Advertisement

Utilization of blood cultures in Danish hospitals: a population-based descriptive analysis

Open ArchivePublished:November 22, 2014DOI:https://doi.org/10.1016/j.cmi.2014.11.018

      Abstract

      This national population-based study was conducted as part of the development of a national automated surveillance system for hospital-acquired bacteraemia and ascertains the utilization of blood cultures (BCs). A primary objective was to understand how local differences may affect interpretation of nationwide surveillance for bacteraemia. From the Danish Microbiology Database, we retrieved all BCs taken between 2010 and 2013 and linked these to admission data from the National Patient Registry. In total, 4 587 295 admissions were registered, and in 11%, at least one BC was taken. Almost 50% of BCs were taken at admission. The chance of having a BC taken declined over the next days but increased after 4 days of admission. Data linkage identified 876 290 days on which at least one BC was taken; 6.4% yielded positive results. Ten species, Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Enterococcus faecium, Enterococcus faecalis, Pseudomonas aeruginosa, Candida albicans, Enterobacter cloacae and Klebsiella oxytoca, accounted for 74.7% of agents for this purpose classified as pathogenic. An increase in BCs and positive BCs was observed over time, particularly among older patients. BCs showed a seasonal pattern overall and for S. pneumoniae particularly. A predominance of male patients was seen for bacteraemias due to S. aureus, E. faecium and K. pneumoniae. Minor differences in BCs and positive BCs between departments of clinical microbiology underpin the rationale of a future automated surveillance for bacteraemia. The study also provides important knowledge for interpretation of surveillance of invasive infections more generally.

      Keywords

      Introduction

      Bacteraemia is a severe condition associated with high mortality [
      • Søgaard M.
      • Nørgaard M.
      • Dethlefsen C.
      • Schønheyder H.C.
      Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study.
      ,
      • Engel C.
      • Brunkhorst F.M.
      • Bone H.G.
      • Brunkhorst R.
      • Gerlach H.
      • Grond S.
      • et al.
      Epidemiology of sepsis in Germany: results from a national prospective multicenter study.
      ,
      • Martin G.S.
      • Mannino D.M.
      • Eaton S.
      • Moss M.
      The epidemiology of sepsis in the United States from 1979 through 2000.
      ,
      • Goto M.
      • Al-Hasan M.N.
      Overall burden of bloodstream infection and nosocomial bloodstream infection in North America and Europe.
      ,
      • Leibovici L.
      Long-term consequences of severe infections.
      ]. Blood cultures (BCs) continue to be the only practical method to diagnose bacteraemia [
      • Kirn T.J.
      • Weinstein M.P.
      Update on blood cultures: how to obtain, process, report, and interpret.
      ]. Since 1 January 2010 the Danish Microbiology Database (MiBa) has collected microbiological test results from all departments of clinical microbiology (DCMs) in Denmark [
      • Voldstedlund M.
      • Haarh M.
      • Molbak K.
      MiBa Board of Representatives
      The Danish microbiology database (MiBa) 2010 to 2013.
      ]. This provided the unique opportunity to study BC utilization on a national level. Combining these data with administrative information from the National Patient Registry (NPR [
      • Lynge E.
      • Sandegaard J.L.
      • Rebolj M.
      The Danish national patient register.
      ]) allowed us to also study BCs in relation to hospital admissions.
      It is of fundamental interest to study the epidemiology and utilization of BCs to evaluate clinical practices and to understand trends observed in surveillance for invasive infections, including those acquired in healthcare. Differences in BC utilization, e.g. between laboratories, patient populations and changes over time, may give rise to artefacts in surveillance systems due to different levels of ascertainment. We conducted a national population-based study describing the utilization of BCs in Denmark to understand to which extent local differences may affect the interpretation of surveillance of bacteraemia. This assessment was done as part of the development of an automated surveillance system for hospital-acquired infections in Denmark; such a system will depend on a meaningful pooling of data from various DCMs.

