Palliative day care is offered to patients with a life-limiting illness. It provides a range of services to meet the varying needs of individuals. Consequently, the service is multidimensional and the outcomes are hard to define. This means this service operates at a level of complexity that is different from other health care services. This presents a challenge for evaluation since the activities differ across individuals and change over time, depending on their needs. This challenge is not unique to palliative care and is relevant to the evaluation of other services. The published guidelines for health economic evaluation have not explicitly taken into account the specific issues relevant to evaluating complex services. It is argued in this thesis that the 'health gain' approach is problematic and a preference-based approach may reveal more useful evidence for policy-makers. This has not been fully considered in empirical studies. A study was undertaken using health-gain and preference based approaches to evaluate a new palliative day care service. The EO-50 health-related quality of life instrument was used to detected differences in outcome between patients who attended a palliative care day centre and those who did not. The instrument did not detect significant differences over time. A choice experiment was also undertaken in four POC centres. This approach estimates the relative preferences that respondents expressed for specific service attributes. The attributes chosen for this study were: opening hours, access, specialist therapies; medical support; hairdressing and bathing. The results showed that specialist therapies were relatively more valued, and that hairdressing and bathing were not important in decision-making. Access to medical care was less important than access to specialist therapies. The thesis critiques health gain approaches for services that have a broad range of hard-todefine goals and aim to meet individual needs. The choice experiment provided insights into how services are valued where these insights cannot be derived from other economic evaluation approaches. Research methods that reflect people's preferences may provide important analysis where very little evidence has previously existed.