Understanding Sex-based Differences in Intensive Care Unit Mortality: Moving Beyond the Biology

Many observational studies have pointed to survival differences between critically ill men and women with varying directionality depending on the disease state (1–3). Women, as compared with men, have displayed worse outcomes in coronary artery disease, cardiac surgery, and after cardiac arrests (4, 5). However, women are more likely to survive than men with certain conditions such as chronic obstructive pulmonary disease and respiratory viral diseases (6, 7).

Although the interplay between the patient’s biological and immunological factors has been speculated as the potential unmeasured drivers for observed sex-based differences in care outcomes (8, 9), the potential contributions of systemic and implicit biases and cognitive errors have been largely understudied. Other factors historically overlooked, such as care environment, caseload, and team dynamics, are being independently investigated and increasingly recognized as important determinants of outcome, particularly for those at a high baseline risk of death (10–12). Similarly, exploring the variation in intensive care unit (ICU) mortality outcomes by the sex of the patient will require dedicated investigation that transcends the lens of acute physiology and comorbidity and other host factors.

In this issue of the Journal, Modra and colleagues (pp. 1353–1360) report their findings from a large, cross-sectional study of adult patients admitted to ICUs in Australia and New Zealand. Modra and colleagues took a deep dive into understanding variation in hospital mortality in men and women on the basis of how frequently a given condition occurred within each sex. The primary exposure variable was “sex balance”, defined as the percentage of patients in a diagnostic group who were women, and the primary outcome was sex difference in adjusted hospital mortality by ICU admission diagnosis. The study was large, encompassing over 1.4 million ICU admissions between 2011 to 2020 in the ANZICS APD (Australia and New Zealand Intensive Care Society Adult Patient Database) (13). Using mixed-effects logistic regressions, the authors adjusted for severity of illness, hours of hospitalization before ICU admission, and year of admission, with hospital site as a random effect.

The key findings were that women displayed better risk-adjusted survival than men in sepsis, respiratory disorders, and in the combined category of metabolic/renal and hematological disorders. On the other hand, women fared worse than men in burns and cardiovascular disorders, with the most marked sex difference observed in the cardiac surgery diagnosis category. The most striking finding was the inverse association between sex balance (defined as the percentage of patients in a diagnostic group who were women) and sex differences in mortality: in diagnostic groups in which there was a lower percentage of women, the women were more likely to die and vice-versa for men. A key strength of the study is the consistency with which this inverse association was observed across a variety of sensitivity analyses, that included dividing the population by hospitals rather than admission diagnosis and adjusting for baseline limitations of care. The authors also evaluated whether patients presenting with less common illnesses spent more time in the hospital before being admitted to the ICU—and they did. This raises the question as to whether preconceived notions of lower probability conditions in a given sex contributed to delayed recognition by clinicians and whether those delays contributed to worse outcomes. Additional strengths of the study were the completeness of the dataset, which included over 90% of all ICU admissions in Australia and New Zealand with minimal missing data, and the quality of the data extracted.

The findings by Modra and colleagues should give pause to clinicians, educators, researchers, and trialists as they imply that variation in ICU mortality is not solely explained by biology but perhaps, because of systemic, implicit, and cognitive biases. Notably, these biases are hurting our patients when they are diagnosed with a condition more prevalent in the opposite sex. The authors clearly show us the limits of our heuristic algorithms and when cognitive biases (e.g., ascertainment, anchoring, and availability biases) are more likely to influence our decision-making, contributing to a delay in diagnosis (e.g., longer pre-ICU admission stays), delay in care (e.g., higher severity of illness on ICU admission), and ultimately resulting in higher adjusted mortality within that diagnosis.

Implicit and cognitive biases in medicine and their impact on patient care and outcomes have long been recognized (14). Despite this, long-term solutions to mitigate bias are still lacking. Given their pervasive nature, successful interventions will entail a multifaceted approach at the individual clinician degree (encouraging individual education and role-modeling) and the institutional degree (increasing workforce diversity and providing rapid solutions to tackle discriminatory behavior). Prior work in medicine and other fields has shown the benefit of diverse teams, especially when dealing with complex tasks such as patients who are critically ill. Diverse teams bring diverse heuristics, knowledge, and representation, resulting in higher quality research and enhanced recruitment into clinical trials (15). Beyond organizational change, there are opportunities within medicine to develop objective, sex-agnostic clinical tools such as biomarkers for pain (16).

There are several notable limitations to the study that should be considered. The study only included patients in the ICU and could not comment on pre-ICU care. In addition, who is admitted to the ICU is complicated and hospital-dependent. Lastly, the study was limited to a binary definition of sex and was unable to study transgender and nonbinary patients in the ICU; these sexual and sex minority groups may experience even more profound medical marginalization, poor representation in or complete exclusion from clinical studies, and as a result, worsened health outcomes (17).

Modra and colleagues have shone a critical spotlight on an insidious problem in medicine that should galvanize the medical community to design solutions and further studies in this field. Moving forward, this study raises urgent questions surrounding the influence of cognitive and implicit biases on medical decision-making for the sexual- and sex-diverse critically ill patient populations and the patients who are not critically ill who are hospitalized or presenting across other clinical domains.