OUP user menu

How should we pay doctors? A systematic review of salary payments and their effect on doctor behaviour

T. Gosden, L. Pedersen, D. Torgerson
DOI: http://dx.doi.org/10.1093/qjmed/92.1.47 47-55 First published online: 1 January 1999


We reviewed the published and unpublished international literature to determine the influence of salaried payment on doctor behaviour. We systematically searched Medline, BIDS Embase, Econlit and BIDS ISI and the reference lists of located papers to identify relevant empirical studies comparing salaried doctors with those paid by alternative methods. Only studies which reported objective outcomes and measures of the behaviour of doctors paid by salary compared to an alternative method were included in the review. Twenty-three papers were identified as meeting the selection criteria. Only one of the studies in this review reported a proxy for health status, but none examined whether salaried doctors differentiated between patients on the basis of health needs. Therefore, we were unable to draw conclusions on the likely impact of salaried payment on efficiency and equity. However, the limited evidence in our review does suggest that payment by salaries is associated with the lowest use of tests, and referrals compared with FFS and capitation. Salary payment is also associated with lower numbers of procedures per patient, lower throughput of patients per doctor, longer consultations, more preventive care and different patterns of consultation compared with FFS payment.


Economic theory and common sense both suggest that the way people are paid affects their working patterns.17 Thus, linking payment directly with output will tend to encourage workers to increase production in order to maximize income. These perceived disadvantages of linking payment with output, or fee for service (FFS), has led, particularly in the UK, to the establishment of a health-care system which either breaks the link between payment and output or at least severely dilutes its effect.

Although economic theory would suggest that FFS payment for doctors may be inefficient8 there are counter-balancing arguments, not least the presence of a formal ethical code to which doctors are expected to adhere. Such a strong system of ethics may dilute, or remove, the incentive for doctors to provide ineffective, dubious or very costly treatments merely to increase their income. Furthermore, other methods of remuneration such as the predominately capitation system in the UK, have other problems. For instance, a capitation system may be efficient as it encourages doctors to compete for patients on the basis of the quality of their care, so that by attracting patients, their income increases.9 This depends on the patient's ability to observe the quality of care, which is a tenuous assumption. Thus, in this instance an adequate level of treatment must rely on a doctor's own ethical code of conduct. This perceived problem with capitation payments prompted the government in England to introduce a range of financial incentives to encourage primary care physicians to meet public health objectives such as cervical screening. Such payments led to a steep increase in physician activity in the targeted areas.10 An alternative to capitation payments is the use of salaried employment. However, whether a salaried system of remuneration facilitates the provision of a more efficient and equitable system of care compared with either capitation or FFS payment is not known.

A salaried system would be more efficient than alternative payment methods in two situations. Firstly, if the level of activity, and therefore cost, of a salaried system is the same as the alternative but the health status of the population of patients under the care of the salaried doctor is improved. Secondly, if the level of cost is lower but the health status of the patients being cared for is the same or higher. A salaried system would be more equitable than an alternative payment method if salaried doctors give the highest priority to those patients with the greatest health needs. For example, capitation payments may encourage doctors to select patients who require lower levels of health care so that the cost of treatment per patient does not exceed the payment.

The government in England has introduced a salaried option for primary care doctors11 to improve the quality of care. The salaried option will operate alongside the existing mixed system of capitation, allowances, target payments and fee for service. It is timely, therefore, to review the international evidence with respect to the effect of salary payment systems on doctor behaviour.


Published and unpublished studies were included in this review if they explicitly and separately reported objective outcomes and measures of the behaviour of doctors paid wholly by salary compared with (either separately or mixed) other payment methods. The other main methods include: FFS, which pays the doctor a pre-determined fee for each item of service they provide; capitation payment, which provides the doctor with a payment for each patient that is registered with them; target payments, which remunerate the doctor on achieving a target level of service provision; allowances, which provide the doctor with a lump sum payment (usually to cover practice expenses) that is independent of the level of service provision; payments can be made for each clinical session (sessional payment); withhold payment involves the withholding of physician income against any budget deficits. We define physician payment method as that which directly affects the real income of the doctor. This will mean that studies of GP fundholding and US studies comparing health maintenance organisations (HMOs) with traditional FFS medical schemes, rather than physician payment methods, will be excluded.

Included studies must explicitly state that all doctors within the study group were paid by the payment systems being compared. This excludes studies in which it is not clear which methods are being used to remunerate doctors.

