Management of multimorbidity using a patient-centred care model: a pragmatic cluster-randomised trial of the 3D approach

Summary Background The management of people with multiple chronic conditions challenges health-care systems designed around single conditions. There is international consensus that care for multimorbidity should be patient-centred, focus on quality of life, and promote self-management towards agreed goals. However, there is little evidence about the effectiveness of this approach. Our hypothesis was that the patient-centred, so-called 3D approach (based on dimensions of health, depression, and drugs) for patients with multimorbidity would improve their health-related quality of life, which is the ultimate aim of the 3D intervention. Methods We did this pragmatic cluster-randomised trial in general practices in England and Scotland. Practices were randomly allocated to continue usual care (17 practices) or to provide 6-monthly comprehensive 3D reviews, incorporating patient-centred strategies that reflected international consensus on best care (16 practices). Randomisation was computer-generated, stratified by area, and minimised by practice deprivation and list size. Adults with three or more chronic conditions were recruited. The primary outcome was quality of life (assessed with EQ-5D-5L) after 15 months' follow-up. Participants were not masked to group assignment, but analysis of outcomes was blinded. We analysed the primary outcome in the intention-to-treat population, with missing data being multiply imputed. This trial is registered as an International Standard Randomised Controlled Trial, number ISRCTN06180958. Findings Between May 20, 2015, and Dec 31, 2015, we recruited 1546 patients from 33 practices and randomly assigned them to receive the intervention (n=797) or usual care (n=749). In our intention-to-treat analysis, there was no difference between trial groups in the primary outcome of quality of life (adjusted difference in mean EQ-5D-5L 0·00, 95% CI −0·02 to 0·02; p=0·93). 78 patients died, and the deaths were not considered as related to the intervention. Interpretation To our knowledge, this trial is the largest investigation of the international consensus about optimal management of multimorbidity. The 3D intervention did not improve patients' quality of life. Funding National Institute for Health Research.


Appendix: Training
Each member of clinical staff delivering the intervention attended two half days of training. These covered key concepts for the 3D approach. The training included case based discussion, demonstration of the interactive electronic template, reflection, evaluation and 'homework'.
The content included: • Overview of the 3D approach • What is patient centred care?
• Why is continuity of care important and how can it be improved?
• The pros and cons of co-ordinated reviews for all health conditions at once • Identifying the patient's main priorities and concerns • Screening for depression and the importance of mental health in multimorbidity • Polypharmacy and medication adherence • Goal setting and health care planning • Use of the interactive 3D template In addition a separate training meeting was held with the practice reception and administrative staff. This covered: • The importance of continuity of care and strategies for how to support this • Offering longer appointments when appropriate • The need to change recall systems, cancelling disease focused reviews for trial patients and inviting them to 3D reviews instead • How to run and export regular searches about progress with undertaking the 3D reviews Appendix a All analyses are multi-level linear regression models adjusted by centre, baseline EQ-5D-5L score, practice list size and deprivation score. Practice is included as a random effect. b Using multiple imputation by chain equations including baseline, 9 month, 15 month and EQ-5D-5L data as available, intervention arm, stratifying/minimisation variables and other covariates that were informative of missingness. c Standard error Appendix -0·05, 0·02 -0.02 (-0.07, 0.03) 0·40 a Numerator includes those who died, who were attributed an EQ-5D-5L score of zero b All analyses are multi-level linear regression models adjusted by centre, baseline EQ-5D-5L score, practice list size and deprivation score. Practice is included as a random effect. c Using participant postcode matched to England IMD data 2010 or Scotland SIMD data from 2012 d p-value from likelihood ratio test comparing model with interaction term against model without interaction term Appendix All analyses are adjusted by centre, baseline EQ-5D-5L score, GP practice list size and GP practice deprivation score. GP practice is included as a cluster effect in the estimation of the variance-covariance matrix. a Combining those in the none and partial compliance groups into the non-compliance group b Combining those in the partial and full compliance group into compliance group

Notes:
Compliance (at the patient level) was defined as 'full' if two GP 3D appointments and two nurse 3D appointments were attended over 15 months; 'partial' -at least one GP or nurse 3D appointment attended, but not full attendance; and 'none' -no GP 3D appointment and no nurse 3D appointment attended.
Using an instrumental variable regression model with randomised group as the instrument and an indicator variable for compliance, the CACE analysis was conducted in two ways: first combining the partial and none compliers into the non-compliance group and, second, combining those in the partial and full compliance group into the compliance group. Both analyses show that there is no evidence of a difference in effect in the intervention group compared with the usual care group. Although there appears to be a trend of greater effect of the intervention in those who had full attendance, there was no difference between trial arms after adjustment because greater attendance was associated with higher EQ-5D-5L at baseline.
The last search for the Cochrane review of interventions for multimorbidity was conducted in September 2015. 9 We conducted searches in Medline and the Cochrane library in August 2017 using a search strategy adapted from that in the Cochrane review to identify trials published since September 2015, and attempted to update the meta-analyses in respect of quality of life and the PACIC measure in the light of these more recent trials and the 3D trial. This updated review identified a further 11 studies. [10][11][12][13][14][15][16][17][18][19][20] and one previously identified study with more recent published data. 21 We have included the trial by Kennedy et al 22 in the above analyses because it was included in the Cochrane review, although is not described by the authors as an intervention for multimorbidity, and patients did not have to have multimorbidity to be included. Similarly several of the other trials are interventions for specific comorbid combinations of conditions, and not appropriate as a general approach to managing multimorbidity.
With respect to quality of life, the Cochrane review identified ten trials with relevant data. [22][23][24][25][26][27][28][29][30][31] These described studies which varied widely in terms of eligible population, setting and outcome measures. The Cochrane review authors were able to enter six of these studies into a meta-analysis, but did not report a pooled effect size due to substantial heterogeneity (I 2 =73%). In our updated review we identified seven further trials reporting quality of life. 11,[13][14][15][16]18,20,21 We have combined the results from the trials from the Cochrane review, the additional trials we identified and the results of the 3D trial and shown these in a Forest plot (see Appendix Figure 1). The data from the individual studies previously included in the Cochrane review are reported slightly differently in this figure from the data used in the original review because this figure is based on the generic inverse variance method which takes account of adjusted rather than unadjusted analyses of effect where these are available.
In extracting data for this analysis we chose any measure described by the authors as a measure of quality of life. Where studies used the SF36 we included data for the Physical Health summary score. Where studies reported data at multiple time-points we used the time-point closest to 12 months follow-up.
References for the studies within this figure are included in the bibliography as follows: Katon, 30 Martin, 21 Kennedy, 22 Boult 20 Barley, 27 Coventry, 29 Garvey, 28 Ekdahl, 18 Muth, 13 Koberlein-Neu, 15 Mercer, 14 Fortin, 16 Gonzalez Ortega, 11 Salisbury (refers to this paper).   Figure 1 provides further evidence of little or no benefit in terms of quality of life, in that the pooled effect estimate is very small and the confidence interval overlaps zero. The updated analysis also shows high levels of heterogeneity so the pooled effect should be treated with considerable caution. There is also the possibility of publication bias, since a funnel plot shows asymmetry with the largest trials showing no evidence of effect ( Figure 2).

Appendix Figure 2 Funnel plot
With regard to the PACIC measure, the Cochrane review identified two studies reporting this outcome. 29,32 We identified one more recent study. 12 The different studies reported data in different ways and were unsuitable for meta-analysis. However, all of the studies which have reported this outcome, including the 3D study, have confirmed that interventions to improve management of multimorbidity have a positive effect on patient-centred chronic care management as measured by PACIC.