      Methods

       Data sources

      MiBa is a real-time database that automatically receives a copy of every electronic microbiology report delivered by all Danish DCMs [
      • Voldstedlund M.
      • Haarh M.
      • Molbak K.
      MiBa Board of Representatives
      The Danish microbiology database (MiBa) 2010 to 2013.
      ]. An extract from MiBa was obtained comprising all BCs with a sampling date between 1 January 2010 and 31 December 2013. This extract included the sampling date and time (the latter if available), cultured microorganisms and the DCM that carried out the test. Each patient was identified in MiBa through the civil registration (CPR) number, a unique identifier given to each person living in Denmark encrypting date of birth and sex [
      • Pedersen C.B.
      • Gøtzsche H.
      • Møller J.O.
      • Mortensen P.B.
      The Danish civil registration system. A cohort of eight million persons.
      ].
      In January 2010 Denmark had 13 DCMs. Although remaining independent DCMs, the laboratory information systems of the DCMs in Herlev and Hvidovre merged in May 2012, and the DCM in Hillerød joined this mutual data server in May 2013. In January 2013 the DCMs in Herning and Viborg merged. For this article, the DCMs were analysed in the new composition (named by their geographic location).
      The NPR includes administrative data on somatic inpatients since 1977 [
      • Lynge E.
      • Sandegaard J.L.
      • Rebolj M.
      The Danish national patient register.
      ]. Individual patients were identified through the CPR number. We used an extract comprising patient administrative data between 1 January 2010 and 31 December 2013. Only those patients who were admitted and discharged within this period were selected; others were excluded, as these would affect analyses on BCs in relation to the number of days since admission. Data included date and time of admission and discharge and the responsible departments and hospitals. The NPR included one record for each admission to a department; each time a patient was transferred to another department, this was registered as a new record. We developed an algorithm relating these inpatient transfers to form a complete course of admission, here referred to as an admission.
      The data from MiBa and NPR were linked using the CPR number. Patients with temporary CPR numbers, such as foreign travellers, were excluded from analysis. Similarly, those CPR numbers derived from MiBa which led to an age calculation of <0 or ≥100 years were excluded, as we could not confirm whether these CPR numbers were correct.

       Definitions

      To enable automatic classification and avoid misclassification of contaminants as pathogens, we considered the following microorganisms as contaminants: Acinetobacter spp., Aerococcus spp. (except A. urinae), Bacillus spp. (except B. anthracis and B. cereus), Corynebacterium spp. (except C. diphtheriae), Lactobacillus spp., Lactococcus spp., Micrococcus spp., Moraxella spp. (except M. catarrhalis), Neisseria spp. (except N. animaloris, N. canis, N. elongate, N. gonorrhoeae, N. zoodegmatis and N. meningitidis), Propionibacterium acnes, Staphylococcus spp. (except S. aureus, S. saprophyticus, S. lugdunensis and S. schleiferi). Most DCMs identify Streptococcus spp. and nonhemolytic streptoccocci to the species level, especially if the microorganism is considered the etiological agent for bacteremia. Thus, when reported at the genus level, findings were assessed as contaminants.
      Microorganisms not listed as contaminants were considered pathogens.
      A blood culture day (BCD) was defined as a day on which a patient had at least one blood sample taken for culture. The reason for this measure was the practice in some DCMs to register each bottle of a BC set as an individual sample, while other laboratories registered a set of bottles. The time of sampling was not always available, making it impossible to distinguish between multiple bottles from one set and sets of bottles drawn at different moments in time on the same day.
      A positive BCD was defined as a BCD on which at least one culture yielded growth of at least one pathogenic microorganism.

       Data analysis

      General demographics were described for patients who had BCs taken. Age at first blood sample and sex were derived from the CPR number. BCDs and positive BCDs were observed over time and by sex and age groups (0–4, 5–24, 25–44, 45–64, 65–74, 75–84 and 85–99 years). When stratifying by age group, the number of BCDs were calculated per 100 000 population, using population data from Denmark Statistics for the first quarter of 2012 (http://www.statistikbanken.dk/statbank5a/selectvarval/saveselections.asp).
      The ten most frequently found pathogenic microorganisms were identified. These were determined by BCD, meaning that different pathogenic species from the same BCD were included but counted only once. Differences in the distribution of these ten pathogenic microorganisms among the DCMs were analysed. Furthermore, the five most frequently found pathogenic microorganisms were observed over time and stratified by age group and sex of the patients.
      BCDs and positive BCDs were studied in relation to admissions and stratified by DCM. On the national level, BCDs, positive BCDs and the percentage of positive BCDs were studied over a period of 30 days since hospital admission, with day 1 being the day of admission. In addition, BCDs and positive BCDs were observed over this 30-day period per 100 patients in hospital on each of the days since admission in order to adjust for the fact that the number of patients decreases on the days after admission.
      Data management and analysis were performed by SAS software (SAS Institute, Cary, NC, USA).

