Current Epidemiology of Septic Shock

Current Epidemiology of Septic Shock The CUB-Reá Network Djillali Annane, Philippe Aegerter, Marie Claude Jars-Guincestre, and Bertrand Guidet for the CUB-Reá Network ˆ pital Raymond Poincare´, Assistance Publique Ho ˆ pitaux de Paris, Garches; Service de Biostatistique Service de Reánimation Me´dicale, Ho et Informatique Me´dicale, Ho ˆ pital Ambroise Pare´, Assistance Publique Ho ˆ pitaux de Paris, Boulogne; Service de Reánimation Me´dicale, ˆ pital Saint Antoine, Assistance Publique Ho ˆ pitaux de Paris, Paris; and INSERM U444, Paris, France Ho

To update the epidemiology of septic shock we analyzed clinical, microbiologic, and outcome variables from 100,554 intensive care unit admissions on the Colle`ge des Utilisateurs de Bases de donneés en Reánimation (CUB-Reá) database, collected from 22 hospitals over a 8-year period, 1993 to 2000. The overall frequency of septic shock was 8.2 per 100 admissions (i.e., 8,251 stays). It increased from 7.0 (in 1993) to 9.7 per 100 admissions (in 2000). The distribution analysis of the sites of infection and of the types of pathogens showed an increase in the rate of pulmonary infection (p ⫽ 0.001) and of multiresistant bacteria-related septic shock (p ⫽ 0.001). The crude mortality was 60.1% and declined from 62.1% (in 1993) to 55.9 (in 2000) (p ⫽ 0.001). As compared with matched intensive care unit admissions without sepsis, the excess risk of death due to septic shock was 25.7 (95% confidence interval, 24.0–27.3) and the matched odds ratio of death was 3.9 (95% confidence interval, 3.5–4.3). The frequency of septic shock is increasing with more multiresistant strains. Its crude mortality rate is decreasing, but patients with septic shock still have a high excess risk of death than critically ill patients who are nonseptic. Keywords: sepsis; epidemiologic studies; case–control studies; intensive care unit; outcome

Sepsis is present when there is clinical suspicion of infection, combined with evidence of systemic inflammation, usually defined by two or more of the following criteria: fever or hypothermia, leukocytosis or leukopenia, tachycardia, and tachypnea (1, 2). Septic shock has sepsis combined with hypotension that is refractory to fluid replacement. Most recent epidemiologic data on septic shock come from surveys conducted in the end of the eighties or the beginning of the nineties or are extrapolated from control groups of randomized trials. In the United States, in 1993, the entity “septicemia” represented the 13th leading cause of death, with an age-adjusted death rate of 7.9 per 100,000 population (3). Some authorities estimate that, in 1993, the hospital incidence of severe sepsis was two cases per 100 admissions with patients in intensive care units (ICUs) accounting for 59% of cases and that septic shock was present at onset of sepsis syndrome in 25% of patients (4). The most recent data were

obtained from 1995 state hospital discharge records from seven states. This large observational cohort study provided national estimates of 751,000 cases of severe sepsis, i.e., 3.0 cases per 1,000 population and 2.26 cases per 100 hospital discharges (5). The overall hospital mortality was 28.6%. The authors estimated that, in 1995, severe sepsis accounted for 9.3% of all deaths in the United States. In France, in 1993, the mean attack rate of severe sepsis was estimated at 6.3 per 100 ICU admissions and the mortality rate at 56% (6). A number of factors may be associated with mortality in severe sepsis and septic shock, e.g., age, severity of patient’s underlying disease, number of organ system dysfunction, severity of illness, hypothermia, thrombocytopenia, lactic acidosis, multiple sources of infection, positive blood stream, type of organism (6), and endocrine function (7). At the dawn of the third millennium, apart from early appropriate antibiotic therapy and source control (8), numerous efforts are still needed to develop specific treatment for septic shock (9). Recent Phase III trials suggested survival benefit from human recombinant activated protein C in patients with severe sepsis and at least one organ dysfunction (10) and from low-dose steroids in patients with septic shock with presumed adrenal dysfunction (11). Several characteristics of septic shock like the site of infection and the type of organisms have significantly changed over time (12). As these factors determine the antibiotic therapy that must be promptly initiated, there is an urgent need to update clinical and microbiologic characteristics of septic shock. Therefore, we analyzed (1) the annual frequency of septic shock treated in adult ICUs from 1993 to 2000, (2) clinical and microbiologic patterns, (3) interventions, (4) attributable mortality risk, and (5 ) factors affecting ICU mortality. METHODS The Database The database included data from 35 ICUs in Paris and its suburb and has been described elsewhere (13, 14) (Appendix 1). Data were extracted for the 100,554 admissions to the 22 ICUs participating in the database since 1993. According to French regulation, the CUB-Reá project was approved by the Comite´ National Informatique et Liberte´.

(Received in original form January 31, 2002; accepted in final form May 7, 2003)

Data

Supported by Assistance Publique—Ho ˆ pitaux de Paris, Paris, France (to CUB-Reá database).

