News Blog readers know that the NIHR Midlands Patient Safety Research Collaboration (PSRC) includes a programme of work on decision aids in maternity. In previous News Blogs we have discussed the role of educating staff in the use of Decision Aids – they cannot replace clinical consultation, and they do not exist in isolation from the services in which they are deployed.[1] Also in your News Blog we have discussed the information paradox – beyond a certain limit, providing more information clouds decision making.[2] This is a particular problem for maternity care because relevant information has to cover decision outcomes relevant to both mother and baby. The amount of information that results in overload – the point where more information clouds rather than elucidates decisions – will vary from one person to the next. People with low health literacy are particularly susceptible to ‘overload’ because they have a lower starting level of knowledge and, arguably, because they do not have an established mental framework that allows new knowledge to be easily assimilated. Our Midlands PSRC is planning a work-package specifically examining how this problem of information overload might be addressed. It is important in planning this work to consider what is already known regarding decision aids, since some of this work is relevant to the problem posed by the information paradox. For this reason, it is worth summarising findings from research over the last 50 years. Much of this information has been described in a series of articles developed with support from the International Patient Decision Aids Standards (IPDAS) Collaboration.[3-5] Here we offer a synopsis.
Probabilities
Probabilities are central to the concept of informed decision making. Following the iconic work of researchers such as Kahneman & Tversky[6] and the great Gigerenzer [7] we know a great deal on how to present probabilities (and how not to). The state of current knowledge is beautifully summarised in a review by David Spiegelhalter.[8] Probabilities should be expressed as percentages (like 5%) or perhaps simple frequencies (like ‘one in twenty’), but not numbers needed to treat or harm (which are confusing).
Visual formats, such as icon panels, are effective in conveying numerical information, especially when explaining the accuracy of diagnostic tests. Icons are also helpful when disclosing very low probabilities like 1 in 2,000. If graphs are used, they should cover 0 to 1 (or 0% to 100%) on the Y axis to avoid risks being over-estimated. It is also worth noting that asking people which format they prefer could be a disservice, because the preferred format is often not the format that maximises understanding. [9] Obviously, presentation should be ‘fair’, using the same denominator when multiple probabilities must be conveyed to cover salient outcomes. This is a particular problem in maternity care, since these outcomes affect both the mother and the baby. The study of how numeracy skills of service users (and service providers) can be considered in decision aids is in its infancy [10] – a point to which we will return. The question of how, or whether, to convey uncertainty in the probabilistic estimates themselves is unsettled. Stochastic uncertainty can be expressed in numbers that equate to 95% credible intervals (for example 5% plus or minus 2%). Even more difficult, is how to express uncertainty that does not arise from imprecision but that arises from inaccuracy when experimental studies have flaws or when we have to rely on non-experimental studies. [11]
Presentation of Information
As stated, icons should be used to present probabilities, particularly small probabilities, such as 1 in 2,000. As stated, the denominator must be defined and the same denominator used in the calculation of all probabilities presented to the decision maker. Further to this, there is good evidence that side-by-side information is better assimilated and appreciated than segmented formats. It goes without saying that aids and scripts should do all they can to avoid heuristic biases where the form of presentations influences the decision taken. The aim (even if it can be realised only imperfectly) is for non-directive counselling. For example, presenting probabilities in both negative and positive formats; 2% of people die after their operation and 98% survive the operation.
Value Clarification
In a previous News Blog, I drew attention to the work of Pauker and Pauker [12] who first elicited people’s values and then used those values (utilities) to calculate expected utilities, which then guided further consultation. The idea is that a clinician should explore the extent to which one outcome is traded-off against another to elicit a numerical value of preference/utility/value. For example, if I would run a 10% risk of death to avoid being rendered infertile, then my utility for infertility is 0.9 on a scale between fertile life (1.0) and death (0.0). This is a standard gamble, and it is but one method to elicit a patient’s preference. This topic has been reviewed recently by Witteman et al.[13] Rather than start with elicitation and utilities, as Pauker did, most seek to get a sense of a person’s preferences as the consultation proceeds. These range from a simple scale of preference across outcomes (A worse than B worse than C) to methods such as standard gamble or time trade-off. There is no clear sense from the literature as to which method is preferred or, indeed, what effect explicit preference measures have on understanding, knowledge or satisfaction. On rare occasions, when still in practice, I would introduce a simple standard gamble at some point in the decision process. Patients seemed to appreciate the clarity the method afforded, and it seemed to help their understanding (see Box). However, I suspect it is not a method that would suit everyone.
Box:
A patient was referred to my obstetric clinic because she wanted a second opinion. She had a condition called kyphoscoliosis where her upper spine was severely bent. Such an anomaly can put pressure on the lungs and hence the right side of the heart. Since the volume of blood passing through the heart increases dramatically (~40%) in pregnancy, a patient with kyphoscoliosis can come to harm. My patient’s general physician had told her “that she should not contemplate pregnancy”. Having formed a rapport with her, we agreed that the nub of the problem concerned her risk of dying as a result of pregnancy. I then asked her to think of the risk that she would be prepared to run to have a child. I asked her to return in a few weeks, during which I asked her physician what he thought the risk of her dying from pregnancy might be. “As much as 2%”, he said. When she came back, I told her this news. Her trade-off figure had been 10%. She was overjoyed and a year or so later I delivered her baby safely.
Breaking Up the Decision Problem – Incremental Approach
In the News Blog concerning the information paradox,[2] I mentioned certain processes to mitigate the problem. Richard Martin and colleagues [14] suggest, first, that information should be provided in aliquots – do not just dump all the information in its entirety. Also, they advise that the pace of information delivery (and perhaps even the amount of information delivered) should be adjusted according to the patient’s ability to assimilate the material. There are risks here, that the clinician will not get it right – judgement is called for. But herein lies the art of practice, making judgments about how much information can be delivered and at what pace. It is here that I think clinicians and members of the public can fruitfully work together to improve practice and where drama may have a role in improving flexible communication skills. This is an area where we plan further research.
