Characteristics and Outcomes of HIV-Infected Patients With Severe Sepsis: Continued Risk in the Post–Highly Active Antiretroviral Therapy Era* Sushma K. Cribbs, MD, MSc1,2; Caroline Tse, MD3; Joel Andrews, RN, BSN4; Neeta Shenvi, MS5; Greg S. Martin, MD, MSc2 Objectives: Although highly active antiretroviral therapy has led to improved survival in HIV-infected individuals, outcomes for HIVinfected patients with sepsis in the post–highly active antiretroviral therapy era are conflicting. Access to highly active antiretroviral therapy and healthcare disparities continue to affect outcomes. We hypothesized that HIV-infected patients with severe sepsis would have worse outcomes compared with their HIV-uninfected counterparts in a large safety-net hospital where access to healthcare is low and delivery of critical care is delayed. Design: Secondary analysis of an ongoing prospective observational study between 2006 and 2010. Setting: Three adult ICUs (medical ICU, surgical ICU, and neurologic ICU) at Grady Memorial Hospital, Atlanta, GA. *See also p. 1779. 1 Pulmonary Medicine, Department of Veterans Affairs, Atlanta, GA. 2 Division of Pulmonary, Allergy and Critical Care, Department of Medicine, Emory University, Atlanta, GA. 3 Department of Medicine, Legacy Meridian Park Medical Center, Tualatin, OR. 4 Emory University, Atlanta, GA. 5 Department of Biostatistics, Rollins School of Public Health, Emory University, Atlanta, GA. Drs. Tse, Andrews, Shenvi, and Martin contributed to study design and database query. Drs. Andrews and Shenvi contributed to database creation and management. Drs. Tse, Andrews, Shenvi, and Martin contributed to analysis and interpretation of data. Drs. Tse, Andrews, Shenvi, and Martin contributed to article preparation. All authors read and approved the final article. Dr. Cribbs received support for article research from the National Institutes of Health (NIH; NIH KL2 TR000455 and NIH UL1 TR000454). Her institution received grant support from the NIH. Dr. Martin served as a board member for Cumberland Pharmaceuticals, Pulsion Medical Systems, the Society of Critical Care Medicine, and Medscape (Critical care editor); consulted for Grifols; and received support for article research from the NIH (NIH P50 AA-013757 and FDA R01 FD-003440). His institution served as board member for the NIH and Federal Drug Administration (data safety monitoring board), Baxter Healthcare, and Abbott Laboratories. The funding sources did not play any role in study design; in collection, analysis, and interpretation of data; in the writing of the article; and in the decision to submit the article for publication. The remaining authors have disclosed that they do not have any potential conflicts of interest. For information regarding this article, E-mail: [email protected] Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved. DOI: 10.1097/CCM.0000000000001003
1638
www.ccmjournal.org
Patients: Adult patients with severe sepsis in the ICU. Interventions: Baseline patient characteristics and clinical outcomes were collected. HIV-infected and HIV-uninfected patients with sepsis were compared using t tests, chi-square tests, and logistic regression; p values less than 0.05 indicated significance. Measurements and Main Results: Of 1,095 patients with severe sepsis enrolled, 165 (15%) were positive for HIV, with a median CD4 count of 41 (8–167). Twenty-two percent of HIV-infected patients were on highly active antiretroviral therapy prior to admission, and 80% had a CD4 count less than 200. HIV-infected patients had a greater hospital mortality (50% vs 38%; p < 0.01). HIV infection (odds ratio = 1.78; p = 0.005) was an independent predictor of mortality by multivariate regression modeling after adjusting for age, history of pneumonia, history of hospitalacquired infection, and history of sepsis. Conclusions: HIV-infected patients with severe sepsis continue to suffer worse outcomes compared with HIV-uninfected patients in a large urban safety-net hospital caring for patients with limited access to medical care. Further studies need to be done to investigate the effect of socioeconomic status and mitigate healthcare disparities among critically ill HIV-infected patients. (Crit Care Med 2015; 43:1638–1645) Key Words: highly active antiretroviral therapy; human immunodeficiency virus; intensive care; pneumonia; sepsis
S
ince the AIDS is epidemic, hospital admission rates and mortality rates have varied considerably (1, 2). In the last decade, overall hospitalization rates have decreased since the advent of highly active antiretroviral therapy (HAART) (1, 3), and survival among HIV-infected patients has improved considerably (2, 4, 5). However, despite these improvements, studies suggest that admission rates to the ICU have not changed and may even be increasing for HIV-infected individuals (6, 7). Sepsis remains among the most common medical conditions in the ICU and continues to be a leading cause of death in HIV-infected patients (8, 9), affecting short- and long-term August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
survival (10). Japiassú et al (11) prospectively studied critically ill HIV/AIDS patients in Brazil and showed that severe sepsis was associated with the highest rate of mortality at 28 days and 6 months. They further discovered that pulmonary infection remained the most common cause for admission to the ICU (11), similar to others (2, 3, 6, 12). In the early HIV era, the most common ICU admitting diagnosis was acute respiratory failure from Pneumocystis jiroveci pneumonia (PCP) (5, 13). In the past decade, the number of patients with AIDS-defining conditions and opportunistic infections admitted to the ICU has decreased considerably in regions with access to HAART. However, respiratory failure from pneumonia remains the most common diagnosis for HIV-infected patients admitted to the ICU. In the era of HAART, the pathogens have merely changed to more bacterial infections (4, 5, 14), and these infectious complications can result in increased rates of sepsis, organ dysfunction, and ICU admissions. HAART has improved hospital mortality among patients admitted to the ICU. Between 1996 and 1999, mortality decreased from 71% in the pre-HAART era to 29% in the early post-HAART era (2, 4, 12, 15). In the ICU, HAART has been associated with a number of predictors for decreased mortality, including higher CD4 count, lower viral load, lower Acute Physiology and Chronic Health Evaluation (APACHE) II score, and higher serum albumin (9). Although survival has improved for individuals living with HIV in the HAART era and some studies suggest that mortality rates are similar to non-HIV patients (10, 16, 17), optimal outcomes have not been achieved. Studies have shown that up to 50% of HIVinfected patients may not be on HAART at admission to the ICU (2, 12). Furthermore, healthcare disparities continue with respect to access to HIV testing, care, and treatment among population groups. HIV disproportionately affects blacks and Hispanics and Latinos. Lack of access to care can result in persons who are unaware of their HIV infection with a greater chance of transmission and persons who are not linked to care after diagnosis resulting in a greater proportion without access to HAART (18, 19). These disparities may also affect characteristics and outcomes of critically ill HIV-infected patients. The objective of this study was to determine baseline demographics, severity of disease, cause of disease, and outcomes of HIV-infected patients with severe sepsis or septic shock at a large county safety-net hospital in Atlanta, GA, between 2006 and 2010, in the era of HAART.
