Wouter Leclercq

Chapter 10 174 quantify patients’ subjective perceptions of the effect that treatment has on their daily lives, in a standardised manner. Furthermore, collection of patient-reported outcome data in clinical settings has been proven to help inform clinical decision making and can be helpful in reflecting the quality of health care. 30 Above all, a patientreported outcome measure has the potential to personalise the outcome, making it possible to measure the effect on each individual patient, which is essential since normal activities are different for every participant. In our study, participants selected the activities that were most relevant for them in daily life; thus, the primary outcome of return to normal activities was specific to outcomes that mattered to participants. Moreover, the measurement process and the main part of the e-health intervention (the possibility to make a convalescence plan) were strongly related to each other, because activities that could be selected for the convalescence plan were (for the greater part) derived from the PROMIS-PF item bank. Therefore, the effect of the intervention could be measured very specifically. Further strengths of our study are the low proportion of dropouts (4%) and the fact we could mask participants to the study hypothesis. This type of masking is unique in the area of e-health research and is the best possible option, because complete masking of participants is not possible since treatments compared are inherently different. 31 All analyses regarding the primary outcome measure, including the per-protocol analysis, showed a significant effect in favour of the intervention group, which shows the consistency of our findings. One might argue that the criterion for including participants in the per- protocol analysis (i.e., creating a convalescence plan) might not be sufficient because we do not know if participants used the plan they created. However, this objectively measurable outcome acts as a proxy to create a clear cutoff. Selection bias is likely to play a part in the per-protocol analysis because a selected subgroup of participants from the intervention group was compared with all participants randomly allocated to the control group. Care should, therefore, be taken when interpreting the results of the per-protocol analysis, which in principle concerns a nonrandomised comparison. Our study has other limitations. First, the proportion of participants included in the study after assessment for eligibility was low (38%). Although we could compare participants and non-participants with respect to age, sex, and surgical procedure, selection bias cannot be ruled out completely. Further, we did not register the reason for non-participation in patients who were not willing to participate. Was the inclusion percentage low because of the study setting and the burden of completing questionnaires during a 6-month period after surgery, or because patients prefer not to use an e-health intervention at all? A survey study in a comparable population showed that 78% would prefer to use e-health in perioperative care; 12 thus, the most