Risk Factors for Community-Acquired Pneumonia among Persons Infected with Human Immunodeficiency Virus
- Thomas R. Navin1,
- David Rimland4,5,
- Jeffrey L. Lennox5,
- John Jernigan5,
- Marty Cetron2,a,
- Allen Hightower1,
- Jacqueline M. Roberts1 and
- Jonathan E. Kaplan3
- 1Division of Parasitic Diseases, US Public Health Service, US Department of Health and Human Services, Atlanta, Georgia
- 2Division of Bacterial and Mycotic Diseases, US Public Health Service, US Department of Health and Human Services, Atlanta, Georgia
- 3AIDS, STD, and TB Laboratory Research, National Center for Infectious Diseases, and Division of HIVIAIDS Prevention Surveillance and Epidemiology, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, US Public Health Service, US Department of Health and Human Services, Atlanta, Georgia
- 4Veterans Affairs Medical Center and Research Center on AIDS and HIV Infection, Atlanta, Georgia
- 5Division of Infectious Diseases, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia
- Reprints or correspondence: Dr. Thomas R. Navin, Mailstop F-22, Division of Parasitic Diseases, Centers for Disease Control and Prevention, 4770 Buford Highway NE, Atlanta, GA 30341 (trnl{at}cdc.gov).
Abstract
Two hundred eleven adults with human immunodeficiency virus (HIV) infection hospitalized for community-acquired pneumonia, including Pneumocystis carinii pneumonia (PCP; patients), and 192 matched HIV-infected hospitalized patients without pneumonia (controls) were interviewed to determine risk factors for pneumonia. Multivariate logistic regression showed that patients were less likely than controls to have used trimethoprim-sulfamethox-azole (TMP-SMZ) prophylaxis (odds ratio [OR], 0.22; 95% confidence interval [CI], 0.12–0.41) and more likely to have been hospitalized previously with pneumonia (OR, 6.25; CI, 3.40–11.5). Patients were also more likely than controls to have gardened (OR, 2.24; CI, 1.00–5.02) and to have camped or hiked (OR, 4.95; CI, 1.31–18.7), but stratified analysis by etiologic agent showed this association only for PCP. These findings reconfirm the efficacy of TMP-SMZ in preventing community-acquired pneumonia. In addition, hospitalization for pneumonia might represent a missed opportunity to encourage HIV-infected patients to enter into regular medical care and to adhere to prescribed antiretroviral and prophylaxis medications.
Pneumonia is the leading cause of death among persons infected with human immunodeficiency virus (HIV) in the United States and accounted for over 9000 deaths in 1992 [1]. Although several studies have identified risk factors for pneumonia, most have focused on pneumonias of specific etiologies, such as those caused by bacteria [2–6] or by Pneumocystis carinii [7–14]. Few studies have examined risk factors for pneumonia in general [15–17], and to our knowledge no studies have reported risk factors associated with pneumonias of unknown cause, a group of infections that in 1992 accounted for 17.7% of HIV-related deaths in the United States. [1].
The Pulmonary Opportunistic Infections Study was conducted to determine the etiologic agents of and risk factors for community-acquired pneumonia (including P. carinii pneumonia [PCP]) in HIV-infected persons in Atlanta, Georgia. The diagnostic and clinical aspects of patients in this study will be published elsewhere. For this report, we interviewed patients with pneumonia, as well as HIV-infected hospitalized patients without pneumonia, to determine risk factors for community-acquired pneumonia in general. We also stratified our analysis by etiologic agent of pneumonia to see whether particular risk factors were associated with one cause more than the others.
Methods
Patient population
Between 1 November 1994 and 15 June 1996, adults (age ⩾18 years) who were known by their medical records to be infected with HIV and who were admitted for possible pneumonia to either Grady Memorial Hospital or the Veterans Affairs Medical Center Hospital (VA) in Atlanta were eligible for enrollment. For this study, a case of community-acquired pneumonia was defined as a patient admitted for an acute onset of a presumed infectious pneumonia who, on subsequent review, was found to have an acute infiltrate on chest radiograph. All eligible patients admitted to the VA and the first eligible patient admitted to Grady Hospital each day (Monday through Thursday) were asked to participate.
Blood cultures, sputum or induced sputum, and bronchoalveolar lavage (BAL), when clinically indicated, were processed at participating hospitals for bacterial and mycobacterial organisms and for P. carinii. Diagnostic studies performed at the Centers for Disease Control and Prevention included culture of nasopharyngeal and throat swabs for viruses, Mycoplasma, and Chlamydia; examination of induced sputum and BAL for P. carinii; analysis of urine specimens for Legionella and Histoplasma antigen; and evaluation of acute and convalescent sera for Legionella, Mycoplasma, Chlamydia, and influenza.