MATERIALS AND METHODS Study Population We performed a secondary analysis of an ongoing prospective observational study, the Emory Sepsis and Acute Respiratory Distress Syndrome (ARDS) Registry, in three adult ICUs (medical ICU, surgical ICU, and neurologic ICU) at Grady Memorial Hospital, a 953-bed county hospital and an affiliate of Emory University, in Atlanta, GA, between 2006 and 2010. The Emory Sepsis and ARDS Registry used a standardized screening protocol to identify patients who met the American College of Chest Critical Care Medicine
Physicians/Society of Critical Care Medicine (20) criteria for severe sepsis. The three ICUs together comprise 60 beds, and patients were screened on a daily basis by designated screeners. This study was approved by Emory University’s Institutional Review Board and the Grady Research Oversight Committee. Patients who were less than 18 years old and those meeting the criteria for severe sepsis for more than 72 hours prior to screening were excluded from the database to enable capture of acutely ill patients. Illness severity (APACHE II) and organ dysfunction scores (Sequential Organ Failure Assessment [SOFA]) were obtained on 355 subjects who were concomitantly enrolled into additional or secondary studies. Questionnaires were completed through direct interview of patients and their families and through review of medical records. Baseline patient demographics, comorbidities, sepsis characteristics, and clinical outcomes were collected from review of medical records. The etiology of severe sepsis was determined by consensus among ICU clinicians through review of clinical, radiographic, and microbiologic findings. Patients were stratified by source of infection—blood, respiratory, genitourinary, gastrointestinal, and other (pleural, cerebrospinal fluid, skin)— and types of organism (Gram positive, Gram negative, fungal, and other). Infectious sources were marked as primary or secondary sources. Patients were labeled based on their primary source of infection which was determined by dates and clinical charts. Information on HIV status, including CD4 count and use of HAART, was also obtained through review of the medical chart and review of charts from the infectious disease department. HAART was defined as a combination of three or more antiretroviral drugs, belonging to at least two classes among the following: nucleoside reverse transcriptase inhibitors, non–nucleoside reverse transcriptase inhibitors, and protease inhibitors. CD4 count and viral loads were determined by reviewing laboratory data in the 3 months prior to admissions. Statistical Analysis Univariate comparisons between HIV subjects and non-HIV subjects were calculated and evaluated for statistical significance at an α of 0.05 using a two-sample t test for continuous variables and a chi-square test for categorical variables. Results were reported as the mean (sd) or as a percentage. To examine the association of HIV infection and hospital mortality, a univariate logistic analysis was done with hospital mortality as the dependent variable and other variables of interest, including HIV, age, history of pneumonia, history of hospital-acquired infection, and history of sepsis. Then, a multivariate logistic model was created with hospital mortality as the dependent variable, HIV (the variable of interest) and variables that were either significant in the univariate analysis between HIV and non-HIV or clinically relevant. Two multivariate models were created (Tables 1 and 2), including all subjects and an additional analysis in the subset of subjects with APACHE II. The odds ratio (OR) and its 95% CI were calculated for each risk factor in the presence of the others in the final multivariable model. The goodness of fit of the logistic model was evaluated using the Hosmer-Lemeshow test. Machine learning algorithms www.ccmjournal.org
1639
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
Table 1. Multivariable Logistic Regression Model for Predictors of Hospital Mortality Without Acute Physiology and Chronic Health Evaluation II (n = 1,095) OR
95% CI
p
Reliability %a
Age (per 1-yr increase)
1.02
1.01–1.02
< 0.0001
98
History of pneumonia
0.92
0.51–1.68
0.79
7
History of hospital-acquired infection
1.22
0.73–2.05
0.45
9
History of sepsis
0.67
0.32–1.40
0.29
16
HIV
1.78
1.19–2.67
0.005
88
Variable
OR = odds ratio. a Percentage of times risk factor appears in 1,000 bootstrap models. Hosmer-Lemeshow goodness-of-fit test χ2 = 0.51.
were used with bootstrap bagging to identify the predictors of mortality. Statistical analysis was performed through Number Crunching Statistical Software (NCSS 2007; NCSS, Kaysville, UT) and SAS version 9.3 (SAS Institute, Cary, NC).
RESULTS Baseline Demographics and Characteristics of Patients With Severe Sepsis Over the 5-year period, 1,095 patients with severe sepsis were enrolled into the Registry, 165 (15%) of whom were positive for HIV, with a median CD4 count of 41 (8–167) and median viral load of 4.8 (2.1–5.1) log copies/mL. HIV-infected patients were predominantly younger (45 ± 9.8 yr vs 54 ± 17.4 yr; p < 0.001), with lower body mass index (24.2 ± 6.4 vs 28.9 ± 9.9; p < 0.001). The majority of patients in both groups were African-American (Table 3). HIV-infected patients were more likely to be on chronic dialysis (5% vs 2%; p = 0.02), have a history of pneumonia (45% vs 12%; p < 0.001), history of previous infection requiring hospital admission (67% vs 10%; p < 0.001), and a history of sepsis (14% vs 3%; p = 0.03). Non–HIV-infected patients were more likely to have coronary artery disease (8% vs 0%; p < 0.001) and diabetes (29% vs 9%; p < 0.001) compared with the HIV-infected patients. There were no differences in the prevalence of chronic obstructive pulmonary disease, renal failure, and cancer between the
groups. Although not statistically significant, HIV-infected patients had greater illness severity at admission to the ICU as determined by APACHE II score (26.3 ± 7.9 vs 24.5 ± 7.7; p = 0.08). HIV-infected patients also had a statistically significant higher A-a gradient (308.0 ± 178.8 vs 379.8 ± 184.1; p = 0.007). There were no differences in baseline SOFA scores between the two groups. There were no differences in the prevalence of shock requiring vasopressors and mechanical ventilation with acute lung injury/ARDS between HIVinfected and non–HIV-infected patients. However, there was a trend toward greater prevalence of acute renal failure (15% vs 11%; p = 0.14) and coagulopathy (16% vs 11%; p = 0.07) in the HIV-infected patients. Outcomes Among Patients With Severe Sepsis HIV-infected patients had a greater hospital mortality (50% vs 38%; p < 0.01) and a shorter ICU length of stay (LOS) (13.7 ± 12.0 vs 16.8 ± 18.6; p = 0.04) compared with HIVuninfected patients (Table 4). There was no significant difference in ICU mortality, duration of mechanical ventilation, or hospital LOS. Among survivors, HIV-infected patients still had a shorter ICU LOS (13.1 ± 12.3 d vs 17.4 ± 18.1 d; p = 0.04), and there remained no differences in duration of mechanical ventilation or hospital LOS. Univariate logistic regression analysis for hospital mortality, which included variables that significantly increased in HIV versus non-HIV,
Table 2. Multivariable Logistic Regression Model for Predictors of Hospital Mortality With Acute Physiology and Chronic Health Evaluation II (n = 355) OR
95% CI
p
Acute Physiology and Chronic Health Evaluation II (per 1-unit increase)
1.08
1.05–1.12
< 0.0001
100
History of pneumonia
0.60
0.23–1.60
0.30
17.6
History of hospital-acquired infection
1.03
0.39–2.73
0.95
5.9
History of sepsis
0.59
0.19–1.77
0.34
28.7
HIV
1.96
1.03–3.75
0.04
44.9
Variable
Reliability %a
OR = odds ratio. a Percentage of times risk factor appears in 1,000 bootstrap models. Hosmer-Lemeshow goodness-of-fit test χ2 = 0.72.
1640
www.ccmjournal.org
August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
Table 3.
Baseline Demographics and Characteristics of All Patients With Severe Sepsis
Variable
N
HIV (–)
HIV (+)
932
165
Age (yr), mean ± sd
54 ± 17.4
Male, n (%)
578 (62)
Body mass index (kg/m2), mean ± sd
28.9 ± 9.9
45 ± 9.8
p
< 0.001
105 (64) 24.2 ± 6.4
0.69 < 0.001
Race Caucasian
197 (21)
12 (7)
African-American
670 (72)
144 (87)
Hispanic
38 (4)
5 (3)
Other
27 (3)
4 (3)
0.002
Baseline comorbidities, n (%) Coronary artery disease
20 (8)
0
< 0.001
Chronic obstructive pulmonary disease
56 (12)
13 (8)
0.16
Renal failure
66 (14)
23 (14)
0.92
Chronic dialysis
17 (2)
8 (5)
0.02
Cirrhosis
39 (8)
7 (4)
0.08
Diabetes
137 (29)
14 (9)
< 0.001
History of pneumonia
32 (12)
71 (45)
< 0.001
History of infection requiring hospital admission
63 (25)
94 (64)
< 0.001
History of sepsis
19 (7)
22 (14)
0.03
History of cancer
59 (13)
19 (12)
0.60
Shock requiring vasopressors
532 (57)
100 (61)
0.38
Mechanical ventilation with acute lung injury/acute respiratory distress syndrome
262 (28)
45 (27)
0.82
99 (11)
24 (15)
0.14
101 (11)
26 (16)
0.07
Sequential Organ Failure Assessment score at ICU admission,a mean ± sd
10.1 ± 3.6
10.3 ± 3.7
0.63
Acute Physiology and Chronic Health Evaluation II score at ICU admission,a mean ± sd
24.5 ± 7.7
26.3 ± 7.9
0.08
1.3 ± 8.5
1.7 ± 0.8
0.66
308.0 ± 178.8
379.8 ± 184.1
0.007
Organ dysfunction, n (%)
Acute renal failure Coagulopathy
Albumin (mg/dL) at ICU admission,a mean ± sd A-a gradient at ICU admission,a mean ± sd CD4 count, median (IQR)
41 (8–167)
Viral load (Log copies/mL), median (IQR) a
4.8 (2.1–5.1)
IQR = interquartile range. a n is different than stated above, changes are noted as follows: Sequential Organ Failure Assessment and Acute Physiology and Chronic Health Evaluation: HIV (–) n = 285, HIV (+) n = 76; albumin: HIV (–) n = 286, HIV (+) n = 72; A-a gradient: HIV (–) n = 232, HIV (+) n = 57; and viral load: HIV (+) n = 71.
showed that HIV infection was associated with an OR of 1.58 (95% CI, 1.13–2.20) (Table 5). Two multivariate logistic regression analyses were performed to test the independent association of HIV with mortality. Adjusting for the clinical covariates age, history of pneumonia, history of Critical Care Medicine
hospital-acquired infection, and history of sepsis, HIV infection was associated with an OR of 1.78 (95% CI, 1.19–2.67) for hospital mortality (p = 0.005) (Table 1). In the clinical model including APACHE II, history of pneumonia, history of hospital-acquired infection, and history of sepsis entered www.ccmjournal.org
1641
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
Table 4.
