Abstract

Background. There is no clear published estimate of the prevalence of diagnosed knee osteoarthritis based on consultation in primary care. Further, little is known about the clinical history of patients who are subsequently diagnosed with knee osteoarthritis.

Objectives. Our aims were (i) to determine the prevalence of knee osteoarthritis diagnosed in primary care; (ii) to compare the medical history of patients currently diagnosed with knee osteoarthritis with that of controls; and (iii) to determine factors associated with X-ray at the time of first diagnosis.

Methods. A case–control study was carried out in one general practice in North Staffordshire. Cases aged ≥45 years were identified from computer searches for knee osteoarthritis between 1 January 1998 and 31 January 2000. Matched controls had no diagnosis of knee osteoarthritis during that same period. The medical records for both groups were examined to identify recorded knee-related symptoms, X-rays and secondary care referrals.

Results. A total of 146 cases and controls were reviewed; 49% of cases and 15% of controls had a previous (pre-1998) diagnosis of knee osteoarthritis. This gives an estimated prevalence of diagnosed knee osteoarthritis in the general population aged ≥45 years of 12.5%. Cases were more likely to have had a prior history of knee-related disorders, X-rays and referral to secondary care than controls. There was no association between an X-ray at first diagnosis and previous knee symptoms (odds ratio 0.98; 95% confidence interval 0.49–1.97) and only weak associations of X-ray with gender (male), age (under 60) of patient and time of diagnosis (pre-1998).

Conclusions. Diagnosed knee osteoarthritis is common in general practice and diagnosed patients often have a long history of knee symptoms prior to diagnosis. Further research might usefully consider what influences GPs' diagnosis of knee osteoarthritis and how diagnosis relates to management and outcome.

Bedson J, Jordan K and Croft P. The prevalence and history of knee osteoarthritis in general practice: a case–control study. Family Practice 2005; 22: 103–108.

Introduction

Epidemiological studies have estimated that symptomatic radiographic knee osteoarthritis (OA) affects 10% of adults >55 years of age.1 Most people with OA of the knee who seek medical help will be cared for in general practice, but there is no published estimate of the prevalence of knee OA based on consultation in that setting. Our first objective was to estimate the prevalence of diagnosed knee OA in a general practice.

There are a growing number of investigations of knee OA based in primary care and the community. Selection of patients for such investigations may be based on GP records of a diagnosis of knee OA, or on patient self-report of a diagnosis received from their GP. Yet very little is known about the way in which GPs may have reached this diagnosis. The onset of the joint disease is usually insidious, and the hallmark symptoms of pain and stiffness have a variable relationship with the presence and severity of X-ray changes. Therefore, there may have been a long history of previous consultation about symptoms and problems related to the knee before a formal diagnosis is made. Furthermore, we do not know whether GPs use the ‘OA’ label for presenting symptoms and signs, or reserve its use only for when radiographic evidence is available. Our second objective was to compare the previous medical records of patients currently recorded as having knee OA with those of controls who have no record of this diagnosis. Comparisons were made regarding past history of consultations for symptoms related to the knee, previous knee X-rays and referrals to secondary care for a knee problem. The third objective was to identify factors associated with a record of an X-ray of the knee at the time of first diagnosis of knee OA.

Methods

This was a case–control study in a single general practice belonging to the North Staffordshire GP Research Network. The practice is situated in an urban area and had a population of 13 731 cared for by six GPs at the time of the study.

Cases were identified from the practice's computer register and consisted of all patients who had received a computerized diagnostic code for knee OA during the 2 year period from 1 January 1998 to 31 January 2000, and were aged ≥45 at the time of that consultation. One control was identified per case, by stratified random selection from all those patients registered with the practice who had the same year of birth and the same gender as the case and who had not been recorded as having knee OA during the same 2 year period as the cases. However, they may have had a previous diagnosis of knee OA. The practice computerized system uses a standard clinical support software for general practice (EMIS; Egton Medical Information System, Leeds, UK), and records morbidity using the Read code system widely available within the British National Health Service. The quality of computerized morbidity coding and recording at the practice has been audited regularly twice per year since 1997 by the Primary Care Sciences Research Centre at Keele University. The practice moved from a paper system for recording consultations to a mainly computerized system at this date although X-rays and referral information were still recorded in the paper system.

