10.1 Introduction

The quality and reliability of Accident and Emergency (A&E) data has historically been poor particularly in comparison to hospital inpatient data.  The cause of this is probably related to the coarse tariff structure used for A&E admissions; only three levels of care complexity and cost exist in this structure. This is in contrast to the relatively complex structure based on healthcare resource groups and length of stay, all with differing tariffs, used for hospital in patients which has resulted in continual improvements in this data quality and completeness.

As the pressure grows on Primary Care Trusts (PCTs) to reduce the volume of patients that require hospital care, both as part of limiting healthcare costs and as part of the government agenda to move care into the community and closer to patient’s homes1, they will increasingly need to know which patients are seeking hospital care and why. Only by being sure of these issues will PCTs be able to put appropriate plans in place to move care out of secondary care.  Whilst this has started to be possible for hospital in patients and out patients the poor quality of the A&E data has presented perennial issues for commissioners in drawing definite conclusions about A&E activity.

£92 million was spent by PCTs in 2007/8 on approximately 1.6million A&E attendances.

The data presented below is taken from the West Midlands region commissioning A&E dataset for the financial year 2007/8 (latest full year available).

10.2 Age and Sex Demographics of A&E Attendees

Directly age standardised attendance rates suggest that it is the very young (0-4 years) and young adults (15-29) have the highest rate of A&E attendance.  This is in contrast to hospital admissions where the elderly have much higher rates (Figure 10.1).  51% of people attending A&E were male.

Figure 10.1: Directly Age Standardised A&E Attendance Rate for the West Midlands in 2007/8

Directly Age Standardised A&E Attendance Rate for the West Midlands in 2007/8

10.3 When are People Attending A&E?

The number of people attending A&E is significantly greater in the spring and autumn than in the winter.  This is initially counter-intuitive since hospital inpatient activity is greatest in the winter (Figure 10.2). However, this pattern may reflect the demographics of those using A&E (i.e. young children and young adults) that are likely to be substantially different from those referred directly into hospital.

Intuition may also suggest that A&E attendances should be greater at weekends because primary care services are often less accessible than during the week.  However, the data shows that there is marked periodicity in A&E attendance with spikes of attendance on almost every Tuesday throughout the year (Figure 10.3).  The average number of A&E attendances is significantly higher on Tuesday than any other day of the week (t-test p-value <0.0001) (Figure 10.4).  The reasons for this periodicity are not immediately clear but have implications for primary care in terms of ensuring satisfactory number of community staff to prevent inappropriate A&E use and for secondary care in terms of capacity.

Figure 10.2: Mean number of A&E Attendances in the West Midlands by Month in 2007/8 with 95% Confidence Intervals

Mean number of A&E Attendances in the West Midlands by Month in 2007/8 with 95% Confidence Intervals

Figure 10.3: Frequency Chart of the Number of A&E Attendances per Day in the West Midlands in 2007/8

Frequency Chart of the Number of A&E Attendances per Day in the West Midlands in 2007/8

Figure 10.4: Mean number of A&E Attendances in the West Midlands by Day of the Week in 2007/8 with 95% Confidence Intervals

Mean number of A&E Attendances in the West Midlands by Day of the Week in 2007/8 with 95% Confidence Intervals

10.4 Do we know what is really happening whilst patients are in A&E?

In order to be able to provide alternative services to A&E or improve access to community services to reduce the need for A&E attendance it is vital to know what problems people are seeking advice for.  In the West Midlands over 31% of all people attending A&E had no diagnosis recorded (i.e. field was blank) and a further 10% had a non-specific diagnosis (either coded as “diagnosis not classifiable” or “no actual diagnosis”.  The completeness of reporting varied substantially between trusts from 100% completeness to less than 30%, which severely limits the ability to compare the type of activity occurring between trusts.  (Figure 10.5).

