Cumulative proportion of patients surviving according to nosocomial and community-associated pneumococcal bacteremia status.
Lyytikäinen O, Klemets P, Ruutu P, Kaijalainen T, Rantala M, Ollgren J, Nuorti JP. Defining the Population-Based Burden of Nosocomial Pneumococcal Bacteremia. Arch Intern Med. 2007;167(15):1635-1640. doi:10.1001/archinte.167.15.1635
The characteristics, risk factors, and outcome of patients with nosocomial pneumococcal bacteremia (NPB) have not been described in large, population-based studies.
All episodes of invasive pneumococcal infections reported by Finnish clinical microbiology laboratories (positive blood or cerebrospinal fluid culture) from January 1, 1995, through December 31, 2002, were linked to data in national health care registries and vital statistics to obtain information on the patient's preceding hospitalizations, comorbidities, and outcome of illness. Pneumococcal bacteremia was defined as nosocomial if the first positive blood culture was obtained more than 2 days after hospital admission, or if the patient had been hospitalized for more than 2 days within 7 days of the first positive blood culture.
Information on hospital admission was available for 4217 of 4357 persons (96.8%) with invasive pneumococcal infections. We identified 387 NPBs (9.7%) among 3973 pneumococcal bacteremias. Patients with NPB were older (median age, 67 years vs 52 years; P < .001) and were more likely to have at least 1 high-risk condition (other than age ≥ 65 years), for which 23-valent pneumococcal polysaccharide vaccine is recommended (59.2% vs 34.6%; P < .001), compared with patients who had community-associated pneumococcal bacteremias. The case fatality proportion at 28 days was higher in patients with NPB than in those with community-associated pneumococcal bacteremias (23.8% vs 10.8%; P < .001). Pneumococcal serotypes included in 23-valent polysaccharide vaccine and 7-valent conjugate vaccine caused 71.5% and 46.1% of NPBs, respectively.
A substantial proportion of pneumococcal bacteremias are health care associated. The high prevalence of conditions for which pneumococcal polysaccharide vaccine is recommended provides opportunities for strengthening prevention efforts in these patients at high risk of illness and death.
Streptococcus pneumoniae predominantly causes community-acquired respiratory tract, central nervous system, and bloodstream infections, but its role in health care–associated infections has not been well defined.1,2 Institutional outbreaks of multidrug-resistant S pneumoniae have been reported, and the prevalence of antimicrobial-resistant strains may be higher in older adults living in long-term care facilities.3- 6 Invasive pneumococcal disease is an important cause of illness and death in older adults living in long-term care facilities, emphasizing the need for better prevention efforts through immunization.7
The epidemiology of nosocomial pneumococcal bacteremia (NPB) has previously been studied primarily in single hospitals, which may not be representative of all health care facilities serving the population.8- 16 No national, population-based studies have compared the rates, patient characteristics, risk factors, and outcome of NPB with those of community-associated pneumococcal bacteremia (CAPB). We analyzed data on all episodes of invasive pneumococcal infections (IPI) identified by Finnish microbiology laboratories performing blood and cerebrospinal fluid (CSF) cultures and reporting isolations of S pneumoniae from blood or CSF during 1995 through 2002. To determine the proportion of pneumococcal infections that were health care associated, we identified preceding hospitalizations for all persons with pneumococcal bacteremias. Surveillance data were linked to Finnish population-based health care registries to determine the incidence, preexisting comorbid conditions, serotype distribution, and outcome of illness in patients with NPB and to compare these characteristics with those of patients who had CAPB.
In Finland (population, 5.2 million), the National Health Care System is organized into 20 geographically and administratively defined health care districts. Five university hospitals provide tertiary care services and have catchment populations ranging from 0.71 million to 1.67 million.
All Finnish clinical microbiology laboratories report all bacterial isolations from blood and CSF, including S pneumoniae, to the National Infectious Disease Register (NIDR). Most laboratory reporting is done electronically. With each notification, the following information is transmitted: date and type of specimen, date of birth, sex, and place of treatment. With this information and a time interval of 3 months, possible multiple positive culture results or notifications of the same person are merged as a single case.
