Statistical methods
We assessed treatment effect for superiority between groups by linear
regression on an intention to treat basis, with results considered
statistically significant if the p-value was <0.05. We
compared baseline characteristics of the final cohort with those of the
group lost to follow-up with linear regression and non-parametric tests.
Data were analysed with IBM SPSS version 26.0 (IBM Corp., Armonk, NY)
and R Studio version 1.2.5033.
The economic evaluation was conducted from a societal perspective,
including direct and indirect medical and non-medical costs over 12
months. Incremental costs per IIALY gained were expressed as an
Incremental Cost-Effectiveness Ratio (ICER). The balance between costs
and QALYs were expressed as an Incremental Cost-Utility Ratio
(ICUR).21 Costs and effects were recorded and
calculated on an individual basis, then the mean differences between the
two study groups were calculated. The ICER and ICUR represent the
average incremental cost needed to be invested to achieve 1 additional
unit of the measure of effect and were computed by dividing the
differences in mean effects and mean costs (as shown in Appendix A). By
performing 5,000 bootstrap replications of the trial data, alternative
confidence intervals were calculated based on the
2.5th and 97.5th centiles.
Cost-effectiveness planes visualise the uncertainty surrounding the ICER
and ICUR. If the app-based treatment saved costs and differences in
effects to be minimal, we would not construct an acceptability curve to
assess the probability of cost-effectiveness, as this would already
imply accurate cost-effectiveness based on the difference in costs.
Additionally, we performed a sensitivity analysis for a scenario with
higher costs for app maintenance and extra costs for annual development.
Data robustness was assessed by using the mean of the follow-up data at
4 and 12 months to estimate costs between 4 and 12 months. Finally, we
performed subgroup analyses with the type of recruitment or type of UI.