The following electronic bibliographic databases were systematically searched for the years: 1966–1997 (inclusive) on Medline; 1980–1997 (inclusive) on BIDS Embase; 1969–1997 for Econlit; 1981–1997 (inclusive) for BIDS ISI. The following keywords were used: physicians; practitioners; doctors (with associated MESH headings). The search was then narrowed by a free-text search using the following words: salaried; salary; and salaries. Papers were then sifted by two reviewers independently to ensure that they fitted the above criteria and then results data were extracted. When non-English language studies were identified, they were translated and included in the review. The bibliographies of located papers were also searched to identify further studies.


Thirty-two papers1243 were identified using the search strategy. Nine3543 papers were excluded from this review because we were unable to obtain a copy of the paper,3841 the outcomes reported were not objective measures of doctor behaviour,3537,42 or not all of the doctors in the study group were remunerated by salary.43 Thus twenty-three papers,1234 which reported the results of nineteen studies, fitted our search criteria and were included in this review. There were nine US studies, four Norwegian, one German, four Canadian and one South African. There was one randomized controlled trial,13 one controlled before and after30, two interrupted time series21,29 and fifteen case control studies.12,1420,2228,3134 Table 1 shows the characteristics of these studies by country and Table 2 summarizes the results by country.

View this table:
Table 1

Characteristics of studies of salary compared with other methods of payment included in the review by country

Study referencesComparatorStudy designDoctor and patient sample
RCT, randomized controlled trial; CC, case-controlled; ITS, interrupted time series.
1Hillman et al. (1989) used data from a survey of HMOs that was then updated in Pauly et al. (1990) and the same regression variables used in two equations for for-profit and not-for-profit HMO ownership status to predict hospitalization rates. This study also compared the effect of bonus payments on hospital utilization, but these results were excluded from this table because they did not reach conventional levels of statistical significance.
2There were 10 HMOs in the original study sample, however, only five explicitly stated doctors were all paid by salary so only the results for this sample are reported.
3These three papers reported results from a single study.
4This study was translated from Norwegian into English.
5This study was translated from German into English.
6The results for this study were translated from French to English.
Epstein et al. 1986 (United States)FFS2 year CC17 internists paid by salary in 2 HMOs and 10 internists in 3 hospitals paid a fee for each ambulatory test they order (FFS).
Hickson et al. 1987 (United States)FFS9 month RCT9 paediatricians in study (salary) and comparison (FFS) groups.
Wolinsky & Marder 1982 (United States)Capitation & FFSCC2521 primary care doctors in HMOs (salary and capitation) and non-federal office-based system (FFS).
Hillman et al. 1989 & Pauly et al. 19901 (United States)Capitation & FFS & withhold income & bonus paymentsCCPrimary-care doctors paid by salary, capitation, income withholds, and bonus payment (sampled in two surveys) in 302 and 317 HMOs and by FFS in traditional indemnity insurance schemes.
Gaus et al. 19762 (United States)FFS1 year CCSalaried doctors in 5 HMOs and FFS doctors in one FFS health plan.
Scitovsky 1981 (United States)FFS1 year CCOne health plan which salaries all its doctors (although they are at risk for hospital costs) and one health plan which pays doctors by FFS.
Perkoff et al. 1974, 1976 (United States)FFSCCSalaried doctors employed by a prepaid health plan were compared to private FFS doctors.
Hemenway et al. 1990 (United States)Salary plus bonus3 year ITS15 doctors (6 family practitioners, 8 specializing in internal medicine, 1 in emergency medicine).
Wilson & Longmire 1978 (United States)Salary plus profit & FFSCCIn 2 hospitals surgeons were paid by salary, 2 hospitals paid their surgeons by salary plus profit share, and 2 hospitals paid their surgeons by FFS. The records of 1724 patients were analysed.
Kristiansen & Hjortdahl 19923
Kristiansen & Mooney 19933
Kristiansen & Holtedahl 19933 (Norway)FFS1 week CC68 salaried General Practitioners (GPs) and 48 FFS GPs.
Bjorndal et al. 1994 (Norway)FFS1 week CC112 salaried GPs and 106 FFS GPs.
Bjorndal et al. 19954 (Norway)Salary/FFSCCGPs in 25 sub-municipals in Oslo.
Johnsen & Holtedahl 19974 (Norway)FFS1 week CC61 salaried GPs and 86 FFS GPs filled in workload diaries and 47 salaried and 56 FFS GPs completed questionnaires.
Burkowitz et al. 19955 (Germany)FFS1 year ITS3 GPs.
Aubin et al. 1994 (Canada)FFS2 year CC before and afterSalaried and FFS doctors treated 425 patients before preventive treatment given and 418 after. These patient groups were matched with 392 patients in a before group and 442 in an after group also treated by salaried and FFS doctors.
Battista & Spitzer 1983 (Canada)Sessional payment & FFSCCFour groups; 165 FFS primary-care doctors in urban settings; 165 FFS primary-care doctors in rural settings; 81 salaried primary care doctors employed in community health centres; 69 physicians of Family Medicine Centres who are paid on a sessional basis.
Maheux et al. 19906 (Canada)FFS1 year CC333 generalists in private practice remunerated on a FFS basis were compared to 283 salaried generalists in publicly funded institutions.
Hastings et al. 1973 (Canada)FFS1 year CC3348 patients were treated by salaried family doctors and specialists and 2052 patients were by their FFS counterparts.
Broomberg & Price 1990 (South Africa)FFS1 year CCGPs and specialists paid by salary in a HMO and by FFS in a traditional medical scheme. 44 324 patients in salaried group and 104 735 in FFS group.
View this table:
Table 2

