African Americans are more likely to be admitted into the ICU presumably due to a higher prevalence of conditions requiring critical care , a higher severity of illness, and an increased number of comorbid conditions that complicate management [22–24]. For example, African American men may be at greater risk for the development of sepsis from Grampositive bacteria and to have at least one acute organ dysfunction upon presentation compared to whites [25]. Using the New Jersey inpatient database, Dombrovskiy et al. found that African American adults with sepsis were younger, had more comorbid conditions, and were more likely to be admitted into the ICU compared to whites [26]. In a cohort study using data from the Acute Respiratory Distress Syndrome Network, African American patients with ARDS had the greatest severity of critical ill ness and were more likely to have complicating comorbidities such as HIV, end-stage renal disease, or cirrhosis [23]. After adjusting for demographic and clinical factors, African Americans and Hispanics with ARDS had a higher mortality rate compared to non-Hispanic whites with ARDS. However, after adjusting for severity of illness, African American race was no longer associated with mortality while the relationship between Hispanic ethnicity and mortality persisted.

A similar trend of increased ICU admission is seen in the pediatric population . However, the reasons for it may not be as clear as in adults. In a study of 4676 pediatric ICU admissions in Shelby County, Tennessee, African American children were more likely to be admitted to the pediatric ICU compared to non-Hispanic white children of similar severity of critical illness (OR: 2.1, 95 % CI: 1.7–2.7) [27]. Interestingly, full-term African American children had higher risks of admission compared to full-term white children (OR: 1.8, 95 % CI: 1.3–2.5) but there was no difference in risk for admission between preterm African American and preterm white children (OR: 1.4; 95 % CI: 0.9–2.2). In the region where the study was conducted, African Americans have a significantly higher prevalence of preterm births (0.1 compared to 0.06 for non-Hispanic whites). Once admitted into the ICU, there was no difference in mortality between full-term African American and white children (4 % vs. 6 %, p = 0.2) or preterm African American and white children (6.8 % vs. 8.9 %, p = 0.5). The authors posit that African American children, particularly those with special needs, are less likely to receive primary care and therefore are more likely to be admitted to the ICU with exacerbations of unmanaged or undermanaged diseases [28–30]. A better understanding of factors affecting ICU admission and outcomes in the pediatric population is needed.

Once admitted into an ICU, African American patients receive disparate care compared to whites, despite no ultimate difference in mortality. In a study of over 15,000 patients admitted to one of 42 ICUs in 40 hospitals, Williams et al. found that African Americans received less technological monitoring, less laboratory testing, and less life-supporting treatments within the first 24 h following admission compared to whites. Adjusted ICU lengths of stay were also shorter [31]. Rapoport et al. found that critically ill African Americans of similar severity of illness to whites received fewer pulmonary artery catheters [32].

There are also ethnic barriers to optimal critical care. Limited English proficiency is a facet of ethnic disparities that has received a considerable amount of recent attention due to the alarming frequency of associated adverse outcomes including delayed care, permanent disability, or even death [33–36]. Some of the most influential factors associated with these poor outcomes include miscommunication, lack of cultural understanding, and poor social support. During the ICU family conference, patients’ surrogates with limited English proficiency may receive less information regarding their loved one’s current illness and proposed treatment and less emotional support from caregivers despite the presence of professional interpreters [37, 38]. It appears that the presence of interpreters attenuates but does not eliminate these adverse outcomes to patients with limited English proficiency [39, 40]. However, some data suggest that outcomes among patients with limited English proficiency may be equal to or better than those of English proficient patients. In a large cohort study of patients admitted to the ICU of two Boston hospitals between 1997 and 2007, patients whose primary language was not English had 31 % lower odds of 30-day mortality compared to patients whose primary language was English [41].

Currently, patients older than 65 years of age comprise 56 % of all ICU days and patients older than 85 years of age comprise 14 % [7, 42]. With the population continuing to age, this percentage is expected to increase [43]. Using prospective data from Australian and New Zealand ICUs, Bagshaw et al. predicted that by 2015 the rate of patients 80 years of age and older will increase by 72 % to approximately 1 in 4 ICU admissions [44].

