Estimating hourly anaesthetic and surgical reimbursement from private medical insurers’ benefit maxima: implications for pricing services and for incentives

Authors


Dr Jaideep Pandit
E-mail: jaideep.pandit@dpag.ox.ac.uk

Summary

While some speculation surrounds annual private practice incomes of anaesthetists, little is known of the hours of work needed to generate any presumed income (the hourly rate). The benefit maxima of five private medical insurers are published in fee schedules and data on the duration of common operations are now also known. In this study we combined these to generate estimates for hourly rates of reimbursement across 78 common operations in eight surgical subspecialties, for anaesthetists and surgeons. We expected to find significant differences between insurers as a result of market competition, and we expected differences between anaesthetists and surgeons. The median (IQR [range]) rate of reimbursement for anaesthetists was £167 (132–211 [68–570]).h−1 with significant variation across subspecialties (p < 0.001); for example, cardiac surgery was best reimbursed at £283 (257–308 [229–398]).h−1 and orthopaedics the least at £146 (133–159 [81–246]).h−1. Contrary to expectations, the rates of payment to anaesthetists by insurers were similar (p > 0.17). Patterns of reimbursement for surgeons were similar to those for anaesthetists, except that surgeons were reimbursed at about twice the rate. We conclude there is a confluence of insurer reimbursement levels and we discuss potential implications of this finding. Our results also have implications for how incentives between the NHS and private practice, or within a private practice group, might be optimally managed.

In the United Kingdom (UK) consultants are contracted to work for the National Health Service (NHS) for a defined number of hours per week (40 h when full time). These hours are apportioned into 4-h aliquots known as ‘programmed activities’ (PAs, or colloquially ‘sessions’), which for anaesthetists include theatre work (or critical care, pain relief or labour ward duties), emergency service and professional activity such as teaching, training and audit [1–4]. Beyond this contractual commitment consultants are free to spend their time in family interests, hobbies, or extra research, audit or teaching (for their own benefit or by association with a university or other body). These activities are rarely remunerated but other activities that generate extra income include additional NHS work (clinical or administrative), or private practice. One inevitable factor governing the choice of how consultants use their free time is the potential financial gain (the incentive) between additional NHS work and private practice.

Additional NHS work can be included as extra regular clinical PAs within a consultant’s job plan, with each extra PA remunerated at a rate of 1/10th the consultant’s base full time NHS salary (currently ∼ £7300–9900 annually, based on seniority [5]). Given absence for leave, and including supporting professional activities, a consultant is likely to deliver ∼300–400 PAs a year [6], so each PA yields ∼£183–330 income, or £46–83.h−1 (all figures are gross, before tax). NHS hospitals also offer temporary sessions ad hoc, outwith the regular job plan, to deal with ‘waiting lists’ (where the queue of patients waiting for surgery threatens to exceed national targets). Remuneration for such ‘waiting list initiative’ work is locally negotiated and so rates may vary (perhaps ∼£400–600 per 4-h session, or £100–150.h−1; up to £200.h−1 for weekend work is common). Regardless of the national average, the local rate will be known to individual anaesthetists and together with extra PAs within a job plan, determines the hourly financial incentive for extra NHS work.

Mean annual income from private anaesthetic practice has been estimated using anonymised tax returns as £26 520 gross in 2003/4 [7], but an important yet unknown element is the time needed to generate this income (that is, the hourly rate in £.h−1). Private work is remunerated very differently from NHS work. Once an anaesthetist is engaged for service by a surgeon, reimbursement may be obtained directly by contract with a fee-paying private patient but since many are insured, the fee paid depends greatly on the maximum amount (‘benefit maximum’) the relevant private medical insurer is prepared to pay (usually direct to the anaesthetist). Where the requested anaesthetic fee exceeds this benefit maximum, a ‘shortfall’ arises for which the patient is liable. Chapman, however, documented that insurers may warn patients and surgeons that engaging an anaesthetist whose fees are consistently high may create a shortfall, which might be avoided by engaging a different anaesthetist [8].

