Skip to main content

This article was first published in the March 2022 issue of On The Risk, the Journal of the Academy of Life Underwriting.

Underwriting manuals have been around for a long time. Originally, they were large that were kept on the underwriter’s desk. They evolved through floppy disk and CDs to the sophisticated e-tools of today live in a cloud somewhere. At their most basic they are a means to serve up multiple documents that record the firm’s underwriting philosophy, but these days they typically go far beyond that. In this article we turn our thoughts to the trends and issues that will influence the development of the underwriting manual of the future – if of course you believe they have a future.

Some of the things that might impact the underwriting manual of the future include:

  • Greater complexity of rating tables and a need for better navigation
  • Increased use of calculators and worksheet wizards
  • More personalization of both navigation and content, including the ability of local administrators to edit content
  • Integration of the manual into other underwriting tools such as workbenches
  • The use of natural language processing and text mining of unstructured data to interpret or summarize electronic health records (EHRs) and attending physician’s statement (APS) information
  • The use of artificial intelligence and machine learning.

In this article we will focus our attention on the first two.

Complexity

The trend is for ratings in manuals to become more complex owing to a desire for greater precision in the ratings for better differentiation between good and bad risks, and the identification, through research, of new medical risk factors that need to be taken into account. Essentially, the rating pages keep getting longer as the number of additional risk factors grows and the risk classification method becomes more complicated.

So underwriters already under pressure to turn cases round more and more quickly will have less time to familiarize themselves with the contents of the manual and less time to slot cases correctly into complex classifications. Arguably, the risk of making errors increases. The manual of the future will have to employ smart design and navigation to increase productivity and maybe reduce the potential for mistakes.

To some degree, calculators and ‘flow chart’-type rating wizards aim to address these issues. They ensure that decisions on complex cases (such as cancer) cannot be underwritten unless all of the critical risk information has been put in. However, these tools can sometimes frustrate senior underwriters who feel that they are not able to use their judgment and experience: calculators are often black and white while underwriting cases can contain an element of grey. Mental health underwriting might be a good case in point, although we do hear that mental health calculators may be introduced to reduce subjectivity and ensure adherence to best practice. Greater consistency should of course be an ever more important objective in these days of increasing regulation.

Underwriting and calculators

On our travels we speak to a lot of people about underwriting manual developments. One topic which comes up time and again is calculators: great invention or the work of the devil?

Some underwriting manuals have had simple calculators for many years, for example for pulmonary function tests, financial ratios and BMI, all of which are designed to help the underwriter by speeding up and simplifying assessment. In recent years calculators have increased in number and complexity. There are calculators for overall cardiovascular risk, coronary heart disease itself, breast and skin cancer, liver function tests, and even occupation risks.

Calculators have a number of advantages:

  • They perform calculations quickly and present ratings without the need to refer to rating tables.
  • They ensure that the underwriting philosophy is implemented consistently by each underwriter on every case that shares the same risk features and across a company with multiple underwriters and multiple sites.
  • In theory it should be easier to defend ratings that are challenged where the decision can be tied to an evidence base and there is less influence of underwriter discretion.
  • Calculators can ensure that important steps in the decision-making process are not missed, such as in the assessment of cancer cases.
  • Where a ‘global’ manual is used in multiple markets, a calculator can be customized to ensure that outcomes are tailored to individual market conditions in terms of disease prevalence and local pricing.

On the other hand, calculators do attract some understandable criticism:

  • Usually they don’t display the ranges or steps in ratings so that an individual who has a BMI that is right on one of a rating range border gets the ‘book’ rating without leaving the underwriter some say in the matter. In contrast to a rating table, calculators only deal with ‘black and white’ and not ‘shades of grey’. When a rating threshold is crossed it is crossed – and that’s it.
  • Underwriters may learn and build up real experience more slowly because the calculator does the thinking for them. And of course, calculators don’t learn either – well not yet anyway.
  • Some calculators deal with risk factors in isolation and don’t account for the interactions between them. Others may give the impression that they do, but in reality just add up the numbers, misleading users into thinking that the calculator is carrying out the full risk evaluation process.
  • Calculators can produce quite inappropriate answers if incorrect data has been keyed in – garbage in, garbage out as the saying goes.
  • Because their workings are unexplained, calculators produce answers in isolation; users don’t know how they were derived and thus cannot question them – assuming he or she were so inclined (see below).
  • Experienced underwriters have learned over the years what is significant and what is not. Calculators may not be able to deal with these subtleties of underwriting (although the counter-argument is that the ‘subtleties’ are subjective, are derived from dubious traditional wisdom and myth, and thus have no place in modern risk assessment).
  • Related to the points above, underwriters are discouraged from thinking about the individual case before them and tend merely to apply whatever the manual says.

Some calculators, especially those dealing with multiple or cardiovascular risk factors, have become increasingly complex; we suggest that the majority of the end-users don’t understand how they work and how they deal with complex interactions between risk factors. In fact, the workings are probably understood only by those who devised them… Furthermore, the algorithms may be regarded as ‘proprietary’ and thus are not made public, so where the calculator stops and where old-fashioned experience and judgment come in are far from obvious. How many users believe that the calculator is always right and never needs to be questioned?

Calculators vary in sophistication. Some algorithms are based on complex interactions between one or more risk factors. Others use simple risk factor counting, with perhaps individual consideration being required when there are four or more factors in play. Some will give decisions for some conditions only when further test are available, for example offering critical illness terms for diabetics only when urinalysis, GFR or cholesterol readings are known.

Cancer calculators tend to be more a process for simplifying decision-making in complex situations and ensuring that pieces of information or vital steps are not omitted.

Whatever their limitations, we believe calculators are here to stay. Underwriters are under increasing pressure to get through more and more cases, and they need tools that help them do so. With a lack of experienced underwriters and less training taking place, perhaps there are not that many underwriters now that are still able to apply ‘old school’ judgment and experience (not that this can be relied on 100%, maybe). However, underwriters do need to remember to engage their brains before using calculators – it’s what they are paid for.

And maybe producers of underwriting manuals need to explain how their calculator algorithms work and what is included so that users don’t have to make assumptions that might be wrong.

Interaction with other underwriting technology

In the modern world a significant percentage of business will be processed by underwriting engines, and with automation comes greater consistency. An underwriter’s view of a case these days will likely not be from a paper file but via some sort of workbench with assorted information being brought together to present a ‘one-stop shop’ of the case details such as:

  • Applicant disclosures
  • Lab test results and physical measurements
  • External data such as prescription histories
  • APS information or summarized APS information
  • Mortality scores or other data from external sources.

The impact of this is that the manual process needs to become slicker, and we believe this will result in the need to integrate the functionality of the underwriting manual and the workbench.

Having this capability would appear to make a lot of sense but the integration could be on a number or levels. Possibilities include:

  • A simple link to the manual with single sign-on
  • Deep linking to the relevant pages of the manual in relation to the medical conditions being underwritten
  • Automated presentation of the specific section of the manual based on the information provided to the workbench; where there are multiple risks this may enable a holistic view of the case
  • Deeper analysis of the various sources of underwriting information referenced by the manual to suggest to the underwriter a case-specific decision; ultimately it may reference decisions made on similar cases assessed previously.

In summary we believe that technology trends will influence the next generation of underwriting manuals – which won’t be disappearing any time soon.