9.1 Introduction
Patterns of Emergency Department (ED) attendance in the Region are highly geographically variable. This study had two aims. Firstly to describe this variation in detail and secondly to explore two commonly cited correlates of attendance, those of deprivation and proximity to hospital. As can be seen from the maps shown in Maps 9.1 – 9.3, hospitals are often located near areas of deprivation. Deprived populations are known to use EDs more. Also it seems reasonable to expect that people are more likely to access a service if they live close to it. However it is not certain how these variables work together in their effect on how a population uses EDs.
9.2 Method
We took the accident and emergency commissioning data set (CDS) and used it to calculate the directly age sex standardised attendance rates for the registered population of the lower level super output areas (LSOAs) in the West Midlands Region. We then mapped the LSOAs by Primary Care Trust (PCT) areas to show those neighbourhoods that had high and low rates of use. These were mapped using standard deviations from the PCT mean attendance rate to show the extent of variation within each area.
We then looked at the income deprivation score for each of the LSOAs (using the income domain of the Index of Multiple Deprivation 2007 and also the Euclidian (straight-line) distance from the geographical centre of the LSOAs to the nearest ED site. These were then entered into a simple linear regression model with the attendance rate as the dependent variable and the strength of association with distance and deprivation was then examined further.
9.3 Results
The urban populations of the region have generally good access to EDs. Approximately 75% of the registered population live in an LSOA whose centre is 5 miles (8.05 km) from an ED. Approximately 10% live in an LSOA whose centre is less than a mile (1.61 km) from an ED. No LSOA centre is more than 22 miles from an emergency department, although road distances may be somewhat further, especially in rural areas.
As can be seen in Maps 9.1– 9.3, in many cases EDs are located close to or even in the midst of some of the more deprived areas of the region.
Also Maps 9.4 – 9.20 show that the distribution of attendance rates by LSOAs within the PCTs of the region. As can be seen, these are very variable and clusters of LSOAs of higher use occur in all of them. Again many of these are very close to the site of an ED. Having described the geographical variation in attendance around the region, we looked in greater detail at the relative importance of distance and deprivation. Intuitively both of these variables seem very important and other studies have shown that ED use increases as distance is lower and deprivation is greater1,2.
Simple linear regression shows strong relationships with both of our variables of interest and attendance rate (Figures 9.1 and 9.2). When we adjusted for linearity and the interaction between the two factors using a simple model, we found that both distance and deprivation remained important factors independent of each other. The deprivation co-efficients was not very intuitive to interpret, as a 1-point difference in income deprivation score is wider than the entire West Midlands range, the mean score being 0.18 and standard deviation being 0.14. For this reason we chose to include the standardised regression co-efficients (Table 9.1).
As can be seen, one standard deviation increase in distance from hospital is associated with just under half a standard deviation decrease in attendance rate. One standard deviation increase in income deprivation is associated with just over half a standard deviation increase in attendance rate. The relationship between deprivation and distance from hospital was
weaker than the maps appear to show, however there is a modest effect and the two variables do interact in a non-constant way (not described here). The overall model fit was good, with an adjusted R2 of 0.64. Note that a square root transformation of the standardised attendance rate was used to derive the model statistics.
9.4 Discussion
This is the first time that we have been able to explore this relationship. It appears that deprivation is the strongest predictive variable, but both deprivation and distance are strong predictors of how much populations use EDs and importantly they appear highly independently predictive. There is further work to be done on how deprivation and distance interact with each other, but this is outside the scope of this brief chapter.
The strength of the model fit was surprising as some other possibly quite important variables were not included such as primary care characteristics and proximity of minor injury units and walk-in centres.
We did identify some further possible problems with the approach that we used. Using road distances rather than euclidian distances may have altered the behaviour of the distance variable, although another study of hospital accessibility found little difference between these two measures as model terms 3. Also there were some possible confounders. Our results related to all attendances although one of our EDs was a children’s ED. We chose to leave this in the model for simplicity. Also many of the LSOAs which were very close to this ED were almost equally close to another major provider that of City Hospital.
Another problem was a lack of homogeneity in our EDs. Some will have separate pathways for minor injuries and primary care problems. One of our EDs was in fact a minor injuries unit (Kidderminster). We included this as we happened to have separately identifiable data for this unit available for analysis. However we do not have data in this model for minor injury units or walk-in centres for the rest of the region. It is also known that the CDS itself has problems. The Department of Health estimated that around 8% of attendances could be missing from CDS returns in the region 4, although this is somewhat smaller than the estimate for missing data nationally. If these data are missing at random, then given the total number of observations, this need not be a serious issue, however if there was systematic undercounting in specific hospitals, this could undermine the analysis.
