Assessing quality of care in nursing homes using discrete choice experiments: How does the level of cognitive functioning impact upon older people's preferences?

Traditionally older people with mild to moderate cognitive impairment have been excluded from preference elicitation studies in health economics. We assessed the impact of the level of cognitive functioning on preference and scale heterogeneity in a discrete choice experiment undertaken with 126 older people living in residential aged care homes in Australia between January 2015 and February 2016. Data was analysed using conditional logit models for sub-groups of participants with mild to moderate cognitive impairment (N = 52) and without cognitive impairment (N = 74), and for the entire study sample using a heteroscedastic conditional logit regression model allowing for scale heterogeneity. The Swait-Louviere test was undertaken to formally test for differences in preference and scale between the two groups. Cognitive impairment was not significant in the scale function of the heteroscedastic conditional logit model (beta = −0.403, SE=0.341, p = 0.237). There were no statistical differences in estimated vector of preference parameters based on the presence or absence of cognitive impairment (Chi-squared = 13, 25 df, p = 0.976). Although there was evidence of a small increase in response variability with increasing cognitive impairment this did not reach statistical significance, and we were able to combine responses for people with and without cognitive impairment. Overall, the findings provide support for the more widespread inclusion of older people with mild to moderate cognitive impairment in such studies.