J.R. Coll. Surg. Edinb., 43, August 1998, 257—261

The case for clinical audit of emergency readmissions after appendicectomy

A. F. BISSET
Senior Registrar in Public Health Medicine, Grampian Health Board, Aberdeen, UK

The aim of this study was to examine how the Scottish Office Clinical Outcome Indicator for emergency readmissions within 28 days of appendicectomy could be made more useful. Scottish Morbidity Record One (SMRI) data on all NHS discharges with a primary operative procedure of appendicectomy, and a linked file of all emergency readmissions within 28 days of discharge in 1993/94 were used in this study. It was found that 8783 appendicectomies were performed in 1993 and 1994, with 403 (4.6%) emergency readmissions within 28 days. A significantly higher proportion of emergency readmissions occurred where a ‘normal’ appendix was removed, and in interval, prophylactic and incidental appendicectomies. Age and length of initial hospital stay did not affect readmission rates. Hospitals performing more appendicectomies tended to have more readmissions [Spearman’s rank correlation coefficient 0.639 (95% CI 0.430—0.782]. None of the five hospitals carrying out less than 17 appendicectomies in 2 years had an emergency readmission. Patients from more deprived areas were significantly more likely to be readmitted. The current performance indicator is useful in identifying patients readmitted to any hospital in Scotland, but needs to be supplemented with more clinical information. The numbers of appendicectomy patients would make this a feasible national clinical audit project, and the lessons learnt should be useful in promoting better practice for considerable numbers of patients of all ages.

Keywords: appendicectomy, audit outcomes, readmissions.

Clinical outcome indicators have been devised in Scotland for emergency readmissions within 28 days of discharge after emergency appendicectomy.¹ Readmission rates are standardized for age, sex and deprivation for the 30 Trusts which discharged 200 or more appendicectomy patients in 3 years between April 1992 and March 1995.¹ It has been suggested that examination of any causal connections between surgery and readmission would be ‘extremely complex’.^1,2 However, there seems little justification for collecting such data unless it can be used to improve patient care. This paper examines the epidemiology of emergency readmissions across Scotland in 1993 and 1994, determining what information is currently available, whether routine data can uncover any risk factors which might predispose to subsequent emergency readmission; and discusses ways of making the indicator more useful clinically.

METHODS

Routine data from Scottish Morbidity Record One (SMR1) on all NHS discharges with a primary operative procedure of appendicectomy (OPCS 4 code HO1-H03) were supplied by the Information and Statistics Division of the Common Services Agency in Edinburgh for the calendar years 1993 and 1994. All readmissions to any Scottish NHS hospital after these discharges were derived from the linked set of Scottish hospital records,³ and emergency readmissions within 28 days were selected.

Data were analysed using SPSS-PC⁴ and the Confidence Interval Analysis Programme.⁵

RESULTS

In 1993 and 1994,4451 and 4332 appendicectomies were carried out, respectively (8783 operations). Sixteen patients (0.2%) died (age range 45—93 years); two died within 28 days of discharge (aged 35 and 58 years); and 403 (4.6%) were readmitted within 28 days as emergencies.

The type of appendicectomy performed affected the proportion of subsequent emergency readmissions. In particular, there was a significantly higher proportion of emergency readmissions in emergency appendicectomies where a ‘normal’ appendix was removed, and in interval, prophylactic and incidental appendicectomies (Table 1).

Table 1 Proportion of readmissions within 28 days by appendicectomy code

The age and sex of emergency readmissions showed the same pattern as first admission appendicectomies: 53% of all appendicectomy patients were male, compared with 51% of emergency readmissions. The age at appendicectomy ranged from 0 to 93 years (median 21 years), and the age of emergency readmission patients ranged from 0 to 85 years (median 21 years). There was no significant difference in rates of emergency readmissions between different age bands, though rates were slightly higher in the youngest and oldest (Table 2).

Table 2 Appendicectomy by age-bands 1993 and 1994

Diagnosis on readmission is shown in Table 3. Of those readmitted, 111 (27.5%) required another operative procedure. The re-operation patients ranged in age from 5 to 85 years (median 24 years), and 58% were male (91 patients) (Table 4).

