INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Despite the success of Pneumocystic carinii pneumonia (PCP) prophylaxis resulting in a declining incidence of PCP and the improvements in antiretroviral therapy, PCP was the most common opportunistic infection for persons with human immunodeficiency virus (HIV) infection in 1997 and occurs during the course of acquired immunodeficiency syndrome (AIDS) for 53% of persons who die with AIDS (1). PCP is also one of the leading causes of pulmonary disease and death for persons with AIDS (2) and remains the most common cause of respiratory failure and intensive care unit (ICU) admission (5, 6). Physicians caring for patients with PCP and impending respiratory failure are confronted with the questions of whether to recommend intensive care and mechanical ventilation. Although previous research has identified several factors associated with survival from PCP (7), there is no prognostic system with sufficient accuracy to identify individual patients for whom mechanical ventilation is futile. Nonetheless, it is important that physicians caring for these patients be aware of outcome data to assist patients and their families in this decision making.

Outcomes of mechanical ventilation for respiratory failure from PCP have changed dramatically since the onset of the AIDS epidemic (12). Early in the epidemic, prior to 1988, multiple reports suggested that the hospital survival was less than 20% (13). Then from 1988 to 1992, multiple studies showed a much higher hospital survival of 40-54%, often attributed to adjunctive use of corticosteroids (16). In yet a third phase from 1992 to 1995, survival after respiratory failure was lower than the prior time period, with hospital survival in the 18-24% range, primarily attributed to increasing severity of PCP among patients receiving mechanical ventilation (6, 10, 21). Another complicating factor in understanding the outcomes of respiratory failure due to PCP is the significant geographic variations in the use of mechanical ventilation that affect hospital survival for ventilated patients (22). Although these data on prior outcomes are helpful to clinicians, they are likely outdated. In this study, we report the utilization and outcomes of HIV-related PCP and respiratory failure in 1995-1997 from seven geographic areas of the United States. This time period represents the beginning of the era of highly active antiretroviral therapy that has resulted in dramatic improvements in morbidity and mortality from HIV infection (23). To our knowledge, these data represent the first report on clinical outcomes for PCP with respiratory failure in the more recent era.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sampling of Patients and Hospitals

Medical records were abstracted for HIV-infected patients with confirmed or presumptive PCP who received a portion of their medical care in a study institution between January 1, 1995 and December 31, 1997. Study institutions included 71 public and private hospitals in seven geographic areas of the countryNew York, Los Angeles, Miami, Chicago, Seattle, Tennessee, and Arizona. For 1996-1997, New York had the largest number of reported AIDS cases in the country, with Los Angeles second, Miami fifth, and Chicago eighth (24).

We have previously described our sampling methodology that employed two hierarchical levels: hospitals within cities and patients within hospitals (25). In brief, within each geographic region, hospitals were randomly selected for participation in the study. A random sample of patients in each hospital was then selected such that the proportion of patients from the hospital relative to the overall number of charts reviewed in a given city was roughly proportional to the square root of the individual hospital caseload divided by the total hospital caseload for the given city. This method of sampling patients was used to ensure enough patients in low experience hospitals to allow comparisons of high and low experience hospitals. Cases were also stratified by whether patients were discharged alive or dead in order to approximate the actual in-hospital mortality for each hospital.

Data Collection

Medical information system analysts at study hospitals identified all hospital discharges that included ICD-9 codes for PCP (136.3) and HIV-related disease (042-044). Retrospective chart reviews were performed by trained registered nurses who were experienced with patients with AIDS and utilization review. Patients were included in the study if they received medical care at a study hospital during the study period, had microbiological confirmation of PCP or physician notes indicating that PCP rather than other causes accounted for their pulmonary process, and if their age was at least 18 yr. The criteria for presumptive PCP were that the physician's notes indicated that PCP was thought to be the cause of the pneumonia and the patient was treated for at least 2 wk (or until death) with an antibiotic and dose appropriate for PCP. Patients were excluded if they received medical care for the current episode of PCP at another hospital or were admitted for other conditions or diagnostic bronchoscopy only. If more than one episode of PCP for an individual patient was identified, only the first episode identified was included. Patients were not excluded for having an episode of PCP prior to the study period. Of the 1,888 medical records reviewed for potential inclusion, 228 were excluded. Of those records excluded, 96% were excluded because the diagnosis of PCP did not meet study criteria, and 4% because patients were transferred to another hospital. The remaining 1,660 records with confirmed (n = 834) or presumptive (n = 826) HIV-related PCP were used for our analyses.

