Child Mortality in England after the Pandemic. Increasing Mortality and Inequalities

Data, Study Design and Population

The National Child Mortality Database (NCMD) commenced data collection on 1^st April 2019, and collects data from all Child Death Overview Panels (CDOPs) across England.(9) In England, all deaths of children (0-17 years old) are reviewed by CDOPs, and data collated by the NCMD, with death notifications required by statute within 48 hours. Data was downloaded on 14^th September 2023. The study population was derived from 2021 Office of National Statistics (ONS) census data(10), and includes measures of ethnicity, sex and age.

Outcome

Deaths of all CYP, who died before their 18^th birthday, and had been born at, or over, 22 weeks of gestation, occurring from 1^st April 2019 until 31^st of March 2023 were identified. In this work, similar to our previous work(4), in order to obtain a provisional category of death, all child deaths reported to NCMD were coded by 3 independent coders (all paediatricians) to identify the likely category of the cause of death. All coders recorded a provisional category of death (see below) or that there was insufficient information provided. For each death, if two or more coders agreed on a category this was taken as the most likely category and where no two coders agreed, the category highest in the following hierarchy was used (based on categorisation used by CDOPs(11), in order of priority: Trauma, Malignancy, Underlying Medical Condition, Intrapartum event, Preterm Birth, Infection, SUDIC (Sudden Unexpected Death in Infant and Childhood).

Covariates

In the initial notification to NCMD, the CDOPs report baseline characteristics of the child; from which the following data were derived: sex of individual (female, male, other (including not known)), age at death, ethnicity, and the child’s home postcode. The postcode was then mapped to a geographic measure, the Middle Layer Super Output Area (MSOA), which is derived at granularity of around 7200 people(12). This code was used to identify the CYP’s region of England, the rural/urban status, and a local measure of deprivation: the Index of Multiple Deprivation (IMD). The IMD is calculated using 7 main domains (income, employment, education, health, crime, access to housing and services, and living environment). The population split into 10 equal sized (by people) deciles 1 being most deprived and 10 least deprived(13). Ethnic Group was derived from NHS data, and is parent reported at source, and classified using ONS guidelines; Asian or Asian British (Bangladeshi, Chinese, Indian, Pakistani, Any other Asian background), Black or Black British (African, Caribbean, Any other Black background), Multi-racial (White and Asian, White and Black African, White and Black Caribbean, Any other Multi-racial background), White or White British (British, Irish, Any other White background, Gypsy or Irish traveller), and Other (Arab, Any other ethnic group).

Statistical Analysis

Deaths of children occurring from 1^st April 2019 until 31^st March 2023, born at, or after 22 weeks of completed gestation, were identified and divided into four 12-month periods; April to March (2019-20, 2020-21, 2021-22 and 2022-23). Initially, the demographics for childhood deaths, split by the year (starting April 1^st) were compared from reported data. Demographic comparisons were made using the Chi² test. Rates of overall death, and for each category, were derived for each year (starting April 1^st).

Next, the underlying population size and characteristics were derived from Office of National Statistics 2021 Census data. Number of deaths were compared across years using the Chi² statistics, and risk of death, for each analysis year (April-March) was calculated per 1,000,000 children, with counts collapsed to number of events per month. A Poisson regression model was used to test for an overall trend across the 4 years, along with the relative risk of death (per year), using counts of death collapsed to number per month. The number of likely additional deaths seen, between April 2020 and March 2023, compared to the number predicted from reference year of 2019-20 data, was derived using a generalised linear model (Poisson family). This was repeated for each provisional category of death. Next, a linear spline model was fitted allowing different trends for April 2019 to March 2021 (Period 1) and April 2021 to March 2023 (Period 2); periods before and after the COVID pandemic’s peak). A p-value was derived to test for a change of trend at this mid-point. This was repeated for each cause of death and the child’s characteristics (above) and interaction p-values (if relevant) were derived as evidence that the profile may differ by the stratified characteristic (e.g. sex).

Finally, to quantify changes in health inequalities across the 4 years, the relative risk of death for children living in the more deprived half of England (Deciles 1-5), was compared to those in the less deprived half (Deciles 6-10), for each of the four years of data, alongside tests for trend across the two a-priori time periods (as above). This analysis was then repeated, comparing mortality for children from non-white ethnicities, and each specific ethnic group, with those from white backgrounds.