For each patient, a control subject was sought from HIV-infected patients hospitalized at the same hospital for a reason other than pneumonia and who was matched by age (± 5 years) and CD4+ T lymphocyte count (matching strata were 0–99, 100–199, 200–499, and 500+).
Patients and controls provided informed consent and were interviewed by study personnel by use of a standardized questionnaire. After a case patient was discharged from the hospital and all diagnostic specimens had been analyzed, a panel of 3 board-certified infectious disease specialists (D.R., J.L.L., and J.J.) reviewed all radiographic, clinical, and laboratory results to determine the most likely cause of pneumonia for each patient.
Analysis
We could not match some patients with controls. To retain as many enrolled patients as possible in the data analysis, patients and controls were pooled into a single patient/control set with 16 strata (the 4 CD4 strata by the 4 age strata [20–29, 30–39, 40–49, and 50+ years] [18]). Subsequent analyses used thesepooled strata for matching patients and controls. Matched univariate odds ratios (MORs) with 95% confidence intervals (CIs) were calculated, and all significant variables were included in subsequent multivariate analyses. Multivariate odds ratios were calculated by using conditional logistic regression matched on age and CD4 count. Multivariate analyses were done for all cases, and separately for cases of laboratory-confirmed PCP, cases of bacterial pneumonia (those with bacteria cultured from blood or, in the absence of positive blood cultures, bacteria cultured from sputum or bronchoal-veolar lavage fluid and with a compatible Gram's stain, chest radiograph, and clinical course), and cases of unknown cause. The pneumonia was considered to have no known cause if no respiratory pathogen was isolated from blood or respiratory specimens and the patient did not have confirmed or probable PCP. We considered a patient to have probable PCP if bronchoalveolar lavage and/or induced sputum was negative or not done but the patient had a compatible radiograph and clinical illness, had no other cause determined, and improved on anti-Pneumocystis medication. Not included in this specific analysis were 69 patients that did not fit into any of these categories (31 with probable PCP, 11 with Mycobacterium tuberculosis, 4 with fungi, and 23 with mixed infections).
Potential effect modifiers (e.g., interactive variables whose effects were not constant across strata), such as socioeconomic status (yearly income<$10,000, not completing high school, staying in a homeless shelter), being jailed, and HIV-risk group were also evaluated by use of conditional logistic regression. All analyses were done with SAS statistical software [19]. Statistical significance was set at α = .05.
Results
A total of 288 patients admitted with the diagnosis of pneumonia were interviewed for the study. Postadmission review of the medical records by 3 infectious disease specialists revealed that 53 patients did not have pulmonary infiltrates on chest radiographs and 5 had noninfectious causes for their infiltrates (4 had Kaposi's sarcoma of the lung and 1 had congestive heart failure). Of the 230 patients left who had community-acquired pneumonia, 18 had no CD4+ T lymphocyte counts obtained at or within 6 months of admission, and 1 fell into a matching stratum with no corresponding control. Thus, we had 211 cases of community-acquired pneumonia available for analysis. The 19 excluded cases were similar to the 211 included cases for age, sex, race, and cause of pneumonia. The 19 excluded cases were, however, somewhat less likely to be men who had sexual intercourse with men (relative risk, 0.18; CI, 0.03–1.22). We identified and interviewed 192 controls, and our subsequent risk factor analysis will be done on the basis of these 211 patients and 192 controls.
In all, we analyzed 79 variables as potential risk factors for community-acquired pneumonia. Patients were similar to controls for several demographic variables (table 1). The mean CD4 count for patients (115.1 cells/μL) was higher than that for controls (75.8 cells/μL), but this difference was controlled for in our matching process, which led to an MOR of 1.00. Patients and controls were mostly black men with annual household incomes <$10,000. About one-third of patients and controls were men who reported their HIV risk group as having sexual intercourse with men (includes bisexual men and male injectiondrug users who have sexual intercourse with men), and about one-quarter of patients and controls reported injection-drug use as their risk group (excludes men who have sexual intercourse with men).
Comparison of demographic variables for patients and controls: results of matched univariate analysis.
Sixteen variables proved to be significant risk factors by univariate analysis (table 2). Patients were significantly more likely than controls to report having had close contact in the 6 months before admission with someone with pneumonia (close contact was defined as having lived with or spent more than 3 h/week in the same room as that person). Two outdoor activities—gardening and camping or hiking—during the 6 months before admission were also significantly associated with pneumonia.
