The annual incidences of severe sepsis in several industrialized nations have recently been reported to be 50–100 cases per 100,000 persons. These numbers exceed the estimated rates for other diseases that hold a heightened public awareness, including breast cancer and acquired immune deficiency syndrome. There are also sex and race differences in the incidence of sepsis. Men are more likely than women to develop sepsis, with a mean annual relative risk of 1.28. Nonwhites are nearly twice as likely to develop sepsis as whites. These race and sex disparities in the incidence of sepsis are likely explained by differences in a variety of factors, including the presence of comorbid conditions. For example, chronic alcohol abuse is associated with a persistent fever, delayed resolution of symptoms, increased rates of bacteremia, increased use of intensive care, prolonged duration of hospital stay, and increased cost of hospitalization for infected patients.
Sepsis remains a major cause of morbidity and mortality in hospitalized patients . Care of patients with sepsis costs as much as $50,000 per patient, resulting in an economic impact of nearly $17 billion annually in the United States alone [2–4]. Between 20% and 50% of patients with sepsis die, and it is the second leading cause of death among patients in noncoronary intensive care units. The Centers for Disease Control and Prevention has estimated that sepsis is the tenth leading cause of death overall in the United States . Furthermore, sepsis is associated with a reduced quality of life in those who survive their acute illness .
Accurate national data on sepsis may be useful for a variety of reasons, including the establishment of health care policy and the allocation of health care resources. However, the diagnostic criteria of sepsis need to be uniformly applied, to accurately compare results from different studies . By consensus, sepsis is now defined as the combination of a pathological infection and physiological changes, known collectively as the systemic inflammatory response syndrome [7, 8]. Patients with coincident acute organ dysfunction are considered to have severe sepsis. There is a limited but growing amount of information concerning the epidemiology of sepsis in a variety of countries around the world . Here, I review recent studies examining the national impact of sepsis in several industrialized countries and explore some of the disparities in the epidemiology of sepsis due to race, sex, and comorbid conditions, such as chronic alcohol abuse.
Global Epidemiology Of Sepsis and Severe Sepsis
To date, 5 studies have reported nationalized epidemiological rates for sepsis. In 2 of these studies, data were collected with multicenter surveys of patients in intensive care units for several weeks, and annual epidemiological data were then extrapolated. In the 3 other studies, large national databases were used, and patients with sepsis were identified on the basis of coding strategies obtained from hospital records.
The EPISEPSIS group conducted a nationwide, prospective, multicenter survey of patients with severe sepsis in 206 French intensive care units over 2 consecutive weeks . During the study, 3738 critically ill patients were screened, and a clinical or microbiologically documented episode of severe sepsis was documented in 546 patients. The overall attack rate for severe sepsis was 14.6% (546/3738). When these findings were extrapolated to the entire population of France (59.6 million), the annual incidence of severe sepsis in French intensive care units was estimated to be 95 cases per 100,000 population. The median age of the patients was 65 years, and >50% had at least one comorbid condition. The Australian and New Zealand Intensive Care Society Clinical Trials Groups studied patients in 23 closed multidisciplinary intensive care units in 21 hospitals throughout Australia and New Zealand . Over a 3-month period in 1999, they identified 3543 intensive care unit admissions for 3338 patients, and 691 met criteria for severe sepsis. The overall adult populations of these countries were obtained from government censuses in the years 1996 and 2001. Assuming a linear population growth, they estimated an annual incidence of 77 cases of severe sepsis per 100,000 population. Consistent with previous studies [2, 11, 12], the mean age of their patients was 61 years, pulmonary and intra-abdominal sources were the most common sites of infection, and positive cultures were present in 58% of cases.
