To describe the development, validation and inter-rater reliability of an instrument to measure the quality of patient decision support technologies (decision aids).
Scale development study, involving construct, item and scale development, validation and reliability testing.
There has been increasing use of decision support technologies – adjuncts to the discussions clinicians have with patients about difficult decisions. A global interest in developing these interventions exists among both for-profit and not-for-profit organisations. It is therefore essential to have internationally accepted standards to assess the quality of their development, process, content, potential bias and method of field testing and evaluation.
Scale development study, involving construct, item and scale development, validation and reliability testing.
Twenty-five researcher-members of the International Patient Decision Aid Standards Collaboration worked together to develop the instrument (IPDASi). In the fourth Stage (reliability study), eight raters assessed thirty randomly selected decision support technologies.
IPDASi measures quality in 10 dimensions, using 47 items, and provides an overall quality score (scaled from 0 to 100) for each intervention. Overall IPDASi scores ranged from 33 to 82 across the decision support technologies sampled (n = 30), enabling discrimination. The inter-rater intraclass correlation for the overall quality score was 0.80. Correlations of dimension scores with the overall score were all positive (0.31 to 0.68). Cronbach's alpha values for the 8 raters ranged from 0.72 to 0.93. Cronbach's alphas based on the dimension means ranged from 0.50 to 0.81, indicating that the dimensions, although well correlated, measure different aspects of decision support technology quality. A short version (19 items) was also developed that had very similar mean scores to IPDASi and high correlation between short score and overall score 0.87 (CI 0.79 to 0.92).
This work demonstrates that IPDASi has the ability to assess the quality of decision support technologies. The existing IPDASi provides an assessment of the quality of a DST's components and will be used as a tool to provide formative advice to DSTs developers and summative assessments for those who want to compare their tools against an existing benchmark.
Citation: Elwyn G, O'Connor AM, Bennett C, Newcombe RG, Politi M, Durand M-A, et al. (2009) Assessing the Quality of Decision Support Technologies Using the International Patient Decision Aid Standards instrument (IPDASi). PLoS ONE 4(3): e4705. https://doi.org/10.1371/journal.pone.0004705
Editor: Kirby Lee, University of California San Francisco, United States of America
Received: July 10, 2008; Accepted: October 28, 2008; Published: March 4, 2009
Copyright: © 2009 Elwyn et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Canadian Institutes of Health Research Group grant; Cardiff University internal funding. No other external funding.
Competing interests: The authors have declared that no competing interests exist.
There has been increasing interest in the use of ‘decision aids’ , defined as adjuncts to the discussions clinicians have with patients during deliberations about decisions: these aids provide information about options and help clarify personal values . These adjuncts range from leaflets through face to face methods such as coaching or counselling to interactive multimedia websites. To describe this generic family of clinician-patient interventions we will use the term decision support technologies (DSTs) , corresponding with the internationally recognised need to assess the impact of ‘health technologies’ . DSTs are complex interventions which require detailed assessment to ensure safe use in healthcare contexts  because they help make options explicit, provide information about harms and benefits, clarify patient values' and provide structured means to help people deliberate when making decisions. Although there are published methods to assess the quality of clinical practice guidelines , DSTs go further and address issues of equipoise for which patients need to deliberate about difficult choices . However, as yet, there are no reliable methods to assure the quality of DSTs development process, content, potential bias, and method of field testing and evaluation – a gap which we address in this study. We did not intend to develop methods to assess how DSTs are used in practice, in the clinical encounter, although we recognise that this is an important area that requires further work.
There are reports that DSTs have achieved a ‘tipping point’ in the US and are widely accessed by increasing numbers of patients . The ability of DSTs to improve the quality of decisions and enable reductions in discretionary surgery and invasive procedures without adverse effects on health outcomes has been demonstrated in clinical trials , . The central role that these technologies will play in future healthcare systems is increasingly recognised , –. Over the last decade, the interest in developing DSTs has moved beyond research groups and has entered the commercial world. A global interest in developing DSTs has emerged among both for-profit and not-for-profit organisations. It is therefore essential to have a set of internationally accepted standards to assess their quality, to assess whether interests are declared and whether they are unduly biased , .