Data abstracted included patient sociodemographics; comorbid conditions; cigarette, alcohol, and drug use history; preadmission use of antiretroviral and prophylactic medications; T-cell count and HIV viral load titer; initial vital signs, arterial blood gas, and laboratory data; treatment medications received; principal and secondary diagnostic and procedure codes; length of stay; and discharge status. Patients were classified as having "wasting" if on Day 1 or 2 of the admission a physician noted that the patient exhibited wasting, cachexia, a greater than 20% weight loss, or a greater than 20 pound weight loss. A preadmission staging system for predicting inpatient mortality was developed as part of a prior report (27). This system was developed using hierarchically optimal classification tree analysis producing an ordered five-category staging system based on three predictors: wasting, alveolar-arterial oxygen difference, and serum albumin level. To assess the effect of hospital experience with HIV-related PCP, hospitals were classified as high experience if there were more than 35 episodes of PCP during the study period.

Data quality was maintained by overreading abstraction forms by two physicians trained in quality assurance for the project. Less than 1% of entries were categorized as possibly inaccurate by the physician overreaders.

Statistical Analysis

The patient was the unit of analysis for this study. We assessed the association between patient characteristics and the use of mechanical ventilation or survival to hospital discharge using the chi-square statistic. Multivariate logistic regression was used to confirm the findings from the bivariate analyses and to determine if there was an independent association between each predictor variable (patient characteristics) and the outcome variables (use of mechanical ventilation or hospital survival). Dependent variables were placed in the model if they had a significant association with the outcome variable in the bivariate analyses at a p value of < 0.05. To control for severity of illness on hospital admission, we used a model developed from these data and described previously (27). For all analyses, we selected a p value of < 0.05 to signify significance.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Of the 1,660 patients hospitalized for HIV-related PCP in these seven geographic areas, 237 (14%) were admitted to an ICU during their hospital stay and 155 (9%) received mechanical ventilation for respiratory failure. Of the 155 patients who received mechanical ventilation for respiratory failure, 38% survived to hospital discharge (95% CI 30.4, 45.6). Of these 155 patients, 99% received adjunctive corticosteroid therapy. The majority of patients (63%) were initially treated with trimethoprim-sulfamethoxazole for the episode of PCP and 16% were initially treated with pentamidine. For those patients who required mechanical ventilation, the mean initial alveolar-arterial oxygen difference was 69.7 mm Hg (SD 24.5) and CD4 cell count 83.7 cells/µl (SD 159.8) with a median number of days until the initiation of steroids of 3 d (range 0- 9). None of these variables was significantly associated with hospital survival (data not shown).

As shown in Table 1, there were several patient characteristics that were associated with the use of mechanical ventilation. African-Americans were significantly more likely to receive mechanical ventilation than Hispanic patients or non-Hispanic whites (p  0.007). Patients with a prior AIDS-defining illness were less likely to receive mechanical ventilation than those without a prior AIDS-defining illness (p  0.01). Furthermore, patients receiving either antiretroviral therapy or PCP prophylaxis prior to admission to the hospital were less likely to receive mechanical ventilation (p  0.02 and 0.01, respectively). There was no association between sex or HIV risk behavior and the use of mechanical ventilation.

We also examined the association between hospital and geographic characteristics and the use of mechanical ventilation (Table 2). Geographic location was associated with significant variation in the use of mechanical ventilation for hospitalized patients with PCP. Although there was a trend toward hospitals with more PCP experience using mechanical ventilation less frequently, this trend was not significant (p = 0.09). Patients who received mechanical ventilation were significantly more likely to have a confirmed diagnosis of PCP.