Table 3 Primary diagnosis on readmission

Table 4 Further operations carried our on readmission

The day of the week on which patients were admitted showed a similar pattern for both groups, with fewer admissions or emergency readmissions on Saturday and Sunday, and a peak on Monday. The time of day at which the operation was carried out (for example during the night),⁶ and grade of medical staff performing the operation, are not included in routine data.

The length of initial hospital stay did not affect emergency readmission rates: the median length of hospital stay during the first admission was the same (4 days) for both groups. The median interval from discharge to emergency readmission was 7 days, and 25% were readmitted within 3 days. The length of readmission stay ranged from 0 to 32 days (median 3 days). One emergency readmission patient died during readmission, 90.3% (364) were discharged home, and 8.7% (35) were discharged to another hospital or speciality.

Patients were readmitted after discharge from 59 different hospitals. Thirty-four (8.4%) of emergency readmissions were admitted to a different hospital. Thirty Trusts were included in the Outcome Indicator, but a further 29 hospitals performed less than 100 appendicectomies in 2 years (and therefore would not be counted in the outcome indicator). There was a moderate correlation between total number of appendicectomy operations in a hospital and numbers of emergency readmissions (Spearman’s rank correlation coefficient was 0.639 [95% CI for Correlation coefficient 0.430— 0.782]). None of the five hospitals carrying out less than 17 appendicectomies in 2 years had an emergency readmission.

The proportion of appendicectomy operations increased slightly (but not significantly) with Carstairs⁷ deprivation quintiles of the Scottish population, but emergency readmissions were significantly more likely to occur in patients with postcodes in the two quintiles with greatest deprivation (190 patients, 47% of all emergency readmissions [95% CI 42.3—52.0%]), compared with patients in the two quintiles with least deprivation (141 patients, 35% of all emergency readmissions [95% CI 30.3—39.6%]) (Table 5).

A further 884 patients were admitted as emergencies between 28 and 365 days. A total of 1150 patients (13%) were readmitted as emergencies within 3 years of their appendicectomy. A further1081 (12.3%) had a waiting list admission within 3 years of their appendicectomy, but only 35 were within 28 days.

Table 5 Deprivation and appendicectomy

DISCUSSION

There were 200 emergency appendicectomy readmissions in Scotland annually, with a maximum of 15 per hospital, in 1993 and 1994. The proportion of emergency readmissions is rising in Scotland (as in Oxford),² although the number of appendicectomies is falling:⁸ emergency readmission rates for appendicectomy were 3.72% with 6374 operations in 1983; 3.50% with 4454 operations in 1990;⁹ and 4.6% with 4332 operations in 1994. Specific risk factors are not obvious from routine data, and the large number of emergency and elective admissions for this group of patients suggest that there is overlap between readmissions directly connected with the appendicectomy, and other conditions.

Rather surprisingly, increasing age and small hospitals are not particular risk factors for readmission, although it is possible that the hospitals with very low numbers of operations may have been miscoded. Appendicectomy contrasts with other emergency readmissions where there is a ‘reasonably straight line association [with age] apart from the youngest age category’.⁹ Larger hospitals and greater deprivation increase the risk of readmission, and more work needs to be done on whether these two aspects are causally linked. Women formed a (slight) minority of the readmissions, though at least 12 of the readmissions were for gynaecological disorders. However, routine data give little information on case-mix, the reasons for readmission, and whether readmission was avoidable.

Table 3 shows that 21 patients had ‘acute appendicitis’ as the primary diagnosis on readmission. Routine data do not explain why 9% needed a further operation on the appendix. Did these patients have an exploratory laparotomy without removal of the appendix on the first admission and subsequently develop full-blown appendicitis? Was the coding incorrect on the first admission (i.e. the patients were admitted with presumptive appendicitis but observed and sent home only to have a recurrence and be readmitted)? Was this simply due to a coding anomaly wherein the code for acute appendicitis recurs for the readmission independent of the actual cause of the readmission? The Quality Assessment and Accreditation Unit at ISD found 96.4% operation coding accuracy at 3-digit level on a sample of 55 Appendicectomy SMR1 forms and 99% accuracy for linkage.¹ We do not know what proportion of emergency readmissions are due to coding errors, and clinical audit would provide useful feedback on coding accuracy.

