Making a picture worth a thousand numbers: recommendations for graphically displaying patient-reported outcomes data

Abstract

Purpose: Patient-reported outcomes (PROs) can promote patient-centered care in multiple ways: (1) using an individual patient’s PRO data to inform his/her management, (2) providing PRO results from comparative research studies in patient educational materials/decision aids, and (3) reporting PRO results from comparative research studies in peer-reviewed publications. Patients and clinicians endorse the value of PRO data; however, variations in how PRO measures are scored and scaled, and in how the data are reported, make interpretation challenging and limit their use in clinical practice. We conducted a modified Delphi process to develop stakeholder-engaged, evidence-based recommendations for PRO data display for the three above applications to promote understanding and use. Methods: The Consensus Panel included cancer survivors/caregivers, oncologists, PRO researchers, and application-specific end-users (e.g., electronic health record vendors, decision aid developers, journal editors). We reviewed the data display issues and their evidence base during pre-meeting webinars. We then surveyed participants’ initial perspectives, which informed discussions during an in-person meeting to develop consensus statements. These statements were ratified via a post-meeting survey. Results: Issues addressed by consensus statements relevant to both individual and research data applications were directionality (whether higher scores are better/worse) and conveying score meaning (e.g., none/mild/moderate/severe). Issues specific to individual patient data presentation included representation (bar charts vs. line graphs) and highlighting possibly concerning scores (absolute and change). Issues specific to research study results presentation included handling normed data, conveying statistically significant differences, illustrating clinically important differences, and displaying proportions improved/stable/worsened. Conclusions: The recommendations aim to optimize accurate and meaningful interpretation of PRO data.