Milbrandt et al. posited that aging “predisposes to critical illness due to lifelong accumulation of molecular and cellular damage leading to decreased physiologic reserves and leaving the individual less able to respond to stressors” [43]. Therefore, it is not surprising that increasing age is positively associated with increasing mortality. In Bagshaw’s study, patients 80 years of age and older had an adjusted odds of ICU and hospital mortality significantly higher than patients between the ages of 18 and 40 (OR: 2.7, 95 % CI: 2.4–3, and OR: 5.4, 95 % CI: 4.9–5.9, respectively). Factors associated with the higher odds of death among patients 80 and older included admission from a chronic care facility, nonsurgical admission, need for mechanical ventilation, comorbid conditions, a longer ICU stay, and a higher severity of illness. However, despite the increased odds of mortality compared to younger patients, approximately 80 % of patients 80 years of age and older survived to hospital discharge. In a cohort study spanning 7 years of ICU patients admitted to a single academic center in Massachusetts, patients over the age of 65 represented more than 45 % of the total ICU population. Mortality (28-day and 1-year) increased with age despite adjustment for gender, comorbidities, severity of critical illness, and presence of do not resuscitate orders [45].

Severe sepsis, a frequent cause of ICU stay and mortality, has enjoyed an increased survival rate over the last decade. Some have attributed the increase to a change in the age distribution or case fatality rate among those affected. Using a cohort of fee-for-service Medicare beneficiaries aged 65 and older, Iwashyna et al. found that the number of incident 3-year survivors of severe sepsis rose 119 % between 1999 and 2008. They attributed this increase in survivorship to an increased rate of organ dysfunction per patient hospitalized with infection rather than a change in the age distribution or better survival among patients. In fact, the 3-year case-fatality rates only decreased from 73 to 71 % over the period of analysis [46].

With the majority of older individuals surviving a hospitalization for severe sepsis, a new problem has emerged—that of chronic disability. Prospective data of 470 patients with severe sepsis admitted to 24 ICUs in Finland revealed a 2-year mortality of 45 % and a lower quality of life compared to age- and sex-adjusted reference values without sepsis [47]. The 2-year mortality was 35 % among those patients older than 55 years of age compared to younger patients (9.8 %, p < 0.001). As age increased, quality of life decreased, while the mean estimated cost per quality-adjusted life year (QALY ) increased ranging from 325€ for those less than 24 years of age to 12,452€ for those over 81 years of age. The findings of significantly impaired quality of life following hospitalization and increased long-term mortality rates were confirmed in a 2010 systematic review of 30 studies [48]. The severity and duration of impairment was well delineated in another study by Iwashyna et al. They prospectively examined participants from the Health and Retirement Study whose data were linked with the Medicare database [49]. The prevalence of moderate to severe cognitive impairment was 11 % greater among the patients who had been hospitalized for severe sepsis compared to those who had been hospitalized for other conditions. Impairments in cognitive and physical functioning persisted for at least 8 years following hospitalization, suggesting that many patients may be unable to ever return to independent functioning.

Despite the fact that the majority of critically ill older patients survive hospitalization, age-related barriers to ICU admission appear to exist. In a prospective cohort study conducted in 15 French hospitals, 2646 patients of age 80 years and older were triaged in the emergency room. The authors used standardized admission criteria to determine patient eligibility for admission. Of the 1426 patients who met definite admission criteria, only 31 % were referred for ICU admission, and only 52 % of those referred were admitted [50]. Increasing age was an independent factor associated with no referral for ICU admission (OR: 1.04, 95 % CI: 1.02–1.07 for every 1 year increase). Another study demonstrated that once admitted into the ICU, elderly patients are less likely than younger patients to receive intensive treatments such as mechanical ventilation and renal replacement therapy, perhaps due to the subjective perception among healthcare providers of a potential lack of benefit from treatment [42]. In both of these studies, it is unclear what role patient preferences may have played in decisions regarding admission and intensity of care.

