Variations in vaginal and abdominal hysterectomy by region and trust in England
Dr A. Bottle, Dr Foster Unit at Imperial, Imperial College London, Department of Epidemiology and Public Health, St Mary's Campus, 16 South Wharf Road, London, W2 1PG UK.
Objective To examine variations between regions and hospitals in the proportion of hysterectomies performed abdominally.
Design Analysis of routine hospital data.
Setting All National Health Service hospitals in England.
Population Women aged 18+ hospitalised between April 1998 and March 2001.
Methods Logistic regression, adjusting for age and diagnosis.
Main outcome measure Use of the abdominal rather than the vaginal route.
Results The adjusted proportion of hysterectomies performed abdominally varied from 75–89% between regions, and from 25–99% between hospitals. Diagnosis accounted for nearly a third of the total variation, dwarfing the contributions of age and hospital. About two-thirds of the variation remained unaccounted for.
Conclusion Despite evidence suggesting that the majority of hysterectomies may be performed vaginally, very few English trust match this.
There is controversy over the variation in the proportion of hysterectomies carried out abdominally. Mulholland et al.1 found that the overall ratio of abdominal to vaginal hysterectomies varied at a regional level in 1993 between 2.7 (Yorkshire) and 6.7 (West Midlands). The abdominal route predominates in the US and UK, whereas the vaginal route is commoner in Europe. Studies have found the vaginal route to have lower mortality and post-operative morbidity as well as lower cost.2 Although vaginal hysterectomies are not usually indicated in malignancy, trials have suggested that other contraindications such as large uterine size are only relative, rather than absolute.3
We use Hospital Episode Statistics (HES) data to describe the variation in hysterectomy routes at trust and regional level within England using 1998/9–2000/1 data. We examine how much of this is attributable to differences in case mix factors available within routine data and show what variation remains after adjustment for these factors.
Elective admissions for five diagnostic groups4 between April 1998 and March 2001 were obtained from HES: endometriosis (ICD10 N80), uterine fibroids (D25), genital prolapse (N81), malignancies (C51–C57, D06, D07) and menstrual irregularities (N92–N94). The main procedure field was used to divide hysterectomies into three categories: abdominal (OPCS4 code Q07), vaginal (Q08) and laparascopically assisted (Q07 or Q08, with Y508 in any of the three secondary procedure fields). Three age groups were used: 18–44, 45–64 and 65 or over. There were eight Regional Offices and 164 English NHS trusts as of April 2001 after excluding nine specialist centres and eight centres carrying out fewer than 30 procedures per year. Laparascopically assisted vaginal hysterectomies are not common, here recorded in only 1.4% of all hysterectomies, so the focus will be on the other two routes.
Variation was examined by including trust or region together with age and diagnosis in logistic regression modelling of the odds of choosing the abdominal route as opposed to the vaginal route. The resulting odds ratios for each region or trust were then multiplied by the England crude odds, and the resulting odds transformed into rates (i.e. the proportion of hysterectomies that were performed abdominally, adjusted for age and diagnosis) by using the following equation:
There were 282,259 admissions with a primary diagnosis within the five diagnostic groups. These admissions accounted for 81% of all hysterectomies during the period. Uterine fibroids had the highest proportion of admissions that included a hysterectomy (73.5%); the proportions for the other conditions were 53.4% (menstrual irregularities), 45.0% (endometriosis), 26.7% (malignancies) and 15.2% (genital prolapse). The overall ratio of abdominal to vaginal hysterectomies was 3.6. The ratio of abdominal to vaginal hysterectomies was understandably greatest for malignancies (19.4); the ratios for the other conditions were 8.6 (uterine fibroids), 7.2 (endometriosis), 3.2 (menstrual irregularities) and 0.1 (genital prolapse).
There was significant variation between regions (Table 1). The abdominal route was most common in the West Midlands and least common in the Eastern region, with 89.3% and 74.6% of procedures using this route, respectively (after adjustment for age and diagnosis). These proportions fell to 87.5% and 71.2% after excluding malignancies, for which there is most consensus on choice of route. Within the same region, there remained considerable interhospital variation in route even after adjustment for age and diagnosis. In all but the South East, there was at least twofold variation in the proportion performed abdominally, with adjusted proportions ranging from about one-fifth to nearly 100% across England.