All cases and controls had their complete ‘paper’ medical record prior to 1 January 1998 reviewed by two individual note summarizers, both health care professionals employed by the practice and with extensive experience of summarizing and recording from patient notes. Both were blinded to each other's findings, and also blinded to which patients' notes were cases or controls.

The record reviews were carried out systematically using pre-designed data forms. Each summarizer was asked to record the following features from the clinical history: (i) the date of the first recorded diagnosis of OA of the knee before January 1998 (if present); (ii) all consultations for knee symptoms of any kind up to the first knee OA diagnosis or up to 31 December 1997 if there was no pre-1998 OA diagnosis, together with their dates; (iii) all X-rays of the knee and their dates; and (iv) the date of first referral to secondary care for any clinical reason relating to the knee.

In addition, the electronic record post 1 January 1998 was examined for the 2 year period to 31 January 2000 by a medical professional familiar with this mode of data entry. In this instance, the presence of all consultations for knee-related symptoms up to and including a first diagnosis of knee OA (or to 31 January 2000 for controls with no past history of knee OA consultation) were recorded with their associated dates.

All information was then transferred in anonymized format to SPSS 11.0 for Windows2 for analysis.

The accuracy and completeness of the summarizers' recording were tested by comparing their combined results, in 10 randomly selected patients' records, with a third health care professional who checked that all the relevant information had been recorded. Out of 31 entries for the 10 patients, only one was inaccurate. This was considered as evidence for the validity of the data extraction.

Statistical analysis

Analysis 1: prevalence of diagnosed knee OA. The first analysis, addressing the first objective, compared all cases and all controls with respect to a previously recorded knee OA diagnosis. The odds ratio (OR) with 95% confidence interval (CI) was calculated using a matched case–control analysis. The prevalence of recorded diagnosis of knee OA (including current and past diagnoses) in the practice was then estimated, using both case and control data and applying this to the age and gender structure of the practice using direct standardization. Estimates of the number of patients with knee OA who had ever had a diagnosis of knee OA in the practice and who would be missed by identifying only currently recorded diagnoses of knee OA were also made, again using direct standardization.

Analysis 2: medical history of patients with current diagnosed knee OA. For the second group of analyses, addressing the second objective, cases were stratified according to the presence or absence of a recorded diagnosis of knee OA prior to January 1998. Controls with a previous recorded diagnosis of knee OA before January 1998 were excluded from further analysis in order to allow a comparison of cases (current diagnosis of knee OA) with controls who had never been attributed this diagnosis currently or in the past. In this instance, controls with no record of knee OA were assumed to represent adults who have never been diagnosed with knee OA. The proportion of this group who had a history of knee symptoms or a history of knee X-ray was compared with the proportion of cases who had a history of knee symptoms before their first recorded diagnosis of knee OA or an X-ray >1 month before such a diagnosis. Comparison was also made between the proportions in each group ever having a referral to secondary care for a knee problem. Absolute differences in these proportions were calculated with 95% CIs. Differences between the groups with respect to the total number of previous knee symptom contacts among those with at least one knee symptom contact were assessed using the Mann–Whitney test. Finally, the median time was calculated, for patients in each group with a record of knee symptoms, between the first record of a knee symptom and either (i) subsequent first record of a knee OA diagnosis (cases) or (ii) 31 January 2000 (controls with no previous history of OA).