A case of IPI was defined as isolation of S pneumoniae from blood and/or CSF from January 1, 1995, through December 31, 2002. Of the total of 4611 IPI episodes identified, only the first episode for each case patient was included in the analysis (n = 4357): 4106 cases were pneumococcal bacteremias, defined as isolation of S pneumoniae from blood only, and 251 cases were pneumococcal meningitis, defined as isolation of S pneumoniae from CSF with or without pneumococcal bacteremia within 7 days.
Data on the hospitalizations of patients with IPI within 7 days before the first positive blood culture specimen for S pneumoniae were retrieved from the National Hospital Discharge Registry (HILMO) by using the case patient's national identity code for database linkage. This registry contains comprehensive health care records from all hospitals and municipal health centers, including outpatient surgery. Each HILMO record includes patient identifying information, admission and discharge dates, health care provider, type of service, medical specialty, the place (home or institution) from which the patient was transferred to the facility, and data on surgical procedures and discharge diagnoses. Data are coded according to the International Classification of Diseases (ICD), Ninth Revision or 10th Revision (from 1996 onward). Information on hospitalizations was available for 4217 (96.8%) of case patients with IPI, of whom 3973 had bacteremias. Pneumococcal bacteremia was defined as nosocomial if the first positive blood culture was obtained more than 2 days after admission, or if the patient had been hospitalized for more than 2 days within 7 days before the first positive blood culture.
To obtain information on comorbidities and underlying conditions for patients with IPI, we linked our database to national population-based health care registries by using the national identity code: the Cancer Registry (diagnosis of hematologic and nonhematologic malignant neoplasm within 1 year before the first positive specimen date), National Infectious Disease Register (human immunodeficiency virus [HIV] infection), the National Social Insurance Institution, and HILMO. Presence of the following chronic underlying diseases was defined as a National Social Insurance Institution record indicating an entitlement for special reimbursement on medications for these conditions: diabetes mellitus, chronic pulmonary disease (chronic obstructive pulmonary disease and asthma), congenital or acquired immunodeficiency, rheumatic and other autoimmune diseases requiring immunosuppressive therapy, solid organ and bone marrow transplantation, cardiac failure, and renal failure. Alcohol-related diseases, chronic liver diseases, diseases of the spleen, and CSF leak were defined as records in HILMO with 1 or more ICD-9 or ICD-10 coded discharge diagnoses within 1 year before the first positive specimen date (codes available at http://www.ktl.fi/portal/12494). The case patients' vital status at 7, 28, and 90 days after the first positive culture of S pneumoniae was determined from the National Population Information System by use of the national identity codes. Research use of data from population-based registries was authorized by the Ministry of Social Affairs and Health, the Finnish Data Protection Authority, and the National Research and Development Center for Welfare and Health.
Streptococcus pneumoniae isolates sent to the reference laboratories at the National Public Health Institute as a part of national surveillance during 1995 through 2002 were serotyped by means of pneumococcal antiserum, and antimicrobial susceptibility testing was performed by Clinical Laboratory Standards Institute methods, as described previously.17,18
We calculated average annualized incidence rates during the surveillance period by using population data from Statistics Finland and the total number of patient-days from HILMO during 1995 through 2002 as denominators. The χ2 test or Fisher exact test was used to assess statistical significance for categorical variables and the Mann-Whitney test for continuous variables. P < .05 was considered statistically significant. Relative risks and 95% confidence intervals were calculated to compare patient and disease characteristics between NPBs and CAPBs; the relative risks were adjusted by multivariate logarithmic binomial regression. Univariate and multivariate logistic regression were used to assess the association of patient and disease characteristics with the outcome of death within 28 days (all-cause mortality). To estimate the hazard ratio of progression to death while adjusting for covariates, we used the Cox proportional hazards regression model in which age was treated as a continuous variable. Data were analyzed by using SPSS for Windows version 14.0 (SPSS Inc, Chicago, Illinois) and Stata, version 9.0 (StataCorp, College Station, Texas).
