Sex differences in COVID-19 mortality in the Netherlands

Data on COVID-19 cases were extracted from the national infectious disease mandatory surveillance database. On January 28, COVID-19 was classified as a mandatory notifiable disease and since then data on patient demographics, comorbidities, source and contact monitoring, hospitalization and possible fatal outcome had to be reported by regional public health services for every laboratory confirmed COVID-19 case. Nursing home residency was estimated based on age and postal code. A confirmed COVID-19 case was defined as a person with a positive PCR result for SARS-CoV-2. The first case in the Netherlands was reported on February 27th. The initial instructions by the Dutch National Institute for Public Health and the Environment (RIVM) was to test hospitalized patients with COVID-19-like symptoms and symptomatic people with a travel history in risk areas or contact history with a laboratory confirmed patient. Early on, this was extended to pneumonia patients in whom no causative agent could be detected and who did not respond to treatment. Due to the rapidly increasing incidence of cases and the scarcity of diagnostic tests, from March 13th testing policy was restricted to suspected cases above 70 years of age and persons with a probable chronical illness, health care workers and patients admitted to the hospital [20]. In clusters of cases, such as in nursing homes, only the first cases were confirmed by PCR. Over time, test capacity grew and testing criteria were broadened and since June 1st, all Dutch citizens can be tested for SARS-CoV-2 in drive-through testing facilities if they have COVID-19-like symptoms.

All SARS-CoV-2 positive cases registered in the national infectious diseases electronic notification database until May 31st 2020 were included for analysis. After this date, as a result of the change in testing policy, data were not comparable to the period before May 31st. Data were last updated on August 22th 2020, to include additions or corrections that could have been made to these notifications afterwards. Healthcare workers were excluded from the analysis since they were over-represented in the data due to the testing policy and are not representative of the sex distribution in the general population. Also, notifications with missing data on sex or age were excluded. Due to the large amount of COVID-19 cases, registration requirements were diminished and data on comorbidities were no longer available for all patients after April 10th 2020. Analyses are performed for the study population in total, for the subgroup of hospitalized patients and for a subset of cases registered before April 10th with complete data on comorbidities.

Whether a notified case had died was reported by the regional public health services. However, the time that cases were followed-up after initial notification, to check for disease outcome, might have differed between regions. Some, but not all regions updated their notifications with information on deaths in the vital registries of municipalities.

Differences between males and females for categorical baseline variables were analyzed using the X²-test. Age was expressed in median (IQR) and analyzed with the Mann–Whitney U-test. Mortality within all notified cases (case fatality ratio, CFR) and mortality within hospitalized notified cases was stratified by 5 year age-groups and sex. The association between sex and mortality was assessed with logistic regression, expressed as an odds ratio (OR) and adjusted for age and nursing home residency as a proxy for frailty. For the subset of cases registered before April 10th, comorbidities were added to the model. The model was also fitted to both sexes separately to indicate differences in the contribution of comorbidities to mortality. To account for anticipated sex-related mortality rate differences in the general population, standardized mortality ratios (SMR) were calculated using indirect standardization by dividing the number of deaths in notified cases by the expected number of deaths in notified cases based on population mortality in 2019. SMRs were adjusted for age and reported as a male–female ratio by 5 year age group. 95% confidence intervals were derived through bootstrapping with 10,000 iterations.

To account for male–female population size differences, we additionally described the number of deaths per 100,000 inhabitants (population mortality rate), reflecting the per-capita risk of becoming infected and dying from infection. Population mortality rates were stratified by 5-year age group and sex. Additionally, male–female mortality rate ratio was calculated for nursing home residents. Male–female mortality rate ratios were compared with rate ratios of all-cause mortality, infectious diseases-specific mortality (ICD10: A00-B99), mortality by pneumonia (ICD10: J12-J18) and mortality by influenza (J09-J11) in the general population over the past 5 years (2015 to 2019). Population numbers for 2020 and population mortality numbers for 2019, stratified by age in 5-year intervals, were obtained from Statistics Netherlands [21]. In addition, the distribution of the absolute mortality over time was analyzed and depicted as the moving average over 14 days, stratified by age and nursing home residency. Data analysis was performed in R (version 4.0.0).