Analysis of private insurer benefit maxima becomes logically central to any assessment of hourly reimbursement rates in private practice. In contrast to NHS ad hoc rates which vary, each insurer’s benefit maximum applies across the UK, with no geographical variation. Private medical insurers are subject to the complex rules of the Fair Trading Act 1973 and the Enterprise Act 2002 [9]: they must compete and cannot conspire to set fee levels. Accordingly, the way in which insurers approach anaesthetic fees seems to vary. The British United Provident Association (BUPA) classes each surgical procedure into one of 25 categories, each category associated with a payment value. Some insurers such as Norwich Union (CGNU; which has recently re-branded many of its operations as AVIVA) and CIGNA use a different process of categorisation into ‘units’. Others (for example, Western Provident Association (WPA), AXA PPP and CS Healthcare) simply publish tables of procedures setting out their benefit maxima for each operation individually. The details are obtainable from their websites. In this way, the same surgical procedure appears to yield somewhat different returns from the different insurers. However, what is unknown – especially as a result of the complex nature of coding, categorisation and unitisation – is whether the hourly rate for reimbursement varies either across the insurers or across the range of surgical subspecialties (that is, whether different procedures of similar duration are treated the same or differently by the insurers).

The main purpose of this study was to use data on duration of common operations to estimate these hourly rates for anaesthetic services for each of five insurers. Our main hypothesis was that because of market competition, hourly rates would differ across insurers. If the data instead showed a confluence of insurer rates then we would need to consider alternative explanations as to why this might be the case.

Our second hypothesis was that hourly rates would be similar across subspecialties, for the following reason. Within any single subspecialty, it is understandable that hourly rates across operations of similar duration might vary based on the complexity of the surgery (for example, transvaginal tape vs sacrospinous fixation within gynaecology). However across subspecialties, there appears no a priori reason to suppose why one subspecialty is inherently more ‘complex’ than another across its whole range of procedures.

For completeness we wished to compare anaesthetic with surgical hourly rates: we suspected hourly rates would be higher for surgeons. Finally, we planned to ascertain the algorithms private medical insurers might use to calculate their reimbursement levels for anaesthetic and surgical fees.

The nature of our analysis brings into objective enquiry a field which was hitherto outside the realm of analytic discourse and previously the focus only of subjective opinion. The data generated could be used locally by individual anaesthetists to make rational choices about whether expected reimbursement (in £.h−1) was sufficiently attractive to constitute a private practice incentive over extra NHS work (or over no work at all).

Methods

We estimated the duration of 78 common operations across eight surgical subspecialties (general surgery, orthopaedics, urology, gynaecology, cardiac surgery, neurosurgery, otorhinolaryngology (ENT) and ophthalmology). We defined ‘common’ procedures as those that collectively constituted more than two thirds of the cases listed over a 2-month period of list analysis for that subspecialty. We defined ‘duration of operation’ as time from the start of anaesthesia to arrival of the patient in recovery (the total of anaesthetic induction time and surgical operative time), using hospital theatre information management systems to analyse ten consecutive cases for each operation. For 54 of the surgical procedures, we sought to confirm that our estimate was within 5% of that of published data in the literature [10–18].

Next, we used the 2008–9 published benefit maxima of five major private medical insurers (termed A–E) to calculate the hourly reimbursement rate for anaesthesia and surgery, respectively, for each of the 78 procedures. We addressed the following specific questions: (i) for anaesthetic fees do hourly reimbursement rates vary by surgical subspecialty, and/or by insurer; (ii) for surgical fees do hourly reimbursement rates vary by surgical subspecialty, and/or by insurer; and (iii) is it possible to ascertain from the data any algorithms that insurers may use to set their hourly fee levels?