During the 8-year study period, 387 patients with NPB were identified (43-53 cases/y), representing approximately 10% of the 3973 hospitalized pneumococcal bacteremia cases (range by year, 7.5%-13.9%). The average annualized NPB incidence rate was 0.9 cases per 100 000 population (range by year, 0.8-1.0; range by tertiary care region, 0.7-1.2) and 0.66 cases per 100 000 patient-days (range by year, 0.60-0.73; range by tertiary care region, 0.49-0.95); the overall annualized CAPB incidence was 9.9 per 100 000 population (range by year, 7.5-11.0).
Among patients with NPB, the positive blood culture was obtained more than 2 days after hospital admission in 266 cases (68.7%). The remaining 121 patients (31.3%) had been hospitalized for more than 2 days within 7 days of the culture or had been admitted directly from another health care institution. Among the 266 cases, the median time from admission to the index blood culture was 8 days (range, 3-1087 days); 85.0% of index blood cultures were obtained within 28 days.
Many characteristics of patients with NPB differed from those of CAPB in univariate analysis (Table 1). Patients with NPB were significantly older than patients with CAPB (median age, 67 vs 52 years; P < .001). More than half of the patients with NPB were 65 years or older, and only 5.2% were younger than 16 years. Of patients with NPB, 229 (59.2%) had at least 1 underlying condition (other than age ≥ 65 years) for which 23-valent pneumococcal polysaccharide vaccine (PPV23) is currently recommended in Finland,19 compared with 1241 of 3586 patients (34.6%) with CAPB (P < .001). This difference was due to significantly greater proportions of chronic pulmonary disease, cardiac failure, diabetes mellitus, alcohol-related disease, nonhematologic and hematologic malignant neoplasms, and immunodeficiency or rheumatic diseases among patients with NPB (Table 1). Of the patients with NPB, none was asplenic or had a CSF leak or known HIV infection. Among 152 patients with NPB who were aged 16 to 64 years, 113 (74.3%) had at least 1 underlying condition for which PPV23 is recommended. Characteristics that were independently associated with higher risk of NPB in the logarithmic binomial regression model included male sex, increasing age, malignant neoplasms, chronic liver disease, alcohol-related disease, and chronic pulmonary disease (Table 1). There were no significant differences between the 266 hospital-associated and 121 health care–associated NPB cases by age or prevalence of underlying conditions (data not shown).
During the 90 days after the first positive blood culture, 121 (31.3%) of patients with NPB died. Of the deaths, 67 (55.4%) occurred within 7 days and 92 (76.0%) within 28 days; 88 (72.7%) were among patients 65 years or older. The overall case fatality proportions at 7, 28, and 90 days were significantly higher in patients with NPB than CAPB (17.3% vs 7.6%, 23.8% vs 10.8%, and 31.3% vs 13.4%, respectively; P < .001 for all comparisons). The Figure illustrates the Kaplan-Meier survival plots for NPB and CAPB. Case fatality proportions were nearly 1.5 times higher in males than in females and almost twice as high among the elderly (those ≥ 65 years) than in patients aged 16 to 64 years (Table 2). None of the children with NPB died. The case fatality proportions were highest among patients with nonhematologic malignant neoplasms, cardiac failure, alcohol-related diseases, hematologic malignant neoplasms, and immunodeficiency or rheumatic diseases (Table 2). Factors significantly associated with death within 28 days after positive blood culture (all-cause mortality) in univariate analyses were increasing age and cardiac failure (Table 3). After adjustment for age and sex in a multivariate logistic regression model, patient characteristics that independently predicted death within 28 days included immunodeficiency or rheumatic diseases (P = .004) and alcohol-related diseases (P = .01). We also evaluated the outcome of all-cause mortality after NPB by using a Cox proportional hazards regression model. In addition to the variables identified in multivariate logistic regression, nonhematologic and hematologic malignant neoplasms independently increased the risk (hazard) of death (Table 4).