Data were extracted for “septicemia,” “sepsis,” “severe sepsis” and “septic shock” (1, 2), for “cardiogenic,” “hemorrhagic,” “obstructive,” and “neurogenic shock,” and for demographic characteristics, dates of ICU admission and discharge, category (medical, scheduled or unscheduled surgical) and type of admission (community, hospital ward, or institution), McCabe class (0 ⫽ nonfatal, 1 ⫽ ultimately fatal or 2 ⫽ rapidly fatal disease) (15), immune status (immune deficiency included human immunodeficiency virus–infection, ongoing malignancy, radiation or chemotherapy, high dose or chronic use of corticosteroids, immune-suppressive drugs), and severity of illness (Simplified Acute Physiologic Score [SAPS] II) (16). For stays before 1996, SAPS II values were obtained from conversion of SAPS I values. Data were

Correspondence and requests for reprints should be addressed to Djillali Annane, M.D., Ph.D., Service de Reánimation Me´dicale, Ho ˆ pital Raymond Poincare´, Faculte´ de Me´decine Paris Ile de France Ouest, Universite´ de Versaille Saint Quentinen-Yvelinnes, 104, Boulevard Raymond Poincare´, 92380 Garches, France. E-mail: [email protected] This article has an online supplement, which is accessible from this issue’s table of contents online at www.atsjournals.org Am J Respir Crit Care Med Vol 168. pp 165–172, 2003 DOI: 10.1164/rccm.2201087 Internet address: www.atsjournals.org

166

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 168 2003

also extracted for “site of infection” and “type of microorganisms,” and from the Omega workload scoring system (14, 17) for “interventions.” The ICU length of stay was calculated using the number of calendar days. Hospital mortality rates were available from 1997 and readmission rates were available from 1999 for all units. To evaluate the excess mortality, a matched study was conducted on a set of septic shock stays versus stays (1 ) without sepsis and (2 ) with nonseptic shock (see additional details for selection of control patients in the online supplement).

Analyses Unpaired t tests and ␹2 statistics were used for comparisons of continuous and nominal variables, respectively. The changes from 1993 to 2000 for relevant variables were analyzed by analysis of variance with the contrasts method and by Pearson ␹2 with the Cochran–Armitage trend test for continuous and nominal variables, respectively. Prognostic variables related to death in univariate analysis with a p value less than 0.2 were introduced in a stepwise multivariate logistic regression model to estimate the odds ratio (OR) of death in the ICU (additional details for the modeling procedure are given in the online supplement). Goodness-of-fit of the logistic model was assessed using the Hosmer– Lemeshow test, whereas the discriminating ability of the model was evaluated by the c-index (18). Adjusted mortality rates and their confidence intervals were calculated (19). In the matched study, comparisons of paired baseline characteristics were performed using the paired Student t test and the McNemar test for continuous and categoric variables, respectively. The matched risk ratio and excess risk were expressed by ratio and difference of mortality rates between exposed and unexposed stays, respectively, and matched OR by the ratio of the frequencies of discordant pairs. The OR of death related to septic shock was adjusted on other prognostic variables by means of a conditional logistic regression model to take into account the matching induced correlations between exposed and unexposed stays. A p value less than 0.01 was considered significant in all multivariate analyses. Analyses were performed with the SAS statistical software (SAS Institute, Cary, NC) and the S-Plus package (Mathsoft, Seattle, WA).

RESULTS Frequency of Septic Shock

In the 22 ICUs population (n ⫽ 100,554), the overall frequency of septic shock was 8.2 per 100 ICU admissions (n ⫽ 8,251). Its annual frequency significantly (p ⬍ 0.001) increased from 7.0 in 1993 to 9.7 per 100 ICU admissions in 2000 (Figure 1). When considering, only first admission similar rates were observed, e.g., 8.8 and 9.7 per 100 ICU admissions in 1999 and 2000, respectively.

Clinical and Microbiologic Patterns

Demographic characteristics. Table 1 compares the demographic characteristics of the 8,251 ICU stays with septic shock with those of the population without septic shock in the same 22 ICUs. The septic shock subset comprised older patients, a greater proportion of males, patients having underlying disease with presumably reduced life expectancy (Figure 2), or having immune depression, and a lower proportion of medical patients, or patients coming from the community. Finally, the septic shock subset represented more severely ill patients, i.e., higher SAPS II and higher crude mortality, than the general ICU population. Site of infection and type of pathogens. The lung was the primary source of infection, followed by the abdomen, and the urinary tract (see Table E1 in the online supplement). Although the proportion of lung infection–related septic shock significantly increased from 1993 to 2000 (p ⬍ 10⫺4), the percentage of septic shock due to urinary tract infection has significantly decreased (p ⬍ 10⫺4), and the proportion of abdominal sepsis remained rather unchanged. In about one fourth of the population, there are multiple sites of infection accounting for septic shock. The relative frequency of positive blood cultures in the septic shock population was about one third and increased with time. The proportion of septic shock from an unknown source of infection significantly decreased with time (p ⫽ 0.001) but still remained about one fifth. The proportion of septic shock with unidentified pathogen decreased with time (p ⬍ 10⫺4) but still remained about one third. The percentage of polymicrobial infection (p ⬍ 10⫺4) as well as the proportion of multiresistant bacteria, like Pseudomonas (p ⬍ 10⫺4) and methicillin-resistant Staphylococci (p ⬍ 10⫺4), has significantly increased with time. Interventions and Outcome

Interventions. Mechanical ventilation, vasopressor therapy, and renal replacement therapy were needed in 80.9, 85.4, and 25.2% of cases. The percentage of cases of septic shock in which a pulmonary artery catheter was inserted was about 42.2% on average but significantly decreased (p ⬍ 10⫺4), whereas the use of a vasopressor therapy increased (p ⬍ 10⫺4). Overall, except for respiratory support, which remained globally unchanged with time, the proportion of patients needing hemodynamic or renal support significantly increased (p ⬍ 10⫺4) (see Table E2 in the online supplement).

Figure 1. Display of the frequency and intensive care unit (ICU) mortality of admissions with septic shock in the subset of 22 units always present in the database since 1993. The frequency of septic shock significantly increased from 1995 to 1999 (p ⬍ 0.001), and ICU mortality significantly decreased (p ⫽ 0.0001).

