General Practitioner, Lecturer in General Practice
Jennifer Wrede MA MPH
Isabel Voigt MA
Research Fellow, Institute of General Practice, Hannover Medical School, Germany
Institute for Epidemiology, Social Medicine and Health Systems Research, Hannover Medical School, Germany
Institute of Biometrics, Hannover Medical School, Germany
Department of General Practice, University Medical Centre Go¨ttingen, Germany
Marie Luise Dierks
Institute for Epidemiology, Social Medicine and Health Systems Research, Hannover Medical School, Germany
Received date: 23 April 2012 Accepted date: 29 August 2012
BackgroundGeneral Practitioners (GPs) often have to simultaneously tackle multiple health problems of older patients. A patient-centred process that engages the patient in setting health priorities for treatment is needed. We investigated whether a structured priority-setting consultation reconciles the often-differing doctor–patient views on the importance of problems. DesignCluster randomised controlled trialwith 40 GPs and their 317 consecutively recruited older patients. Procedure Following a geriatric assessment, patients and doctors independently rated the importance of each uncovered problem. GPs then selected priorities with their patients in a consultation. Trained intervention GPs held a structured consultation and utilised the list of uncovered patient problems with their importance ratings to agree priorities. Untrained control GPs only used the patient’s problem list without importance ratings. Main outcome Doctor–patient agreement on independent importance ratings two weeks after the priority-setting consultation. Analysis Weighted kappa (_w) and multilevel logistic regression model. Results Intervention GPs and their patients determined mutual priorities for 20% of individual problems. In this process, GPs often succeeded in convincing their patients of the importance of vaccinations, lifestyle and cognitive issues. Likewise, patients convinced their GPs to prioritise their social and functional problems. Further treatment plans ensued in 84% of these priority areas. The regression model adjusting for clusters and baseline characteristics did not demonstrate significant differences in doctor–patient agreement on problems between groups a two weeks later. ConclusionDiffering views on the importance of health problems between GPs and older patients were not sustainably reconciled. The special consultation facilitated identification of priority problems for treatment despite differences in perceived importance of problems between patients and GPs. German clinical trials register drks 00000792.
family practice, health priorities, multimorbidity, older people, patient-centred care
In older patients with multimorbidity, many health problems need to be treated at the same time. In these instances, priority setting becomes necessary.
There is no tried and tested method on how to set priorities in agreement with older patients. Instead, priorities are traditionally determined in two ways. Either the patient sets priorities by deciding on the agenda of the consultation or the general practitioner (GP) weighs up what best to treat in the presence of multiple health problems.
These two practices both have disadvantages. If, on the one hand, GPs react solely to patients’ agendas at each contact, they may be tempted to add treatments and lose sight of previous problems and medications. If, on the other hand, GPs weigh up the eﬃciencies and dangers of treatments, especially in the case of poly-pharmacy, they run the risk of neglecting what is most important to the patient. Compatibility of medi-cations tends to take precedence over patients’ health preferences.
Recent changes in primary healthcare have not been able to solve the problem of how to treat patients with multimorbidity in a comprehensive and person-centred manner. Evidence-based guidelines and disease management programmes have become common practice. They focus on single diseases and are there-fore unable to officer solutions for the complexity of health.[4,5] Another recent development in European primary care is the emergence of geriatric assessment programmes, which provide a comprehensive health overview, but give no guidance on how to translate this complex health information into a holistic care plan.
Because of the lack of care strategies for people with multimorbidity, organisations, like the World Health Organization (WHO) and the European General Prac-tice Network, encourage research into person-centred and holistic care,[8,9] and this entails priority setting.
As a general rule, priorities are determined by decision-makers, who take competing interests into account. In the consultation, the decision-makers are the doctor and the patient. Doctors need to ensure that the preferences of patients are considered and that decisions on priorities are agreed upon. Although not easily achievable, this is recognised as an ethical necessity and a standard for the quality of care that patients experience.
