Stable incidence and continued improvement in short term mortality of Staphylococcus aureus bacteraemia between 1995 and 2008

The register-based matched cohort was created by merging data from four national registries based on the unique civil registration number assigned to each Danish citizen [8]. The study was approved by the Ethics Committee for the Municipalities of Copenhagen and Frederiksberg (01–369 ⁄ 93) and the Danish Data Protection Agency (2009-41-4179).

The Danish Staphylococcal Bacteraemia Study Group has carried out a continuous, nationwide registration of SAB in Denmark since 1956 [6, 9]. The present study defined a case as an individual with a first episode of SAB identified via the registry between January 1, 1995 and December 31, 2008. Discharge records for each admission with SAB were reviewed [6, 9]. We defined hospital-associated SAB (HA SAB) as an infection acquired after 48 hours of hospitalisation, an infection in a patient in ambulatory care in a specialized hospital department, or a case residing in a nursing home. Community-acquired SAB (CA SAB) was defined as an infection present at the time of hospitalization in patients not meeting the definition of HA SAB. Methicillin resistance was determined by antimicrobial susceptibility testing and confirmed by the presence of the mecA gene [10, 11].

Information on the number of persons alive in Denmark on January 1 from 1995 through 2008 was obtained from Statistics Denmark (Danmarks Statistik at http://​www.​dst.​dk). These included total number and subgroups defined by age and gender.

The Danish Civil Registration System which is updated daily, tracks changes in vital status and migration for the entire Danish population [8]. For each case, we randomly selected 10 control subjects, matched by age (year of birth) and sex. We used the risk set sampling technique (i.e. eligible control subjects had to be alive and at risk of a first hospitalization with SAB on the date the corresponding case was admitted [12]).

Data on comorbidity was collected from the National Patient Registry. This registry is updated monthly and holds information on all hospital admissions and discharge diagnoses of all patients treated in a Danish hospital since January 1, 1977 as well as all outpatient contacts since January 1, 1995. The International Classification of Diseases (ICD)-8 was used from 1977 through 1993; and ICD-10 from 1994. Comorbidity prior to the SAB admission was assessed by estimating the Charlson comorbidity index (CCI) score [13, 14]. Briefly, the CCI is a validated score system used to asses patients’ comorbid conditions in longitudinal studies [15], and takes into account both the number and severity of comorbid disease. We calculated the CCI score for cases and controls by gathering diagnoses from the National Patient Registry from up to 10 years prior to the SAB index date and excluding diagnoses from the admission with SAB. To avoid the influence of re-registration of diagnoses, each of the 17 Charlson diagnostic categories could only contribute one time to the overall index. We defined three levels of comorbidity for categorizing patients in this study: none (patients without underlying disease), intermediate (weighted index of comorbidity CCI 1–2) and high (CCI ≥ 3). Alcohol related comorbidity is not included in the CCI score and was assessed separately as previously described [16, 17]. We considered the following diagnosis codes: for ICD-10 classification: F10, G31.2, G62.1, G72.1, I42.6, K29.2, K86.0, K70, R78.0, T51, Z72.1, and for ICD-8 classification: 291 (Alcoholic psychosis), 303 (Alcoholism), 979 (Alcohol in combination with specified medicinal agents), 980 (Toxic effect of alcohol), 577.10 (Chronic pancreatitis due to alcohol), 571.09 (Cirrhosis of the liver due to alcohol), 571.10 (Steatosis of the liver due to alcohol). ICD-10 K70 (cirrhosis) is included in the CCI score but was only assessed for alcohol-related conditions.