Annane, Aegerter, Jars-Guincestre, et al.: Epidemiology of Septic Shock

167

TABLE 1. DEMOGRAPHIC CHARACTERISTICS OF 8,251 INTENSIVE CARE UNIT STAYS WITH SEPTIC SHOCK FROM 1993 TO 2000 Variables Baseline descriptors Age, yr Sex, female Admission category, medical Type of admission, direct Reason for admission, shock Reason for admission, infection McCabe ⬎ 0 Comorbidity Immune deficiency Diabetes mellitus Chronic renal failure Hematologic malignancy Cancer HIV-related disease Chronic pulmonary disease SAPS II ALI–ARDS Outcomes ICU stay, d median (quartiles) ICU crude death rate Hospital crude death rate† ICU readmission rate‡

Septic Shock (n ⫽ 8,251)

Population without Septic Shock (n ⫽ 92,293)

61.4 ⫾ 16.6 36.7 (35.6–37.7) 78.2 (77.3–79.1) 36.7 (35.7–37.7) 33.7 (32.7–34.8) 35.4 (34.3–36.4) 47.2 (46.1–48.2)

53.9 ⫾ 19.5 42.0 (41.6–42.3) 87.7 (87.5–87.9) 58.1 (57.8–58.4) 6.5 (6.3–6.6) 7.6 (7.5–7.8) 28.1 (27.8–28.4)

p p p p p p p

⬍ ⬍ ⬍ ⬍ ⬍ ⬍ ⬍

0.001 0.001 0.001 0.001 0.001 0.001 0.001

21.9 (21.0–22.7) 2.2 (1.9–2.5) 3.2 (2.9–3.6) 8.4 (7.8–9.0) 6.9 (6.3–7.4) 5.8 (5.3–6.3) 9.2 (8.6–9.8) 58.3 ⫾ 23.9 22.2 (21.3–23.1)

11.4 (11.2–11.6) 2.8 (2.7–2.9) 3.7 (3.5–3.8) 2.5 (2.4–2.6) 3.7 (3.6–3.8) 3.3 (3.1–3.4) 13.5 (13.3–13.7) 34.4 ⫾ 20.32 3.0 (2.9–3.1)

p p p p p p p p p

⬍ ⫽ ⫽ ⬍ ⬍ ⬍ ⬍ ⬍ ⬍

0.001 0.002 0.05 0.001 0.001 0.001 0.001 0.001 0.001

15.2 ⫾ 21.3 8 (3–19) 60.1 (59.0–61.1) 61.2 (59.8–62.7) 4.2 (3.4–5.0)

6.9 ⫾ 11.4 4 (2–7) 13.2 (13.0–13.4) 17.2 (16.8–17.5) 3.5 (3.2–3.7)

p ⬍ 0.001

p Value*

p ⬍ 0.001 p ⬍ 0.001 NS

Definition of abbreviations: ALI–ARDS ⫽ acute lung injury–acute respiratory distress syndrome; HIV ⫽ human immunodeficiency virus; ICU ⫽ intensive care unit; NS ⫽ not statistically significant at ␣ ⫽ 0.05; SAPS II ⫽ Simplified Acute Physiologic Score II. Values are given as mean ⫾ SD for continuous variables except stated otherwise and as percentage and 95% confidence interval for categoric variables. * Unpaired comparison between septic shock and all other stays without septic shock among the same 22 ICUs since 1993. † Hospital death rates calculated on 54,966 admissions in the years 1997–2000. ‡ Readmission rate calculated on 25,144 admissions in the years 1999–2000.

Outcomes

ICU mortality. The crude ICU mortality was on average 60.1% and significantly decreased since 1997 (p ⫽ 0.001) (see Table E2 in the online supplement). Table 2 displays the factors that independently affect ICU mortality. Among these, the need for mechanical ventilation was the factor associated with the highest OR of dying, i.e., 4.72 (95% confidence interval, 4.10–5.42). The use of vasopressors was also associated with an increased risk of death. Positive bacterial cultures (from any site) favorably impacted on survival (OR of dying 0.74; 95% confidence interval,

0.66–0.83), whereas fungus infections were associated with an increased risk of death (OR of dying 1.81; 95% confidence interval, 1.36–2.41). When taking into account these prognostic factors, the adjusted death rate was significantly lower between the 1993 to 1996 period and the next one (see Table E2 in the online supplement). ICU length of stay. The ICU length of stay was on average 15.2 days (median 8 days) and significantly increased since 1997 (p ⫽ 0.0001). However, this trend was mainly observed among nonsurvivors (see Table E2 in the online supplement).

Figure 2. Display of the frequency and the intensive care unit (ICU) mortality of admissions with septic shock according to age and McCabe score in the subset of 22 units always present in the database since 1993: 0 ⫽ nonfatal (open squares) versus more than 0 ⫽ rapidly or ultimately fatal (closed squares).

168

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 168 2003 TABLE 2. FACTORS THAT AFFECT INTENSIVE CARE UNIT SURVIVAL IN 8,251 INTENSIVE CARE UNIT STAYS WITH SEPTIC SHOCK OVER THE 8-YEAR PERIOD 1993–2000: MULTIVARIATE LOGISTIC REGRESSION ANALYSIS Variables Period, 1997–2000 Demographic characteristics Age, by 10 yr McCabe ⬎ 0 Admission category, medical Type of admission, direct Severity of illness SAPS II ln (1 ⫹ SAPS II) Interaction term: age ⫻ (McCabe ⬎ 0) ALI–ARDS Acute renal failure Characteristics of infection Unknown site of infection Positive bacterial strains Fungus Interventions Mechanical ventilation Vasopressors* Renal support

