Comorbidity Cohort (2C) study: Cardiovascular disease severity and comorbid osteoarthritis in primary care

Our study will be carried out in ten general practices, from North Staffordshire, Stoke-on-Trent and Cheshire. These practices are part of a local research network, the Primary Care Musculoskeletal Research Consortium. These practices, supported by the Primary Care Research West Midlands North (PCR WMN) network, cover a wide range of socio-economic groups and includes practices that have actively participated in routine collection of clinical data using computer records for the purposes of epidemiological study. Clinical information relating to all morbidity and drug therapies is recorded using standard classifications of Read codes [39] and BNF (British National Formulary) respectively [40]. Ethics permission was sought and given by the Cheshire Research Ethics Committee (REC ref no: 09/H1017/40).

The cohort study is based on the recruitment of four main groups aged 40 years and over who have either a record or consultation for: (i) No CVD or OA (reference); (ii) CVD without OA; (iii) OA without CVD and (iv) CVD and OA (comorbid cohort). The sampling of these groups from the ten general practices was in a three-year period beginning November 2006 and ending January 2010.

CVD cohorts: Using the CVD registers and historical information, all adults aged 40 years and over who had a record (main Read code Chapter G: “Cardiovascular system diseases”) and on active registration at the end of the 3-year study time-period will be identified. Cohorts will then be based on the three CVD severity categories of hypertension (Read and daughter codes beginning with G20), ischaemic heart disease (Read and daughter codes beginning with G3.) and heart failure (Read and daughter codes beginning with G58 and heart failure codes related New York Heart Association (NYHA) classification), as these conditions are currently part of the national Quality Framework of clinical recording, a means by which General Practitioners (GP) are reimbursed as part of chronic disease management. The CVD groups will be organised into three exclusive severity groups, which means that allocation to a cohort group will be based on the most severe category e.g. if an individual had consulted for hypertension and heart failure, they would be in the heart failure cohort. Previous studies have indicated that the recording quality of such information is likely to be high [41, 42]. These three cohorts will be separated into two sub-groups of (i) patients who had not consulted for OA in the same time period (index CVD cohort) and (ii) patients who had also consulted for OA in the same time period (comorbid cohort). All those in the three CVD severity groups who had comorbid OA were sampled, as were those with heart failure or ischaemic heart disease but no OA. However, those with hypertension but without OA were randomly sampled (stratified by practice) given the high prevalence of this group (see Sample size).

There are 3 phases to the study data collection. The first phase (survey population) will use postal questionnaires to obtain self-reported health information at baseline, monthly interval and 12-month follow-up. The second phase (linkage population) will link the survey data to consultation data, for patients who will give written consent to access their clinical records from general practice in the baseline questionnaire. The third phase (denominator population) will construct an anonymised database for the whole cohort for the total five-year time-period. So the denominator cohort constitutes the whole sample from which the survey sample will be drawn. The denominator sample will provide the basis for a distinct cohort sample in its own right, as well as addressing selection comparisons between people who took part in the survey and those who did not, and people who gave consent to record review and who did not.

Overall, study participants will be invited to complete 12 questionnaires – baseline and 12-month follow up, and between these time-points, 10 monthly short-interval questionnaires (Figure 1). A pilot study in one practice on 500 patients was carried out to test the feasibility of sampling and testing of the questionnaire methods.

The identified sampled cohorts will be sent a baseline questionnaire (Q1), which will include generic and specific-health measures (Table 1). Measures have been identified on the basis of the primary focus on CVD and OA diseases and other specific measures selected on the basis of conceptual links between the disease and outcome of interest in the follow-up phase.

For the primary outcome of interest, we will use the Short-Form-12 (SF-12) (version 2) health survey as a generic measure of physical health, specifically the Physical Component Summary (PCS) score [43]. We will also include an assessment of physical activity [44] and Hospital Anxiety and Depression (HAD) questionnaire will be included as a measure of psychological status, that influences both CVD and OA [45].

