a prospective cohort study - Keele Research Repository

Downloaded from ard.bmj.com on April 12, 2011 - Published by group.bmj.com

Extended report

The influence of consulting primary care on knee pain in older people: a prospective cohort study M Blagojevic, C Jinks, K P Jordan Primary Care Musculoskeletal Research Centre, Primary Care Sciences, Keele University, Keele, Staffordshire, UK Correspondence to: M Blagojevic, Primary Care Musculoskeletal Research Centre, Primary Care Sciences, Keele University, Keele ST5 5BG, UK; [email protected] Accepted 20 January 2008 Published Online First 31 January 2008

1702

ABSTRACT Objective: To investigate whether consulting a general practicioner (GP) in the 3 years after reporting knee pain is linked to better knee pain outcomes (reduced presence or severity of knee pain) at the end of the 3 year period. Methods: We undertook a population-based cohort study linking baseline (2000) and follow-up (2003) surveys to primary care medical records. The cohort comprised 1577 adults aged 50 and over registered at 3 general practices in North Staffordshire, UK, who reported knee pain in a baseline survey. The main outcome measures were selfreported prevalence of knee pain and severity of knee pain in the follow-up survey. The relationship between consultation and future knee pain status was adjusted for an individual’s propensity to consult given related demographic and health-related factors. Results: In persons who consulted for knee pain, 91% reported knee pain at 3 years, compared with 73% of those who did not consult (adjusted odds ratio (OR) 2.25; 95% CI 1.56 to 3.26). Among persons reporting severe knee pain or disability at baseline (n = 669), those who consulted for knee pain were more likely to report severe knee pain or disability at 3 years than those who did not consult (82% v 65%, adjusted OR 1.93; 95% CI 1.27 to 2.93). Conclusions: Older adults with knee pain continue to have persistent problems regardless of whether they consult primary care or not. Further research is needed to identify more effective means of reducing the burden of knee pain in the community.

problem and to previous use of health care services for knee problems.7 However, the decision to consult healthcare for a problem is complex and may be influenced by many factors; not just the morbidity itself but factors that may predispose a person to consult about a problem (for example, gender, psychological status), or that encourage or prevent them from consulting (previous experience of health services or social circumstances).8 Previous research has suggested decisions to consult a GP for knee pain are linked to perceptions of ageing, wear and tear, beliefs about treatments and their effectiveness and perceived seriousness of the condition.9–12 There may also be influences on the course of knee pain,13 14 which may be related to the decision to consult, for example, on other joint pain or health problems. What is not known is whether those who choose to consult their GP have different long-term outcomes to those who do not seek primary care. In particular, whether those who do not consult might benefit from seeking primary care, and if those who do consult appear to benefit. We have compared 3-year outcomes among consulters and non-consulters in order to evaluate whether health care is currently shifting the natural history of clinical osteoarthritis in an older population sample. We have used the method of propensity scoring to allow for the potential differences in risk of poor outcomes among those who do and do not consult,

Musculoskeletal problems are a common reason for consultation to general practice in the UK with an estimated 20% of the adult population having a recorded consultation during the course of a year.1 Osteoarthritis is the most frequent of these problems in the older population. Despite the wide availability of interventions such as analgesics, anti-inflammatory drugs and physiotherapy, it is not clear whether current care of osteoarthritis is improving the long-term outcome of the condition. Knee pain in older adults is usually attributed to osteoarthritis and is a common and a major source of disability.2–5 We have previously reported that nearly half of adults aged 50 and over living in the community reported having knee pain during the previous year, but only a third of these reported consulting their general practicioner (GP) for this problem during that time period. In persons who reported severe knee pain or limitations with physical function, half had no recorded primary care consultation for knee pain during a 3-year period.6 Future consultation for knee pain among older people appears to be linked to chronicity of the

METHODS This was a population-based prospective cohort study linking self-reported data from baseline and 3-year follow-up postal surveys to primary care medical records. The study was approved by the North Staffordshire Local Research Ethics Committee.

The surveys The baseline survey in April 2000 was mailed to all people aged 50 and over (n = 8995) registered at three general practices in the Keele GP Research Partnership in North Staffordshire, UK.6 15 In the UK most people are registered with a general practice regardless of whether they use the service or not. Participants were asked for consent to link survey responses with their medical records of consultations. The follow-up survey took place 3 years later. Both surveys included demographic questions, the Knee Pain Screening Tool (KNEST),6 15 the Short Form-36 (SF36),16 the Hospital Anxiety and Depression Scale (HADS)17 and the Western Ontario and McMaster Universities Osteoarthritis Ann Rheum Dis 2008;67:1702–1709. doi:10.1136/ard.2007.080259


Extended report Table 1 Potential covariate baseline factors assessed for inclusion in propensity score modelling Category

