- Top of page
OBJECTIVE: To describe the presentation, resolution of symptoms, processes of care, and outcomes of pneumococcal pneumonia, and to compare features of the bacteremic and nonbacteremic forms of this illness.
DESIGN: A prospective cohort study.
SETTING: Five medical institutions in 3 geographic locations.
PARTICIPANTS: Inpatients and outpatients with community-acquired pneumonia (CAP).
MEASUREMENTS: Sociodemographic characteristics, respiratory and nonrespiratory symptoms, and physical examination findings were obtained from interviews or chart review. Severity of illness was assessed using a validated prediction rule for short-term mortality in CAP. Pneumococcal pneumonia was categorized as bacteremic; nonbacteremic, pure etiology; or nonbacteremic, mixed etiology.
MAIN RESULTS: One hundred fifty-eight (6.9%) of 2,287 patients (944 outpatients, 1,343 inpatients) with CAP had pneumococcal pneumonia. Sixty-five (41%) of the 158 with pneumococcal pneumonia were bacteremic; 74 (47%) were nonbacteremic with S. pneumoniae as sole pathogen; and 19 (12%) were nonbacteremic with S. pneumoniae as one of multiple pathogens. The pneumococcal bacteremia rate for outpatients was 2.6% and for inpatients it was 6.6%. Cough, dyspnea, and pleuritic pain were common respiratory symptoms. Hemopytsis occurred in 16% to 22% of the patients. A large number of nonrespiratory symptoms were noted. Bacteremic patients were less likely than nonbacteremic patients to have sputum production and myalgias (60% vs 82% and 33% vs 57%, respectively; P < .01 for both), more likely to have elevated blood urea nitrogen and serum creatinine levels, and more likely to receive pencillin therapy. Half of bacteremic patients were in the low risk category for short-term mortality (groups I to III), similar to the nonbacteremic patients. None of the 32 bacteremic patients in risk groups I to III died, while 7 of 23 (30%) in risk group V died. Intensive care unit admissions and pneumonia-related mortality were similar between bacteremic and nonbacteremic groups, although 46% of the bacteremic group had respiratory failure compared with 32% and 37% for the other groups. The nonbacteremic pure etiology patients returned to household activities faster than bacteremic patients. Symptoms frequently persisted at 30 days: cough (50%); dyspnea (53%); sputum production (48%); pleuritic pain (13%); and fatigue (63%).
CONCLUSIONS: There were few differences in the presentation of bacteremic and nonbacteremic pneumococcal pneumonia. About half of bacteremic pneumococcal pneumonia patients were at low risk for mortality. Symptom resolution frequently was slow.
Pneumonia due to Streptococcus pneumoniae continues to be a challenge. In the preantibiotic era it accounted for most cases of pneumonia. 1 Currently it causes about 6% to 10% of all cases of community-acquired pneumonia (CAP) 2–4; however, it causes 60% of all cases of bacteremic pneumonia.
Although pneumococcal pneumonia has been studied for decades, there are still many unanswered questions about this illness, including the apparent differences in mortality rate between countries and the failure of antibiotic therapy to influence the high early mortality rate. In 1 study, the mortality rate from bacteremic pneumococcal pneumonia was 26% in Huntington, WV, while it was only 5% in Stockholm, Sweden. 5 The mortality rate from bacteremic S. pneumoniae pneumonia is highest in the first 48 hours of hospitalization 6 and is unchanged from the preantibiotic era. 7 Furthermore, some investigators have found that even treatment in an intensive care unit (ICU) does not lower early mortality. 8 Therapy of pneumococcal pneumonia used to be simple, but the emergence of antibiotic-resistant S. pneumoniae strains has complicated treatment of this common infection. 9,10
From October 1991 through March 1994, we conducted a prospective observational study of patients with ambulatory and hospitalized CAP at 1 site in Halifax, Nova Scotia, Canada; 2 sites in Boston, Mass, and 2 sites in Pittsburgh, Pa. Our objectives were to examine a variety of outcomes in patients with pneumococcal pneumonia (resolution of symptoms, time to return to work, time to return to usual activities) that have not been examined in previous studies. An additional objective of our study was to compare the clinical and other features of bacteremic pneumococcal pneumonia with nonbacteremic S. pneumoniae as a single pathogen, and cases of pneumonia in which S. pneumoniae was part of a polymicrobial etiology.