The influence of insurer and subspecialty on hourly reimbursement was assessed using factorial ANOVA at significance level p < 0.05, with the outcome being the mean values of ‘hourly reimbursement’ and three factors being ‘discipline’ (anaesthetics or surgery), ‘surgical subspecialty’ (eight levels, one for each subspecialty), and ‘insurer’ (five levels, one for each insurer). Post hoc t-tests with Bonferroni correction for multiple tests were used to locate any potentially significant differences as indicated by ANOVA (for comparisons across insurers, the level of significance was taken at p < 0.05/10 = p < 0.005; for comparisons across subspecialties it was taken at p < 0.05/27 = p < 0.0019). For curve fitting, we used linear and non-linear regression on median values using spss ver. 10.0 for Windows (SPSS Inc, Chicago, IL, USA), which uses least squares estimation. To explore possible associations of unrelated variables we used Pearson’s correlation coefficient and where correlations existed, we fitted appropriate regression lines or curves (p < 0.05 was taken as significant).

Results

We analysed results from 27 operations in general surgery, ten in urology, 16 in orthopaedics, 12 in gynaecology, four in cardiac surgery, three in neurosurgery, four in ENT and two in ophthalmology. The medians, range and interquartile range (IQR) of the duration of operations in general surgery, urology, orthopaedics, gynaecology and cardiac surgery were all within 5% of the times previously published [10–18]. Table 1 shows our new time estimates for those operations in neurosurgery, ENT and ophthalmology without previous reference data.

Table 1.  Median (IQR [range]) duration of operations by subspecialty, for those operations where no previous published data appear to exist.
Subspecialty and operationDuration; min
Neurosurgery
 Craniotomy for excision of tumour224 (143–307 [121–333])
 Trans-sphenoidal hypophysectomy98 (73–123 [69–131])
 Ventriculo-peritoneal shunt100 (65–135 [61–142])
Ophthalmology
 Unilateral phacoemulsification and lens prostheses32 (26 38 [24–43])
 Vitrectomy128 (72–184 [68–190])
Otorhinolaryngology (ENT)
 Septorhinoplasty63 (50–76 [45–83])
 Unilateral parotidectomy152 (68–236 [65–252])
 Tonsillectomy44 (28–61 [26–68])
 Microlaryngoscopy44 (29–59 [27–63])

Anaesthetic and surgical reimbursement by subspecialty and insurer

Figure 1 shows the hourly reimbursement rates by each surgical subspecialty, for each of the five insurance companies studied, and Table 2 the aggregate data by surgical subspecialty. For most subspecialties, reimbursement was in the range £100–200.h−1, with neurosurgery and ENT a little higher and cardiac surgery substantially higher (at ∼£300.h−1). Ophthalmology was also higher but the range was wide simply because only two operations were studied and priced very differently by insurers. There seemed little difference between insurers across the subspecialties (Fig. 2a) – for all subspecialties combined, the median (IQR [range]) of reimbursement was £167 (132–211 [68–570]).h−1.

Figure 1.

 Box plot of rates of anaesthetic reimbursement by insurer and subspecialty. The horizontal line in each box indicates the median, the edges of the box are the 25th and 75th quartiles, and the error bars the 10th and 90th deciles. The subspecialties are: surgery (GEN SURG), orthopaedics (ORTHO), gynaecology (GYNAE), cardiac, ENT, neurosurgery (NEURO) and ophthalmology (EYES). For each subspecialty, there are five boxes indicating the rates of insurers A–E in turn. The dark-filled box indicates the aggregates for each subspecialty. The dark-filled box at the extreme right indicates the aggregate for the data as a whole (TOTAL).