The International Patient Decision Aid Standards (IPDAS) Collaboration produced a checklist for the assessment of DSTs . The checklist was rigorously developed in a two stage web-based Delphi process using online rating process to enable international collaboration. A total of 122 individuals from four stakeholder groups (researchers, practitioners, patients, policy makers) representing 14 countries reviewed background evidence summaries, and rated the importance of 80 criteria in 12 quality dimensions. Second round participants received feedback from the first round and repeated their assessment of the 80 criteria plus three new ones. The IPDAS checklist enabled broad assessments in 12 dimensions: systematic development process, providing information about options; presentation of probabilities; clarification of values; use of patient stories; information about guiding or coaching; disclosure of interests; providing internet access; balanced presentation of options; use of plain language; use of up-to-date evidence; and effectiveness. The IPDAS checklist allows users, developers and others to assess whether these technologies contain the suggested components and judge whether they underwent rigorous development and evaluation. It has been used in updating the Cochrane systematic review of DSTs and to guide the development of DSTs , .
However, the checklist was not designed to provide precise, quantitative assessments, such that judgements could be made about the quality of DSTs, either at item, dimension or global levels. In addition, because not all checklist items were applicable to every DST, comparability, even at the checklist level, was not possible. Given interest in being able to assess these DSTs at a more precise level of detail — in terms of how they were developed and field tested, whether their content was valid and whether effectiveness had been evaluated with patients facing relevant decisions — the IPDAS Collaboration agreed that achieving this objective would require an instrument capable of quantitatively assessing the quality of DSTs. The aim of this article is to describe the development, validation and inter-rater reliability of an IPDAS instrument (IPDASi), built on the existing framework.
IPDASi was developed in four stages.
Stage 1 Refinement and preparation of instrument (IPDASi v1)
The published IPDAS checklist required transformation into a quantitative instrument, although we agreed to adopt the dimension-item framework. As part of this preparation, a group of researchers (GE, DS, RT, CB, SB, TW) used the existing checklist and dimension-item framework to score three purposefully selected DSTs, representing different design approaches and where our prior overall assessments indicated variable quality. These were Healthwise's Breast Cancer Surgery (BCS), web-based information, Bastian&McBrides Hormone Replacement Therapy (HRT), an illustrated booklet, and Wolf et al's Prostate Specific Antigen (PSA) screening, a brief text-based script. A binary (yes/no) and ‘not applicable’ scale was proposed; comments were collected on item applicability. Tabulations and qualitative analyses were performed but inter-rater correlations were not calculated.
Stage 2 IPDASi Confirmation of items (IPDASi v2)
On the basis of the results of Stage 1, a refined version IPDAS instrument (IPDASi v2) was designed and used in Stage 2. The non-applicable option was removed, and in this and all subsequent versions, a 4-point rating scale was used for each item, with possible responses as follows: strongly agree = score 4 (the issue is addressed clearly and comprehensively); agree = score 3 (the issue is addressed but with room for improvement); disagree = score 2 (the DST fails to clearly address the issue); strongly disagree = score 1 (the DST totally fails to address the issue). In common with the binary (yes/no) scale it replaced, the scale intentionally does not include a midpoint expressing neutrality. Items in the ‘balance’ dimension were integrated into the ‘information’ dimension. The web dimension was not applicable to all DSTs, therefore removed. A website was created for data collection (http://www.ipdasi.org/). Scale anchor point descriptions were developed for all items.
Five raters, two in the UK (MA-D and SS, Cardiff) and three in North America (ED and SK in Ottawa and MP in Providence) were familiarised with IPDASi v2, prior to using it to score the three previously selected DSTs, and asked to comment on item phrasing. Members of the IPDASi development group were asked to view the IPDASi instrument online and comment on item phrasing. For IPDASi v2 and subsequent versions, item scores were rescaled to be 0 to 100. At Stage 2, only an unweighted average of all items was calculated, as our focus was not on dimension scores. Analysis included inter-rater reliability using intraclass correlations for two way random effects at item and global score levels .
Stage 3 IPDASi Validation Study
Based on the results of Stage 2, a third version, IPDASi v3 was designed. This retained the majority of items from Stage 2, albeit with changes to phrasing. It comprised 47 items representing 10 dimensions. 9 dimensions applicable to all DSTs relate to Information (8 items); Probabilities (8 items); Values (4 items); Decision Guidance (2 items); Development (6 items); Evidence (5 items); Disclosure (2 items); Plain language (1 item); Evaluation (2 items). One additional dimension (9 items) relates to decisions based around tests or screening. Feedback from the comments resulted in more detailed anchor scale descriptions and standardization of descriptions.
