It is important to realise that up to this point, the inTrigue methodology often has no examples of confirmed patients with disease to learn from. Typically, these would be referred to as 'labels'. At this point it is useful to get the first 'gold-standard' input and validation of the model performance. The ideal person to do that is a specialist clinician.

Volv works with your local medical scientific leads to involve a specialist clinician often from a tertiary care centre, who will, in a blinded fashion, review the anonymised health records of the cohort that the model has highlighted as being at risk. The clinician's assessment is then compared with the results of the model to

• input as gold-standard feedback to further improve model performance
  
• confirm accuracy of the model up to this point
• confirm prevalence estimates for the country data set in question

A scientific publication is the ideal outcome at this point.

Once the review has taken place and the models have been further improved, we can then consider how to embed the algorithm in the medical system.