Acceptability, reliability and validity of the Patient Reported Experiences and Outcomes of Patient Safety in Primary Care (PREOS-PC) instrument

The problem

Despite the significance of patient safety for primary care, the knowledge base in this context is still sparse, particularly in relation to patient reported experiences and outcomes. We aimed to examine the psychometric properties of the Patient Reported Experiences and Outcomes of Patient Safety in Primary Care (PREOS-PC), a newly developed patient reported instrument to measure patient safety experiences and outcomes in primary care.

The approach

The PREOS-PC questionnaire was posted to 6,675 adult patients from 45 purposefully selected general practices distributed across five regions in England (Manchester, Nottingham, Keele, Southampton and Birmingham). The questionnaire contained the following eleven scales: 1) safe environment; 2) perception of patient safety-A; 3) perception of patient safety-B; 4) trustworthiness; 5) patient safety patient activation; 6) impact of harm on overall health; 7) impact of harm on specific aspects of health; 8) impact of harm on specific aspects of needs; 9) time to recover from overall harm; 10) time to recover from specific types of harm; and 11) amount of overall harm experienced. Acceptability (item-specific response rate, floor/ceiling effect), reliability (internal consistency, test-retest), structural validity (confirmatory factor analysis, correlation among constructs), and ability to discriminate among practices (intra-cluster correlation) were examined.

Findings

1,244 (18.6%) completed questionnaires were returned. Respondents were more likely to be female, elderly and of ‘‘white'' ethnicity. Item-specific response rate ranged from 98.9% to 39.1% (median 91.3%, interquartile range 28.0%). Ceiling effects were observed for four scales (scales 1 - 4), and floor effects for two scales (7 and 8). The five multi-item scales (1, 5, 7, 8 and 10) showed evidence of high internal consistency, with Cronbach's ? ranging from 0.80 to 0.96, and inter-item correlations from 0.51 to 0.83. The test-retest intra-class correlation coefficients were above the expected 0.70 standard only for the scales 1, 2, 5, 7, and 9. In terms of structural validity, item loadings derived from confirmatory factor analysis were generally high, ranging from 0.43 to 0.94 on all the scales. Statistically significant correlations confirmed our pre-specified hypothesis for construct validity (e.g. positive correlation between the scales 1 and 2 (r=0.61), and negative correlation between the scales 1 and 7 (r=-0.44)). Intra-cluster correlation coefficients (statistically significant for all except two scales (10 and 11)) ranged from 0.02 to 0.20, indicating small to moderate differences in patient responses across practices.

Consequences

Preliminary psychometric analyses support the acceptability, reliability, and construct validity of PREOS-PC for the measurement of patient perceptions of processes and outcomes of patient safety in General Practices. Further work is planned to produce shorter versions, explore alternative administration methods, and investigate the potential utility of the instrument in informing primary care service planning and configuration.

Credits

• Ignacio Ricci-Cabello, Faculty of Medicine & Health Sciences, University of Nottingham, Nottingham, UK
• Anthony Avery, Institute for Population Health, University of Manchester, Manchester, UK
• David Reeves, Institute for Science and Technology in Medicine. University of Keele, Keele, UK
• Umesh Kadam, Institute for Health Research, University of Exeter Collaboration for Academic Primary Care
• Jose María Valderas