IPDASi v3 was then used in a validation study to assess the quality of a sample of DSTs. Two approaches were used to achieve a sample of DSTs. First, five major producers of publically available DSTs were identified (The Foundation for Informed Medical Decision Making, Healthwise, Mayo Clinic, Midwives Information and Resource Service (MIDIRS) and Ottawa Health Decision Centre (OHDeC). Three DSTs from each producer were chosen at random, giving a total of 15. Second, 66 English-language DSTs, for which contact details were available, were chosen at random from the Cochrane inventory maintained by the University of Ottawa (http://decisionaid.ohri.ca/cochinvent.php), and their developers were approached and asked:
- Whether the DST was in current use and free of charge to clients;
- For consent to assess the DST using IPDASi; and
- For copies or information about documentation (published reports or peer reviewed articles) about the development or evaluation of the DST.
Each DST included in the sample was prepared for assessment in a standardised way. Background documents (relevant publications, reports) and all DST content were made available online (either in pdf or html formats; videos were converted into Windows Media Video format) for raters to assess. Table 1 provides details of the DSTs that were included in the sample, and the results of the IPDASi assessments.
Eight raters with diverse backgrounds and training were trained to undertake independent ratings: four in the UK (MA-D, MS, NJ, SS in Cardiff) and four in North America (SK, ED, AS in Ottawa; MP in Providence). Each DST was scored by two raters, one chosen randomly from each location, such that one rating was done in UK and the other in North America. New raters were asked to pilot the instrument on a ‘test’ DST and new raters also had access to raters who had completed the Stage 2 assessment if they required advice on item interpretation.
As in Stage 2, each item was scored on a 4-point scale, rescaled from 0 to 100, and dimension means were calculated. Two overall scores were calculated, scaled 0 to 100: the unweighted mean of all items (38 or 47, depending on whether the DST addressed a treatment or a test/screening decision) and the weighted mean score, a mean of the 9 or 10 dimension-specific means. The latter score upweights items belonging to dimensions comprising few items and downweights items from dimensions with many, but each dimension contributes an equal weight into the final score.
Summary statistics were calculated for dimension scores and unweighted and weighted overall means. Weighted means were modelled by rater and tool in a two-way balanced incomplete ANOVA model. Intraclass correlations and Cronbach's alpha, by each rater and by dimension means, were also calculated. The quality of each DST was then characterised by the average of the weighted mean scores from the two raters, adjusted by the model to take account of their personal propensity to give higher or lower scores. We wanted to predict the degree of accuracy if others used IPDASi in the future, considering one or two raters, known to us (i.e. one of the existing eight raters) or unknown to us. To achieve this, components of variation were determined by Bayesian modelling (Markov chain Monte Carlo) using WinBugs software , to arrive at estimated confidence interval half-widths for differing future rating situations. The raters' qualitative comments were summarised.
Stage 4 Agreement on IPDASi-SF (short form)
A core set of items was also chosen to develop a ‘short form’ (IPDASi-SF) aiming to test whether a ‘minimum’ quality threshold could be established. By agreement in the development group, these criteria were chosen based on having an equimedian score of 9 (i.e. maximum agreement) in the IPDAS consensus process . The equimedian is designed to represent the cumulative distribution function for a population with equal numbers in each of the four stakeholder groups . In addition, core-set items represented key concepts for each dimension. The 19 items selected for the IPDASi-SF consisted of 3 items for tests/screening and 16 others for all DSTs including: Information (4 items: options available, positive features, negative features, and fair comparison); Probabilities (3 items: reference class, event rates, compare probabilities); Values (1 item: personal importance); Development (3 items: patients' needs, impartial review, tested with patients) ; Disclosure (1 item: information about funding); Evaluation (2 items: knowledge, improved decision quality); Evidence (2 items: citations to studies, production date). The three items selected for the test/screening dimension included: next steps, chances of detection, non-symptomatic. These SF items were not highlighted for special attention during the rating process. Unweighted mean scores were calculated (i.e. all SF items and not the means related to their respective dimensions), and correlations (Pearson) with the IPDASi overall mean adjusted weighted score (Table 2).
Table 2 provides a synopsis of the different versions, detailed in the four stages.
