Table 1.

Conjoint Analysis process

Conjoint Analysis requirements	Corresponding stage in process
Conjoint Analysis relies on the development of a set of attributes or criteria that describe a given product.
	Stage 1: Identifying plausible and meaningful innovation attributes that could be used to characterize healthcare innovations (for example, financial cost).
Levels of each attribute (such as £0, £100, £1,000) for each criterion are assigned. These need to be meaningful and able to be traded off.	Stage 2: Operationalizing the attributes of innovations and their levels. Twelve postnatal depression treatments being considered by the Trust for implementation were described using these attributes.
Hypothetical scenarios with different combinations of attribute levels are included in the questionnaire.	Stage 3: A fractional factorial design is used to identify the number of hypothetical scenarios to be included.
Eliciting stakeholder preferences. In choosing or rating, respondents must trade off some elements of the innovation (for example, cost) for an increase in another attribute (for example, quality); a process known as the ‘marginal rate of substitution,’ [12]. Analysis of the choices made yields estimates of how much respondent stakeholders are prepared to trade off in their preferred attributes in order to receive their preferred combination of attributes.	Stage 4: Information about clinician preferences for innovation attributes is collected using a questionnaire. Respondents rate hypothetical (but feasible) innovations, products or services described using these criteria. Analysis reveals the importance of each attribute, and the clinician preferences for each attribute at each of its component levels. Alternative methods such as Choice Based Conjoint are available that quantify individuals’ values in terms of their willingness to pay (WTP) for an innovation, but rating scales have been shown to perform well in eliciting preferences for healthcare services [27].
Estimating utilities to determine the importance of each attribute to stakeholders.	Stage 5: Analysis of data using Sawtooth software provides preference scores (utilities) for each attribute.
Independent scoring of innovations.	Stage 6: ‘Scoring’ of potential postnatal depression treatments using the attributes and levels identified.
Matching clinician preferences for innovations with the scored innovations	Stage 7: The 12 innovations were ranked according to the preferences of the clinicians who would have to implement them.