Risk factors for community-acquired pneumonia: results of matched univariate analysis.
We asked patients whether they recalled breathing a number of potential chemical irritants, such as insect spray, gasoline, or paint fumes (whether a patient smoked cigarettes or crack cocaine was considered separately). Having breathed one or more of these potential chemical irritants was significantly associated with pneumonia, and 32 (34.4%) of the 93 patients who recalled breathing these chemicals reported becoming short of breath at the time. We did not ask the patients whether breathing these irritants led directly to their hospitalization but to determine whether this exposure might have caused prolonged breathing problems, we looked at whether exposed patients were more likely at the time of admission to report prolonged (>7 days) prodromal cough. Although almost half (44.1%) of the 93 patients who reported breathing these chemicals also reported prolonged cough at the time of admission, more than half (52.5%) of the 118 patients who did not report breathing these chemicals also had a prolonged cough.
Over half (60.0%) of the patients had been hospitalized in the previous 6 months for a separate episode of pneumonia. This high rate of contact with the medical establishment, however, did not guarantee that patients were receiving adequate PCP prophylaxis. Of the 117 patients who had been hospitalized in the previous 6 months, 93 had CD4 counts <200 cells/μL, and were, therefore, candidates for PCP prophylaxis (the number who had CD4 counts <200 cells/μL at the time of the previous admission was not known). Of these 93, only 47 (50.5%) were taking any prophylaxis, and only 37 (39.8%) reported that they adhered regularly to their prescribed prophylactic regimen. Patients reported significantly lower rates of use than did controls for a number of outpatient medications, as well as for influenza vaccinations.
Not listed in table 2 are several important negative findings. Reported rates of pneumococcal vaccination were comparably low for both patients and controls; 34 (16.1%) of patients and 41 (21.4%) of controls reported having received a pneumococcal vaccination in the previous 5 years (MOR, 0.68; CI, 0.41–1.15). None of the following variables was significantly associated with pneumonia: being born or having traveled outside the United States or owning dogs, cats, or birds.
Table 3 lists the results of our multivariate matched logistic regression analysis for the 9 significant variables found to be independently associated with pneumonia. The use of prophylactic fluconazole or TMP-SMZ, as well as the use of any antiretroviral agent, were associated with a decreased risk of pneumonia. The protective effect seen for influenza vaccine in the univariate analysis was not apparent in the multivariate analysis. Having been hospitalized for pneumonia in the previous 6 months or having close contact with someone with pneumonia were both significant risk factors. Both of the outdoor activities we asked about (camping or hiking and gardening) were found to be independent risk factors, and both breathing chemical irritants and smoking crack cocaine were associated with pneumonia. Smoking cigarettes, on the other hand, was not a significant risk factor. No evidence of effect modification was seen between any of the risk factors listed in table 3 and patients' HIV-risk group or variables associated with socioec-onomic status.
Multivariate matched logistic regression of all patients and their matched controls.
To determine whether the risk factors listed in table 3 were associated with cases of pneumonia in general or only with cases of PCP, bacterial pneumonia, or pneumonias of unknown cause, we repeated our multivariate matched logistic regression analysis for each of these 3 groups, analyzing the 9 variables shown to be significant in our analysis of all cases. For the 36 cases of PCP, we could match 178 of the 192 controls; for the 51 cases of bacterial pneumonia, we could match 188 of the 192 controls; and for the 55 cases of unknown cause, there were 186 matched controls.
Seven of the 9 variables that were significant risk factors for all patients were also significant risk factors for cases of PCP; the 2 exceptions were having had close contact with someone with pneumonia and having smoked crack cocaine (table 4). We also found no statistical association between being a PCP patient and having previous close contact with someone with PCP (MOR, 0.96; CI, 0.08–15.0), and only 3 of the 36 PCP patients reported such contact (data not shown in table 4). For cases of bacterial pneumonia, only 4 risk factors remained significant: using fluconazole or TMP-SMZ prophylaxis, having been hospitalized previously for pneumonia, and having breathed chemical irritants. For cases of pneumonia of unknown cause, 4 risk factors were also significant: using fluconazole prophylaxis, having been hospitalized previously for pneumonia, having close contact with someone with pneumonia, and having smoked crack cocaine.
Multivariate matched logistic regression stratified by cause of pneumonia.
The analyses shown in table 4 compare the various risk factors between patients by the causes of their pneumonias. To see whether a specific risk factor was more likely to be associated with 1 of the 3 etiologies considered, we used matched logistic regression analysis to compare the 36 cases of PCP with the 106 cases of other etiologies (51 cases of bacterial pneumonia and 55 cases of unknown cause), the 51 cases of bacterial pneumonia with the 91 cases of other causes, and the 55 cases of unknown cause with the 87 cases of other causes. Risk factors that were significantly different for 1 cause than the other 2 are footnoted in table 4.
