- Top of page
Over-running operating lists are known to be a common cause of cancellation of operations on the day of surgery. We investigated whether lists were overbooked because surgeons were optimistic in their estimates of the time that operations would take to complete. We used a questionnaire to assess the estimates of total operation time of 22 surgeons, 35 anaesthetists and 16 senior nursing staff for 31 common, general surgical and urological procedures. The response rate was 66%. We found no difference between the estimates of these three groups of staff, or between these estimates and times obtained from theatre computer records (p = 0.722). We then applied the average of the surgeons' estimates prospectively to 50 consecutive published surgical lists. Surgical estimates were very accurate in predicting the actual duration of the list (r2 = 0.61; p < 0.001), but were poor at booking the list to within its scheduled duration: 50% of lists were predictably overbooked, 50% over-ran their scheduled time, and 34% of lists suffered a cancellation. We suggest that using the estimates of operating times to plan lists would reduce the incidence of predictable over-runs and cancellations.
Between 10% and 40% of booked elective operations may be cancelled before surgery takes place [1–8]. The reasons include cancellation by the patient, cancellation for poorly optimised medical conditions, or cancellations due to poor organisation [3, 4, 7, 8]. We previously concluded that a nurse-led pre-assessment system to optimise patients medically was effective in approximately halving the number of cancellations . The most common cause of the remaining cancellations was ‘lack of theatre time’ (i.e. over-running operating lists). The Audit Commission has estimated that in about 5% of hospitals in the UK, the majority of operating lists were consistently over-running .
An operating list may over-run because of delayed starts, slow turnover, unanticipated surgical/anaesthetic problems or staff shortages. Many of these are difficult to quantify. However, one relatively easily measured factor is the possibility that some operating lists are predictably ‘over-booked’.
In North America, the provision of surgical-anaesthetic services is virtually unlimited in time because of the way in which health care is financed. Theatres are often utilised for as long as needed for operations and cancellations due to ‘over-running’ are rare. Instead, over-runs add to overall costs, which can pose different but equally important problems for the hospital [9–14].
In contrast, in the UK, each surgical team is offered a dedicated operating room for a predetermined length of time. Usually, 4 h is a single ‘session’ (or contractually a ‘programmed activity’[15, 16]). Thus a ‘half-day list’ is usually 4 h; a ‘full-day list’ is 8 h (lunch breaks or rests are excluded from these times). Over-booking occurs when the expected duration of all the listed operations combined is anticipated to exceed these allocated times.
Surgical lists are most commonly currently booked by surgical teams and our initial prejudice was that surgeons were optimistic about the times taken for operations. We planned to obtain estimates (using a questionnaire) from surgeons for times taken for common operations and compare these estimates with those of other staff groups (i.e. anaesthetists and theatre nursing staff). A second aim of the study was to compare these subjective estimates with more objective data obtained from the theatre computer. Finally, we wished to apply these time estimates to actual published operating lists, to assess whether the total duration of the surgical list (and hence any patient cancellations) could be predicted.
- Top of page
Contrary to our expectations, surgeons are not more optimistic than anaesthetists or nurses in their estimates of the time needed for common operations. Furthermore, these subjective estimates are all in agreement with more objective data from the theatre computer. Predictable over-booking, rather than any unanticipated influence, appears to be the main factor underlying cancellation.
Before we consider the impact of this data on the management of operating lists, it is important to consider some potential limitations and strengths of this study.
The response rate to the questionnaires across all staff groups was good. However, for some specific operations, the number of responses was low, e.g. transanal resection of tumour (Table 3) and PCNL (Table 4), probably because only a few specialists regularly undertake these procedures. Similarly, very little data was extracted from TIMS for some operations, e.g. biopsy/sampling axillary lymph nodes (Table 2), because by chance very few such procedures were recorded during our data collection period. However, low returns were not uniform across all staff groups. For example, PCNL yielded responses from the only two surgeons undertaking this procedure but 10 anaesthetic responses, which was acceptable (Table 4). We decided therefore to include all such data sets in our analysis, and also present a ‘combined estimate’, which averaged the estimates of all staff groups.
Surgeons' responses may have been so accurate because they guessed the purpose of this study and did not wish to be ‘outliers’, and this may have been a source of bias. However, we think this unlikely: there were still large differences between estimates of staff groups for some individual operations (e.g. PCNL and total cystectomy, Table 4).