Stage 1 Refinement and preparation of instrument (IPDAS v1)
Results of the seven raters were compared. The number of comments made at the interpretation level and the wide variation in scoring indicated a need for further item development. In addition some items had double criteria. In October 2006, five researchers met (AC, AOC, DS, CB&GE) and, using the results of this Stage, judged each item against two criteria, clarity and feasibility of measurement. All item phrasings were modified and it was decided to base the development of IPDASi on the following assumptions.
- All items should be applicable to the assessment of all DSTs. This enables the computation of a standard quality score per DST with no adjustment for specific content. An exception was made for DSTs designed to guide deliberations about undertaking diagnostic or screening tests. This type of DST would be subject to an additional dimension of items relating specifically to information on test characteristics.
- All items should meet the criterion of measurement feasibility. At this Stage, we decided to have 10 dimensions in IPDASi, mirroring the dimensions agreed in the IPDAS consensus process. Further information on dimension and items is presented in Stage 3.
Stage 2 Refinement and preparation of instrument (IPDAS v2)
Mean scores on a 0–100 scale for the three DSTs were as follows, with SDs reflecting inter-rater variation: HRT 68.7 (6.9); BCS 46.0 (6.5); PSA 38.5 (6.4). The intraclass correlation coefficient was 0.89. These results provided sufficient confidence to refine the instrument for a larger reliability study (Stage 3). Qualitative comments revealed where more specific item anchors descriptors were required, achieved collaboratively using a shared online spreadsheet. Discussions regarding dimension weighting led to agreement that the mean of each dimension should contribute equally to the total score.
Stage 3 Dual rater assessments of 30 DSTs (IPDAS v3)
Table 1 describes the sample of DSTs and provides the results. Table 3 lists the items used in IPDAS v3. Three DSTs were assessed from each of the five selected major producers. The other 15 were obtained by approaching 36 developers (representing 47 DSTs). Eighteen developers did not respond and we found that five of the DSTs were no longer in use. After repeated contacts, 13 developers (representing 15 DSTs) agreed to participate in the study, resulting in an overall sample of 30 DSTs.
The time taken to assess a DST varies considerably, dependent on its complexity. A simple DST comprising a leaflet could be completed in two hours but assessing multimedia web-based DST required at least 8 hours. A weighted overall score (scaled from 0 to 100) for each DST is shown, averaged over two raters, and then adjusted for the pair of raters. Adjusted IPDASi scores ranged widely from 33 to 82 (Table 2). The intraclass correlation for the weighted overall score was 0.80. Correlations of dimension scores with the weighted overall score were all positive (0.31 to 0.68). Cronbach's alpha values for the 8 raters ranged from 0.72 to 0.93. Cronbach's alphas based on the means in the 9 dimensions ranged from 0.50 to 0.81, indicating that the dimensions, although relatively well correlated, measure different aspects of DST quality. Calculations of the standard deviation (SD) presenting imprecision using a Bayesian model based on the existing eight raters, and projected for different number of known (one of the existing eight raters used) and unknown raters, for whom we have no information about their scoring tendencies, resulted in the following estimates: two known raters, 6.6; one known rater, 9.4; two unknown raters, 9.3; one unknown rater, 13.1. Qualitative comments were received on some items, requesting clarifications. This was achieved by adding examples and more descriptive elements to the anchor statements.
Stage 4 Agreement on IPDASi short form
The mean unweighted score for the short-form 16 item IPDASi was 56.1, similar to 56.3 for all items. The correlation of the unweighted IPDASi-SF to the overall mean weighted score (IPDASi score in Table 2) is 0.87 (CI 0.79–0.92). The ranking of the DSTs according to the SF version are very similar, with adjusted scores ranging from 34.5 to 83.1. DST number 32 still ranks highest, but the order shifts at the lower end of the scale. However, the aim of the IPDASi-SF was not to rank DSTs in order of quality but to determine whether or not a limited set of IPDASi items may be useful in determining minimal levels of quality.
This work demonstrates that IPDASi has the potential to assess the quality of DSTs. The four stage process revealed the need to make significant changes in the IPDAS checklist and modifications to the set of assumptions so that a measurement tool could be applied across the range of all possible DSTs. Having undertaken this work, we also suggest that IPDASi could provide formative feedback about dimensions in which DST developers could make improvements to subsequent versions. A short-form may also support the development of rapidly applicable quality standards. In addition, the study demonstrated the high correlation between IPDASi and IPDASi-SF, demonstrating support for the instrument's ability to provide correspondence between scores that indicate high quality at detailed dimension assessment and a version with focus on fewer items.
