Using a ‘talk’ model of shared decision making to propose an observation-based measure: Observer OPTION^{5 Item}

Abstract

Objective

To propose a revised Observer OPTION measure of shared decision making.

Methods

We analyzed published models to identify the core components of a parsimonious conceptual framework of shared decision making. By using this framework, we developed a revised measure combining data from an observational study of clinical practice in Canada with our experience of using Observer OPTION^{12 Item}.

Results

Our conceptual framework for shared decision making composed of justifying deliberative work, followed by the steps of describing options, information exchange, preference elicitation, and preference integration. By excluding items in Observer OPTION^{12 Item} that were seldom observed or not aligned to a robust construct, we propose Observer OPTION^{5 Item}.

Conclusion

Although widely used, Observer OPTION^{12 Item} did not give sufficient attention to preference elicitation and integration, and included items that were not specific to a core construct of shared decision making. We attempted to remedy these shortcomings by proposing a shorter, more focused measure.

Practice implications

Observer OPTION^{5 Item} requires evaluation; we hope that it will be useful as both a research tool and as a formative measure of clinical practice.

Introduction

Measuring shared decision making is a challenge, despite reasonable consensus about its core characteristics [1]. To achieve shared decision making, providers should ensure that patients are informed about options, and supported to deliberate about those options. Both patients and providers should consider the role of evidence as well as each other's preferences as they deliberate. Ideally, end-users, that is, patients, would be asked to assess whether or not they experienced shared decision making. However, despite many efforts over the last decade to develop valid, reliable patient-reported measures of shared decision making, it has proved difficult to be confident that these measures are accurate assessments of the target phenomenon [2], [3].

The difficulty probably lies in the novelty of the construct. When asked whether they have been ‘involved in decisions’ patients tend to interpret the question as one about satisfaction, either with their provider or with their experience of care [3], [4]. The data from patient-reported measures of shared decision making typically provide scores at the upper end of scales, with relatively little variation [5], [6]. However, data from ratings of audio and videotapes, i.e. observational process assessments such as the Observer OPTION^{12 Item} measure [7], show that practitioners seldom achieve high levels of performance in tasks characteristic of shared decision making, a finding confirmed by Couët's recent review of 29 studies [8], and by studies using other observational measures [9], [10], [11], [12]. Over the last decade, further consensus has been achieved about the characteristics of shared decision making [1], [13]. We therefore considered it opportune to revisit the issue of observational measurement to evaluate whether improvements could be made to the existing Observer OPTION^{12 Item} [7].

There are a number of observational measures of shared decision making or efforts to provide ‘decision support’ to patients. These include Braddock's Informed Decision Making Scale [9], the Rochester Participatory Decision Making Scale [11] and Stacey's Decision Support Analysis Tool [12]. The Observer OPTION^{12 Item} measure is used to assess audio- or video-recordings of clinical encounters. Twelve items are scored and re-scaled to provide a score between 0 and 100, indicating “the extent to which providers involve patients in [shared] decision making” [7]. However, application of the measure over the last decade has revealed areas where improvement could be made. Observers, using Observer OPTION^{12 Item} to assess provider performance have noted the relative lack of attention to the elicitation of patient preferences, an important weakness of the measure [7].

Studies show significant variation in inter-rater agreement, particularly at item rather than global-score levels [8]. There are items that are not specific to the construct of involving patients in decision making, and some items specify behaviors that seem to be idealized constructions rarely observed in routine practice. For example, item 3 in the original instrument, see Table 1, specifies whether providers ask patients about their preferred information format, e.g. risk portrayal in graphs or number formats. Although plausible, this level of checking is unlikely in routine care. The inclusion of items that are never observed contributes to floor effects in the measure.

We appreciate that a shorter tool risks not being able to cover the specified construct of shared decision making fully, and might therefore be less valid and reliable, and this risk will need to be empirically tested in future studies. Our hypothesis is that a more focused, briefer measure would offer the following potential benefits:

• Improved construct validity, given a focus on a set of behaviors specific to shared decision making;

• Improved reliability because raters would be required to assess fewer, more relevant, and defined, observable behaviors;

• Increased efficiency because of shorter completion time.

The aim of this study therefore is to propose a shorter version of Observer OPTION^{12 Item} based on our analysis of published shared decision making models.