- Top of page
About 7% of the 2,287 CAP patients in our study had S. pneumoniae. This is probably an underestimate of the number of cases due to this microorganism since diagnostic studies were pursued at the discretion of the attending physician. This rate of pneumococcal pneumonia as a percentage of all cases of CAP is far below the 20% to 60% rate quoted for North America and the 60% to 75% rate for Britain. 14 There are several reasons for this discrepancy. The studies quoted probably do not reflect the current situation. A carefully done study by Mundy et al. 3 at Johns Hopkins Hospital in Baltimore from November 1990 to November 1991 found that 15.1% of 201 HIV-uninfected patients with CAP requiring hospitalization had pneumococcal pneumonia. This was a marked decrease in the proportion of pneumonia due to S. pneumoniae, compared with a study done at Johns Hopkins Hospital during 1965-1966 in which 62% of patients had pneumococcal pneumonia. 3 Another reason for the low rate of pneumococcal pneumonia in our study is that outpatients with pneumonia were included in the study population and this group had few diagnostic tests performed, so we do not know the true rate of S. pneumoniae pneumonia in this group. Among our inpatients, only 4.6% had bacteremic pneumococcal pneumonia; however, this compares favorably with 6.3% in Mundy's study, 3 and with the 5.5% rate in a study of 2,776 patients with CAP requiring hospitalization in Franklin and Summit Counties, Ohio in 1991. 15 The percentage of cases of CAP due to S. pneumoniae depends upon the diagnostic techniques used, the assiduousness with which specimens are collected, and the rapidity with which they are transported to the microbiology laboratory and processed for culture.
The most sensitive diagnostic test for diagnosis of pneumococcal pneumonia is detection of pneumolysin by polymerase chain reaction, or by detecting immune complexes of pneumolysin antigen-antibody 16,17 or pneumolysin antibodies. 17 Using these techniques, pneumococcal pneumonia was diagnosed in 32% to 55% of patients with CAP. 16,17 In our study, the rate of pneumococcal pneumonia was highest at the MGH, Boston site. The higher rate at the Boston site reflects a higher rate of sputum specimens processed for culture within 24 hours of presentation at this site; 50% of the patients hospitalized at MGH had such a specimen compared with 15% at VGH, Halifax. The pneumococcal bacteremia rate among inpatients was 6.4% at Halifax and 5.4% at MGH, Boston, indicating that the overall rate of pneumococcal pneumonia was probably the same at both sites.
Forty-one percent of the 158 patients with pneumococcal pneumonia were bacteremic. In Burman's study, 16 in which a number of serological techniques were employed to diagnose pneumococcal pneumonia, 21% of the patients were bacteremic.
Streptococcus pneumoniae is found in the upper respiratory tract of healthy persons, 18 so its isolation from sputum cannot be considered proof that the pneumonia is due to the pneumococcus. For this reason, some investigators do not use the results of sputum culture to assign etiology. 19 However, in practice, clinicians equate isolation of S. pneumoniae from the sputum with pneumococcal pneumonia. It is thus important to compare definite (i.e., bacteremic) pneumococcal pneumonia with presumptive (i.e., nonbacteremic) pneumococcal pneumonia. There is also a third group of patients—those who have S. pneumoniae and other respiratory pathogens implicated. We compared all 3 groups of patients. There were no differences in demographic features, comorbidities, or severity of illness at presentation.