The study also displayed the levels of measurement imprecision when two raters assess each tool, and points to the need to ensure rater calibration and training in the use of IPDASi prior to assessment. We propose that IPDASi ratings should therefore be undertaken by raters who are familiar with DST development and use and who have undergone calibration training.
Strengths and weaknesses
The instrument design is based on prior international consensus which provided a framework in which to assess DST quality, and in addition, a set of criterion-based ‘items’ for a new instrument. Secondly, the work was planned by researchers who followed a detailed protocol and met regularly. Thirdly, a staged approach was used, adopting the principles of instrument development . Limitations of the study included the limited size of the sample and our focus on only DSTs developed in English, a constraint imposed by resource availability. There are also further opportunities to examine the validity of IPDASi, for example by examining whether low IPDASi scores for the ‘probability information’ dimension are associated with low patient knowledge about probabilities, when measured in controlled trials. Additionally, the raters used in the second and third stages were all researchers in the DST field and had some content expertise, so it is likely that raters with more diverse backgrounds may not perform as well. There was no opportunity in this study to provide intensive group training to all raters to ensure tight calibration and standardisation of item interpretation. To mitigate against this weakness, a detailed online manual that provided details about scale anchor definitions was available. Nonetheless, the results indicate that there is room to improve inter-rater reliability.
Results in context
Two other studies have used the IPDAS checklist. Coulter et al undertook a detailed assessment of 40 information materials to support people in making decisions about their health and health care . They found that the overall quality of information was poor and no systematic processes were adopted to give attention to presentational issues, such as readability or to ensure the validity of evidence. O'Connor et al used the checklist to assess the registered trials and found that several IPDAS process measures had not been used . Williams used IPDASi v2 to assess DSTs for genetic testing for breast cancer . We are not aware of any other work that has developed a quantitative measure of DST quality.
IPDASi, and IPDASi-SF, will be available as a quality assessment method to developers, researchers and purchasers, and given a recognised need to set standards and achieve benchmarks, will be subject to further development. The existing IPDASi provides an assessment of the quality of a DST's components, and in the absence of any other method, will be used as a tool to provide formative advice to DSTs developers and as a summative assessment for those who want to compare their tools against existing benchmarks (http://www.ipdasi.org). In due course, data from these assessments might form a platform for potential certification but questions remain. There is for instance only one dimension on evaluation outcomes. The items in this dimension cannot be scored unless the developers have actually conducted an evaluation. It is likely that developers may assert that not all DSTs require evaluation, provided they meet other requirements. However, we contend that research in this field is at an early stage. There is no agreement as yet on the essential ‘active’ components of DSTs ; moreover the theoretical underpinning for both their mode of action, measurement models and implementation strategies needs strengthening , . Further work is needed to assess which DSTs designs are superior to one another. Prospective studies that compare theoretically derived DSTs components and deliberation tools are required to help explore these areas.
The IPDAS collaboration and the resulting instruments (IPDASi and IPDASi-SF) need to meet the following challenges: How can new dimensions and items be considered? How are valid ‘option menus’ in DSTs derived and agreed when there are complex debates about equity, economics and evidence? Should there be items that assess the use of theory in the development of these methods, given that these are examples of ‘complex interventions’ and deserve attention to frameworks of design and mode of action . These challenges provide an agenda for future research.
What this paper adds
What is already known on this subject.
Interest in decision support technologies is rapidly increasing and they are being accessed by ever larger number of patients, especially in the United States.
A quality checklist for decision support technologies has been published by the International Patient Decision Aid Standards Collaboration.
The checklist was not designed to provide precise, quantitative assessments about the quality these interventions.
Conceived and designed the experiments: GE AO AE. Performed the experiments: GE AO CB MAD ED NJ SK AS SS MS. Analyzed the data: GE CB RGN MP MAD ED NJ SK AS SS MS. Contributed reagents/materials/analysis tools: GE. Wrote the paper: GE AO CB RGN MP MAD ED NJ SK AS SS MS SJB NC AC KE MH MHR NM DS RT TW TvdW AE.
- 1. O'Connor AM, Wennberg J, Legare F, Llewellyn-Thomas H, Moulton B, et al. (2007) Towards the ‘tipping point’: decision aids and informed patient choice. Health Affairs 26: 716–725.
