Objectives Behaviour change interventions often target 'important' beliefs. The literature proposes four methods for assessing importance of attitudinal beliefs: elicitation frequency, importance ratings, and strength of prediction (bivariate and multivariate). We tested congruence between these methods in a Delphi study about selective decontamination of the digestive tract (SDD). SDD improves infection rates among critically ill patients, yet uptake in intensive care units is low internationally. Methods A Delphi study involved three iterations ('rounds'). Participants were 105 intensive care clinicians in the United Kingdom, Canada, and Australia/New Zealand. In Round 1, semi-structured interviews were conducted to elicit beliefs about delivering SDD. In Rounds 2 and 3, participants completed questionnaires, rating agreement and importance for each belief-statement (9-point Likert scales). Belief importance was assessed using elicitation frequency, mean importance ratings, and prediction of global attitude (Pearson's correlations; beta-weights). Correlations between indices were computed. Results Participants generated 14 attitudinal beliefs. Indices had adequate variation (frequencies: 4-94, mean importance ratings: 4.93-8.00, Pearson's correlations: ±0.09 to ±0.54, beta-weights: ±0.01 to ±0.30). SDD increases antibiotic resistance was the most important belief according to three methods and was ranked second by beta-weights (behind Overall, SDD benefits patients to whom it is delivered). Spearman's correlations were significant for importance ratings with frequencies and correlations. However, other indices were unrelated. The top four beliefs differed according to the measure used. Conclusions Results provided evidence of congruence across three methods for assessing belief importance. Beta-weights were unrelated to other indices, suggesting that they may not be appropriate as the sole method. Statement of contribution What is already known on this subject? Attitudinal beliefs (specific beliefs about the consequences of performing an action) are key to designing interventions to change intentions and behaviour. The literature reports four methods for assessing the importance of attitudinal beliefs: frequency of elicitation in interviews, importance ratings in questionnaires, and strength of prediction (bivariate and multivariate) of global attitude scores. The congruence between these measures of importance is not known. What does this study add? Four indices of importance were examined in a multi-professional, international study about the use of selective digestive decontamination to prevent infection in intensive care settings. Three indices were correlated with one another. Each method used to assess importance produced a different subset of the most important beliefs. Selection of the most important beliefs should use multiple assessment methods. This evidence suggests that multiple regression approaches may not be appropriate as the sole method for assessing belief importance.