Table 1. Variations in percentage of hysterectomies performed abdominally between trusts for each region, adjusted for age and diagnosis, with and without malignancies.
|Eastern||1||74.6 (71.2)||25.3 (19.6)||84.4 (81.2)|
|Northern and Yorkshire||1.08 (1.01–1.15)||76.0 (72.6)||41.3 (35.1)||92.3 (90.7)|
|South West||1.23 (1.14–1.31)||78.3 (74.8)||33.7 (28.4)||96.1 (95.4)|
|London||1.69 (1.56–1.81)||83.2 (81.0)||42.4 (38.8)||94.5 (93.2)|
|South East||2.00 (1.87–2.14)||85.5 (83.0)||60.0 (55.7)||95.3 (94.5)|
|North West||2.09 (1.95–2.24)||86.0 (83.4)||41.5 (35.4)||98.8 (98.7)|
|Trent||2.10 (1.95–2.26)||86.1 (83.4)||43.2 (37.3)||93.0 (91.4)|
|West Midlands||2.84 (2.63–3.06)||89.3 (87.5)||58.3 (53.5)||96.1 (95.4)|
|All England||–||78.4 (75.3)||25.3 (19.6)||98.8 (98.7)|
From logistic regression, age and diagnostic group accounted for 32.1% (adjusted R2= 0.321) of the variation (two-way interactions did not usefully improve the model fit). Quantitatively, the most important factor was diagnosis group, accounting for 29.0% of the variation by itself. The removal of the adnexa was recorded in 67% of abdominal and 14% of vaginal procedures in our data (although this does not necessarily indicate adnexal pathology); similarly, one or both ovaries was 2.5 times more likely to be removed with an abdominal hysterectomy than a vaginal hysterectomy. Adjustment for adnexa or ovary removal improved the model fit by less than 1% and neither factor was included in the final regression.
Lengths of stay were longest with the abdominal route in each age group. The median stay for an abdominal procedure was five days for ages 18–44, six days for ages 45–64 and seven days for ages 65 or more; median stays for vaginal procedures were one day shorter at each age group. This enables an estimation of the savings in hospital bed days that might be made if more procedures were performed vaginally. Trusts in this study, which were selected to be comparable to a ‘typical’ District General Hospital in size and specialty mix, performed an average of 669 procedures over the three years, of which 78% (535) used the abdominal route and 22% (134) the vaginal route. If this hospital was to do 80% of its procedures vaginally, it would save about 390 bed days over three years; the saving potentially made by all trusts in this study combined was over 66,000 bed days.
There is great variation across England in the route of hysterectomy. Nearly a third of this variation is explained by age and diagnosis (32.1%), of which the most important quantitatively is the primary diagnosis, that is, the indication for hysterectomy. The national odds of using the vaginal route between 1998 and 2000 are little changed from those in 1990–1993 in work by Mulholland et al.1 and use of the abdominal route is still most frequent in the West Midlands region. Combining three years gave narrow confidence intervals so the findings are probably not due to chance. Reasons for the remaining variation may be divided into trust and patient factors.
The quality of the coding of clinical information is known to vary between trusts. However, HES data have improved greatly since their inception and the early 1990s5 and are unlikely to account for either the effects of age, diagnosis or trust seen in this study or for a large part of the unexplained variation. Although there are no published figures on the accuracy of the coding of route of hysterectomy (i.e. Q07 vs Q08), to our knowledge, this is probably high. Less predictable is the completeness of recording Y50.8 to indicate laparascopic assistance, as in national coding guidelines. Some loss of admissions is likely to be due to vague diagnostic coding, with an unpredictable but probably modest effect on the results.
The abdominal route is preferred in patients with a large uterus, whereas the vaginal route is preferred for metrorrhagia, genital prolapse or small fibroids, for example. Unfortunately, HES data have no information on uterus size except for the presence of hypertrophy (if diagnosed) and it is known that secondary diagnoses are less frequently recorded in the UK than in the US,5 where there is a financial incentive for doing so. As it is likely that relevant comorbidities were under-recorded in at least some trusts, and hence their effect under-estimated, their inclusion in the regression was not felt to be worthwhile.
The five diagnostic groups chosen here accounted for 81% of all hysterectomies and have been used elsewhere.4 The inclusion of further groups would have accrued more cases but the law of diminishing returns suggests that any improvement in model fit is likely to be offset by reduced precision in model estimates.
More importantly, this study has not been able to address intra-trust differences, in particular, surgeon-specific odds. This is probably why so much variation remained despite including trust in the model; the trust of operation is a proxy for a number of factors within it—in most hospitals more than one surgeon will perform the procedures. HES data now include consultant team code but this was not made available to us for our analysis.
There are a number of reasons for preferring the abdominal route, other than large uterus, malignancy and adnexal pathology already considered, for which there are no data in HES. These include previous caesarean section or surgery, the need to perform oophorectomy, contracted bony pelvis with narrow vagina and lack of uterine descent. The vaginal route may be preferred in obese patients, but HES does not hold body mass index information. Routine hospital data also include no details about the severity of disease or intra-operative findings on the appropriateness of the route chosen.
We estimate that if 80% or more hysterectomies could be performed vaginally,6 the NHS could save as many as 22,000 bed days annually in addition to the benefits of lower mortality and post-operative morbidity.
HES data are useful in examining variation in care within the NHS. Large variations in hysterectomy route exist between trusts, even after accounting for the large effect of diagnosis and the smaller but important effect of age. Differences between gynaecologists within the same trust could not be investigated using the data available to us (although this will be possible in future) but are likely to be important.
The authors would like to thank Dr Deidre Lyons, consultant in colposcopy at St Mary's Hospital, for her helpful comments on the manuscript.