Analysis 3: associations with a recorded X-ray at time of first diagnosis (cases only). The final analysis, relating to the third objective, studied the associations between having a knee X-ray at the time of first recorded diagnosis of knee OA (defined as a recorded date of X-ray within a month of the diagnosis) and the following factors: gender, age at diagnosis, time of diagnosis (pre- or post-1998, representing the median time of diagnosis of the cases) and previous knee symptom consultation. All cases were included. Unadjusted ORs with 95% CIs were calculated, and then a multivariate analysis was performed using logistic regression.

Analyses were undertaken using SPSS for Windows 11.02 and CIA.3

Results

At the time of the study, there were 6102 patients aged ≥45 registered at the practice, which represented 44% of the total registered population. From these, 161 cases were identified from the practice computerized register as having a diagnosis of knee OA between 1 January 1998 and 31 January 2000. A total of 146 (91%) had their records reviewed. Of the remaining 15, eight had died, six had left the practice and for one the ‘paper’ notes were not available.

One hundred and sixty-one controls were initially identified. Fifteen were excluded because they had been matched to the 15 cases whose records were not reviewed. Eleven further controls had to be replaced because five had died, five had left the practice and one record was not available.

The 15 excluded cases were older (median age 76 at 31 January 2000) than the 146 (median age 68) who had their records reviewed (Mann–Whitney test, P = 0.002). There were no differences in gender or in the time when, during the 2 year period (1 January 1998 to 31 January 2000), they had first consulted with knee OA.

Ninety-seven (66%) of the 146 included cases were female. The age range was from 46 to 88 years (mean 66.8, SD 8.6, median 68). The majority of cases (69%) had their first knee OA consultation between the ages of 55 and 74 years (Table 1).

Table 1

Age at first knee OA diagnosis (cases)

Age at diagnosis
 
n (%)
 
<45 6 (4%) 
45–54 28 (19%) 
55–64 51 (35%) 
65–74 49 (34%) 
75+ 12 (8%) 
Age at diagnosis
 
n (%)
 
<45 6 (4%) 
45–54 28 (19%) 
55–64 51 (35%) 
65–74 49 (34%) 
75+ 12 (8%) 

n = 146.

Analysis 1: prevalence of diagnosed knee OA

Seventy-one (49%) cases had had a diagnosis of knee OA recorded by the GP prior to 1998 compared with a previous recorded diagnosis of OA in 22 (15%) controls (OR 5.45; 95% CI 2.84–11.51). Estimates of the prevalence of diagnosed OA in general practice can be calculated. In a practice of 10 000, with a similar age and gender distribution to that of the study practice, identifying only those patients who had consulted and been given a diagnosis of OA within the previous 2 years would identify an estimated 106 patients with currently recorded knee OA but miss an estimated 447 (95% CI 268–626) patients with previously diagnosed OA. The prevalence of currently recorded diagnosis of knee OA is 1.1%, and the estimated prevalence of all those currently registered with the practice who have had knee OA diagnosed at some point is 5.5%. If these calculations are restricted to the registered practice population aged ≥45, these figures rise to 2.4 and 12.5%, respectively.

Analysis 2: medical history of patients with current diagnosed knee OA

Table 2 assesses the association between current diagnosis of knee OA and previous features of the medical history, comparing all cases (n = 146) with those controls (n = 124) without a previous recorded diagnosis of knee OA. The proportion of cases with a knee X-ray in the record >1 month pre-diagnosis (25%) is greater than the proportion of controls with a knee X-ray history (11%). There was also a higher proportion of cases with previous knee symptoms (cases 62%, controls 30%). More cases had been referred to secondary care with a knee problem than controls (cases 34%, controls 5%).