Factor

Demographic

Age at time of survey (grouped into 50–64, 65–74 and 75+) Gender General practice registered with Further education after leaving school (self-report from survey) Cohabiting with spouse or partner (self-report from survey) Deprivation (based on Townsend score with deprived rated as being above upper tertile); socio-economic classification (based on higher classified of respondent’s and their partner’s current/most recent job)24 Body mass index rating of normal, underweight, overweight or obese (from self-report of height and weight from survey); anxious or depressed, based on being above top tertile on HADS anxiety or depression scale (from survey) Widespread pain, as shaded on manikin (from survey) Physical functioning scale (from SF36) Social functioning scale (from SF36) Body pain scale (from SF36) General health scale (from SF36) Previous knee injury ever that required consultation with GP (from KNEST) Pain in one or both knees (laterality; from KNEST) Knee pain for 3 months or longer in past year (chronic; from KNEST) Self-report of use of non-GP services for knee pain in 12 months prior to survey (from KNEST) Severity of knee pain (from WOMAC) Consultation for knee pain in 18 months before baseline (from records) Non-knee pain musculoskeletal consultation between baseline and follow-up (from records) Comorbidity: number of Read Code chapters (excluding musculoskeletal chapter) for which the GP had entered a code between baseline and follow-up: grouped into none/low (0–2 chapters), medium (3–5 chapters), high (6+ chapters) (from records)

General health

Knee-related

Consultation behaviour

HADS, Hospital Anxiety and Depression Scale; KNEST, Knee Pain Screening Tool; SF36, Short Form-36; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index.

Index (WOMAC).18 A further question asked subjects whether they consented to viewing of their medical records. The KNEST contains a question asking respondents to report whether they have had pain in or around the knee within the past 12 months. Subjects responding positively continued to answer further questions about the laterality (one or both knees) and chronicity of their knee pain, their use of GP and non-GP services for knee pain in the previous 12 months, and they also completed the WOMAC. The WOMAC has 24 questions covering knee-related pain, stiffness and physical function over the previous 48 h. Respondents were defined as having severe pain or physical function difficulty if they reported ‘‘severe’’ or ‘‘extreme’’ on at least one item on the

Table 2 Baseline characteristics of those reporting knee pain at baseline

n Age: 50–64 65–74 75+ Female Obese (BMI.30) Knee pain for more than 3 months in past 12 months Severe knee pain Pain in both knees Self-reported GP consultation for knee pain in 12 months before baseline

All, n (%)

Responders to follow-up, n (%)

2732

1577

1434 849 449 1609 558 1390

(52) (31) (16) (59) (21) (53)

1236 (47) 1345 (51) 850 (32)

834 519 224 927 330 843

(53) (33) (14) (59) (22) (55)

689 (45) 775 (51) 493 (32)

BMI, body mass index; GP, general practicioner.

Ann Rheum Dis 2008;67:1702–1709. doi:10.1136/ard.2007.080259

pain or physical function scales. Subjects who did not report any severe or extreme problem and answered at least 4 out of 5 pain items and 14 of the 17 physical function items were rated non-severe.19

The medical record linkage The participating practices undergo a cycle of assessment, feedback and training in their use and quality of computerised coding of the morbidity of their patients.20 Clinicians use Read Codes to record morbidity information from consultations. Read Codes are a commonly used morbidity coding system in the UK and form a hierarchy of diagnostic and process of care codes.21 GPs can also add textual information about a consultation (‘‘consultation text’’) alongside the code. Knee consultations among survey participants were identified from the medical records using knee-related Read Codes and through a search of the consultation texts by two researchers independently. Knee-related Read Codes were determined following classification by four GPs. The time period covered was the 3-year period from return of baseline questionnaire to return of the follow-up questionnaire. Our previous work has suggested there is no influence of completing a knee pain questionnaire on consulting for knee pain.22 A further search of the medical records was performed for all musculoskeletal consultations (defined as a Read Code under chapter N, ‘‘Musculoskeletal and Connective Tissue Diseases’’, of the Read Code hierarchy)21 during that same time period.

Statistical analysis Only persons who reported knee pain at baseline were included in the analysis. The outcome was reporting of knee pain at follow-up. Comparing the effectiveness of care in prospective observational studies is difficult due to the non-randomised 1703


Extended report Table 3 Unadjusted associations of baseline covariates with reporting of knee pain at follow-up Knee pain at follow-up, n (%) Demographic Age:* 50–64 65–74 75+ Gender: Male Female General practice:* A B C Further education: No Yes Cohabitating: Yes No Deprived area: No Yes Social class: Managerial Intermediate Routine General health BMI:* Normal Underweight Overweight Obese Most anxious or depressed:* No Yes Widespread pain:* No Yes Physical function,* mean (SD){ Social function,* mean (SD){ Body pain,* mean (SD){ General health,* mean (SD){ Knee related Previous knee injury:* No Yes Laterality:* Unilateral Bilateral Chronicity:* ,3 months .3 months Use of non-GP services:* No Yes Severe WOMAC pain:* No Yes Consultation behaviour Pre-base knee consultation:*

No knee pain at follow-up, Odds ratio n (95% CI)

640 (77) 414 (80) 183 (82)