Fine et al. 11 derived a prediction rule that stratifies patients with CAP into 5 classes with respect to the risk of death within 30 days. The mortality rates for the entire PORT study cohort of patients in risk classes I, II, and III were 0.1%, 0.6% and 0.9%, respectively. For classes IV and V, the rates were 9.3% and 27.0%. 11 Almost half (49.2%) of the patients with bacteremic pneumococcal pneumonia in the current study were in classes I to III, and none of them died within 30 days of presentation. In contrast, 33% of those in class V died. This is the first time that this prediction rule has been applied to patients with bacteremic pneumococcal pneumonia. It clearly illustrates that some of these patients fall into a low risk group for mortality. It is likely that our data underestimate the proportion of patients with bacteremic pneumococcal pneumonia who are in risk classes I to III, since most of the patients treated on an ambulatory basis did not have blood cultures done. The 2 outpatients with bacteremic pneumococcal pneumonia in our study responded well to treatment with oral antibiotics. However, additional data are needed before outpatient therapy of bacteremic pneumococcal pneumonia with oral antibiotics can be recommended.
The patients with bacteremic pneumococcal pneumonia were more likely to be treated with natural penicillins than were the other two groups. The sequence of events in terms of antimicrobial therapy was that at the time of admission empiric therapy usually consisted of 2 antibiotics. 20 The choice varied according to geographic site; e.g., cefuroxime plus erythromycin was most commonly used in Halifax, while cefuroxime plus gentamicin was most commonly used in Boston. 20 When the blood culture result was obtained, 64.6% of the bacteremic patients were then switched to penicillin alone. In contrast, patients with nonbacteremic, pure etiology, pneumococcal pneumonia were rarely treated with penicillin only. The reasons for this cannot be clearly ascertained from our study, but it is likely that physicians remained uncertain that S. pneumoniae was the cause of the pneumonia.
The current increase in penicillin-resistant S. pneumoniae in the United States has prompted infectious disease clinicians to change the way they treat pneumococcal infections. Seventy-two percent of respondents to an Infectious Diseases Society of America questionnaire indicated that they would select cefotaxime and/or ceftriaxone to treat patients with a presumptive diagnosis of pneumococcal pneumonia and probable bacteremia; 19% would select vancomycin. 21 There was no high-level penicillin resistance among the 50 isolates tested from bacteremic patients; 2 (4%) isolates had intermediate resistance. Since our study enrolled patients from October 1991 through March 1994, the pneumococcal penicillin susceptibility data are already dated. Currently there is rapid evolution of penicillin-resistant S. pneumoniae in North America. A study by Butler et al. 22 of pneumococcal isolates recovered from normally sterile body sites of patients at 12 hospitals in 11 states during 1993-1994 found that 14.1% of 740 isolates were penicillin-nonsusceptible (intermediate resistance MIC ≥ 0.1 μg/mL) and 3.2% had high-level resistance (MIC ≥ 2.0 μg/mL). One of the participating hospitals in Butler's study was in Pennsylvania.
Overall, 13.3% of our patients were admitted to an ICU and 15.4% of bacteremic patients were admitted to an ICU, compared with 12.2% and 10.5% for the pure etiology and mixed etiology of two nonbacteremic groups, respectively. In studies that have focused on severe CAP requiring admission to ICU, S. pneumoniae is the most common etiology, accounting for 18 to 37% of the cases. 23,24
The pneumonia-related mortality rate was lowest (but not statistically signficant) for the nonbacteremic, pure etiology group at 2.7% and highest for the bacteremic group at 7.7%. The mortality rate in bacteremic pneumococcal pneumonia has been related to factors that also apply to mortality for pneumonia in general, such as age, 25 alcoholism, 26 and involvement of two or more lobes radiographically. 8,27 There are also organism-specific factors that influence mortality. In Austrian and Gold's study of 529 patients, the mortality rate for those infected with S. pneumoniae capsular polysaccharide type I was 8%, and for type III, it was 50%. 6
The complication rate among the bacteremic patients was higher than among the nonbacteremic groups, but only achieved statistical significance for anemia and renal insufficiency. Tilghman and Finland, 27 in a study of 1,586 cases of pneumococcal pneumonia (582 were bacteremic) from 1929 to 1935 (preantibiotic era), showed that complications occurred twice as often in the bacteremic group. Twelve percent of their entire group had a suppurative complication; empyema was most common at 7%, followed by pulmonary abscess (2.7%), endocarditis (1%), and meningitis (1.2%). Our suppurative complication rate was 3.8% overall and 6.2% for the bacteremic group. Our empyema rate was 1.9%, while the rate of endocarditis was 0.6%. None of our patients had a complication of meningitis.