The CVD-specific measures will include the: Seattle Angina Questionnaire [46], Kansas City Cardiomyopathy questionnaire [47], Rose angina questionnaire [48], and symptoms of palpitations [49] and dizziness [50].

The OA-specific measures will include the: Knee injury and Osteoarthritis Outcome Score (KOOS – physical function component) [51], Hip injury and Osteoarthritis Outcome Score (HOOS – physical function component) [52], pain scale, pain manikin [53], joint pain, and other links previously reported in chronic disease literature in relation to the symptoms of tiredness [54] and sleep [55].

Survey participants will also be requested for consent to the monthly short-interval questionnaires and for permission to the subsequent review of their clinical records. Participants will be “tagged” in their general practice registers as ‘2C’ study participants.

With baseline consent, participants will be sent 10 monthly short-interval questionnaires (4 pages), including the SF-12, pain measures and CVD symptom measures. At the end of the study period, all study participants will be sent a 12-month follow-up postal questionnaire (Q12) using the same measures as used in the baseline survey.

The linked consultation data covers a total five-year time-period, for 3 years before the baseline survey, 12 months of the survey, and the 12 months after the full follow-up survey. We will measure the clinical (GP or nurse) decision process associated with the survey cohort follow-up. Using the same study samples for both phases allows the linkage of the self-reported health status, morbidity and the decision process experienced by that individual patient in the 12-month period.

After the completion of the survey phase, the clinical record data for cohort groups will be downloaded with prior patient consent. Clinical decisions in consenting patients will be measured following the baseline survey on the basis of: (i) drug treatment changes (new, repeat, dose, type), (ii) investigations such as blood tests or X-rays, (iii) referrals (in-practice and external, investigation or second-opinion) or (iv) no change in any of these measures. Drug use will be measured on the basis of new index-related treatments (i.e. analgesia for OA and CVD-related therapies), or change in doses used for those on pre-existing drug treatments. Investigations will cover index-specific indications, cardiological (e.g. Electrocardiogram (ECG), Cholesterol) or rheumatological (e.g. X-Rays). Referrals will include those relevant to the index conditions (i.e. physiotherapists and allied therapists, rheumatologists, orthopaedics and cardiologists) and all other referrals. Other measurements for the cohorts will include all other comorbidity using the Kadam severity classification [56, 57], specific morbidities such as cerebrovascular disease and peripheral arterial disease, and co-drug therapies used.

Phase 1, survey population method is defined by the responders who take part in the study, phase 2, the survey-consultation method is subject to the consenters who will give permission for this link to occur. Therefore, phase 3 will focus on the total cohort denominator population and the anonymised clinical data within the medical records of each individual patient to allow methodological assessment of response biases in the cohort samples. Patients who explicitly stated to their general practice that they did not wish to take part in any research or share their clinical data are excluded from this phase.

For the total denominator population, invited to take part at baseline, anonymised clinical data with patient diagnostic information, prescribed drugs, blood tests and investigations, referrals and linked healthcare activity will be collected. This anonymised data archive covers a total five year time-period (between 2006 and 2011) from 3 years prior to the baseline survey to 12 months after the 12 month follow-up survey (Figure 2).

The pilot study was used to guide the sample size needed. The primary outcome of the study is the mean change in PCS score from baseline to 12-month follow-up. In order to compare the change in PCS score between the whole comorbid (OA and CVD) cohort and the OA index cohort we estimated that at least 394 patients responding at 12 months per cohort were needed (confidence level 0.05; power 80%) to detect a between cohort effect size of 0.2. Based on expected responses of 60% at baseline and 70% at 12 months, this required 939 patients to be invited with comorbid OA and CVD and 939 with OA alone. Assuming 75% consent to medical record review, this number would also give 423 people in each cohort responding at baseline and consenting to record review. However, to maximize numbers in the index CVD cohorts (hypertension, ischaemic heart disease and heart failure), OA index and comorbid groups and in order to allow comparison between the CVD severity comorbid cohorts with their related index CVD cohorts, we invited all eligible patients with these conditions to participate. The exceptions were the index hypertension cohorts and index OA cohorts where we randomly sampled patients stratified by practice due to the large size of these groups. Similarly, we also randomly sampled patients from the reference group without CVD or OA.