194 105 41

1 1.20 (0.91 to 1.56) 1.35 (0.93 to 1.97)

496 (77) 741 (80)

154 186

1 1.24 (0.97 to 1.58)

364 (77) 533 (76) 340 (84)

107 169 64

1 0.93 (0.70 to 1.22) 1.56 (1.11 to 2.20)

1053 (79) 139 (77)

289 41

1 0.93 (0.64 to 1.35)

895 (78) 334 (80)

256 84

1 1.14 (0.86 to 1.50)

820 (78) 414 (80)

238 101

1 1.19 (0.92 to 1.55)

340 (76) 274 (78) 546 (79)

106 79 144

1 1.08 (0.78 to 1.51) 1.18 (0.89 to 1.57)

344 32 528 285

(71) (80) (79) (86)

144 8 137 45

1 1.67 (0.75 to 3.72) 1.61 (1.23 to 2.11) 2.65 (1.83 to 3.84)

693 (77) 527 (81)

210 121

1 1.32 (1.03 to 1.70)

983 (76) 306 1 254 (88) 34 2.33 (1.59 to 3.40) 50.49 (29.90) 62.91 (29.89) 71.25 (30.13) 47.51 (24.32) 54.29 (23.23)

78.30 (26.86) 58.56 (25.13) 60.96 (23.69)

631 (74) 561 (84)

224 106

1 1.88 (1.45 to 2.43)

553 (73) 656 (85)

206 119

1 2.05 (1.60 to 2.64)

472 (69) 737 (87)

212 106

1 3.12 (2.41 to 4.05)

713 (73) 524 (88)

268 72

1 2.74 (2.06 to 3.63)

606 (72) 607 (88)

235 82

1 2.87 (2.18 to 3.78)

Continued

1704

Table 3

Continued

No Yes Non-knee musculoskeletal consultation: No Yes Comorbidity:* None/low Medium High

Knee pain at follow-up, n (%)

No knee pain at follow-up, Odds ratio n (95% CI)

975 (77) 262 (86)

296 44

1 1.81 (1.28 to 2.55)

552 (77) 685 (80)

166 174

1 1.18 (0.93 to 1.51)

371 (76) 615 (79) 251 (83)

120 167 53

1 1.19 (0.91 to 1.56) 1.53 (1.07 to 2.20)

*Included in propensity score modelling; {p,0.001. BMI, body mass index.

nature of the groups (defined here by recorded consultation for knee pain during the 3-year period) and the likelihood that there will be baseline differences between the groups prior to consultation. It is important, therefore, to take into account the possible confounding effects of such differences on the association of consultation with long-term outcome of knee pain as precisely as possible. This might be performed using logistic regression, which adjusts for each potential confounder (covariate). The method of propensity scores is an attractive alternative because it avoids the issue of building a parsimonious model with the covariates.23 Instead, it is based on using all the potential baseline confounders to determine a single propensity score for each responder. This score reflects the likelihood (or propensity) of a responder seeking primary care for knee pain given their baseline covariate characteristics. This propensity score can then be adjusted for in the final analysis assessing the effect of consultation on outcome. The details of the propensity score approach are given in the Appendix. Essentially, all responders are placed into one of five groups based on their score relating to their propensity to consult. There should be no difference on baseline characteristics between those who consult and those who do not within each propensity score group. The covariates examined here are listed in table 1 and included those related to the knee problem (including prior use of health care, severity and chronicity), general health status (including comorbidity) and demographic factors. These covariates were based on those previously associated with severity of knee pain in cross-sectional analyses,19 or which were significant predictors of consultation7 based on the Anderson–Newman model8 that suggests predisposing, enabling and need factors as determinants of healthcare use. For the first analysis, those who responded to the follow-up survey, had reported knee pain at baseline and had consented to medical record review were dichotomised based on whether they had a record of a knee consultation between baseline and 3year follow-up. Logistic regression was then used to assess the association of consultation status with knee pain at follow-up, adjusting solely for the propensity to consult. Knee pain at follow-up was the dependent variable with consultation status and propensity score group as the explanatory variables. In order to explore whether the closeness in time of the consultation to the baseline or follow-up survey date affects outcome, we then further dichotomised consulters in two ways. First, we split consulters into ‘‘early’’ consulters (those who consulted for the first time after the baseline survey within a Ann Rheum Dis 2008;67:1702–1709. doi:10.1136/ard.2007.080259


Extended report Table 4 Unadjusted associations of baseline covariates with reporting of severe knee pain at follow-up

Demographic Age:* 50–64 65–74 75+ Gender: Male Female General practice:* A B C Further education: No Yes Cohabitating: Yes No Deprived area:* No Yes Social class: Managerial Intermediate Routine General health BMI:* Normal Underweight Overweight Obese Most anxious or depressed:* No Yes Widespread pain:* No Yes Physical function,* mean (SD){ Social function,* mean (SD){ Body pain,* mean (SD){ General health,* mean (SD){ Knee related Previous knee injury:* No Yes Laterality:* Unilateral Bilateral Chronicity:* ,3 months .3 months Use of non-GP services:* No Yes Consultation behaviour Pre-base knee consultation:*