Odds Ratio

95% Confidence Interval

p Value

0.61

0.55–0.68

⬍ 10⫺4

1.23 4.01 1.75 0.84

1.18–1.29 2.70–5.94 1.53–2.01 0.75–0.95

⬍ ⬍ ⬍ ⬍

10⫺4 10⫺4 10⫺4 10⫺2

1.07 0.16 0.89 1.70 1.43

1.05–1.09 0.07–0.40 0.83–0.95 1.47–1.97 1.28–1.61

⬍ ⬍ ⬍ ⬍ ⬍

10⫺4 10⫺4 10⫺3 10⫺4 10⫺4

1.28 0.74 1.81

1.12–1.47 0.66–0.83 1.36–2.41

⬍ 10⫺3 ⬍ 10⫺4 ⬍ 10⫺4

4.72 1.52 2.03

4.10–5.42 1.28–21.80 1.77–2.32

⬍ 10⫺4 ⬍ 10⫺4 ⬍ 10⫺4

Definition of abbreviations: ALI–ARDS ⫽ acute lung injury–acute respiratory distress syndrome; ln ⫽ natural log, term introduced to improve calibration; SAPS II ⫽ simplified acute physiologic score II. Results of stepwise selection procedure and resampling validation. C-index ⫽ 0.82. * Vasopressors were considered when the patients received for at least 1 h dopamine of more than 5 ␮g/kg/min or any dose of norepinephrine, epinephrine, or phenylephrine.

Septic Shock–related Mortality

Matched comparison with ICU stays without sepsis. A total of 5,473 ICU admissions with septic shock could be matched to 5,473 ICU admissions without sepsis, according to all seven factors: unit, year, age (⫾ 10 years), sex, type of admission, admission category, and SAPS II (⫾ 2) (Table 3). The former represented 59% of the admissions with septic shock, and, as compared with the 2,778 admissions with septic shock who were not matched, were older (62.0 years vs. 60.3, p ⬍ 10⫺4), had more frequent medical (nonsurgical) stays (86.8% vs. 61.5, p ⬍ 10⫺4), and had lower SAPS II values (50.9 vs. 72.9, p ⬍ 10⫺4). The septic shock matched subset had a lower death rate (53.8% vs. 72.2, p ⬍ 10⫺4) and a longer mean length of stay (16.8 days vs. 12.0, p ⬍ 10⫺4) than the septic shock unmatched subset. Septic shock admissions were more likely to have a McCabe score above 0, or immune deficiency, to be on ventilator, on vasopressor therapy, on renal replacement therapy, or to have a right-heart catheter. Finally, as compared with the matched control patients, septic shock admissions had a higher daily workload score (mean Omega score) and mortality rate. Septic shock was associated with a matched excess risk of death in the ICU of 25.7 (95% confidence interval, 24.0–27.3). The matched risk ratio was 1.9 (95% confidence interval, 1.8–2.0), and the matched OR was 3.9 (95% confidence interval, 3.5–4.3). Septic shock remained significantly associated with death in the ICU (OR of dying of 2.0; 95% confidence interval, 1.8–2.3) after adjusting on other prognostic variables (acute respiratory distress syndrome [ARDS], acute renal failure, respiratory support, hemodynamic support, and renal replacement therapy) (see Table E3 in the online supplement). Matched comparison with ICU stays with nonseptic shock. To further refine the estimate of mortality related to septic shock, we compared 2,350 ICU stays with septic shock with 2,350 ICU stays with nonseptic shock (Table 4). The control patients were in cardiogenic, hemorrhagic, and neurogenic shock in 43, 33, and 17% of cases, respectively. Septic shock admissions were still

more likely to have immune deficiency and a need for mechanical ventilation, vasopressor therapy, renal replacement therapy, or a right-heart catheter. Finally, they had a higher mortality rate. Septic shock was associated with a matched excess risk of death in the ICU of 11.6 (95% confidence interval, 9.0–14.1). The matched risk ratio was 1.26 (95% confidence interval, 1.2–1.3), and the matched OR was 1.8 (95% confidence interval, 1.6–2.0). Septic shock remained significantly associated with death in the ICU with an OR of dying of 1.3 (95% confidence interval 1.1 to 1.5) after adjusting on other prognostic variables (ARDS, respiratory support, hemodynamic support, and renal replacement therapy) (see Table E4 in the online supplement).

DISCUSSION As compared with the latest published large study on epidemiology of severe sepis (5), the current study provided important additional information. First, this study focused on septic shock patients admitted to the ICU and used a common definition for septic shock (2). Second, this is the first large longitudinal study that covers almost all the nineties, whereas the study by Angus and coworkers was restricted to data from 1995. Finally, the current study provides more detailed information on demographic data and comorbidities, sites and types of infection, types of pathogens, and interventions. To avoid bias related to the addition of data from new ICUs, we restricted the analyses to the data from the 22 ICUs that contributed to the database since its initiation in 1993. This analysis of the CUB-Reá database showed a significant increase in the annual frequency of septic shock treated in the ICUs over the 8-year period 1993 to 2000. In 2000, the mean attack rate of septic shock was 9.7 per 100 ICU admissions. Analyses of data from the whole database (35 units) showed slightly lower annual frequencies, a similar trend with time, and a mean attack rate of septic shock for 2000 of 8.8 per 100 ICU admissions. The annual frequency of septic shock of 6.9 per 100 ICU admissions

