The key here is 'clinically validated data'. The algorithms...

The key here is 'clinically validated data'. The algorithms should improve if 'retrained' with this additional data - although more data has diminishing returns. This clinically validated data requires the input (cough and attributes) and the output (clinically validated diagnosis). General use will not improve the algorithm, but it may lead to further insights that are needed during clinical stages.

On the flip side, the algorithm can degrade. All predictive models assume the future is sufficiently like the past. When the future is different, say an entirely new virus takes hold, the algorithm may decline in accuracy.

The main thing for RAP to do is become the authority on lung sound diagnosis. Once perched as the authority, it will be hard for the company to be dislodged.