Non-severe or no Severe knee pain at follow- knee pain at follow-up, n up, n (%)

Odds ratio (95% CI)

202 (69) 179 (74) 106 (77)

89 62 31

1 1.27 (0.87 to 1.86) 1.51 (0.94 to 2.41)

169 (72) 318 (73)

67 115

1 1.10 (0.77 to 1.56)

111 (67) 215 (72) 161 (79)

56 83 43

1 1.31 (0.89 to 1.97) 1.89 (1.19 to 3.01)

427 (73) 40 (70)

160 17

1 0.88 (0.49 to 1.60)

339 (71) 145 (76)

135 45

1 1.28 (0.87 to 1.89)

271 (29) 216 (24)

113 68

1 1.33 (0.93 to 1.88)

102 (71) 104 (69) 243 (74)

41 46 85

1 0.91 (0.55 to 1.50) 1.15 (0.74 to 1.78)

90 13 212 154

(58) (72) (74) (83)

65 5 76 31

1 1.88 (0.64 to 5.53) 2.02 (1.33 to 3.04) 3.59 (2.18 to 5.92)

178 (64) 299 (79)

99 82

1 2.03 (1.43 to 2.87)

342 (71) 139 145 (77) 43 27.65 (22.93) 42.25 (28.19) 52.37 (30.68)

69.95 (29.39)

30.81 (18.89) 41.56 (21.98)

41.52 (20.67) 51.21 (22.34)

1 1.37 (0.93 to 2.03)

213 (68) 244 (76)

100 79

1 1.45 (1.02 to 2.05)

176 (65) 306 (79)

96 81

1 2.06 (1.45 to 2.92)

73 (58) 406 (77)

54 122

1 2.46 (1.64 to 3.69)

187 (66) 300 (78)

96 86

1 1.79 (1.27 to 2.53)

Continued

No Yes Non-knee musculoskeletal consultation:* No Yes Comorbidity:* None/low Medium High

Non-severe or no Severe knee pain at follow- knee pain at follow-up, n up, n (%)

Odds ratio (95% CI)

340 (70) 147 (80)

146 36

1 1.75 (1.16 to 2.65)

163 (69) 324 (75)

73 109

1 1.33 (0.94 to 1.89)

102 (71) 245 (71) 140 (77)

41 100 41

1 0.99 (0.64 to 1.52) 1.37 (0.83 to 2.27)

*Included in propensity score modelling; {p,0.001. BMI, body mass index.

year of this baseline survey) and ‘‘late’’ consulters (the remainder) and assessed association with knee pain at followup. Second, we split consulters into those who had their final consultation between the surveys within a year of the follow-up survey (to match the recall period for knee pain) and those who had their final consultation more than a year before follow-up. These analyses were repeated on the subgroup of respondents who reported severe (based on the WOMAC) knee pain at baseline, with the outcome being severity of knee pain at follow-up. Finally, the effect of consulting for any (knee or non-knee related) musculoskeletal consultation on future knee pain status was assessed by dichotomising subjects into consultation status based on any musculoskeletal problem (rather than just knee problems) and analysed using the same process. Statistical analysis was performed using SPSS V 14.0 for Windows (SPSS Inc., Chicago, Illinois, USA).

RESULTS Response A total of 6792 persons responded to the baseline survey (adjusted response 77%), of whom 5784 were alive and still registered at the practices at the time of the follow-up survey. Of these 5784, 2592 had reported knee pain in the baseline survey and 1577 (61%) of these responded to the follow-up survey, answered the knee pain question and consented to medical record review. Response to the baseline survey was higher in females (79%) than males (75%, p,0.001) and responders were slightly older than non-responders (mean difference 1.5 years, 95% CI 1.0 to 2.0). Among persons reporting knee pain in the baseline survey, those who responded to the follow-up survey and consented to medical record review were slightly younger (mean difference 1.2 years, 95% CI 0.5 to 2.0), more likely to have chronic knee pain at baseline (55% v 50%, p = 0.02), but less likely to have severe knee pain at baseline (45% v 50%, p = 0.01). There were no baseline differences by gender, body mass index, history of knee injury, laterality of knee pain, or in self-reported consultation or use of other services for knee pain prior to baseline. Table 2 compares baseline characteristics of the 1577 subjects who form the cohort for this analysis with all subjects who reported knee pain at baseline.

Persistence of knee pain Continued


Table 4

A total of 507 respondents (32%) had a recorded knee consultation between baseline and follow-up. In all, 1237 1705


Extended report Table 5 Association of consultation in primary care with reporting of any knee pain at 3 years in those reporting any knee pain at baseline Knee pain at follow-up

No consultation Consultation for No consultation Consultation for

for knee problem knee problem for any musculoskeletal problem any musculoskeletal problem

Yes, n (%)

No, n

OR* (95% CI)

OR{ (95% CI)

776 461 381 856

294 46 151 189

1.00 3.80 (2.73 to 5.29) 1.00 1.80 (1.40 to 2.30)

1.00 2.25 (1.56 to 3.26) 1.00 1.31 (0.92 to 1.87)

(73) (91) (72) (82)

*Unadjusted; {adjusted for potential confounders using propensity score modelling. OR, odds ratio.