The leukopenia rate was 2.7 times higher in the bacteremic group than in the nonbacteremic groups (7.7% vs 2.7%), but the difference was not significant. In a review of 93 episodes of pneumococcal bacteremia, Perlino and Rimland 28 noted that 15 patients were leukopenic and 12 (80%) died compared with 40 (51%) of 78 of the nonleukopenic patients (P < .05). They also found that 12 of the 15 leukopenic patients had alcoholism; the bacteremia was fatal in 10 of these 12 patients. Three of our 5 bacteremia patients with leukopenia had alcoholism.
The Centers for Disease Control recently reported three outbreaks of pneumococcal pneumonia among residents of nursing homes in Massachusetts, Oklahoma, and Maryland. 29 Almost 8% of the bacteremic pneumococcal pneumonia patients in our study were from a nursing home. This compares with 14% for the entire group of 1,343 inpatients with pneumonia. These data indicate that pneumococcal pneumonia is an important problem in nursing homes despite the policy of offering pneumococcal vaccine to nursing home residents. Pneumococcal vaccine was infrequently utilized (at least according to self-reports) by our patients.
While most of the discussion has focused on a comparison between bacteremic and nonbacteremic pneumococcal pneumonia, we also defined the symptoms of pneumococcal pneumonia. The slow resolution of these symptoms is reflected by the finding that 63% of patients complained of fatigue at 30 days, and about half still had symptoms of cough, dyspnea, and sputum production. Thirteen percent still had pleuritic chest pain. In a previous study, we found that older patients with pneumonia complained of fewer symptoms; patients aged 45-64, 65-74, and ≥75 years had 1.4, 2.9, and 3.3 fewer symptoms, respectively, than patients aged 18-44. Persistence of symptoms at 30-day follow-up was common. 30
Our study has several strengths and limitations. The strengths are the prospective nature of the study, its large size, geographic diversity (United States and Canada), use of a validated pneumonia-specific severity of illness scoring system, 30-day follow-up, and attention to important outcomes, such as time to return to usual activities, that have not been studied previously in this group of patients. The limitations are the observational nature of the study (some important investigations, such as blood and sputum cultures, were performed at the discretion of the attending physicians) and the incomplete enrollment (some patients did not wish to participate). These limitations may affect 1 of our key findings, namely that 50% of the bacteremic pneumococcal pneumonia patients were in the low risk group for mortality. However, a large multicenter study of pneumococcal bacteremia 31 found that 51% had an APACHE II score (an acute physiology and chronic illness score) of <12 and a mortality rate of 2.6%.
In conclusion, among patients with CAP, there is a low rate of bacteremic pneumococcal pneumonia, and 50% of these patients are at low risk for mortality. The time to return to usual household activities is significantly longer for bacteremic than nonbacteremic patients with pneumococcal pneumonia. Antibiotic therapy seems to be influenced by the results of blood cultures, since a significantly greater number of the bacteremic group received penicillin therapy. Finally, resolution of symptoms in patients with pneumococcal pneumonia can take more than 1 month.
- Top of page
This work was conducted as part of the Pneumonia PORT project, funded by a grant from the Agency for Health Care Policy and Research (R01 HS-06468)
Dr. Fine is supported as a Robert Wood Johnson Foundation Generalist Physician Faculty Scholar.
We thank research assistants Rhonda Grandy, RN, Dawn Menon, GN, Linda Kraft, RN, Jackie Cunning, RN and Maxine Young, RN (Halifax); Mary Ungaro, RN, Leila Hadad, AB (Boston); Mary Walsh, RN, and Donna Polenick RN (Pittsburgh). Karen Lahive, MD, coordinated the study at Harvard Community Health Plan, Boston and Elmer Holzinger, MD, coordinated the study at St. Francis Medical Center, Pittsburgh.