Table 2.  Hourly rates of reimbursement (all insurers combined) for each surgical subspecialty. Values are medians (IQR [range]).
 Anaesthetic fee; £.h−1Surgical fee; £.h−1Ratio of surgical: anaesthetic fee
General surgery157 (125–194 [80–236])343 (293–411[140–693])2.18:1
Orthopaedics144 (128–163 [91–192])356 (322–424[200–748])2.47:1
Gynaecology181 (158–205 [141–345])451 (412–544 [320–1324])2.49:1
Urology173 (96–236 [68–264])314 (277–365[176–764])1.82:1
Cardiac270 (253–307 [249–393])624 (553–661[435–903])2.31:1
ENT239 (148–267 [114–382])526 (464–609 [396–779])2.20:1
Neurosurgery224 (173–332 [131–368])495 (280–1060 [335–1260])2.21:1
Ophthalmology278 (115–374 [84–570]654 (281–860 [234–1458])2.35:1
Total167 (132–211 [68–570])385 (320–492 [140–1458])2.31:1
Figure 2.

 Box plots of reimbursement rates by insurer across all surgical procedures examined, with the last value representing the total for all insurers combined. The order of insurers from left to right on the x-axis is the same as in Fig. 1; (a) Anaesthetic fee; (b) Surgical fee; * indicates that Insurer D differed from Insurer E (p = 0.026), but no other comparisons were significant (p > 0.053).

These impressions were confirmed by ANOVA with post hoc testing (Bonferroni correction), with no difference between insurers (all comparisons p > 0.17). ‘Surgical subspecialty’ was, however, influential (p < 0.001) and the interactive term of ‘insurer’ and ‘surgical subspecialty’ was not significant (p = 0.958), indicating that reimbursement levels differ across the subspecialties regardless of the insurer. Reimbursement levels for cardiac surgery, ophthalmology and neurosurgery were similar (p > 0.083), but these were higher than the other five subspecialties (p < 0.001; Table 2). In summary, statistical testing indicated that insurers’ reimbursement maxima for anaesthetic services were similar regardless of the surgical subspecialty. For all insurers, cardiac, neuro- and ENT surgery were remunerated at a higher level compared with other subspecialties.

Figure 3 shows the hourly reimbursement rates for surgeons by each surgical subspecialty, for each of the five insurance companies studied, and Table 2 the aggregate data by surgical subspecialty. Levels of reimbursement were rather higher than for anaesthetic reimbursement, at ∼ £300–400.h−1 for most subspecialties, again with cardiac, neuro- and ENT surgery higher still (at ∼ £600.h−1). Ophthalmology was also higher but as with anaesthetic reimbursement for this subspecialty, the range was also considerably wider. For all subspecialties combined, the median reimbursement from insurers was £385 (320–492 [140–1458].h−1, (Fig. 2b).

Figure 3.

 Box plot of rates of surgical reimbursement by insurer and subspecialty. The legend for this plot is as for Fig. 1. Note, however, the different y-axis scale.

Analysis of variance confirmed that fees for surgeons were higher than for anaesthetists (p < 0.001) and also suggested that insurers varied in their fees for surgical services (p < 0.02). Post hoc testing (Bonferroni correction) located this significant effect to the single comparison of Insurer D vs Insurer E reimbursement surgical rates (significant at the p = 0.026 level), but all other comparisons were not significant (p > 0.053; Fig. 2b). Within surgical fees, ‘subspecialty’ was influential (p < 0.001) and the interactive term of ‘insurer’ and ‘specialty’ was not significant (p = 0.909), indicating that subspecialties were reimbursed differently, regardless of insurer. Reimbursement levels for ophthalmology, cardiac surgery, neurosurgery, gynaecology and ENT were all similar (p > 0.378), but different from the other three subspecialties (p < 0.001; Table 2). In summary, statistical testing indicated – as with anaesthetic fees – little difference between insurance companies but some significant differences by subspecialty.

Analysis of insurers’ possible algorithms for reimbursement

Figure 4 shows the benefit maxima set by each of the five companies studied, plotted against the median duration of the 78 procedures. The spread of data points appears similar for all insurers, raising the possibility that – regardless of the reimbursement set by each insurer – the final level is simply a singular function of the duration of operation. This notion is supported by correlation and least squares regression, which indicates that linear fits were the most suitable (p < 0001), r2 values lying in a narrow range (0.517–0.624) with similar values for slope (£2.46–2.62 .min−1). The slope for Insurer D was a little higher at £3.12 .min−1. Figure 4 is, therefore, sufficient to estimate the reimbursement provided by the two insurers (Insurers D and E) who set an individual fee for each procedure.