Annane, Aegerter, Jars-Guincestre, et al.: Epidemiology of Septic Shock

169

TABLE 3. COMPARISON BETWEEN INTENSIVE CARE UNIT STAYS WITH SEPTIC SHOCK AND MATCHED INTENSIVE CARE UNIT STAYS WITHOUT SEPTIC SHOCK Variables Baseline descriptors Age, yr Sex, female Type of admission, direct Admission category, medical SAPS II McCabe ⬎ 0 Comorbidity Immunodeficiency Diabetes mellitus Chronic renal failure Hematologic malignancy Cancer HIV-related disease Chronic pulmonary disease ALI–ARDS Interventions Mechanical ventilation Time on ventilator, d† Vasopressors‡ Right-heart catheterization Renal support Mean daily Omega score Outcomes ICU length of stay, d†; mean ⫾ SD, median (quartile) ICU crude mortality rate


Control Patients (n ⫽ 5,473)

p Value*

61.9 ⫾ 15.6 33.4 37.7 86.8 50.9 ⫾ 19.5 45.4

62.1 ⫾ 15.5 33.4 37.7 86.8 50.9 ⫾ 19.5 43.2

NS

21.1 2.4 3.1 8.0 6.8 4.9 11.5 21.6

17.1 3.3 6.2 6.2 6.8 4.4 14.3 5.7

⬍ ⬍ ⬍ ⬍

76.9 15.0 ⫾ 20.6 83.5 37.0 23.5 18.1 ⫾ 9.6

61.7 8.9 ⫾ 13.6 43.2 11.9 12.9 14.8 ⫾ 9.5

⬍ ⬍ ⬍ ⬍ ⬍ ⬍

16.8 ⫾ 22; 2 9 (4–21) 53.8

10.1 ⫾ 15; 0 5 (3–11) 28.2

⬍ 10⫺4 ⬍ 10⫺4

NS ⬍ 10⫺2 10⫺4 10⫺2 10⫺4 10⫺4 NS NS ⬍ 10⫺4 ⬍ 10⫺4 10⫺4 10⫺4 10⫺4 10⫺4 10⫺4 10⫺4

Definition of abbreviations: ALI–ARDS ⫽ acute lung injury–acute respiratory distress syndrome; HIV ⫽ human immunodeficiency virus; ICU ⫽ intensive care unit; NS ⫽ nonstatistically significant at ␣ ⫽ 0.05; SAPS II ⫽ Simplified Acute Physiologic Score II. Continuous variables are expressed as mean ⫾ SD, and categoric variables as proportion. * Paired comparison by Student’s t test for continuous variables and McNemar test for categoric variables. † Comparison after logarithmic transformation of the variable. ‡ Vasopressors were considered when the patients received for at least 1 h dopamine of more than 5 ␮g/kg/min or any dose of norepinephrine, epinephrine, or phenylephrine.

observed in 1995 in our study is very consistent with the 10% incidence in the ICU of severe sepsis reported for the same year by Angus and coworkers (5), given that septic shock is a subset of severe sepsis. Similarly, the frequency of septic shock of 7.0 per 100 ICU admissions observed for 1993 is also consistent with a previous French cohort study, which reported an incidence of confirmed severe sepsis of 6.3 per 100 ICU admission (6). The mean SAPS II value of 56 and the crude hospital mortality of 59% observed suggested that the authors were reporting data for septic shock rather than severe sepsis. The observed increase in the frequency of septic shock with time might have been explained by change in ICU admission decisions with the less sick patients not being admitted anymore, as suggested by the increase in mean SAPS II values. However, the proportion of patients with comorbidities did not dramatically change with time, and even the proportion of patients with the McCabe score above 0 significantly decreased from 49.8 per 100 ICU admissions in 1993 to 42.4 in 2000. In addition, the size of units were constant (apart for 1999 when two units merged into a new one), and the number of stays increased. Thus, it seems very unlikely that the observed increase in septic shock frequency was related to changes in ICU admission decisions. This study confirms that, as compared with the general ICU population, the septic shock subset is made of older patients, a higher proportion of patients with previous diseases reducing life expectancy or having immune deficiency, and a higher proportion of surgical patients (5, 6, 20). This study also confirms that the lung is the primary source of infection, whereas the

number of urinary tract infection–related shocks has considerably decreased (5, 6, 12, 20, 21). In addition, the proportion of Pseudomonas and Staphylococcus resistant to methicillin-related septic shock has dramatically increased with time, to reach a worrying proportion of one fifth to one fourth of cases in 2000. This result is in keeping with previous findings in European ICUs (22) and might result from the high “antibiotic pressure” in ICUs (23). By contrast, septic shock crude mortality and adjusted mortality decreased during the 8-year period from 63 (in 1,993) to 58% (in 2000). This absolute 5% reduction in mortality represented almost 60 additional rescued patients a year. This drop in mortality might result from more aggressive diagnostic testing, as suggested by the reduction in the proportion of unknown source of infection, and by more aggressive therapies. In the last decade, the Socie´te´ de Reánimation de Langue Française has made tremendous efforts to standardize the care of the critically ill patients, notably to manage septic shock. Guidelines were established for fluid resuscitation, the use of catecholamines, and for renal replacement therapy (24–26). Whether these efforts explain the observed reduction in mortality with time remains speculative. Interestingly, after stratification on quartiles of SAPS II, we observed that the reduction in mortality mainly occurred in the two lower strata (SAPS II under 54), whereas in the upper fourth (SAPS II over 72) mortality remained constant about 90%. This study allowed the estimation that infection-related shock was associated with an excess risk of death of 25.7% when compared with matched critically ill patients without sepsis, and with

170

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 168 2003 TABLE 4. COMPARISON BETWEEN INTENSIVE CARE UNIT ADMISSIONS FOR SEPTIC SHOCK AND INTENSIVE CARE UNIT ADMISSIONS FOR NONSEPTIC SHOCK Variables Baseline descriptors Age, yr Sex, female Type of admission, direct Admission category, medical SAPS II McCabe ⬎ 0 Comorbidity Immunodeficiency Diabetes mellitus Chronic renal failure Hematologic malignancy Cancer HIV-related disease Chronic pulmonary disease ALI–ARDS Interventions Mechanical ventilation Time on ventilator, d† Vasopressors‡ Right-heart catheterization Renal support Mean daily omega score Outcomes ICU length of stay, d†; mean ⫾ SD median (quartile) ICU mortality rate


