Article Text

This article has a correction. Please see:

Download PDFPDF
42 Preference-sensitive multi-criterial decision support can avoid the over-testing produced by mono-criterial guidelines: the case of bone health
  1. Mette Kjer Kaltoft1,2,
  2. Jack Dowie1,2,3
  1. 1University of Southern Denmark, Odense, Denmark
  2. 2University of Sydney, Sydney, Australia
  3. 3London School of Hygiene and Tropical Medicine, London, UK


Background In preference-sensitive person-centred care any ‘over’ or ‘under’ – testing, detecting, diagnosing, or treating - has meaning only in relation to what is optimal for the individual, irrespective of what guidelines or clinical policies suggest. Guidelines typically involve the imposition of a value judgement-based action threshold on a single criterion, usually that perceived to be the main benefit. Given that multiple things matter to the person, the other criteria, especially the possible harms, need to be approached just as analytically in multi-criterial decision support, ideally provider-independent.

Method The decision support toolMinKnoglehelbredsbeslutning is built within value-based, compensatory Multi-Criteria Decision Analysis (MCDA) as implemented in the Annalisa template. It has three criteria (Avoiding Fracture, Avoiding Treatment Side Effects, and Avoiding Treatment Burden) and 18 options (11 medications, 6 lifestyle changes, plus ‘doing nothing new other than watchful waiting’). Implemented at the individual level, the performance Ratings for the Fracture criterion are the person’s personalised 10 year risks for Major Osteoporotic Fracture (MOF) and HIP fracture. These are obtained from Frax online and adjusted by the effectiveness of each option in fracture risk reduction, drawing on synthesised Network Meta-Analyses, published evidence, and expert opinion. The average Ratings for the Side Effects of each option were elicited from local experts via OPEN Redcap. The person/patient themselves enters their estimates of the Treatment Burden they anticipate experiencing from each option. Combining the Ratings and Weightings generates an expected value Score for each option. We take as exemplar cases 4 hypothetical women attending the COPD clinic at Odense University Hospital Svendborg (Odense University Hospital Svendborg): 60 and 70 year olds with BMI of either 28.6, the average in the study, or 20, selected as a low BMIto increase the justification for DXA. They are assigned two of the highest Frax risk factors (Parent Fractured Hip and Corticosteroid use) for the same reason. Their 10 year percentage fracture risks emerge as follows - 60yo (28.6): MOF 24, Hip 3; 60yo (20): MOF 28, Hip 6.6; 70yo (28.6): MOF 35, Hip 15; 70yo (20): MOF, 44 Hip 28. The average criterion importance weights from the 32 patients in the underlying study - Avoiding Fracture 43%, Avoiding Side Effects 35%, and Avoiding Treatment Burden 22% - were assigned to them, along with the average Treatment Burden ratings.

Result All four such women would be referred for a DXA under three conditions: (i) Current OUHS Guideline:Forced Expiratory Volume 1(FEV1) <50% with no Risk Factors; (ii) Current OUH Guideline FEV1 >50% with ≥1 Risk Factors; (iii) Frax risk ≥ 20% MOF or ≥ 3% HIP. (Under the UK NOGG: guidelines all four would actually be offered treated without a DXA.) In contrast, if they accepted the opinion of the PDST, all four would decline a DXA – and hence all medications which require a scan.

Conclusion A preference-sensitive multi-criteria decision support tool is often likely to produce an opinion out of line with guidelines and can help avoid over-testing and its consequences for the person concerned.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles