Severity of non-typhoid salmonellosis as a predictor of 0-30 and 31-365 day mortality


Kim O. Gradel, Tove Ejlertsen, Henrik C. Schønheyder, Henrik Nielsen



To evaluate whether hospitalization, ordering of blood cultures, or detection of bacteremia were predictors of mortality in patients with non-typhoid salmonellosis (NTS).


Population-based registry study comprising all patients with a first-time NTS detected in stool cultures, in North Jutland County (Denmark) from 1995 through 2003. Patients were categorized into four groups: group 1) not hospitalized; group 2) hospitalized 30 days within their NTS without blood cultures (BC) being ordered; group 3) hospitalized with negative BC only; group 4) hospitalized with bacteremia. We obtained data in the county’s hospital discharge registry on comorbidity as recorded from 1977 and hospitalization. Complete follow-up was possible through the Danish civil registration system. Age and comorbidity adjusted Cox’s regression analyses were used to compute mortality rate ratios (MRR) with 95% confidence intervals (CI) at 0-30 and 31-365 days.    


Among 1,764 NTS patients, 1,082 (61.3%) were not hospitalized, 344 (19.5%) were hospitalized without BC being ordered, 245 (13.9%) were hospitalized with negative BC, and 93 (5.3%) had bacteremia (Salmonella in 81, other pathogens in 12). Within one year, 2 (0.2%) died in group 1, 14 (4.1%) in group 2, 20 (8.2%) in group 3, and 19 (20.4%) in group 4. Using group 2 as reference, 0-30 day adjusted MRR (95% CI) could not be calculated in group 1 (no deaths), whereas they were 1.8 (0.6-5.7) and 1.3 (0.4-4.6) in groups 3 and 4, respectively. For 31-365 days, adjusted MRR (95% CI) were 0.16 (0.03-0.78) in group 1, 1.6 (0.69-3.9) in group 3, and 2.3 (1.0-5.5) in group 4.


The general practitioner’s decision to hospitalize NTS patients was a predictor of short-term and long-term mortality independent of age and comorbidity. When hospitalized, the physician’s decision to obtain blood cultures or the detection of bacteremia were independent predictors of mortality, albeit the statistical precision was low.