We might have used TIMS data alone and avoided the need for any subjective questionnaire estimates, but we regarded TIMS data only as ‘pseudo-objective’ and prone to errors. We encountered missing data, implausible times and operations that did not match those on the published list. Our publication of the ‘combined estimate’(Tables 1–4) may have some value in adjusting for deficiencies of the two methods (questionnaire and computer), and also for the problem of low returns for some operations.
A particular strength of our study was to emphasise the total duration of an operation. Some studies measure only the time from skin incision to application of a surgical dressing [18–20], which has less value in planning an operating list.
Our definition of 20 min as a list over- or under-run is larger than that of Widdison  (10 min) and Durani et al.  (1 min), but is more useful as it resembles the duration of some short surgical procedures such as cystoscopy.
We were careful to record data for published lists over two separate time periods over the year, to avoid potential problems related to fluctuations in case mix, staffing, or absences due to staff leave over the year .
Some factors may have led us to overestimate the duration of lists. Our simple sum of the estimates for individual operations should yield the list duration, but list duration can sometimes be shorter if, for example, one anaesthetist is present to induce anaesthesia, and another anaesthetist awakens or recovers the previous patient .
Other factors may have led us to overestimate the cancellations. Some cancellations may instead have occurred for other reasons (e.g. medical), but by chance on a list that was over-booked. Furthermore, patients may have been postponed for other reasons before arriving in hospital, but their name somehow appeared in error on the published operating list: our methodology would have recorded this as a ‘cancellation’.
Our mathematical analysis is relatively simple, Previous studies have used sophisticated techniques such as upper prediction bound analysis , bin packing algorithms and fuzzy constraints , and ‘L1’ regression . Such complex tools may have enabled us to analyse our data in more detail, but a strength of our approach is that it is intuitive, relevant and readily applied by others. Indeed, Dexter & Macario have argued that a simple average of the past 39 durations for an operation, combined with surgeons' mean estimates for that operation, yields a good measure of future operating times .
Why lists are over-booked
We can only speculate on the reasons why as many as 50% of lists are predictably over-booked.
Surgeons may experience pressure from managers to reduce waiting lists by adding more patients to an already full list. Surgeons who book fewer patients on their lists (albeit appropriately) may fear criticism (e.g. not ‘working hard’) from their clinical colleagues . We assumed that all surgeons book their own lists: it may be that in some other hospitals, non-clinical managers are now more directly involved in planning surgical lists, and so managers' estimates may be more relevant.
If surgeons are asked about the ‘average’ time for a certain operation in a questionnaire, they may respond quite accurately (in agreement with estimates of other staff groups). However, they may feel that they themselves always perform better than ‘average’, so are justified in booking more patients than the norm. ‘Intentionally’ over-booking a list might cause the team as a whole to work harder and our data suggests that this may be the case: although ∼ 50% of lists are predictably over-booked, only ∼ 34% suffered a cancellation. However, this is a very ‘high cost’ strategy: 14% of booked patients were cancelled, which is unacceptable.
Finally, lists may be booked more randomly and less by design, than we might suppose. Surgeons may (subconsciously) ‘know’ how long a given operation will take, but they then do not ‘consciously’ use this knowledge to plan the list.
Why lists are under-booked
We were surprised to find that about a third of lists were apparently under-booked. This may simply be because surgeons are poor at planning a list, and this leads to both frequent over-booking and under-booking. Some lists may have been intentionally under-booked to accommodate planned absences for meetings, etc. We might then have expected parallel reductions in allocation of anaesthetic/nursing staff, but this did not occur.
Some patients may have been cancelled for surgical or medical reasons after they arrived in hospital, but before the list was published. Since we analysed only the published list, this would have given the appearance of under-booking.
A final explanation for apparent under-booking is that surgeons may have felt some particular ‘routine’ operations would actually turn out to be more complex, and so allowed themselves more time. Again, our study design would not have detected this.