We therefore developed a supplementary consul-tation in which, on the basis of a health overview (accomplished by a geriatric assessment), the patient and their GP determine the priority of problems for a holistic healthcare plan. This consultation had two features: person-centred communication and the set-ting of mutual health priorities to bring about reflec-tion or change in care plans. We tested this priority-setting consultation in a cluster randomised trial with the following objectives:
•to describe what was achieved in the intervention consultations in terms of priority setting and treat-ment planning,
•to determine the impact that a priority-setting consultation had on doctor–patient agreement of the importance of patient problems.
The study was a cluster randomised, general practice-based pragmatic clinical trial of a complex interven-tion. Randomisation was carried out in a two-level cluster design, in which the patients were nested within the GP surgeries. The Ethics Committee of the Hannover Medical School approved the study (No 5096, 2008).
In a phased approach, we identified, developed and refined the active components of this healthcare intervention. Table 1 gives an outline of the theoretical phase and preparatory studies. The concept of the intervention is an exchange of the patient’s and GP’s views on the importance of health problems (from an assessment) in order to jointly decide on the priority of problems for further attention or treatment. The main area of development lay in gaining a better understanding of what ‘importance’ means to doctors and patients when they consider health problems, and also its applicability as an outcome.
We contacted all GPs (N = 277) from selected postal districts in and around Hannover in four recruitment waves. In each practice, only one GP was to participate to avoid further clustering. Non-respondents were fol-lowed up by telephone. Positive responders received information material and a practice visit to explain the study aims and procedures. Participating doctors were allocated 1:1 into the intervention and control group using random block sizes of four.
Staffic in each practice approached up to 12 con-secutive patients aged 70 years and over registering at the reception desk in a predetermined recruitment week. Exclusion criteria were: a high level of physical dependence that ruled out independent living (care status II), limited mental capability, insuﬃcient language skills, current participation in another trial, or non-availability by telephone.
Doctors and patients were informed about the procedures for their own trial arm only.
All study patients received an assessment as an overview of their health problems. Patients and GPs independently rated every problem according to its perceived importance. The patient problems became the subject of the subsequent priority-setting consul-tations: trained intervention doctors obtained patients’ problem lists with importance ratings and followed a priority-setting protocol, whereas control doctors received the problem list without importance ratings. All doctors and patients once again rated the import-ance of each problem a fortnight later.
Our study nurses undertook a computer-aided assess-ment (STEP) with the patients in the practice during the week following recruitment. STEP was developed in a European Concerted Action to gain a compre-hensive view of older patients’ health issues. It consisted of 38 questions and eight basic examin-ations/laboratory results (e.g. fasting glucose, clock-drawing test). The 46 items were grouped into ten health domains: functional status, social environment, physi-cal symptoms, laboratory findings, mood, lifestyle, vaccination, problematic medication use, cognitive function and additional self-named problems not covered by the assessment (e.g. any type of cancer).
Baseline ratings (t0)
Directly after the assessment, patients received a list of uncovered health problems. They then rated the importance of each problem, which was in turn fed into the computer. The GPs independently received the patient’s problem list and rated the importance of each problem.
Immediately after these baseline ratings, the inter-vention or control consultation commenced.
Final ratings (t1)
Two weeks later, patients returned to the practice. Patients and GPs once more separately rated the importance of problems on the list, blinded to the previous baseline ratings. The follow-up period of 14 days was chosen to verify a sustained efficect on agree-ment between patients and doctors, if there was any.
The intervention consisted of an approximately 30-minute GP-training session, held on a one-to-one basis with a member of the research team, in prepar-ation for the priority-setting consultations. In particu-lar, GPs received training in the use of an individual patient-related PrefCheck guide (check patients’ pref-erences) for each consultation.
The PrefCheck guide was developed based on the findings of a pilot study with nine doctors and 35 patients (Table 1) and tested in consultations with eight patients and three GPs. PrefCheck is a three-part document that facilitates priority setting in a patient-centred way:
•Preparation phase – part I of the guide displayed a list of the individual patient problems uncovered by STEP, together with the patient’s importance ratings for each problem (‘important/not import-ant’). GPs could add any patient problems that they wanted to bring up in the consultation.