Patient and provider decisions regarding care of the critically ill older patient may be based on incomplete or faulty information. In 1995, there were 215,000 deaths attributable to severe sepsis representing 9.3 % of all deaths in the United States and equivalent to the number of deaths attributed to acute myocardial infarction (AMI). The burden of severe sepsis is significant among the older population. The incidence of severe sepsis is 26/1000 among those 85 years and older compared to 5/1000 for adults between that ages of 60 and 64. Moreover, mortality from severe sepsis is 38 % among those ≥85 years of age and <30 % for those between 60 and 64. Despite the increased burden, observational studies of severe sepsis and clinical trials of sepsis therapies often exclude the elderly due to perceptions of increased risk of death or lack of response to treatment. It is important to note that the majority of elderly patients admitted to the ICU with severe sepsis are discharged alive. In addition, as pointed out by Angus et al., with the elderly comprising a substantial proportion of the critically ill population, excluding them from such studies threatens external validity and prevents a comprehensive public policy approach from being created [43, 51].

It is also important to consider the effect of multiple demographic factors on patient outcomes. For example, age and race may be interacting to uniquely affect health outcomes. In a study using data from the National Hospital Discharge Survey, Martin et al. found that African American men presented with the highest rates of sepsis (331 cases/100,000), the youngest age at onset (47 years), and the highest mortality (23 %) [52]. The reasons were not explored, but the authors presented several possible mechanisms including genetic, social, and clinical differences, and called for further investigation to be performed.

Although women comprise a larger proportion of the US population, the evidence suggests that men have a higher incidence of critical ill nesses such as sepsis (mean annual relative risk: 1.3, 95 % CI: 1.2–1.3) [52]. Despite the increased incidence, men do not appear to have a higher case-fatality rate compared to women [25]. In surgical ICUs in the US, for example, men were found to have a higher incidence of sepsis and septic shock compared to women but no difference in hospital or postdischarge mortality [53]. Men are also at risk of prolonged ICU stays compared to women of a similar severity of illness [54]. This may represent differences in end-of-life care, personal preferences regarding care, or other unmeasured factors.

In Europe, some data suggest women have a higher severity of illness and receive a lower overall intensity of care compared to men, but there are no apparent differences in mortality by gender. For example, among 25,998 adults admitted to one of 31 ICUs in Austria, women had higher severity of illness scores (SAPS II 28 vs. 26, p < 0.001) and in-hospital mortality rates (18 % vs. 17 %, p = 0.04) compared to men [55]. However, after adjustment for severity of illness, the mortality rate did not differ between men and women. Men received more intensive care compared to women including mechanical ventilation, vasoactive medication, placement of central venous and pulmonary artery catheters, and renal replacement therapy compared to women. These results suggest that the SAPS II score did not fully capture patient severity of illness or other factors besides severity of illness are associated with disparate receipt of intensive therapies between men and women.

Gender differences in receipt of therapies are found in other parts of the world as well. In the U.S., evidence suggests that critically ill men are more likely to receive thrombolytic therapy, emergent surgery, mechanical ventilation, and even coronary artery bypass graft surgery more frequently than critically ill women [56]. Similar results were found in Canada [57]. In a retrospective examination of almost 25,000 critically ill patients admitted to Ontario hospitals over a 2-year period, women were less likely to be admitted into an ICU compared to men (40 % compared to 60 %, p < 0.001). In fact, older women (≥50 years of age) had 32 % lower odds of being admitted compared to older men. Older women were also less likely to receive mechanical ventilation and pulmonary artery catheterization, and they had shorter ICU stays but longer overall stays in the hospital. Most concerning was the fact that ICU and in-hospital mortality rates were greater for older women compared to older men.

While race, ethnicity, age, and gender all play substantial roles in the development of disparate outcomes, other factors linked to these demographic indices may be equally if not more important. Studies that have explored the root causes of disparities have often found that much of the effect attributed solely to race, ethnicity, gender, or age is significantly attenuated upon consideration of potential confounders. Unfortunately, few studies have incorporated detailed adjustments of these confounding factors into their analyses. A deeper understanding of the effects of these confounding factors is warranted.

