Measuring Advance Care Planning: Optimizing the Advance Care Planning Engagement Survey (TH307C)

Abstract

•To describe a systematic process to reduce survey items on the validated Advance Care Planning Engagement Survey.•To describe a systematic process to validate progressively shorter versions of the Advance Care Planning Engagement Survey. A validated 82-item Advance Care Planning (ACP) Engagement Survey measures a broad range of behaviors (i.e., discussions and documentation). However, concise surveys are needed. To validate progressively shorter versions of the ACP Engagement Survey. The original Survey includes 57 Process Measures (i.e., knowledge, contemplation, self-efficacy, and readiness) and 25 Action Measures across 4 domains (surrogates, quality of life, flexibility, and asking questions). For item reduction (participants from 4 Canadian and 3 US sites), we systematically eliminated questions based on face validity, item non-response, redundancy, ceiling effects, and factor analysis. For construct validity (English- and Spanish-speaking participants from 3 ongoing US randomized ACP trials), we assessed internal consistency using Cronbach’s alpha, cross-sectional correlations, and the ability of the progressively shorter surveys to detect change one week after exposure to an ACP intervention using Pearson’s correlation coefficients. 501 participants were included in item reduction (age 69 years (±10), 41% non-white). Due to high correlations (mean 95% correlation, ±6) between readiness and action items, all action-items were removed. Due to high correlations and ceiling effects, 2 additional process-items were removed. Successive factor analysis then created 55, 34, 15, 9, and 4-item versions. 664 participants were included in the construct validity analysis (age 65 years (±8), 72% non-white, 34% Spanish-speaking). Cronbach’s alpha was high for all versions (4-item, 0.84 to 55-item, 0.97). Compared to the original survey, cross-sectional Pearson’s correlations were high (4-item, 0.85 to 55-item, 0.97) as were correlations detecting change (4-item, 0.68 to 55-item, 0.93). Correlations remained high for English and Spanish-speakers. Shorter versions of the ACP Engagement Survey are valid, internally consistent, and able to detect change across a broad range of ACP behaviors for English and Spanish speakers.