Outcomes in Patients With Severe Sepsis
Variable
p
HIV (–)
HIV (+)
932
165
Duration of mechanical ventilation (d), mean ± sd
14.7 ± 17.0
12.9 ± 11.9
0.26
Hospital LOS (d), mean ± sd
30.5 ± 31.4
29.7 ± 26.0
0.77
ICU LOS (d), mean ± sd
16.8 ± 18.6
13.7 ± 12.0
0.04
Home
302 (32)
52 (32)
0.03
Rehab
62 (7)
5 (3)
128 (14)
15 (9)
80 (9)
11 (7)
ICU mortality
295 (32)
58 (35)
0.38
Hospital mortality, n (%)
358 (38)
82 (50)
0.007
n
Discharge status, n (%)
Skilled nursing Other healthcare
LOS = length of stay.
as potential confounders into the model, HIV infection was associated with an OR of 1.96 (95% CI, 1.03–3.75) for increased hospital mortality (p = 0.04) (Table 1). Characteristics and Outcomes Among Severe Sepsis Patients With HIV As shown in Table 6, 22% of HIV-infected patients were on HAART before admission, 15% of patients had HAART initiated during admission, and 63% of patients were never on HAART. There were 19 patients with a new diagnosis of HIV. Stratifying by HAART status at admission, mortality rates were 51%, 39%, and 50% for those patients who were on HAART at admission (n = 35), those who had HAART started at admission (n = 23), and those who never received HAART, respectively (n = 96) (Table 6). Given that HAART started at admission may have limited effect, we compared mortality rates between individuals on HAART at admission versus those who were not on HAART and did not find a statistically significant difference (51% vs 47%; p = 0.62). Stratifying by CD4 count, 80% of HIV-infected patients had a CD4 count less than 200 and 65% had a CD4 count less than 100. There Table 5.
was no significant difference in mortality rates between HIVinfected patients with a CD4 count less than 100 versus greater than 100 (52% vs 46%; p = 0.45). There was also no significant difference in mortality rates between HIV-infected patients with a CD4 count less than 200 versus greater than 200 (50% vs 47%; p = 0.73). Primary Infection and Microbiology The primary infection source was known for 162 HIV-infected patients and for 870 HIV-uninfected patients, and specific microbiology was known for 91 HIV-infected patients and 180 HIV-uninfected patients. For both HIV-infected and HIV-uninfected patients, the primary source of infection was respiratory (61% vs 48%, respectively; p = 0.002). Among 91 HIV-infected and 343 HIV-uninfected subjects, a Gram-positive source was the most common (39% vs 53%, respectively; p = 0.01). HIV-infected patients had a greater percentage of fungal infections compared with HIV-uninfected patients (33% vs 4%; p < 0.0001). In HIV-infected patients, 79% of all fungal infections were due to PCP, 10% were due to Candida, and 10% were due to Cryptococcus.
Univariate Logistic Regression Model for Predictors of Hospital Mortality n
OR
95% CI
1,095
1.01
1.01–1.02
0.0007
355
1.08
1.05–1.12
< 0.0001
History of pneumonia
1,095
1.28
0.85–1.92
0.24
History of hospital-acquired infection
1,095
1.31
0.93–1.84
0.12
History of sepsis
1,095
0.86
0.45–1.63
0.63
HIV
1,095
1.58
1.13–2.20
0.007
Variable
Age (per 1-yr increase) Acute Physiology and Chronic Health Evaluation II (per 1-unit increase)
p
OR = odds ratio.
1642
www.ccmjournal.org
August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
Characteristics and Outcomes of Severe Sepsis Patients With HIV Table 6.
n (%)
Median CD4 Count (IQR)
Mortality, n (%)
On HAART at admission
35 (22)
121 (25–244)
18 (51)
HAART started this admission
23 (15)
16 (6–111)
9 (39)
Not on HAART
96 (63)
37 (6–152)
48 (50)
Variable
HAART characteristics
CD4 count (within 3 mo of admission) CD4 < 100
106 (65)
55 (52)
CD4: 100–200
24 (15)
11 (46)
CD4 > 200
34 (21)
16 (47)
IQR = interquartile range, HAART = highly active antiretroviral therapy.
DISCUSSION The results from this study conducted in a large urban safetynet hospital showed that HIV-infected patients with severe sepsis have higher hospital mortality when compared with HIV-uninfected patients with severe sepsis. This finding was robust after adjustment for cofounders, with HIV infection still associated with almost two-fold greater odds of hospital mortality. In addition, HIV-infected patients also had a shorter ICU LOS compared with HIV-uninfected patients. This was not due to difference in ICU mortality, which was not significantly different between the two groups. However, despite a shorter ICU LOS, there were no significant differences in hospital LOS, suggesting that HIV-infected patients still remained in the hospital, perhaps due to secondary conditions that required additional hospital care prior to discharge. Among the HIV-infected patients, there was also a greater prevalence of infectious baseline comorbidities, including chronic dialysis, a history of pneumonia, a history of infection requiring ICU admission, and a history of sepsis. Further, in our cohort, there were a large number of patients who were never on HAART. This could also reflect the large percentage of HIV-infected subjects with a CD4 count less than 100. These results suggest that significant healthcare disparities still exist in large urban areas of the United States where there are many HIV-infected individuals with limited access to care and who are unable to take HAART, potentially resulting in increased morbidity and mortality. More recent studies have shown substantial differences in ICU and hospital mortality rates among HIV-infected patients with the widespread use of HAART compared to the pre-HAART era. In developed countries, survival for critically ill HIV-infected patients has continued to improve in the HAART era (2, 5, 17, 21). Some have even noted that survival Critical Care Medicine
for HIV-infected patients in the ICU was comparable to general medical patients (22). Similar to our study, Akgün et al (6) showed that HIV-infected veterans experienced a higher 30-day mortality compared with HIV-uninfected patients. This study used the Veterans Aging Cohort Study (VACS) Risk Index (VACS Index) score, which included measures of HIV-specific and general organ injury developed to predict all-cause mortality in HIV-infected patients (21). However, unlike our study, many of the subjects were on HAART. Comparatively, many recent studies in underdeveloped countries have shown higher mortality rates ranging from 45% to 80% (11, 23), where AIDS-related diseases are more common and access to HAART is difficult. Mortality rates among HIVinfected patients in our institution mirror the rates in the pre-HAART era at other institutions. Independent predictors of mortality have been poor functional status, more than 1 year between HIV diagnosis and ICU admission, age, higher illness severity, low serum albumin, mechanical ventilation, and respiratory failure (17, 24, 25). This is likely due to a large number of HIV-infected individuals in the Atlanta area who lack access to medical care and health insurance. The Ponce de Leon Center in Atlanta, GA, is one of the largest facilities dedicated to the treatment of HIV/AIDS in the United States. Many of the patients who are admitted to Grady Hospital are taken care in this facility where it is known that approximately 40% of all patients are uninsured. These data may reflect the low percentage (22%) of HIV-infected patients with severe sepsis who were on HAART at admission to the hospital. Studies have shown that critically ill uninsured patients are more likely to be admitted to an ICU when hospitalized and more likely to die in the ICU (26). Delay of care can also result in greater burden of comorbid illness and higher severity of illness, which has been noted in our cohort. The impact of HAART on ICU outcomes is still unclear. Many studies have noted that HAART use in the ICU was not associated with survival (5, 22, 25), but associated with increased CD4 counts. HAART administration in the ICU is a challenge for a number of reasons. Difficulties with medication delivery, drug interactions, and toxicities, in addition to concerns for viral resistance, pose challenges to all healthcare providers in the ICU. In addition, lack of screening, resources for HAART initiation, and maintenance may also play a role. In our study, the current ICU admission resulted in a new diagnosis of HIV in approximately 14% of the HIV-infected patients. Other studies have shown that even though HAART use specifically was not associated with survival, survival rates have improved in the HAART era (17). In our cohort, HIV-infected patients had a high mortality regardless of HAART. This is likely because approximately 80% of our cohort had a CD4 count less than 200 and, further, approximately 64% of patients in our cohort had a CD4 count less than 100. This suggests that, regardless of HAART, many of these patients had advanced AIDS, which contributed to their poor outcomes. However, given the small subgroups, this cohort is insufficient to test HAART initiation in the ICU. www.ccmjournal.org
1643
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
