Journal of Infection (2015) 70, 111e126
www.elsevierhealth.com/journals/jinf
Bacteremia is associated with excess long-term mortality: A 12-year population-based cohort study Stig Lønberg Nielsen a,*, Annmarie Touborg Lassen b, Kim Oren Gradel c,d, Thøger Gorm Jensen e, Hans Jørn Kolmos e, Jesper Hallas f, Court Pedersen a a
Department of Infectious Diseases, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark b Department of Emergency Medicine, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark c Center for Clinical Epidemiology, South, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark d Research Unit of Clinical Epidemiology, Institute of Clinical Research, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark e Department of Clinical Microbiology, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark f Research Unit of Clinical Pharmacology, University of Southern Denmark, Sdr. Boulevard 29, 5000 Odense C, Denmark Accepted 28 August 2014 Available online 9 September 2014
KEYWORDS Bacteremia; Epidemiology; Mortality; Population-based; Causes of death
Summary Objectives: Little is known about long-term outcomes following bacteremia. We investigated long-term mortality and causes of death among bacteremia patients compared with population controls. Methods: Population-based cohort study of bacteremia patients and population controls matched on sex, year of birth, residency and calendar time, in Denmark during 2000e2008. We calculated absolute mortality and adjusted mortality rate ratios (MRRs) in predefined follow-up periods. Results: The absolute mortality for bacteremia patients (n Z 7783) and population controls (n Z 38,906) was 22.0% vs. 0.2% (30 days), 41.4% vs. 2.6% (1 year) and 75.8% vs. 36.6% (10 years). For bacteremia patients, the MRR was 115.3 (95% CI, 88.2e150.9) 0e30 days after bacteremia and 2.1 (95% CI, 1.8e2.3) from 5 years to end of follow-up. The most common causes of death were cancer and cardiovascular diseases. Within one year of bacteremia,
* Corresponding author. Tel.: þ45 65412738; fax: þ45 66117418. E-mail address: [email protected] (S.L. Nielsen). http://dx.doi.org/10.1016/j.jinf.2014.08.012 0163-4453/ª 2014 The British Infection Association. Published by Elsevier Ltd. All rights reserved.
112
S.L. Nielsen et al. the relative risk of death was highest for genitourinary diseases and infectious diseases. Among one-year survivors of bacteremia, the relative risk of death was increased for all major causes of death. Conclusions: Bacteremia is associated with a poor prognosis and considerable excess long-term mortality compared with the general population. The most common causes of death after bacteremia are cancer and cardiovascular diseases. Summary: This population-based cohort study reports an excess long-term mortality among 7783 bacteremia patients compared with matched population controls during 12 years of follow-up. We identified patients in particular risk of death and reported novel information on causes of death. ª 2014 The British Infection Association. Published by Elsevier Ltd. All rights reserved.
Introduction Bacteremia is an important cause of mortality1,2 and ranks among the top seven causes of death in Europe and North America.3 Whereas the association between bacteremia and a high 30-day mortality is well-established,1e3 less is known about long-term mortality. Previous studies have reported mortality rates of 25e48% at 1 year,4e8 49% at 3 years5 and 63% at 4 years.6 These studies were either hospital-based,5e7 restricted to community-onset bacteremia,4 or had limited time of follow-up to one year.8 To date, mortality beyond 1 year remains unreported in a non-selected population of all adult bacteremia patients. Knowledge on the survival of bacteremia patients compared with the general population is important to inform health policy, plan patient follow-up and identify predictors of an unfavorable outcome. We are aware of only two studies that have made such a comparison. Leibovici et al. found that bacteremia was associated with excess mortality during 4 years of follow-up.6 In contrast, Skogberg et al. found that the survival of bacteremia patients was comparable to that of the general population after 60 days.2 However, none of the studies used a matched cohort design or adjusted for potential confounders. Whereas early deaths are likely to be related to bacteremia,6,7 late deaths are more likely to be influenced by new onset or worsening of pre-existing comorbidity.6,9 Knowledge on causes of death following bacteremia could help to determine if increased mortality is a direct consequence of the infection or whether bacteremia is a marker of new or existing illness such as solid cancers, hematological malignancies or cardiovascular diseases. However, previous studies have rarely reported causes of death and no comparison has been made with causes of death in the general population.4,6 We therefore conducted a population-based cohort study of adults with first-time bacteremia during 2000e2008 with follow-up until the end of 2011 to investigate long-term mortality and causes of death compared with the general population.