Because the degree of immunodeficiency is considered to be strongly associated with the development of opportunistic infections, we repeated our analyses for the 47 patients and 24 controls with CD4 counts ⩾200 cells/μL. Of the 47 patients, 1 had confirmed PCP, 6 had probable PCP, 17 had bacterial pneumonia (12 with Streptococcus pneumoniae), 5 had tuberculosis, 6 had mixed infections (2 with S. pneumoniae), and 12 had pneumonias of unknown cause. Two variables were associated with significant risk by matched multivariate analysis. Twenty-three (50.0%) of the 46 patients (1 patient did not answer this question) reported having been hospitalized for pneumonia in the previous 6 months, compared with 3 (12.5%) of the 24 controls (MOR, 4.98; CI, 1.15–21.5). Twenty-one (44.7%) of 47 patients reported breathing chemical irritants in the previous month, compared with 3 (12.5%) of the 24 controls (MOR, 6.18; CI, 1.22–31.4). Only 11 (23.4%) of the 47 patients had received pneumococcal vaccination in the previous 5 years, and only 5 (20.8%) of the 24 controls had been vaccinated (MOR, 0.51; CI, 0.11–2.33).
Discussion
The results of this study indicate that TMP-SMZ is a very effective prophylactic agent for preventing community-acquired pneumonia. That TMP-SMZ protects against PCP is well known [20–32], but our results corroborate those of other recent studies showing that TMP-SMZ also is associated with a lower risk of bacterial pneumonia [2, 24]. Comparison of data from the Pulmonary Complications of HIV Infection Study, however, suggests that the effect of TMP-SMZ against bacterial pneumonia is more to delay its onset as HIV disease progresses, rather than to prevent it altogether [2, 33, 34]. Nevertheless, TMP-SMZ's positive effect, as well as findings from other studies that TMP-SMZ might be effective in preventing encephalitis caused by reactivation of Toxoplasma gondii infection [24, 29, 31, 35], points to the remarkable benefit of this drug for persons with HIV infection.
Some authors have suggested that, because TMP-SMZ is so effective in preventing PCP, current recommendations for prophylaxis with TMP-SMZ should be liberalized to include patients with CD4 counts above 200 cells/μL [36, 37]. Our results suggest another reason in favor of liberalizing these recommendations: that it would protect against bacterial pneumonias, which are more common than PCP at higher CD4 counts. The potential benefit of prescribing the drug to more patients, however, must be weighed against potential drawbacks, such as increased cost, the possibility of adverse drug reactions, and that increased drug use might provide pressure toward the development of antimicrobial resistance. If a subset of persons could be identified who were at particularly high risk for developing pneumonia despite their high CD4 counts, recommendations targeting this group might overcome many of the possible drawbacks of more general recommendations.
Our results might identify such a high-risk subset. Half (51.1%) of the 47 patients with community-acquired pneumonia in our study with CD4 counts ⩾200 cells/μL reported a previous hospitalization for pneumonia. The risk associated with previous hospitalization for pneumonia was also apparent for each of the 3 groups of pneumonia that we analyzed. We had little detailed information about the antecedent hospitalizations; therefore, we could not determine whether this effect was due to a predisposition to pneumonia in these patients, possible pulmonary damage caused by the antecedent infection, failure to completely clear the organism followed by a recurrence caused by the same organism, or possible nosocomial transmission at the time of the antecedent hospitalization.
Whatever the pathophysiologic explanation of the risk posed by previous hospitalization, however, our findings point to an opportunity to easily reach this high-risk population. We believe that patient education programs that emphasize the benefits of regular medical care and adherence to prescribed medications should be developed and targeted at HIV-infected patients who are hospitalized for pneumonia. For hospitalized patients with CD4 counts <200 cells/μL, these programs should put special emphasis on the importance of regular adherence to PCP prophylaxis. Our results that 60% of our previously hospitalized patients with CD4 counts <200 cells/μL reported not receiving or not adhering to prescribed PCP prophylaxis provide empiric data that effective education is needed, although we do not know what percentage of these patients had CD4 counts <200 cells/μL at the time of their previous hospitalization, or even whether they were HIV positive at that time. For hospitalized patients with CD4 counts ⩾200 cells/μL, such programs should evaluate the cost-benefit ratio of recommending prophylaxis with TMP-SMZ. Because no patients with CD4 counts above 200 cells/μL in our study received TMP-SMZ, we could not calculate the percentage of subsequent pneumonias that would be prevented by TMP-SMZ. Other studies have indicated, however, that TMP-SMZ is more effective at preventing PCP at higher, as compared with lower, CD4 counts, so it may be very effective in preventing PCP, as well as other pneumonias, in patients with CD4 counts above 200 cells/μL.
