Socioeconomic deprivation and ethnicity inequalities in disruption to NHS hospital admissions during the COVID-19 pandemic: a national observational study

Introduction Hospital admissions in many countries fell dramatically at the onset of the COVID-19 pandemic. Less is known about how care patterns differed by patient groups. We sought to determine whether areas with higher levels of socioeconomic deprivation or larger ethnic minority populations saw larger falls in emergency and planned admissions in England. Methods We conducted a national observational study of hospital care in the English National Health Service (NHS) in 2019–2020. Weekly volumes of elective (planned) and emergency admissions in 2020 compared with 2019 were calculated for each census area. Multiple linear regression analysis was used to estimate the reductions in volumes for areas in different quintiles of socioeconomic deprivation and ethnic minority populations after controlling for national time trends and local area composition. Results Between March and December 2020, there were 35.5% (3.0 million) fewer elective admissions and 22.0% (1.2 million) fewer emergency admissions with a non-COVID-19 primary diagnosis than in 2019. Areas with the largest share of ethnic minority populations experienced a 36.7% (95% CI 24.1% to 49.3%) larger reduction in non-primary COVID-19 emergency admissions compared with those with the smallest. The most deprived areas experienced a 10.1% (95% CI 2.6% to 17.7%) smaller reduction in non-COVID-19 emergency admissions compared with the least deprived. These patterns are not explained by differential prevalence of COVID-19 cases by area. Conclusions Even in a healthcare system founded on the principle of equal access for equal need, the impact of COVID-19 on NHS hospital care for non-COVID patients has not been spread evenly by ethnicity and deprivation in England. While we cannot conclusively determine the mechanisms behind these differences, they risk exacerbating prepandemic health inequalities.

Note: the maximum of each quintile is larger than the minimum of the next quintile because the percentages in the table are calculated using rounded sums to prevent disclosure rather than the raw sums used to compute the quintiles Regression specifications is the outcome variable of interest for Middle Layer Super Output Area (MSOA) in week . is an indicator for the COVID period. is an indicator for socio-economic deprivation quintile .
is an indicator for ethnicity quintile . is a vector of week fixed effects. is MSOA characteristic .
is the MSOA COVID rate, lagged by one week.

Coefficient transformations
To calculate the percentage point fall for each quintile from specifications (5) and (6) we sum the implied COVID period coefficient, relevant quintile coefficient and an adjustment term to account for the different base group when both sets of quintile indicators are included. This adjustment ensures that the omitted group when interpreting the deprivation quintile coefficients is the lowest deprivation quintile with the average (for the group) percent ethnic minority quintile and vice versa. The implied COVID period coefficient is the difference between the mean week fixed effect during and before the COVID period. For specification (5) this is -0.3805 for elective admissions and -0.2164 for emergency admissions. The adjustment term is based on the joint distribution of the two groups of quintiles. For calculating falls for socio-economic deprivation quintiles, the adjustment term is equal to the average ( ) ( ) and for ethnicity quintiles, the adjustment term is For specification (5) this is -0.0164 for the socio-economic gradient in elective admissions, -0.0138 for the socio-economic gradient in emergency admissions, 0.0069 for the ethnicity gradient in elective admissions and 0.0312 for the ethnicity gradient in emergency admissions. Confidence intervals on the percentage differences between different percentage points changes are calculated using the delta method.
To calculate the approximate absolute change in elective and emergency admissions in the discussion section, we multiply the coefficients for the fifth quintiles from specification (5) by the number of admissions for the same quintile in March to December 2019. For elective admissions, this was 1,606,000 for the fifth socio-economic quintile and 1,452,000 for fifth ethnicity quintile. For emergency admissions, this was 1,332,000 for the fifth socio-economic quintile and 1,005,000 for the fifth ethnicity quintile. We then divide by the population of the quintile, which was 11,258,170 for the fifth socio-economic quintile and 11,257,630 for the fifth ethnicity quintile. 1

Robustness tests
Income deprivation and ethnicity deciles Same specifications as (5) and (6) but and refer to socio-economic and ethnicity deciles respectively.   (5) and (6), but using 2011 census data to classify the ethnicity quintiles.

Additional local need controls
Specifications are the same as (5) and (6), but all include the MSOA's population mean age in 2019 ( ) squared.

Alternative measures of local COVID rates
Number of COVID patients from the local authority ( ) and in the local hospital ( ) are defined analogously to the number of COVID patients in MSOA used in the specification (6). We map MSOAs to local authorities (2019 boundaries) using an ONS mapping (source to population data). We map MSOAs to a unique local hospital defined as the hospital that received the most emergency admissions from the MSOA in 2019.

Results split by diagnosis type
Only primary diagnoses with at least 100 emergency admissions in 2019 are included. We classify primary diagnoses in two ways. First by their in-hospital mortality rates (controlling for patient age year dummy variables and sex), with diagnoses classed as high if their mortality rate is above the 75th percentile. Second by their deferability status, following the methodology of Card, Dobkin and Maestas (2009 (6) High (6) Deferrable (6) Non-Deferrable (6)