We will also determine cumulative onset of disability defined on the basis of the generic and disease-specific health measures. The rate of progression to severe disability in the comorbid groups will be estimated compared to the other groups using Cox regression. Attributable fractions for health status and clinical decisions (referrals and drug use), and other variables (age, gender, deprivation, obesity, consultation comorbidity) for onset or progression of disability will be determined. We will also assess time to change in clinical decisions.

We will compare consultation patterns and management for the anonymised denominator cohort to the survey-consultation cohort to assess generalisability of our findings and estimate the likely extent of response bias.

Initial data collection has already been completed and the following data is presented to follow the sequence of sampling stages in an epidemiological cohort study (Figure 2).

In the denominator cohort population, there were 9,676 people aged 40 years and over (Table 2) who were identified as baseline participants. The denominator study groups were as follows: 2,535 (26%) without CVD or OA; CVD index groups without OA – 1,322 (14%) with hypertension, 2,036 (21%) with ischaemic heart disease, 259 (3%) with heart failure; 1,317 (13%) in OA index group without CVD, and CVD groups with comorbid OA – 1,644 (17%) with hypertension, 490 (5%) with ischaemic heart disease, 73 with heart failure (1%).

Reference cohort: The reference cohort without CVD or OA (Table 3) was composed of the younger population, and around 69% were aged 59 years or younger. It had an equivalent number of men and women, and around 20% in the top or bottom tertiles of deprivation.

Index CVD cohorts without OA: The youngest age groups were in the hypertension cohort (30% aged 59 years and younger), and the oldest groups were in the heart failure cohort (69% aged 70 years and over) (Table 3). There were more men than women in both the IHD and heart failure cohorts, and a quarter of the heart failure cohort had the most deprived status.

Index OA cohort without CVD: Most of this cohort was aged between 50 and 80 years of age, and there were more women than men (Table 3). The age figures were in contrast to the index CVD cohorts who had a higher number in the age group 70 years and over, but the deprivation figures were similar.

Comorbid CVD & OA cohort: The comorbid groups were relatively older than the index groups and the age proportion for the group 70 years and over were: 57% in the hypertension comorbid group, 73% in the IHD comorbid group and 93% in the heart failure comorbid group (Table 3). There were more women than men with CVD and comorbid OA for all three groups, and IHD comorbid group had relatively fewer numbers in the top and bottom deprivation tertiles.

In the baseline survey population, there were 5,426 (56%) people aged 40 years and over who responded to the questionnaire. The lowest response was in the reference cohort (46%) and the highest was in the comorbid heart failure cohort (63%). The proportion of responders in the index OA cohort and the CVD comorbid cohorts were higher than the index CVD cohorts (Table 2).

The survey responders were broadly similar in terms of age, gender and deprivation characteristics, compared to the denominator population (Table 3 and 4). When comparing the responder and non-responder groups (Tables 4 and 5), non-responders were likely to be the younger groups, women and those people with the most deprived status. In the reference cohort, 74% of the non-responder group were 59 years or younger. The age proportions of the index hypertension and index OA cohort were similar, but the proportions of age groups 70 years and over were higher in index IHD (47%) and heart failure (61%) cohorts, and highest in the IHD (70%) and heart failure (93%) comorbid cohorts. Apart from the reference cohort and the index IHD cohort, there were more women than men non-responders, and around a quarter of the most deprived groups were non-responders in all study cohorts.

Of the survey responders, 4,030 (74%) gave permission to access and link their clinical records (Table 2), and this linked clinical data will be anonymised for the purposes of analyses. The lowest consent proportion was in the reference group (70%) and the highest in the CVD comorbid groups (76%).