Padkin et al.  used a database compiled in England, Wales, and Northern Ireland to determine the incidence of severe sepsis in these countries. A total of 91 adult intensive care units, or ∼39% of all intensive care units in these countries, contribute data from the first 24 h of intensive care unit admission. In 1997, data were available for 56,673 adult intensive care unit admissions. Overall, 27% of patients (15,362) had severe sepsis during their first 24 h after admission to the intensive care unit. When the results were modeled for England and Wales, the annual incidence of severe sepsis was 51 cases per 100,000 population. Angus et al.  constructed a similar patient database from discharge records from hospitals in 7 states in the United States in 1995. To identify patients with severe sepsis, they selected all acute-care hospitalizations with a hospital discharge code for a bacterial or fungal infection and a diagnosis of acute organ dysfunction. In total, they used 1286 distinct infection codes. However, only 225 of these infection codes were needed to identify 99% of the patients. After adjusting for age and sex, they estimated a national incidence for severe sepsis of 300 cases per 100,000 persons. However, an editorial expressed concern that this estimate may overstate the incidence of severe sepsis by 2–4-fold, given that the estimated deaths would exceed the number of deaths reported in association with nosocomial bloodstream infections and septic shock combined . More recently, our group used the National Hospital Discharge Survey database and identified >10 million cases of patients with sepsis in the United States during 1979–2000 . We have subsequently updated this information through 2002. When severe sepsis is defined as a diagnosis of sepsis and acute organ dysfunction, the incidence of severe sepsis during 1997–2002 was 91 cases per 100,000 population.
The results from these national epidemiological studies illustrate that severe sepsis, defined as infection along with systemic inflammatory response syndrome and acute organ dysfunction, is a common disorder . For example, the frequency of cases and deaths related to severe sepsis exceeds the numbers of persons with other diseases that hold a heightened public awareness, such as breast cancer and AIDS [17, 18]. It is hoped that the results of these studies will translate into improvements in health policy, resource allocation, and the distribution of funding for sepsis research. In addition, these epidemiological studies demonstrate that sepsis occurs in elderly patients. Treatment trials for patients with severe sepsis have sometimes excluded elderly patients because of concerns that they are less likely to respond to the therapeutic intervention and are at a higher risk of dying . Future clinical trials must assess the efficacy and cost-effectiveness of investigational medications in this large group of patients, which will most likely receive these therapies. With the recent approval of activated protein C, a treatment is available that reduces mortality in patients with severe sepsis . The high cost of caring for patients with sepsis makes comprehensive epidemiological information useful for hospitals to develop accurate fiscal budgets.
On the basis of 4 of these 5 studies, the incidence of severe sepsis consistently lies between 50 and 100 cases per 100,000 persons in industrialized countries  ( table 1). Some of the variability in reported incidence is related to differences in study design. For example, the diagnostic criteria for severe sepsis vary between those studies that prospectively identified patients and those that used hospital records. In addition, variability in the percentage of intensive care unit admissions related to severe sepsis may reflect differences in national and hospital-specific bed availability and admissions policies in intensive care units. Finally, the study of Padkin et al.  identified cases of severe sepsis only during the first 24 h after intensive care unit admission and, therefore, likely underestimated the true incidence. However, there may be specific biological reasons for the different incidences of sepsis in the various countries. For example, there may be disparities in distribution of specific comorbid conditions or genetic polymorphisms between the patients enrolled into the studies that, when present, alter the probability of developing severe sepsis.
Race And Sex Differences In Sepsis
There are clearly discrepancies in the incidence and severity of several medical conditions on the basis of race and sex. With regard to race, African Americans have a higher prevalence of essential hypertension . There have been several well-publicized differences in cardiovascular care and outcome according to race . In the oncology literature, African American women have a higher risk of developing ovarian cancer and a worse outcome from breast cancer [22, 23]. With regard to pulmonary diseases, sarcoidosis has been estimated to be >10 times more prevalent in African Americans than in whites in the Unites States [24, 25]. In addition, African American men consistently had a higher incidence of and mortality due to lung cancer than did white men during the 1970s and 1980s .