Summary of findings from evaluations comparing salaried doctors (SDs) with those paid by alternative methods by country

Study referenceSalaried payment associated withNotes
Epstein et al. 1986 (United States)50% fewer electrocardiograms No difference in other testsControlled for patient characteristics such as age, sex, duration and severity of hypertension; practice type; and doctor characteristics.
Hickson et al. 1987 (United States)0.86 fewer visits per patient and 0.43 well child visits per patient 0.1 more emergency room visits per patient Lower percentage of patients they saw themselves (8.3 percentage points) No differences in number of patients enrolled and seen after adjusting results for doctor interest in private practicePatients were similar in age, mother's age, number of children at home and fathers in home. Paediatricians randomized to FFS or salary according to year of training and day of the week in which their clinics are held and were based in the same practice.
Wolinsky & Marder 1982 (United States)0.37 min longer consultationsNo patient characteristics were reported. Regression analysis was used to adjust 1979 net doctor incomes for differences in medical specialty, workload, sex and experience.
Hillman et al. 1989 Paul et al. 1990 (United States)13.1% reduction in average number of hospital days per 1000 enrollees per year The reduction for capitation paid doctors was 7.5% Lowest rates of inpatient hospitalization per 1000 enrollees per yearRegression analysis controlled for doctor, patient, HMO type and market characteristics (number of HMOs per capita, characteristics of local community hospitals, the percentage of unemployed and the number of physicians graduated in state per capita).
Gaus et al. 1976 (United States)30% and 56% lower amounts of preventive care (baby check-ups and prenatal visits) over a one-month period Lower annual rates of hospitalization in 2 HMOs Lower average length of stay in 2 HMOs Lower days of care in 3 HMOs No differences in ambulatory care usePatients were matched by prevalence of chronic conditions, perceptions of their health status and their health consciousness but differed in their disability rating. No controlling for HMO type or doctor characteristics.
Scitovsky 1981 (United States)0.24 lower number of visits for preventive care per patient 4.1 fewer laboratory tests per office visit 0.025 fewer X-rays per office visit Similar numbers of patient and doctor initiated visits, and levels of hospital and ambulatory care useThere was no controlling for HMO type or doctor characteristics. Patient groups were controlled retrospectively for age and sex in the statistical analysis of study results.
Perkoff et al. 1974, 1976 (United States)21% more diagnostic tests and X-rays 25% more laboratory tests 18% more preventive services 22% higher number of office visits and consultations 33% more ambulatory surgery 53% more out-patient psychiatric visitsNo doctor and patient characteristics were reported.
Hemenway et al. 1990 (United States)23% fewer laboratory tests 16% increase in the number of X-rays per patient visit 12% increase in average number of patient visits per monthNo patient characteristics data were collected for the `before' and `after' periods. No contemporaneous doctor controls were used.
Wilson & Longmire 1978 (United States)Highest percentage of preoperative X-rays Lowest rates of surgery and tended to operate later Highest wound infection rate of 11%Patients in one hospital with salaried surgeons had younger patients while the other had fewer female patients than other hospitals. No information was reported on surgeon characteristics and their numbers.
Kristiansen & Hjortdahl 1992 Kristiansen & Mooney 1993 Kristiansen & Holtedahl 1993 (Norway)More likely to provide urine microscopy Less likely to schedule home visits 1.1 min longer patient consultations 14 fewer surgery-based consultations per week No differences in weekly number of hours spent consulting and the proportion of return visitsPatient populations were similar with respect to: doctor turnover, size, travel time to doctor, travel time to hospital but were not similar with regard to GP to population ratio. GPs were comparable with respect to age, sex, number of years since graduation, place of training and job satisfaction but not number who collected patient co-payments.
