1 Introduction

The debate about health care rationing, in terms of both its justification and mechanism, is conducted largely in the absence of data. The relevant literature mainly consists of assertion, exploration of ethical principles and political analysis. In particular, an epidemiological basis for the assumption that demand for effective treatments will invariably exceed supply is lacking. Waiting list figures suggest that health services are not satisfying demand in particular areas, but this does not mean that demand is generally insatiable.

Total hip and knee replacements are effective interventions for patients with severe joint disease, resulting in large improvements in patient-related outcome measures for the majority of those undergoing these procedures. Although the number of operations performed in England has been rising each year over the last two decades [1], it has been suggested that there is still a large unmet need. However, the current evidence base is limited. Some prevalence data on severe joint disease in the community are available [2], but incidence data to estimate the annual population requirement for hip and knee replacement are lacking. Consensus criteria for case selection for total joint replacement (TJR) have been published [3, 4], but data on the impact of these criteria on annual rates are unavailable, and the implications of different thresholds for surgery, patient preference and other modifiers of the decision to recommend surgery are unclear.

We used prevalence data from the Somerset and Avon Survey of Health (SASH) [5–7] to estimate the annual population requirement for primary hip and knee replacement in England, also comparing health care utilisation by people with hip and knee disease.

2 Patients and Methods

2.1 Sampling of Patients

SASH is a population-based cross-sectional study described elsewhere [5, 7]. We used a multistage sampling strategy [8]. Forty general practices were selected from Avon and Somerset; from each practice, 702 people aged 35 years and over were randomly selected using age/sex stratification, resulting in a sample of 28,080 people with the numbers of men and women in each 10-year age band reflecting the population distribution of Avon and Somerset [5]. After exclusion of 2,034 people who had moved out of the study area, suffered from a severe mental illness or a terminal illness, or were deceased, 26,046 people were included in the study. Approval was obtained from the relevant ethics committees [7].

2.2 Screening Process

All 26,046 people were sent a screening questionnaire comprising questions on general health, utilisation of health services and symptoms of hip and knee disease. Non-respondents were sent two reminders and contacted by telephone, if necessary [5]. We screened people for hip and knee pain using a modified version of the questions used in the first National Health and Nutrition Examination Survey [9]: “During the past twelve months, have you had pain in or around either of your hips (knees) on most days for one month or longer?” Participants who reported hip or knee pain were invited for further examination either at a clinic or by home visit. Examinations were organised in two phases by location of participating practices.

2.3 Orthopaedic Assessment

Interviewer-administered questionnaires were completed on hip and knee pain and stiffness, activities of daily living, use of health services and referral to specialist care. Participants were asked whether they had received drug therapy for their joint pain in the previous year, whether they suffered from symptoms that might make them unfit for surgery (chest tightness, wheeze, breathlessness, chest pain or palpitations), and whether they would accept surgery if it were offered, with a follow-up question to elicit the reasons for their view. A clinical examination of hip, knee and lower back was carried out by a physician and a team of nurses with orthopaedic experience who had undergone a standard training programme.

2.4 Criteria for Case Selection

The New Zealand priority criteria for major joint replacement surgery [4] (New Zealand Score) were used for case selection for primary TJR. In a pilot study, agreement of the developed criteria with overall clinical judgement was found to be excellent [10]. The final composite score included sub-scores on pain (40 points), disability (20 points), clinical findings (20 points), and multiple joint disease and ability to live independently (20 points), and ranged from 0 to 100 with higher scores reflecting more severe disease. No agreed cut-off point for case selection were proposed. To reflect severe and moderate disease, we chose a priori threshold scores of 55 (primary cut-off) and 43 (secondary cut-off), respectively. Examples for degrees of pain and disability associated with these cut-off points have been published previously [7].

2.5 Statistical Methods

Incidence and prevalence calculations were carried out in an identical way for hip and knee disease [6, 7]. We calculated age- and sex-specific prevalence of already replaced joints for those responding to the screening questionnaire. Then we estimated the prevalence of joint disease warranting TJR (cut-off points 43 and 55 on the New Zealand Score) using extrapolations from the examined group of participants to the overall group reporting symptoms with age/sex-specific sampling fractions and the assumption that attendees and non-attendees were similar. New Zealand Scores were not assigned to joints which had already been replaced.

Incidence was calculated by the method of Leske et al. [11], using the increase in prevalence between consecutive age bands to calculate age-specific incidence. Its assumptions have been discussed before [7]. We considered prevalence data for replaced and severely diseased joints (New Zealand Scores ≥43 or 55 points) separately and smoothed both by fitting quadratic models across age-groups in logistic regression analysis. Then, we included age/sex-specific death rates. The annual number of TJRs needed in the population of England was calculated by multiplying point estimates and 95% confidence intervals of age/sex-specific incidences by population figures for England. This number was modified by excluding those assumed to be unfit for surgery (self-reported chest tightness, wheeze, breathlessness, chest pain or palpitations many times a day or all the time), those who had not had a trial of medical therapy in the past year, and those who indicated they might not accept surgery if offered it.

For exploratory analyses, we defined index joints as the symptomatic hip or knee with the highest New Zealand score (1,302 index hips and 2,056 index knees). We used logistic regression models to compare the use of health services and referral to specialist care for knee versus hip disease, using robust standard errors, which allowed for correlation within participants who suffered from both knee and hip disease, and adjusting for disease severity. In a further analysis, we also adjusted for age, gender, and willingness and fitness for surgery; because of missing data, this analysis was based on 2,928 index joints only. Finally, we extracted the annual number of primary total hip and knee replacements performed in English NHS hospitals from first episodes of 1997 Hospital Episode Statistics [1]. The number of procedures performed independently was estimated using data from a 1997 national survey of private hospitals in England [12].

