Comprehensive, up-to-date information on HIV/AIDS treatment, prevention, and policy from the University of California San Francisco

HIV-infected patients may require critical care for a number of reasons. Acute respiratory failure accounts for approximately 25-50% of intensive care unit (ICU) admissions in HIV-infected patients.(1-6,7) Respiratory failure usually is secondary to infectious causes, with Pneumocystis jiroveci pneumonia (PCP) and bacterial pneumonia (BP) being the most common etiologies.(1,2,6) Other common indications for ICU admission include central nervous system (CNS) dysfunction and sepsis. Patients also may require ICU admission for gastrointestinal (GI) bleeding, among other GI conditions, and for cardiovascular disease. The spectrum and outcome of critical illness in HIV patients is changing in the setting of effective antiretroviral therapy (ART). Current data suggest that many HIV-infected patients are presenting to the ICU with medical or surgical issues unrelated to their HIV infection.(6)

This chapter reviews critical care of HIV-infected patients, including causes of ICU admission and patient outcomes. Consideration is given to the changing spectrum of disease in the setting of ART and to the potential impact of ART on the critical care of HIV-infected patients. Special emphasis is placed on the etiology and management of respiratory failure, particularly that due to PCP, which carries an especially high mortality risk. Finally, the importance of advance directives and end-of-life care in patients with HIV is discussed.

Although the rate of hospitalizations for HIV-infected patients has decreased significantly in the setting of effective ART, the rate of ICU admissions generally has remained constant (7,8) or has increased (9,10) or decreased only slightly (6) at individual centers. Approximately 5-12% of hospital admissions for HIV-infected patients in U.S. and European studies involve ICU care.(8,9,11) Possible reasons why ICU admissions have remained relatively constant include the fact that, in many studies, a large proportion of HIV-infected patients continue to be admitted to the ICU without prior known HIV infection (range 28-40%).(7,9) In addition, approximately 50% of patients are not on effective ART at the time of admission.(7,9,11) Furthermore, as overall survival has improved in HIV-infected patients on effective ART, the number of persons living with HIV has increased. Given this improved survival, providers may be more likely to admit patients to the ICU and pursue aggressive life-support measures.(8,9)

Studies of critically ill HIV-infected patients indicate that the spectrum of diseases requiring ICU admission is changing in the setting of ART. Early in the epidemic, the majority of patients were admitted with an AIDS-associated condition, most often PCP. Increasingly, patients with HIV infection were admitted with a non-AIDS-associated condition. For example, in a study conducted at 4 hospitals in England, only 14% of ICU admissions between 1993 and 1997 were due to non-AIDS-associated diagnoses.(3) In contrast, a review of all HIV-infected patients admitted to the ICU at San Francisco General Hospital in the early years of effective ART use (between 1996 and 1999), found that the majority of patients (63%) were admitted with non-AIDS-related conditions.(6) Additional studies have compared the spectrum of ICU admission before and after effective ART at single institutions. In a study from France, the proportion of admissions for non-AIDS-related conditions increased substantially from 42% to 63% when comparing admissions between 1995-1996 and 1997-1999.(7) Likewise, in a study from Beth Israel Medical Center in New York City, ICU admissions for non-HIV-related disease increased substantially from 12% in 1991-1992 to 67% of all admissions in 2001.(10)

In terms of the specific diagnoses requiring ICU admission, acute respiratory failure is the most common, accounting for approximately 25-50% of ICU admissions (Table 1).(1,2,4,5,7,10) P jiroveci has been identified as the responsible pathogen in up to 25-50% of these patients,(2,5,6,12) although more recent studies have demonstrated PCP in only 3-12% of patients.(7,10,11) Bacterial pneumonia also is a frequent cause of acute respiratory failure, and in some studies is as common (6) or a more common cause (10) of respiratory failure than PCP in the era of effective ART. In addition, sepsis is an increasingly common diagnosis, in 1 study increasing from 3% to 23% of all admissions for HIV-infected patients in the era of ART.(7) Other frequently reported causes of ICU admission include CNS dysfunction (11-27%), GI bleeding (6-15%), and cardiovascular disease (8-13%).(1,2,5,6,7,12) Reasons for ICU admission unrelated to immunodeficiency include trauma, routine postoperative care, noninfectious pulmonary diseases including asthma and chronic obstructive pulmonary disease, renal failure, metabolic disturbances, and drug overdose.(2,5,6) Given the frequent coinfection with hepatitis C among patients with HIV, liver disease is increasing as a cause of death,(13,14) and complications related to cirrhosis often require ICU admission. In addition, solid organ transplantation (liver, kidney) currently is being studied in HIV-infected patients; thus, these patients also may be encountered in the ICU setting.

