Opinion: Groundbreaking innovation is possible in pharmacy

Three years ago, Titan X was launched as the first artificial intelligence system for automating pharmacists’ clinical checks. At the time, the concept was revolutionary for the profession, and was met with equal measures of excitement and scepticism.

Since its launch, Titan X has become one of the most popular apps on Titan’s Marketplace, having been installed in 80 per cent of all Titan sites that process more than 8,000 items a month. It has become the staple of busy pharmacies that now depend on this technology for their existence and the viability of their business.

In the month of April 2024 alone, 1.9 million prescriptions were clinically checked by Titan X, and then labelled, packaged and handed out to patients using Titan’s workflow process, with zero pharmacists involved. Over the past three years, over 20 million prescriptions have been processed by Titan X. The reported error rate has been one in one million, none of which has caused harm to a patient.

This is quite an achievement by any standard. Perhaps the greatest testimony to Titan’s success is the fact that almost every other PMR system has jumped onto the bandwagon and introduced their own version of automation for clinical checks.

Of course, a flag that says ‘don’t show me this repeat again for six months’ is not exactly artificial intelligence – but it does demonstrate the appetite for using technology to reduce pharmacist tasks.

Foundation first

To build Titan X, we didn’t take a legacy system and add AI to it. That would be like adding a battery to a petrol car. Yes, you can get a few miles on electric and call it a hybrid, but it is not the same as a full EV. In the same way, Titan started from a blank canvas and rebuilt the pharmacy workflow from the ground up. We took the whole end-to-end process and broke it down into clinical checks, picking, labelling, packaging and dispatch.

For each of these distinct elements the process was designed around appropriate skill mix. There were no final checks given that accuracy was baked into the step using barcoding, which prevented the next step happening until the previous step was safely passed. Crucially, the pharmacist was moved to a single part of this production line – the clinical check.

“Titan X is now able to predict what a typical pharmacist would do in most scenarios to a high degree of accuracy”

As pharmacists, we inherently know which drugs carry more risk so, from the outset, we identified a set of drugs that would always be excluded from Titan X. All controlled drugs, valproate preparations, cytotoxics, anticoagulants and many other categories are eliminated from automation and go to the pharmacist at all times.

Titan X also excludes items where there is an interaction, or where there are notes on the patient’s record calling for a review. In this way, only low risk items go through Titan X, leaving the pharmacist to focus on high-risk items and scenarios.

Machine learning

With Titan having successfully moved the pharmacist to a single task – that of clinically checking prescriptions – we were able to gather vast amounts of data. We were holding millions of records of what pharmacists did each time a prescription came in. We recorded which doses they changed, which they didn’t, and when they made an intervention.

Titan X was born out of being able to successfully convert this wealth of data into a machine learning algorithm. As more and more pharmacies have joined and dispensed more and more prescriptions, the capabilities of the system have grown stronger and stronger. Titan X is now able to predict what a typical pharmacist would do in most scenarios to a high degree of accuracy.

Military grade audit

We know bullets used in guns are made in factories using a manufacturing process. We also know there isn’t a sergeant major standing at the end of the production line doing a final check on every bullet. So how does the military audit the machinery for something that literally could kill people?

The answer is a concept called Acceptance Quality Limits (AQL). This is a mathematical model that uses sample sizes, inspection levels and confidence levels to determine whether a batch should be accepted or rejected. We incorporated this model into Titan X and adapted it for pharmacy use. In doing so, it meant Titan X could function as a black box, while having an audit that could assess its performance.

The Titan X audit process uses the concept of a burst that lasts up to four hours to provide a sample size upon which an audit is carried out using AQL methodology. You can’t move to the next burst until the previous one has been audited. During the audit, an item can be marked as accepted, minor concern or major concern. The results of these audits then influence future sampling. 

In other words, as confidence grows, the system requires less audit – but if confidence reduces, the audits become more intense. Titan’s audit process ensures a pharmacist remains in control of the dispensary and prevents them from switching the system on and then leaving the pharmacy.

Governance

It is understandable that some people may find it difficult to trust technology that regulates itself and the performance of its systems. To reassure them, I invited a group of seasoned Titan users to form an independent Clinical Safety Group to review the performance of Titan X.

Since its establishment, the CSG has met every six months to review the results of minor and major audits and provide external feedback over the performance of the system. They have also advised on updating the exclusion list as well as providing a steer on refining the algorithm to improve accuracy and patient safety.

In addition to our own safety group, Titan X has been registered as a medical device with MHRA and accredited against the DCB0129 clinical safety standard.
 
The future of innovation

Innovation is desperately needed in pharmacy. The pressures are growing year on year, so we need to release time to deliver more workload. Yet innovation in pharmacy is particularly challenging due to the nature of the products and services we provide – and is made even harder due to our general reluctance to change.

The solution is not just about building a piece of clever tech. It is also about creating a model of practice that upholds our values as a profession. Our journey with Titan X over three years has demonstrated that groundbreaking innovation is possible in pharmacy. We can create solutions that help solve the big challenges of our time, while meeting high standards that build trust among the profession, stakeholders and patients alike.