       Ethical considerations

      This study was approved by the Danish Data Protection Authority as part of the development of the Danish Hospital Acquired Infections Database (registration number 2012-41-1269).

      Results

      A total of 408 179 unique patients had at least one BC taken within the study period. Of these, 413 were excluded because their CPR numbers led to an age calculation outside the 0- to 99-year age range. Finally, there were 407 766 patients, 205 417 (50.4%) men and 202 349 (49.6%) women. The mean age was 57 years (median 64 years).

       Distribution by time and person

      The total number of BCDs was 876 290; 55 992 (6.4%) of these were positive BCDs. The number of BCDs increased slightly over time for both men and women (Fig. 1). In addition, seasonal variation occurred, with more BCDs in the winter months. A slight general increasing trend was also seen for positive BCDs, mostly for men. However, there was no seasonal variation of positive BCDs, which means that the percentage of positive BCDs was lower during winter months. Overall, the proportion of positive BCDs for men was 6.8% and for women 5.9%. Over the entire period, more positive BCDs were recorded for men.
      Figure thumbnail gr1
      FIG. 1Number of blood culture (BC) days and positive BC days stratified by sex between 2010 and 2013.
      Considering the incidence of BCDs among age groups, the increase in BCDs and positive BCDs over the 4 years was mainly seen among patients aged 65 years and over. In these age groups, seasonal variation was also observed, with a particularly large peak in the winter of 2013. The youngest age group of children (0–4 years) also showed a seasonal pattern in BCDs, with more consistent increases in the winter. Seasonal variation in positive BCDs was not distinctive in any age group.
      Trends over time varied among DCMs. The general increase in BCDs was observed for all, except for the DCM at the university hospital and national referral centre, Rigshospitalet, in Copenhagen, where a slight decrease was seen. The increases were most considerable for the DCMs in Herlev/Hvidovre/Hillerød, in Slagelse, Odense and in Aarhus. These DCMs, as well as those in Aalborg and Herning/Viborg, also showed most marked seasonal variation in BCDs. The DCMs in Herlev/Hvidovre/Hillerød and Slagelse were the largest contributors to the increase in positive BCDs.

       Microbiological findings

      The ten most frequently isolated pathogenic microorganisms are presented in Table 1. Together, the top ten accounted for 74.7% of all pathogenic microorganisms. Of the bacterial species that were considered contaminants, 77.2% were coagulase negative staphylococci (excluding S. saprophyticus, S. lugdunensis and S. schleiferi, as these were considered pathogens).
      TABLE 1Ten most common pathogenic microorganisms per blood culture day according to department of clinical microbiology between 2010 and 2013
      Departments of clinical microbiologyEscherichia coliStaphylococcus aureusKlebsiella pneumoniaeStreptococcus pneumoniaeEnterococcus faeciumEnterococcus faecalisPseudomonas aeruginosaCandida albicansEnterobacter cloacaeKlebsiella oxytocaTotal
      n%n%n%n%n%n%n%n%n%n%n
      Aalborg181441.273816.73959.03918.91984.52405.42024.61543.51393.21373.14408
      Aarhus204037.093116.95089.24768.64377.93306.02294.12173.91793.21733.15520
      Esbjerg62839.827917.71398.81227.7674.21026.5694.4855.4422.7442.81577
      Herlev/Hvidovre/Hillerød533942.1220117.311108.711459.07025.58126.44393.52982.33172.53262.612 689
      Herning/Viborg133238.063418.13459.83459.81604.62858.11213.4702.0972.81203.43509
      Odense161632.878716.04158.43246.656811.54378.92114.32334.71823.71473.04920
      Rigshospitalet56720.061021.532711.5782.752318.42458.61144.01685.91324.6782.72842
      Slagelse267642.1107716.96309.95869.22724.33635.72303.61522.41782.81983.16362
      Sønderborg57745.222417.51169.11088.5483.8776.0372.9221.7362.8322.51277
      Vejle92344.232415.51798.61879.01276.11185.6904.3391.9472.2552.62089
      Total17 51238.7780517.341649.237628.331026.930096.717423.914383.213493.013102.945 193
      The DCM at Rigshospitalet saw a lower occurrence of E. coli and S. pneumoniae than the other DCMs and a higher occurrence of E. faecium (Table 1). Otherwise, there were only small variations in the distribution between the DCMs.
      Increasing trends were seen for E. coli and S. aureus (Fig. 2). These increasing trends affected both men and women, but only in age groups of 65 years and older. S. pneumoniae showed a clear seasonal variation, with increases during winter. This trend was seen among men and women and in all age groups. S. aureus, E. faecium and K. pneumoniae were more often found among men (male:female ratio of 1.7:1.0, 1.6:1.0 and 1.6:1.0, respectively), while E. coli and S. pneumoniae were found equally among the sexes.
      Figure thumbnail gr2
      FIG. 2Frequency of five most common pathogens per blood culture day between 2010 and 2013.

