Impact of information and deliberation on the consistency of preferences for prioritization in health care – evidence from discrete choice experiments undertaken alongside citizens’ juries

Figures & data

Figure 1. Overview of research design.

Table 1. Summary of approach to addressing the research objectives and related hypotheses, and summary of findings (last column).

Research objective	Hypothesis	Data used	Analytic approach/evidence	Finding
Impact of information and deliberation on preference structure (that is, consistency of juror preferences before (pre) and after (post) the jury)	The choices jurors make will be less consistent between the pre and post survey than for controls	DCE of jurors and controls pre- and post- jury	Comparison of raw choice data responses between surveys pre and post, for jurors and controls. Comparison of rank order for preference weights for jurors and controls pre- and post-.	Hypothesis partially supported. The jurors’ preferences became more differentiated after the jury; a pattern that was not observed for controls. The juror’s preferences did change between the pre and post surveys but only in terms of a reversal of rank order for one attribute. This was also the case for controls, whose preferences changed in the opposite direction to the jurors.
Impact of information and deliberation on preference consistency of the juror preferences	The jurors will be more consistent in their choices after the jury than before the jury. The jurors will be more consistent in their choices after the jury than will controls.	DCE of jurors pre- and post- jury. DCE of jurors and controls post- jury.	Comparison of consistency to repeat choice task in raw choice data responses pre and post, for jurors. Comparison of scale of MNL model (which is inversely related to variance of the error term) for jurors pre- and post-. Comparison of scale of MNL model (which is inversely related to variance of the error term) for jurors and control post.	Hypothesis supported. Overall, this data suggests that the jurors became more convicted (less random) with their prioritization choices over time, and that this was not the case for controls who were not exposed to the jury. However, the finding that jurors were less consistent to the repeat choice task within survey post jury than controls is incongruent with this.
Consistency of the preferences of an uninformed public with the preferences of jurors after undertaking a CJ	Preferences of the uninformed general public as measured by a DCE will differ to those of an informed jury.	DCE in public and Jury recommendations (Qld and SA)	Descriptive comparison of rank order for preference weights for juror post- and public DCEs	Hypothesis supported. The juror preferences were similar to that of the general public before they undertook the jury, but they changed during the jury and differed from those of the public after the jury, leading to an increase in the priority that should be given to a person who already had obesity related conditions.
Alignment of the recommendations made by each CJ with the aggregated preferences of jurors after deliberation and with the preferences of the general public	Preferences of the jurors aggregated in a preference model will be consistent with the recommendations reached by the jury after deliberation. Preferences of the public will be less consistent with the recommendations reached by the jury after deliberation.	DCE in Jurors post-jury and Jury recommendations (Qld and SA)	Descriptive comparison of rank order for preference weights for jurors post- and for public DCE with recommendations of the juries (Qld and SA)	Hypothesis partially supported. Overall, the criteria considered most important are broadly consistent for both the juror and public DCE and juries but with some variation in the rank order of importance. There is slightly greater similarity to the jury recommendations for the DCE post jury than for the public DCE. However, there is also a similar level of variation in the importance of criteria between the two juries to that observed between the jury recommendations and the DCE.

Table 2. Characteristics of participants in the two jury, control and public samples.

	Jury Brisbane (n = 17)	Jury Adelaide (n = 11)	Control (n = 21)	Public (n = 1,994)
Female (%)	12 (70.5%)	8 (72.7%)	13 (61.9%)	1,038 (52.1%)
Age (mean, SD)	54.1 (15.2)	42.6 (14.7)	45.2 (16.1)	46.6 (16.5)
Age
18–34 yrs	2 (11.8%)	3 (27.3%)	5 (23.8%)	569 (28.5%)
35–54 yrs	4 (23.5%)	5 (45.5%)	10 (47.6%)	746 (37.4%)
55+ yrs	11 (64.7%)	3 (27.3%)	6 (28.6%)	679 (34.1%)
English not main language	2 (11.8%)	1 (9.1%)	4 (19.0%)	118 (5.9%)
Post-secondary education	12 (70.5%)	9 (81.8%)	16 (76.2%)	892 (44.7%)
BMI> =25.0	12 (70.5%)	10 (90.9%)	11 (52.4%)	1,171 (60.10%)
Previous bariatric surgery (self or family)	0^a	0^a	0^a	141 (7.1%)
Employed (full-or part-time)	10 (58.9%)	5 (45.5%)	13 (61.9%)	1,085 (54.4%)
Household annual income
<$42,000	5 (29.4%)	5 (45.5%)	5 (26.3%)	512 (29.8%)
$42,000–$100,000	9 (52.9%)	3 (27.3%)	7 (36.8%)	831 (48.4%)
>$100,001	3 (17.7%)	3 (27.3%)	7 (36.8%)	374 (21.8%)
Married/defacto	13 (76.5%)	6 (54.5%)	13 (61.9%)	1,288 (64.6%)

In instances of missing data, valid percentages are reported.

^a This was an exclusion criterion for the jury.

Figure 2. Priority weights for the juror and control samples, pre and post CJ deliberations (attributes are presented on the horizontal axis; possible priority scores range from 0 to 100 on the vertical axis).

Chance per % represents the priority score associated with a 1% increase in the chance of maintaining a substantial (at least half) reduction in excess weight.

Table 3. Rank order of priorities for juror and control DCE, pre and post.

	Rank
	Control	Control	Juror	Juror	Public
Attribute level	PRE	POST	PRE	POST	Baseline
Shown commitment to lifestyle change	1	1	1	1	1
Very severe obesity (BMI > 50kg/m^2)	3 n.s.	2	2	3	2
Already has obesity-related comorbidity	2	3	3	2	3
Severe obesity (BMI 40 to 50 kg/m^2)	4 n.s.	4 n.s.	4 n.s.	4 n.s.	4 (p < .1)
Family history of obesity	5 (rev*) n.s.	7 n.s.	5 (p < .1)	5 n.s.	5
1 month additional on wait list	6	5	6	6	6
1% additional chance of maintaining substantial weight loss	7	6	7	7	7
Age of recipient (20–50yrs)	n.s.	n.s.	n.s.	n.s.	n.s.

Parameters significant at the 5% level, except where indicated: p < .1 parameter significant at 10% level; n.s. Parameter not significant at the 10% level; Age is not ranked as it was not significant at the 10% level across all samples.

Rev* For control PRE the preference for this attribute was in the opposite direction, i.e. to prioritize those with no family history of obesity over those with a family history, but the coefficient was not significant at the 10% level (p = .39).

Figure 3. Comparison of the Jury recommendations and the DCE ranks for the jury post and public samples.

For the DCE, the attribute levels were significantly associated with the choices the respondents made at the 5% level (p < .05) with the exception of severe obesity and family history of obesity (which was only significant at the 10% level in some models, p < .1) and age of the recipient, which was not significantly associated with choice (n.s. p > .1).