Free tax-funded healthcare was provided at one university hospital with 1100 beds and one smaller community hospital. All specialties were represented and only patients requiring allogenic bone marrow or solid organ transplantation (except kidney) were referred out of Funen County for care.
Study population We identified all adult (15 years old) first-time bacteremia patients during 2000e2008 in a research database named The Danish Observational Registry of Infectious Syndromes (DORIS) that contains data on all positive blood cultures drawn in Funen County from May 1999 through December 2008. Blood cultures consisted of two aerobic/anaerobic bottle sets and were all analyzed at the Department of Clinical Microbiology, Odense University Hospital, using methods as described previously.10 We defined bacteremia as (1) recognized pathogens detected in 1 blood culture, or (2) common skin contaminants (coagulase-negative staphylococci, Bacillus spp, Propionibacterium spp, Corynebacterium spp, viridans group streptococci, Aerococcus spp, or Micrococcus spp) detected in 2 blood culture sets within 5 days.11e13 The date of the first positive blood culture set was regarded as the date of bacteremia. Polymicrobial bacteremia was defined as isolation of 2 different microorganisms within 2 days.14 Bacteremias were classified according to place of acquisition. Nosocomial acquisition was defined as draw of the first positive blood culture >2 days after admission. Healthcare-associated acquisition was defined as draw of the first positive blood culture 2 days after admission and either discharge from a hospital or attendance at an outpatient clinic (hematology, nephrology, or oncology) within 30 days prior to the admission. Community-acquired acquisition was defined as draw of the first positive blood culture 2 days after admission while not fulfilling the criteria for healthcareassociated acquisition.1
Population control cohort
Methods Study setting The study was conducted in Funen County, Denmark, with approximately 390,000 adult (15 years old) residents.
For each bacteremia patient, we randomly sampled 5 population controls matched on sex, year of birth and residency (Funen County) on an index date identical to the date of bacteremia of their corresponding case. To avoid immortal time bias, bacteremia patients were eligible as population controls until their first bacteremia. Three
Long-term mortality following bacteremia bacteremia patients (0.04%) had no controls and were excluded, and 13 bacteremia patients (0.16%) had less than 5 controls. Accordingly, the bacteremia: control ratio differed slightly from 1:5.
Data sources We used the unique Danish personal identification number15 assigned to all residents of Denmark at birth or immigration to retrieve information on characteristics and outcome of bacteremia patients and population controls by crosslinkage with the following registries: From The Danish Civil Registration System15 we extracted data on place of residence, marital status, and date of loss to follow-up (mainly emigration) or death. From The Danish Register of Causes of Death16 we extracted data on the underlying cause of death, which was grouped as shown in Appendix S1. From The Danish Cancer Registry17 we extracted data on all diagnosed cancers including date of diagnosis and characteristics. From The Danish National Hospital Register18 and The Danish Psychiatric Central Research Register,19 we extracted data on dates of admission, discharge and outpatient contacts to define place of acquisition. Further, we extracted data on discharge diagnoses (classified according to The International Classification of Diseases, 10th Revision) to define the patients’ comorbid status. Finally, we extracted data on redeemed disulfiram from Odense Pharmacoepidemiological Database20 to define a history of alcohol dependency.
Outcome Study outcomes were death from any cause (all-cause mortality) and death from a specific underlying cause of death (cause-specific mortality).