As in prior studies, respiratory infection was the most common etiology of sepsis in our HIV-infected patients (4, 5, 11, 12). Although PCP has become less frequent in many areas of the world, bacterial pneumonia has become increasingly more common (2, 4, 5, 14, 27). In our ICU patient population, infection was predominantly due to a Gram-positive source in both HIV-infected and HIV-uninfected patients. However, fungal causes of infection, mostly secondary to PCP, were much higher in the HIV-infected population. Overall, the prevalence of PCP has declined with the introduction of HAART and trimethoprim-sulfamethoxazole prophylaxis and treatment (28); however, we showed that in our large urban safety-net hospital, that PCP infections continue to occur at a high rate. PCP has been noted in other studies to occur mostly among persons unaware of their HIV infection, those who fail to seek medical care, and those who fail to adhere or respond to HAART (26, 29). Our study has several limitations. First, it was a secondary analysis of an ongoing prospective cohort study. Limited by the available data collected in this study, we were unable to determine certain aspects of ICU care that may have impacted patient outcomes. In this complete case analysis, we have presented two regression models, with APACHE II score substituted for age in the second model. Although APACHE II is a clinically significant variable, including this variable in the final multivariate model limits the sample size, resulting in a model with lower reliability. However, excluding this important variable creates clinical bias. Therefore, both models have been included and, importantly, the ORs for HIV are quantitatively the same and clinically relevant in both models. Second, our sample size of HIV-infected patients was moderate even though our overall severe sepsis sample size was large. Third, larger numbers of enrollment are needed in order to determine the impact of HAART in our ICU as the subgroups of patients initiated and not on HAART were small. In addition, given the retrospective nature of the study, information on viral loads, duration of HIV illness, and HAART are unknown on a number of patients. Therefore, possibility of HIV resistance to HAART is also unknown and could be a significant bias. Furthermore, this is a single-center study, subject to local ICU admission criteria and ICU practices, which may not necessarily be generalized. However, data on an urban inner city hospital in the current HAART era are lacking and therefore informative of the nature of critically ill HIV-infected patients. Finally, we are not able to draw conclusions about posthospital outcomes, including 60- and 90-day mortality, even though deaths may continue to accrue differentially after discharge, and this outcome is potentially more important to patients, providers, and healthcare organizations.
CONCLUSIONS Our study shows that HIV-infected patients with severe sepsis are still at risk of poor outcomes compared with HIV-uninfected patients in a large urban hospital setting composed of uninsured patients who lack access to medical care. Further studies need to be done to investigate the effect of socioeconomic status and mitigate healthcare disparities among critically ill HIV-infected patients. 1644
www.ccmjournal.org
ACKNOWLEDGMENTS We thank Celeste Sarmiento, MD, and the Grady Research Team for their assistance with patient screening, enrollment, and database management and Laura Ward, MSPH, for her review of the statistical methods.
REFERENCES
1. Buchacz K, Baker RK, Moorman AC, et al; HIV Outpatient Study (HOPS) Investigators: Rates of hospitalizations and associated diagnoses in a large multisite cohort of HIV patients in the United States, 1994-2005. AIDS 2008; 22:1345–1354 2. Narasimhan M, Posner AJ, DePalo VA, et al: Intensive care in patients with HIV infection in the era of highly active antiretroviral therapy. Chest 2004; 125:1800–1804 3. Nickas G, Wachter RM: Outcomes of intensive care for patients with human immunodeficiency virus infection. Arch Intern Med 2000; 160:541–547 4. Morris A, Creasman J, Turner J, et al: Intensive care of human immunodeficiency virus-infected patients during the era of highly active antiretroviral therapy. Am J Respir Crit Care Med 2002; 166:262–267 5. Powell K, Davis JL, Morris AM, et al: Survival for patients with HIV admitted to the ICU continues to improve in the current era of combination antiretroviral therapy. Chest 2009; 135:11–17 6. Akgün KM, Tate JP, Pisani M, et al: Medical ICU admission diagnoses and outcomes in human immunodeficiency virus-infected and virusuninfected veterans in the combination antiretroviral era. Crit Care Med 2013; 41:1458–1467 7. Coquet I, Pavie J, Palmer P, et al: Survival trends in critically ill HIVinfected patients in the highly active antiretroviral therapy era. Crit Care 2010; 14:R107 8. Croda J, Croda MG, Neves A, et al: Benefit of antiretroviral therapy on survival of human immunodeficiency virus-infected patients admitted to an intensive care unit. Crit Care Med 2009; 37:1605–1611 9. Huang L, Quartin A, Jones D, et al: Intensive care of patients with HIV infection. N Engl J Med 2006; 355:173–181 10. Medrano J, Alvaro-Meca A, Boyer A, et al: Mortality of patients infected with HIV in the intensive care unit (2005-2010): Significant role of chronic hepatitis C and severe sepsis. Crit Care 2014; 18:475 11. Japiassú AM, Amâncio RT, Mesquita EC, et al: Sepsis is a major determinant of outcome in critically ill HIV/AIDS patients. Crit Care 2010; 14:R152 12. Casalino E, Wolff M, Ravaud P, et al: Impact of HAART advent on admission patterns and survival in HIV-infected patients admitted to an intensive care unit. AIDS 2004; 18:1429–1433 13. Rosen MJ, Clayton K, Schneider RF, et al: Intensive care of patients with HIV infection: Utilization, critical illnesses, and outcomes. Pulmonary Complications of HIV Infection Study Group. Am J Respir Crit Care Med 1997; 155:67–71 14. Khouli H, Afrasiabi A, Shibli M, et al: Outcome of critically ill human immunodeficiency virus-infected patients in the era of highly active antiretroviral therapy. J Intensive Care Med 2005; 20:327–333 15. Afessa B, Green B: Clinical course, prognostic factors, and outcome prediction for HIV patients in the ICU. The PIP (Pulmonary complications, ICU support, and prognostic factors in hospitalized patients with HIV) study. Chest 2000; 118:138–145 16. Casalino E, Mendoza-Sassi G, Wolff M, et al: Predictors of short- and long-term survival in HIV-infected patients admitted to the ICU. Chest 1998; 113:421–429 17. van Lelyveld SF, Wind CM, Mudrikova T, et al: Short- and long-term outcome of HIV-infected patients admitted to the intensive care unit. Eur J Clin Microbiol Infect Dis 2011; 30:1085–1093 18. Giordano TP, Gifford AL, White AC Jr, et al: Retention in care: A challenge to survival with HIV infection. Clin Infect Dis 2007; 44:1493–1499 19. Hall HI, Holtgrave DR, Maulsby C: HIV transmission rates from persons living with HIV who are aware and unaware of their infection. AIDS 2012; 26:893–896 20. Bone RC, Balk RA, Cerra FB, et al: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 1992; 101:1644–1655 21. Justice AC, McGinnis KA, Skanderson M, et al; VACS Project Team: Towards a combined prognostic index for survival in HIV infection: The role of ‘non-HIV’ biomarkers. HIV Med 2010; 11:143–151 22. Dickson SJ, Batson S, Copas AJ, et al: Survival of HIV-infected patients in the intensive care unit in the era of highly active antiretroviral therapy. Thorax 2007; 62:964–968 23. Amâncio FF, Lambertucci JR, Cota GF, et al: Predictors of the shortand long-term survival of HIV-infected patients admitted to a Brazilian intensive care unit. Int J STD AIDS 2012; 23:692–697 24. Akgün KM, Pisani M, Crothers K: The changing epidemiology of HIVinfected patients in the intensive care unit. J Intensive Care Med 2011; 26:151–164
Critical Care Medicine
25. Greenberg JA, Lennox JL, Martin GS: Outcomes for critically ill patients with HIV and severe sepsis in the era of highly active antiretroviral therapy. J Crit Care 2012; 27:51–57 26. Danis M, Linde-Zwirble WT, Astor A, et al: How does lack of insurance affect use of intensive care? A population-based study. Crit Care Med 2006; 34:2043–2048 27. Vincent B, Timsit JF, Auburtin M, et al: Characteristics and outcomes of HIV-infected patients in the ICU: Impact of the highly active antiretroviral treatment era. Intensive Care Med 2004; 30:859–866 28. Weverling GJ, Mocroft A, Ledergerber B, et al: Discontinuation of Pneumocystis carinii pneumonia prophylaxis after start of highly active antiretroviral therapy in HIV-1 infection. EuroSIDA Study Group. Lancet 1999; 353:1293–1298 29. Restrepo MI, Mortensen EM, Rello J, et al: Late admission to the ICU in patients with community-acquired pneumonia is associated with higher mortality. Chest 2010; 137:552–557
www.ccmjournal.org
1645
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
1638
www.ccmjournal.org