Table 2

Comparison of cases and controls with no history of knee OA

 Cases
 
  Controls All cases versus controls with no history of knee 

 
Previous knee OAa (n = 71, 49%)
 
No previous knee OA (n = 75)
 
Total (n = 146)
 
No history of knee OA (n = 124)
 
OA Difference in proportions (95% CI)
 
Prior history of X-rayb 19 (27%) 18 (24%) 37 (25%) 14 (11%) 14.1 (4.8 to 22.9) 
Prior history of knee symptomsc 50 (70%) 41 (55%) 91 (62%) 37 (30%) 32.5 (20.7 to 43.0) 
Referral to secondary care ever 35 (49%) 15 (20%) 50 (34%) 6 (5%) 29.4 (20.5 to 37.8) 
Median (IQR) time from symptom to diagnosis (cases) or 31 January 2000 (controls) 10.4 years (4.9–16.1) 8.0 years (2.5–23.3) 10.0 years (3.9–17.7) 10.2 years (1.9–23.5) −0.8 (−5.2 to 2.6)d 
 Cases
 
  Controls All cases versus controls with no history of knee 

 
Previous knee OAa (n = 71, 49%)
 
No previous knee OA (n = 75)
 
Total (n = 146)
 
No history of knee OA (n = 124)
 
OA Difference in proportions (95% CI)
 
Prior history of X-rayb 19 (27%) 18 (24%) 37 (25%) 14 (11%) 14.1 (4.8 to 22.9) 
Prior history of knee symptomsc 50 (70%) 41 (55%) 91 (62%) 37 (30%) 32.5 (20.7 to 43.0) 
Referral to secondary care ever 35 (49%) 15 (20%) 50 (34%) 6 (5%) 29.4 (20.5 to 37.8) 
Median (IQR) time from symptom to diagnosis (cases) or 31 January 2000 (controls) 10.4 years (4.9–16.1) 8.0 years (2.5–23.3) 10.0 years (3.9–17.7) 10.2 years (1.9–23.5) −0.8 (−5.2 to 2.6)d 
a

First knee OA diagnosis before 1998.

b

First X-ray >1 month before diagnosis (cases) or ever (controls).

c

Symptoms before diagnosis (cases) or ever (controls).

d

Median difference (95% CI).

Extrapolating the figure for previous symptoms in controls who have never had a record of knee OA to a practice population of 10 000, with 4444 registered patients ≥45 years, we can estimate that 1364 (31%) of this adult group would have consulted about a knee symptom or knee problem at some time but not had a GP diagnosis of knee OA.

Among those with at least one recorded consultation about a knee symptom, the median length of time from first symptom to first OA diagnosis for cases was 10.0 years [interquartile range (IQR) 3.9–17.7]. It was slightly less for those first diagnosed post-1998 (i.e. with no previous diagnosis of OA) than pre-1998 (Table 2). Controls with a history of a knee symptom had a median time of 10.2 years since their first consultation with a knee symptom to 31 January 2000. Cases also had more contacts with their GP for knee symptoms than controls (P = 0.002, Table 3).

Table 3

Number of knee symptom contacts for subjects with at least one knee symptom

 Cases Controlsa 
No. with knee symptom 91 37 
1 contact 39 (43%) 28 (76%) 
2 contacts 33 (36%) 5 (14%) 
≥3 contacts 19 (21%) 4 (11%) 
 Cases Controlsa 
No. with knee symptom 91 37 
1 contact 39 (43%) 28 (76%) 
2 contacts 33 (36%) 5 (14%) 
≥3 contacts 19 (21%) 4 (11%) 

Mann–Whitney test, P = 0.002.

a

With no history of knee OA.

Analysis 3: associations with a recorded X-ray at time of first diagnosis (cases only)

Sixty (41%) cases had a knee X-ray within a month either side of the first diagnosis of knee OA. Table 4 shows that having an X-ray at around the time of diagnosis had weak but not significant associations with being under 60 years old at the time of diagnosis (OR 1.46; 95% CI 0.73–2.93), having the first diagnosis prior to 1998 (i.e. having a previous diagnosis of OA: OR 1.29; 95% CI 0.65–2.57) and male gender (OR 1.31; 95% CI 0.65–2.63). No apparent relationship was found between X-ray at time of diagnosis and history of previous knee symptoms (OR 0.98; 95% CI 0.49–1.97).