The reported in-hospital mortality for AIDS patients admitted to the ICU is approximately 25-40%, with a median ICU length of stay of 5-11 days.(2,4-6,7,9) The overall mortality rate for HIV-infected patients requiring critical care has improved during the course of the AIDS epidemic, particularly from the early 1980s when the mortality rate was approximately 70%. Some studies have suggested a further decline in ICU mortality after the introduction of effective ART, although this has not been demonstrated in all reports. In analyses of ICU admissions at a hospital in Paris, France, mortality while in the ICU was not changed when comparing admissions prior to effective ART with admissions after the introduction of effective ART.(7,9) However, in studies from San Francisco General Hospital, ICU mortality decreased from 37% in 1992-1995 to 29% in 1996-1999.(15) Likewise, studies comparing the mortality rate between 1991-1992 and 2001 from Beth Israel Medical Center in New York City demonstrated a decrease in mortality from 51% to 29%.(10)

Mortality in the ICU is clearly related to the reason for ICU admission. The highest mortality rates for HIV patients requiring ICU admission are associated with sepsis and respiratory failure, especially if due to PCP; in these patients, mortality rates remain as high as 50-68%.(2,5,11,16,17) For AIDS patients admitted to the ICU for other HIV-related reasons, the reported mortality generally is lower. For example, the reported mortality rate for CNS dysfunction is 20-48%,(11) whereas the rate for GI disease is approximately 30-35%.(2,4-6) However, patients admitted with non-HIV-related conditions may have better outcomes: in one study, patients admitted with a non-AIDS-associated diagnosis were 3.7 times more likely to survive than patients admitted with an AIDS-associated condition.(6) In a New York study, ICU admission with an HIV-related illness was independently associated with increased mortality (odds ratio [OR]: 4.2; 95% confidence interval [CI]: 2.0-9.0; p < .001).(11)

Mortality during hospitalization also is related to the severity of the acute illness. Predictors of increased hospital mortality include need for mechanical ventilation and disease severity (as assessed by scoring systems such as the Simplified Acute Physiology Score I [SAPS I] and the Acute Physiology and Chronic Health Evaluation II [APACHE II] score).(1,5,6,7,11) Patients who have a decreased serum albumin value or a history of weight loss also may have a higher mortality risk.(1,5,6) The CD4 T-cell count and the HIV RNA level generally have not been predictive of short-term mortality during ICU stay.(4-6) However long-term mortality after ICU admission is related to the underlying severity of HIV disease.(1,5,12) Compared with the era prior to effective ART, long-term survival following ICU discharge is improved in the era of effective ART.(7,9)

The full impact of ART on outcome of HIV patients in the ICU is unclear, as prospective, randomized trials assessing the initiation of ART on outcome in critically ill patients have not been completed. Two retrospective studies conducted at San Francisco General Hospital suggest that ART may improve outcomes in critically ill HIV patients. In a review of all HIV-infected patients admitted to the ICU between 1996 and 1999, patients receiving combination ART at the time of ICU admission were less likely to present with 2 conditions associated with decreased survival: AIDS-associated diagnosis and decreased serum albumin levels.(6) In a study of all HIV-infected patients with PCP who were admitted to the ICU at San Francisco General Hospital between 1996 and mid-2001, patients who were on combination ART at the time of ICU admission had an improved survival compared with patients not receiving ART.(18) However, in another study from New York City, ICU mortality was not different in patients admitted between 1997 and 1999 when comparing patients receiving effective ART with those not on ART.(11) Further, the prior use of combination ART was not associated with differences in overall hospital mortality or length of stay.(11) Conclusions regarding the impact of ART on outcome are limited by the nonrandomized nature of these retrospective studies and by the inability to measure potential bias in the selection of which patients received ART. In addition, these studies do not address treatment failure, drug resistance, or medication nonadherence prior to ICU admission, all of which also influence outcome.(11) The question of whether critically ill HIV patients who require ICU admission for AIDS-related conditions should be started on ART while in the ICU can be answered definitively only by a randomized, prospective study.