Relatively few clinical studies have investigated race and sex differences in patients in intensive care units, and almost none have directly studied patients with sepsis [27–30]. In the study of Padkin et al. , there was a predominance of men (58.8%) in their cohort of patients with severe sepsis. Similarly, there were 59.6% men in the Australian and New Zealand study and twice as many men as women in the French EPISEPSIS study [9, 10]. In the study of Martin et al. , after adjustment for sex in the population of the United States, men were more likely to have sepsis than women in every year of their 22-year study, with a mean annual relative risk of 1.28  (figure 1). Racial disparities in the incidence of sepsis were even more striking. Both African Americans and other nonwhites had a similarly elevated risk of sepsis, compared with whites (mean annual relative risk of 1.89 and 1.90, respectively). Most concerning was the risk among African American men: they had the highest rate of sepsis during the study period (330.9 cases per 100,000 population), the youngest age at onset (mean age, 47.4 years), and the highest mortality (23.3%) . Race and sex disparities in sepsis related to the source of infection have also been examined [31, 32]. Overall, 32% of the cases of sepsis were from a respiratory source, 32% from a genitourinary source, and 22% from a gastrointestinal source. With regard to race, African Americans and other nonwhites had higher incidences of sepsis for all sources of sepsis. Similar, men had higher incidences than did women for all sources of sepsis expect genitourinary sources (figure 2).
Although the disparities in the incidence of sepsis are likely explained by a variety of factors, including genetic polymorphisms, exposure to resistant organisms, and access to health care, there are race and sex differences in specific comorbid conditions in patients with sepsis. In our cohort of 10 million cases of sepsis, race and sex differences were present in the distribution of several common comorbid conditions, including alcohol abuse, diabetes, HIV, end-stage renal failure, and cancer. For example, African Americans and nonwhite patients with sepsis were more likely to have several concurrent diagnoses that are associated with an increased risk of infection, including diabetes, end-stage renal disease, HIV, and alcohol abuse. Similarly, male patients with sepsis were more likely to have chronic obstructive pulmonary disease, cancer, alcohol abuse, or HIV than were female patients with sepsis.
Chronic Alcohol Abuse and Infection Or Pneumonia
Alcohol abuse is a common behavioral condition that has been reported to alter immune function and increase susceptibility to infection . Alcohol is the most frequently abused drug in the world . In the United States, ∼50% of the adult population regularly consumes alcohol, and 15–20 million persons carry the diagnosis of alcohol abuse or dependence . Alcohol is the third leading cause of preventable mortality and is associated with an estimated 100,000 deaths per year in the United States. The annual economic cost of alcohol abuse in 1990 was $100 billion in the United States, and >10% of this cost was directly related to medical services [35–37]. In addition, 20%–40% of patients admitted to general hospitals have alcohol-related disorders, and hospitalizations resulting from alcohol abuse are as common among elderly patients as are hospitalizations resulting from myocardial infarctions . In some industrialized countries, alcohol consumption is decreasing . However, alcohol consumption is increasing in developing nations and is a problem of special concern in areas of central and eastern Europe .
Since the late 1700s, clinicians have postulated that excessive use of alcohol is associated with an increase risk of infection . In 1905, Sir William Osler  postulated that alcohol abuse was the single most potent predisposing condition for the development of bacterial pneumonia. Usually, 25%–50% of patients with bacterial pneumonia have a prior history of alcohol abuse [40–43]. In 1965, Nolan  reviewed 900 consecutive admissions over a 5-month period and classified 124 patients as being alcoholic when defined by psychological criteria . The incidence of acute bacterial pneumonia was significantly higher among alcoholic patients (17.0%) than among patients who were not alcoholics (6.5%). More recently, 2 longitudinal studies have examined the association between alcohol abuse and subsequent hospital admissions in enlisted men [46, 47]. Persons with a primary diagnosis of alcohol psychosis or alcoholism were matched with a control group of persons who were not alcoholics and were chronologically followed for both the quantity and etiology of their subsequent hospital admissions. During the first year of service, younger enlisted men who were alcohol abusers had a higher incidence of respiratory system diseases than did controls subjects who were not alcohol abusers . For older personnel, alcohol abusers were twice as likely to be hospitalized with the admission diagnosis of pneumonia as were persons who were not alcohol abusers . In a European study, 50 patients with community-acquired pneumonia were matched by age and sex with control subjects . The patients with pneumonia had a significantly higher daily alcohol intake prior to hospitalization and had used alcohol chronically for a longer period of time. After adjustment for the presence of cirrhosis and cigarette smoking, excessive alcohol intake was the only independent risk factor positively associated with the development of community-acquired pneumonia.
