A new RGA study demonstrates differing mortality insights from LabPiQture and medical claims depending on which medical conditions are present. This provides further evidence that these two underwriting evidences are complementary in mortality risk assessment.
Peanut butter. Cheese. Cookies. Wine. Jelly. Milk. Individually, these six have some appeal. But pair them up correctly – peanut butter and jelly, wine and cheese, cookies and milk – and each is taken to a whole new level.
LabPiQture (LP) and medical claims (MC) data work in much the same way. Useful on their own in underwriting, their full risk assessment potential emerges when combined.
RGA has conducted extensive research to enhance the assessment of mortality impacts associated with digital underwriting evidence (DUE). A recent RGA white paper presents a comparative analysis of the mortality impacts among three primary types of DUE: LP, MC, and EHR. The new findings demonstrate that, while the mortality impact observed with LP and MC individually may not match that of EHR, these data sources are complementary and, when combined, can provide mortality insights comparable to those of EHR in certain circumstances.
This study examined distinctions between LP and MC, exploring factors that may account for their complementary relationship. The aim was to better understand differences in risk differentiation between LP and MC for different medical conditions. The findings from this analysis provide information about the comparative effectiveness of DUE types for specific medical conditions.
The analysis included life insurance applicants from a major carrier, selecting only individuals with both LP and MC data available at underwriting. Because the carrier did not order LP for cases declined by MC, cases declined by either evidence were excluded to reduce bias. The final dataset consisted of 7,527 applicants with both data sources and no record of being declined.
Underwriting rules applied to both LP and MC data classified applicants by risk classes (e.g., “NT super preferred,” “NT standard,” “T preferred,”) with each class assigned a specific relative mortality risk (RR). The RR for NT standard was set as 1. This analysis considered only laboratory test results associated with LP; diagnosis codes related to LP lab tests were not included.
23 medical conditions were identified via application questionnaire:
Among 7,527 cases, the average relative risks (RRs) for LP and MC were 0.897 and 0.894, respectively, indicating that both sources show similar levels of adverse risk. These results are consistent with previous research demonstrating comparable mortality impacts for LP and MC among approved cases.
Figure 1 displays RRs by age and sex. The total LP and MC risks are similar, as previously noted; however, LP risk is higher among males, while MC risk is higher among females. Additionally, LP shows a stronger age dependency compared to MC. Potential factors contributing to these differences include:
Figures 2 and 3 illustrate the mean relative risks (RRs) for LP and MC across 23 self-disclosed medical conditions. For each condition, the average RR for individuals who disclosed their diagnosis was compared to those who did not.
For instance, Figure 2 demonstrates that the RR for LP is substantially higher among participants reporting diabetes than those indicating no diabetes, suggesting that although LP risk assessment may not specifically identify diabetes, it is strongly associated with the condition. Figure 3 reveals that, in most of the 23 conditions, MC RRs were elevated when the conditions were disclosed, indicating a significant correlation between MC risk and self-reported medical issues, notably depression and mental health disorders.
In contrast, only a few conditions showed higher LP RRs for affirmative responses compared to negative ones on the condition questionnaire; diabetes and blood clotting disorders were prominent examples, though only 26 cases met criteria for the latter as reported in Figure 4.
Although the overall risk is similar between MC and LP, the risk identified by LP appears to correlate less strongly with self-reported medical issues, except in cases where laboratory tests are highly relevant – such as diabetes. The comparison between LP and MC demonstrates that different data sources can reveal distinct types of risk. While LP may be more effective in certain conditions and MC in others, this finding underscores that LP and MC function effectively as complementary tools in comprehensive risk assessment.
The findings of this study offer additional support for earlier conclusions, as detailed in a prior white paper, indicating that the mortality effects of MC and LP are similar in scale and generally complementary.
Evidence shows LP is most effective in identifying risk when the applicant discloses diabetes, hypertension, and high cholesterol (Figure 2), while MC works better for cases with self-disclosure of depression, OCD, and other mental health disorders (Figure 3).
A primary strength of this study is its comparative analysis of MC and LP with respect to self-reported medical conditions. The results demonstrate a significant association between MC risk and these conditions. It may suggest lower exclusivity in the protective value of MC. Although this research does not constitute a formal assessment of protective value, the topic warrants more comprehensive investigation.
One limitation of this study is that the LP and MC data were not collected in parallel. The carrier elected to request LP only when MC had not produced a declined decision, preventing direct comparison between LP and MC in cases where MC rendered a decline. To mitigate potential bias, cases in which LP resulted in a declined decision were also excluded from the analysis. As such, findings are limited to comparisons between LP and MC outcomes where neither produced a declined result.
Declined decisions represent risk selection, whereas non-declined decisions pertain to risk classification. Therefore, this study is confined to evaluating LP and MC within the scope of risk classification. Notably, previous research suggests that much of the protective value associated with MC may be realized through effective declines. Accordingly, this study possibly underestimates the total value provided by MC.
When applying underwriting rules to LP data, only laboratory results and procedures were considered, while diagnosis codes are omitted. This method is based on the possibility that diagnosis codes may not be fully adjudicated and could include differential diagnoses. As a result, this approach may not reflect the complete value of LP.
Finally, just as the flavor of a fine wine may be enhanced or diminished by the cheese with which it is paired, care must be taken by each carrier to ensure their specific use of MC and LP is a strong match for their unique products, market segment and business goals. The results of this study may not be representative of what any other carrier may experience, and fine tuning the use of MC and LP – in conjunction with other underwriting evidence – is key to a successful underwriting program.
This work highlights critical insights from RGA’s ongoing research on digital underwriting evidence (DUE). It aims to provide carriers with a deeper understanding of the value of emerging DUE, which is expected to play an essential role in the future of life insurance underwriting.
Contact us today to learn more about how studies such as these can help your business better incorporate DUE into its underwriting processes.
