Risky business: underwriting medical conditions

The customer isn’t always right

The underwriter has the unenviable job of balancing two conflicting impulses: the customer’s desire for a product that is affordable and comprehensive, and the insurer’s desire not to go bankrupt

For starters, customers are not always the most reliable narrators when it comes to their own conditions, often omitting key information about their medical history – or potentially something less obvious. While it might seem clear that their age, build and predilection for smoking and/or alcohol consumption will be a factor, it may not occur to them that their residential and travel history will also contribute to their risk profile – for example, insurers will want to know whether a customer has travelled to or lived in a country with high rates of HIV/AIDS, even if the customer wasn’t actually aware of that fact.

Therefore, it is essential to communicate to them upfront that they need to give their insurer all relevant (and seemingly irrelevant) information, in order to avoid problems down the line.

Specific documentation from health practitioners might be needed in some cases, though generally it is hoped that the customer can provide all necessary data in the first instance.

And even seemingly obvious omissions are not necessarily intentional. Dipesh Patel, Senior Travel Underwriter at Allianz Partners, told ITIJ: “Occasionally, a customer may unintentionally omit information about their medical conditions; not everyone keeps an updated list of the various pills they need to manage their condition, so it can sometimes be overlooked.”

Customers may also be confused sometimes about exactly what does and doesn’t constitute a pre-existing medical condition – a sensitive area that requires careful navigation. For example, they may not realise that just because they are taking medication to deal with an ongoing condition, that doesn’t magically make the condition invisible or irrelevant for the purposes of insurance.

“Addressing the misconception that managed conditions, such as controlled blood pressure, do not constitute a pre-existing medical issue is crucial in ensuring proper disclosure for insurance purposes,” said Patel. “Too many people think that because they’re regularly taking blood-pressure pills to stabilise their blood pressure, it doesn’t count as a medical condition; but it still needs to be declared.”

It’s also essential for customers to understand that this process is an ongoing dialogue, and that they must update their insurer regularly on any changes to their medical profile. This enables insurers to either confirm that the cover they’ve purchased remains valid, or inform them that the policy needs to be tweaked or otherwise updated.

In order to do this, Patel explained, underwriters will collaborate with other business partners, such as banking associates or tour operators, to make sure the disclosure requirement is made explicit during the sales journey.

Ultimately, our view is that someone with a medical condition that is being treated is potentially less of a risk to cover provided the condition is stable and well controlled

“Ultimately,” said Patel, “our view is that someone with a medical condition that is being treated is potentially less of a risk to cover provided the condition is stable and well controlled. This is another reason why it is important customers declare their whole medical history.”

But how are risks actually calculated once the underwriter has all the information they need?

A complex equation

“Underwriters often use third-party medical risk assessment systems to help evaluate declared medical conditions and their potential impact during travel,” said Patel. “These risk assessment systems are useful tools for underwriters, helping them make informed decisions based on a standardised set of criteria. While each underwriter may have their own appetite towards risk, these systems provide a consistent framework for evaluating and managing risk.”

But in order to make this process as comprehensive and efficient as possible, both the quantity and quality of the information received needs to be high. Patel explained: “Insurers need access to granular detail in order to analyse the loss ratio by risk score. This granular detail helps underwriters to understand the relationship between risk and claims, which in turn informs underwriting decisions, pricing strategies, and risk management practices.”

Then, once the underwriter has access to all that necessary granular detail (which may require further questions for the customer, so let’s hope they’re feeling cooperative), it all needs to be compiled and analysed individually before an overall risk profile can be established.

The lingering effects of the Covid-19 pandemic, the consequences of rising inflation, and persistently ballooning prices will all conspire to squeeze profits from every angle

“Typically insurers would want to evaluate loss ratio performance split by variables such as screening score, policy type, age and destination,” Patel told us. “From there, the insurer can better establish more accurate pricing for customers with specific conditions.”

The age-old question

Underwriters will face all manner of hurdles over the next few years. The lingering effects of the Covid-19 pandemic, the consequences of rising inflation, and persistently ballooning prices (including the cost of essential drugs) will all conspire to squeeze profits from every angle.