Categories such as ‘complications of surgical care’ give no indication as to whether they were preventable, or whether they were more common with certain techniques, for example laparoscopy. There was no evidence that short lengths of stay affected readmission rates. One can only guess which of the ‘other operations’ had nothing at all to do with the first operation For example, ‘fractured skull’ and ‘fractured facial bones’ seem likely to be due to intercurrent trauma unrelated to the first admission — but were nor coded as emergency readmissions due to injury or accident. Reasons for some readmissions going to another hospital¹⁰ also need to be explored. Was this due to lack of beds, dissatisfaction with the first admission, inadequate facilities to cope with the readmission, or just miscoding?

Readmissions are being used as a proxy for avoidable adverse outcomes,¹⁰ but this makes many sweeping assumptions. We have no idea how sensitive or specific an indicator it is for this purpose. The data in this study cover only those patients where appendicectomy was the primary operation, missing out those where it was incidental to another primary operation. The Outcome Indicator covers only emergency appendicectomies, but emergency readmissions were higher for incidental appendicectomies in this study. The highest readmission rate was for emergency removal of ‘normal’ appendices. Was this due to misdiagnosis of other conditions, which needed subsequent treatment? If post-operative wound infection is an important avoidable outcome, then many patients who were treated in the community would be excluded. Reassuringly, however, Scotland’s figures for mortality after appendicectomy (18 deaths within 28 days for 8783 patients) compare favourably with Oxford figures of 28 deaths in 0—29 days for 16046 patients in 1980—6¹¹ More research is needed on whether the indicator should be limited to emergency readmissions within the first 7 days, to make the measure more specific. A team of assessors in North-East Thames Region found that surgical admissions at 0—6 days after discharge were more likely to be assessed as avoidable than those at 21—27 days, and surgical readmissions were thought to be more avoidable than medical ones.¹²

One criticism of this article could be that it raises more questions than it answers. It might be possible to answer quite a number of these questions by an external survey: for example, visiting hospitals all round Scotland, retrieving case notes (which may well not be available, or may not supply sufficient information), and interviewing records staff to find out why they used vague diagnostic codes such as ‘complications of surgical care’, or ‘gastrointestinal symptoms’ or ‘symptoms of the abdomen’ (rather than specifying the particular problem). This, however, would miss the whole point of this paper — namely, that the surgeons themselves need to do this work if it is to have any influence on future surgical management. The surgeons concerned need to reflect on the readmission, and whether anything could have been done differently. If the coding is inaccurate or vague, then the surgeons need to consider why. For example, was the discharge letter vague? Did a secretary or untrained officer do the coding? Only in this way (and in the best audit tradition) can surgeons understand, improve and take ownership of readmission rates. An external survey by Public Health Medicine, or a researcher, would not be an ‘audit’ in the true sense, i.e. clinicians examining their own practice. An external survey would be impracticable, historical, and would tend to cause antipathy among clinicians who would feel that they had been misrepresented and that ‘the data are rubbish’.

Coding and information play very little part in most surgical training, but even busy clinicians should appreciate that routine data are the public and national face of what they and their hospital do, and, therefore, need to be as accurate as possible — only the surgeons concerned know exactly which procedures were carried out, by whom, and why. Surgeons need to help ensure that the correct data are captured. Indeed, the more meticulous surgeons keep personal card-indexes of operations anyway. But how many of them routinely check their own data against their hospital statistics?