Figure 4.

 Scatter plot of the median reimbursement paid by each of the five insurers vs median duration of operation for anaesthetic fees. Linear regression lines are shown.

In contrast, the fee algorithms of Insurers A, B and C function differently. These insurers first assign each procedure a ‘unit value’ that is then converted into a monetary value. The value of each Insurer A unit is ∼£12 (Fig. 5a) and each minute of operating is assigned ∼0.2 units (Fig. 5b). The value of each Insurer B unit is ∼£77 (Fig. 5c) and each minute of operating is assigned ∼0.03 units (Fig. 5d). Thus for an operation of duration ∼1½ h, Insurer A’s algorithm assigns 18 units and reimburses ∼£216; Insurer B’s algorithm assigns 3 units and reimburses ∼£231 (Fig. 5a–d). Notwithstanding some variation for individual procedures (Fig. 4), the final reimbursement is therefore very similar between the two companies and it is unclear what the process of ‘unitising’ adds to the process of calculation.

Figure 5.

 Anaesthetic reimbursement calculations for three insurers (A, B, C). Panels (a), (c), (e): plots of median reimbursement by insurer vs scale used by insurer in its calculations for each operation. Panels (b), (d), (f): plots of scale used by insurer vs median duration of each operation.

Insurer C places each operation into one of 25 ‘categories’: five main categories (called ‘minor’, ‘inter’, ‘major’, ‘major plus’ and ‘complex major’); and within each of these, five levels (ranging from 1–5). In contrast to the units of Insurers A and B, the monetary value of each category (or ‘Insurer C unit’) is not constant (Fig. 5e): a polynomial relationship seems to fit this data best. In turn, Insurer C’s units are related to average operation time also in a non-linear (probably logarithmic) manner (Fig. 5f). Again, it is unclear what this process adds to pricing structure, since the final pricing outcome is very similar to Insurers A and B in terms of hourly rate, as shown in Fig. 4.

The algorithms by which surgical reimbursement is calculated appear similar to anaesthetic reimbursement but at higher overall rates. Across all operations, reimbursement is related linearly to duration of operation with high levels of correlation (r2 between 0.607–0.646; regression slopes 6.29–6.80 £.min−1; Fig. 6). Figure 6 is therefore sufficient to estimate the reimbursement provided by the two insurers (D and E) who set an individual surgical fee for each procedure.

Figure 6.

 Scatter plot of the median reimbursement paid by each of the five insurers vs median duration of operation for surgical fees.

For those insurers using more complex algorithms, the ‘units’ used by Insurers A and B are linearly related to reimbursement (Fig. 7a and c), as they are with anaesthetic fees for these insurers. Insurer A assigns the same value per unit (£12 .unit−1) to surgical as anaesthetic services (Fig. 5a vs Fig. 7a), but rewards surgeons with more units for the same procedure (Fig. 5b vs Fig. 7b), reimbursement being linearly related to duration of operation. By contrast, Insurer B rewards surgeons and anaesthetists the same units for the same procedure (Fig. 5d vs Fig. 7d) but assigns a higher value to a ‘surgical unit’ (£193 .unit−1; Fig. 7c) than it does to an ‘anaesthetic unit’ (£77 .unit−1; Fig. 5c). Similar to anaesthetic fees, the model used by Insurer C for surgical fees seems to fit best with non-linear relationships (Fig. 7e and f). Thus, as with anaesthetic fees for all these insurers, translating data into ‘units’ does not reach a different value for reimbursement for these three companies than if the simpler linear relationships of Fig. 6 were used to estimate reimbursement.

Figure 7.

 Surgical reimbursement calculations for three insurers (A, B, C). Panels (a), (c), (e): plots of median reimbursement by insurer vs scale used by insurer in its calculations for each operation. Panels (b), (d), (f): plots of scale used by insurer vs median duration of each operation.