Table 4

Associations with X-ray within 1 month of first diagnosis of knee OA; cases only


 
X-raya
 
No X-ray
 
ORb (95% CI)
 
ORc (95% CI)
 
n 60 (41%) 86   
Female 38 (39%) 59 1.00 1.00 
Male 22 (45%) 27 1.27 (0.63–2.54) 1.31 (0.65–2.63) 
Aged ≥60 at diagnosis 33 (37%) 56 1.00 1.00 
Aged <60 at diagnosis 27 (47%) 30 1.53 (0.78–3.00) 1.46 (0.73–2.93) 
Diagnosed 1998 or after 28 (37%) 47 1.00 1.00 
Diagnosed before 1998 32 (45%) 39 1.38 (0.71–2.67) 1.29 (0.65–2.57) 
No knee symptom prior to diagnosis 22 (40%) 33 1.00 1.00 
Knee symptom prior to diagnosis 38 (42%) 53 1.08 (0.54–2.13) 0.98 (0.49–1.97) 

 
X-raya
 
No X-ray
 
ORb (95% CI)
 
ORc (95% CI)
 
n 60 (41%) 86   
Female 38 (39%) 59 1.00 1.00 
Male 22 (45%) 27 1.27 (0.63–2.54) 1.31 (0.65–2.63) 
Aged ≥60 at diagnosis 33 (37%) 56 1.00 1.00 
Aged <60 at diagnosis 27 (47%) 30 1.53 (0.78–3.00) 1.46 (0.73–2.93) 
Diagnosed 1998 or after 28 (37%) 47 1.00 1.00 
Diagnosed before 1998 32 (45%) 39 1.38 (0.71–2.67) 1.29 (0.65–2.57) 
No knee symptom prior to diagnosis 22 (40%) 33 1.00 1.00 
Knee symptom prior to diagnosis 38 (42%) 53 1.08 (0.54–2.13) 0.98 (0.49–1.97) 
a

Within 1 month of diagnosis.

b

Unadjusted.

c

Adjusted for other presented variables.

Discussion

Our study has estimated the prevalence of documented knee OA in this general practice in those aged ≥45 years to be 12.5%. Previous epidemiological studies of knee OA estimate a prevalence of 6.1% for radiologically diagnosed OA in adults aged ≥30.4 Our figure can be interpreted as an estimate of clinical knee OA for which primary care consultation has been sought and a GP diagnosis of OA made. In addition, we estimated that 31% of adults in this age group will consult with knee problems which are not given the diagnostic label of OA at the time or subsequently. Our findings are only from one general practice and we cannot therefore assume that they represent primary care as a whole.

The majority of cases with currently diagnosed knee OA occurred in the 55- to 74-year-old age group (69%) and in females (66%). This is comparable with findings from studies of radiographic knee OA.4,5 A few cases did, however, have an initial diagnosis of knee OA before the age of 45, which is consistent with previous studies.4

We found that 49% of patients with a recently recorded diagnosis of knee OA had an earlier recorded diagnosis of knee OA, a substantially higher proportion than among those who had not consulted with knee OA in the same recent period. This emphasizes the high proportion of general practice patients with knee OA who present a long-term problem of care, and also highlights the issue for general practice intervention studies that only around a half of currently recorded cases will represent newly diagnosed knee OA. In addition, the burden of care relating to those with diagnosed knee OA is greater as measured by the frequency of contacts with their GP for knee disorders when compared with those controls who have presented with a knee-related disorder but who do not subsequently receive a diagnosis of knee OA.