Emerging evidence suggests that genetic predisposition may play a role in many disorders affecting the critically ill. However, the extent to which genetic predisposition plays a role in the development of health disparities has not been well described. Perhaps the best evidence linking genetics and premature mortality among the critically ill was from a case–control study of 976 adult Danish nonfamilial decedent adoptees and their biological and adoptive parents [58, 59]. Sørensen et al. found an increased mortality among the biological parents of decedent children but not among their adoptive parents. The associated causes of death included all of the major sources of critical illness: infectious causes (HR: 1.9, 95 % CI: 1.1–3.5), vascular causes (HR: 2.0, 95 % CI: 1.2–3.1), and even natural causes (HR: 1.2, 95 % CI: 1.0–1.4). However, there was no adjustment for demographic or socioeconomic factors. Other studies have identified only a few heritable mutations predisposing to critical illness that are limited mainly to a handful of families. For example, Picard et al. described 3 unrelated children with inherited interleukin-1 receptor-associated kinase (IRAK-4) deficiency rendering them susceptible to recurrent pyogenic bacterial infections [60]. Recent genetic epidemiologic studies have focused on the more prevalent genetic variations [61]. Differing allelic frequencies have been found by both race and gender in patients with ARDS and sepsis. The myosin light chain kinase gene (MYLK) encodes a multifunctional protein involved in the inflammatory response [62]. Different single-nucleotide polymorphisms of MYLK were found to be associated with sepsis and sepsis-associated ARDS among African Americans and whites. The functional T-46C polymorphism in the Duffy antigen/receptor for chemokines (DARC) gene is found almost exclusively in persons of African descent and associated with worse clinical outcomes among African Americans with ARDS, perhaps due to an increase in circulating IL-8 [63].

Part of the difficulty in identifying genetic influences on health disparities is due to the wide variability in genetic variants between people of different ancestries. When diverse populations are studied, the associations with a clinical phenotype may be mistaken for being associated with the presence of multiple specific genetic variants determining a predisposing genotype, rather than with an association with prevalence/incidence due to patient ancestry [61]. This spurious association confounding can be overcome by stratifying the case and control groups with different fractions of ancestry from each ancestral subpopulation [64]. Unfortunately, this level of detail is often missing from genetic epidemiology studies in critical illness.

Vulnerable critically ill patients are at high risk of experiencing poor health outcomes because of poor access to acute and chronic care, lower socioeconomic status, lower levels of education, higher rates of unemployment, and a higher burden of chronic disease compared to the majority of patients [65]. The imbalances in the geographical distribution of resources, available technological advancements, and distribution of wealth have intentional and unintentional repercussions that have left increasing numbers of the general population unprotected. In the United States, for example, minority populations frequently live clustered together in neighborhoods separated from white populations. Due to the need for emergent care, critically ill patients are often cared for in hospitals nearest to their homes. As a result, critically ill minority patients are more likely to receive care in different hospitals compared to critically ill white patients. The resources available to persons living in minority-predominant neighborhoods are often fewer compared to majority-predominant neighborhoods . This is true in health care as well. Indeed, a recent analysis of Medicare data revealed that only 25 % of hospitals in the United States care for almost 90 % of elderly African American patients [66]. These hospitals tended to be larger and more often were teaching hospitals situated in the southern United States. They also tended to have worse measures of quality of care including treatment of AMI, heart failure, and pneumonia compared to hospitals caring for lower proportions of African American patients.

Hospital level factors may influence health disparities more than patient-level factors. For example, in a study of patients admitted to 28 hospitals for community-acquired pneumonia , African American patients were less likely to receive timely or guideline-adherent antibiotics [67]. Within each hospital, African American and white patients received a similar quality of care. However, among hospitals serving a greater proportion of African American patients, African American and white patients with community-acquired pneumonia were less likely to receive timely antibiotics (OR = 0.8, 95 % CI: 0.8–0.9) and were more likely to receive mechanical ventilation (OR = 1.6, 95 % CI: 1.0–2.4). In a retrospective population-based cohort study including six U.S. states (Florida, Massachusetts, New Jersey, New York, Virginia, and Texas), African Americans had the highest age- and sex-standardized population-based incidence of severe sepsis and hospital-acquired infections and the highest ICU case fatality rates compared to Hispanic and non-Hispanic whites [22]. However, adjustment for clinical characteristics and the treating hospital fully explained the higher case fatality rate.