(78.4%; 95% CI 76.3% to 80.4%) of the 1577 respondents reported knee pain in the follow-up survey; 91% of those who consulted and 73% of those who did not consult reported knee pain at follow-up (unadjusted OR 3.80; 95% CI 2.73 to 5.29). Table 3 details the covariates that were individually related to knee pain status at follow-up and were included in the propensity score modelling. Following the propensity score modelling, consulting for knee pain had a reduced, but still significantly increased association with reporting of knee pain at follow-up (OR 2.25; 95% CI 1.56 to 3.26). A total of 234 (46%) of all consulters consulted within a year of the baseline survey. There was no difference in likelihood of reporting knee pain at follow-up between those who first consulted within a year of the baseline survey and those who consulted first more than a year afterwards (unadjusted odds ratio (OR) 0.70; 95% CI 0.38 to 1.28). In all, 229 (45%) of all consulters had their last consultation within 1 year of the follow-up survey, and they were at a significantly increased risk of reporting knee pain at follow-up compared with consulters whose last consultation was more than 1 year before the followup survey (unadjusted OR 3.26; 95% CI 1.58 to 6.72). However, those whose last consultation was more than a year before the follow-up survey were still more likely to report knee pain at follow-up than non-consulters (unadjusted OR 2.55; 95% CI 1.75 to 3.71).

Severity of knee pain In total, 689 persons had severe knee pain at baseline and 669 of these could be classified in terms of severity of knee pain at follow-up. A total of 297 (44%) consulted primary care between baseline and follow-up. Of the 669 respondents, 487 (72.8%; 95% CI 69.3% to 76.0%) reported severe pain at follow-up. Of those who consulted, 82% reported severe knee pain at followup compared to 65% of those who did not consult (unadjusted OR 2.44; 95% CI 1.70 to 3.52). Table 4 lists covariates individually associated with presence of severe knee pain or physical function at follow-up and hence included in the

propensity score modelling. The propensity score adjusted OR for association of consultation status with severe knee pain at follow-up was 1.93 (95% CI 1.27 to 2.93). A total of 151 (51%) of persons who reported severe knee pain at baseline and consulted, did so first within a year of the baseline survey. Those who first consulted within a year of baseline were no more likely to report severe knee pain at 3 years than those who consulted first more than a year after the baseline survey (unadjusted OR 1.20; 95% CI 0.66 to 2.17). In all, 136 (46%) of all consulters had their last consultation within 1 year of the follow-up survey. There was an increased but non-significant likelihood of reporting severe knee pain at follow-up in those who last consulted within a year of the follow-up survey compared to those who consulted last more than a year beforehand (unadjusted OR 1.65; 95% CI 0.89 to 3.05).

Consulting for any musculoskeletal consultation A total of 1045 (66%) persons with knee pain at baseline consulted either for a knee problem or for another musculoskeletal consultation during the 3-year period. In those with severe knee pain at baseline, 79% consulted for knee pain or another musculoskeletal problem. Consulting with any musculoskeletal problem did not significantly increase the likelihood of reporting knee pain (adjusted OR 1.31; 95% CI 0.92 to 1.87, table 5) at 3-year follow-up. However, there was an increased likelihood of reporting severe knee pain (adjusted OR 2.06; 95% CI 1.15 to 3.69, table 6) at 3-year follow-up compared to persons who did not consult for any musculoskeletal problem.

DISCUSSION This study has assessed the effect of consultation in primary care for knee pain on long-term reporting of knee pain. Those who consulted primary care appeared more likely to have continuing persistence of knee pain, and persistence of severe knee pain. A half of respondents reporting knee pain at

Table 6 Association of consultation in primary care with reporting of severe knee pain at 3 years in those reporting severe knee pain at baseline Severe knee pain at follow-up

No consultation for knee problem Consultation for knee problem No consultation for any musculoskeletal problem Consultation for any musculoskeletal problem

Yes, n (%)

No, n

OR* (95% CI)

OR{ (95% CI)

243 (65)

129

1.00

1.00

244 (82)

53

2.44 (1.70 to 3.52)

1.93 (1.27 to 2.93)

88 (62)

54

1.00

1.00

399 (76)

128

1.91 (1.29 to 2.83)

2.06 (1.15 to 3.69)

*Unadjusted; {adjusted for potential confounders using propensity score modelling. OR, odds ratio.

