We conducted a population-based cohort study in the three Danish counties of North Jutland, Aarhus and Funen from January 1999 through December 2008 (average population 1.6 million, i.e. approximately one third of the Danish population). The demographic composition was representative for all of Denmark.
The National Health Service provides tax-supported health care free of charge to all Danish residents, provided by general practitioners and public hospitals. People with Danish citizenship or permanent residency in Denmark are uniquely identifiable by a 10-digit civil registration number, which is recorded at all contacts with the health service allowing accurate linkage between health registries.
Local microbiological databases
In each county a single diagnostic laboratory processed all bacteriological samples submitted by general practitioners and hospitals. Information on blood isolates of Salmonella serovars were retrieved from the electronic laboratory information systems in each laboratory.
The National Registry of Enteric Pathogens (REP)
Diagnostic laboratories mandatorily report Salmonella infections to the national reference center at Statens Serum Institut (SSI) and this information is recorded in the REP. If a Salmonella serovar is isolated more than once from the same person within six months, only the first positive sample is registered in the REP. Sample type (feces, blood or another usually sterile site) is not available for all isolates. For the years 2004-2008 the database holds information on international travel provided by the local laboratory on a voluntarily basis. In 2008 full travel information in REP was obtained by telephone interviews . The travel information collected by telephone interviews and notified voluntarily has been found comparable with regard to the distribution of travel destinations. We identified patients with Salmonella infection who were residents in any of the three counties in the REP.
For patients hospitalized with non-typhoidal Salmonella bacteremia, we extracted the following information: travel history, comorbidity, any extraintestinal site of infection, dates of admission and discharge, length of stay and time of death. Comorbidity was classified according to the Charlson comorbidity index and three levels of comorbidity were defined: low (0), corresponding to patients with no recorded underlying diseases; medium (1-2) and high (≥3) .
Number of citizens in the three counties according to age groups were available from Statistics Denmark and used as denominators to calculate incidence rates .
Data on international travel
Statistics Denmark obtains information on international travel from a random sample of approximately 500 Danish citizens contacted by telephone each month. Information in the database includes age groups, travel destinations, month of travel, type of travel (business or pleasure), and length of travel. We restricted our analyses to pleasure travellers as business travellers may have a different travel form and risk profile. To account for a large number of short trips to neighboring countries, we included only travellers staying abroad for at least three nights. We used the information in the database to estimate travel patterns in the study population. Estimated numbers by age group and country of travel were used as denominators to calculate the risk of bacteremia per 100,000 travellers.
Blood culture systems and nominal volumes of blood per culture differed between regions (North Jutland County: BacT/Alert (bioMérieux, Marcy I'Etoile, France) with a blood volume of 3 × 10 mL; Aarhus County: BacT/Alert (bioMérieux, Marcy I'Etoile, France); 2 × 20 mL; Funen County: ESP (Trek Diagnostic Systems, Cleveland, Ohio, USA) until 2001, and BACTEC (Becton Dickinson, Franklin Lakes, NJ, USA) thereafter; 2 × 20 mL).
In North Jutland County fecal cultures were performed by the regional diagnostic laboratory throughout the study period . In the counties of Aarhus and Funen the regional laboratories carried out the cultures from 2004 and 2006, respectively. Prior to that, fecal cultures were performed by SSI, Copenhagen.
All Salmonella isolates obtained locally were referred to the national reference laboratory at SSI for serotyping according to the Kauffman-White scheme .
We defined travel-related Salmonella bacteremia (henceforth TRB) on the basis of information in the hospital chart regarding foreign travel prior to admission and lack of any indication of domestic exposure. The remaining bacteremias were defined as domestically acquired (henceforth DAB). If a patient had visited more than one destination, we defined the main travel destination as the country where the patient had had an episode of diarrhea. If this information was missing, the last visited country was defined as the travel destination.
A patient with Salmonella gastroenteritis was defined as a patient recorded with non-typhoidal Salmonella in the REP and no blood culture isolate recorded in the local microbiological database.
Clinical data retrieved from patient charts were tabulated in a spreadsheet. Relative prevalence proportions (RPP) with corresponding 95% confidence intervals were calculated to estimate differences in age, sex and comorbidity among patients with TRB compared to DAB. Risk estimates per 100,000 travellers were calculated according to travel destination and age groups (children 0-14, young adults 15-24, adults 25-44, middle-aged 45-64, and elderly ≥65 years).
In the subset of patients diagnosed in 2004-2008, we estimated the risk of TRB compared to travel-related gastroenteritis by calculation of odds ratios (ORs) with 95% confidence intervals (CI). Risk associated with travel region was estimated by logistic regression and age group (<65, ≥65 years) and gender were included as potential confounders.
STATA version 9.2 (College Station, Texas) was used for the statistical analyses.
The study was approved by The Danish National Board of Health and the Danish Data Protection Agency (Record no. 2008-54-0474).