docet

Key Points

• There is currently no formal screening for either chronic open angle glaucoma or ocular hypertension.
• Following the publication of NICE guideline CG85, referrals to HES for suspected COAG and OHT doubled.
• Referral refinement and revision schemes reduce false positives but vary in nature throughout England.
• Bayesian artificial intelligence algorithms can be used to calculate the probability of an eye condition like COAG being present from its prevalence and any new evidence that arises from diagnostic tests.
• A study compared the outcomes based on a Bayesian matrix prediction of glaucoma with the decision made by three specialist optometrists in Kent
• Accuracy of data can be approximated by using the same data to both build and to test the system in use and accuracy levels can be established using randomised tenfold stratified cross validation – this is believed to give the best accuracy of data.
• A weighted average accuracy of 95.4% was reached – of just over 1000 patients studied, 34 people followed up or referred by the clinicians were discharged by the Bayesian method while 31 patients were referred or followed up who would have been discharged by the clinicians.
• This technique may have application with other diseases.