Elodie de Bock1, Kevin Dolgin2, Benoit Arnould1, Guillaume Hubert2, Aaron Lee3, John D Piette4. 1. Patient Centred Outcomes, ICON plc, Lyon, France. 2. Observia, Paris, France. 3. Department of Psychology, University of Mississippi, Oxford, MS, USA. 4. Department of Health Behavior and Health Education, University of Michigan School of Public Health, Ann Arbor, MI, USA.
Abstract
OBJECTIVE: The SPUR (Social, Psychological, Usage, and Rational) Adherence Profiling Tool is a recently developed adaptive instrument for measuring key patient-level risk factors for adherence problems. This study describes the SPUR questionnaire's psychometric refinement and evaluation. METHODS: Data were collected through an online survey among individuals with type 2 diabetes in the United States. 501 participants completed multiple questionnaires, including SPUR and several validated adherence measures. A Partial Credit Model (PCM) analysis was performed to evaluate the structure of the SPUR tool and verify the assumption of a single underlying latent variable reflecting adherence. Partial least-squares discriminant analyses (PLS-DA) were conducted to identify which hierarchically-defined items within each dimension needed to be answered by a given patient. Lastly, correlations were calculated between the latent trait of SPUR adherence and other patient-reported adherence measures. RESULTS: Of the 45 candidate SPUR items, 39 proved to fit well to the PCM confirming that SPUR responses reflected one underlying latent trait hypothesized as non-adherence. Correlations between the latent trait of the SPUR tool and other adherence measures were positive, statistically significant, and ranged from 0.32 to 0.48 (p-values < .0001). The person-item map showed that the items reflected well the range of adherence behaviors from perfect adherence to high levels of non-adherence. The PLS-DA results confirmed the relevance of using four meta-items as filters to open or close subsequent items from their corresponding SPUR dimensions. CONCLUSIONS: The SPUR tool represents a promising new adaptive instrument for measuring adherence accurately and efficiently using the digital behavioral diagnostic tool.
OBJECTIVE: The SPUR (Social, Psychological, Usage, and Rational) Adherence Profiling Tool is a recently developed adaptive instrument for measuring key patient-level risk factors for adherence problems. This study describes the SPUR questionnaire's psychometric refinement and evaluation. METHODS: Data were collected through an online survey among individuals with type 2 diabetes in the United States. 501 participants completed multiple questionnaires, including SPUR and several validated adherence measures. A Partial Credit Model (PCM) analysis was performed to evaluate the structure of the SPUR tool and verify the assumption of a single underlying latent variable reflecting adherence. Partial least-squares discriminant analyses (PLS-DA) were conducted to identify which hierarchically-defined items within each dimension needed to be answered by a given patient. Lastly, correlations were calculated between the latent trait of SPUR adherence and other patient-reported adherence measures. RESULTS: Of the 45 candidate SPUR items, 39 proved to fit well to the PCM confirming that SPUR responses reflected one underlying latent trait hypothesized as non-adherence. Correlations between the latent trait of the SPUR tool and other adherence measures were positive, statistically significant, and ranged from 0.32 to 0.48 (p-values < .0001). The person-item map showed that the items reflected well the range of adherence behaviors from perfect adherence to high levels of non-adherence. The PLS-DA results confirmed the relevance of using four meta-items as filters to open or close subsequent items from their corresponding SPUR dimensions. CONCLUSIONS: The SPUR tool represents a promising new adaptive instrument for measuring adherence accurately and efficiently using the digital behavioral diagnostic tool.
Entities:
Keywords:
Digital questionnaire; Rasch; adherence; drivers of non-adherence; partial credit model; psychometric validation