This work was reviewed by the Chair of the Central Bristol NHS Research Ethics Committee who confirmed that NHS ethical approval, including obtaining individual consent, was not required. Data are presented as number (%), risk per 100,000 person year (95% CI), relative risk ratio (95% CI), or excess deaths (number) with 95% CI. All tests were two-sided. Reporting follows the STROBE reporting guideline for observational studies. Analysis was performed using Stata Version 17.

Patient and Public Involvement

Parent and public involvement guided the design and setting up of the NCMD at establishment and real-time child mortality surveillance system at the beginning of the COVID-19 pandemic.

Relative Risks are per year with 95% confidence intervals. Risk per month smoothed over 5 month period

A similar profile was seen for deaths from underlying disease, from infection, and from substance abuse across the time period (Infection, p<0.001; Underlying Disease, p<0.001; Substance Abuse, p=0.021) (Figure 2). In contrast, for deaths after Intrapartum events, an increasing risk in the first period (RR) was followed by a reduction in the second (RR) (p=0.004); and a similar, if weaker, pattern was seen with deaths after suicide (p=0.073).

• Download figure
• Open in new tab

Figure 2.

Risk of death per month (per 1,00,000 person years) and trends of risk across study period, split by category of death

Deaths from preterm birth, trauma and SUDIC, had increases in risk across the entire 4 year study period, with little evidence of a change in trajectory during the pandemic (Preterm birth, p=0.562; Trauma, p=0.358; SUDIC, p=0.281). There was no change in the risk of death from Malignancy overall (p=0.831), or any change in trend during the pandemic (p=0.947).

Relative Risks are per year with 95% confidence intervals. Risk per month smoothed over 5 month period

Repeating the analysis, split by child characteristics, suggested that patterns of changing mortality were not modified by sex (p=0.098), area of residence (p=0.176) or region of England (p=0.256). However, the profile did differ by age (p<0.001), with mortality clearly dropping and then rising for children between 1 and 4 years old (p<0.001). A similar pattern was seen in children under 1, and between 5 and 15 years, although evidence of an initial drop was weaker in both groups (Infants RR 0.97 (0.93-1.01); 5-15 years RR 0.95 (0.93-1.18)). The risk appeared to increase across the whole time period for children in the oldest age group (16 and 17 years old) (RR 1.08 (1.03-1.14)). The pattern of mortality also appeared to differ by ethnicity with children from Asian or Asian British (p<0.001), Black or Black British (p=0.012) and Other (p=0.003) groups having a measurable reduction in mortality across the first period, and then an increase in the second. For children from White and Mixed background, any possible trend did not appear to be affected by the pandemic, with a general risk increase across the 4 years for white CYP (RR 1.05 (1.03-1.07)) and only weak evidence of an increase for those from a Mixed background (RR 1.05 (0.99-1.12)). For measures of deprivation, there was evidence of a change in risk trajectory for the 3 most deprived quintiles across the pandemic, with a decrease in the first half, and rise in the second. Children in the 2 least deprived quintiles had similar trends throughout the 4 years, with a general increase for the least deprived group (Deciles 9/10, RR 1.06 (1.01-1.10), or little evidence of change (Deciles 7/8, RR 1.02 (0.98-1.07)).

Finally, the relative risk of dying in the more deprived half of England remained similar across the first two years of this study (p_trend=0.526) (Figure 3, S3 Appendix), but there was weak evidence that inequalities increased in the second half (p_trend=0.093), with the highest RR seen in the final year (2022-23; RR 2.47 (2.30-2.66)). The relative risk of dying in children from Non-white backgrounds, compared to White children, reduced over the 1^st period (p_trend=0.021), but then increased (p_trend<0.001), to, again, the highest RR seen across the 4 years (2022-23; RR 1.70 (1.59-1.83).

• Download figure
• Open in new tab

Figure 3.

