Sepsis Epidemiology and Outcomes in Asia: Advancing the Needle

Sepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection (1), contributes to 11 million deaths annually, representing one-fifth of all causes of death (2). Severe and critical coronavirus disease (COVID-19), which are manifestations of viral sepsis, have further illustrated the burden, with an estimated 6 million deaths globally to date (3). In addition to mortality, sepsis contributes to significant short- and long-term morbidity (4, 5) and is associated with higher resource use and healthcare costs (6). In recognition of this burden, the 2017 World Health Assembly adopted a resolution (7) with an outline of key priorities and suggested action plans for member states.

In this issue of the Journal, Li and colleagues (pp. 1107–1116) report on the epidemiology and outcomes of sepsis in intensive care units (ICUs) from nations of differing macroeconomic wealth across Asia (8). In this point prevalence study conducted across 4 days from 386 adult ICUs across 22 Asian countries in 2019, the investigators collected a wealth of data on the processes and outcomes of care of patients with sepsis. The overall prevalence of sepsis, defined using Sepsis-3 (Third International Consensus Definitions for Sepsis and Septic Shock) criteria, was 22.4% and ranged from 20.9% in low-income and lower-middle-income countries (LICs/LMICs) to 24.5% in upper-middle-income countries. Respiratory, intraabdominal, and urinary sepsis were the most common causes, and more than half of pathogens were gram-negative bacteria. Most patients (>70%) received mechanical ventilation. More than one-third of patients had died by 90 days, with higher adjusted mortality in LICs/LMICs than in high-income countries. The authors also report adherence of 21.5% for the 1-hour sepsis bundle and 36.6% for the 3-hour bundle, as recommended by the Surviving Sepsis Campaign (9). Of the bundle elements, delays in antibiotic administration beyond 3 hours and absence of antibiotic administration within 24 hours were associated with higher 90-day mortality.

Several insights emerge from this study. First, the reported overall prevalence of sepsis was lower than in a recent study from India that used similar methods (10) and found a Sepsis-3–based prevalence of 33%. The lower prevalence in the present study may be due to different inclusion criteria, types of participating sites, or interpretations of the sepsis case definitions, or it may represent an underestimate of the true prevalence across Asia.

Second, although overall compliance with sepsis bundles was low, the highest compliance was in LICs/LMICs that had the highest adjusted mortality, a finding that is counterintuitive. Although the authors have postulated reasons such as the higher proportion of emergency department admissions in LICs/LMICs, the primary driver of compliance in LICs/LMICs appears to be the higher use of antibiotics. The reason for higher early use of antibiotics may reflect heightened awareness of worse outcomes from delays in antibiotic administration in the context of a high prevalence of multidrug-resistant organisms (10) in LICs/LMICs. Alternatively, Time 0 may have occurred later in LIC/LMIC hospitals using paper-based records if ICU admission time was recorded more reliably than earlier times of emergency department triage or ward deterioration. Such a bias would lead to inflated estimates of bundle adherence in LICs/LMICs; however, the absence of an interaction between bundle adherence and county income class mitigates this concern. Finally, the association between country income status and bundle compliance may be spurious and related to collider bias due to inclusion of patients admitted to the ICU, as opposed to all patients with sepsis in the emergency department or hospital ward, regardless of final location of care. Thresholds for ICU admission are likely to vary among ICUs, given marked differences in ICU capacity among Asian countries (11).

Third, in contrast to what we might expect, the proportion of patients admitted with “tropical” infections was low in this study, possibly reflecting the predominantly urban location of participating sites and unmeasured seasonal variation despite spacing the 4 days of data collection throughout the year. Finally, epidemiologic differences and outcome disparities within countries with a substantial rural or semiurban population, which include many LICs/LMICs in this study, would be expected on the basis of differential public access to hospitals. Elucidating outcomes in low health resource areas disconnected from urban academic centers, where routinely collected health administrative data do not exist, may require alternative methods such as systematic death surveys and verbal autopsy (12).

The study by Li and colleagues has several strengths: the large sample size, the effort to build a hospital cohort representative of the population, careful consideration of causal pathways to design analyses, and the clear insight into the importance of early antibiotic therapy alone as the bundle element associated with lower mortality. It also has limitations. Participating sites were predominantly urban and academic and may have been more likely to be high performing. Like other point prevalence studies, the data represent a snapshot in time, and estimates of prevalence are likely conservative because of the short duration of illness due to either recovery or death of incident cases of sepsis. Included patients were in ICUs and likely represented the minority of hospitalized patients with sepsis. In addition, patients had an admission diagnosis of sepsis, which excludes those developing sepsis in the course of their ICU stay. In settings that offer intensive treatments such as surgery for cancer or organ transplant, or that care for patients with traumatic injuries, sepsis developing after initial ICU admission for an unrelated reason is common. Globally, one in four cases of sepsis in hospitals and one in two cases of sepsis in the ICU result from healthcare-associated infection (13).

There remain several knowledge gaps in sepsis epidemiology and outcomes. Clinicians and policy makers interested in health system improvement need better mechanisms to measure the true incidence of sepsis, especially in semiurban and rural settings. One solution is the creation of health administrative data “from the ground up” by deploying clinical quality registries that integrate surveillance tools. More data are required on pathogen-specific clinical characteristics and outcomes, particularly for nonbacterial pathogens. The finding of benefit with early antibiotic administration requires more nuanced understanding of patient subgroups most likely to benefit and appropriate antibiotic choices to mitigate risks of antimicrobial overuse and resistance. Improving sepsis outcomes will require national investments in performance improvement for early recognition and treatment and in improving access to primary, secondary, and tertiary health care. In addition, sepsis incidence and outcomes act as a barometer of a country’s overall state of public health. Any improvements based on clinical care alone are likely to be small unless parallel actions improve sanitation and hygiene, access to clean drinking water, vaccination, and the burden of poverty.