Another thorny problem – though one that is by no means unique to insurance – is the issue of ageing populations. Businesses and politicians in practically every country in the world are figuring out how to deal with the challenge of more and more people living ever longer lives, and requiring ever more complex care as a result. In terms of medical underwriting, it’s one of many million-dollar questions, and an area that will likely necessitate constant agility and adaptation.

“With an ageing population,” Patel said, “we are increasingly seeing that individuals’ propensity to travel does not diminish as they get older. It is true that some insurers will need to adapt to the behaviour change. In the future we may see a shift towards more personalised underwriting approaches, with insurance premiums being tailored based on the individual’s specific health profile, travel history and lifestyle factors. There is a trend for customers to use health apps to gather real-time health data, which may be useful to adjust premiums dynamically. [Overall], in the context of ageing populations, the future of medical underwriting is likely to involve more personalised, data-driven approaches.”

We have the technology

As is often the case, the way forward is at least partly a technological one. Automated underwriting, for example, could help to streamline the process via the algorithmic analysis of vast quantities of medical data, potentially paving the way for a simpler, more consistent offering.

Artificial intelligence (AI) and machine learning (ML) have demonstrated proven results in various industries, including travel insurance, and many major companies have made strategic investments in these technologies to enhance their products.

Technologies like machine learning or generative AI allow underwriters to consider a very large number of data points in their assessment

“The importance of AI algorithms in the underwriting process has increased in recent years,” Stephan Ebbeler, Chief Underwriting Officer at Allianz Partners, told us. “Technologies like machine learning or generative AI allow underwriters to consider a very large number of data points in their assessment. From an underwriting perspective, we can differentiate between the application of AI tools in data preparation and their application in the final pricing/underwriting decision.”

Data preparation, said Ebbeler, is an area particularly ripe to benefit from the adoption of AI-based processes: “Picture recognition tools and large language models can help underwriters to assess risks without manual effort by processing and evaluating unstructured data and information, but based on very standardised assessment. AI models are also being used to standardise data and adjust or correct data errors in the data sets being assessed. As a consequence, underwriters are able to work with a more standardised database for the underwriting process, which provides increased accuracy and better-informed decision-making.”

By implementing AI, underwriters are also able to expand the scope of what they can do. Additional data points can now be integrated that might previously have been out of reach, as it was either impractical or impossible to address such huge volumes of external and public data in a standardised form. “Being able to use this data will help to improve the quality of underwriting and provide more value-adding insurance products for customers,” said Ebbeler.

Finally, AI is also being tested and implemented in the field of pricing algorithms. “Here, the application is rather broad,” Ebbeler said, “from using machine learning models to identify the best generalised linear models [statistical models used to determine premiums] for a specific product in order to find the relationship between different variables – claim cost and age, for example – to using AI algorithms to best price parts of a product. This can simplify the work of the actuary to find the right model, which can sometimes be very time-consuming. However, a good understanding of the applied GBMs [gradient boosting machines – specialised machine learning algorithms used to amalgamate various predictions] is required to ensure that the obtained pricing model is accurate and will lead to fair pricing.”

While leveraging new technology will be essential to the continued evolution and refinement of travel insurance products, adoption of such systems comes with its own risks

Onwards and upwards?

Of course, while leveraging new technology will be essential to the continued evolution and refinement of travel insurance products, adoption of such systems comes with its own risks.

To point to just one high-profile case outside the world of insurance, Google’s previously impeccable search engine has become riddled with AI-produced hallucinations and provably wrong answers – a problem with no clear fix on the horizon.

And while it might be mildly amusing when a search engine is telling someone to use glue to stick the cheese to their pizza, it’s somewhat less so if an unreliable machine is analysing sensitive health data.

Ultimately, the hybrid approach seems the most sensible, and fortunately insurers seem to be approaching the situation with the appropriate level of caution. The human touch will likely remain a core part of the process, and with good reason.

“The application of AI models in the underwriting process presents many opportunities and will only develop further in the future,” said Ebbeler. “These models will help the underwriter to better assess risk, find more accurate pricing and improve the speed of quotations. To fully leverage these advantages, it is necessary to upskill the underwriting and actuarial departments and integrate data scientists and AI experts.”

After all, if anybody is aware of the potential impact of risk, it’s an underwriter.