•Consultation phase – part II gave guidance on how to set priorities: (1) exploration of problems (go through the list of important patient problems and help patients to talk about their perspectives, give your perspective, bring in any additional problems important to you as a doctor); (2) negotiate and jointly determine priority problems; (3) if necessary, discuss and agree on further course of action.
•Documentation phase – part III involved GPs docu-menting which problems were discussed, which priorities set and any treatments planned for each problem. Treatments were grouped according to counselling, diagnostic procedures, treatments, referrals, coordination of care, watchful waiting.
Intervention GPs were given these patient-related PrefCheck guides immediately before the consul-tation.
The untrained control GPs were also asked to set health and treatment priorities. They only obtained the individual STEP problem lists for their consul-tations but were blinded to the patients’ views on the importance of their problems.
•Description of the intervention consultation in terms of determined priorities and planned treat-ments,
•Doctor–patient agreement on the importance of individual health problems. Our hypothesis was that 14 days after the consultation (at t1), GPs trained in person-centred priority setting would have greater agreement with their patients on the importance of problems compared with untrained GPs of the control arm, who had set priorities without training.
Ratings and measurement
Doctors and patients rated the importance of prob-lems on an ordinal four-point rating scale of 1, ‘not important’; 2, ‘slightly important’; 3. ‘quite import-ant’; and 4, ‘very important’. The importance ratings of doctors and patients were paired for each problem: ‘importance to patient–importance to doctor’s care’. These problem-related pairs were used for testing the agreement at t1 (the main outcome).
Agreement was measured by applying the weighted kappa ( Kw) to all paired ordinal importance ratings for each doctor–patient relationship. The mean of the resulting Kws over all patients was used to analyse a difficerence between the intervention and control group.Kw = 0 indicates no agreement among raters (other than would be expected by chance). Values between 0 and 1 range from slight (0.0–0.2), fair (0.21–0.4), moderate (0.41–0.6) and substantial (0.61–0.8) to almost perfect agreement (0.81–1).
Our sample size calculation was based on a moderate efficect size of Cohen’s d = 0.5 for the outcome ‘doctor–patient agreement’. We determined a mean kappa of 0.205 with a standard deviation (SD) of 0.35 in a pilot study. Therefore, a difficerence in the kappa mean of 0.175 was required. To demonstrate such a difficerence using a two-sided t-test with 80% power, an alpha of 0.05, on the basis of a patient cluster of 8 and an ICC of 0.05, 88 patients would be needed in 11 practices per group. Assuming a 20% loss of follow up, we would have to include 106 patients per group.
For the main outcome, we used the weighted kappa ( Kw) based on the ordinal four-point rating scale for all paired importance ratings per doctor–patient re-lationship. We tested significant difficerences between the intervention and control groups using a two-sample t-test for the mean Kws of each group (unadjusted analysis). A normal distribution ofK ws permitted its application.
To simplify the interpretation of importance ratings for intervention doctors during their consultations with study patients and also for the ease of result presentation, we transformed the ordinal importance ratings into dichotomous responses: ‘not important’ (ratings 1–2) and ‘important’ (ratings 3–4). Hence, apart from the analysis of weighted kappas, all other results on importance ratings are based on dichot-omous findings.
A multilevel logistic regression approach was ap-plied in addition to the kappa statistics in order to account for baseline difficerences and cluster efficects. The dependent variable ‘patient–doctor agreement on importance at t1’ was generated out of the paired dichotomous importance ratings for each problem. Patients and doctors entered as random efficects. Adjusting variables are listed in Table 3.
Descriptive analysis involved the prevalence of im-portant problems as assessed by doctors and patients at baseline and t1. To investigate to which extent the perceived importance of a problem was dependent on the type of problem, we grouped the dichotomous doctor and patient ratings into the 10 health domains provided by STEP.
Activities in the intervention consultations are presented as the frequency of problems that were discussed, prioritised and that led to treatment plan-ning. These activities were also related to the perceived importance of problems at t0. StatXact v. 8, Stata 12 and SPSS 18 were used.