       BCs in relation to admissions

      Between 2010 and 2013 there were 4 587 295 admissions; in 506 797 of these (11%) there was at least one BCD. As patients may be transferred to other hospitals, several DCMs may be involved in culturing blood samples during the course of one admission. For 498 245 admissions (98.3%), only one DCM was recorded, for 8337 (1.7%) two, for 205 (0.04%) three and for ten (<0.01%) four. If two DCMs were involved on the same day, only the one that tested the first sample was included in Table 2. Of the total of 876 290 BCDs, 827 106 (94.4%) occurred during an admission and 432 164 (49.3%) coincided with the day of admission. The 5.6% of BCDs that were not taken during an admission were taken during the 3 days before an admission or up to 30 days after an admission. The percentage of positive BCDs did not vary substantially between DCMs.
      TABLE 2Number of BCs and percentage of positive BC days in relation to admission according to department of clinical microbiology between 2010 and 2013
      Departments of clinical microbiologyBC days totalBC days during admissionBC days at day of admission
      n% positiven% positiven% positive
      Aalborg82 7626.679 1756.642 0707.6
      Aarhus125 2055.5116 9535.659 5986.7
      Esbjerg31 9915.930 4876.017 4896.6
      Herlev/Hvidovre/Hillerød216 7177.2205 9327.3113 9217.9
      Herning/Viborg77 1885.673 5575.642 6186.6
      Odense91 0506.783 5926.938 0098.0
      Rigshospitalet67 6605.562 2255.715 1528.1
      Slagelse113 7666.9109 8426.965 8687.6
      Sønderborg26 1046.024 4986.213 2487.5
      Vejle43 8475.940 8456.124 1917.2
      Total876 2906.4827 1066.5432 1647.4
      BC, blood culture.
      The number of BCDs decreased more after admission than the number of positive BCDs, resulting in a percentage of positive BCDs that showed an initial decrease but a steady increase from day 4 since admission (Fig. 3A). The number of BCDs and positive BCDs per 100 admitted patients also showed an initial decrease, followed by a steady increase from day 4 (Fig. 3B). When observing the five most common pathogens in relation to the admission, all showed a marked decrease after the day of admission. Only E. faecium increased again after day 5 and decreased after day 9.
      Figure thumbnail gr3
      FIG. 3(a) Number of blood culture (BC) days, positive BC days and percentage of positive BC days in relation to days since admission between 2010 and 2013. (b) Number of BC days and positive BC days per 100 admitted patients in relation to days since admission between 1 January 2010 and 31 December 2013.