By May 31st 2020 46,507 cases of COVID-19 were registered, of which 16,906 cases concerned health care workers and were excluded from analyses. Furthermore, 62 cases with missing data on age or sex were excluded (0.2%), (supplement A). The remaining study population consisted of 29,539 cases of which 49.5% were males. Of these cases, 11,227 were hospitalized of which 62.9% were males. An overview of the main characteristics of the study population is presented in Table 1. The median age of notified COVID-19 cases was 78 [IQR, 60–87] in females and 71 [IQR, 56–81] in males. Most pronounced was the higher proportion of notified cases aged over 85 in females compared to males. Age-difference between males and females was less pronounced in the hospitalized subgroup. The proportion living in an institution (e.g. nursing home) was much larger in females than in males (37.4 versus 18.8%). Most frequently reported comorbidities were cardiovascular disease or hypertension (39.5%), chronic lung disease (18.6%) and diabetes (17.2%). There were no consistent differences between males and females in the presence of comorbidities.

Of all cases notified by May 31st, 6070 died (20.5%). Mortality among notified cases was 22.9% for males and 18.3% for females (OR of 1.33 [95% CI 1.26–1.41]). Of hospitalized cases, 2466 died (22.0%), also with a significant difference between males (23.5%) and females (19.4%); OR of 1.27 [95% CI 1.16–1.40]. For both males and females, the case fatality rate among all cases as well as among hospitalised cases increased with age (supplement B). For all age-groups, case fatality was higher among males than among females, for all notified cases and for hospitalized cases (Fig. 1). Odds ratios adjusted for differences in age were higher than the unadjusted odds ratios (Table 2). Supplementing this age-adjusted model with the comorbidities did not change the odds ratios. We found cardiovascular disease or hypertension, diabetes, chronic neuromuscular disease, renal impairment, chronic lung disease and malignancy to be independent risk factors for mortality, also for both sexes separately (supplement C). Adding nursing home residency to the model slightly increased the adjusted odds ratio for case fatality.

The standardized mortality ratio (SMR) among notified COVID-19 cases was 4.8 in males [95%CI 4.64–4.96] and 2.8 in females [95%CI 2.71–2.93] with a male–female ratio of 1.7 [95%CI 1.62–1.78]. Stratified by age, this male–female difference was only significant above one for age-groups above 80 years (Fig. 2). Remarkably, the point estimate of the SMR was highest in the 55–59 year-age group (1.48 [95% CI 0.97–2.48]) but significantly below one among 70–74 years old (SMR 0.81 [95%CI 0.70–0.94]).

The peak in total COVID-19 deaths was much higher for males than females and was reached towards the end of March, slightly earlier for males than females. The number of deaths decreased faster in males than in females and after April 8th, the daily number of reported deaths was slightly higher in females (supplement D). The majority of deaths in females consisted of nursing home residents, a group in which the number of deaths increased more slowly. The population mortality rate was 35.1 per 100.000, with a male–female rate ratio of 1.25 [95% CI 1.18–1.31]. Among nursing-home residents only, the male–female mortality rate ratio was 1.88 [95%CI 1.74–2.02]. When stratified by age, the absolute number of deaths was higher in males younger than 90 years than females younger than 90 (supplement E). Number of deaths in males was highest in the age group 75–80 years (739 deaths) and in females in the age group 85–89 years (663 deaths). Mortality rates increased by age in both sexes (supplement F). In all age-groups from 55 to 94 years, the population mortality rates were significantly higher (range 1.53 – 3.21) in males than females. In the age-group 95 + years (with low population numbers) and in the age-group 0–54 years (with few COVID-19 deaths), the rate was also higher in males, but not significantly so (Fig. 3). Between the ages of 55–69 and 75–94 years, male–female ratios for COVID-19 mortality were significantly higher than for overall mortality in the general population. Also compared to mortality due to any infectious disease in general, male–female ratio was higher, though this was only significant between the age of 75 and 89. Male–female discrepancies in COVID-19 mortality were less distinct from pneumonia mortality and only significant in the age-group 55–59 and 75–79 (Fig. 4, supplement F).