Intensive care providers should be aware of potential complicating issues in patients who are already on ART or who initiate ART in the ICU. Antiretroviral and other HIV-related medications are involved in numerous drug-drug interactions that can alter serum drug levels and result in adverse effects.(19) Thus, the decision to initiate ART in a critically ill patient requires careful consideration. Each patient's medications should be reviewed carefully to prevent or minimize adverse effects when initiating new therapies in the ICU. Administration of ART can be complicated by unpredictable GI absorption of medications, lack of parenteral formulations for many medications, and complicated dosing regimens.(15) Clinicians also need to be acutely aware of complications that may result directly or indirectly from the use of ART. Patients may develop drug hypersensitivity reactions (perhaps more common with nonnucleoside reverse transcriptase inhibitors [NNRTIs] and with the nucleoside reverse transcriptase inhibitor [NRTI] abacavir than with others), resulting in fever, hypotension, and, as reported for abacavir, acute interstitial pneumonitis with respiratory failure.(20) Fatal lactic acidosis caused by NRTIs has been described.(20,21) (See HIV InSite Knowledge Base chapter "Metabolic Complications of HIV Therapy.") Concurrent use of zidovudine and corticosteroids may result in severe myopathy and respiratory muscle dysfunction.(20)

In addition, patients started on ART may experience worsening of a previously recognized or unrecognized underlying opportunistic infection, such as cryptococcosis, tuberculosis, or PCP, attributable to increased inflammatory responses related to immune reconstitution.(22-26,27,28) (See HIV InSite Knowledge Base chapter "Clinical Implications of Immune Reconstitution in AIDS.") A paradoxical increase in disease severity resulting from ART-associated immune reconstitution may be challenging to distinguish clinically from treatment failure or from a superimposed respiratory infection. In an HIV-infected patient who has recently initiated ART, this phenomenon of paradoxical worsening should be a diagnosis of exclusion. A careful evaluation should assess for worsening of underlying infection due to drug resistance or inadequate drug levels, and should rule out a new infection, drug toxicity, or noninfectious complication. At times, immune reconstitution syndromes can result in complications that require ICU care, such as from acute respiratory failure and worsening of CNS disease.

Sepsis is an increasingly common cause of ICU admission for HIV-infected patients and is associated with a high mortality.(7,17) In general, the management of sepsis in HIV-infected patients is similar to that in HIV-uninfected patients, and includes appropriate volume resuscitation, early institution of broad spectrum antibiotics, and vasopressors as needed to maintain adequate blood pressure. In choosing initial empiric antibiotics, consideration should be given to a patient's CD4 count and prophylactic antibiotics (which may select for bacterial resistance), as well as other risk factors for particular infections, such as intravenous drug use and exposures to endemic fungi and mycobacteria. As with other conditions, consideration also needs to be given to potential drug interactions with ART, particularly for protease inhibitors and NNRTIs with such classes of drugs as azoles and macrolides.

Although therapies such as corticosteroids and recombinant human activated protein C have been shown to improve survival in mainly HIV-uninfected populations with sepsis, their use cannot be extrapolated to all critically ill HIV-infected persons. PROWESS, the randomized controlled trial assessing the efficacy and safety of activated protein C in critically ill patients,(77) excluded HIV-infected patients with CD4 counts <=50 cells/µL. Additional exclusion criteria that are particularly relevant to HIV-infected patients included a history of CNS mass lesion and a platelet count of <30,000/µL. Low-dose steroids also have been demonstrated to improve survival in HIV-uninfected patients with severe sepsis requiring vasopressors who have a <9 mcg/dL increase in serum cortisol levels in response to corticotropin stimulation testing.(78) Whether steroids improve outcome in HIV-infected patients with sepsis is unknown, as this trial excluded HIV-infected patients. Nonetheless, the prevalence of adrenal insufficiency is high in HIV-infected critically ill patients, and thus recommendations have been made to consider all HIV-infected patients in the ICU for evaluation and treatment of adrenal insufficiency.(79)

