Table 1:

Summary of Consensus Statements (by Issue and Application) ^a,b

		Meeting #1				Meeting #2
Issue		Application 1: Individual Patient Data		Application 2: Research Results Presented to Patients (i.e. Educational Materials and Decision Aids)		Application 3: Research Results Presented to Clinicians (i.e. PeerReviewed Publications)		Comments
Directionality of PRO Scores		There is no easy solution to the issue of directionality. There is a split in the “intuitive” interpretation of symptom scores, with some people expecting that higher scores would be “better” and others expecting that higher scores would be “more” of the symptom (and, therefore, worse).				The Consensus Panel acknowledges the challenges associated with directionality. There is a split in the “intuitive” interpretation of symptom scores, with some people expecting that higher scores would be “better” and others expecting that higher scores would be “more” of the symptom (and, therefore, worse).
		The Consensus Panel warned against trying to change current instruments – even if only how the data are displayed (e.g., “flipping the axes” where required for symptom scores so that lines going up are always better).				The Consensus Panel recommends against changing the scoring of current instruments.		Meeting #2 was more comfortable with the notion of changing the directionality in some situations in journal publications.
		PRO data presentation should avoid mixing score direction in a single display.				PRO data presentation should avoid mixing score direction in a single display. In cases where this is not possible, authors should consider changing the directionality in the display to be consistent.
		Mixed directionality between domains can cause confusion for both clinicians and patients. There is a need to address this potential confusion by using exceptionally clear labeling, titling, and other annotations.				There is a need for exceptionally clear labeling, titling, and other annotations.		Meeting #2 did not see the need to emphasize confusion.
Conveying Score Meaning		Descriptive labels (e.g., none/mild/moderate/severe) along the y-axis are helpful and should be used when data supporting their location on the scale are available.
		At a minimum, anchors for the extremes should be included (e.g., none/severe), as these labels also help with the interpretation of directionality. Labels for the middle categories (e.g., mild/moderate) should be included if evidence is available to support the relevant score ranges for each label.						Meeting #2 determined that this statement was redundant with the one above.
		In addition to the descriptive y-axis labels, reference values for comparison populations should be included if they are available.				In addition to the descriptive y-axis labels, reference values for comparison populations should be considered for inclusion if they are available.		The change during Meeting #2 was made to be consistent with the recommendation for displaying reference population norms below.
Score Representation		When presenting individual patient PRO scores, there is value in using consistent representation (i.e., line graphs, bar charts, etc.).						Because the display of research results is driven by the analytic strategy, both mean scores over time and proportions were addressed for research data
		Line graphs are the preferred approach for presenting individual patient PRO scores over time.
Conveying Possibly Concerning Results (Absolute Scores)		It is very important to show results that are possibly concerning in absolute terms, assuming the data to support a concerning range of results are available.
		The display of possibly concerning PRO results should be consistent with how possibly concerning results for other clinical data (e.g., lab tests) are displayed in the institution (comparison with other data in the electronic health record was uniquely considered for this issue).
Conveying Possibly Concerning Results (Change in Score)		Patients tend to value an indication of score worsening that is possibly concerning.
Normed Scoring				PRO data presentation needs to accommodate instruments the way they were developed, with or without normed scoring.
				One can decide if/when to show the norm visually (with a line on the graph), understanding that displaying it might provide additional interpretive value, but at the cost of greater complexity.		One can decide if/when to show the reference population norm visually (e.g., with a line on the graph), understanding that displaying it might provide additional interpretive value, but potentially at the cost of greater complexity.
				Comparison to the norm might be less relevant in the context where the primary focus is the choice between treatments.		Display of the norm might be less relevant in the context where the primary focus is the choice between treatments.
				If a norm is displayed: • It is necessary to describe the reference population and label the norm as clearly as possible (recommend “average” rather than “norm”) • It also requires deciding what reference population to show (to the extent that options are available). • It will need to be explained to patients that this normed population may not be applicable to a given patient.		If a norm is displayed: • It is necessary to describe the reference population and label the norm as clearly as possible (recommend “average” rather than “norm”). • It also requires deciding what reference population to show (to the extent that options are available).		The bolded text was seen as appropriate for patient educational materials/decision aids, but not for journal publications.
Clinically Important Differences				Patients may find information regarding clinically important differences between treatments to be confusing, but it is important for them to know what differences “matter” if they are going to make an informed decision.		Clinically important differences between treatments should be indicated with a symbol of some sort (described in a legend). The use of an asterisk is not recommended (as it is often used to indicate statistical significance).
						If there is no defined clinically important difference, that also needs to be in the legend and/or the text of the paper.
Conveying Statistical Significance						The data suggest that clinicians and others appreciate p-values; however, the Consensus Panel recognizes a move away from reporting them (and toward the use of confidence limits to illustrate statistical significance). Regardless of whether p-values are reported, confidence intervals should always be displayed.
Proportions Changed				Pie charts are the preferred format for displaying proportion meeting a responder definition (improved, stable, worsened), so long as the proportion is also indicated numerically.		Responder analysis results should be displayed visually.		The data supporting pie charts were stronger for presentation to patients than for presentation to clinicians/researchers
						Reasonable options include bar charts, pie charts, or stacked bar charts.

^aUnderlining indicates minor textual differences between statements across the different applications.

^bBolding indicates substantive differences between statements across the different applications.