We investigated potential sex differences in a convenience sample of hospitalized patients during the first wave in the Netherlands. Given the specific cultural and health policy approach to intensive care (29), we specifically looked at patterns of ICU care preferences between female and male patients. In our study population, male patients were significantly more likely to need ventilatory support, be admitted to the ICU, and die in the hospital. Around a fifth of the patients died, which is consistent with other studies (30, 31). While we identified no sex differences in the prevalence of most risk factors, such as age, LDH level, and lower peripheral oxygen saturation, shorter symptom duration and an elevated number of comorbidities were associated with in-hospital mortality only in female patients.
Approximately a third of the patients had a non-ICU policy. People with a restricted policy were significantly older and had more comorbidities. There was a significant rise in non-ICU policies in female patients in the second half of the first wave compared to male ones, which could not be explained by age or comorbidities.
Our study confirmed the previously reported role of higher age, elevated LDH, lower diastolic blood pressure and lower peripheral oxygen saturation as predictors of in-hospital mortality in both male and female patients (22, 23, 25, 32–39). However, shorter symptom duration before hospitalization and more comorbidities were associated with in-hospital mortality in female patients only, contrary to most other studies that emphasized the role of comorbidities as risk predictors for COVID-19 mortality primarily in male ones (34, 40, 41). In our study population male patients reported a higher prevalence of cardiovascular comorbidities, while female patients were affected more often by autoimmune conditions or taking immunosuppressive medications. Overall, the role of immunosuppression might have been more relevant than cardiovascular risk factors, highlighting the need for a potential stratification of risk factors for prediction of severe COVID-19. Furthermore, different co-morbidities might play a different role in female and male patients. Although SARS-CoV2´s main target appears to be the vascular rather than the pulmonary system (42), the influence of the immune system is most essential in the development of the disease. This might especially apply to female patients. The role of sex differences in the susceptibility and progression of COVID-19 has been reviewed elsewhere (43). In general, female patients appear to experience an immune advantage due to genetics, hormonal influences or differences in innate immunity (44). If these fundamentally protective mechanisms are impaired, it might affect the disease susceptibility in female patients more substantially than in male ones leading to the potentially higher mortality we have identified in our cohort.
Another aspect leading to these differences might be the reported disease duration before admission. In our population, symptom duration before admission was significantly associated with in-hospital mortality only in female patients. Overall female patients reported a slightly shorter symptom duration before admission than men; a phenomenon that has been previously associated with in-hospital mortality (45, 46). The identified pattern could be due to two distinct factors: delay in health seeking by male patients or delayed symptom development in female patients. Health seeking delay has been extensively reported in male patients (47, 48) and could potentially skew the reported symptom duration. However, symptom development might also be different in female and male patients. Sex differences in symptomatology have been reported in many conditions. Especially in the field of cardiovascular health, the absence of symptoms in female patients (49) has notoriously led to delayed diagnosis and avoidable mortality (50). Given that COVID-19 is a vascular disease with predominantly pulmonary symptoms, a similar phenomenon could be at play. A recent meta-analysis has confirmed the lower prevalence of symptomatic COVID-19 in female patients (3) and their overall prognosis could thus be different than the one of male patients reporting the same severity or duration of symptoms upon admission to the hospital. This potential difference should be investigated in large cohorts, as it could be an essential sex-specific consideration for patient triage upon admission.
We then focused on ICU care preferences in female and male patients. A third of the patients refused ICU care. This is a significant percentage of the hospitalized and a phenomenon possibly unique to the Netherlands and few other countries with open societal discussions about end-of-life (EOL) choices (51–53). The two main factors influencing this choice were older age and comorbidities (54). However, over the course of the first wave we saw a proportional increase in the female patients with a non-ICU policy and a decrease in male patients. The average age of male patients in the second half of the first wave decreased contributing to the lower proportional rate of non-ICU policies, however this age difference alone could not explain the identified sex differences. Given the lack of differences in co-morbidities, other factors might have been involved.
First, it appears as if a higher number of younger male patients were hospitalized in the second half of the first wave, as a potential consequence of more strict community-care triage of cases. Unfortunately, our current data does not allow the investigation of this aspects. Focusing on the female patients and the higher relative restricted non-ICU policy rates, we hypothesize the effect of multiple factors. Firstly, health literacy can impact choice of care. At the beginning of the pandemic there was a broad public debate about COVID-19, its disease course, and about what ICU care entails in the Netherlands. Given previous reports of higher health literacy in women (55), better understanding of ICU care and its consequences might have contributed to more reticence in accepting it. Second, gender differences have been reported in the attitude towards life-prolonging treatment options. For example, a previous study reported that more men chose to continue treatment if placed in a hypothetical coma with only a slight chance of recovery (56). Analogous reactions might have played out as primary care physicians discussed ICU options openly with their patients during the first wave (57). Third, gendered expectations might affect patients’ choices. At the beginning of the pandemic, the possibility of ICU bed shortage was openly discussed in the media. Gender stereotypes include the characterization of “femininity” as being caring, nurturing and self-sacrificing (58). These stereotypes might be one of the underlying reasons for the higher rates of female living organ donors (59, 60). These expectations could lead to a potentially higher willingness in women to forgo an ICU treatment to benefit younger patients. Fourth, gender-specific communication differences between physicians and patients could impact choices. Research on do not resuscitate (DNR) orders showed that overall female physicians were more inclined to have this discussion with their patients and female physicians were 1.5 times more likely to write DNR orders for their female than for their male patients (16). Patient-provider dyads might, thus, have had a potential influence on ICU care choices. Last, gender norms and relations might impact EOL preferences. Active choice to forgo invasive treatment with long-term consequences might be an informed choice rather than any expression of discrimination by the health care provider or system. Little is known about the underlying reasons for gender differences in EOL preferences, but current literature suggests more willingness to spontaneously discuss death by women with their healthcare providers (61) and more openness to discuss the emotional aspects of dying (62).
The current study offers a realistic perspective of hospitalization patterns in the Netherlands during the first wave, but some limitations should be acknowledged. In approximately 7% of the cases, we did not have information about mortality, mainly due to transfer of patients. However, the overall patterns recapitulate the situation in the country at the time, so this should not represent a meaningful source of bias. Second, the high non-ICU policy rates in the included population offered an ideal setting to study these patterns, but limited the ability to predict the potential outcomes of patients that might have experienced a positive outcome if given ICU care. Hence, the calculated risk factor associations might be influenced by a subgroup of patients that willingly chose to forgo ICU care. Last, we investigated the impact of sex, age and comorbidities on ICU choices but did not include other potentially influential factors, such as e.g. ethnicity, living situation/marital status and education (17–19). These data were not available in the current database and should be investigated in future research.
Perspectives and Significance
Overall, we confirmed the previously reported higher incidence of severe disease and mortality in male compared to female patients with COVID-19. However, we also identified sex differences in the impact of symptom duration before admission and markers of immune disfunction on prognosis. Given the potential prognostic value of the speed of symptom development and immune dysfunction in female patients, this should be specifically investigated to possibly inform sex-specific triage protocols for COVID-19. Last, sex and gender differences in ICU care preferences should be further investigated. Our data highlights some unexplained differences in ICU care policy, which cannot be solely explained on clinical grounds. Further studies are needed to evaluate the potential role of gender in care choices and active priority setting at the end-of-life (EOL). Furthermore, patient-provider communication at EOL in the context of health care urgency and restricted availability such as the COVID-19 first wave should be further explored.