Patients with neurologic manifestations of HIV may require critical care for intractable seizures or for an altered level of consciousness ranging from lethargy to coma; these conditions often can be associated with secondary respiratory failure that requires intubation and mechanical ventilation. The HIV-related neurologic processes that can cause an acute deterioration generally are mass lesions, meningitides, or, rarely, myelopathies. The most common etiologies of mass lesions are T gondii infection and primary CNS lymphoma, whereas the most common cause of meningitis is C neoformans infection.(80) Other major CNS infections encountered in the setting of HIV infection include progressive multifocal leukoencephalopathy, a progressive demyelinating disease; CMV; and herpes simplex virus (HSV) encephalitis.(81) Because HIV is a neurotropic virus, a variety of neurologic insults can occur, including myelopathy (occasionally causing spastic paralysis), peripheral neuropathy (rarely causing respiratory failure from a Guillain-Barré-like syndrome), and HIV encephalopathy or AIDS dementia. The clinical manifestations and evaluation of HIV-infected patients with neurologic disease are discussed in more detail in the HIV InSite Knowledge Base chapter "Neurologic Manifestations of HIV."

In a case series of all AIDS patients admitted to one ICU, 17% of admissions were for "neurologic failure," defined as an altered level of consciousness or seizures requiring ICU management.(82) Most of these patients required mechanical ventilation because of the severity of their neurologic dysfunction. The most frequent diagnosis was CNS toxoplasmosis. The 3-month mortality rate was 68%; profound coma was the cause of death in 50%. Independent predictors of death were a Glasgow Coma Score of <7 and clinical signs of brain stem involvement. The specific neurologic diagnosis, CD4 count, and prior health status were not predictive of outcome.

CNS dysfunction has a strong impact on a patient's ability to make decisions regarding health care, and can be particularly challenging when patients are hospitalized in the ICU and unable to communicate regarding end-of-life decisions. Approximately 40% of ART-untreated patients experience some neurologic process (typically AIDS dementia complex or cryptococcal meningitis) that limits decision-making capacity.(80) Clearly, whereas neurologic dysfunction affects a patient's ability to participate in clinical decision-making, it also affects the patient's perceived quality of life. In a study of outpatients in an AIDS clinic, 53% reported that they would want life-sustaining treatment if required, but only 19% reported that they would want this level of care if they also had coexistent AIDS dementia.(76) This study highlights the importance of establishing advance directives and designation of surrogate decision-makers early in the course of patient care, prior to the onset of clinical deterioration.

A variety of GI disorders can either necessitate admission to an ICU or complicate the course of care in an ICU. Although relatively infrequent compared with other GI complications of HIV outside the ICU, GI bleeding is the most common GI diagnosis necessitating ICU admission in HIV-infected patients.(2,4,5,83,84) Bleeding in HIV-infected patients is more likely to occur in the upper GI tract than in the lower GI tract.(84) Approximately half of the cases of upper GI bleeding in HIV-infected patients are attributable to HIV-related conditions, including infectious esophagitis and ulcers, Kaposi sarcoma, and AIDS-associated lymphoma.(84) HIV-infected patients also can present with non-HIV-associated causes of upper GI bleeding such as gastric or duodenal ulcers, variceal bleeding, or erosive gastritis. In lower GI bleeding, approximately 70% of cases are related to the underlying HIV infection.(84) CMV colitis and idiopathic colon ulcers are the most common causes. Other etiologies include Kaposi sarcoma, AIDS-associated lymphoma, Mycobacterium avium complex, and other HIV-associated infectious causes.(83)

The management of patients with HIV and acute GI bleeding is similar to that of non-HIV-infected individuals, and consists of fluid resuscitation, identification of the bleeding source, attainment of hemostasis, and prevention of recurrent bleeding. Bleeding may be exacerbated by concomitant HIV-associated thrombocytopenia. Furthermore, observational studies suggest that HIV-infected patients may have a higher rate of rebleeding than non-HIV-infected patients, although direct comparative studies have not been done.(84)