It appears that when it comes to health care in general in the US, separate may not be equal. For example, risk-adjusted mortality after AMI is higher among African American and white patients admitted to hospitals caring for the highest proportion of African American patients compared to those caring for the lowest [68]. Hospitals with large proportions of African American patients also have worse cardiac arrest outcomes compared to hospitals with predominantly white patients [69, 70]. This might explain the disparities in survival following cardiac arrest noted between African Americans and whites. Among patients discharged to home following evaluation in emergency rooms in Arizona, Massachusetts, and Utah, African American and Asian patients had lengths of stay ranging from 2 to 14 % shorter than white patients in teaching hospitals, and 1.6 to 16 % longer than white patients in nonteaching hospitals, potentially leading to incomplete clinical evaluations [71]. Finally, in critical care at the end of life, Barnato et al. found differences in ICU use between African Americans, Hispanics, and whites that were attributed to admission into different hospitals with varying ICU utilization patterns at the end of life rather than effects of patient race or ethnicity on ICU use within the hospitals [72]. These studies provide a compelling case to adjust for type and locations of hospitals in all studies of health disparities affecting the critically ill.

The Department of Housing and Urban Development conducted a study that provided insight into successful interventions that may overcome the harmful effects of poverty and segregation on health outcomes [73]. Between 1994 and 1998, 4498 women and children living in public housing in high-priority urban census tracts were randomized to one of three groups. The first group was assigned to receive housing vouchers which could be redeemed only if the participant family moved to a census tract where <10 % of the residents were impoverished and if the participant received counseling regarding moving. Participants in the second group were assigned to receive unrestricted, traditional vouchers with no additional counseling on moving, and participants in the third group served as a control group that received neither vouchers nor counseling. Ten to 15 years later, participants were contacted to determine their body mass index and glycated hemoglobin levels as proxies for the development of high-risk morbid conditions. Participants who had received the vouchers to move to low poverty census tracts combined with counseling on moving were less likely to be obese and had lower glycated hemoglobin levels than participants in the control group. There were no differences in body mass index or glycated hemoglobin among participants in the unrestricted voucher group and participants in the control group. Whether this or similar interventions will have an effect on critical illness outcomes remain to be seen.

Comorbid conditions have a significant effect on critical care outcomes [74], and differential prevalence of comorbid conditions as well as differential receipt of treatment of such conditions may explain a significant portion of observed racial and ethnic differences in critical care outcomes. African Americans are more likely to be hospitalized for ambulatory care-sensitive conditions—conditions for which appropriate ambulatory care could prevent hospitalizations—compared to whites [75, 76]. Among patients admitted to ICUs in 35 California hospitals, Erikson et al. found no racial or ethnic differences in in-hospital mortality or ICU length of stay after adjusting for severity of illness, socioeconomic status, and insurance status [77]. They did find that African American patients were more likely to be admitted with a higher severity of illness and more metabolic derangements suggesting poor access to care and poor control of comorbid conditions prior to admission. The lack of a difference in mortality when compared to white patients could be due to the fact that providing initial care for exacerbations of chronic diseases altered the trajectory of the critical illness. For example, African Americans between the ages of 45 and 64 are 2.5 times more likely to die of heart failure compared to whites of similar age [78]. An African American patient may be admitted to the ICU with an acute exacerbation of previously untreated heart failure with reduced ejection fraction due to poor access to ambulatory care and consequently an inability to initiate routine first-line therapy such as diuretics or ACE inhibitors. With prompt initiation of these agents upon ICU admission, his ICU mortality may improve even though his acute severity of illness on presentation was high, as these agents have previously been demonstrated to be effective in reducing mortality among patients with his degree of CHF. A white patient presenting with a similar CHF exacerbation and an equal acute severity of illness may have an equal or worse mortality as he may have had better access to evidence-based treatments for heart failure while in the ambulatory setting and therefore may already be taking several medications that are indicated for the treatment of CHF. The current measures of ICU severity of illness such as APACHE and SAPS do not account for severity of chronic illness nor do they account for degree of optimization of comorbid conditions. The African American patient may appear sicker according to such severity of illness measures, but require less aggressive treatment from the care team and have a lower ICU mortality. Another explanation for the lack of mortality difference between African American and white patients in this study may be that the participating hospitals were located in the west coast, which care for a higher proportion of white patients and may therefore deliver superior care compared to hospitals in other regions of the country which serve predominantly African American patients.

