69É«ÇéƬ's Mike Cusumano explains how through cross-functional collaboration insurers can apply data analysis to refine and improve complex case underwriting. Read Part II of this article series to learn how additional and more complex analyses can provide more precise – and more actionable – underwriting insights.
The more complex an applicant’s medical history, the more difficult it is for underwriters to assess mortality risk.
Determining the risk of a single impairment is generally straightforward, allowing for a more traditional numerical rating system. However, assessing an applicant with multiple comorbidities, such as diabetes and heart disease, and evaluating how those conditions affect one another is more complicated. Depending on the combination of impairments and other related factors, the resulting mortality risk may be more or less than the sum of its parts.
Traditional underwriting approaches often prove insufficient for such complex cases. Clinical literature, which tends to focus on narrow populations and specific disease conditions, provides limited value. An insurer may have some of its own mortality experience tied to impairments, but such data is often non-existent or too thin to draw credible conclusions. With at least 25% of the U.S. population having two or more comorbidities and with that percentage on the rise, determining the interrelationships among conditions is critical.1
But there is some good news. Insurers have greater access to third-party data, such as prescription and medical claims histories, than ever before. When linked to mortality outcomes, this data can allow for flexible and customized analysis to derive credible conclusions. Through cross-functional collaboration, insurers have a fresh opportunity to refine and improve complex case underwriting.
Progress starts with a multi-disciplinary approach, including key roles for underwriters, medical directors, actuaries, and data scientists.
The new paradigm revolves around manipulating large, complex datasets and identifying patterns and trends. As a result, the expertise of actuaries and data scientists is essential. At the same time, guidance from risk assessment experts remains critical throughout the process. The role of underwriters and medical directors starts with posing key questions: In what areas might meaningful risk assessment conclusions be derived from the data? What is the most effective way to apply the data to uncover actionable underwriting insights?
All contributors need to communicate clearly and remain aligned around priorities and methodology, lending their own perspectives and expertise from the start. For example, while medical directors bring clinical insights and experience, underwriters understand the real-world application of risk assessment tools and information. Working together, these domain experts can provide a more holistic and even intuitive sense of what may be reasonable to expect from a given dataset. Meanwhile, actuaries and data scientists outline the opportunities and limitations of available data before diving straight into the analysis. This gives underwriters and medical directors an early opportunity and a sound basis to provide guidance and course corrections.
Open communication should remain ongoing through project completion. Once data analysis is complete, results need to be clearly delivered to the domain experts to avoid misunderstandings and confirm alignment with the real-world application defined from the beginning. At this point, underwriters and medical directors can evaluate the results to assess whether the questions initially posed have been answered and whether the findings are reasonable.
The need for clear communications increases as cases become more complex. There are several ways in which a relationship between two or more conditions can manifest itself.
In many cases, the combined risk of two impairments is not equal to the sum of their individual contributions. That is, 2+2 does not always equal 4. Impairments may have an adverse synergistic relationship, resulting in a combined risk that is greater than the sum of the parts (i.e., 2 + 2 > 4). In other cases, the impairments interact in a way that results in lower mortality risk than if the impairments were evaluated separately.
Access to data sources such as prescription and medical claims histories, coupled with the ability to connect that data to mortality outcomes, enables the evaluation of a wide range of comorbidities. The robustness and flexibility of the data further allow for customized analysis targeting the situations most relevant to life insurers.
One such situation is the relationship between diabetes (condition A) and coronary artery disease, or CAD (condition B). The first step in determining this relationship is to define criteria for identifying individuals with relevant risk profiles. What medications or diagnosis codes need to be present in an individual’s history to classify them as likely to have condition A or condition B? Some people will fit the criteria for condition A only (diabetes, but no CAD); others will fit the criteria for condition B only (CAD, but no diabetes); and still, others will fit the criteria for both conditions A and B (the comorbidity of diabetes alongside CAD). Those that we expect to have neither condition serve as the baseline for assessing the relative mortality risk of the other three categories.
The relationship between the two evaluated conditions can be assessed from here. Do the findings suggest that each condition can be assessed independently, without needing to account for the co-existence with the other condition? Or is a more complex relationship present, such as an adverse synergy that results in the comorbid mortality exceeding the sum of each condition’s contribution to excess mortality?
Figure 1 provides a visual example of an 69É«ÇéƬ assessment for diabetes and CAD. Relative to the Neither Condition group, the mortality risk of the Diabetes Only group results in an additional 59 debits, while those with CAD Only have an additional 65 debits. Treating these conditions independently and stacking their debits results in a total of 124. This falls well short of the 185 debits actually observed for the Both Conditions group. This points to a more complex relationship – an adverse synergy between diabetes and CAD resulting in additional mortality beyond each condition’s independent contribution. In other words, 2 + 2 > 4.
Figure 1: CAD and Diabetes
1
The simple but effective analytical approach outlined above provides valuable insights into the relationship between two conditions. However, the true value of mortality data analytics is realized only when insights are effectively communicated to those who can put them into practice – the underwriters. By empowering underwriters to make more informed decisions, cross-functional collaboration delivers a more complete approach to risk assessment.
Read Part II of this article series to learn how additional and more complex analyses can provide more precise – and more actionable – underwriting insights.
The methodology presented in this article focuses on an additive or debit-based, approach to mortality. It is an approach that is intuitive to underwriters, easy to visualize, and generally consistent with how risk is evaluated for individuals applying for a life insurance policy. However, many actuaries and data scientists would instead reflex to a multiplicative assessment. Rather than asking whether 2 + 2 = 4, they would tend to ask whether 2 x 2 = 4.
Both approaches strive to understand the relationship between the actual combined mortality relative to each condition’s independent contribution but can arrive at different conclusions. This is possible even when the inputs are exactly the same – same underlying data, same definitions for conditions, and same mortality outcomes.
It is imperative that, regardless of the analytical approach, communication remains clear. All functions should be on the same page regarding any actionable insights the analysis does, and does not, provide. The application of such insights, including updates to underwriting guidance, must be consistent with the methodology used.
