Exploring Patient Perspectives of Statistically Derived Item Weights for Composite Scores in Spinocerebellar Ataxia
Author(s)
Chris N. Buckley, MSc1, Naomi Suminski, MS2, Rinchen Doma, BA, BS, MPH3.
1Clinical Outcomes Assessment Science, PAREXEL, London, United Kingdom, 2Clinical Outcomes Assessment Science, PAREXEL, Massachusetts, CA, USA, 3Clinical Outcomes Assessment Science, PAREXEL, Massachusetts, NC, USA.
1Clinical Outcomes Assessment Science, PAREXEL, London, United Kingdom, 2Clinical Outcomes Assessment Science, PAREXEL, Massachusetts, CA, USA, 3Clinical Outcomes Assessment Science, PAREXEL, Massachusetts, NC, USA.
OBJECTIVES: To compare patients’ perspectives with statistically derived item contribution of the Spinocerebellar Ataxia Composite Score (SCACOMS).
METHODS: SCACOMS was developed using partial least squares (PLS) regression, selecting items from three parent instruments that were most sensitive to SCA progression in natural history datasets. Item weights reflected their sensitivity to disease progression. To gather patient perspectives on SCACOMS items and weights, qualitative interviews were conducted with 24 SCA patients in the US. Participants reviewed items for relevance and importance, then assigned weights (percentages out of 100%) to each item based on its perceived importance in predicting SCA progression. Interactive pie charts representing SCACOMS items facilitated discussions. Participants adjusted pie segments to reflect their assigned weights, thinking aloud to explain their rationale. Subsequently, a pie chart showing PLS-derived weights was presented for feedback. Participants discussed discrepancies between their perspective and PLS-derived weights and rated their agreement with the latter. Post-interview analyses assessed the level of agreement between patient perspectives and PLS-based weightings.
RESULTS: All patients were able to understand, engage with, and provide responses to the exercise. Patient-perceived weightings closely aligned with PLS regression weightings for certain symptom items (Stance: 17% PLS-derived vs. 16% patient; Sitting: 8% PLS-derived vs. 13% patient) while larger differences were observed for other items (Gait: 12% PLS-derived vs. 33% patient; Speech: 10% PLS-derived vs. 25% patient; Clinician Global Impression-Improvement [CGI-I] 53% PLS-derived vs. 14% patient).
CONCLUSIONS: This approach integrating interactive visual aids to interviews, proved effective in eliciting patient-perceived item importance and comparison with statistically derived weights. Weighting differences for the CGI item indicated patients viewed clinician assessment as less important than symptom items for predicting SCA progression. This underscores potential misalignment between statistical methods and patient opinion, highlighting the importance of triangulating results. Doing so may allow for more patient-centric outcome measurement, enhancing research and clinical practice.
METHODS: SCACOMS was developed using partial least squares (PLS) regression, selecting items from three parent instruments that were most sensitive to SCA progression in natural history datasets. Item weights reflected their sensitivity to disease progression. To gather patient perspectives on SCACOMS items and weights, qualitative interviews were conducted with 24 SCA patients in the US. Participants reviewed items for relevance and importance, then assigned weights (percentages out of 100%) to each item based on its perceived importance in predicting SCA progression. Interactive pie charts representing SCACOMS items facilitated discussions. Participants adjusted pie segments to reflect their assigned weights, thinking aloud to explain their rationale. Subsequently, a pie chart showing PLS-derived weights was presented for feedback. Participants discussed discrepancies between their perspective and PLS-derived weights and rated their agreement with the latter. Post-interview analyses assessed the level of agreement between patient perspectives and PLS-based weightings.
RESULTS: All patients were able to understand, engage with, and provide responses to the exercise. Patient-perceived weightings closely aligned with PLS regression weightings for certain symptom items (Stance: 17% PLS-derived vs. 16% patient; Sitting: 8% PLS-derived vs. 13% patient) while larger differences were observed for other items (Gait: 12% PLS-derived vs. 33% patient; Speech: 10% PLS-derived vs. 25% patient; Clinician Global Impression-Improvement [CGI-I] 53% PLS-derived vs. 14% patient).
CONCLUSIONS: This approach integrating interactive visual aids to interviews, proved effective in eliciting patient-perceived item importance and comparison with statistically derived weights. Weighting differences for the CGI item indicated patients viewed clinician assessment as less important than symptom items for predicting SCA progression. This underscores potential misalignment between statistical methods and patient opinion, highlighting the importance of triangulating results. Doing so may allow for more patient-centric outcome measurement, enhancing research and clinical practice.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
CO9
Topic
Clinical Outcomes
Topic Subcategory
Clinical Outcomes Assessment
Disease
SDC: Neurological Disorders