Ietje Perfors

182 Chapter 7 we designed the GRIP evaluation in a RCT design. This brought several methodological challenges. The RCT showed a higher drop-out in the intervention group (23%) compared to the control group (10%), which may have caused selection bias. Reasons for drop-out in the intervention group were ‘felt to well’ (n=5) and ‘no extra care needed’ (n=3). This may have left selected patients in the intervention group with lower baseline functioning, resulting in worse outcome in the intervention group. However, we did not find differences in quality of life between the two study groups. The GRIP trial focuses on the patient experiences and uses a wide range of subjective outcomes measures. The use of PROMs in this study may have affected the results. Patients were not blinded, which could have influenced the outcomes, especially the subjective ones. We hypothesized that the intervention may have raised expectations about GP involvement in the intervention group, that were notmet in practice since the homecare oncology nurse was usually the primary care professional involved. This may have caused disappointment among patients in the intervention arm, resulting in an extra negative evaluation. We did not assess these expectations, but literature shows they have an influence on satisfaction measurements. 8, 9 Chow et al. 2009 even suggest a paradox between patient satisfaction and quality of care, in which patient satisfaction might also decrease if quality of care increases, due to the higher expectations. 10 In future evaluations optimal blinding, e.g. by using a cluster randomized or stepped wedge design, and the measurement of expectations should be aimed for, especially when assessing satisfaction as an outcome. Some constructs that might be affected by the intervention are difficult tomeasure. The hypothesis behind GRIP was that the intervention would result in more patient centred care and in more continuity of care. Both of these constructs proved difficult tomeasure. 11, 12 Even though RCTs are reputed for providing themost robust effect evaluation alternative designs could be considered. To prevent contamination of the control group we discussed a cluster randomized trial.