41% of A&E attendances had no diagnosis recorded.  This equates to about 660,000 attendances that PCTs paid for (£37.7 million) but had no indication of why the patient attended.

Figure 10.5: Percentage of A&E Attendances that have no diagnosis or a non-specific diagnosis recorded

Percentage of A&E Attendances that have no diagnosis or a non-specific diagnosis recorded

36.2% of people attending A&E did not have any investigation recorded.  A further 12.8% did not require any investigation. Therefore up to 49.1% of people attending A&E may not have required any investigation.

41.0% of people attending A&E did not have the treatment they received recorded. 10.1% required no treatment and 20.0% required advice only. Therefore up to 71.2% of people attending A&E may not have required any treatment other than advice.  This equates to over 800,000 attendances that PCTs paid for but had no indication of what treatment was received.

10.5 Admissions from A&E

The discharge of patients from A&E is a well recorded field and is very complete.  The proportion people admitted to hospital following their A&E attendance differs significantly between trusts across the region and by a factor of over 2.5. Admissions to hospital, even short stays, are substantially more expensive than A&E attendances therefore commissioners should investigate assessment, referral and treatment pathways in departments with significantly higher than average admission rates.

Figure 10.6: Statistical Process Control Chart of the Proportion of A&E Attendances that End in Admission to Hospital in the West Midlands in 2007/8

Statistical Process Control Chart of the Proportion of A&E Attendances that End in Admission to Hospital in the West Midlands in 2007/8

10.6 Deaths in A&E

The proportion of deaths occurring in A&E departments across the region does not differ significantly between trusts.

Figure 10.7: Statistical Process Control Chart of the Proportion of A&E Attendances that End in Death in the A&E Department in the West Midlands in 2007/8

Statistical Process Control Chart of the Proportion of A&E Attendances that End in Death in the A&E Department in the West Midlands in 2007/8

10.7 Waiting Times

The A&E four-hour wait target is an important and high profile performance measure for hospital trusts.  Waiting time refers to the time between the patient arriving in A&E and being discharged or transferred and the national target is that 98% of patients wait less than four hours².  On average, 98.16% of patients waited less than four hours in the West Midlands in 2007/8 according to the results of the Healthcare Commission’s Annual Health Checks³.

Examining the data however reveals that 23.8% of people have insufficient details recorded to allow the waiting time to be calculated e.g. no arrival or discharge time and 6.7% have a waiting time of zero recorded.  These are equivalent to about 492,000 people.  With this level of data incompleteness there is no way that a commissioner can be reassured that their local trust’s proportion of people waiting less than four hours is accurate.

The four-hour wait target has produced a spike in the frequency distribution of waiting times coming to a peak at 3 hours 59 minutes with a remarkably step drop off after 4 hours.  The average number of people discharged in the five minutes before 4 hours (6849 per minute) and the five minutes after (130 per minute) are significantly different (t test p value <0.0001).

Figure 10.8: Frequency Distribution of A&E waiting times in the West Midlands in 2007/8

Frequency Distribution of A&E waiting times in the West Midlands in 2007/8

10.8 Conclusions

• The current A&E dataset is of variable quality and completeness but despite this it can produce useful information that can be used to help plan services.


• Completeness and accuracy particularly around diagnostics/investigations/treatments is very poor.


• Validity of 4-hour waits is questionable due to data completeness issues.


• There were large and significant differences in the rate of admission to hospital from A&E across hospital trusts.


• There was no significant difference between the death rate of patients in A&E departments across the region.

References

1. Darzi A. High Quality Care for All: NHS Next Stage Review Final Report. London: Department of Health 2008


2. Healthcare Commission. The Annual Healthcheck 2008. London: Heathcare Commission. 2008


3. Healthcare Commission. Complete Dataset of existing national target indictator results for 2007/8. London: Healthcare Commission. 2008. Available at: www.cqc.org.uk/publications.cfm?fde_id=1254 (Accessed 10th July 2009).