review

Abstract

Background: Salmonellosis and shigellosis are mandatorily notifiable diseases in Greece. Underreporting of both diseases has been postulated but there has not been any national study to quantify it. The objective of this study was to: a) estimate underreporting of hospitalised cases at public Greek hospitals in 2011 with a capture-recapture (C-RC) study, b) evaluate the accuracy of this estimation, c) investigate the possible impact of specific factors on notification rates, and d) estimate community incidence of both diseases.

Methods: The mandatory notification system database and the database of the National Reference Laboratory for Salmonella and Shigella (NRLSS) were used in the C-RC study. The estimated total number of cases was compared with the actual number found by using the hospital records of the microbiological laboratories. Underreporting was also estimated by patients' age-group, sex, type of hospital, region and month of notification. Assessment of the community incidence was based on the extrapolation of the hospitalisation rate of the diseases in Europe.

Results: The estimated underreporting of salmonellosis and shigellosis cases through the C-RC study was 47.7% and 52.0%, respectively. The reporting rate of salmonellosis significantly varied between the thirteen regions of the country from 8.3% to 95.6% (median: 28.4%). Age and sex were not related to the probability of reporting. The notification rate did not significantly differ between urban and rural areas, however, large university hospitals had a higher underreporting rate than district hospitals (p-value < 0.001). The actual underreporting, based on the hospital records review, was close to the estimated via the C-RC study; 52.8% for salmonellosis and 58.4% for shigellosis. The predicted community incidence of salmonellosis ranged from 312 to 936 and of shigellosis from 35 to 104 cases per 100,000 population.

Conclusions: Underreporting was higher than that reported by other countries and factors associated with underreporting should be further explored. C-RC analysis seems to be a useful tool for the assessment of the underreporting of hospitalised cases. National data on underreporting and under-ascertainment rate are needed for assessing the accuracy of the estimation of the community burden of the diseases.

Figure 1

Schematic description of the distribution of the total number of cases in a two-source capture-recapture model (Venn diagram). n is the total number of cases in the population, L1 and L2 the number of cases captured by each one of the two systems, m the number of missing cases from both systems and d the duplicates.