In this nationwide study, we evaluated trends of SAB incidence in the population of Denmark in an 8-year period, corresponding to >40 million patient-years. We report a 48% increase in SAB incidence during 2008–2015. Although the short-term death rate remained unchanged throughout the period, population-based death rates increased more than 50% because of the increase in SAB incidence.

The reported increase in SAB incidence was, in particular, a result of the major and consistent increase among persons >80 years of age: the SAB rate among the oldest old (>80 years) rose with an alarming 90% during 2008–2015, corresponding to an estimated annual increase of 8%–13% in incidence rate. In comparison, a 36% increase in the incidence rate was found for persons <80 years of age.

The differential increase in the SAB rate among age groups stresses the importance of comprehensive data that enable age-stratified analyses when investigating temporal trends in infectious diseases. Significant changes in death and incidence rates within subgroups of a population may otherwise not be apparent.

The increasing incidence rate observed in this study is contrary to most other recent studies. Stable or decreasing rates of SAB were reported in several large observational studies (2,4–6,14,15). However, increasing rates of both community- and healthcare-acquired SAB were reported in a recent study in Finland (16). A higher occurrence of SAB is somewhat to be expected in an increasingly elderly population. However, our data showed an increase in SAB occurrence that exceeded contemporary changes in the population demographic profile.

Multiple factors may have influenced the changes in SAB incidence. Concurrent conditions form a strong risk factor associated with development of SAB; the increasing burden of concurrent conditions associated with aging could explain the higher SAB rate among the oldest elderly (2). The study population generally developed more concurrent conditions during the study period; thus, the number of persons at risk of acquiring SAB increased. However, because concurrent conditions increased similarly for SAB cases and population controls, the increasing incidence of SAB could not be explained solely by an increase in these conditions.

A more liberal use of invasive hospital procedures and immune modulating treatments may also contribute to a higher occurrence of SAB. Although the proportion of SAB cases with 90-day prior hospital contact remained unchanged throughout the study, we do not know whether rates of invasive procedures and immune-suppressive medication increased in the same period. Thus, rates of hospital contacts might not be an accurate measure; recent reports showed an increasing use of chemotherapy and invasive procedures over time, in particular among the elderly (17,18).

The average life expectancy in Denmark increased during the study period, leading to an increasingly older and potentially more fragile population (19). There may be biologic risk factors of infectious diseases associated with aging that have not yet been identified. Physiologic age-dependent changes, such as immunological senescence, are likely attributable, in part, to the higher vulnerability to infectious diseases among elderly persons (20,21).

Higher rates of hospitalization and test activity and longer in-hospital stays may be other potential explanations for the increasing SAB rate (22). Blood culture activity increased by 31% during the study period but the increase in the SAB positivity rate was 8%, suggesting that the increase in activity alone could not explain the observed changes. Further, a systematic increase in culture activity or higher sensitivity of culture systems would not explain a differential increase in incidence among age groups. Higher test rates would likely lead to the identification of milder cases of SAB. If so, a contemporary decline in the overall 30-day CFR would be expected, given higher survival rates among the less severe cases, but this was not the case.

Annual hospital admission rates increased by 15% and the number of days hospitalized declined by 16% during the study period, suggesting that changes in hospitalizations did not explain the increased rates of SAB. The increase in rates of SAB remained after adjustment for hospitalizations and hospital days. An increase in hospital-acquired infections is an unlikely explanation for the increased SAB rates because the proportion of SAB patients with and without 90-day prior hospital contact remained unchanged throughout the study.

A recent study found that the increase in sepsis rates and decline in death rates in California were associated with up-capture of less severely ill patients after introduction of guidance on coding based on the International Classification of Diseases, Ninth Revision (23). Our study, however, relied on blood culture positive cases of SAB; therefore, changes in coding practices are an unlikely explanation of our finding. Further, death rates were unchanged over time in our study.

In spite of a recent report of a decline in death associated with infectious diseases in general, SAB survival has not improved markedly during the past decades (24). The 30-day CFR of 24% found in our study is in line with other reports (25–27). However, the population death rate rose 1.5-fold, to 7.60/100,000 person-years by 2015, because of the contemporary increase in SAB incidence. Thus, population death rates may represent a more accurate measure of the actual disease burden, as changes in incidence are reflected despite an unchanged 30-day CFR. Few studies have reported SAB death rates as population death rates for comparison and none have reported age-specific death rates (28,29). Tom et al. reported a death rate from community-acquired SAB of 3.4/100,000 person-years (28). A recent study from Norway on bloodstream infections found a death rate for SAB similar to ours, of 7/100,000 person-years (29).

A particular finding in our study was the close association between the changes in the overall death rate and the increase in the age-specific death rate for the oldest age groups. The higher incidence rates for the oldest age groups were directly reflected by higher population death rates. Death rates from SAB will likely increase even more over the next decades owing to the increasingly older population, which may also be the case for other invasive bloodstream infections. However, temporal population death rates from nonstaphylococcal bloodstream infections have not been reported; further studies are warranted to address this matter.

Contrary to the case with most bloodstream infections, female sex has been associated with an inferior outcome in SAB (25). In agreement with several previous observations, we found a significantly higher death rate among women compared with men (7,25,30,31). The mechanisms underlying the observed gender differences in SAB death rates are not fully understood; a recent study did not find any gender-specific differences in clinical management, patient characteristics, or severity of the disease between men and women (32).

This study benefits from the large number of observational years, nationwide settings, and standardized registration. Nonetheless, some limitations must be noted. First, the increasing life expectancy during the study period could have led to an increasing median age within age groups and, consequently, an underestimation of the effect of age on incidence rates, in particular for the oldest persons. Second, the SAB definition was based solely on microbiological findings and not related to criteria of clinical infection such as systemic inflammatory response syndrome and sequential organ failure assessment score. Thus, we cannot preclude that a minor proportion of the positive blood cultures may be from contamination rather than clinical infection. Still, this possibility does not explain the age-dependent increase in SAB. Third, we were not able to stratify SAB origin by community-, healthcare-, and hospital-acquired infections, because these data were not accessible. Instead, we used a validated approach to access temporal changes in healthcare/hospital- and community-acquired infections. Fourth, data on blood culture activity were not available for 2008–2009; thus, we were not able to analyze trends in culture rates for the whole study period. Finally, we did not have access to clinical data regarding the primary focus of the infection, the severity of the disease, and the applied treatment strategies, as well as the effect of these potential risk factors on SAB death.

MRSA bacteremia is infrequent in Denmark (1.3% of all SAB cases), which could theoretically limit the validity of the results of this study to settings with higher MRSA prevalence. Nevertheless, in most populations, methicillin-susceptible S. aureus (MSSA) bacteremia has remained prevalent despite the emergence of MRSA and, as such, several studies have concluded that MRSA bacteremia adds to the total burden of SAB rather than replacing MSSA bacteremia (16,33). Thus, we believe that the observed changes in MSSA bacteremia in our study may be applicable to other populations.

In conclusion, SAB incidence in Denmark increased by 48% during 2008–2015. SAB rates increased the most among the oldest age group, for whom the age-specific incidence rate nearly doubled and where increases in rates exceeded the contemporary changes in the population demographic profile. Furthermore, the short-term prognosis of SAB did not improve within the study period and, combined with the increasing incidence, population death rates rose significantly. Our results stress that infection prevention initiatives and improved care are warranted to reduce SAB incidence and improve outcomes. In addition, examinations of the burden of bloodstream infections caused by MSSA must be prioritized in future research; a specific focus should be on the frail elderly population.