Data analysis plan The analysis will be conducted on an intention-to-treat selleck screening library (ITT) basis.

Exploratory analysis will be conducted first for outcome and patient background variables; descriptive statistics of each variable will be presented separately for each group at each follow-up point, with means and SD for normally distributed variables, medians (IQR) for skewed variables and frequency (percentage) for categorical variables. Missing values will be checked and reported. Multiple imputation will be used to hand missing values, based on a multilevel modelling approach. To compare the number of visits needed to achieve an ADHD diagnosis (either confirmed or excluded) between groups, Poisson regression with binary group status as the explanatory variable will be implemented. To compare clinician’s confidence in their diagnostic decisions, multilevel modelling with patient as a level

2 unit will be used to take into account the non-independence within patient data due to repeated measures.41 κ Statistics will be used to reflect the stability of diagnosis between first confirmed diagnosis and diagnosis rerated at 6-month follow-up time. κ Statistics will be reported for each group and the stability of diagnosis will be compared between arms using logistic regression. The same analysis approach will be implemented to explore the stability of diagnosis confidence between time of first confirmed diagnosis and 6-month follow-up. To assess the diagnosis accuracy, the sensitivity, specificity, likelihood ratio (LR) ve+, LR ve−, positive predictive value (PPV) and negative predictive

value (NPV) will be reported for each group and the test performance will be compared between QbO and QbB arms.42 43 Receiver operating characteristic curve analyses will be used to obtain the best predictive model based on QbTest scores that discriminates between ADHD ‘positive’ and ADHD ‘negative’ gold standard DAWBA diagnoses. For treatment related outcomes (phase 2) outcome measures such as SNAP-IV, side effects scale, SDQ and C-GAS scores, multilevel modelling with patient as a level 2 unit will be again Batimastat applied to quantify the difference between QbO and QbB arms. For time to event variables such as time to diagnosis (in days), survival analysis using log-rank test will be performed for group comparison and Kaplan-Meier survival curves will be displayed for each group. Logistic regression will be used to compare the proportion of normalisation between two groups at 6-month follow-up time. For all regression modelling to explore the difference between arms, group status will be included as explanatory variables. Data transformation would be needed for skewed outcome variables. Health economic evaluation Economic evaluation will be completed primarily from a health service perspective but in addition from a societal perspective. A cost-effectiveness and cost utility analysis of the treatment options will be conducted.