1706



Extended report Table 7 Covariate balance between consulters for knee pain and non-consulters in quintiles derived from propensity scores Outcome: knee pain at follow-up, p value

Outcome: severe knee pain at follow-up, p value

Covariate

Consultation

Consultation*quintile

Consultation

Consultation*quintile

Age group: 65–74(75+) General practice: B (C) Deprived area BMI: underweight (overweight) (obese) Most anxious or depressed Widespread pain Physical function Social function Body pain General health Previous knee injury Laterality Chronicity Use of non-GP services Severe WOMAC pain Pre-base knee consultation Non-knee musculoskeletal consultation Comorbidity: medium (high)

0.20 (0.74) 0.81 (0.33) – 0.41 (0.87) (0.66)

0.13 (0.60) 0.82 (0.26) – 0.35 (0.77) (0.58)

0.68 (0.38) 0.28 (0.79) 0.88 0.85 (0.63) (0.80)

0.84 (0.31) 0.30 (0.59) 0.82 0.78 (0.55) (0.67)

0.67 0.95 0.54 0.90 0.51 0.85 0.05 0.47 0.94 0.43 0.93 0.91 –

0.61 0.92 0.61 0.82 0.54 0.83 0.05 0.42 0.97 0.32 0.79 0.93 –

0.58 0.59 0.64 0.60 0.74 0.50 0.18 0.90 0.58 0.47 – 0.86 0.97

0.51 0.59 0.60 0.53 0.66 0.49 0.19 0.85 0.50 0.61 – 0.78 0.87

0.94 (0.90)

0.84 (0.96)

0.38 (0.60)

0.53 (0.54)

p>0.05 indicates satisfactory balance. BMI, body mass index; GP, general practicioner; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index.

baseline who did not consult for knee pain during the 3-year period did consult for another musculoskeletal problem. However, those who did not consult for any musculoskeletal problem were no more likely to have persistent knee pain than those who had consulted for a musculoskeletal problem. Further, those who did not consult for a musculoskeletal problem were at a decreased risk of persistent severe knee pain at 3-year follow-up. Non-consulters may have been seeking help elsewhere (for example, physiotherapy, acupuncture). However, we have previously identified low use of other services or private services in this population.6 In those who did not consult, 75% continued to have knee pain at 3 years, and two-thirds with severe knee pain at baseline reported severe knee pain at 3 years. Despite the levels of persistence found, our results suggest certain sections of older adults are effectively self-managing without need for recourse to healthcare professionals. An area of future research would be to identify the specific ways in which this group are successfully self-managing. Although we specifically examined severity of knee pain, it is possible that those who consulted for knee pain had greater problems with their knee than we were able to identify. However, the conclusion remains that the problem does not seem to have been resolved by consulting primary care. We have not examined the content of the consultation, nor the management of knee pain in respondents who consulted for their knee pain. Guidelines for management of knee osteoarthritis and knee pain do exist.25 26 However, it is not known how closely current primary care management of knee pain follows these guidelines. This was a large prospective cohort study of older adults living in the community. As with all longitudinal observational studies, there was some attrition but this was not linked to selfreported consultation. The associations found here are unlikely to be removed by a reduced loss to follow-up. The practices are part of the Keele GP Research Partnership and, as such, have high levels of quality of morbidity coding. We Ann Rheum Dis 2008;67:1702–1709. doi:10.1136/ard.2007.080259

have previously shown disparity between self-reported and recorded consultation for knee pain.27 This is partly due to GPs sometimes recording knee pain in the free text of a consultation rather than coding it specifically, particularly if substantial comorbidity exists, or because they are coding knee pain as part of a widespread pain syndrome such as multiple arthralgia or generalised osteoarthritis. It may be that GPs are more likely to code knee pain if it is a major problem, again suggesting that the poorer outcome of those recorded as having consulted may be due to some unmeasured marker of severity. GPs may also be less likely to record consultations for chronic conditions that are often consulted for. However, it is unlikely that someone consulting several times for knee pain would not get at least one record of it during our exposure time period of 3 years. In any case, we also searched the consultation text, and separately analysed musculoskeletal consultations. The KNEST question enquires about knee pain in the previous 12 months, and therefore the last recorded consultation could have been after the last occurrence of pain. Indeed, there were some differences by time of consultation. Those who consulted last within the year prior to follow-up were more likely to report knee pain as having occurred in that year than those who last consulted more than a year previously. However, the latter group were still more likely to report knee pain in the previous year at the follow-up survey than non-consulters. Additionally, analysis of a body pain manikin in the follow-up survey showed 80% of those consulting in the year before follow-up had indicated knee pain in the previous month (compared to 70% of those consulting more than a year prior to the follow-up survey). Further, severity of knee pain is based on the past 48 h so this would not be an issue for the analysis of severity. Clinical insight and statistical expertise are required when constructing an efficient propensity score model. Previous studies have demonstrated that those variables related to outcome should always be included in the propensity score model.28 The propensity score modelling balanced baseline 1707