Comparison with previous work
Much of the work examining the management of operating lists has come from North America [1, 9–14, 18, 21–24]; UK data is quite sparse. Barr et al. reported that ∼ 21% of lists over-ran and ∼ 33% under-ran . More recently, an analysis of general surgical lists in a district general hospital found that ∼ 56% of lists over-ran and 70% under-ran . Over-booking contributed to 42% of over-runs, and both over-runs and under-runs could be reliably predicted , as in our current study. In a very large US study examining 56 000 cases retrospectively, Gordon et al. found that ∼ 31% of lists were predictably over-booked and ∼ 30% predictably under-booked . One important shortcoming of these studies is that they did not publish actual times for durations of specific operations. Thus, although their conclusions are broadly similar to our own, further detailed comparison is limited.
The Audit Commission's report on operating theatres emphasised ‘utilisation rate’, defined as the time for which a set of operating theatres is used, as a proportion of the time for which it is available (a target of 92.5% was proposed) . This target can be quite readily achieved (simply by the mathematics of averaging) enabling some hospitals to report utilisation rates > 100% (indicating regular, systematic over-runs) . This will not help reduce cancellations. As we have seen in our current study, an under-running list does not compensate for an over-running list: our crude utilisation rate was 93.8%, which seems superficially acceptable.
In a North American study, Wright et al.  compared the ability to predict the end-time of a surgical list using surgeons' own estimates with those using computer-based estimates. However, they did not examine ‘over-runs’ or ‘patient cancellations’ (these notions may not have been relevant to their US model). They concluded (as we have done) that surgeons' estimates of the likely duration of a list were similar to (and indeed a little more accurate) than computer estimates. In addition, they found that some surgical operations appear inherently more variable than others. Laparoscopic surgery has a co-efficient of variation of 42%, compared with hernia repair, abdominal hysterectomy and transurethral prostatectomy which are all more predictable with a co-efficient of variation of just 4%. This suggests that any system which uses estimates of operation time to plan lists might be more accurate for the latter group of operations than the former.
While many studies analyse different methods of operating list management, very few papers report actual data for the duration of common operations. Abouleish et al. have also observed that very little publicly accessible data exists on surgical duration . Table 5 shows that, for the few data we have been able to retrieve, different publications generally agree on the operation times [19, 28]. This in turn suggests that it might indeed be possible to generate (within limits) universal measures for many common operations and use these to plan lists.
Table 5. Comparison of published times from three other sources with our own study (data from Tables 1–4). Times in minutes.
| || Widdison*||Kendall et al. ||Centers for Medicare and Medicaid Services*,†||This study (Tables 1–4)|
|Breast lumpectomy|| 20||–|| 58|| 54|
|Inguinal hernia repair|| 41||70||–|| 63|
|Laparoscopic cholecystectomy|| 65||–||–|| 86|
|Lymph node biopsy|| 24||–||–|| 50|
|Unilateral varicose veins|| 32||73||–|| 86|
|Cystoscopy||–||–|| 34|| 33|
Implications of the study for clinical practice
Anecdotally, a number of factors are cited as being responsible for over-running lists (e.g. late starts, staff absence, variations in speed of operating or administering anaesthesia) [4, 24]. Our data suggest that there is actually seldom a reason to invoke any of these factors: over-booking is the main problem. Dexter et al. reached the similar conclusion that issues such as case turnover times have negligible impact on theatre efficiency .
We might plausibly use the data in Tables 1–4 to plan and manage operating lists, and it would be important to assess if this will reduce cancellations.
These operating time data are also potentially useful for costing the service; a factor which is becoming more important with the introduction of ‘Payment-By-Results’ in the National Health Service . In brief, this complex scheme requires hospitals to know the costs of delivering procedures, as they will be re-imbursed according to a single reference cost. The largest proportion of any hospital cost is invariably in staff salary [9, 13, 14, 28] and, therefore, procedure duration directly influences cost. Our data might help in calculating some average costs.
Our data also have potential implications for the management of waiting lists. Traditionally, waiting lists are described in terms of the number of patients waiting for surgery (and for how long). A more relevant method might be to focus instead on the total time required to conduct the operations for which patients are waiting. For example, our data suggest that 100 patients added per month to a waiting list for a hernia repair will need ∼ 10 days of surgical operating capacity for hernias per month for the waiting list to remain static.
Finally, it will be important for others in different surgical centres to repeat our approach to assess whether the times we present in Tables 1–4 are representative, or if they are instead more typical of a teaching hospital.