      Discussion

      The availability of national microbiology data in combination with admission data on all Danish patients created the unique opportunity to make a population-based descriptive analysis of the utilization and general results of BCs in Denmark and to analyse trends. This study showed increases in BC activity and seasonality over the period between 1 January 2010 and 31 December 2013. It provided important insights in differences between DCMs and among the patient population in terms of age and sex. The BCDs and percentage of positive BCDs were shown to decrease in the first 4 days of admission and to increase after that.
      Our study was subject to several limitations. Because of varying registration practices, we were unable to assess the blood volume of each BC and distinguish culture sets (several bottles for a BC obtained at one time point) from individual culture bottles. The measure we used as a proxy, the BCD, may therefore underestimate blood-culturing activity. However, it does reflect the daily decision by the clinical team to test a patient for bacteraemia. The definition of positive BCDs will also give rise to an underestimation, as the positive rate is dependent on the volume of blood that was used for culture. Another reason for underestimation is that the BCDs do not include BCs where likely contaminants were repeatedly isolated and hence could be ruled to be clinically relevant. Classifying all nonspeciated streptococci as contaminants may also add to an underestimation of positive BCDs, as may the classification of species commonly considered contaminants in Denmark, which in rare cases may be etiologic agents (e.g. Acinetobacter baumanii). This was done to be certain that BCDs counted as positive reflected well-defined and true cases of bacteraemia and not just contaminated samples. Conclusions drawn from automated systems should be judged with caution, and our rationale was to prioritize specificity while being aware that absolute numbers of BCDs may be underestimated.
      Lastly, numbers of positive vs. negative BCDs late in the course of admission may be biased by clinicians’ attempts to take repeated BCs until a negative appears, e.g. in cases of endocarditis or candidemia. The magnitude and direction of this bias is difficult to assess.
      For these reasons, the measures of BCDs and positive BCDs may be difficult to compare to other studies in absolute numbers. They do, however, standardize the data and therefore allow following trends over time and studying differences by age group, sex and DCM.
      To further understand differences between DCMs, it would have been useful to assess patients’ clinical characteristics which led to the decision to take a BC. However, these data were not available from the sources used in the present study. It was also not possible to group departments to which patients were admitted by their specialty, e.g. to identify intensive care units, departments of surgery, internal medicine and paediatrics. The NPR and MiBa do provide the name of each specific department. Nonetheless, these names do not always contain information on the specialty, and the registries do not contain an unambiguous classification of the type of department. It would also have been informative to analyse the findings for each DCM in relation to the patient population it serves—for example, in terms of number of patients and length of stay. This information was, however, only available on a national level, not by DCM. As a follow-up study, it would be of interest to apply a comorbidity index based on ICD-10 codes from the NPR.
      At the beginning of 2010, the increases observed in BCDs and positive BCDs may have been due to the starting up phase of MiBa, but the BC activity showed a steady increase over all 4 years, suggesting a real increase. The increasing trend is also in line with other studies that have seen substantial increases in incidence rates of bacteraemia over longer periods of time [
      • Søgaard M.
      • Nørgaard M.
      • Dethlefsen C.
      • Schønheyder H.C.
      Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study.
      ,
      • Madsen K.M.
      • Schønheyder H.C.
      • Kristensen B.
      • Sørensen H.T.
      Secular trends in incidence and mortality of bacteraemia in a Danish county, 1981–1994.
      ,
      • Uslan D.Z.
      • Crane S.J.
      • Steckelberg J.M.
      • Cockerill 3rd, F.R.
      • St Sauver J.L.
      • Wilson W.R.
      • et al.
      Age- and sex-associated trends in bloodstream infection: a population-based study in Olmsted County, Minnesota.
      ,
      • Skogberg K.
      • Lyytikäinen O.
      • Ruutu P.
      • Ollgren J.
      • Nuorti J.P.
      Increase in bloodstream infections in Finland, 1995–2002.