To control for confounding among variables associated with the use of mechanical ventilation, we used logistic regression to assess the independent effect of these characteristics. In addition, we used a preadmission staging system (described in METHODS) to control for severity of illness for these patients. The results of the logistic regression (Table 3) show that whereas severity of PCP at hospital admission, ethnicity, confirmed diagnosis of PCP, and geographic location were independently associated with the use of mechanical ventilation, the other variables (prior AIDS-defining illness, antiretroviral use, and PCP prophylaxis) were not.

Of the patient and hospital characteristics described above, only severity of illness at hospital admission, use of PCP prophylaxis prior to hospital admission, and presumptive diagnosis of PCP were significantly associated with survival for those patients who received mechanical ventilation for respiratory failure. We used logistic regression to control for severity of illness and found that prior use of PCP prophylaxis remained significantly associated with decreased hospital survival (OR 2.32; 95% CI 1.14, 4.72).

We also evaluated the association between timing of treatment and outcomes. First, patients who developed respiratory failure after 5 or more days of treatment for PCP with antibiotics and corticosteroids were slightly less likely to survive to hospital discharge than those who developed respiratory failure less than 5 d after initiation of therapy, but this difference was not significant (see Table 1; p = 0.5). In addition, the duration of mechanical ventilation was significantly associated with hospital survival. Of the 99 patients who received mechanical ventilation for less than 1 wk, 42 (42%) survived to hospital discharge compared with 11 of the 33 (33%) who received mechanical ventilation for 1-2 wk and 3 of the 17 (17%) patients who received mechanical ventilation for greater than 2 wk (p = 0.05; chi square for trend).

There were 73 patients who died in the hospital without receiving ICU care and with a do not resuscitate (DNR) order in place (4% of all patients and 39% of in-hospital deaths). These patients likely represent patients for whom ICU care was withheld. There was no significant association between withholding ICU care and the demographic variables in Tables 1 and 2 (data not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Treatment of HIV infection has undergone marked changes in the past 5 yr, with advances in antiretrovial and prophylactic therapy and associated improvements in survival. With these advances, the incidence of PCP has significantly declined, but it is still the most common opportunistic infections for patients with AIDS (1) and one of the most common causes of respiratory failure in these patients (5, 6). We found that approximately 10% of patients with HIV-related PCP received mechanical ventilation for PCP and that 38% of those receiving mechanical ventilation survived to hospital discharge. This survival proportion is similar to the 44% from a recent single-center study of patients hospitalized between 1991 and 1996 (28). These authors also showed that if patients survived to hospital discharge, subsequent survival time was the same for patients with HIV-related PCP whether or not they had developed respiratory failure.

Geographic and ethnic variations in the use of mechanical ventilation for patients with HIV-related PCP were identified during 1987 to 1990, immediately after the introduction of azidothymidine (AZT) (22). The current study found similar variations during 1995-1997, in the time period associated with the use of newer antiretroviral therapies. It is important to examine variation in the utilization of mechanical ventilation for HIV-related PCP before interpreting outcomes data because if there is significant variation in the utilization of mechanical ventilation, it may have an important effect on the outcomes. The current data demonstrate that large geographic variations in utilization of mechanical ventilation persist, confirming the importance of reviewing outcomes data from multiple centers. During both time periods, African-Americans were more likely to receive mechanical ventilation after controlling for severity of disease. This may reflect a preference for more life-sustaining therapy in these patients (29, 30). From 1987 to 1990, African-American patients with PCP were less likely to have DNR orders, and when these orders were written, they were more likely to be later in the hospitalization (31). Finally, we found that patients who received mechanical ventilation were significantly more likely to have a confirmed diagnosis of PCP. The likely explanations for this latter finding are increased concern about establishing a definitive diagnosis in patients requiring mechanical ventilation and the fact that bronchoscopy is an easier procedure in patients who are intubated.

It is interesting that just under half of the hospitalized patients and approximately one-quarter of patients who received mechanical ventilation for respiratory failure did not receive a definitive diagnosis of PCP. Although there is controversy about the role of empiric treatment for PCP without obtaining a definitive diagnosis (32, 33), recent expert opinion suggests that among patients with disease severe enough to be hospitalized, obtaining a definitive diagnosis should be a standard of care (33, 34). This represents an important area for physician education.