Recruitment and characteristics of GPs
The study was carried out between July 2009 and July 2010. Participation in the first three recruitment waves included 18.6% (39/210) of all GPs; the fourth wave was stopped prematurely because only three more practices were required. Two GPs dropped out before the baseline survey so that 40 practices provided data (Figure 1). The characteristics of the 21 intervention GPs did not significantly differ from the 19 control GPs (Table 2). Data from the Regional Association of Statutory Health Insurance Physicians from 2010 indicated that our participating GPs were on average slightly younger than the average GP in this region (50.1 vs 53.8 years), and the percentage of female GPs in our sample was also higher (47.5 vs 34.3%).
Recruitment and baseline characteristics of patients
Three hundred and forty-seven patients initially agreed to participate (projected participation rate 57%; Figure 1). Twenty-five patients dropped out before baseline data collection due to the withdrawal of the two GPs, patients’ withdrawals of consent or interfering illness. Five further patients dropped out before any importance rating at t1 was completed (main outcome). This left 317 patients for analysis (Figure 1).20 Baseline characteristics are given in Table 2. Importantly, control patients had significantly more health prob-lems uncovered by the STEP assessment than inter-vention patients and were more worried about their health.
Baseline importance ratings of health problems and their doctor–patient agreement (t0)
Independent of randomisation, patients considered about half of their problems uncovered by STEP to be important, and GPs rated even more problems as important for their care. The group-specific analysis revealed that the importance ratings of control doc-tors significantly exceeded those of intervention doc-tors, although GPs had been allocated to groups at random (Table 1).
The type of problem had a bearing on what mattered to patients and to doctors. Figure 2 illus-trates the difficerent perspectives on the importance of problems for the health domains of STEP. Missed vaccinations, a failed clock-drawing test, medication-and lifestyle problems were frequently important to doctors’ care. Patients mostly considered self-named problems (problems not covered by the assessment and therefore added by the patient), social and func-tional everyday problems as more important (Figure 2).
Agreement was > 50% for all paired problem ratings on importance regardless of arm assignment. The weighted kappa indicated that, if chance agree-ments were disregarded, agreement accounted for just 6% of all rated problems (Table 1).
Importance ratings on health problems and their doctor–patient agreement (t1)
Two weeks after the initial consultation, intervention and control patients rated more problems as important
(58%) than at baseline (53%). GPs found about the same proportion of problems important for their care (62% compared with 63% at t0). The marked difficer-ence between intervention and control doctors observed in t0 prevailed at t1 (Table 3).
Agreement on the importance of health problems for the rating pairs ‘important to patient- to doctor’s healthcare’ amounted to 58% compared with baseline findings (53%), and surprisingly higher in the control compared to the intervention group (Table 3).
The multilevel logistic regression model, adjusting for clusters and baseline difficerences, demonstrated that assignment to the intervention or control group did not significantly predict doctor–patient agree-ment on health problem importance at t1 (Table 4).
What happened in the intervention consultation?
In order to avoid a Hawthorne efficect for the control GPs, only intervention GPs documented their actions for the patient problems in each consultation. They determined priorities with 70% of patients and planned at least one treatment for 90%.
Twenty per cent of all uncovered health problems were agreed as priority problems (369/1827). The health domains with the greatest proportion of pri-orities were ‘function’, ‘vaccination’ and ‘physical symptoms’ (Table 5). In one-fifth of all priority prob-lems (70/369), doctors had rated these as unimportant beforehand. Thus in the priority-setting consulta-tions, they accepted the patients’ difficering perspective. Similarly, patients accepted as priorities 20% of prob-lems which they had initially considered unimportant (72/369). Table 5 also shows those health domains with which patients and GP convinced the other party in the priority-setting process: patients emphasised social, self-named and functional problems, whereas GPs prioritised failed clock-drawing tests, vaccinations, lifestyle and medication issues. The top seven problems that patients and GPs managed to unilaterally identify as a priority are presented in Table 6.
Nearly all determined priorities received treatment planning (84% as opposed to 37% for non-priority problems). Again, 25% of problems that entered treat-ment planning were found important by GPs alone and 21% solely by patients. Priority problems received more active treatment planning (diagnostic testing, referral, start or change of treatment) than non-priority problems (watchful waiting) (Figure 3).