      ]. This can partly be explained by an aging population. Organizational changes in acute care in Danish hospitals, which have taken place during the same period, may also have contributed to the increase. Thirdly, the introduction of sepsis packages in several Danish hospitals since April 2010, describing interventions to reduce mortality from sepsis through correct and timely diagnostics and treatment (http://www.patientsikkertsygehus.dk/pakker/alle-pakker/sepsispakken.aspx), can also be a reason for the increase in BC activity. The differences in trends over time we observed between DCMs may be due to differences in the patient population the DCMs served, but other factors, such as BC methodology, also need to be taken into account in the interpretation of time trends of bacteraemia [
      • Søgaard M.
      • Engebjerg M.C.
      • Lundbye-Christensen S.
      • Schønheyder H.C.
      Changes in blood culture methodology have an impact on time trends of bacteraemia: a 26-year regional study.
      ]. The general increase in positive BCDs was mainly due to an increase in E. coli bacteraemias and to a lesser extent to S. aureus bacteraemias. It would be interesting to further investigate if the increase in E. coli included an increase in extended-spectrum β-lactamase-producing bacteria. The observed trends over time will need to be confirmed using time-series analysis when data become available for a longer period of time.
      The seasonal trends in BC activity with increases in the winter coincided with the influenza seasons. The seasons of 2010–2011 and 2012–2013 had a higher influenza activity in Denmark than the season of 2011–2012 [
      • Krause T.
      • Knudsen L.
      • Emborg H.
      • Gubbels S.
      • Mazick A.
      • Nielsen J.
      • et al.
      Week 23, 2013. Epi-News.
      ]. Our data on the BC activity show a similar trend with more marked increases in the seasons of 2010–2011 and 2012–2013. An explanation could be that patients admitted with fever and respiratory symptoms are commonly tested for bacteraemia. The finding that the seasonal variation was more marked in some DCMs may suggest that some hospitals received more patients with fever and respiratory symptoms than others and/or that some hospitals have different practices for testing these patients. The increase in BC testing did not lead to an increase in positive BCs during winter. Therefore, the percentage of positive BCs was lower, suggesting that the decision to take a BC may be influenced by expectations of the clinical team or by the presence of fever rather than signs and symptoms of bacterial vs. viral infections. The occurrence of S. pneumoniae did show a seasonal trend with peaks in the winter. This was not notable in the overall picture of positive BCs because S. pneumoniae bacteraemias accounted for a small percentage of the total number of positive BCs.
      The number of positive BCDs was higher among men, which is in line with other studies showing higher incidences of bacteraemia among men [
      • Uslan D.Z.
      • Crane S.J.
      • Steckelberg J.M.
      • Cockerill 3rd, F.R.
      • St Sauver J.L.
      • Wilson W.R.
      • et al.
      Age- and sex-associated trends in bloodstream infection: a population-based study in Olmsted County, Minnesota.
      ,
      • Laupland K.B.
      • Kibsey P.C.
      • Gregson D.B.
      • Galbraith J.C.
      Population-based laboratory assessment of the burden of community-onset bloodstream infection in Victoria, Canada.
      ]. The higher number of positive BCDs among men was mainly due to S. aureus, E. faecium and K. pneumoniae. As mentioned before, once data are available for a longer period of time, time-series analysis will allow confirmation of these trends.
      The finding of E. coli, S. aureus and K. pneumoniae as the most common microorganisms was to be expected. Although an upsurge of E. faecium is well described in Denmark, it was a surprise that E. faecium was found to be more frequent than E. faecalis [
      • Lester C.H.
      • Sandvang D.
      • Olsen S.S.
      • Schønheyder H.C.
      • Jarløv J.O.
      • Bangsborg J.
      • et al.
      Emergence of ampicillin-resistant Enterococcus faecium in Danish hospitals.
      ,
      • Pinholt M.
      • Ostergaard C.
      • Arpi M.
      • Bruun N.E.
      • Schønheyder H.C.
      • Gradel K.O.
      • et al.
      Incidence, clinical characteristics and 30-day mortality of enterococcal bacteraemia in Denmark, 2006–2009: a population-based cohort study.
      ]. This may have to do with the way we analysed the data, i.e. including a microorganism only once per BCD.
      Differences in the percentage of positive BCDs between the DCMs were small. This finding is of importance for the construction of a system for automated surveillance of nosocomial bacteraemia. The upcoming automated system is based on the rationale that data from different DCMs and patient populations can be pooled meaningfully. The main outlier was the DCM at Rigshospitalet, which is a tertiary hospital with highly specialized national functions. This DCM had a different ranking of pathogens, which we cautiously ascribe to a different patient population with more comorbidities, complications and susceptibility to opportunistic infections. Other variations between the DCMs may be due to differences in sample size, timing of antimicrobial treatment, antibiotic treatment policies and blood volume taken.