People with Low Health Literacy
In a review of this topic, Muscat, et al.[15] considered mainly use of language to accommodate those with low reading age and involving consumers/public in the design of the decision aid. I could find little evidence on methods to prevent information overload specifically in people with low health literacy. However, some of the general methods mentioned above may help: tailoring the amount of information delivered at any one time; providing more time/sessions to support the decision maker; and involving family/friends in the consultation. It seems right to ask the person (as you proceed) whether they want more information and providing plenty of space and opportunity for them to express opinions and ask questions. Where time is a constraint, decision aids delivered on the computer, can be recommended for use between sessions. The use of computers to capture the medical history can release clinical time for more complex consultation tasks. [16]
Personal Stories
Some people argue that providing personal stories (testimonials) can augment the more abstract presentation of data thereby clarifying issues for the decision maker. This topic is reviewed by Shaffer, et al.[17] At first assessment, providing accounts of other people’s ‘lived experience’ might seem helpful. However, including narratives of individual choices in decision aids is rightly contentious. They improve knowledge in some studies (compared to aids with no narrative). However, they can result in slanted recall. The literature shows very varied responses in terms of how they affect decisions taken, decisional conflict and knowledge acquired. My concern arises from the fact that they can be very persuasive. It follows that unless a narrative pointing one way can be accurately titrated against a narrative pointing in the opposite direction, then they are directive and hence must undermine the very reason for use of a clinical decision aid. Thus, when we are trying to persuade people to live healthy lives (stop smoking) use narratives all you want. However, be circumspect in the use of narrative when you are offering choice – planned home versus facility births, for example. Stories can easily degrade non-directive counselling into persuasion.
Overarching Principle
It goes without saying that presenting choice cannot be hurried or done in a mechanical, dispassionate way. The clinician must first earn the patient’s trust and display empathy throughout – a point that comes out very strongly in the Spiegelhalter review cited above. Chat to your patient about their immediate family. Ask them how they are feeling. Being able to express one’s emotion is not only psychotropic – it also helps clear the mind for the decision task at hand. The neurophysiology shows that a person needs emotion – without it you just cannot make a decision. However, too much emotion and the brain becomes overwhelmed and reasoning is crowded out. Indeed, supporting non-directive decision making blends art and science.
In conclusion I hope you found this romp through decision aid design interesting. In this article I concentrated mostly on the decision aid artefact. In a forthcoming blog, I shall discuss the implementation of decision aids given that, as stated, they do not exist in isolation of the service itself.
— Richard Lilford, NIHR ARC West Midlands Director; NIHR Midlands PSRC Co-Director
References:
- Lilford RJ. Informing and Facilitating Choice in Maternity Care: What Do We Know & Where Are the Research Gaps? NIHR ARC West Midlands News Blog. 30 June 2023; 5(6):3-6.
- Lilford RJ. The Information Paradox at the Heart of Non-Directive Counselling. NIHR ARC West Midlands News Blog. 20 September 2024; 6(4).
- Stacey D & Volk R (for the Evidence Update Leads). The International Patient Decision Aid Standards (IPDAS) Collaboration: Evidence Update 2.0. Med Dec Mak. 2021;41(7):729-33.
- Trevena L. Commentary on History of IPDAS. Med Dec Mak. 2021;41(7):734-5.
- Witteman HO, et al. Systematic Development of Patient Decision Aids: An Update from the IPDAS Collaboration. Med Dec Mak. 2021;41(7):736-54.
- Tversky A & Kahneman D. Judgement under Uncertainty: Heuristics and Biases. Science 1974; 185: 1124-31.
- Gigerenzer G. How to Make Cognitive Illusions Disappear: Beyond “Heuristics and Biases”. Eur Rev Soc Psychol. 1991;2(1):83-115.
- Spiegelhalter D. Risk and Uncertainty Communication. Annual Rev Stat Appl. 2017; 4:31-60.
- Trevena LJ, et al. Presenting quantitative information about decision outcomes: a risk communication primer for patient decision aid developers. BMC Med Inform Decis Mak. 2013;13(s2):S7.
- Bonner C, et al. Current Best Practice for Presenting Probabilities in Patient Decision Aids: Fundamental Principles. Med Dec Mak. 2021; 41(7): 821-33.
- Dowswell T, et al. Should there be a trial of home versus hospital delivery in the United Kingdom? BMJ. 1996; 312(7033):753-7.
- Pauker SP & Pauker SG. Prenatal diagnosis: a directive approach to genetic counselling using decision analysis. Yale J Biol Med. 1977; 50(3): 275-89.
- Witteman HO, Ndjaboue R, Vaisson G, Chipenda S, et al. Clarifying Values: An Updated and Expanded Systematic Review and Meta-Analysis. Med Dec Mak. 2021;41(7):801-820.
- Martin RW, et al. Providing Balanced Information about Options in Patient Decision Aids: An Update from the International Patient Decision Aid Standards. Med Dec Mak 2021;41(7):780-800.
- Muscat DM, et al. Addressing Health Literacy in Patient Decision Aids: An Update from the International Patient Decision Aid Standards. Med Dec Mak. 2021;41(7):848-69.
- Lilford RJ & Chard T. Microcomputers in antenatal care: a feasibility study on the booking interview. BMJ. 1981;283(6290):533-536.
- Shaffer VA, et al. Do Personal Stories Make Patient Decision Aids More Effective? An Update from the International Patient Decision Aid Standards. Med Dec Mak. 2021;41(7):897-906.