A total of 319 S pneumoniae isolates from the 387 patients with NPB (82.4%) were available for serotyping. The most common serotypes/groups were 6 (13.2%), 14 (11.0%), 19 (9.4%), 4 (9.1%), and 23 (8.5%). Of the pneumococcal isolates, 228 (71.5%) were serotypes included in PPV23, and 147 (46.1%) were included in the 7-valent pneumococcal conjugate vaccine. Antimicrobial susceptibility testing results were available for 138 (35.7%) nosocomial S pneumoniae isolates: 6 isolates (4.3%) were resistant to erythromycin and 5 (3.6%) were intermediately resistant to penicillin; none of the patients with a nonsusceptible isolate died within 28 days.
Data from our national, population-based study indicate that about 10% of all pneumococcal bacteremias were health care associated and that mortality among patients who had NPB was more than twice as high as among patients with CAPB. Most of the patients with nosocomial disease had underlying conditions for which PPV23 is recommended, emphasizing the importance of strengthening prevention efforts in these patient groups.
This study provides a comprehensive evaluation of the burden of NPB, highlighting how its characteristics differ from those of CAPB. Our estimates from national laboratory-based surveillance are representative of the entire population of Finland. The standard case definition minimized selection bias due to different case mixes and provides accurate rates for comparison over time and place. The registry-based design enabled the use of accurate denominators for NPB rates (ie, total hospital patient-days) and differentiation between traditional nosocomial and health care–associated cases. Although there was little annual variation in rates of NPB per 100 000 population and patient-days during the study period, rates varied in different tertiary care regions, possibly reflecting small numbers of cases or local disease clusters.
The proportion of pneumococcal bacteremias among hospitalized patients that were nosocomial varied from 8% to 14% annually. This proportion is smaller than that previously reported from Spain (25%-41%)9,11 and the United States (27%-59%)8,10,12 but similar to that reported in France (10%)15 and recent reports from Spain (10%-14%).13,14,16 The reasons for these differences may be related to differences in study populations, small numbers of cases from individual hospitals,8- 14 and the case definitions used, but also to admission criteria and differences in blood culture sampling practices from patients with pneumonia between those who are hospitalized and outpatients.
The definitions used for nosocomial acquisition of disease have also varied in previous reports; some included episodes that manifested 48 hours or more after hospitalization,14 while others used a longer, 72-hour cutoff point.12,15,16 In addition to the episodes in which the index blood culture was obtained more than 2 days after hospitalization, our analysis included those that may have been associated with previous admission in the same or another health care setting within a 7-day time window. These potentially health care–associated infections accounted for about one-third of NPBs, and the patient characteristics in the 2 groups were similar. The median period of hospitalization before the index blood culture, 8 days, is consistent with previous reports, indicating that NPB tends to occur after a relatively prolonged hospitalization.14,16
Patients with NPB had significantly more chronic pulmonary diseases, cardiac failure, diabetes mellitus, alcohol-related disease, malignant neoplasms, and immunodeficiency or rheumatic disease than did patients with CAPB. No HIV-infected patients with NPB were identified, reflecting the low prevalence of HIV infections in Finland (10-16 per 100 000 population in 1995-2002).20 Nosocomial pneumococcal bacteremia has been previously associated with severe underlying conditions, including neoplasia, chronic obstructive pulmonary disease, heart failure, and cirrhosis.16 A case-control study in elderly veterans found that diagnoses of respiratory or hematologic malignant neoplasm, anemia, chronic obstructive pulmonary disease, and coronary artery disease on admission to the hospital were independent clinical predictors of NPB.12
In previous reports, definitions used for assessing mortality associated with NPB have been variable: some evaluated all-cause (or crude) mortality,12,14- 16 while others evaluated only mortality considered directly attributable to bacteremia.13 Analysis of vital status at discharge (in-hospital mortality), however, can be influenced by the length of stay and differences in the health care delivery system. By using data from national vital statistics, we were able to accurately determine the all-cause mortality within 90 days of the first blood culture positive for S pneumoniae, providing a more comprehensive assessment of this outcome. Although we were unable to distinguish between pneumococcal bacteremia-related and unrelated mortality, one-quarter of patients with NPB died within the first month and almost one-third had died at 3 months. The all-cause mortality was significantly higher in NPB than in CAPB. The risk of death increased with age, being 34% among persons 65 years or older and nearly 1.5 times higher in males than in females. Our time-dependent model showed that, in addition to age, male sex, and malignant neoplasms, risk of death was independently associated with immunodeficiency or rheumatic diseases and alcohol-related diseases. Previous studies on NPB have reported much higher case fatality proportions, ranging from 38% to 76%.8,9,13 This may be due to differences in study populations, as some included only adults12,13,16 or elderly men.10 In some studies, up to 57% of the patients with NPB had malignant neoplasms,9 likely reflecting a selection bias in hospital-based case ascertainment. Our data suggest that, in studies evaluating the clinical outcome of pneumococcal infections, mortality rates should be stratified according to nosocomial and community-associated cases.