In our study, prophylaxis with the anti-fungal agent flucon-azole was also associated with a protective effect for pneumonia in general, as well as for PCP and pneumonias caused by bacteria or of unknown causes. Recent data from genetic studies indicate that P. carinii is more closely related to fungi than to protozoa [38, 39], as was once thought, but in at least one study, fluconazole had no effect against P. carinii [40].
Patients reported inhaling a variety of chemical irritants more often than did controls. It is possible that these irritants predisposed patients to pneumonia by damaging the respiratory epithelium. Smoking crack cocaine was also a risk factor in our study, perhaps for the same reason. Because this was a case-control study and our data were based on patient interviews, however, we could not corroborate what patients reported with independent observations of what they actually had inhaled. It is possible, therefore, that patients hospitalized for pneumonia were more likely to recall inhaling a chemical irritant, such as gasoline fumes, than were controls, who were hospitalized for another reason.
There has been recent interest in whether infection with P. carinii can be transmitted from one person to another [41]. Current epidemiologic theory holds that most PCP in adults represents reactivation of earlier infection, and current guidelines for preventing opportunistic infections do not recommend respiratory isolation for cases of PCP [42]. Evidence that person-to-person transmission is common would, therefore, lead to reexamination of that theory and those guidelines. Although our study was not designed to address this issue directly, our data do not provide evidence in support of person-to-person transmission. Not only did we fail to find a statistical association between being a PCP patient and having previous close contact with someone with PCP, but also only 3 of our 36 PCP patients reported such contact. In contrast, PCP patients did report being hospitalized in the previous 6 months more often than did their controls, and without knowing it they could have been exposed to patients with PCP during a previous hospi-talization. Previous hospitalization, however, was an equally strong risk factor for cases of bacterial pneumonia and cases of unknown cause.
Our finding that histories of recent gardening and hiking or camping were strongly associated with PCP but not with bacterial pneumonia or pneumonia of unknown cause might indicate that P. carinii can be acquired from a recent environmental exposure. Additional evidence that adults can develop new infections with P. carinii has come from comparing molecular types of P. carinii isolates in patients who are rehospi-talized for second episodes of PCP [43]. In approximately 30% of such rehospitalizations, the P. carinii isolate from the second episode had a genetic sequence distinct from the P. carinii isolate obtained at the first episode, suggesting not only that de novo infection might occur but also that previous infections do not necessarily provide protection against subsequent infections.
Our study had several limitations. It included mostly poor black men who were admitted to nonprivate hospitals, and our results may not apply to other patient populations. Specifically, we did not include in this study patients with community-acquired pneumonia who were treated as outpatients. Although the patients in general seemed knowledgeable about their medications and activities during our interviews, we did not independently verify patients' actual medication use or whether or not they actually had been vaccinated or had participated in the activities they reported. Finally, we could not match some of our patients who had CD4 counts >200 cells/μL with appropriate controls, and this limited our ability to identify risk factors in this important patient population.
In summary, our findings point to the remarkable efficacy of TMP-SMZ prophylaxis in preventing community-acquired pneumonia. This effect is so strong that we believe the benefit of extending current guidelines for TMP-SMZ prophylaxis to HIV-infected patients with CD4 counts >200 cells/μL should be studied in relationship to the costs of such a recommendation. In addition, our study suggests that hospitalization for pneumonia might represent a missed opportunity for health education. We believe that special health education programs should be tested that target HIV-infected patients while they are hospitalized for pneumonia, to encourage them to receive appropriate vaccinations, take appropriate prophylaxis and antiretroviral therapy, and enter into or remain in regular medical care. Finally, all of these recommendations presuppose that a patient knows his or her HIV status, so that high-risk patients who are hospitalized for pneumonia should also be counseled to be tested for HIV infection.
Acknowledgments
We thank Dan Colley, Suzanne Binder, Anne Schuchat, and Martha Rogers for critical review of the manuscript.
Footnotes
-
Informed consent was obtained from all patients. Human experimentation guidelines of the US Department of Health and Human Services and those of participating institutions were followed in the conduct of this research.
-
Financial support: the Centers for Disease Control and Prevention.
- Received June 1, 1999.
- Revision received September 20, 1999.
- © 2000 by the Infectious Diseases Society of America