Extended report differences on potential confounders between consulting and non-consulting groups across the propensity score quintiles. The set of potential confounder variables included was large but there may be other important unobserved covariates. These may include, for example, proximity to the surgery and belief as to whether the GP is likely to help. Further, our definition of comorbidity (based on number of recorded Read Code chapters) does not reflect comorbidity occurring within the same chapter. However it does reflect the wide range of coexisting morbidities with which patients present in primary care. There has been little previous work on the association of GP consultation with long-term outcome of knee pain. One study showed more deterioration of knee symptoms over an approximately 7-year follow-up period in those who reported consulting their GP for their knee problem.13 Our findings for knee pain reflect a similar pattern to that found in prospective studies of shoulder pain. Badcock et al29 reported that among all participants with persistent shoulder–neck pain at 2 years, consulters were 1.5 times more likely to have shoulder related disability at follow-up than non-consulters. In a hospital-based study of knee and shoulder referrals,30 referral was not associated with an improvement in pain or function at 1 year, and for shoulder problems referral was linked to worse outcomes. Knee pain and severe knee pain are persistent problems for older adults, regardless of whether they consult primary care. Future research could beneficially identify means of more effectively reducing the burden of knee pain in the community. Acknowledgements: We would like to thank the doctors and patients of the three health centres involved and also the administration team in Primary Care Sciences at Keele University who helped with the surveys. We would also like to thank Professor Peter Croft and Dr Mark Porcheret for comments on the draft of this paper.

REFERENCES 1.

2.

3. 4.

5.

6.

7. 8.

9.

10.

11.

12. 13.

1708

Jordan K, Clarke AM, Symmons DP, Fleming D, Porcheret M, Umesh K, et al. Measuring disease prevalence: a comparison of musculoskeletal disease using four general practice consultation databases. Br J Gen Pract 2007;57:7–14. Jinks C, Jordan K, Croft P. Osteoarthritis as a public health problem: the impact of developing knee pain on physical function in adults living in the community: (KNEST 3). Rheumatology (Oxford) 2007;46:877–81. McAlindon TE, Cooper C, Kirwan JR, Dieppe PA. Knee pain and disability in the community. Br J Rheumatol 1992;31:189–92. O’Reilly SC, Muir KR, Doherty M. Knee pain and disability in the Nottingham community: association with poor health status and psychological distress. Br J Rheumatol 1998;37:870–3. 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;60:91–7. Jinks C, Jordan K, Ong BN, Croft P. A brief screening tool for knee pain in primary care (KNEST). 2. Results from a survey in the general population aged 50 and over. Rheumatology (Oxford) 2004;43:55–61. Jordan K, Jinks C, Croft P. A prospective study of the consulting behaviour of older people with knee pain. Br J Gen Pract 2006;56:269–76. Anderson R, Newman JF. Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc 1973;51:95–124. Ballina GF, Hernandez MR, Martin LP, Fernandez SJ, Cueto EA. Epidemiology of musculoskeletal complaints and use of health services in Asturias, Spain. Scand J Rheumatol 1994;23:137–41. Jinks C, Ong BN, Richardson JC. A mixed methods study to investigate needs assessment for knee pain and disability. Population and individual perspectives. BMC Musculoskelet Disord 2007;8:59. Sanders C, Donovan J, Dieppe P. Unmet need for joint replacement: a qualitative investigation of barriers to treatment among individuals with sever pain and disability of the hip and knee. Rheumatology 2004;43:353–7. Verbrugge LM, Ascione FJ. Exploring the iceberg. Common symptoms and how people care for them. Med Care 1987;25:539–69. Peters TJ, Sanders C, Dieppe P, Donovan J. Factors associated with change in pain and disability over time: a community-based prospective

14. 15. 16. 17. 18. 19. 20.

21. 22. 23. 24. 25.

26.

27. 28. 29. 30. 31.