Almost 100 million people worldwide are forced into poverty each year because of catastrophic household medical expenses [79]. As evident in other areas of healthcare, lack of adequate health insurance adversely affects critical care outcomes. For example, low-income and uninsured individuals residing in large metropolitan areas are much less likely to visit with a physician compared to those with higher income or health insurance [80]. Uninsured patients experiencing new serious or morbid symptoms are less likely to receive medical care even though they think they need it [81]. In a systematic review of 29 studies examining the association between insurance status and critical care delivery and outcomes, uninsured patients were less likely to receive critical care services than those who were insured [82]. Following admission, uninsured patients also received fewer procedures compared to insured patients. Most importantly, lack of insurance was associated with an increased risk of death. In a more recent study not included in the systematic review, Lyon and colleagues performed a retrospective review of patients admitted to Pennsylvania hospitals in 2005 and 2006 [13]. They performed comprehensive patient clinical and demographic adjustments while also considering hospital-level effects and found an increased 30-day mortality among uninsured patients compared to privately insured patients (5.7 % vs. 4.6 %, p < 0.001). Uninsured patients were also less likely to receive a central venous catheter (7.3 % vs. 9.8 %, p < 0.001), acute hemodialysis (0.7 % vs. 1.1 %, p < 0.001), or tracheostomy (8.6 % vs. 22 %, p < 0.001). As pointed out by the authors, adjusting for hospital-level effects in the analysis allowed comparisons between uninsured and private patients cared for at the same hospitals. Therefore, the lower receipt of critical care procedures and higher mortality rate seen among the uninsured compared to private patients in this study were most likely due to factors occurring within each care setting.

With expansion of insured care under the Affordable Care Act , one might expect greater utilization of critical care services and even a decrease in observed mortality among the critically ill. However, in a comprehensive analysis of Massachusetts data before and after healthcare reform compared to four states that did not enact reform (New York, Washington, Nebraska, and North Carolina), no difference was noted in ICU utilization, discharge destination, or hospital mortality [83]. However, the number of critically ill patients with insurance increased, as was expected. The authors cited several hypotheses for their lack of observed mortality difference, including the unique patient demographics of Massachusetts where only 9 % of patients were uninsured compared to a national average of 17 %. The population of Massachusetts also has a higher baseline socioeconomic status and less racial and ethnic diversity compared to the rest of the nation. Another possibility is that the association between lack of health insurance and mortality observed in prior studies may have been due to other unmeasured factors for which lack of insurance served as a proxy (i.e., poverty). These unmeasured factors may not have changed immediately following insurance expansion. In a study comparing 5 years before Medicaid to have expansion to 5 years after expansion in New York, Maine, and Arizona were found to have a relative reduction in all-cause mortality of 6.1 % or 20 deaths per 100,000 adults compared to neighboring states that did not undergo Medicaid expansion [84]. Mortality reductions were greatest among older adults, nonwhites, and residents of poorer counties.

In 2008, Oregon used a lottery system to select from those individuals waiting for Medicaid expansion. About 2 years after the lottery, 6387 adults who had been selected to apply for Medicaid coverage were compared with 5842 adults who had not been selected [85]. Medicaid coverage was associated with a decreased likelihood of a positive screen for depression (−9 %, 95 % CI: −17 to −1.6 %, p = 0.02), increased use of many preventive services, and nearly complete elimination of potentially catastrophic out-of-pocket medical expenditures. Among those covered by Medicaid, more cases of diabetes were diagnosed (3.8 % increase, 95 % CI: 1.9–5.7 %) and a higher proportion of patients were using diabetes medications (5.4 % increase, 95 % CI: 1.4–9.5 %). There was no difference in the use of medications for hypertension or hyperlipidemia, nor was there a difference in average glycated hemoglobin levels. While physical health failed to improve in the first 2 years following expanded Medicaid coverage, there was greater healthcare utilization and reduced financial strain.

Work may offer many people a sense of accomplishment and well-being, but for some individuals, work can have deleterious effects on health [86]. Over the last half-century, women have had increased representation in the labor force. Despite this increased representation, African American and white women have had significantly different work trajectories. Black women are more likely to work in jobs with lower earnings, little room for advancement, and high risk of termination [87]. Using data from the National Longitudinal Survey of Mature Women, Shippee et al. found that Black working women who had felt that their work had progressed in the past 10 years had a 24 % lower mortality risk compared to those who felt that their work was static or had regressed. This result persisted despite adjusting for personal demographics, type of occupation, health characteristics, family life, and personal and household wealth. We were unable to find any studies exploring the effects of work trajectory on critical care outcomes.