Discussion

Our novel finding is that although insurers use different algorithms to determine their benefit maximum for anaesthetic services (Fig. 5), the final payment does not differ significantly from the simpler process of just using the duration of operation as the primary determinant (Fig. 4). Thus, contrary to our expectations, hourly rates for anaesthetic services did not differ across insurers (Fig. 2a). There was significant variation in hourly reimbursement across surgical subspecialties that was not influenced by insurer (these differences by subspecialty were consistent across all insurers; Fig. 1). Surgical rates were overall much higher than anaesthetic (Fig. 3 vs Fig. 1).

Before we consider the practical implications of these results, we consider some potential limitations of our analysis, how we handled them and how these present opportunities for further research. Our analysis may be of little interest to those who eschew private practice, or to those whose private practice scope is limited to just one or two procedures, or those exclusively involved in critical care or pain management, but it is relevant to those undertaking a wide variety of procedures. It is also important to the wider notion of ‘incentivisation’ of NHS over alternative types of work, something of potential interest to all anaesthetists.

The hourly rates we calculate rely greatly on the accuracy of our estimates for durations of operation. We generated these timings from a UK NHS hospital. It is theoretically possible that operation times differ in UK private hospitals, but unlikely as our times were all close to previously published data from different types of NHS hospitals (teaching and district) [10, 11, 17, 18], Finland [16], a variety of US hospitals [13, 14] and also from New Zealand private practice [19, 20]. With such widespread agreement (which we have previously noted in correspondence [12]) it would be unlikely that UK private hospital data alone would differ, but even if they did, it would not change our main conclusion, since these times would apply to all insurers equally. Nonetheless, further refinement of timing data is clearly possible. Although we examined a wide range of subspecialties we did not study them all, and for some (such as ENT and ophthalmology) we used only very few procedures. Although these were representative of practice in our hospital, they may not reflect the balance of procedures undertaken elsewhere.

We used insurers’ fee schedules to estimate the hourly rates, yet several factors could make the actual reimbursement received higher than our estimates. Multiple procedures (including epidural analgesia for postoperative pain relief) attract enhancements of ∼ 25% to the maxima and higher rates for complex or high-risk procedures can be negotiated. Equally, other factors will work to reduce our estimates. Time required for pre- and postoperative work lengthens the workday (2 h from ‘knife to skin’ to ‘skin closure’ may in fact take a further hour or more of the anaesthetist’s time) but attracts no extra fee. The enhanced rates for multiple procedures are not in strict proportion to the extra time needed, and there is a limit (usually three) on how many concomitant procedures are reimbursed by insurers. We suspect the net influence of the above opposing factors over a wide range of operations is neutral.

Studies set in private hospitals to verify our speculation would be ideal, but robust data are sparse for several possible reasons. There is no requirement for non-public bodies to disclose data. Since private hospitals compete for services (in contrast to NHS hospitals) some data, including timing data, might be market-sensitive. There are fewer doctors working in the private sector, which might limit wider applicability of some results, and because they rarely work as teams, data could potentially identify individuals. The median estimates we report may not reflect the practice of those with consistently faster or slower operating speeds for a particular surgery. However, it is unlikely that this will be the case for all their procedures, so using the median for a set of operations – especially for group practices that service a wide range of teams – seems appropriate.

Recently, the Association of Anaesthetists of Great Britain and Ireland (AAGBI) reported after a questionnaire survey that its members charge (and presumably receive) a median ∼17% higher than insurer norms [21]. Fuller publication of this data is awaited, but this raises several possibilities: anaesthetists may be somewhat successful in recouping shortfalls from patients; benefit maxima might be applied less rigidly than often supposed (for example, insurers may relax their maxima for company benefit schemes as opposed to individual policies); or uninsured, self-funding patients form a large proportion of consumers and are charged higher than insurers’ maxima.

