It is well described that there are more male than female deaths in most analyses of mortality data in infants and children. However several recent studies have suggested greater female than male mortality in intensive care 22–25. Given that a sizable proportion of childhood deaths in the UK occur in PICU, we deemed PICU admissions a relevant cohort in which to examine potential sex differences in mortality. Our study differs from previous published literature in that it is the largest study in which the primary aim is to address the sex mortality difference in PICU, with careful consideration to the causal relationships between common variables and their relationship to the outcome. In particular, by drawing causal graphs connecting measured and unmeasured relevant variables, we clarified the data and conditions required to draw causal estimates of the direct effect of sex on mortality in PICU and highlighted the limitations of the current analysis.
The main finding from our analysis is that female sex is associated with higher mortality in PICU for infants up to 12 months of age. This finding is supported by strong evidence from both analysis methods we have used. There was no evidence of effect modification by the subgroups of infants under/over the age of 56 days, thus suggesting no clear effect of primary immunisation, or with a primary diagnosis of infection. When compared to the relative mortality in the general population where male infants are reported to have higher mortality, both the CSHR and the OR are in the opposite direction. Higher odds/hazard of mortality for female infants in PICU persisted across all subgroups of infants.
Over the 11-year duration of the study we observed a decline in PICU mortality for both sexes, but with a sharper decline for females. However, the sex mortality difference between females and males that we have described remained.
Using simple probability calculations, the higher number of male admissions to PICUs, together with the lower rate of male infant deaths in PICU leads to posterior odds of 0.79 when assessing PICU deaths. This is still overall in favour of female survival and consistent with the mortality rate ratio calculated in the overall population of 0.82 based on published national statistics figures (ONS death summary tables of 2016, https://www.ons.gov.uk/) for under one year of age female to male mortality per 1000 live births. Hence the data we report do not contradict the numerically greater overall mortality of males (see calculations of posterior probabilities in supplementary materials). Rather, it warrants further investigation into the reasons for the greater proportion of males admitted and the greater mortality of females in PICUs.
Our results are in line with four other studies, two of which indirectly investigated the effect of sex on mortality and others observing this effect as a secondary finding, in children with varying age ranges. These studies were carried out in the United Sates, 2011 22, Spain, 2015 23, Sweden, 2017 24, and the UK 25. The US study had a large sample size of over 80,000 cases which included children up to 18 years. It was a multi-centre study with over 31 PICUs and spanning 2005-2008. The focus of that study was to assess if ethnicity had an effect on PICU mortality and the effect of being female, which was only used as an adjustment confounder, was estimated as 12% higher odds of PICU mortality over males (p=0.019) 22. The Swedish study analysed data for 21,972 children over an eight year period (2008-2015), and included all PICU admissions in Sweden <16 years. They reported a sex difference in the HR of mortality of 0.91 for boys, p=0.035 24. The Spanish single centre study of children 0–18 years admitted to PICU over a period of three years showed an in-hospital mortality advantage of males over females (3.3% versus 4.9%, p=0.042)23. The UK study, with a sample size of 154,667 and aged <18 years, investigated mortality in PICU for children with and without life-limiting conditions. Sex was used as an adjustment variable with a mortality odds ratio 1.09 (p=0.002) for females compared to males 25.
All these studies were conducted in high-resource settings with access to advanced critical care facilities for children. In these studies, gender was primarily used as an adjustment variable. Our current study confirms that the higher PICU mortality rate for females is consistent with that previously described over multiple healthcare settings i.e. socialised systems and private healthcare facilities. Strengths of our study are that it directly addresses the impact of sex on mortality and infants and it spans a longer duration than the other studies, with a primary focus on sex differences in mortality.
It is unclear what is driving the greater mortality risk for female infants in PICU. One hypothesis is that there are biological differences in the response to critical illness or in the response to interventions. Another implication from our study is that the quantification of severity of illness using PIM2R, which is calculated within the first hour of admission to PICU and is calculated independent of sex, should be recalibrated to account for sex. Since the PIM2R score had similar distributions between male and female infants, it is possible that female infants have a different trajectory of illness. Males and females may present with the same PIM2R score but at different stages of their illness, with males admitted to PICU earlier in the course of their illness.
Limitations
Our study does not address any potential explanations for the observed sex difference in mortality as we did not have access to data factors that may influence PICU admission, such as maternal or birth factors.
PICANet data collection is carried out by individual PICUs and thus may vary in data collection. However, data collection is closely audited and sex and outcome are expected to be consistently recorded.
In addition, data is censored beyond discharge so it is not possible to assess if the mortality difference continues beyond discharge from PICU.