Confounders
113 and 10 years for both bacteremia patients and population controls. For all-cause mortality, we reported mortality rates (MRs) as deaths per 1000 person years and used Cox regression models to calculate mortality rate ratios (MRRs) in predefined follow-up periods after the index date (0e30 days, 31e90 days, 91e365 days, 1 yeare5 years and 5þ years). In a sub-analysis, we excluded bacteremia patients and population controls who were diagnosed with cancer within /þ 1 year of the index date. Further, we performed subgroup analyses for sex, place of acquisition, clinical department, Charlson Index, age groups (15e39, 40e64, 65e79, 80þ years), and groups of microorganism(s) in follow-up periods of 0e1 year, 1e5 years and 5þ years. For cause-specific mortality, we computed cause-specific mortality rates as deaths per 1000 person-years and used Cox regression models to calculate MRRs for each specific cause of death 0e1 year and 1þ year after the index date.23 To account for the matched design, we used Cox regression models (stratified on matched sets) and adjusted for the above-mentioned potential confounders. The proportional hazard assumptions were assessed graphically and found appropriate. Finally, we compared the proportions of cancer-specific deaths in bacteremia patients and population controls using the chi-squared test. P values
www.elsevierhealth.com/journals/jinf
Bacteremia is associated with excess long-term mortality: A 12-year population-based cohort study Stig Lønberg Nielsen a,*, Annmarie Touborg Lassen b, Kim Oren Gradel c,d, Thøger Gorm Jensen e, Hans Jørn Kolmos e, Jesper Hallas f, Court Pedersen a a
Department of Infectious Diseases, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark b Department of Emergency Medicine, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark c Center for Clinical Epidemiology, South, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark d Research Unit of Clinical Epidemiology, Institute of Clinical Research, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark e Department of Clinical Microbiology, Odense University Hospital, Sdr. Boulevard 29, 5000 Odense C, Denmark f Research Unit of Clinical Pharmacology, University of Southern Denmark, Sdr. Boulevard 29, 5000 Odense C, Denmark Accepted 28 August 2014 Available online 9 September 2014
KEYWORDS Bacteremia; Epidemiology; Mortality; Population-based; Causes of death
Summary Objectives: Little is known about long-term outcomes following bacteremia. We investigated long-term mortality and causes of death among bacteremia patients compared with population controls. Methods: Population-based cohort study of bacteremia patients and population controls matched on sex, year of birth, residency and calendar time, in Denmark during 2000e2008. We calculated absolute mortality and adjusted mortality rate ratios (MRRs) in predefined follow-up periods. Results: The absolute mortality for bacteremia patients (n Z 7783) and population controls (n Z 38,906) was 22.0% vs. 0.2% (30 days), 41.4% vs. 2.6% (1 year) and 75.8% vs. 36.6% (10 years). For bacteremia patients, the MRR was 115.3 (95% CI, 88.2e150.9) 0e30 days after bacteremia and 2.1 (95% CI, 1.8e2.3) from 5 years to end of follow-up. The most common causes of death were cancer and cardiovascular diseases. Within one year of bacteremia,
* Corresponding author. Tel.: þ45 65412738; fax: þ45 66117418. E-mail address: [email protected] (S.L. Nielsen). http://dx.doi.org/10.1016/j.jinf.2014.08.012 0163-4453/ª 2014 The British Infection Association. Published by Elsevier Ltd. All rights reserved.
112
S.L. Nielsen et al. the relative risk of death was highest for genitourinary diseases and infectious diseases. Among one-year survivors of bacteremia, the relative risk of death was increased for all major causes of death. Conclusions: Bacteremia is associated with a poor prognosis and considerable excess long-term mortality compared with the general population. The most common causes of death after bacteremia are cancer and cardiovascular diseases. Summary: This population-based cohort study reports an excess long-term mortality among 7783 bacteremia patients compared with matched population controls during 12 years of follow-up. We identified patients in particular risk of death and reported novel information on causes of death. ª 2014 The British Infection Association. Published by Elsevier Ltd. All rights reserved.
Introduction Bacteremia is an important cause of mortality1,2 and ranks among the top seven causes of death in Europe and North America.3 Whereas the association between bacteremia and a high 30-day mortality is well-established,1e3 less is known about long-term mortality. Previous studies have reported mortality rates of 25e48% at 1 year,4e8 49% at 3 years5 and 63% at 4 years.6 These studies were either hospital-based,5e7 restricted to community-onset bacteremia,4 or had limited time of follow-up to one year.8 To date, mortality beyond 1 year remains unreported in a non-selected population of all adult bacteremia patients. Knowledge on the survival of bacteremia patients compared with the general population is important to inform health policy, plan patient follow-up and identify predictors of an unfavorable outcome. We are aware of only two studies that have made such a comparison. Leibovici et al. found that bacteremia was associated with excess mortality during 4 years of follow-up.6 In contrast, Skogberg et al. found that the survival of bacteremia patients was comparable to that of the general population after 60 days.2 However, none of the studies used a matched cohort design or adjusted for potential confounders. Whereas early deaths are likely to be related to bacteremia,6,7 late deaths are more likely to be influenced by new onset or worsening of pre-existing comorbidity.6,9 Knowledge on causes of death following bacteremia could help to determine if increased mortality is a direct consequence of the infection or whether bacteremia is a marker of new or existing illness such as solid cancers, hematological malignancies or cardiovascular diseases. However, previous studies have rarely reported causes of death and no comparison has been made with causes of death in the general population.4,6 We therefore conducted a population-based cohort study of adults with first-time bacteremia during 2000e2008 with follow-up until the end of 2011 to investigate long-term mortality and causes of death compared with the general population.