Annual data were divided into five periods (1995–1997, 1998–2000, 2001–2003, 2004–2006 and 2007–2008) for the assessment of temporal changes, and age was divided into 6 groups (<1, 1–15, 16–35, 36–55, 56–75, >75). Using the cumulative incidence proportion method to calculate incidence rates, that assumes that all individuals of a given birth cohort remain alive for a given year, indicated that incidence rates increased over time. However, annual survival rates for individuals older than 65 improved significantly during the study, leading to an average increase in days at risk of SAB in the older age strata in more recent time periods. E.g. among population controls >75 years the 30-day- and 1-year mortality proportion decreased from 7,0% and 51,4% in 1995–1997 to 1,2% and 12,8% in 2007–2008, respectively. Therefore, we adjusted the number of days at risk in each gender- and age stratum by a correction factor derived from annual survival among the population controls. Incidence rates with 95% CIs were calculated based on these adjusted data. Categorical variables were compared by chi-square test. Ninety days survival curves were constructed by the Kaplan-Meier method, and the proportional hazards assumption was evaluated visually on survival plots. Mortality rate ratios with 95% CIs were calculated using Cox proportional hazards models. Unvariate and multivariate Cox regression was performed to determine the mortality rate ratios within 30 days of SAB. All variables in the multivariate analysis were analysed using the forced entry procedure. Analyses were performed using PASW v.18 (SPSS Inc., Chicago, IL, USA).

Fifty-seven percent of cases were HA and 31% were CA. In 13% of cases the place of acquisition could not be determined. The characteristics of HA and CA cases are shown in Table 1. Overall, patients were in their mid-sixties and the majority were male.

The proportion of cases older than 75 years increased from 23.2% in 1995–1997 to 29.8% in 2007–2008 (p = 0.0001), and the proportion of cases aged 1–55 years decreased during the same period from 37.1% to 28.1% (p = 0.0001). The proportion of cases aged <1 years and 56–75 years were stable.

The proportion of cases who had a CCI score > 0 increased during the study period from 60.8% in 1995–1997 to 69.5% in 2007–2008 (p = 0.0001), of which the proportion of individuals with a CCI score > 2 increased from 19% to 32% (p = 0.0001). The proportion of cases with a CCI score increased with age although fewer cases older than 75 years compared to the 56–75 year old stratum had a CCI score greater than 0 (72.0% vs. 74.8%, p = 0.001). Patients with HA SAB had higher CCI scores than CA SAB. Among HA SAB 70,0% had a CCI greater than 0 compared to 54,8% among CA SAB (p = 0.0001).

The proportion of cases with pre-admission alcohol-related diagnoses increased from 7% to 12% between 1995–1997 and 2007–2008 (p = 0.0001). Alcohol-related diagnoses were more frequent among CA SAB compared to HA SAB (p = 0.0001).

The distribution of CCI (p = 0.0001) and alcohol-related disorders (p = 0.0001) among the age- and sex matched controls differed from cases. For controls 23% had a CCI score of 1–2 and only 6.0% had a CCI score > 2. Over time the proportion of controls with a CCI score > 0 decreased from 31.4% in 1995–1997 to 28.0% in 2007–2008 (p = 0.0001), while the proportion of controls with pre-admission alcohol-related diagnoses increased from 1.5% to 2.1% (p = 0.0001).

The adjusted overall IR (22.7/100 000 population (95%CI: 22.4-23.1/100 000)) of first time SAB remained constant from 1995 to 2008 (Table 2). IRs increased with age and were higher among males (27.9/100.000 population (95%CI: 27.4-28.5/100 000)) than females (17.6/100.000 population (95%CI: 17.2-18.1/100.000)). The IRs decreased slightly from 1995 to 2008 among persons aged 1–75 years and were stable among cases aged <1 and >75 years (Table 2).

The overall cumulative mortality among cases within 30 days from SAB was 25.7%. The median survival of HA, CA and undetermined acquisition SAB was 556 days (501–610), 836 days (721–951) and 476 days (386–566), respectively.

There was a significant impact of origin, age, gender, CCI, alcohol-related disorders and time period on mortality (Tables 3 and 4). Due to few deaths among SAB patients <36 years the periodic 30 day mortality proportions in these age groups were not presented. The highest impact on mortality was noted with age (Figure 1). The impact diminished in multivariate analyses, but remained the most influential. Second and third highest impact was seen with alcohol-related disorders and comorbidity assessed by CCI (Figure 2). By multivariate analyses the impact of alcohol-related disorders increased after adjustment for age whereas the impact of CCI decreased. This was seen among both HA and CA SAB. Multivariate analyses showed improved survival over time particularly among cases of HA but also CA bacteraemia. Gender had the lowest impact on mortality and men had lower mortality in the adjusted analysis.