Through a rigorous matched cross-over study, RGA validated the substantial time-saving potential of DigitalOwl's AI-powered attending physician statement summaries for underwriting, with even greater efficiencies observed for complex, high page count cases.
The insurance industry is on the cusp of a transformative shift driven by the rapid advancement of artificial intelligence (AI) and machine learning technologies.
These tools, including RGA’s industry-leading partnership with DigitalOwl, hold great potential for underwriting, a core function that has traditionally relied heavily on human expertise and manual processes.
DigitalOwl’s APS summary service is designed to reduce the time needed to review potentially lengthy attending physician statements (APS) during underwriting. However, it is challenging to quantify such efficiency gains due to the many factors affecting underwriting times, including case complexity, APS page count, total death benefit, number of underwriting requirements, disparate information, and the underwriter’s training. These factors can overlap on each underwriting case, making it difficult to isolate the time savings attributable to DigitalOwl.
To examine DigitalOwl’s actual efficiencies for APS summaries and maximize control of those other factors, we performed a study using a clinical trial methodology called a “matched cross-over study design.” Here’s how it worked: We selected eight underwriters to form four matched pairs. For each pair, underwriter A performed underwriting on a case using traditional APS without a DigitalOwl summary. Underwriter B performed underwriting on the same case using APS plus a DigitalOwl summary. Each case was underwritten twice. At the mid-point of the study, underwriters A and B switched their methods (the cross-over).
The objectives of this study were:
During the study, eight underwriters underwrote 218 cases on 109 unique facultative cases, recording the time spent from taking up the case to issuing an underwriting decision. While performing underwriting with DigitalOwl, underwriters were instructed to use their own judgment to decide whether partial or full APS review was still needed, or if they could make a decision based solely on a DigitalOwl summary.
The study found an average 19-minute – or 25% – underwriting time reduction with DigitalOwl compared with traditional APS review. A statistical significance testing shows with 93% confidence (or p=0.07) this time reduction is not due to random error (Fig 1).
We also observed that time savings are more concentrated among cases with high page counts (fig 2) and significant impairments that resulted in rated, postponed, or declined underwriting decisions (Fig 3).
We observed significant bifurcated variations by underwriters in terms of DigitalOwl effectiveness. Significant time saving was observed in four underwriters, while no time differences were recorded among the other four underwriters.
The study found some minor discrepancies in underwriting decisions reached with and without DigitalOwl. For example, three cases were “Declined or Postponed” using traditional APS while “Accepted” by underwriting with DigitalOwl. Nine cases “Accepted” using underwriting with DigitalOwl were “Declined or Postponed” with traditional APS. Further case reviews among all 12 of those cases indicated none of the different underwriting decisions were attributed to DigitalOwl per se.
We observed overall efficiency gains with DigitalOwl, with a magnitude of approximately 25% , although, given the small sample size, the difference did not reach statistical significance with P=0.07.
The results are notable and encouraging given the following:
1) Underwriters included in the study on average had only eight weeks of training on utilizing the summaries.
2) No direct guidance was given nor rules employed in how to utilize the tool (or defer to APS). This was solely at underwriter discretion. We believe and early data supports the tool can best be leveraged by incorporating nuance around page count, medical condition, and complexity
3) Cases were drawn exclusively from facultative referrals, known almost by definition to be among the most complex in the industry
4) The study was conducted via traditional PDF-based summary. As a web-based interface is introduced, RGA anticipates material gain efficiencies as underwriters shift from a “scrolling” approach to the more surgical capabilities provided by the online platform.
The study showed significant bifurcated variation in efficiency gains by underwriters. Multiple reasons for this can be hypothesized. This may be due to the fact that DigitalOwl adoption at RGA is at an early stage and underwriters differ in their respective levels of confidence or familiarity with the product. Alternatively, there may be subsets of underwriters who are more change-resistant and still lean on traditional APS records. This suggests that as the DigitalOwl product improves over time, along with more underwriter training inclusive of organizational change management (OCM) techniques, DigitalOwl’s expected efficiency gains will become even more significant.
The study results also support the assumption that use of DigitalOwl summaries impacts underwriting time but not the underwriting decisions.
As insurers continue to embrace AI technologies, they will unlock opportunities to streamline processes, enhance risk assessment capabilities, and deliver tailored solutions that meet the evolving needs of policyholders.
Revolutionize risk assessment with RGA and DigitalOwl. Contact us to learn more.