Income has a significant effect on health outcomes. Low income is associated with low birth weight, poor educational outcomes, unemployment, work disability, lack of medical insurance, increased medical expenditures, smoking, and sedentary activity [88]. It should come as no surprise, therefore, that income inequality is also associated with differences in all-cause age-adjusted mortality [89]. Across Europe, countries with a lower proportion of their population in relative poverty have higher average life expectancies [90].

Bein et al. prospectively administered a questionnaire that assessed patient socioeconomic status (level of education, occupation, income, marital, and health insurance status) to the surrogates of 1006 patients in a 24-bed surgical ICU of a tertiary hospital in Germany [91]. They found patients of lower socioeconomic status had a higher adjusted odds for ICU length of stay and a lower adjusted odds for visits from friends and family compared to patients with higher socioeconomic status. This result has not been replicated in the United States. In the previously mentioned study involving multiple hospitals in California, Erickson et al. found that socioeconomic status (and higher admission severity of illness) attenuated the increased ICU length of stay identified in African Americans [77]. This demonstrates the importance of including multiple patient level factors in disparities studies.

In spite of the previously mentioned associations between low income and increased mortality, a retrospective study of 38,917 patients admitted to either of two academic medical centers in Boston between 1997 and 2007 found that the percentage of census tract residents below the federal poverty line was not associated with all cause 30-day, 90-day, or 1-year mortality [92]. It was also found to not be associated with 90- and 365-day mortality postcritical care initiation. The study did not include severity of illness information based on physiologic parameters but did include comorbid conditions.

In summary, several factors may confound the relationship between race, ethnicity, age and gender, and health outcomes, including genetic predisposition, geographic location, chronic illness, access to care, and socioeconomic status. Such factors should be adequately addressed in any study of health disparities before valid conclusions can be made.

The emergency department remains the primary source of ICU admissions. Despite this, there is a paucity of data regarding emergency department care of the critically ill [93]. To our knowledge, there are no national databases tracking critically ill patients in the emergency department. This leads to an inability to accurately assess the proportion of emergency department patients that are critically ill, the quality of the care that is delivered to them, and how our care delivery in this setting has changed over time.

Even with the lack of national emergency department data on critical illness, important work in health disparities has been performed in the emergency department. A seminal study in health disparities was conducted in an emergency department in Los Angeles in 1993, involving chart review for 139 Hispanic and non-Hispanic white adult patients who presented to the emergency department of a level I trauma center with isolated long-bone fractures [94]. Hispanics remained more likely than non-Hispanic whites to not receive pain medication for their acute fractures after adjustment for several patient and physician characteristics (odds ratio: 7.5, p < 0.01). A follow-up study in the same setting found that despite Hispanics receiving less analgesia they did not differ from non-Hispanic whites in their delineation of pain and their physicians rated their pain similar to non-Hispanic whites [95]. This suggests that other factors were responsible for lack of an equitable receipt of analgesia among Hispanic patients.

Factors affecting triage of critically ill patients in the emergency department may also influence disparities, and prehospital therapy may influence outcomes in critical illness by affecting appropriate triage. In one study of patients with sepsis, patients that arrived by ambulance had a higher likelihood of receiving immediate care (including a shorter time to first antibiotics and a shorter time to initiation of early goal directed therapy) compared to “walkins” [96]. For patients who are unable to afford the cost of an ambulance and instead present to their local emergency department by their own means, this can significantly affect their survival. How this further modifies existing racial, ethnic, gender, and age-related disparities is unclear.

Delayed transfer of patients from the ED to the ICU also has a significant impact in outcomes. A study conducted using the Project IMPACT database demonstrated increased ICU and inhospital mortality rates and prolonged hospital lengths of stay following ICU discharge for patients with at least a 6 h delay in ICU transfer from the ED [97]. Unfortunately, few patient level demographics were available to look for associations with health disparities. However, similar findings were demonstrated in a Brazilian study [98]. Such delays are not uncommon and are related to availability of critical care beds in the same institution, the need for interhospital transfer due to need for higher level resources, and physician and nursing staffing in the emergency department and ICU.

The intensive care unit is the setting for the majority of studies evaluating health disparities among the critically ill. Common conditions encountered in the ICU have received significant attention.