Patients: Adult patients with severe sepsis in the ICU. Interventions: Baseline patient characteristics and clinical outcomes were collected. HIV-infected and HIV-uninfected patients with sepsis were compared using t tests, chi-square tests, and logistic regression; p values less than 0.05 indicated significance. Measurements and Main Results: Of 1,095 patients with severe sepsis enrolled, 165 (15%) were positive for HIV, with a median CD4 count of 41 (8–167). Twenty-two percent of HIV-infected patients were on highly active antiretroviral therapy prior to admission, and 80% had a CD4 count less than 200. HIV-infected patients had a greater hospital mortality (50% vs 38%; p < 0.01). HIV infection (odds ratio = 1.78; p = 0.005) was an independent predictor of mortality by multivariate regression modeling after adjusting for age, history of pneumonia, history of hospitalacquired infection, and history of sepsis. Conclusions: HIV-infected patients with severe sepsis continue to suffer worse outcomes compared with HIV-uninfected patients in a large urban safety-net hospital caring for patients with limited access to medical care. Further studies need to be done to investigate the effect of socioeconomic status and mitigate healthcare disparities among critically ill HIV-infected patients. (Crit Care Med 2015; 43:1638–1645) Key Words: highly active antiretroviral therapy; human immunodeficiency virus; intensive care; pneumonia; sepsis
S
ince the AIDS is epidemic, hospital admission rates and mortality rates have varied considerably (1, 2). In the last decade, overall hospitalization rates have decreased since the advent of highly active antiretroviral therapy (HAART) (1, 3), and survival among HIV-infected patients has improved considerably (2, 4, 5). However, despite these improvements, studies suggest that admission rates to the ICU have not changed and may even be increasing for HIV-infected individuals (6, 7). Sepsis remains among the most common medical conditions in the ICU and continues to be a leading cause of death in HIV-infected patients (8, 9), affecting short- and long-term August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
survival (10). Japiassú et al (11) prospectively studied critically ill HIV/AIDS patients in Brazil and showed that severe sepsis was associated with the highest rate of mortality at 28 days and 6 months. They further discovered that pulmonary infection remained the most common cause for admission to the ICU (11), similar to others (2, 3, 6, 12). In the early HIV era, the most common ICU admitting diagnosis was acute respiratory failure from Pneumocystis jiroveci pneumonia (PCP) (5, 13). In the past decade, the number of patients with AIDS-defining conditions and opportunistic infections admitted to the ICU has decreased considerably in regions with access to HAART. However, respiratory failure from pneumonia remains the most common diagnosis for HIV-infected patients admitted to the ICU. In the era of HAART, the pathogens have merely changed to more bacterial infections (4, 5, 14), and these infectious complications can result in increased rates of sepsis, organ dysfunction, and ICU admissions. HAART has improved hospital mortality among patients admitted to the ICU. Between 1996 and 1999, mortality decreased from 71% in the pre-HAART era to 29% in the early post-HAART era (2, 4, 12, 15). In the ICU, HAART has been associated with a number of predictors for decreased mortality, including higher CD4 count, lower viral load, lower Acute Physiology and Chronic Health Evaluation (APACHE) II score, and higher serum albumin (9). Although survival has improved for individuals living with HIV in the HAART era and some studies suggest that mortality rates are similar to non-HIV patients (10, 16, 17), optimal outcomes have not been achieved. Studies have shown that up to 50% of HIVinfected patients may not be on HAART at admission to the ICU (2, 12). Furthermore, healthcare disparities continue with respect to access to HIV testing, care, and treatment among population groups. HIV disproportionately affects blacks and Hispanics and Latinos. Lack of access to care can result in persons who are unaware of their HIV infection with a greater chance of transmission and persons who are not linked to care after diagnosis resulting in a greater proportion without access to HAART (18, 19). These disparities may also affect characteristics and outcomes of critically ill HIV-infected patients. The objective of this study was to determine baseline demographics, severity of disease, cause of disease, and outcomes of HIV-infected patients with severe sepsis or septic shock at a large county safety-net hospital in Atlanta, GA, between 2006 and 2010, in the era of HAART.
MATERIALS AND METHODS Study Population We performed a secondary analysis of an ongoing prospective observational study, the Emory Sepsis and Acute Respiratory Distress Syndrome (ARDS) Registry, in three adult ICUs (medical ICU, surgical ICU, and neurologic ICU) at Grady Memorial Hospital, a 953-bed county hospital and an affiliate of Emory University, in Atlanta, GA, between 2006 and 2010. The Emory Sepsis and ARDS Registry used a standardized screening protocol to identify patients who met the American College of Chest Critical Care Medicine
Physicians/Society of Critical Care Medicine (20) criteria for severe sepsis. The three ICUs together comprise 60 beds, and patients were screened on a daily basis by designated screeners. This study was approved by Emory University’s Institutional Review Board and the Grady Research Oversight Committee. Patients who were less than 18 years old and those meeting the criteria for severe sepsis for more than 72 hours prior to screening were excluded from the database to enable capture of acutely ill patients. Illness severity (APACHE II) and organ dysfunction scores (Sequential Organ Failure Assessment [SOFA]) were obtained on 355 subjects who were concomitantly enrolled into additional or secondary studies. Questionnaires were completed through direct interview of patients and their families and through review of medical records. Baseline patient demographics, comorbidities, sepsis characteristics, and clinical outcomes were collected from review of medical records. The etiology of severe sepsis was determined by consensus among ICU clinicians through review of clinical, radiographic, and microbiologic findings. Patients were stratified by source of infection—blood, respiratory, genitourinary, gastrointestinal, and other (pleural, cerebrospinal fluid, skin)— and types of organism (Gram positive, Gram negative, fungal, and other). Infectious sources were marked as primary or secondary sources. Patients were labeled based on their primary source of infection which was determined by dates and clinical charts. Information on HIV status, including CD4 count and use of HAART, was also obtained through review of the medical chart and review of charts from the infectious disease department. HAART was defined as a combination of three or more antiretroviral drugs, belonging to at least two classes among the following: nucleoside reverse transcriptase inhibitors, non–nucleoside reverse transcriptase inhibitors, and protease inhibitors. CD4 count and viral loads were determined by reviewing laboratory data in the 3 months prior to admissions. Statistical Analysis Univariate comparisons between HIV subjects and non-HIV subjects were calculated and evaluated for statistical significance at an α of 0.05 using a two-sample t test for continuous variables and a chi-square test for categorical variables. Results were reported as the mean (sd) or as a percentage. To examine the association of HIV infection and hospital mortality, a univariate logistic analysis was done with hospital mortality as the dependent variable and other variables of interest, including HIV, age, history of pneumonia, history of hospital-acquired infection, and history of sepsis. Then, a multivariate logistic model was created with hospital mortality as the dependent variable, HIV (the variable of interest) and variables that were either significant in the univariate analysis between HIV and non-HIV or clinically relevant. Two multivariate models were created (Tables 1 and 2), including all subjects and an additional analysis in the subset of subjects with APACHE II. The odds ratio (OR) and its 95% CI were calculated for each risk factor in the presence of the others in the final multivariable model. The goodness of fit of the logistic model was evaluated using the Hosmer-Lemeshow test. Machine learning algorithms www.ccmjournal.org
1639
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
Table 1. Multivariable Logistic Regression Model for Predictors of Hospital Mortality Without Acute Physiology and Chronic Health Evaluation II (n = 1,095) OR
95% CI
p
Reliability %a
Age (per 1-yr increase)
1.02
1.01–1.02
< 0.0001
98
History of pneumonia
0.92
0.51–1.68
0.79
7
History of hospital-acquired infection
1.22
0.73–2.05
0.45
9
History of sepsis
0.67
0.32–1.40
0.29
16
HIV
1.78
1.19–2.67
0.005
88
Variable
OR = odds ratio. a Percentage of times risk factor appears in 1,000 bootstrap models. Hosmer-Lemeshow goodness-of-fit test χ2 = 0.51.
were used with bootstrap bagging to identify the predictors of mortality. Statistical analysis was performed through Number Crunching Statistical Software (NCSS 2007; NCSS, Kaysville, UT) and SAS version 9.3 (SAS Institute, Cary, NC).