Bjorndal et al. 1994 (Norway)0.31 fewer consultations per hour of patient contact 0.35 more telephone consultations per hour Similar levels of patient turnover and consultation lengths More time spent on: administrative work, meeting with other personnel, antenatal and well-baby clinics, schools and nursing home visits and in consultation with patientsThe patients treated by the two groups of doctors were similar with regard to sex, age, place of birth, status of employment and whether they lived alone. Both groups of doctors were similar in age, registration of ancillary personnel but there were more female GPs in the salaried group who had fewer patient consultations compared with their male colleagues.
Bjorndal et al. 1995 (Norway)Lower numbers of consultations Larger waiting lists Lower incomeNo information on the characteristics of GPs or patients in either payment groups was reported.
Johnsen & Holtedahl 1997 (Norway)Similar average numbers of working hours GPs (51.3 h vs. 54.5 h) 4.3 more out of hours/emergency visits per week 6.6 fewer face-to-face 2.8 fewer telephone consultations 2.9 more home visits per week 2.3 min longer consultationsRegression analysis was used to control for GP sex, age, practice setting. No patient characteristics were reported.
Burkowitz et al. 1995 (Germany)134% fewer home visitsNo details were available about patient and doctor characteristics or sample sizes. Home visits by nurses over the same period were not affected by the change in GP remuneration.
Aubin et al. 1994 (Canada)More likely to provide hypertension screening (odds ratio of 3.67)Patients being treated in salaried and non-salaried groups were matched according to age, sex and type of visit. Logistic regression was used to control for doctor age and number of years in practice.
Battista & Spitzer 1983 (Canada)The following activities more than FFS doctors but less than doctors paid per session (SP): mammographies; testing for occult blood in stools; pap tests; anti-smoking counselling; teaching of breast self-examination activities Fewer chest X-rays and sputum cytology regarded as not effective in the detection of lung cancerNo patient characteristics were reported. Results were adjusted for age, sex and language of doctor, type of practice (solo vs. group), number of patients seen per hour, education activities.
Maheux et al. 1990 (Canada)More community health activities Less emergency care and care in hospitals and external clinicsNo patient characteristics were reported. Salaried doctors tended to be younger, more often female, have different attitudes, work in group rather than solo practices and in less densely populated areas than FFS doctors.
Hastings et al. 1973 (Canada)5.8 more check-ups per 100 person years 9.2 more immunizations Greater use of radiology services (7% hospital and 59% ambulatory) Greater use of laboratory testing (28% in-hospital and 92% ambulatory) Lower percentage of pre-school children who did not receive any doctor (8 percentage points) 24% lower hospital use 24 percentage points higher rate of circumcision Three times lower rates of tonsillectomy and adenoidectomyThe two patient populations were similar with regard to education, family income, religion, country of birth, and length of residence in area. Populations were not matched according to age, sex or family composition. No doctor characteristics were reported.
Broomberg & Price 1990 (South Africa)1 less consultation per person per year 0.33 fewer radiology ordered per person per year 0.38 fewer pathology investigations ordered per person per year 0.02 lower hospital admissions per person per year 0.11 fewer hospital days per person per year 0.2 lower average length of stay per person per yearPatient populations were similar with respect to age, sex, race and income distribution. No information reported on doctor characteristics or on their number within the study and control groups.