With the co-operation of surgical teams and sufficient funding, it would be feasible to undertake a national clinical audit¹³ of all future emergency readmissions (other national audits have been undertaken successfully by the Scottish Otolaryngological Society for tonsillectomy and glue ear surgery).¹⁴ Anonymous clinical information similar to that for confidential enquiries into maternal deaths and peri-operative deaths (CEPOD)¹⁵ could be gathered on a trial basis, particularly as record linkage allows follow up of those patients readmitted to another hospital. Two hundred patients per year in Scotland would make this a manageable project. Research elsewhere found that a panel of expert assessors with access to clinical information were able to categorize readmissions as avoidable or not.¹¹ More detailed information on surgical techniques used (for example, whether the operation was carried out by laparoscopy), patient management (for example, the use of prophylactic antibiotics) and the reasons for readmission, might provide an opportunity to improve local practice, make comparisons with other centres, and contribute to updating guidelines on good practice. If it were determined that certain routine data codes identify subgroups of patients where the readmission is much more likely to be related to a quality-of-care problem, then the audit could be targeted in future years and made more efficient.

One argument for starting such an audit with appendicectomy is that in 1990 only prostatectomy had a higher number of operations and higher emergency readmission rate, and prostarectomy affects a much narrower age (and sex) range.⁹ Lessons learned from appendicectomy may well apply to other operations as well, particularly as appendicectomy covers a much wider age range than many other operations.

In conclusion, the current performance indicator is useful in identifying patients readmitted to any hospital in Scotland, but needs to be supplemented with more clinical information if it is to contribute to improving patient care. The numbers of appendicectomy patients involved would make this a feasible project with national funding, and the lessons learnt should be useful in promoting the best possible practice for considerable numbers of patients of all ages.

ACKNOWLEDGEMENTS

I am very grateful to Steve Kendrick and Marion Bain for supplying data, and for advice; to Isla Imray, Librarian; and to Sir David Carter and Sir Robert Shields, for commenting on an earlier draft.

REFERENCES

1. Clinical Resource and Audit. Clinical Outcome Indicators. Edinburgh: The Scottish Office, 1996.
2. Henderson J, Goldacre MJ, Graveney MJ, Simmons HM. Use of medical record linkage to study readmission rates. BMJ 1989; 299: 709—13.
3. Kendrick 5, Clarke J. The Scottish record linkage system. Health Bull 1993; 51: 72—9.
4. SPSS/PC’. Chicago, IL. SPCC Inc. 1990.
5. Gardner SB, Winter PD, Gardner MI. Confidence Interval Analysis, Version 1.0. BMJ, 1989.
6. McLean AD, Stonebridge PA, BradhuryAW, Rainey JB, Macleod DAD.Time of presentation, time of operation, and unnecessary appendicectomy. BMJ 1993; 306: 307.
7. Carstairs V, Morris K Deprivation and Health in Scotland. Aberdeen:Aberdeen University Press, 1991.
8. Bisset AF. Appendicectomy in Scotland: a 20-year epidemiological comparison. J Public Health Med 1997; 19 (2): 213—8.
9. ISD Health Briefing (No 93/23). Trends in Hospital Readmission Rates in Scotland 1983—90. Edinburgh: Information and Statistics Division, 1993.
10. Milne R, Clarke A. Can readmission rates be used as an outcome indicator? BMJ 1990; 301:1139-40.
11. Seagroatt V, Goldacre M. Measures of early post-operative mortality:beyond hospital fatality rates. BMJ 1994; 309: 361—6.
12. Clarke A. Are readmission rates avoidable? BMJ 1990; 301: 1136—8.
13. Black N, Moore L. Comparative audit between hospitals: the example of appendectomy. IntJ Health Care Quality Assurance 1994; 7 (3):11—5.
14. Blair RL, MeKerrow WS, Carter NW, Fenton A. The Scottish tonsillectomy audit. JLaryngol Otol 1996; 110 (Suppl. 20): 1—25.
15. National Confidential Enquiry into Peri -operative Deaths. 35—43 Lincoln’s Inn Fields, London WC2A 3PN.

Paper accepted on 21 December 1997

© 1998 The Royal College of Surgeons of Edinburgh. J.R. Coll. Surg. Edinb., 43, August 1998, 257-261