RESULTS Baseline Demographics and Characteristics of Patients With Severe Sepsis Over the 5-year period, 1,095 patients with severe sepsis were enrolled into the Registry, 165 (15%) of whom were positive for HIV, with a median CD4 count of 41 (8–167) and median viral load of 4.8 (2.1–5.1) log copies/mL. HIV-infected patients were predominantly younger (45 ± 9.8 yr vs 54 ± 17.4 yr; p < 0.001), with lower body mass index (24.2 ± 6.4 vs 28.9 ± 9.9; p < 0.001). The majority of patients in both groups were African-American (Table 3). HIV-infected patients were more likely to be on chronic dialysis (5% vs 2%; p = 0.02), have a history of pneumonia (45% vs 12%; p < 0.001), history of previous infection requiring hospital admission (67% vs 10%; p < 0.001), and a history of sepsis (14% vs 3%; p = 0.03). Non–HIV-infected patients were more likely to have coronary artery disease (8% vs 0%; p < 0.001) and diabetes (29% vs 9%; p < 0.001) compared with the HIV-infected patients. There were no differences in the prevalence of chronic obstructive pulmonary disease, renal failure, and cancer between the
groups. Although not statistically significant, HIV-infected patients had greater illness severity at admission to the ICU as determined by APACHE II score (26.3 ± 7.9 vs 24.5 ± 7.7; p = 0.08). HIV-infected patients also had a statistically significant higher A-a gradient (308.0 ± 178.8 vs 379.8 ± 184.1; p = 0.007). There were no differences in baseline SOFA scores between the two groups. There were no differences in the prevalence of shock requiring vasopressors and mechanical ventilation with acute lung injury/ARDS between HIVinfected and non–HIV-infected patients. However, there was a trend toward greater prevalence of acute renal failure (15% vs 11%; p = 0.14) and coagulopathy (16% vs 11%; p = 0.07) in the HIV-infected patients. Outcomes Among Patients With Severe Sepsis HIV-infected patients had a greater hospital mortality (50% vs 38%; p < 0.01) and a shorter ICU length of stay (LOS) (13.7 ± 12.0 vs 16.8 ± 18.6; p = 0.04) compared with HIVuninfected patients (Table 4). There was no significant difference in ICU mortality, duration of mechanical ventilation, or hospital LOS. Among survivors, HIV-infected patients still had a shorter ICU LOS (13.1 ± 12.3 d vs 17.4 ± 18.1 d; p = 0.04), and there remained no differences in duration of mechanical ventilation or hospital LOS. Univariate logistic regression analysis for hospital mortality, which included variables that significantly increased in HIV versus non-HIV,
Table 2. Multivariable Logistic Regression Model for Predictors of Hospital Mortality With Acute Physiology and Chronic Health Evaluation II (n = 355) OR
95% CI
p
Acute Physiology and Chronic Health Evaluation II (per 1-unit increase)
1.08
1.05–1.12
< 0.0001
100
History of pneumonia
0.60
0.23–1.60
0.30
17.6
History of hospital-acquired infection
1.03
0.39–2.73
0.95
5.9
History of sepsis
0.59
0.19–1.77
0.34
28.7
HIV
1.96
1.03–3.75
0.04
44.9
Variable
Reliability %a
OR = odds ratio. a Percentage of times risk factor appears in 1,000 bootstrap models. Hosmer-Lemeshow goodness-of-fit test χ2 = 0.72.
1640
www.ccmjournal.org
August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
Table 3.
Baseline Demographics and Characteristics of All Patients With Severe Sepsis
Variable
N
HIV (–)
HIV (+)
932
165
Age (yr), mean ± sd
54 ± 17.4
Male, n (%)
578 (62)
Body mass index (kg/m2), mean ± sd
28.9 ± 9.9
45 ± 9.8
p
< 0.001
105 (64) 24.2 ± 6.4
0.69 < 0.001
Race Caucasian
197 (21)
12 (7)
African-American
670 (72)
144 (87)
Hispanic
38 (4)
5 (3)
Other
27 (3)
4 (3)
0.002
Baseline comorbidities, n (%) Coronary artery disease
20 (8)
0
< 0.001
Chronic obstructive pulmonary disease
56 (12)
13 (8)
0.16
Renal failure
66 (14)
23 (14)
0.92
Chronic dialysis
17 (2)
8 (5)
0.02
Cirrhosis
39 (8)
7 (4)
0.08
Diabetes
137 (29)
14 (9)
< 0.001
History of pneumonia
32 (12)
71 (45)
< 0.001
History of infection requiring hospital admission
63 (25)
94 (64)
< 0.001
History of sepsis
19 (7)
22 (14)
0.03
History of cancer
59 (13)
19 (12)
0.60
Shock requiring vasopressors
532 (57)
100 (61)
0.38
Mechanical ventilation with acute lung injury/acute respiratory distress syndrome
262 (28)
45 (27)
0.82
99 (11)
24 (15)
0.14
101 (11)
26 (16)
0.07
Sequential Organ Failure Assessment score at ICU admission,a mean ± sd
10.1 ± 3.6
10.3 ± 3.7
0.63
Acute Physiology and Chronic Health Evaluation II score at ICU admission,a mean ± sd
24.5 ± 7.7
26.3 ± 7.9
0.08
1.3 ± 8.5
1.7 ± 0.8
0.66
308.0 ± 178.8
379.8 ± 184.1
0.007
Organ dysfunction, n (%)
Acute renal failure Coagulopathy
Albumin (mg/dL) at ICU admission,a mean ± sd A-a gradient at ICU admission,a mean ± sd CD4 count, median (IQR)
41 (8–167)
Viral load (Log copies/mL), median (IQR) a
4.8 (2.1–5.1)
IQR = interquartile range. a n is different than stated above, changes are noted as follows: Sequential Organ Failure Assessment and Acute Physiology and Chronic Health Evaluation: HIV (–) n = 285, HIV (+) n = 76; albumin: HIV (–) n = 286, HIV (+) n = 72; A-a gradient: HIV (–) n = 232, HIV (+) n = 57; and viral load: HIV (+) n = 71.
showed that HIV infection was associated with an OR of 1.58 (95% CI, 1.13–2.20) (Table 5). Two multivariate logistic regression analyses were performed to test the independent association of HIV with mortality. Adjusting for the clinical covariates age, history of pneumonia, history of Critical Care Medicine
hospital-acquired infection, and history of sepsis, HIV infection was associated with an OR of 1.78 (95% CI, 1.19–2.67) for hospital mortality (p = 0.005) (Table 1). In the clinical model including APACHE II, history of pneumonia, history of hospital-acquired infection, and history of sepsis entered www.ccmjournal.org
1641
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
Table 4.