Control Patients (n ⫽ 2,350)

p Value*

64.6 ⫾ 13.3 30.1 39.3 89.9 54.1 ⫾ 20.0 45.3

64.9 ⫾ 13.4 30.1 39.3 89.9 54.1 ⫾ 20.1 46.4

⬍ 10⫺2

20.4 2.6 3.4 8.0 6.9 3.5 12.3 22.3

13.3 3.2 5.1 4.7 5.3 1.3 8.2 7.6

⬍ 10⫺4 NS ⬍ 10⫺2 ⬍ 10⫺4 ⬍ 0.05 ⬍ 10⫺4 ⬍ 10⫺4 ⬍ 10⫺4

79.1 15.0 ⫾ 20.2 84.8 36.7 25.1 18.4 ⫾ 9.4

69.1 8.6 ⫾ 13.4 75.9 25.7 16.4 18.2 ⫾ 11.3

16.8 ⫾ 21; 6 10 (4–22) 57.1

9.8 ⫾ 15; 1 5 (3–11) 45.4

NS NS

⬍ ⬍ ⬍ ⬍ ⬍

10⫺4 10⫺4 10⫺4 10⫺4 10⫺4 NS

⬍ 10⫺4 ⬍ 10⫺4

Definition of abbreviations: ALI–ARDS ⫽ acute lung injury–acute respiratory distress syndrome; HIV ⫽ human immunodeficiency virus; ICU ⫽ intensive care unit; NS ⫽ not statistically significant at ␣ ⫽ 0.05; SAPS II ⫽ Simplified Acute Physiologic Score II. Continuous variables are expressed as mean ⫾ SD, and categoric variables as proportion. * Paired comparison by Student’s t test for continuous variables and McNemar test for categoric variables. † Comparison after logarithmic transformation of the variable.

an excess risk of death of 11.6% when compared with patients with nonseptic shock. Given, the high frequency of multiresistant strains, this excess risk of death of patients with septic shock is in keeping with previous reports (27–30). As compared with the “general” ICU patients and with patients with nonseptic shock, patients with septic shock required more frequent mechanical ventilation, vasopressor therapy, renal replacement therapy, and heavier daily workload. In addition, the use of the life-supporting therapies increased with time in the treatment of septic shock, suggesting a more aggressive management. Notably, fewer pulmonary artery catheters were used with time, probably resulting from the impact of the worrisome report of increased risk of death associated with their use in critically ill patients (31) and the increasing use of noninvasive tools. Nonetheless, as compared with their use in matched nonseptic shock, pulmonary artery catheters are still more frequently used in septic shock. Study Limitations

First, we should acknowledge that 18 of the 22 ICUs belong to teaching hospitals and only two are surgical ICUs. Accordingly, the conclusions drawn from this study might not become generalized. Moreover, ICUs in nonteaching hospitals differed from those in teaching hospitals by a lower frequency of septic shock (5.8 vs. 8.9 per 100 admissions, p ⬍ 10⫺4) and fewer immunecompromised patients (14% vs. 24%, p ⬍ 10⫺4). Although SAPS II values were similar, the death rate was higher in ICUs in nonteaching hospitals (64% vs. 59%, p ⬍ 0.01). Surgical ICUs differed from medical ICUs by higher frequency of septic shock admissions (11.5 vs. 8.1 per 100 admissions, p ⬍ 10⫺4) and fewer

immune-compromised admissions (14% vs. 23%, p ⬍ 10⫺4). Otherwise, they had similar SAPS II values and death rates. As far as the matching procedures are concerned, the time to onset of septic shock was not available for all patients, preventing the selection of control patients on minimal duration of exposure. The reason for ICU admission was “shock” in about one third of septic shock stays and “infection” in about 35%. We must acknowledge that only one reason for admission was entered in the database and that there were no recommendation for coding this variable. Nonetheless, the introduction in multivariate analysis of the reason for admission (“shock” or “infection”) did not lower the strength of the relation between septic shock and death (data not shown). When we restricted the analysis to the pairs being stayed in the ICU at least two days, or even 10 days, the matched OR of death remains about 4 and over 3, respectively. In addition, in pairing, for lack of witnesses, one drew aside the most seriously ill patients and therefore the excess of risk may be underestimated. An interesting although costly alternative to the use of sampling and statistical methods is the appropriateness evaluation protocol, where experts conclude on explicit criteria whether each day of stay or even death for patients with septic shock is related to septic shock. In conclusion, septic shock is more and more frequent in the ICU. Its crude mortality rate remains about 60%, and its attributable mortality risk is about 25%. The physicians who care for patients with septic shock must take into account a high likelihood for the patient to have (1) comorbidity that affects life expectancy or immune suppression, (2) multiresistant bacteria involved, and (3 ) need for life-support therapies.