Some limitations should be considered in interpreting our results. First, because of the registry-based study design, our analysis of mortality lacked chart review data to assess severity of illness and admission to intensive care.13,16 Second, information on some of the patients' comorbid conditions (alcohol-related diseases, chronic liver diseases, disease of the spleen, and CSF leaks) was based on ICD codes and could not be verified by chart reviews. Although the diagnostic codes in the National Social Insurance Institution's database are based on standard criteria, the ICD codes in hospital discharge data may be incomplete and could be subject to misclassification. Also, we did not have information on ongoing treatments for malignant neoplasms and therefore considered only malignant neoplasms diagnosed within 1 year of the culture date. Third, when classifying NPB cases as hospital acquired or health care associated, some cases in ambulatory care may have been misclassified, and it was not possible to evaluate the presence of symptoms of respiratory illness at admission. Fourth, we did not have information on receipt of pneumococcal polysaccharide vaccination and cigarette smoking status of case patients, as some of the associations we found with higher risk of NPB (eg, alcohol-related diseases and chronic obstructive pulmonary disease) could be confounded by smoking.21 However, the coverage of 23-valent polysaccharide vaccination among elderly persons and persons with high-risk conditions in Finland is very low (3%), and our data may be considered to represent a prepneumococcal vaccination baseline.22
More than 70% of the serotypes of the pneumococcal isolates from patients with NPB were those included in the PPV23, and 74.3% of the patients with NPB aged 16 to 64 years had at least 1 underlying condition for which PPV23 is currently recommended. Together with the high mortality, these data warrant increased efforts to improve the use of the PPV23 in persons with chronic illness in Finland. However, some of these persons may not respond well to polysaccharide vaccine.23 The routine 7-valent pneumococcal conjugate vaccine program in children in the United States has also substantially benefited adults through herd immunity, but persons with underlying illness may benefit less than healthier ones from the indirect effects of the vaccine.24 Because a substantial proportion of NPBs in Finland were caused by conjugate vaccine serotypes, a childhood immunization program might also have the potential to reduce NPB through indirect effects. Targeting persons with chronic conditions for administration of polysaccharide vaccine remains an established strategy for reaching those at highest risk for invasive infection. Our study offers a comprehensive assessment of disease burden, risk factors, and outcome of NPB in a well-defined, representative population. This information may be useful for policymakers in Finland and other industrialized countries who are assessing the benefits, cost-effectiveness, and appropriate target groups to strengthen immunization programs.
Correspondence: Outi Lyytikäinen, MD, PhD, Department of Infectious Disease Epidemiology, National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland (firstname.lastname@example.org).
Accepted for Publication: April 9, 2007.
Financial Disclosure: None reported.
Author Contributions:Study concept and design: Lyytikäinen and Nuorti. Acquisition of data: Klemets and Kaijalainen. Analysis and interpretation of data: Lyytikäinen, Klemets, Ruutu, Kaijalainen, Rantala, Ollgren, and Nuorti. Drafting of the manuscript: Lyytikäinen, and Nuorti. Critical revision of the manuscript for important intellectual content: Klemets, Ruutu, Kaijalainen, Rantala, Ollgren, and Nuorti. Statistical analysis: Lyytikäinen, Ollgren, and Nuorti. Administrative, technical, and material support: Klemets, Ruutu, Kaijalainen, and Ollgren. Study supervision: Nuorti.
Previous Presentations: This study was presented in part at the 15th Annual Scientific Meeting of the Society for Healthcare Epidemiology of America; April 9-12, 2005; Los Angeles, California (abstract 155).