Our results suggest that a first diagnosis of knee OA is preceded by a history of knee problems presented to the GP, an increased use of knee X-rays and increased referrals to secondary care for knee-related problems. This confirms that the diagnosis of knee OA tends not to be an initial response to the presentation of knee symptoms. Knee symptoms in the practice population aged over 45 are proportionately more frequent in those who eventually get a GP diagnosis of knee OA than in those who do not get this diagnosis. Patients commonly present with these symptoms well before a diagnosis of knee OA is finally made, a median of 10 years in our study. However, knee symptoms which do not progress to a diagnostic label of OA over a similar median period are also common, affecting an estimated 31% of the practice population studied here aged ≥45.

In those cases and controls with a previous history of knee symptoms, the similarity of the median time from first knee symptom to eventual OA diagnosis (cases) and from first knee symptom to 31 January 2000 (controls), i.e. 10.0 and 10.2 years, respectively, is evidence against the possibility that controls with symptoms differentially represent those in the early stages of developing knee OA. This is particularly so, given that these cases and controls were similar in age at their first knee symptom (mean difference −2.18 years; 95% CI −7.21 to 2.85).

The use of an X-ray within a month of diagnosing knee OA was modestly though not significantly associated with being aged under 60 and male gender. This might occur because knee OA is more commonly associated with elderly females.4,5 The use of X-rays in a younger male group might therefore reflect uncertainty about the diagnosis and confirmation in terms of a radiological diagnosis is being sought. There was some evidence that those diagnosed with knee OA before 1998 were more likely to have had an X-ray requested at or around the time of diagnosis than those diagnosed post-1998. This change may reflect the introduction of guidelines regarding the use of X-rays in managing knee-related disorders during the 1990s, such as those of the UK Royal College of Radiologists which argue against routine use of X-rays in diagnosing knee problems.6 Knee symptoms pre-diagnosis did not, however, appear to be associated with the use of X-rays at the time of diagnosis. This suggests that the decision to X-ray is related to the current presentation and not to an assessment of pre-history.

In conclusion, this study has estimated the prevalence of knee OA recorded by GPs in a community sample, and confirmed that this is an intermittent problem from a GP's point of view, and one in which the diagnostic label is used with caution in those who present with knee symptoms in primary care. Patients who eventually do get the diagnosis of knee OA from their GP are likely to have already presented with knee-related symptoms for some time prior to the diagnosis being applied. Given the frequency of consultations about knee symptoms, further study of what influences GPs in their diagnostic process, and the nature of the symptoms which occur in those who eventually develop knee OA, would help in the development of practical methods for appropriately diagnosing and managing knee OA.

Declaration

Funding: John Bedson has a clinical research fellowship funded by the NHS R and D Capacity Development Programme through the North Staffordshire Primary Care Research Consortium.

Ethical approval: the original work was conducted as part of the practice's auditable procedures during 2001.

Conflicts of interest: None.

We wish to thank Mrs R Rushton and Ms T Waller for all their hard work in the extraction of data for this study, and Mrs T Whitehurst for creating the computer searches to identify cases and controls.

References

1
Peat G, McCarney R, Croft P. Knee pain and osteoarthritis in older adults: a review of community burden and current use of primary health care.
Ann Rheum Dis
 
2001
;
60
:
91
–97.
2
SPSS for Windows. Release 11.0.0., 2001. Chicago (IL): SPSS Inc.
3
Altman DG, Machin D, Bryant TN, Gardner MJ (eds). Statistics with Confidence, 2nd edn. London: BMJ Books;
2000
.
4
Felson DT, Zhang Y. An update on the epidemiology of knee and hip osteoarthritis with a view to prevention.
Arthritis Rheum
 
1998
;
41
:
1343
–1355.
5
Felson D, Zhang Y, Hannan MT et al. Risk factors for incident radiographic knee osteoarthritis in the elderly: the Framingham Study.
Arthritis Rheum
 
1997
;
40
:
728
–733.
6
The Royal College of Radiologists. Making the Best Use of a Department of Clinical Radiology. Guidelines for Doctors, 4th edn. London: The Royal College of Radiologists;
1998
.