Annane, Aegerter, Jars-Guincestre, et al.: Epidemiology of Septic Shock References 1. Bone RC. Sepsis, the sepsis syndrome, multi-organ failure: a plea for comparable definitions. Ann Intern Med 1991;114:332–333. 2. Members of the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Crit Care Med 1992;20:864–874. 3. Center for Disease Control and Prevention. National Center for Health Statistics: annual summary of births, marriages, divorces and deaths, 1993. Mon Vital Stat Rep 1994;42:5–10. 4. Sands KE, Bates DW, Lanken PN, Graman PS, Hibberd PL, Kahn KL, Parsonnet J, Panzer R, Orav EJ, Snydman DR. Epidemiology of sepsis syndrome in 8 academic medical center. JAMA 1997;278:234–240. 5. Angus D, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001; 29:1303–1310. 6. Brun-Buisson C, Doyon F, Carlet J, Dellamonica P, Gouin F, Lepoutre A, Mercier JC, Offenstadt G, Regnier B for the French ICU Group for Severe Sepsis. Incidence, risk factors, and outcome of severe sepsis and septic shock in adults: a multicenter prospective study in intensive care units. JAMA 1995;274:968–974. 7. Annane D, Se´bille V, Troche´ G, Raphael JC, Gajdos P, Bellissant E. A three-level prognostic classification in septic shock based on cortisol levels and cortisol response to corticotropin. JAMA 2000;283:1038– 1045. 8. Weinstein MP, Reller LB, Murphy JR, Lichtenstein KA. The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II: clinical observations with special reference to factors influencing prognosis. Rev Infect Dis 1983;5:54–70. 9. Wheeler AP, Bernard GR. Treating patients with severe sepsis. N Engl J Med 1999;340:207–214. 10. Bernard GR, Vincent JL, Laterre PF, LaRosa SP, Dhainaut JF, LopezRodriguez A, Steingrub JS, Garber GE, Helterbrand JD, Ely EW, et al. Efficacy and safety of recombinant human activated protein C for severe sepsis. N Engl J Med 2001;344:699–709. 11. Annane D, Sebille V, Charpentier C, Bollaert PE, Francois B, Korach JM, Capellier G, Cohen Y, Azoulay E, Troche G, et al. Effect of treatment with low doses of hydrocortisone and fludrocortisone on mortality in patients with septic shock. JAMA 2002;288:862–871. 12. Friedman G, Silva E, Vincent JL. Has the mortality of septic shock changed with time? Crit Care Med 1998;26:2078–2086. 13. Aegerter P, Auvert B, Buonamico G, Sznajder M, Beauchet A, Guidet B, Legall JR, CUB-Reá. Mise en œuvre et e´valuation d’une base de donneés commune aux services de reánimation d’Ile-de-France. Rev Epide´miol Sante´ Publique 1998;46:226–237. 14. Sznajder M, Leleu G, Buonamico G, Auvert B, Aegerter P, Merliere Y, Dutheil M, Guidet B, LeGall JR. Estimation of direct cost and resource allocation in intensive care: correlation with Omega system. Intensive Care Med 1998;24:582–589. 15. McCabe WA, Jackson GG. Gram negative bacteriemia. Arch Intern Med 1962;110:847–855. 16. Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study. JAMA 1993;270:2957–2963. 17. Catalogue des Actes Me´dicaux. Champs Beta et Omega. Fascicule Spećial, Ministe`re des Affaires Sociales de la Sante´ et de la Ville, Bulletin Officiel. Paris, France: Ministry of Health; 1995 No. 95-4bis. 18. Harell Jr FE. Regression modelling strategies. New York: Springer Verlag; 2001. 19. Hosmer DW, Lemeshow S. Confidence interval estimates of an index of quality performance based on logistic regression models. Stat Med 1995;14:2161–2172. 20. Knaus WA, Harrell FE, Fischer CJ, Wagner DP, Opal SM, Sadoff JC, Draper EA, Walawander CA, Coboy K, Grasela TH. The clinical evaluation of new drugs for sepsis: a prospective study design based on survival analysis. JAMA 1993;270:1233–1241. 21. Pittet D, Thie´vent B, Wenzel RP, Li N, Auckenthaler R, Suter PM. Bedside prediction of mortality from bacteremic sepsis : a dynamic analysis of ICU patients. Am J Respir Crit Care Med 1996;153:684–693. 22. Vincent JL, Bihari D, Suter PM, Bruining HA, White J, Nicolas-Chanoin MH, Wolff M, Spencer RC, Hemmer M. The prevalence of nosocomial infection in intensive care units in Europe: the results of the EPIC study. JAMA 1995;274:639–644.

171 23. Wenzel RP, Edmond MB. Managing antibiotic resistance. N Engl J Med 2000;343:1961–1963. 24. Raphael JC, Antony I, Bellissant E, Bonmarchand G, Cotting J, Fierobe L, Gouyon JB, Hasselmann M, Leleu G, Misset B, et al. Utilisation des cathećolamines au cours du choc septique (adultes-enfants). Reá Urg 1996;5:441–450. 25. Recommandations pour la Pratique Clinique: remplissage vasculaire au cours des hypovole´mies relatives ou absolues. Reá Urg 1997;6:335–341. 26. Charbonneau P, Auriant I, Berger M, Calvat S, Fourrier F, Girardin G, Gobeaux RF, Lamer C, Lepape A, Leverve X, et al. Epuration extrareńale continue en reánimation (a` l’exclusion de la dialyse pe´ritoneále). Reá Urg 1998;7:174–182. 27. Leu HS, Kaiser DL, Mori M, Woolson RF, Wenzel RP. Hospital-acquired pneumonia attributable mortality and morbidity. Am J Epidemiol 1989; 129:1258–1267. 28. Fagon JY, Chastre J, Hance AJ, Montravers P, Novara A, Gibert C. Nosocomial pneumonia in ventilated patients: a cohort study evaluating attributable mortality and hospital stay. Am J Med 1993;94:281–288. 29. Rello J, Jubert P, Valles J, Artigas A, Rue´ M, Niederman MS. Evaluation of outcome for intubated patients with pneumonia due to Pseudomonas aeruginosa. Clin Infect Dis 1996;23:973–978. 30. Rello J, Rue´ M, Jubert P, Muses G, Sonora R, Valles J, Niederman MS. Survival in patients with nosocomial pneumonia: impact of the severity of illness and the etiologic agent. Crit Care Med 1997;25:1862–1867. 31. Connors AF Jr, Speroff T, Dawson NV, Thomas C, Harrell FE Jr, Wagner D, Desbiens N, Goldman L, Wu AW, Califf RM, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients: SUPPORT investigators. JAMA 1996;276:889–897.