Outcomes in Patients With Severe Sepsis
Variable
p
HIV (–)
HIV (+)
932
165
Duration of mechanical ventilation (d), mean ± sd
14.7 ± 17.0
12.9 ± 11.9
0.26
Hospital LOS (d), mean ± sd
30.5 ± 31.4
29.7 ± 26.0
0.77
ICU LOS (d), mean ± sd
16.8 ± 18.6
13.7 ± 12.0
0.04
Home
302 (32)
52 (32)
0.03
Rehab
62 (7)
5 (3)
128 (14)
15 (9)
80 (9)
11 (7)
ICU mortality
295 (32)
58 (35)
0.38
Hospital mortality, n (%)
358 (38)
82 (50)
0.007
n
Discharge status, n (%)
Skilled nursing Other healthcare
LOS = length of stay.
as potential confounders into the model, HIV infection was associated with an OR of 1.96 (95% CI, 1.03–3.75) for increased hospital mortality (p = 0.04) (Table 1). Characteristics and Outcomes Among Severe Sepsis Patients With HIV As shown in Table 6, 22% of HIV-infected patients were on HAART before admission, 15% of patients had HAART initiated during admission, and 63% of patients were never on HAART. There were 19 patients with a new diagnosis of HIV. Stratifying by HAART status at admission, mortality rates were 51%, 39%, and 50% for those patients who were on HAART at admission (n = 35), those who had HAART started at admission (n = 23), and those who never received HAART, respectively (n = 96) (Table 6). Given that HAART started at admission may have limited effect, we compared mortality rates between individuals on HAART at admission versus those who were not on HAART and did not find a statistically significant difference (51% vs 47%; p = 0.62). Stratifying by CD4 count, 80% of HIV-infected patients had a CD4 count less than 200 and 65% had a CD4 count less than 100. There Table 5.
was no significant difference in mortality rates between HIVinfected patients with a CD4 count less than 100 versus greater than 100 (52% vs 46%; p = 0.45). There was also no significant difference in mortality rates between HIV-infected patients with a CD4 count less than 200 versus greater than 200 (50% vs 47%; p = 0.73). Primary Infection and Microbiology The primary infection source was known for 162 HIV-infected patients and for 870 HIV-uninfected patients, and specific microbiology was known for 91 HIV-infected patients and 180 HIV-uninfected patients. For both HIV-infected and HIV-uninfected patients, the primary source of infection was respiratory (61% vs 48%, respectively; p = 0.002). Among 91 HIV-infected and 343 HIV-uninfected subjects, a Gram-positive source was the most common (39% vs 53%, respectively; p = 0.01). HIV-infected patients had a greater percentage of fungal infections compared with HIV-uninfected patients (33% vs 4%; p < 0.0001). In HIV-infected patients, 79% of all fungal infections were due to PCP, 10% were due to Candida, and 10% were due to Cryptococcus.
Univariate Logistic Regression Model for Predictors of Hospital Mortality n
OR
95% CI
1,095
1.01
1.01–1.02
0.0007
355
1.08
1.05–1.12
< 0.0001
History of pneumonia
1,095
1.28
0.85–1.92
0.24
History of hospital-acquired infection
1,095
1.31
0.93–1.84
0.12
History of sepsis
1,095
0.86
0.45–1.63
0.63
HIV
1,095
1.58
1.13–2.20
0.007
Variable
Age (per 1-yr increase) Acute Physiology and Chronic Health Evaluation II (per 1-unit increase)
p
OR = odds ratio.
1642
www.ccmjournal.org
August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations
Characteristics and Outcomes of Severe Sepsis Patients With HIV Table 6.
n (%)
Median CD4 Count (IQR)
Mortality, n (%)
On HAART at admission
35 (22)
121 (25–244)
18 (51)
HAART started this admission
23 (15)
16 (6–111)
9 (39)
Not on HAART
96 (63)
37 (6–152)
48 (50)
Variable
HAART characteristics
CD4 count (within 3 mo of admission) CD4 < 100
106 (65)
55 (52)
CD4: 100–200
24 (15)
11 (46)
CD4 > 200
34 (21)
16 (47)
IQR = interquartile range, HAART = highly active antiretroviral therapy.
DISCUSSION The results from this study conducted in a large urban safetynet hospital showed that HIV-infected patients with severe sepsis have higher hospital mortality when compared with HIV-uninfected patients with severe sepsis. This finding was robust after adjustment for cofounders, with HIV infection still associated with almost two-fold greater odds of hospital mortality. In addition, HIV-infected patients also had a shorter ICU LOS compared with HIV-uninfected patients. This was not due to difference in ICU mortality, which was not significantly different between the two groups. However, despite a shorter ICU LOS, there were no significant differences in hospital LOS, suggesting that HIV-infected patients still remained in the hospital, perhaps due to secondary conditions that required additional hospital care prior to discharge. Among the HIV-infected patients, there was also a greater prevalence of infectious baseline comorbidities, including chronic dialysis, a history of pneumonia, a history of infection requiring ICU admission, and a history of sepsis. Further, in our cohort, there were a large number of patients who were never on HAART. This could also reflect the large percentage of HIV-infected subjects with a CD4 count less than 100. These results suggest that significant healthcare disparities still exist in large urban areas of the United States where there are many HIV-infected individuals with limited access to care and who are unable to take HAART, potentially resulting in increased morbidity and mortality. More recent studies have shown substantial differences in ICU and hospital mortality rates among HIV-infected patients with the widespread use of HAART compared to the pre-HAART era. In developed countries, survival for critically ill HIV-infected patients has continued to improve in the HAART era (2, 5, 17, 21). Some have even noted that survival Critical Care Medicine
for HIV-infected patients in the ICU was comparable to general medical patients (22). Similar to our study, Akgün et al (6) showed that HIV-infected veterans experienced a higher 30-day mortality compared with HIV-uninfected patients. This study used the Veterans Aging Cohort Study (VACS) Risk Index (VACS Index) score, which included measures of HIV-specific and general organ injury developed to predict all-cause mortality in HIV-infected patients (21). However, unlike our study, many of the subjects were on HAART. Comparatively, many recent studies in underdeveloped countries have shown higher mortality rates ranging from 45% to 80% (11, 23), where AIDS-related diseases are more common and access to HAART is difficult. Mortality rates among HIVinfected patients in our institution mirror the rates in the pre-HAART era at other institutions. Independent predictors of mortality have been poor functional status, more than 1 year between HIV diagnosis and ICU admission, age, higher illness severity, low serum albumin, mechanical ventilation, and respiratory failure (17, 24, 25). This is likely due to a large number of HIV-infected individuals in the Atlanta area who lack access to medical care and health insurance. The Ponce de Leon Center in Atlanta, GA, is one of the largest facilities dedicated to the treatment of HIV/AIDS in the United States. Many of the patients who are admitted to Grady Hospital are taken care in this facility where it is known that approximately 40% of all patients are uninsured. These data may reflect the low percentage (22%) of HIV-infected patients with severe sepsis who were on HAART at admission to the hospital. Studies have shown that critically ill uninsured patients are more likely to be admitted to an ICU when hospitalized and more likely to die in the ICU (26). Delay of care can also result in greater burden of comorbid illness and higher severity of illness, which has been noted in our cohort. The impact of HAART on ICU outcomes is still unclear. Many studies have noted that HAART use in the ICU was not associated with survival (5, 22, 25), but associated with increased CD4 counts. HAART administration in the ICU is a challenge for a number of reasons. Difficulties with medication delivery, drug interactions, and toxicities, in addition to concerns for viral resistance, pose challenges to all healthcare providers in the ICU. In addition, lack of screening, resources for HAART initiation, and maintenance may also play a role. In our study, the current ICU admission resulted in a new diagnosis of HIV in approximately 14% of the HIV-infected patients. Other studies have shown that even though HAART use specifically was not associated with survival, survival rates have improved in the HAART era (17). In our cohort, HIV-infected patients had a high mortality regardless of HAART. This is likely because approximately 80% of our cohort had a CD4 count less than 200 and, further, approximately 64% of patients in our cohort had a CD4 count less than 100. This suggests that, regardless of HAART, many of these patients had advanced AIDS, which contributed to their poor outcomes. However, given the small subgroups, this cohort is insufficient to test HAART initiation in the ICU. www.ccmjournal.org
1643
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Cribbs et al
As in prior studies, respiratory infection was the most common etiology of sepsis in our HIV-infected patients (4, 5, 11, 12). Although PCP has become less frequent in many areas of the world, bacterial pneumonia has become increasingly more common (2, 4, 5, 14, 27). In our ICU patient population, infection was predominantly due to a Gram-positive source in both HIV-infected and HIV-uninfected patients. However, fungal causes of infection, mostly secondary to PCP, were much higher in the HIV-infected population. Overall, the prevalence of PCP has declined with the introduction of HAART and trimethoprim-sulfamethoxazole prophylaxis and treatment (28); however, we showed that in our large urban safety-net hospital, that PCP infections continue to occur at a high rate. PCP has been noted in other studies to occur mostly among persons unaware of their HIV infection, those who fail to seek medical care, and those who fail to adhere or respond to HAART (26, 29). Our study has several limitations. First, it was a secondary analysis of an ongoing prospective cohort study. Limited by the available data collected in this study, we were unable to determine certain aspects of ICU care that may have impacted patient outcomes. In this complete case analysis, we have presented two regression models, with APACHE II score substituted for age in the second model. Although APACHE II is a clinically significant variable, including this variable in the final multivariate model limits the sample size, resulting in a model with lower reliability. However, excluding this important variable creates clinical bias. Therefore, both models have been included and, importantly, the ORs for HIV are quantitatively the same and clinically relevant in both models. Second, our sample size of HIV-infected patients was moderate even though our overall severe sepsis sample size was large. Third, larger numbers of enrollment are needed in order to determine the impact of HAART in our ICU as the subgroups of patients initiated and not on HAART were small. In addition, given the retrospective nature of the study, information on viral loads, duration of HIV illness, and HAART are unknown on a number of patients. Therefore, possibility of HIV resistance to HAART is also unknown and could be a significant bias. Furthermore, this is a single-center study, subject to local ICU admission criteria and ICU practices, which may not necessarily be generalized. However, data on an urban inner city hospital in the current HAART era are lacking and therefore informative of the nature of critically ill HIV-infected patients. Finally, we are not able to draw conclusions about posthospital outcomes, including 60- and 90-day mortality, even though deaths may continue to accrue differentially after discharge, and this outcome is potentially more important to patients, providers, and healthcare organizations.
