The use of capture-recapture methods to provide better estimates of the burden of norovirus outbreaks from seafood in England, 2004–2011

Abstract

Norovirus (NoV) is the greatest cause of infectious intestinal disease in the UK. The burden associated with foodborne outbreaks is underestimated in part because data are dispersed across different organisations. Each looks at outbreaks through a different lens. To estimate the burden of NoV from seafood including shellfish we used a capture-recapture technique using datasets from three different organisations currently involved in collecting information on outbreaks. The number of outbreaks of NoV related to seafood including shellfish in England was estimated for the period of 2004–2011. The combined estimates were more than three times as high (N = 360 using Chao's sample coverage approach) as the individual count from organisation three (N = 115), which captured more outbreaks than the other two organisations. The estimates were calculated for both independence and dependence between the datasets. There was evidence of under-reporting of NoV outbreaks and inconsistency of reporting between organisations, which means that, currently, more than one data source needs to be used to estimate as accurately as possible the total number of NoV outbreaks and associated cases. Furthermore, either the integration of reporting mechanisms or simplifying the process of reporting outbreaks to organisations is essential for understanding and, hence, controlling disease burden.

Using Capture-Recapture Methods to Estimate Influenza Hospitalization Incidence Rates

“…By comparison, in a C-R study of norovirus cases, the combined databases yielded incidence at a rate 2.5 times the level of the rate of the highest individual database. 9 Moreover, in the current study, the overlap between databases was large, resulting in high disease coverage estimates.…”

“…C-R has been adapted from ecological studies and used in a wide array of health-related studies including Alzheimer's disease, heart attack, HIV infection, gun injury, pediatric disease surveillance, gastric cancer, and norovirus infections. [4][5][6][7][8][9] Direct C-R uses two databases and is believed to result in better estimates than indirect C-R, which uses three or more databases. In this study, estimates were nearly identical between observed and estimated incidence.…”

“…This study used capture-recapture methods to calculate influenza incidence among hospitalized patients and then used two methods for population burden: CDC’s HAIVEN burden estimate methods and a combination of HAIVEN and C-R estimates. C-R has been adapted from ecological studies and used in a wide array of health-related studies including Alzheimer’s disease, heart attack, HIV infection, gun injury, pediatric disease surveillance, gastric cancer, and norovirus infections [6-11]. The Petersen C-R method using two sources of data in this study is a special case of the generalized C-R method for estimating burden using multiple data sources.…”

“…In this study, estimates were nearly identical between observed and estimated incidence. By comparison, in a C-R study of norovirus cases, the combined databases yielded incidence at a rate 2.5 times the level of the rate of the highest individual database [11]. The assumptions of the Petersen estimator allow for any amount of overlap in the databases.…”

Using capture‐recapture methods to estimate influenza hospitalization incidence rates

“…This study used C‐R methods to calculate influenza incidence among hospitalized patients and then used two methods for population burden: CDC's HAIVEN burden estimate methods and a combination of HAIVEN and C‐R estimates. C‐R has been adapted from ecological studies and used in a wide array of health‐related studies including Alzheimer's disease, heart attack, HIV infection, gun injury, pediatric disease surveillance, gastric cancer, and norovirus infections 7–12 . The Petersen C‐R method using two sources of data in this study is a special case of the generalized C‐R method for estimating burden using multiple data sources.…”