In our study, older patients and their GPs showed poor agreement on the importance of patients’ health problems. Our training programme for a structured consultation did not significantly improve agreement on the importance of health issues.
This study, however, highlighted that a structured patient-centred consultation provides the foundation for an exchange of views and negotiated care. Both doctors and patients altered their original views in 20% of cases during the process of jointly prioritising problems which in turn facilitated treatment plan-ning. For nearly all priority problems, treatment planning took place; this was in marked contrast to non-priority problems, suggesting that if doctors and patients discuss their views on the importance of problems and agree on priorities, unmet needs will be exposed and clarified.
Agreement: does it matter?
Our baseline analysis on doctor–patient agreement reveals that ‘normal’ consultations do not bridge the divide between patient and GP perspectives despite long-standing relationships. Patients tend to rank health problems that afficect their well-being and every-day life higher than GPs, who focus on medical problems and risk factors.[18,23]
It has been shown that doctor–patient concordance cannot be taken for granted. Divergent views are evident for disease relevance,[18,23] disease impact,[23,24] quality of life, exchange of information on diagnosis and prognosis, patient expectations[27,28] and treatment preferences. Doctor–patient agreement is an indicator of the adequacy and efficectiveness of information sharing. It lays the foundation for a trusting doctor– patient relationship, for better patient adherence and for improved health outcomes. Agreement is also an ethical requirement for shared decision-making and for deciding individual priorities.
Our intervention, involving a patient-centred con-sultation, did not lead to an improvement in doctor– patient agreement. Why this occurred is not obvious. On the one hand, it is possible that the role of the medical expert and that of the patient entail insolvable disagreement; On the other hand, shortcomings in our study design may have contributed to the only small overall improvement in agreement.
Limitations of our study
In contrast to the kappa statistic applied for measuring concordance of ratings of importance, importance ratings themselves are not established outcomes in healthcare research. We may have expected too much of the GPs by asking them to re-evaluate the importance of as many as, on average, 12 problems per patient after the consultation, although they could make use of their notes.
Our originally planned longer training course for intervention GPs was met with considerable resistance because of time concerns and so had to be altered to briefer face-to-face training. Also due to feasibility concerns, no extra session times for the priority-setting consultations were arranged. A busy atmosphere during normal sessions may have interfered with resolving difficerences in doctor–patient views even though priority-setting consultations were longer than average. The randomisation of doctors did not grant us evenly balanced importance ratings between inter-vention and control doctors at t0 – possibly due to the small number of clusters.
Finally, our participating GPs difficered in character-istics from the average GP practising in this area.
Implications for future research and practice
A health overview, as provided by the STEP assess-ment, provides an opportunity for doctors and par-ticularly patients to identify health topics that really matter. This type of doctor–patient communication takes the complexity of multimorbidity into account and helps decisions on priorities for management.
We investigated whether a patient-centred priority-setting approach was able to reconcile difficerences in views of patients and doctors on the importance of health problems. Our intervention was not successful in showing a sustained efficect on improved doctor– patient agreement for individual health problems. However, the practice of negotiating priorities increased awareness of problems previously only important to one party. Unmet needs were discovered, as reflected in the fact that treatments were planned for nearly all priority problems. The new awareness of what was important to the other party extended to the planning of treatments.
Our priority-setting approach is work-intensive. Further research should explore difficerent methods of priority setting. There is also a need to show whether a more sustained priority-setting approach impacts on long-term health and care experiences. A simplified iterative consultation with joint priority setting, re-flection and re-evaluation, may be the way forward.
We thank all general practitioners who participated in the study. We also thank Dr. Peter Walker (FRCGP) for revising the manuscript in English.
The trial was funded by the Federal Ministry of Education and Research (Project No 01GX0744). Participating GPs received a compensation of e90 per patient.
The Ethics Committee of the Hannover Medical School approved the study (No 5096, 2008).
Not commissioned; externally peer reviewed.