Addressing objective 2: to identify the features of transitional health
care that are effective and efficient
This WP examines the economic aspects of transition. It has two parts: 1. discrete choice experiment
2. economic modelling.
Discrete choice experiment
Discrete choice experiments are regularly used in health economics to elicit preferences for health-care
products and programmes, and in the valuation of preferences for health states.90–92Further details are in
Appendix 7.
Aim
The aim was to quantify the strength of young people’s preferences for aspects of health services for
transition using a DCE.
Methods
A DCE describes a service in terms of a number of characteristics, or‘attributes’ (e.g. the flexibility of
appointments, parental involvement). The extent to which an individual values an intervention is expected
to depend on the‘level’ these attributes take (e.g. can appointments be made outside office hours or
not?). Thus, a DCE explores relative preferences about how services can be organised, by choosing
between services with differing levels of their attributes.93
The design and conduct of the DCE involved four steps.
Step 1: identifying attributes and levels
Attributes and levels were informed by the Q-sort (WP 1.3). Other information was drawn from previous
research about important features of transition.94The design and presentation of the DCE were
undertaken with the members of the transition research programme’s Collaborative Group and with UP.
The chosen attributes and levels are shown in Table 7.
Step 2: experimental design
The number of attributes and levels described in Table 7 give 128 possible combinations. Statistical manipulation, using Ngene software (2017; ChoiceMetrics, Sydney, NSW. Australia), reduced the number
of scenarios. The resulting D-efficient design had 24 questions, which were further reduced by‘blocking’
them into three groups, so that each respondent needed to complete only eight questions. Young people also answered questions about their current care, defined by the same attributes and levels used in the DCE. For each question, respondents chose between two hypothetical ways in which a service might be organised (see Appendix 7 for specimen questions).
Step 3: data collection
The DCE was completed at visit 3 of the longitudinal study (WP 2.1). Several decision aids were piloted and then used to assist young people who had difficulty completing the DCE (see Appendix 7).
Step 4: data analysis and interpretation
Data were analysed within a random utility model framework using logistic regression techniques. The analysis estimated the predicted uptake of the service: the more an attribute is preferred, the higher the predicted uptake.
Analysis
In total, 247 participants completed at least one DCE question (fully completed, n = 223; partially completed, n = 24); 101 participants did not complete the DCE as they had withdrawn or were lost to follow-up. Young people with ASD were significantly less likely to respond than those with diabetes
mellitus [16/87 vs. 7/112 (χ2= 10.50, degrees of freedom = 2; p≤ 0.001)].
Appendix 7 presents detailed results. In summary, 43 respondents always chose current care and 60 never chose current care. Those not currently receiving any service were significantly less likely to choose current care (p < 0.001). Each attribute and level was often preferred by many young people, as was current care. Table 8 shows that in a service where none of the attributes was present, the uptake of the service would be 78%. Adding a new service attribute, such as flexible appointments, predicted an uptake of 81%. When demographic characteristics (gender, age at data collection and condition) were taken into account, all attributes, except flexibility of appointments, were preferred to a service with none of the attributes (see Appendix 7, Table 23).
Preference for current care was stronger among young men than among young women. Preference for
current care was also stronger among those who had not transferred from children’s to adults’ services
than among those who had. Young men had a stronger preference for out-of-hours clinics than young women did, and a weaker preference for making their own decisions about treatment. Young people with ASD had weaker preferences for parental involvement, and for being the ones to make decisions about their treatment, than those with cerebral palsy.
TABLE 7 Attributes and levels of the DCE
Attribute
Number
of levels Description
Flexibility 2 0 = does not offer appointments outside office hours
1 = offers appointments outside office hours
Staff at appointments 4 0 = do not see the same staff at my appointments and do not have a key worker 1 = do not see the same staff at my appointments but have a key worker 2 = see the same staff at my appointments but do not have a key worker 3 = see the same staff at my appointments and have a key worker
Staff communication 2 0 = yes
1 = no
Parental involvement 2 0 = discouraged
1 = welcomed‘if I want it’
Decisions about care 2 0 = staff discuss my care with me but they make the decisions 1 = staff give me choices but expect me to make the decisions
Extra support 2 0 = no
1 = yes
WORK PACKAGE 2.3: DISCRETE CHOICE EXPERIMENT AND ECONOMIC ANALYSIS
NIHR Journals Library www.journalslibrary.nihr.ac.uk
Strengths and limitations
A DCE can be demanding to complete. Although most young people were able to respond, proportionately fewer with ASD did so. It is possible that these young people have different preferences from those who did complete the DCE. Furthermore, it is possible that the DCE failed to accurately capture the preferences of those young people who completed it. On this latter issue, some reassurance was provided by the broad similarities between the DCE findings and the ranking exercise. The advantage of the DCE over the ranking exercise was that the DCE provided an indication of the relative importance of each attribute.
Discrete choice experiments are widely accepted as a methodological approach preferred by many organisations. For example, they form the basis of eliciting EQ-5D-5L (EuroQol-5 Dimensions, five-level
version) population tariffs throughout the world95and the Center for Devices and Radiological Health
has stated that DCEs are a suitable method of eliciting patient preferences.96The approach we adopted
was consistent with best practice. The aids to completion of the DCE might have infringed strict DCE methodology, but they enabled young people, with a wide range of confidence, intellectual ability and
flexibility of thinking, to complete it. Indeed, the use of aids to completion has been advocated.97–99
A fixed-effects modelling assumed that respondents were consistent in their choices over time. Although such an approach was not incorrect, more sophisticated econometric modelling could be used to explore heterogeneity and whether or not participants considered only a subset of attributes when choosing among alternatives. This might help explain why an appreciable proportion of respondents always chose current care.
Key findings
l Young people with long-term conditions, including those with ASD, could complete a DCE.
l In total, attributes were preferred to them not being present, except for flexibility of appointments.
l Preference for current care was strong and more pronounced among young men than among young
women. This preference was stronger among those who had not transferred.
l Most young people (and especially women) least valued being able to have appointments outside
office hours.
l Young people valued services in which information was passed to the right person, parental
involvement was welcomed, staff offered choice and allowed young people to make decisions about their care, the same staff were seen at each clinic appointment and extra support was available for preparation for everyday life. Young people valued seeing the same staff at each clinic appointment more than having a key worker.
TABLE 8 Predicted probabilities of uptake of services as attributes are added
Attribute Probability of uptake (95% confidence interval)
Current service (none of the attributes) 0.78 (0.75 to 0.81)
New service with flexible appointments 0.81 (0.78 to 0.84)
New service with a key worker 0.79 (0.74 to 0.85)
New service seeing the same staff at each appointment 0.90 (0.85 to 0.94) New service seeing the same staff and having a key worker 0.88 (0.84 to 0.92)
New service with good staff communication 0.93 (0.91 to 0.95)
New service with parental involvement 0.93 (0.90 to 0.95)
New service offering young people decisions about treatment 0.91 (0.88 to 0.93) New service offering extra support to young people for their future 0.86 (0.83 to 0.89)
Inter-relationship with other parts of the programme
From WP 1.1, UP contributed much to the design of the DCE questionnaire. DCEs have not been undertaken often with young people. The understanding of the choice options was therefore discussed and piloted with UP.
The content of the DCE was, in part, informed by the Q-sort study of WP 1.3. The DCE was administered at visit 3 in the longitudinal study WP 2.1.
This WP, 2.3.1, informed WP 2.3.2 on economic analysis.
The results of this WP led directly to one of the implications of the programme, implication 7, concerning elements of service that would make it more likely that transition services were taken up by young people.