Other GI disorders requiring ICU admission include hepatic encephalopathy (often due to end-stage liver disease and underlying viral hepatitis), peritonitis, and bowel perforation. In AIDS patients, the most common cause of life-threatening abdominal pain is peritonitis from small bowel or colon enteritis caused by CMV, with or without demonstrable perforation.(85) Bowel perforations also have been associated with Kaposi sarcoma, lymphoma, and mycobacterial infection.(83) Patients also may present with pancreatitis, which can be caused by antiretroviral medications, pentamidine, or other drugs. AIDS cholangiopathy, caused by a variety of infectious and neoplastic processes, can range in presentation from asymptomatic cholestasis to fulminant biliary sepsis.(83) Critically ill patients with probable cholangitis should receive fluids and broad-spectrum antibiotics directed against gram-negative bacteria. Endoscopic retrograde cholangiopancreatography with sphincterotomy may be palliative in septic patients with sonographic evidence of common bile duct dilation.(85) Acute abdominal problems unrelated to immunosuppression also must be considered in evaluating the HIV-infected patient.

Decisions with regard to ICU care of AIDS patients, as with all ICU admissions, involve estimation of short-term prognosis and goals, benefits, and patient wishes regarding life-sustaining treatment. A fundamental part of a patient's right to autonomy includes the right of a competent and informed patient to refuse care in terminal as well as nonterminal diseases. An advance directive is the expression of a patient's preferences, determined while he or she possesses decision-making capacity, concerning future medical care, including life-sustaining measures.(86) Generally, advance directives take the form of either a living will or the designation of some trusted person (referred to as a "surrogate" or, in some states, "durable power of attorney for health care") with decision-making authority. Because living wills, as written instruments, may not address all medical scenarios encountered, a surrogate may provide more flexibility.(86) Additionally, the surrogate, by using his or her perception of the patient's wishes, can include any new relevant medical advances in weighing decisions.

Because it is seldom possible to predict when an event that impairs decision-making capacity may occur, patients should be encouraged to discuss their wishes about life-sustaining treatment with their designated surrogate as well as with their primary care provider as early as possible in the course of care. In patients with HIV, such discussions may be inappropriately delayed. In one study, almost 75% of AIDS patients wanted to discuss life-sustaining treatment with their physicians and thought that the most appropriate time to discuss issues of future care was in the outpatient setting.(76) Only one third of individuals surveyed in an outpatient AIDS clinic, however, had ever held such a discussion with their physicians. Patients who had been hospitalized within the past year were more likely to have a living will than those never hospitalized and also were more likely to have discussed life-sustaining treatments, at least with family members and partners.

A number of perceived barriers may delay discussions regarding end-of-life care. In another study, 57 AIDS patients and their primary care clinicians were interviewed to identify barriers and facilitators to patient-clinician communication about end-of-life care.(87) Clinicians reported more barriers to communication than did patients; some common reasons that clinicians did not discuss end-of-life care with patients included the perception that the patient had not yet experienced severe illness, the impression that the patient was not yet ready to discuss end-of-life care, and the lack of time available at appointments. End-of-life discussions may be facilitated using a "patient-centered" approach that incorporates an "HIV-specific advance directive," which describes specific scenarios relevant to HIV/AIDS as part of treatment choices.(88) Continued education and interventions are needed to improve the quality of communication between clinicians and patients regarding advance directives and end-of-life issues.

HIV-infected patients can require critical care for a number of reasons. Many patients may not be identified as HIV infected at the time of ICU admission, and providers must remain aware of the possibility of a new diagnosis of HIV infection among their critically ill patients. Acute respiratory failure remains the most common ICU-admitting diagnosis and usually is due to infectious causes, particularly PCP or BP. Other common indications for ICU admission include sepsis, CNS dysfunction, and GI complications, especially GI bleeding. As discussed, the spectrum and the outcome of critical illness in HIV patients appears to be changing in the setting of effective ART, and more patients are likely to present to the ICU with medical or surgical issues unrelated to their HIV infection. Overall, the reported in-hospital mortality rate of AIDS patients admitted to the ICU remains approximately 30-40%. However, patients who are admitted with a non-AIDS-associated diagnosis may have a decreased mortality risk compared with patients who are admitted with an AIDS-associated condition. Many of the challenges in the critical care of HIV-infected patients are similar to those encountered in caring for non-HIV-infected patients; as with all patients, discussion of end-of-life issues is a crucial part of this care.