Among those patients who developed PCP and respiratory failure requiring mechanical ventilation, prior use of PCP prophylaxis was associated with decreased hospital survival. After controlling for severity of illness at hospital admission, this association remained significant. There are a number of potential explanations for the association between PCP prophylaxis and lower survival from PCP with respiratory failure including residual confounding by severity of immunosuppression or severity of PCP or by a history of nonadherence with medication. Nonetheless, the findings raise the question of whether P. carinii can develop resistance to antibiotics, an hypothesis suggested by small studies showing that patients who develop PCP while on prophylaxis have mutations in the drug's target genes (35, 36). Furthermore, a recent cohort study showed decreased survival in patients with these mutations (37). Further research is necessary to confirm and explain this association for patients with acute respiratory failure from PCP.

During the 1995 to 1997 time period, 39% of all patients that died with HIV-related PCP had ICU care withheld. This is a slightly higher proportion than the 30% described from 1987 to 1990 in an era when AZT was the only antiretroviral agent available (22). This increase may reflect secular changes associated with increased withholding and withdrawing intensive care therapies (38) or may reflect changing views of patients and physicians on the outcomes of PCP and respiratory failure (12). Given the rapid changes in antiretroviral therapy and long-term outcomes of HIV infection, it will be important for ICU clinicians to continually assess the attitudes of patients, their families, and their primary care providers on use of life-sustaining therapy.

Our study has important implications for physicians who are considering withdrawing respiratory support for severely ill patients who fail to improve with supportive care. Although two studies from one institution found that patients who developed respiratory failure after 5 or more days of treatment for PCP with antibiotics and steroids had a significantly lower survival than those who developed respiratory failure earlier in treatment (21, 28), our study did not corroborate these findings. Although our numbers were relatively small, they were larger than either of these prior reports. Also, although there was a significant relationship between the duration of mechanical ventilation and hospital survival such that longer time of mechanical ventilation was associated with worse survival, we found that those patients requiring mechanical ventilation for more than 2 wk had a hospital survival of 17%. Two prior studies found that patients who received mechanical ventilation for more than 2 wk never survived to hospital discharge (10, 39). Our findings suggest that survival for those receiving mechanical ventilation longer than 2 wk, although low, is not zero. Physicians who recommend withdrawing mechanical ventilation from severely ill PCP patients after 2 wk of mechanical ventilation on the basis of reported zero survival should reconsider this strategy.

There are several limitations to this study. First, despite studying 1,660 hospitalized patients with PCP, only 155 patients received mechanical ventilation. The sample size may have limited our ability to show important predictors of hospital survival in these patients. In particular, there were only 40 patients in this group who were receiving antiretroviral therapy at the time of hospital admission. Although antiretroviral therapy was not associated with survival, a larger sample size will be necessary to exclude the possibility that combination therapy with the new potent agents does not influence survival from PCP and respiratory failure. Second, although we used the strongest predictors of severity of illness available in these data (27), there may be residual confounding by severity of illness. Finally, only half of the patients in this study had the diagnosis of PCP cytologically confirmed. Although those patients treated presumptively for PCP might have had other diagnoses and even those with PCP may have had a significant coinfection (33, 40), a potential strength of this study is that it describes the outcomes of all hospitalized patients treated by physicians for PCP. Although patients who required mechanical ventilation were more likely to receive a definitive diagnosis, there was no association between confirmation of diagnosis and survival after controlling for severity of illness. In addition, to minimize the risk that our results were confounded by presumptive versus confirmed diagnosis, we controlled for confirmed diagnosis using multivariate analyses.

In the current era of AIDS treatment, hospital survival after HIV-related PCP and respiratory failure requiring mechanical ventilation is approximately 40%, representing a twofold improvement from the years 1992 to 1995. Survival may be lower for patients on PCP prophylaxis prior to hospitalization, but further studies are needed to confirm this finding. Several factors previously shown in single-site studies to be predictive of survival in patients with PCP, duration of therapy prior to development of respiratory failure and duration of mechanical ventilation, were not significantly associated with survival in this multisite study. Physicians caring for patients with HIV-related PCP and impending respiratory failure should be aware of the improvements in short-term outcomes for this disease.