- 2. O'Connor AM, Stacey D, Entwistle V, Llewellyn-Thomas H, Rovner D, et al. (2004) Decision aids for people facing health treatment or screening decisions (Cochrane Review) Issue 1. The Cochrane Library., Chichester, UK: John Wiley&Sons, Ltd.
- 3. O'Connor AM, Stacey D, Légaré F (2008) Coaching to support patients in making decisions. BMJ 336: 228–9.
- 4. Banta D (2003) The development of health technology assessment. Health Policy 63: 121–32.
- 5. Cluzeau FA, Burgers JS, Brouwers M, Grol R, Mäkelä M, et al. (2003) Development and validation of an international appraisal instrument for assessing the quality of clinical practice guidelines: the AGREE project. Quality and Safety in Health Care 12: 18–23.
- 6. Elwyn G, Edwards A, Kinnersley P, Grol R (2000) Shared decision making and the concept of equipoise: defining the competences of involving patients in healthcare choices. BJGP 50: 892–899.
- 7. Coulter A, Ellins J (2007) Effectiveness of strategies for informing, educating, and involving patients. BMJ 335: 24–27.
- 8. O'Connor AM, Llewellyn-Thomas HA, Flood AB (2004) Modifying unwarranted variations in health care: shared decision making using patient decision aids. pp. 63–72. Health Aff (Millwood), Suppl Web Exclusive:VAR.
- 9. Holmes-Rovner M, Nelson WL, Pignone M, R.D. Elwyn G, O'Connor AM, et al. (2007) Are patient decision aids the best way to improve clinical decision making? Report of the IPDAS Symposium. Med Decis Making (27): 599–608.
- 10. Godlee F (2007) BMJ 334: 1.
- 11. Elwyn G, O'Connor A, Stacey D, Volk R, Edwards A, et al. (2006) Developing a quality criteria framework for patient decision aids: online international Delphi consensus process. BMJ 333: 417–21.
- 12. Evans R, Elwyn G, Edwards A, Watson E, Austoker J, et al. (2007) Toward a model for field-testing patient decision-support technologies: a qualitative field-testing study. J Med Internet Res 9: e21.
- 13. O'Connor AM, Bennett C, Stacey D, Barry MJ, Col NF, et al. (2007) Do patient decision aids meet effectiveness criteria of the international patient decision aid standards collaboration? A systematic review and meta-analysis. Med Decis Making 27: 554–74.
- 14. Yaffee R (1998) Enhancement of reliability analysis: application of intraclass correlations with SPSS/Windows v.8. New York: New York University. Available from: URL: http://www.nyu.edu/its/socsci/docs/intracls.html.
- 15. Lunn DJ, Thomas A, Best N, Spiegelhalter D (2000) WinBUGS – a Bayesian modelling framework: concepts, structure, and extensibility. Statistics and Computing 10: 325–337.
- 16. Streiner N, Norman GR (1995) Health Measurement Scales: A practical guide to their development and use. 2nd ed. Oxford: Oxford University Press.
- 17. Coulter A, Ellins J, Swain D, Clarke A, Heron P, et al. (2006) Oxford: Picker Institute Europe. Assessing the quality of information to support people in making decisions about their health and healthcare http://www.pickereurope.org/Filestore/Publications/Health-information-quality-web-version-FINAL.pdf.
- 18. Williams L, Jones W, Elwyn G, Edwards A (2008) Interactive patient decision aids for women facing genetic testing for familial breast cancer: a systematic web and literature review. J Eval Clin Pract 14: 70–4.
- 19. Charles C, Gafni A, Whelan T, O'Brien MA (2005) Treatment decision aids: conceptual issues and future directions. Health Expect 8: 114–25.
- 20. Durand MA, Stiel M, Boivin J, Elwyn G (2008) Where is the theory? Evaluating the theoretical frameworks described in decision support technologies. Patient Educ Couns. Jan 30; [Epub ahead of print].
- 21. Gravel K, Legare F, Graham ID (2006) Barriers and facilitators to implementing shared decision-making in clinical practice: a systematic review of health professionals' perceptions. Implement Sci 1: 16.
- 22. Campbell M, Fitzpatrick R, Haines A, Kinmonth AL, Sandercock P, et al. (2000) Framework for design and evaluation of complex interventions to improve health. BMJ 321(7262): 694–696.