London Market Monitor – 31 December 2021
Learn more on Solvency II discount rates, and the state of the local and global markets at month end.
This is part 1 in a series on body part coding in workers' compensation.
Approximately 15% of claims are coded as multiple body parts, a vague description that does not identify the specific injured body parts. The complexity and risk profile of a claim varies significantly between a finger/hand injury and neck/head injury, both of which would be coded as multiple body parts. Multiple body part claims can be translated into specific combinations of body parts by extracting data from unstructured data (e.g., adjuster notes) using text mining algorithms.
Figure 1: Body part coding accuracy - single vs multiple body parts
Consider the following:
Over 40% of body part coding is inaccurate. Hover over the graphic above to explore coding accuracy of body parts.
Claims that reference multiple body parts in the unstructured data cost significantly more on average. Hover over the graphic below to explore the average severity of multiple body part claims.
Figure 2: Average severity by number of body parts
Text mining identifies the specific body parts underlying a multiple body part claim. Hover over the graphic below to explore the body parts most frequently referenced in multiple body part claims.
Figure 3: Most frequently referenced body parts for multiple body part claims
It is crucial to understand the individual body parts underlying a multiple body part claim. Hover over the graphic below to explore the average severity of the most frequent combinations of body parts in a multiple body part claim.
Figure 4: Average severity by body part combinations
Significant value is lost in the data by coding as multiple body parts. Text mining helps translate body part coding into meaningful data points for adjuster interpretation as well as for predictive analytics.
* All values above are representative of a workers' compensation claim population.