We also had possible denominator problems with some LSOAs at the edges of the region. Our denominators were registered West Midland's patients by LSOA. We used these as it is believed that for some important groups of attenders (especially younger adults in deprived areas), that this is a more accurate estimate than others that are available. However there were a few isolated LSOAs where most of the residents were registered just outside the region and yet the preferred ED of attendance was inside it. In theses cases and artificially large ED usage rate was produced. As this happened very infrequently, we left these values in the analysis rather than trying to model in a more accurate population for these areas.
There was not a strong relationship between proximity to hospital and deprivation. The fact that both are important factors but the fact that they are strongly independently predictive is important. This means that in PCTs where there are deprived communities living near EDs will see exceptionally high demand in these areas.
Map 9.1: Income deprivation by LSOA and location of Emergency Departments in Birmingham and the Black Country
Map 9.2: Income deprivation by LSOA and location of Emergency Departments in the Stoke-on-Trent area
Map 9.3: Income deprivation by LSOA and location of Emergency Departments in the Coventry area
Map 9.4: Emergency Department attendance rate variation in Birmingham East and North PCT, 2007/2008
Map 9.5: Emergency Department attendance rate variation in Coventry PCT, 2007/2008
Map 9.6: Emergency Department attendance rate variation in Dudley PCT, 2007/2008
Map 9.7: Emergency Department attendance rate variation in Heart of Birmingham PCT, 2007/2008
Map 9.8: Emergency Department attendance rate variation in Hereford PCT, 2007/20
Map 9.9: Emergency Department attendance rate variation in North Staffordshire PCT, 2007/2008
Map 9.10: Emergency Department attendance rate variation in Sandwell PCT, 2007/2008
Map 9.11: Emergency Department attendance rate variation in Shropshire County PCT, 2007/2008
Map 9.12: Emergency Department attendance rate variation in Solihull PCT, 2007/2008
Map 9.13: Emergency Department attendance rate variation in South Staffordshire PCT, 2007/2008
Map 9.14: Emergency Department attendance rate variation in South Birmingham PCT, 2007/2008
Map 9.15: Emergency Department attendance rate variation in Stoke-on-Trent PCT, 2007/2008
Map 9.16: Emergency Department attendance rate variation in Telford and Wrekin PCT, 2007/2008
Map 9.17: Emergency Department attendance rate variation in Walsall PCT, 2007/2008
Map 9.18: Emergency Department attendance rate variation in Warwickshire PCT, 2007/2008
Map 9.19: Emergency Department attendance rate variation in Wolverhampton City PCT, 2007/2008
Map 9.20: Emergency Department attendance rate variation in Worcestershire PCT, 2007/2008
Figure 9.1: Simple linear regression of distance from ED on standardised ED attendance rate of LSOAs in the West Midlands 2007/2008
Figure 9.2: Simple linear regression of income deprivation on standardised ED attendance rate of LSOAs in the West Midlands 2007/2008
Table 9.1: Summary of breast screening units drive times analysis
| |
Sq. root rate |
Coef. |
Std. Err. |
t |
P>t |
Standardised
coefficient |
Km from ED |
-0.251 |
0.010 |
-25.49 |
0.000 |
-0.474 |
Income deprivation |
10.869 |
0.350 |
31.06 |
0.000 |
0.509 |
Distance / income |
-0.123 |
0.067 |
-1.85 |
0.066 |
-0.035 |
Constant |
15.944 |
0.079 |
202.41 |
0.000 |
|
|
|
| |
“This work is based on data provided with the support of the ESRC and JISC and uses boundary material which is copyright of the Crown and the ED-LINE Consortium” |
|
|
References:
- Hull S, Jones I, Moser J. Factors influencing the attendance rate at accident and emergency departments in East London: the contributions of practice organization, population characteristics and distance. Health Serv Res Policy. 1997; 2(1):6-13
- Beattie T, Gorman D, Walker J. The association between deprivation levels, attendance rate and triage category of children attending a children's accident and emergency department. Emergency Medicine Journal. 2001; 18:110-111
- Fone D, Christie S, Lester N. Comparison of perceived and modelled geographical access to accident and emergency departments: a cross-sectional analysis from the Caerphilly Health and Social Needs Study. International Journal of Health Geographics. 2006; 5(16) available from
http://www.ijhealthgeographics.com/content/5/1/16
- Health and Social care Information Centre. Accident and Emergency Attendances in England (Experimental Statistics) 2007-08, 2009, appendix 3, 51-5
| 