Relative Risk of childhood death for children in the more deprived (compared to the less deprived) areas, and children of non-white ethnicities (compared to those from white background), by year

Limitations

This work is based on the ongoing data collection from the NCMD, and previous work has shown good validation and coverage(14). While we had some missing data on some demographics (e.g. ethnicity), data completeness was generally good, and the primary analysis of risks was based on the statutory reporting of deaths. We have used ONS data to derive the population at risk, and this was estimated at a mid-point of the study. Changes may have biased our estimates (e.g. significant, differential migration, during the pandemic), and consequently absolute risk in particular should be interpreted with caution. However, the population came from a recent National Census and numbers of deaths are relatively small; making changes in the denominators unlikely to change the conclusions here. Fortunately, child death in England is rare, particularly for causes such as suicide and substance abuse deaths, which may further limit interpretation. Finally, the category of death in this work is provisional, based on information available at 48 hours, and further CDOP investigations may modify this in some cases.

In Context

Overall, the number of deaths of CYP appears to be increasing, and is now above the level seen pre-pandemic(3, 4, 15), although four patterns across the 4 years appeared to exist. Two categories of death appeared to follow (and likely drive) this profile: deaths from infection and underlying disease. Both metrics are likely driven by a reduction in circulating pathogens, and public health initiatives to shield vulnerable CYP. These deaths are often in vulnerable CYP, either due to underlying health problems, or in vulnerable social groups; and the transient reduction in deaths across the pandemic appeared to have benefited these individuals most, reducing health inequalities across England. However, any benefits, temporary resources or learning from the pandemic, which helped reduce deaths of these children, appear to have been lost. Changes in behavior (e.g. enhanced hand washing)(16) or wider, state-based enhanced social support, which benefited our most vulnerable families, appear to be not delivering their previous benefits. As well as medically vulnerable children, we know that underlying mortality is higher in those from most deprived areas(17), and for some minority ethnic groups(18, 19). The benefits to these groups appeared to be the largest seen in this work, but again, any benefit appears to have disappeared, with rising mortality in those groups where, we saw transient reductions in health inequities.

Two further profiles of mortality were seen in this work. For two conditions (suicide and intrapartum events) we saw an increase in the number during the pandemic, and then a reduction leaving it. For both of these categories, a lack of data prior to 2019-20 limits the interpretation, and the trajectory of these deaths, had the pandemic not occurred, is unclear. We know suicide in CYP has risen in many countries over the last 10 years(20), and the apparent peak during lockdown in this work may represent the existing trend, or an effect of the social restrictions(21). However, the subsequent reduction was unexpected and further close monitoring is essential. For intrapartum events, the increase in deaths was unexpected, on a background of national initiatives to reduce brain injury, and death, around birth (e.g. the Each Baby Counts program)(22). Clearly, the subsequent reduction back to pre-pandemic levels is welcome, and may represent a lag in these initiatives embedding in practice. Perhaps the most surprising findings here were the conditions where, despite COVID, lockdown and broad societal change, little impact of the pandemic was apparent. Apparently indifferent to the pandemic impacts, we saw a general rise across the 4 years for deaths caused by preterm birth, trauma and SUDIC. Death by trauma is now the 4^th most common cause of childhood deaths in England, and while changes in practice around preterm birth(23); may influence some of the trend seen here, death after preterm birth continues to increase to the highest level since NCMD started collecting data, and remains a substantial contributor to the further widening of health inequalities in England(18, 24).

Finally, while malignancy, a common cause of childhood death, continues to show little, to no, overall influence from the pandemic, or generally over the time-period studied. Despite concerns related to later diagnosis or initiation of treatment, deaths from malignancy (the 3^rd highest category in this work) remains similar to 2019-20 levels, although we do know that this group was the one of the most vulnerable(2, 25). As with preterm birth, since the underpinning risk of death in this group is high(26) the effect of the pandemic may not have been considerable enough to be identified.

Conclusion

Any mortality gains, often to vulnerable children, seen during the wide-ranging social changes put in place to reduce the spread and impact of COVID-19, have now dissipated. Overall child mortality in 2022-23 in England is now higher than before the pandemic. Three causes of death, trauma, SUDIC and preterm birth, appear to have continued to increase in incidence over the 4 years, with only Suicide and Intrapartum event deaths showing evidence of a reducing trend (although both of which may have increased during the pandemic). For all other causes of death, rates are either static, or worsening. In addition, the reductions in health inequalities seen during the pandemic in the most deprived children in England, and in our highest risk minority ethnic groups, have now reversed, with higher measures of inequalities than before the pandemic, for most vulnerable groups.