Are local public expenditure reductions associated with increases in inequality in emergency hospitalisation? Time-series analysis of English local authorities from 2010 to 2017

Background Reductions in local government funding implemented in 2010 due to austerity policies have been associated with worsening socioeconomic inequalities in mortality. Less is known about the relationship of these reductions with healthcare inequalities; therefore, we investigated whether areas with greater reductions in local government funding had greater increases in socioeconomic inequalities in emergency admissions. Methods We examined inequalities between English local authority districts (LADs) using a fixed-effects linear regression to estimate the association between LAD expenditure reductions, their level of deprivation using the Index of Multiple Deprivation (IMD) and average rates of (all and avoidable) emergency admissions for the years 2010–2017. We also examined changes in inequalities in emergency admissions using the Absolute Gradient Index (AGI), which is the modelled gap between the most and least deprived neighbourhoods in an area. Results LADs within the most deprived IMD quintile had larger pounds per capita expenditure reductions, higher rates of all and avoidable emergency admissions, and greater between-neighbourhood inequalities in admissions. However, expenditure reductions were only associated with increasing average rates of all and avoidable emergency admissions and inequalities between neighbourhoods in local authorities in England’s three least deprived IMD quintiles. For a LAD in the least deprived IMD quintile, a yearly reduction of £100 per capita in total expenditure was associated with a yearly increase of 47 (95% CI 22 to 73) avoidable admissions, 142 (95% CI 70 to 213) all-cause emergency admissions and a yearly increase in inequalities between neighbourhoods of 48 (95% CI 14 to 81) avoidable and 140 (95% CI 60 to 220) all-cause emergency admissions. In 2017, a LAD average population was ~170 000. Conclusion Austerity policies implemented in 2010 impacted less deprived local authorities, where emergency admissions and inequalities between neighbourhoods increased, while in the most deprived areas, emergency admissions were unchanged, remaining high and persistent.


What is an AGI?
An AGI is an Absolute Gradient Index of inequality.It is the modelled gap in outcomes between the most and least deprived neighbourhoods in the country.You can use any outcome that can be broken down by deprivation groups, for example, avoidable emergency admissions, GPs per population, rates of early stage cancer diagnosis, waiting times for psychological therapies, and many others.
Why a "modelled" gap rather than an "actual" gap?
Because it summarises information about everyone in the whole population, not just a few people at the two extremes.Actual gaps between extreme groups can be unstable and misleading.
Why the most and least deprived "in the country" rather than "in my local area"?Because local inequality in your area can then be compared with local inequality in other areas on a like-for-like basis.For instance, the most deprived fifth of people in Buckinghamshire are considerably better off than the most deprived fifth in Liverpool.Inequality based on national deprivation rank can be compared between local areas, but not inequality based on local deprivation rank.

How is the modelling done?
Simple linear regression at neighbourhood level.We take all the neighbourhoods (technically "Lower Layer Super Output" areas) within your locality (e.g.clinical commissioning group or local authority).We then graph an outcome versus the national deprivation rank for each neighbourhood, and draw a straight line through those points considering the local population size (bigger neighbourhood have more influence on the direction of the line).That straight line is the "social gradient" in outcomes.The "modelled gap" is the difference between the top and bottom of the social gradient, representing the most and least deprived neighbourhoods in the country.

What about simpler equity measures?
Useful simple equity measures for non-specialists include: • The actual gap between best-and worst-off fifth (or tenth) based on national deprivation • The trend in outcomes for the worst-off fifth (or tenth) based on national deprivation

When is the AGI misleading?
The AGI is unreliable for localities with a narrow deprivation range (fewer than 3 quintile groups) such as Bradford City CCG, where almost everyone is in the most deprived national quintile group.
In such cases, it is more helpful to look at trends in the worst-off quintile group(s).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)      Note: This excludes capital expenditure and housing benefit, which is not a local government service but a cash benefit administered by local government under central government rules.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)   Indirectly age-sex standardised rates per 100,00 population BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  Indirectly age-sex standardised rates per 100,00 population BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

-
Equation multilevel linear model for predicting expected expenditure -Command for estimating the multilevel linear model for predicting expected expenditure mixed expenditure-per-capita year || LAD_id: year, mle covariance(unstructure) -Example: -Command for estimating the fixed effects panel linear regression: xtset LAD_id year xtreg emergency-admissions need-adjusted-expenditure-change##IMD15_quintile_group, fe vce(robust) -Test of correlation between outcome variable and error term

Figure
Figure S1 Local Authority Service Expenditure by Category, 2017-18

Figure S2 -
Figure S2 -Trends in Absolute Gradient Index of Inequality (AGI) between neighbourhoods within a local authority for (I) avoidable and (II) all-cause emergency admissions, by local authority deprivation quintile.

Figure S3 -
Figure S3-Trends in local government expenditure per head for (I) total, (II) services, and (III) social care, by local authority deprivation quintile.

Figure S4 -
Figure S4-Comparison of avoidable emergency admission rates for years 2010 and 2017 by quintile groups of deprivation of lower super output areas (LSOA) for (I) most deprived LADs, and (II) least deprived LADs.

Figure S5 -
Figure S5-Comparison of all-cause emergency admission rates for years 2010 and 2017 by quintile groups of deprivation of lower super output areas (LSOA) for (I) most deprived LADs, and (II) least deprived LADs.

Figure S6 -
Figure S6-Maps for 2017 for (I) LADs deprivation, (II) Total expenditure reductions, (III) Standardised rates of any avoidable admissions per 100,000 population, and (IV) Absolute gradient index of inequality for any avoidable admissions.