      The number of BCDs during admission showed an expected pattern, in which close to 50% of BCDs occurred on the day of admission, followed by a decrease over the following days. Moreover, the results of BCDs per 100 patients showed that the risk of having a BC taken increases with longer admissions. As the positive BCDs per 100 patients and the percentage positive of BCDs also continued to increase, it can be concluded from our data that the risk of getting bacteraemia also increases with the length of stay. Many patients with suspected invasive illness at admission will receive antimicrobial treatment cover after 48 hours’ admission, and this may in part explain the turning point we observed. Another explanation for the increase from day 4 may be the occurrence of hospital-acquired bacteraemia. The finding that E. faecium, which is often found in hospital-acquired bacteraemias, increased at day 5 of admission may further support this. Computer algorithms to determine hospital-acquired bacteraemia showed 48 hours to be a useful threshold [
      • Trick W.E.
      • Zagorski B.M.
      • Tokars J.I.
      • Vernon M.O.
      • Welbel S.F.
      • Wisniewski M.F.
      • et al.
      Computer algorithms to detect bloodstream infections.
      ,
      • Gradel K.O.
      • Knudsen J.D.
      • Arpi M.
      • Ostergaard C.
      • Schønheyder H.C.
      • Søgaard M.
      Classification of positive blood cultures: computer algorithms versus physicians’ assessment—development of tools for surveillance of bloodstream infection prognosis using population-based laboratory databases.
      ,
      • Pokorny L.
      • Rovira A.
      • Martín-Baranera M.
      • Gimeno C.
      • Alonso-Tarrés C.
      • Vilarasau J.
      Automatic detection of patients with nosocomial infection by a computer-based surveillance system: a validation study in a general hospital.
      ]. Other studies have suggested 72 hours [
      • Bellini C.
      • Petignat C.
      • Francioli P.
      • Wenger A.
      • Bille J.
      • Klopotov A.
      • et al.
      Comparison of automated strategies for surveillance of nosocomial bacteremia.
      ]. Our approach of showing positive BCDs per 100 patients and percentage of positive BCDs cannot give a clear suggestion for the number of hours to be used as a threshold, but the increases we observed after day 4 may strengthen the previous assumptions that a threshold in a computer algorithm can be used.
      In conclusion, we observed an overall increase in BC utilization and positive BCDs, most prominently among the older age groups, and mostly caused by E. coli and to a lesser extent S. aureus. The activity in terms of BCDs showed a seasonal pattern, driven by negatives rather than positives, possibly related to the influenza seasons and the seasonality of S. pneumoniae. The distribution of pathogens differed among DCMs, possibly due to differences in patient populations. A predominance of men was seen for bacteraemias due to S. aureus, E. faecium and K. pneumoniae. The proportion of positive BCDs was similar between DCMs. It decreased in the first 4 days of admission and increased after that.
      These trends and differences provide important insights, which will soon be used to create a nationwide electronic surveillance system for hospital-acquired bacteraemia, which will show data for the whole country, as well as by region, hospital and each individual clinical department. The minor differences in the BCDs and positive BCDs among DCMs underpin the rationale and meaningfulness of such a surveillance system. Nonetheless, we found differences, particularly in the findings of specific pathogens, which suggest that factors such as clinical characteristics of the patient population are influencing the results. The different patterns between age groups, sex and DCMs that serve tertiary hospitals further illustrated this. This is particularly important to keep in mind when attempting to compare results from departments or hospitals with each other or to regional and national results. It shows the need for age, sex and comorbidity adjustment when standardizing national surveillance statistics.
      This type of analysis, in which routine databases are linked, may also create opportunities for public health surveillance in other countries, especially when they face restrictions on the use of clinical databases for public health purposes. The study provides important baseline data for the interpretation of surveillance data for invasive infections more generally, in particular when the aim is to compare surveillance figures from various populations and healthcare systems, and where case ascertainment is highly dependent on diagnostic practices.