3 Results

A total of 22,978 individuals responded to the screening questionnaire, 22,217 completed the question on hip pain and 22,379 the question on knee pain. A total of 6,416 participants reported hip or knee pain, or both (28.7%). Of those, 4,304 were invited for further examination either at a clinic or by home visit (67.1%) and 2,703 attended (62.8%). Prevalence of joint pain increased with age and was higher in women than in men [5, 6]. Attendees were more likely than non-attendees to have sought care for joint disease [5, 6].

The prevalence of primary TJRs was 22.0 joints per 1,000 people aged 35 years and over for hips (95% CI 20.1–24.0), and 8.4 joints per 1,000 people for knees (95% CI 7.2–9.5). Using the primary cut-off of 55 points on the New Zealand Score, we found an estimated population prevalence of hip disease requiring TJR of 18.2 joints per 1,000 people aged 35 years and over (95% CI 16.0–21.2), increasing to 36.5 with a cut-off of 43 points (95% CI 34.0–40.8). Respective figures for knee disease were 27.4 joints per 1,000 people aged 35 years and over (95% CI 24.7–30.1), increasing to 64.3 with a cut-off of 43 points (95% CI 60.8–67.7).

Prevalence figures translated into an estimated incidence of 54,000 hips (95% CI 35,900–72,100) for a cut-off of 55 points, and 64,800 hips for a cut-off of 43 points (95% CI 44,600–85,000), respectively, requiring total hip replacement annually in England. The annual number of primary total hip replacements actually performed in England in 1997 were 31,300 in the NHS [1] and an additional estimated 10,100 in independent hospitals [12]. Figure 10.1 shows respective numbers for knees for a cut-off of 55 points and the effect on estimates for hip and knee replacements when suitability for surgery and patients’ preferences were accounted for, comparing hips with knees [6]. The number of hips requiring TJR in England decreased only moderately to 46,200 after exclusion of those people who were unfit for surgery, had not had a trial of medical therapy, or indicated they did not want a surgical intervention (95% CI 27,500–64,900). In contrast, the number of knees requiring TJRs decreased to about half of the initial estimate after using the same criteria for exclusion of people unsuitable or unfit for surgery (29,100, 95% CI 16,300–41,900). Willingness to consider surgery had a greater impact on incidence estimates for knees than of hips, indicating that patients may be more reluctant to undergo knee replacement than hip replacement. After adjustment for severity of disease, those with hip disease were less likely than those with knee disease to have sought care from their GP [odds ratio (OR) 0.78, 95% CI 0.66–0.93), but more likely to have been referred to specialist care (OR 1.23, 95% CI 1.04–1.45), to have consulted an orthopaedic surgeon (OR 1.42, 95% CI 1.09–1.86), or to be on a waiting list for joint replacement (OR 2.39, 95% CI 1.22–4.68). Differences remained after additional adjustment for age, gender, willingness and fitness for surgery [6].

Fig. 10.1
figure 1

Estimated annual number of incident hip and knee disease requiring primary total joint replacement surgery for those aged 35–85 years in England. Expressed as 1,000s of joints required in England

Full size image

4 Discussion

The majority of people with hip and knee pain do not suffer from severe disease. A variety of outcome instruments are available to help grade the severity of symptoms, such as Lequesne’s index [13] or WOMAC [14], but these instruments were not developed for case selection for TJR. In contrast, the priority criteria from Ontario, Canada [3], and New Zealand [4] were produced for this very reason. The New Zealand priority criteria provide a numerical score which can be used to define indications for surgery from population-based data [5]. Using these criteria, we calculated the prevalence of hip and knee disease severe enough to warrant TJR. However, interventions such as TJR are provided on an incident basis, so we converted prevalence figures to incidence data. Using the primary cut-off point this translated to a calculated annual need for 54,000 total hip replacements and 55,800 total knee replacements. When estimates were adjusted for potential modifiers of the decision to recommend surgery (fitness, willingness and previous drug therapy), the incidence estimates decreased to approximately 75% for hip replacements and 50% for knee replacements. Resulting estimates of 46,200 hips and 29,100 knees requiring TJR annually corresponded approximately to the observed 41,400 hips and 29,300 knees actually replaced in England in 1997.

Rates vary greatly in different countries [15]. Surprisingly, the actually observed provision of total hip replacements exceeds the rates found in the USA [16], where adequate provision is assumed: 52 per 100,000 of the overall population, corresponding to about 32,000 operations that would be performed annually in England. In contrast, the observed provision of knee replacements amounts to only 50% of that observed in the USA, 92 per 100,000 of the overall population, corresponding to about 58,000 operations performed annually in England.

We also found interesting differences between hip and knee disease in the pattern of health services utilisation along the whole pathway to joint replacement, with a lower provision of care for knee disease. Referral patterns by primary care physicians may be of particular importance. Differences remained, however, when we adjusted for referral to specialist care, indicating that differential provision of care may also be an issue at the level of specialists. Part of this difference may be due to the fact that hip disease often deteriorates rapidly, with a relatively sudden increase in the level of pain [17], while knee disease usually deteriorates slowly, giving patients and physicians more time to adjust to and accept increasing severity [15].

The limitations of our approach to estimate population requirements for joint replacement were discussed previously [6, 7]. Despite these limitations, our data suggest that there is at most only a moderate underprovision of total hip replacements, but a considerable underprovision of total knee replacements in England. Our study indicates that this may not simply be due to a failure of the National Health System to satisfy demand, but also because of reluctance by patients and doctors to consider surgery in some instances. While the satisfaction of demand for total hip replacement, given agreed criteria, appears to be a realistic objective in England, a review of policies for the management of knee disease is needed.