Severity studies. The effects of alcohol abuse on a variety of outcome measures for patients with bacterial pneumonia have been examined, including duration of stay, recurrence, intensive care unit stay, and hospital cost. In a study of 358 cases of pneumococcal pneumonia, prolonged fever, slower resolution, and a higher rate of empyemas were noted in patients with chronic alcoholism . In a Scandinavian study of 277 patients with community-acquired pneumonia, alcoholism was associated with delayed recovery, defined as the return to normal activity at 8 weeks after hospital admission . In 1985, 312 patients admitted because of community-acquired pneumonia were followed at a public municipal hospital over a 12-month period . In this study, a higher incidence of positive results of blood cultures was present in the 118 patients with a history of alcohol abuse than in patients without a history of alcohol abuse. In the case-control study discussed above, the authors followed the 50 patients with pneumonia during their hospital stay . The durations of fever (4.3 vs. 2.1 days) and hospital stay (10.3 vs. 6.9 days) were significantly longer in the alcoholic patients. In addition, alcoholic patients had a higher frequency of persistent pulmonary infiltrates on chest radiography at 1 week.
Recently, the effect of alcohol abuse on the development of nosocomial pneumonia has been studied. Bercault and Boulain  studied 135 patients who developed ventilator-associated pneumonia and 135 matched control subjects. The matching process was extremely well done and was based on a variety of criteria, including cause of hospital admission, indication for ventilatory support, immunologic status, cardiac status, age, Glasgow coma score, and probability of death, as determined by an admission severity-of-illness score. The primary goal of this study was to determine whether nosocomial pneumonia was independently associated with death in the intensive care unit. However, in the multivariate analysis, chronic alcohol abuse was also independently associated with mortality in the patients with nosocomial pneumonia.
Chronic alcohol abuse has also been reported to be an independent risk factor for the development of acute respiratory distress syndrome (ARDS). Initially, surgeons from the University of Washington examined the effects of acute and chronic alcohol abuse on the morbidity and mortality of trauma, a known at-risk diagnosis for the development of ARDS . The risk of respiratory failure, defined as the requirement of mechanical ventilation, was higher among trauma patients with evidence of chronic alcohol abuse. Hudson et al.  identified 695 critically ill patients with 1 of 7 at-risk diagnoses for the development of ARDS. ARDS occurred in 179 patients (26%). In patients with both prior alcohol-related disease and a low arterial pH, the risk of developing ARDS (71.4%) was twice that observed in patients with a normal pH and no history of alcohol-related disease (38.7%). Subsequently, 351 medical and surgical intensive care unit patients with 1 of 7 at-risk diagnoses were followed for the development of ARDS . Patients with a prior diagnosis of chronic alcohol abuse were nearly twice as likely to develop ARDS as were patients without a history of alcohol abuse (figure 3). The association between chronic alcohol abuse and ARDS has been confirmed in a prospective trial of 220 patients with septic shock . In addition, the effects of chronic alcohol abuse on the incidence of postoperative ARDS have been studied in patients undergoing lung resection surgery . A total of 21% of patients with a history of alcohol abuse developed ARDS after their surgery, compared with only 2% of postoperative patients without a history of alcohol abuse [57, 58].
In an attempt to determine the economic impact of alcohol abuse on hospitalizations for pneumonia, Saitz et al.  examined a statewide database over a 1-year period. Only 4% of the 23,198 cases of pneumonia were classified as alcohol related, on the basis of appropriate hospital discharge codes. The overall hospital mortality for pneumonia was 10%, and 12% of all admissions included a requirement for intensive care. In a risk-adjusted analysis, the hospital charges ($11,179 vs. $9886), the total duration of hospital stay (increased by 0.6 days), and the requirement for intensive care (18% vs. 12%) were higher for the patients with a history of alcohol abuse.
Alcohol and immune function. There are many potential mechanisms by which alcohol abuse can increase susceptibility to infection, bacterial pneumonia, and sepsis, including an increased risk of aspiration, poor dental hygiene, suppression of a normal cough reflex, malnutrition, and physical proximity to other infected people . However, both acute and chronic exposure to alcohol have direct effects on the immune system . Alcohol abuse has been reported to cause alterations in neutrophil and macrophage function and abnormalities in ciliary and surfactant function in the lung .