      Transparency declaration

      All authors report no conflicts of interest relevant to this article.

      Acknowledgements

      We thank K. S. Nielsen and S. Jakobsen, Department IT Development, and projects at Statens Serum Institut, the MiBa Board of Representatives and all departments of clinical microbiology in Denmark.

      References

        • Søgaard M.
        • Nørgaard M.
        • Dethlefsen C.
        • Schønheyder H.C.
        Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study.
        Clin Infect Dis Off Publ Infect Dis Soc Am. 2011; 52: 61-69
        • Engel C.
        • Brunkhorst F.M.
        • Bone H.G.
        • Brunkhorst R.
        • Gerlach H.
        • Grond S.
        • et al.
        Epidemiology of sepsis in Germany: results from a national prospective multicenter study.
        Intensive Care Med. 2007; 33: 606-618
        • Martin G.S.
        • Mannino D.M.
        • Eaton S.
        • Moss M.
        The epidemiology of sepsis in the United States from 1979 through 2000.
        N Engl J Med. 2003; 348: 1546-1554
        • Goto M.
        • Al-Hasan M.N.
        Overall burden of bloodstream infection and nosocomial bloodstream infection in North America and Europe.
        Clin Microbiol Infect Off Publ Eur Soc Clin Microbiol Infect Dis. 2013; 19: 501-509
        • Leibovici L.
        Long-term consequences of severe infections.
        Clin Microbiol Infect Off Publ Eur Soc Clin Microbiol Infect Dis. 2013; 19: 510-512
        • Kirn T.J.
        • Weinstein M.P.
        Update on blood cultures: how to obtain, process, report, and interpret.
        Clin Microbiol Infect Off Publ Eur Soc Clin Microbiol Infect Dis. 2013; 19: 513-520
        • Voldstedlund M.
        • Haarh M.
        • Molbak K.
        • MiBa Board of Representatives
        The Danish microbiology database (MiBa) 2010 to 2013.
        Euro Surveill Bull Eur Sur Mal Transm Eur Commun Dis Bull. 2014; 19
        • Lynge E.
        • Sandegaard J.L.
        • Rebolj M.
        The Danish national patient register.
        Scand J Public Health. 2011; 39: 30-33
        • Pedersen C.B.
        • Gøtzsche H.
        • Møller J.O.
        • Mortensen P.B.
        The Danish civil registration system. A cohort of eight million persons.
        Dan Med Bull. 2006; 53: 441-449
        • Madsen K.M.
        • Schønheyder H.C.
        • Kristensen B.
        • Sørensen H.T.
        Secular trends in incidence and mortality of bacteraemia in a Danish county, 1981–1994.
        APMIS. 1999; 107: 346-352
        • Uslan D.Z.
        • Crane S.J.
        • Steckelberg J.M.
        • Cockerill 3rd, F.R.
        • St Sauver J.L.
        • Wilson W.R.
        • et al.
        Age- and sex-associated trends in bloodstream infection: a population-based study in Olmsted County, Minnesota.
        Arch Intern Med. 2007; 167: 834-839
        • Skogberg K.
        • Lyytikäinen O.
        • Ruutu P.
        • Ollgren J.
        • Nuorti J.P.
        Increase in bloodstream infections in Finland, 1995–2002.
        Epidemiol Infect. 2008; 136: 108-114
        • Søgaard M.
        • Engebjerg M.C.
        • Lundbye-Christensen S.
        • Schønheyder H.C.
        Changes in blood culture methodology have an impact on time trends of bacteraemia: a 26-year regional study.
        Epidemiol Infect. 2011; 139: 772-776
        • Krause T.
        • Knudsen L.
        • Emborg H.
        • Gubbels S.
        • Mazick A.
        • Nielsen J.
        • et al.
        Week 23, 2013. Epi-News.
        Statens Serum Institut, Copenhagen2013
        • Laupland K.B.
        • Kibsey P.C.
        • Gregson D.B.
        • Galbraith J.C.
        Population-based laboratory assessment of the burden of community-onset bloodstream infection in Victoria, Canada.
        Epidemiol Infect. 2013; 141: 174-180
        • Lester C.H.
        • Sandvang D.
        • Olsen S.S.
        • Schønheyder H.C.
        • Jarløv J.O.
        • Bangsborg J.
        • et al.
        Emergence of ampicillin-resistant Enterococcus faecium in Danish hospitals.
        J Antimicrob Chemother. 2008; 62: 1203-1206
        • Pinholt M.
        • Ostergaard C.
        • Arpi M.
        • Bruun N.E.
        • Schønheyder H.C.
        • Gradel K.O.
        • et al.
        Incidence, clinical characteristics and 30-day mortality of enterococcal bacteraemia in Denmark, 2006–2009: a population-based cohort study.
        Clin Microbiol Infect Off Publ Eur Soc Clin Microbiol Infect Dis. 2014; 20: 145-151
        • Trick W.E.
        • Zagorski B.M.
        • Tokars J.I.
        • Vernon M.O.
        • Welbel S.F.
        • Wisniewski M.F.
        • et al.
        Computer algorithms to detect bloodstream infections.
        Emerg Infect Dis. 2004; 10: 1612-1620
        • Gradel K.O.
        • Knudsen J.D.
        • Arpi M.
        • Ostergaard C.
        • Schønheyder H.C.
        • Søgaard M.
        Classification of positive blood cultures: computer algorithms versus physicians’ assessment—development of tools for surveillance of bloodstream infection prognosis using population-based laboratory databases.
        BMC Med Res Methodol. 2012; 12: 139
        • Pokorny L.
        • Rovira A.
        • Martín-Baranera M.
        • Gimeno C.
        • Alonso-Tarrés C.
        • Vilarasau J.
        Automatic detection of patients with nosocomial infection by a computer-based surveillance system: a validation study in a general hospital.
        Infect Control Hosp Epidemiol Off J Soc Hosp Epidemiol Am. 2006; 27: 500-503
        • Bellini C.
        • Petignat C.
        • Francioli P.
        • Wenger A.
        • Bille J.
        • Klopotov A.
        • et al.
        Comparison of automated strategies for surveillance of nosocomial bacteremia.
        Infect Control Hosp Epidemiol Off J Soc Hosp Epidemiol Am. 2007; 28: 1030-1035