We might have increased the accuracy of our overall estimate by weighting the relative market share of each insurer (for example, BUPA has a ∼40% share of the market so its rates should have a greater influence on any analysis) [22]. However, there may be some regional variation in market share and even this adjustment would not take into account those uninsured patients who choose to pay for their own private care or who choose fixed price packages offered by some hospitals. Our study could not assess some large insurers (for example Standard Life) who do not appear to publish benefit maxima.

Insurers occasionally revise their fees and since this paper was accepted for publication, Insurers A, B and D revised their fees for our cohort of operations by £23, £-1 and £20 respectively for anaesthesia fees and £27, £-13 and £5 respectively for surgeons’ fees. These changes were even more negligible when changed to hourly rates and do not alter the conclusions.

If we assume that all the available session time can be filled with cases (which is not always the case since the volume of work can vary), one session (4 h) of private practice will yield ∼£644. The annual leave and other absences of anaesthetist or surgeon leaves ∼40 working weeks in the year [6], so one regular session of private work per week is likely to generate ∼£26 700 .year−1. Two sessions (1 day) per week yields ∼£51 500 .year−1 and so on. The actual income clearly depends on several factors [7, 23] but nonetheless, we asked four group practices in the south of England for their gross (pre-tax) annual incomes per partner for 3 years 2006–9 and also asked them to estimate the average hours per week a partner devoted to private practice. While some practices appeared to have fairly predictable hours of contribution to private practice, Fig. 8 shows that others could only offer an estimate of these hours. Overall, however, the data fit our prediction rather well (but not perfectly). Despite the theoretical shortcomings discussed above, our analysis seems therefore reasonably accurate.

Figure 8.

 Estimated gross annual income per partner against number of regular 4-h sessions in private practice for four group practices. The solid line indicates the median income predicted from our data and the dotted lines the interquartile range. The filled circles (•), hollow circles (o) and crosses (+) represent three successive years of annual incomes from three group practices. One practice estimated its activity as a range (horizontal arrows indicate the range; the short vertical line indicates the median).

A comprehensive analysis of surgical reimbursement was not a prime aim of this study but we included it for completeness and comparison. Some differential may be justified since surgeons possibly deliver greater care postoperatively than do anaesthetists (although the reverse may be true for day surgery). It is interesting that the range of ratios of surgical:anaesthetic fees per hour for the UK (∼1.8–2.5; Table 2) was a little narrower than found in New Zealand (range ∼1.9–5.0) [19, 20]. Our study offers some data to inform this debate, but we do not discuss these issues further. In contrast to anaesthetists who often work across a range of surgical subspecialties, it is inappropriate to generate a similar ‘average’ for surgery since surgeons on the whole confine themselves to just one subspecialty. Also, outpatient clinic work (and minor procedures such as colonoscopy or refraction) can greatly influence surgical incomes.

Work from the United States (US) has analysed how anaesthetic groups might maximise their income (termed ‘productivity’ or ‘compensation’) [24–26]. Simple extrapolation of these studies is limited by differences between systems and working practices, yet the conclusions are broadly similar to ours: first, hourly rather than absolute income is recognised as the relevant statistic; second, any differences in hourly rates between insurers or subspecialties are important in determining the group’s total income [24–29]. Studies from New Zealand conclude the average anaesthetic reimbursement from private practice is ∼NZ$372 .h−1 (∼£136 .h−1) [19, 20]. Interestingly, in slight contrast to our result (and in contrast to the finding for New Zealand surgeons), there appears to be little variation in hourly fee levels across subspecialties for anaesthetists (except in ophthalmology where fees are higher than for other specialties, as we also found in our study) [19, 20].