After exposure to bacterial toxins, macrophages normally secrete TNF, other inflammatory cytokines, and reactive oxygen intermediates. Acute alcohol administration has a suppressive effect on release of proinflammatory cytokines, such as TNF, IL-1, and IL-6, from monocytes and macrophages [62, 63]. This suppressive effect of acute alcohol exposure on the sepsis-induced TNF response in alveolar macrophages occurs at a posttranscriptional level . In addition, acute alcohol exposure suppresses the release of several chemokines, including IL-8, macrophage inflammatory protein—2, and cytokine-induced neutrophil chemoattractant. In addition, diminished hydrogen peroxide production by alveolar macrophages has been demonstrated in a septic rat model of acute alcohol abuse . These acute effects of alcohol lead to decreased neutrophil recruitment, impaired bacterial clearance, increased dissemination of bacteria outside of the lung, and increased mortality in animal models of sepsis . Some of the intracellular mechanisms responsible for the acute effects of alcohol have been elucidated. Human monocytes exposed to alcohol in vitro express inhibited lipopolysaccharide-induced nuclear factor—κB activation by decreasing DNA binding of the p65/p50 heterodimer . In addition, alcohol prevents nuclear translocation of p65 and, to a lesser extent, p50 subunits.
Acute alcohol intoxication also decreases the clearance of bacteria from the lung or after an intradermal injection [68, 69]. When mice were exposed to aerosolized Staphylococcus aureus, the clearance of bacteria from the non—alcohol-exposed lung followed a logarithmic curve, with 87% of the bacteria removed within 4 h and only 1% of the bacteria remaining at 24 h. In contrast, mice acutely exposed to alcohol intraperitoneally 15–30 min before intratracheal administration of bacteria cleared ∼50% of the bacteria at 4 h. This alteration in clearance is likely due to several factors, including depression of ciliary function. Laurenzi and Guarneri  quantitated ciliary function by measuring the time required to move carbon particles a distance of 5 mm in the intact tracheas of kittens. Normally, carbon particles moved at a fast rate of 3.12 s/mm, yet acute alcohol exposure markedly slowed ciliary function to 10.6 s/mm.
Chronic alcohol ingestion causes different abnormalities in immune function. In animal models, chronic exposure to alcohol has been reported to increase the release of proinflammatory cytokines and chemokines, such as TNF, IL-1β, and IL-8. This proinflammatory response is likely related to an increased production of reactive oxygen species . However, isolated alveolar macrophages from chronic alcoholic patients released less TNF in response to stimulation than did alveolar macrophages from healthy persons . In addition, chronic alcohol abuse does not impair neutrophil mobilization or adherence . Surfactant has a potent bactericidal activity against invading pulmonary pathogens, which is believed to be secondary to the detergent-like activity of surfactant long-chain free fatty acids . Chronic alcohol abuse causes deficiencies in the amount of these free fatty acids produced by alveolar type II cells, induces the release of surfactant inhibitors, and leads to decreased opsonization of microorganisms and impaired bactericidal activity [75, 76].
Chronic alcohol exposure has been reported to enhance intestinal permeability, leading to abnormalities in intestinal epithelial barrier function . Increased intestinal permeability can facilitate bacterial translocation, a process by which gut flora or bacterial products traverse the abnormal intestinal barrier and, ultimately, reach mesenteric lymph nodes and the portal circulation. These translocating bacteria can then disseminate and cause sepsis.
In summary, we have identified specific populations that are at increased susceptibility to develop sepsis and severe sepsis. Other specific risk factors are associated with a poor prognosis for patients who have developed sepsis. Some of these persons are distinguished by race or sex and others by other diseases or comorbid conditions that they might have acquired. The recognition of responsive populations in sepsis is important for several reasons. This information will be helpful to clinicians in answering questions regarding prognosis in family meetings. It is also possible that some of these specific populations might respond differently to the therapies that are now available for patients with sepsis. Finally, the identification of specific patients who are predisposed to developing sepsis raises many questions regarding the reasons for increased susceptibility. By understanding the mechanisms that cause these alterations in responsiveness, the pathogenesis of sepsis will be better elucidated. Therefore, physicians will be able to provide better care for all patients with sepsis.
Financial support. National Institutes of Health (grants AA-014435 AA-013757).
Potential conflicts of interest. M.M.: no conflicts.