APPENDIX 1: DATABASE CHARACTERISTICS Briefly, in 1992 on the initiative of the Socie´te´ de Reánimation de Langue Française, the CUB-Reá network was set up among intensivists of hospitals in Paris and its suburbs. Initially made of by 22 ICUs the number of participating centers increased up to 30 in 1996 and 35 in 1999. The number of beds per unit ranged from 8 to 20, for a total of nearly 500 beds. Among these, twothirds belong to teaching hospitals (60.4% of stays in 2000), whereas four are surgical ICUs (10.0% of stays in 2000). Each year, about 20,000 stays are added to the database. In 2000, the mean SAPS II was 37.1, mean length of stay was 7.4 days, whereas ICU and hospital death rate were 17.3 and 21.2%, respectively, giving a mean standardized mobility ratio of 0.78. A steering committee meets every month. The Steering Committee has established a house-style book, is responsible for quality controls, training of ICUs’ representatives, and the follow-up of the database. A representative of each ICU has been trained and follows yearly updating courses. A common set of data and thesauri have been defined. All data are recorded prospectively in each unit with standardized database software, whereas methods of coding are harmonized through yearly meetings. Quality assurance procedures, including a computer program operating fifty rules dedicated to consistency between diagnoses and procedures, are applied both to local data within each unit and to data in the common database. Quality controls involving 25 units assessed the reliability of this database of case mix and outcome information (13), showing, for instance, that evaluation of severity of illness, as assessed by SAPS II was highly reproducible (intraclass correlation coefficient ⫽ 0.89). ␬ coefficients measuring agreement for patient sex, type of admission, and outcome were 0.99, 0.78, and 0.97, respectively, whereas reliability of the McCabe and Jackson classification appeared moderate (␬ ⫽ 0.51, when considering modality “nonfatal comorbidity” vs. “other”). Agreement was 89% (␬ ⫽ 0.76) for the coding of cardiovascular failure, 79% (␬ ⫽ 0.56) for respiratory failure, and 90% for renal failure (␬- ⫽ 0.58). Agreement was even higher when considering organ support. The intraclass correlation coefficients for length of stay and total workload were 0.99 and 0.96, respectively. Quality controls were again performed during the year 2000 and confirmed overall reliability of the data, notably showing good

172

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 168 2003

agreement for septic shock (90%, ␬ ⫽ 0.67) and excellent ones for comorbidities such as hematologic malignancy (␬ ⫽ 0.76), cancer (␬ ⫽ 0.81), or human immunodeficiency virus–related disease (␬ ⫽ 0.93).

APPENDIX 2: THE CUB-Reá NETWORK Members of CUB-Reá database are (* indicates ICUs participating in the database since 1993): F. Jardin, B. Page (Hoˆpital Ambroise Pare´*), J. P. Bedos, P. Guezennec (Hoˆpital Andre´ Mignot*), F. Brivet (Hoˆpital Antoine Bećle`re*), Y. Cohen, J. Ph. Fosse (Hoˆpital Avicenne*), Cl. Gibert (Hoˆpital Bichat*), B. Regnier, Ph. Auburtin (Hoˆpital Bichat*), Ch. Richard, J. Depre´-Vassal (Hoˆpital Biceˆtre*), J. Labrousse, E. Guerot (Hoˆpital Boucicaut*), J. Y. Fagon (Hoˆpital Europeén Georges Pompidou*), J. F. Dhainaut, A. Cariou (Hoˆpital Cochin*), F. Fraisse, G. Moret (Hoˆpital Delafontaine), P. Kalfon (Hoˆpital Diaco-

nesses), F. Blin (Hoˆpital Gonesse*), F. Lemaire, Ch. BrunBuisson (Hoˆpital Henri Mondor*), A. Rabbat (Hoˆpital Hotel Dieu*), G. Nitenberg, F. Blot (Institut Gustave Roussy), J. L. Pourriat, R. Gauzit (Hoˆpital Jean Verdier*), F. Baud, D. Goldgran-Toledano (Hoˆpital Lariboisie`re*), D. Dreyfuss (Hoˆpital Louis Mourier*), A. Tenaillon (Hoˆpital Louise Michel), J. L. Pallot, E. Obadia (Hoˆpital Montreuil), J. M. Coulaud, L. Donetti (Hoˆpital Montfermeil), H. Bismuth (Hoˆpital Paul Brousse), T. Similowski (Hoˆpital Pitie´-Salpe´trie`re), F. Bolgert (Hoˆpital Pitie´Salpe´trie`re, H. Outin (MICU, Hoˆpital Poissy/St. Germain), J. P. Terville (SICU, Hoˆpital Poissy/St. Germain), Ph. Gajdos, M. C. Jars-Guincestre (Hoˆpital Raymond Poincare´*), F. Hilpert, P. Manet (Hoˆpital Robert Ballanger*), G. Offenstadt, B. Guidet (Hoˆpital Saint-Antoine*), J. Carlet, B. Misset (Hoˆpital Saint-Joseph), J. R. Le Gall, G. Leleu (MICU, Hoˆpital Saint-Louis*), L. Jacob (SICU, Hoˆpital Saint-Louis*), C. Mayaud, A. Parrot (Hoˆpital Tenon), G. Bleichner, H. Mentec (Hoˆpital Victor Dupouy*).