Our conclusion, that there is little variation in pricing anaesthetic services across insurers (Fig. 2a), does not imply failure of market competition or collusion in price-fixing. Insurers mainly compete to offer packages of cover for medical care at attractive prices to increase their customer base [30]. Since insurers do not compete for the services of medical practitioners, it is understandable that little competition exists in setting reimbursement levels. Regulatory authorities have made clear that the public interest is best served by keeping the overall price of insurance low [31, 32], and one logical means of achieving this is to keep costs low (which includes the fees paid to healthcare providers). Therefore, each insurer may have independently reached similar conclusions on the level of anaesthetic fees that maintains its own competitiveness. Indeed, setting an hourly rate for medical services (whether naturally through market forces or by negotiation) is one rational and established method of remuneration in the private sector: it forms the basis of many ‘relative value schemes’ adopted in the US [24–29], New Zealand [19, 20] and elsewhere, and was an element of proposals supported by AAGBI in negotiations with UK insurers [33, 34]. By contrast to this understandable confluence of hourly rates, the larger variations in reimbursement for procedures of similar duration across surgical subspecialties seem less rational. Perhaps enhanced rates for cardiac and neurosurgery acknowledge greater anaesthetic peri-operative input such as high dependency care. However, this does not explain why operations in ENT or ophthalmology should generally be paid higher per hour of work than those in orthopaedics or general surgery (Fig. 1).

Analysing hourly rates in an objective way might assist in the rational setting of anaesthetic reimbursements or provide a quantitative basis for negotiating both private and NHS-related fees, including the complex arrangements surrounding Extended and Free Choice Networks [35]. Logically, it will have to be an indispensable part of developing any relative values scheme in the future. Relevant to the NHS is the notion that incentivisation can help direct any spare consultant capacity. Superficially, the financial incentive to undertake additional NHS work within a regular weekly job plan seems small (∼£56 .h−1), as compared with our estimate of income from private practice (∼£167 .h−1), and our data suggest that ad hoc NHS work would need to attract £100–200 .h−1 to be more competitive with the private practice incentive. However, these comparisons are crude as the NHS rate applies also to time spent in peri-operative care; the private estimate is specific for time spent in theatre since peri-operative work attracts no extra fee. Thus those anaesthetists who spend more time undertaking peri-operative care are likely to be penalised in terms of hourly earnings in the private sector but not in the NHS. Elective work in unsocial hours (evenings and weekends) normally attracts a premium for NHS work but not in the private sector.

What is viewed as incentive depends greatly on individual circumstances. Remuneration is not the only motivation: working conditions, freedom, personal satisfaction, interest, etc, and balancing these against family commitments and relaxation are all important. Our study is not a recipe for happiness, nor is does it provide a comprehensive analysis for each individual, but our data do form a basis for rational decision-making.

Our data could also be used to help plan incentives or avoid friction within group practices [36–38]. Partnerships may be constituted on an equity basis (all income pooled and divided equally) or alternatively, a partnership may be simply a means of allocating work (a ‘chambers’ where each partner retains the fees). Figure 1 indicates that because hourly rates seem to vary by subspecialty, chambers should best allocate work equitably by subspecialty, as well as by quantity. On the other hand, the same data suggest that equity partnerships wishing to avoid conflicts should focus on ensuring partners work or are available for similar hours, rather than analyse income-earned-per-partner. As regards overall expectations, we conclude that any anaesthetist or group devoting one full, regular day per week in general private practice is likely to earn ∼£50 000 gross annually (Fig. 8). It would be interesting if future studies, or the personal experiences of individuals, confirm this prediction.

Conflicts of interest and acknowledgements

JJP undertakes private practice and until 2005 and 2006 respectively, MEW and JJP were members of an anaesthetic group private practice. Authors or close members of their families hold policies, investments, mortgages or loans with some of the companies (or their subsidiaries) studied in this paper. MEW served over several years as member of AAGBI Council, Vice-President, and Chair of AAGBI Independent Practice Committee.

We thank Professor Alex Macario, Professor of Anesthesia & Health Research Policy, Stanford University, California, USA, for helpful comments on US reimbursement methods. We also thank the four group practices who each generously provided us 3 years of their annual incomes per partner for our study; they wish to remain anonymous.

Ancillary