CONCLUSIONS Our study shows that HIV-infected patients with severe sepsis are still at risk of poor outcomes compared with HIV-uninfected patients in a large urban hospital setting composed of uninsured patients who lack access to medical care. Further studies need to be done to investigate the effect of socioeconomic status and mitigate healthcare disparities among critically ill HIV-infected patients. 1644
www.ccmjournal.org
ACKNOWLEDGMENTS We thank Celeste Sarmiento, MD, and the Grady Research Team for their assistance with patient screening, enrollment, and database management and Laura Ward, MSPH, for her review of the statistical methods.
REFERENCES
1. Buchacz K, Baker RK, Moorman AC, et al; HIV Outpatient Study (HOPS) Investigators: Rates of hospitalizations and associated diagnoses in a large multisite cohort of HIV patients in the United States, 1994-2005. AIDS 2008; 22:1345–1354 2. Narasimhan M, Posner AJ, DePalo VA, et al: Intensive care in patients with HIV infection in the era of highly active antiretroviral therapy. Chest 2004; 125:1800–1804 3. Nickas G, Wachter RM: Outcomes of intensive care for patients with human immunodeficiency virus infection. Arch Intern Med 2000; 160:541–547 4. Morris A, Creasman J, Turner J, et al: Intensive care of human immunodeficiency virus-infected patients during the era of highly active antiretroviral therapy. Am J Respir Crit Care Med 2002; 166:262–267 5. Powell K, Davis JL, Morris AM, et al: Survival for patients with HIV admitted to the ICU continues to improve in the current era of combination antiretroviral therapy. Chest 2009; 135:11–17 6. Akgün KM, Tate JP, Pisani M, et al: Medical ICU admission diagnoses and outcomes in human immunodeficiency virus-infected and virusuninfected veterans in the combination antiretroviral era. Crit Care Med 2013; 41:1458–1467 7. Coquet I, Pavie J, Palmer P, et al: Survival trends in critically ill HIVinfected patients in the highly active antiretroviral therapy era. Crit Care 2010; 14:R107 8. Croda J, Croda MG, Neves A, et al: Benefit of antiretroviral therapy on survival of human immunodeficiency virus-infected patients admitted to an intensive care unit. Crit Care Med 2009; 37:1605–1611 9. Huang L, Quartin A, Jones D, et al: Intensive care of patients with HIV infection. N Engl J Med 2006; 355:173–181 10. Medrano J, Alvaro-Meca A, Boyer A, et al: Mortality of patients infected with HIV in the intensive care unit (2005-2010): Significant role of chronic hepatitis C and severe sepsis. Crit Care 2014; 18:475 11. Japiassú AM, Amâncio RT, Mesquita EC, et al: Sepsis is a major determinant of outcome in critically ill HIV/AIDS patients. Crit Care 2010; 14:R152 12. Casalino E, Wolff M, Ravaud P, et al: Impact of HAART advent on admission patterns and survival in HIV-infected patients admitted to an intensive care unit. AIDS 2004; 18:1429–1433 13. Rosen MJ, Clayton K, Schneider RF, et al: Intensive care of patients with HIV infection: Utilization, critical illnesses, and outcomes. Pulmonary Complications of HIV Infection Study Group. Am J Respir Crit Care Med 1997; 155:67–71 14. Khouli H, Afrasiabi A, Shibli M, et al: Outcome of critically ill human immunodeficiency virus-infected patients in the era of highly active antiretroviral therapy. J Intensive Care Med 2005; 20:327–333 15. Afessa B, Green B: Clinical course, prognostic factors, and outcome prediction for HIV patients in the ICU. The PIP (Pulmonary complications, ICU support, and prognostic factors in hospitalized patients with HIV) study. Chest 2000; 118:138–145 16. Casalino E, Mendoza-Sassi G, Wolff M, et al: Predictors of short- and long-term survival in HIV-infected patients admitted to the ICU. Chest 1998; 113:421–429 17. van Lelyveld SF, Wind CM, Mudrikova T, et al: Short- and long-term outcome of HIV-infected patients admitted to the intensive care unit. Eur J Clin Microbiol Infect Dis 2011; 30:1085–1093 18. Giordano TP, Gifford AL, White AC Jr, et al: Retention in care: A challenge to survival with HIV infection. Clin Infect Dis 2007; 44:1493–1499 19. Hall HI, Holtgrave DR, Maulsby C: HIV transmission rates from persons living with HIV who are aware and unaware of their infection. AIDS 2012; 26:893–896 20. Bone RC, Balk RA, Cerra FB, et al: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. August 2015 • Volume 43 • Number 8
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
Clinical Investigations The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 1992; 101:1644–1655 21. Justice AC, McGinnis KA, Skanderson M, et al; VACS Project Team: Towards a combined prognostic index for survival in HIV infection: The role of ‘non-HIV’ biomarkers. HIV Med 2010; 11:143–151 22. Dickson SJ, Batson S, Copas AJ, et al: Survival of HIV-infected patients in the intensive care unit in the era of highly active antiretroviral therapy. Thorax 2007; 62:964–968 23. Amâncio FF, Lambertucci JR, Cota GF, et al: Predictors of the shortand long-term survival of HIV-infected patients admitted to a Brazilian intensive care unit. Int J STD AIDS 2012; 23:692–697 24. Akgün KM, Pisani M, Crothers K: The changing epidemiology of HIVinfected patients in the intensive care unit. J Intensive Care Med 2011; 26:151–164
Critical Care Medicine
25. Greenberg JA, Lennox JL, Martin GS: Outcomes for critically ill patients with HIV and severe sepsis in the era of highly active antiretroviral therapy. J Crit Care 2012; 27:51–57 26. Danis M, Linde-Zwirble WT, Astor A, et al: How does lack of insurance affect use of intensive care? A population-based study. Crit Care Med 2006; 34:2043–2048 27. Vincent B, Timsit JF, Auburtin M, et al: Characteristics and outcomes of HIV-infected patients in the ICU: Impact of the highly active antiretroviral treatment era. Intensive Care Med 2004; 30:859–866 28. Weverling GJ, Mocroft A, Ledergerber B, et al: Discontinuation of Pneumocystis carinii pneumonia prophylaxis after start of highly active antiretroviral therapy in HIV-1 infection. EuroSIDA Study Group. Lancet 1999; 353:1293–1298 29. Restrepo MI, Mortensen EM, Rello J, et al: Late admission to the ICU in patients with community-acquired pneumonia is associated with higher mortality. Chest 2010; 137:552–557
www.ccmjournal.org
1645
Copyright © 2015 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