observational study of hip and knee osteoarthritis. Br J Gen Pract 2005;55:205–11. Mallen CD, Peat G, Thomas E, Lacey R, Croft PR. Predicting poor functional outcome in community-dwelling older adults with knee pain: prognostic value of generic indicators. Ann Rheum Dis 2007;66:1456–61. Jinks C, Lewis M, Ong BN, Croft P. A brief screening tool for knee pain in primary care. 1. Validity and reliability. Rheumatology (Oxford) 2001;40:528–36. Ware JE, Snow K, Kosinski M, Gandek B. SF36 health survey. Manual & interpretation guide. Boston, Massachusetts, USA: Health Institute, New England Medical Centre, 1993. Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale. Acta Psychiatr Scand 1983;67:361–70. Bellamy N. WOMAC osteoarthritis index. A users guide. London, Ontario, Canada: University of Western Ontario, 1996. Jinks C, Jordan K, Croft P. Measuring the population impact of knee pain and disability with the Western Ontario and McMaster Universities Osteoarthritis index (WOMAC), Pain 2002;100:55–64. Porcheret M, Hughes R, Evans D, Jordan K, Whitehurst T, Ogden H, et al. Data quality of general practice electronic health records: the impact of a program of assessments, feedback, and training. J Am Med Inform Assoc 2004;11:78–86. NHS Information Authority. The clinical terms version 3 (the Read Codes). Birmingham, UK: NHS Information Authority, 2000. Jeffery A, Jinks C, Jordan K. The influence of completing a health-related questionnaire on primary care consultation behaviour. BMC Health Serv Res 2006;6:101. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41–55. Office for National Statistics. National statistics socio-economic classification (NSSEC 2002). London, UK: Stationery Office, 2002. Jordan KM, Arden NK, Doherty M, Bannwarth B, Bijlsma JWJ, Dieppe P, et al. EULAR recommendations 2003: an evidence based approach to the management of knee osteoarthritis: report of a task force of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT). Ann Rheum Dis 2003;62:1145–55. Roddy E, Zhang W, Doherty M, Arden NK, Barlow J, Birrell F, et al. Evidence-based recommendations for the role of exercise in the management of osteoarthritis of the hip or knee – the MOVE consensus. Rheumatology (Oxford) 2005;44:67–73. Jordan K, Jinks C, Croft P. Health care utilization: measurement using primary care records and patient recall both showed bias, J Clin Epidemiol 2006;59:791–7. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Sturmer T. Variable selection for propensity score models. Am J Epidemiol 2006;163:1149–56. Badcock LJ, Lewis M, Hay EM, Croft PR. Consultation and the outcome of shoulder-neck pain: a cohort study in the population. J Rheumatol 2003;30:2694–9. Solomon DH, Bates DW, Schaffer JL, Horsky J, Burdick E, Katz JN. Referrals for musculoskeletal disorders: patterns, predictors, and outcomes. J Rheumatol 2001;28:2090–5. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study. Stat Med 2007;26:734–53.

APPENDIX Propensity score modelling There is debate as to whether the included covariates in propensity score modelling just need to be related to the outcome (ie, knee pain at follow-up), or should also be related to the exposure (ie, consultation).28 31 As, in our case, the set of covariates (in unadjusted analysis) related to knee pain were substantially the same as those related to knee pain and consultation, covariates related to outcome were included. Unadjusted associations of each potential confounding factor with knee pain status at follow-up were assessed using odds ratios for categorical variables and Student t tests for continuous variables. Factors that were significant at the 5% level or had odds ratios (for categorical variables) greater than 1.30 or less than 0.77 were included in the propensity score modelling. The propensity score model was then estimated by fitting a logit model with consultation status between baseline and follow-up as the dependent variable and all other factors included as the explanatory variables. This model was used to determine the propensity score (ie, the propensity to consult) for each respondent and these scores were used to group respondents into five groups base on the quintile scores.23 Covariate balance between consulters and nonconsulters within quintile groups was evaluated by using linear regression models for continuous variables and logistic regression for categorical variables, with each covariate as the dependent variable and the propensity score quintiles groups and consultation status, with an interaction term, as independent variables. Significance of the main effect of consultation status or the interaction term indicated imbalance for that covariate. Ann Rheum Dis 2008;67:1702–1709. doi:10.1136/ard.2007.080259


Extended report For the outcomes of any knee pain and severe knee pain, checking of covariate balance revealed no significant main effects of consultation or the interaction between consultation and propensity score quintile at the 5% level (table 7), suggesting that propensity scores have been well estimated with regards to balancing covariates in the quintiles. This suggests that every non-consulter has a consulter who is comparable with regards to having a similar estimated probability of consulting.

In the instance of consultation for any musculoskeletal problem for the case of severe knee pain at follow-up, the check for covariate balance revealed some imbalances that were satisfactorily improved by incorporating three further variables (interaction of general practice and deprived area; interaction of pre-baseline knee consultation and widespread pain; squares of social function SF36 score and body pain SF36 score).

Access the latest content chosen by our Editors BMJ Journals editors select an article from each issue to be made free online immediately on publication. Other material is free after 12 months to non-subscribers. Access the Editor’s Choice from the home page—or expand your horizons and see what the other BMJ Journals editors have chosen by following the links on any BMJ Journal home page.


1709


The influence of consulting primary care on knee pain in older people: a prospective cohort study M Blagojevic, C Jinks and K P Jordan Ann Rheum Dis 2008 67: 1702-1709 originally published online January 31, 2008

doi: 10.1136/ard.2007.080259

Updated information and services can be found at: http://ard.bmj.com/content/67/12/1702.full.html

These include:

References

This article cites 27 articles, 11 of which can be accessed free at: http://ard.bmj.com/content/67/12/1702.full.html#ref-list-1

Article cited in: http://ard.bmj.com/content/67/12/1702.full.html#related-urls

Email alerting service

Topic Collections

Receive free email alerts when new articles cite this article. Sign up in the box at the top right corner of the online article.

Articles on similar topics can be found in the following collections Pain (neurology) (30699 articles) Degenerative joint disease (9787 articles) Musculoskeletal syndromes (17359 articles) Epidemiology (5369 articles)

Notes

To request permissions go to: http://group.bmj.com/group/rights-licensing/permissions

To order reprints go to: http://journals.bmj.com/cgi/reprintform

To subscribe to BMJ go to: http://group.bmj.com/subscribe/