Rules of the Trade

Rules of the Trade

AI and Technology
in Healthcare

terena-bell

Christophe Djaouani

Christophe Djaouani joined SDL in 2018 with the acquisition of Donnelley Language Solutions.
He successfully led the integration of the regulated industries team providing solutions to the financial, legal, life sciences, and investor relations markets.

The previous edition of “Rules of the Trade” examined the role of patient centricity in clinical trials. I’d like to flip the coin and delve further into the growing role of AI and technology in healthcare.

One element that will drive the success of all up-and-coming developments in healthcare is the involvement of the language industry — it has a lot to offer to the healthcare sector by way of technology to speed up and enhance many of the processes where language is involved (hint: everywhere). This unrealized potential is on its way, especially with the events of the past 12 months resulting in the rapid shift towards digital in all industries. Adoption of AI and technology is understandably accelerating.

Barriers to the adoption of AI and tech

One could say that the life sciences and the healthcare services sectors are notoriously slow on the uptake of new ways of working, but this is not without good reason. To safeguard their patients, these sectors are heavily regulated, often weighed down with onerous regulations and processes.

Healthcare service providers and insurers have come under increased regulatory pressure in recent years to provide language-appropriate materials and services for individuals who do not understand the native language of the country they are in. This requirement further amplifies an already major pain point — time-to-market.

Thanks to the transition to digital communications (web, social, video, print, and so on), we have witnessed an unprecedented explosion of content volumes over the past decade, leaving many organizations struggling to keep up. And in this new context, manual translation alone is simply no longer an appropriate response.

The post-COVID landscape

In an unexpected twist of events, the barriers listed above were upended when COVID-19 forced the life sciences and healthcare sectors to embrace new ways of working. Adoption of medical technology and pharmaceutical services was boosted by the COVID-19 crisis through the sudden increase in production of diagnostic testing supplies, personal protective equipment (PPE), and acute care devices such as ventilators. Over the last year, the entire world has been holding its breath waiting for the successful development of vaccines, which have thankfully arrived and are being administered at varying rates around the world.

In addition, the acceleration towards telemedicine and tele-health models means we are arriving at a new state, known as the new health economy, earlier than predicted. Language technology can be a major part of that transformation.

GLOSSARY AND TERMS FOR FURTHER LEARNING

To familiarize yourself with the possible applications of language AI and technology in content manage-ment for life sciences, here are some key terms.

Machine translation: The field of computational linguistics that automates translating text from one language to another.

Statistical machine translation (SMT): This approach replaced rule-based translation (RBT) and example-based translation (EBMT). SMT generates translations based on statistical models that are trained over time by comparing parallel texts of a source language and a human-translated target language.

Neural machine translation (NMT): NMT can be applied to any language pair and adapted to particular writing styles and formats, such as those applied in pharmaceuticals, medical devices, diagnostics, and other regulated documents. One of its disadvantages is lack of consistency, which is a real issue in healthcare where accuracy is a must.

Post editing: An additional step undertaken by a human translator to review and correct any errors in a piece of machine translation by comparing the original and the machine output.

Natural language processing (NLP): NLP is a branch of AI that enables machines to understand human language. This is what allows us to create chatbots or apply spellchecks to our work.

Reinforcement learning: Useful when dealing with enormous volumes of data, RL is an effective way to train an NMT system. RL is considered the best way of achieving true intelligence.

Conversational AI: The use of chatbots is increasing as a way for businesses to stay in touch with their customers 24/7. More progress is needed to human-ize interactions with them, with the way language is formulated being one of the pain points — users often report that the repetitive responses make it clear the experience is very much automated.

GLOSSARY AND TERMS FOR FURTHER LEARNING

To familiarize yourself with the possible applications of language AI and technology in content manage-ment for life sciences, here are some key terms.

Machine translation: The field of computational linguistics that automates translating text from one language to another.

Statistical machine translation (SMT): This approach replaced rule-based translation (RBT) and example-based translation (EBMT). SMT generates translations based on statistical models that are trained over time by comparing parallel texts of a source language and a human-translated target language.

Neural machine translation (NMT): NMT can be applied to any language pair and adapted to particular writing styles and formats, such as those applied in pharmaceuticals, medical devices, diagnostics, and other regulated documents. One of its disadvantages is lack of consistency, which is a real issue in healthcare where accuracy is a must.

Post editing: An additional step undertaken by a human translator to review and correct any errors in a piece of machine translation by comparing the original and the machine output.

Natural language processing (NLP): NLP is a branch of AI that enables machines to understand human language. This is what allows us to create chatbots or apply spellchecks to our work.

Reinforcement learning: Useful when dealing with enormous volumes of data, RL is an effective way to train an NMT system. RL is considered the best way of achieving true intelligence.

Conversational AI: The use of chatbots is increasing as a way for businesses to stay in touch with their customers 24/7. More progress is needed to human-ize interactions with them, with the way language is formulated being one of the pain points — users often report that the repetitive responses make it clear the experience is very much automated.

Applications of language AI and technology

Patients justifiably prefer to receive information in their native language, which means companies must get their content translated into multiple languages — for example, outreach materials to assist with patient recruitment and retention. Recruitment has seen an increase in the use of e-consent software to facilitate and enhance the enrollment process, so this is an obvious place for large scale gains to be made with technology.

Another area is clinical trial design and optimization. Particularly in the last year or so due to COVID-19, virtual clinical trials have become an alternative in response to limitations on physical interactions. Service providers in this area are using operational data to drive AI-enabled clinical trial analytics and language technology has a critical role to play in ensuring the correct interpretation of captured findings.

A third example is the use of language technology to optimize patient monitoring and medication adherence, especially with the transition to provision of healthcare via digital platforms. Rapidly delivering accurate multilingual content here would directly improve the ongoing health management of individuals through more accurate diagnoses of their conditions. This impact cannot be overstated.

What’s in it for the end customer?

LSPs are pouring huge resources into the R&D of language technology with the aim of alleviating several key pain points for customers:

  • Patient experience. The experience for a patient to whom materials are presented in their native language is incomparable to that of an individual forced to wrestle with content in another language. Again, the positive impact on trust, reputation, and ultimately loyalty and retention from this consideration is hard to measure.
  • Enhanced security. Adherence to critical data security regulations like HIPAA have become table stakes to consumers, while providers fully appreciate the risk to their reputation and trust from the threat of patient data being compromised. Secure language technology has a major role to play in the overall security ecosystem by which patient data is protected.
  • Cost reduction. Where healthcare is a paid-for service, efficiencies through implementation of technology can be passed onto patients and also have the added benefit of enabling the healthcare provider to be more competitive in the market.
  • Turnaround time reduction. Using a human translator or interpreter can be time-consuming and significantly increase turnaround time. This can in some cases be alleviated by using machine translation with postediting, or if using an interpreter, over the phone interpreting (OPI).
  • Increased accuracy. Currently, machine translation does not provide accurate enough results to allow it to be deployed independently of humans, but this can be offset by adding pre- and post-editing steps, resulting in overall efficiency gains.

RELATED ARTICLES