OBJECTIVES: Person-Centered-Care requires knowledge about patient preferences. For People-living-with-Dementia (PlwD) evidence on quantitative preferences is limited. The Analytic-Hierarchy-Process (AHP) has been suggested suitable for preference-elicitation among PlwD. An important aspect in an AHP-survey is the assessment of acceptability, reliability and consistency.

METHODS:

Data from the PreDemCare-study, including an AHP-survey (6x2-(sub)criteria experimental-design) with n=50 PlwD and n=25 physicians. AHP-weights were calculated with principal-eigenvector-method and group aggregated by Aggregation-of-Individual-Priorities-mode in ExpertChoice-Comparion® and R. Acceptability was assessed by missing values and survey-difficulty evaluated with a 5-point-scale (easy-to-difficult). Reliability included descriptive analyses of contradiction and Intra-Class-Correlation-(ICC)-analyses for agreement between re-test-questions. Individual consistency ratios (CRs) at criteria-level as a measure of logical judgement-performance in AHP were calculated and aggregated per group, with an accepted inconsistency-threshold of patient’s CR≤0.3, physician’s CR≤0.2. We tested for correlation of CR by severity of cognitive impairment (DemTect, Mini-Mental-Status-Examination), diagnosis (yes/no), status (patient/physician) and by sum of values ≥5/ =1 value-judgements with one-way Analysis-of-Variance, Kruskal-Wallis-test, Mann-Whitney-U-test, Kendall’s and Spearman’s correlation.

RESULTS: The AHP-survey had zero missings. Two-third of patients rated the survey as rather easy/neutral, no one as difficult. 12/25 physicians rated the survey as easy/neutral, and 4/25 as rather easy/difficult respectively. Descriptive analyses of re-test-questions showed some contradiction, only n=3 (6%) patients and n=2 (8%) physicians chose different (sub)criteria. ICCs showed moderate-agreement at criteria- and substantial-agreement at sub-criteria-level in both groups. Mean-CRs were acceptable for patients=0.261 and physicians=0.181. No significant differences of CR by sociodemographic/clinical characteristics could be found. CR increased significantly with higher sum of ≥5 value-judgments and decreased significantly with sum of =1 value-judgements.

CONCLUSIONS: The AHP showed good acceptability, moderate-substantial agreement in reliability, and both group’s consistency below defined threshold. Findings should be tested in larger samples, which may include adjustment of the AHP-judgement-scale to avoid use of high values and related risk of inconsistency.