Natural Language Generation and its applicability in an environment where an ambulance just can't reach
Let us introduce you…meet Alan and Sarah. Alan lives in the UK and his job is cabin crew for a European airline. 6,000 miles away, Sarah lives in Botswana. She works as a Park Ranger and spends her days providing safari experiences for international tourists.
Both Alan and Sarah share a common responsibility. They are first responders, not doctors, paramedics or healthcare professionals but someone with just a few days of first aid training. Should they face a medical emergency in their job, they are tasked with delivering best possible care and communicating the incident until such time that they handover to medical professionals. These responders can be significantly remote – in time or distance – from professional and/or definitive healthcare.
The medical emergencies they face can vary; for Alan he may deal with an in-flight cardiac arrest or a passenger with potential stroke. Sarah may see the same, or she may have to manage a patient who has been crushed by an animal or has a gaping wound.
For professionals (like paramedics) these types of medical scenarios are hard enough. As well as managing the patient they complete multiple pieces of paperwork. An ambulance clinician could resort to writing on the back of their glove to quickly record information before transferring this on to the ‘proper’ paperwork, be it real paper or an electronic patient report form. This can be inefficient and time consuming, but at least as professionals they can interpret patient’s vital signs and have the knowledge and regular experience to systematically communicate on-scene data.
But what about Sarah and Alan? How can they determine if a patient is deteriorating or improving? How do they record this data for handover to emergency services or for their own organisation to use for post-incident audit, governance or training?
With devices like smartwatches increasing in popularity, it is becoming easier to automatically record a person's vital signs such as heart rate and breathing rate. Clinically-proven versions of wireless medical sensors can support the responder to do their job without having to stop what they're doing and take measurements using the very limited paper space they have.
Data-mining techniques can detect trends in the data such as if someone's heart rate is increasing or decreasing. The information detected by these techniques can be communicated using natural language generation (NLG). NLG reads data or text, and produces a textual output which can be adapted depending on the audience and purpose.
One example of this being used in the real world is BabyTalk. This is an NLG system used to continuously monitor premature babies. Babies who are born before they are fully developed are more likely to have health problems and need to be monitored closely so issues are identified as soon as possible. The quicker problems are detected and dealt with, the less likely the issue will cause serious harm to the baby.
Using NLG means that more than one report style could be used to present the same information. A report for doctors is very factual, has a flat tone and contains more medical terminology and details of the baby's health, whereas reports for parents need less technical detail and a more caring tone.
Going back to our company vision of using digital technologies to support emergency first responders, we are already seeing the use of NLG in hospital environments. For example, Suregen is a system used in a German hospital to create cardiology reports. Data is entered into the system, and it produces an NLG report describing the data. Using a system to generate text saves time spent on paperwork. They can even be more accurate as computers are better at tedious jobs such as report writing, in which humans may lose interest and/or make unintended errors.
As a company, MIME Technologies has been validating the use of NLG in pre-hospital care with first responders and in differing remote care environments. The technology is already proving it works in the hospital space, hence the opportunities for use in pre-hospital patient monitoring are plentiful!