Free tax-funded healthcare was provided at one university hospital with 1100 beds and one smaller community hospital. All specialties were represented and only patients requiring allogenic bone marrow or solid organ transplantation (except kidney) were referred out of Funen County for care.
Study population We identified all adult (15 years old) first-time bacteremia patients during 2000e2008 in a research database named The Danish Observational Registry of Infectious Syndromes (DORIS) that contains data on all positive blood cultures drawn in Funen County from May 1999 through December 2008. Blood cultures consisted of two aerobic/anaerobic bottle sets and were all analyzed at the Department of Clinical Microbiology, Odense University Hospital, using methods as described previously.10 We defined bacteremia as (1) recognized pathogens detected in 1 blood culture, or (2) common skin contaminants (coagulase-negative staphylococci, Bacillus spp, Propionibacterium spp, Corynebacterium spp, viridans group streptococci, Aerococcus spp, or Micrococcus spp) detected in 2 blood culture sets within 5 days.11e13 The date of the first positive blood culture set was regarded as the date of bacteremia. Polymicrobial bacteremia was defined as isolation of 2 different microorganisms within 2 days.14 Bacteremias were classified according to place of acquisition. Nosocomial acquisition was defined as draw of the first positive blood culture >2 days after admission. Healthcare-associated acquisition was defined as draw of the first positive blood culture 2 days after admission and either discharge from a hospital or attendance at an outpatient clinic (hematology, nephrology, or oncology) within 30 days prior to the admission. Community-acquired acquisition was defined as draw of the first positive blood culture 2 days after admission while not fulfilling the criteria for healthcareassociated acquisition.1
Population control cohort
Methods Study setting The study was conducted in Funen County, Denmark, with approximately 390,000 adult (15 years old) residents.
For each bacteremia patient, we randomly sampled 5 population controls matched on sex, year of birth and residency (Funen County) on an index date identical to the date of bacteremia of their corresponding case. To avoid immortal time bias, bacteremia patients were eligible as population controls until their first bacteremia. Three
Long-term mortality following bacteremia bacteremia patients (0.04%) had no controls and were excluded, and 13 bacteremia patients (0.16%) had less than 5 controls. Accordingly, the bacteremia: control ratio differed slightly from 1:5.
Data sources We used the unique Danish personal identification number15 assigned to all residents of Denmark at birth or immigration to retrieve information on characteristics and outcome of bacteremia patients and population controls by crosslinkage with the following registries: From The Danish Civil Registration System15 we extracted data on place of residence, marital status, and date of loss to follow-up (mainly emigration) or death. From The Danish Register of Causes of Death16 we extracted data on the underlying cause of death, which was grouped as shown in Appendix S1. From The Danish Cancer Registry17 we extracted data on all diagnosed cancers including date of diagnosis and characteristics. From The Danish National Hospital Register18 and The Danish Psychiatric Central Research Register,19 we extracted data on dates of admission, discharge and outpatient contacts to define place of acquisition. Further, we extracted data on discharge diagnoses (classified according to The International Classification of Diseases, 10th Revision) to define the patients’ comorbid status. Finally, we extracted data on redeemed disulfiram from Odense Pharmacoepidemiological Database20 to define a history of alcohol dependency.
Outcome Study outcomes were death from any cause (all-cause mortality) and death from a specific underlying cause of death (cause-specific mortality).
Confounders
113 and 10 years for both bacteremia patients and population controls. For all-cause mortality, we reported mortality rates (MRs) as deaths per 1000 person years and used Cox regression models to calculate mortality rate ratios (MRRs) in predefined follow-up periods after the index date (0e30 days, 31e90 days, 91e365 days, 1 yeare5 years and 5þ years). In a sub-analysis, we excluded bacteremia patients and population controls who were diagnosed with cancer within /þ 1 year of the index date. Further, we performed subgroup analyses for sex, place of acquisition, clinical department, Charlson Index, age groups (15e39, 40e64, 65e79, 80þ years), and groups of microorganism(s) in follow-up periods of 0e1 year, 1e5 years and 5þ years. For cause-specific mortality, we computed cause-specific mortality rates as deaths per 1000 person-years and used Cox regression models to calculate MRRs for each specific cause of death 0e1 year and 1þ year after the index date.23 To account for the matched design, we used Cox regression models (stratified on matched sets) and adjusted for the above-mentioned potential confounders. The proportional hazard assumptions were assessed graphically and found appropriate. Finally, we compared the proportions of cancer-specific deaths in bacteremia patients and population controls using the chi-squared test. P values