There are two important results. Firstly, that there is a paucity of studies which reported health status outcomes. Only one of the studies attempted to identify health status as an outcome, and the measure in this study was wound infection rates. Secondly, there were no studies that examined whether different patient groups were differentially treated by salaried doctors. Mostly studies reported type and quantity of care outcomes such as volume of tests, consultations, referrals, and days spent in hospital. The included studies did show that salaried doctors tended to differ in terms of the quantity of care they provided, compared to doctors paid by FFS. Nine studies showed that salaried doctors had lower volumes of consultation than FFS doctors. Five out of the nine studies that looked at the volume of tests and X-rays requested found that salaried doctors ordered fewer tests than FFS doctors. In four studies from the US, Canada and South Africa, salaried doctors were associated with lower levels of hospital use compared with FFS doctors. Two studies found that salaried surgeons undertook less surgery, however, one found that salaried surgeons had a higher wound infection rate than their FFS and salary plus profit colleagues. The two studies that compared salary payment with capitation found that salaried doctors had lower volumes of consultations, ordered fewer tests and X-rays and were associated with lower levels of hospital use. These studies showed that FFS was associated with the highest levels of service provision compared with salary and capitation.

Salaried employment was associated with different patterns of consultation compared with FFS. Norwegian evidence shows that salaried doctors may not differ from FFS doctors in the proportion of telephone to face-to-face consultations, but they provided more emergency/out of hours care in one study. Two of the three studies from Norway and Germany indicates that salaried doctors tended to have surgery-based consultation, rather than home visits, particularly for elderly people. Furthermore, US and Norwegian evidence suggests that salaried doctors tended to have longer consultations compared with FFS. In one Norwegian study salaried doctors spent more time on administration than FFS doctors. Four out of the five studies that investigated the provision of preventive care found that salaried doctors provided more than FFS doctors.


This paper set out to review the literature on the effect of salaried payment on the clinical behaviour of doctors. Whilst some broad trends can be identified from the studies included in this review, the evidence is incomplete. Firstly, most of the included studies only compared salary payment with FFS. Therefore, our review does not provide sufficient evidence on the efficiency and equity implications of salary compared with, for example, capitation payment. Secondly, most of the studies reported type and quantity of care outcomes. Since in most studies lower levels of care are provided, the implication is that the cost of health care would be lower under a salary system. However, only one of the studies identified reported a proxy for the health status of patients. In addition, none of the studies examined differences in resource use by patients with either similar or different health status. Therefore, we cannot make any conclusions about the efficiency or equity of salary payment.

The internal and external validity of the studies in this review determine the extent to which the findings can be generalized to other settings and countries. Table 2 indicates whether each study controlled for physician, patient and practice setting characteristics, either in the design or the data analysis. In the two studies with no control group, it is possible that the same changes in behaviour may have been observed in other physicians and that it was not due to the change in payment. For those studies that did have a control group, eight did not control for doctor characteristics. Eight of the studies in this review also did not control for patient characteristics. In some of the US studies, lower rates of utilization may be due to healthier patients self-selecting to HMOs. Ten studies failed to control for any aspects of practice setting. Doctor behaviour may be influenced by other incentives such as organisational level payments, limited drug lists, therapeutic protocols and high levels of peer review.

Although the quality of studies in this review is not necessarily high, there are two main reasons why non-randomized and uncontrolled studies merit serious consideration. Firstly, if the more robust studies systematically differ in their respective results, then there is no justification for discarding any less rigorous evidence. In this review, the variability of the results of a study was independent of the rigour of its design. Secondly, there is a limit to what kind of studies can be undertaken and therefore, to what we can reasonably know about salaried systems. Randomized controlled trials are the most rigorous type of study design, however, there are two main reasons why they are impractical in this research area. Firstly, few physicians would be willing to participate in randomized trials where their income was randomly determined. Secondly, it would not be possible to blind the physician to the way in which they were paid. Therefore, our call for more high-quality research is a cautious one.

Overall, it was thought worthwhile to identify broad trends across the studies, given that this may be the highest quality evidence attainable within the methodological constraints inherent in this particular subject area. The finding in this review that salary payment is associated with lower levels of care compared with capitation is in line with prior expectations. Since in order to increase income doctors paid by capitation need to attract patients to their practice, perhaps by offering more services as an indicator of higher quality.44 We would also expect that salary payment resulted in lower levels of care compared with FFS payment since salaried doctors cannot increase their income but may minimize effort44 by seeing the minimum number of patients.

We would not expect differences between FFS and salary payment in the amount of preventive care provided unless fees were payable for specific items of service. However, several Canadian studies found evidence that salaries resulted in higher levels (although the US studies were contradictory). This may be for two reasons: firstly, some doctors with a preference for this kind of care are attracted more to salaried posts; and secondly, prevention might be seen as a way of reducing future workload.

To conclude, this review has shown that there is a dearth of high-quality studies evaluating the impact of salaried payment on efficiency and equity. The results show that salary payment reduces activity compared with FFS and that capitation appears to have a similar but subdued effect, although more research is required in this area. Therefore, the results of our review suggest that if cost containment is a key policy aim of government then salaried payment systems are more likely to achieve this compared with FFS and possibly more effective than capitation systems. However, cost containment by itself may be inefficient if it results in the provision of sub-optimal care. Indeed, salaried systems may not be ideal for pursuing public health policy such as population immunization which depends upon high quantities of care being provided.


We would like to thank Dr Peter Bower, Professor Hugh Gravelle, Professor David Wilkin and Andrew Lloyd for their comments on previous drafts of this paper. The usual disclaimers apply.


  • Address correspondence to T. Gosden, National Primary Care Research and Development Centre, 5th Floor Williamson Building, University of Manchester, Oxford Road, Manchester M13 9PL


View Abstract