Exposing Impediments to Insurance Claims Processing Robert Andrews, Moe Wynn, Arthur H.M. ter Hofstede, Jingxin Xu, Kylie Horton, Paul Taylor, Sue Plunkett-Cole
INTRODUCTION
Case scenario Processing injury-compensation claims, such as compulsory third party (CTP) claims, is complex, as it involves negotiations among multiple parties (e.g., claimants, insurers, law firms, health providers). Despite the relevant legislation mandating milestones for claims processing, the Nominal Defendant sees significant behavioral and performance variations in CTP claims processing, affecting the costs and durations of claims.
SITUATION FACED
Background The Motor Accident Insurance Act (1994) provides overarching framework for establishing and managing compensation claims from persons injured in a motor vehicle accident. The Nominal Defendant, an arm of the Queensland government’s Motor Accident Insurance Commission, determines liability for claims when the vehicle “at fault” is unregistered or unidentified and manages such claims from injured persons. Schematic of the general CTP claims-management process
Challenges The reasons for variations in claim duration and costs are poorly understood. The contention is that contextual factors account for the variations Aims Identify process-related context factors that affect claim duration Investigate process differences for cohorts of claims of interest to the ND
ACTION TAKEN
The Case Study Process identification Domain familiarisation Focus was on process identification, discovery and analysis phases of BPM Lifecycle Context data identification Claims event log preparation Data quality assessment & log cleaning Process discovery Process modelling Context modelling Process Monitoring & Controlling Process analysis Performance Context Applied process mining methodology and techniques Process implementation Process redesign
The Case Study Domain familiarization – Researchers familiarized themselves with CTP legislation, existing process models, process-support documentation Context data set – Guided by literature and interviews with ND staff, a set of process-related context factors were identified for further investigation Claims data set – Historical records of claims processing (finalized between Sep 2012 and Nov 2015 or lodged after Sep 2012 but not yet finalized) Data quality and log cleaning – Raw data for both context and claims analyzed for quality issues and cleaned. Many issues were found including missing values, incorrect timestamps, inconsistent activity names, inclusion of irrelevant events and concept drift Process discovery – Process models in the form of workflow nets were derived Performance analysis - Various performance measures were derived that relate to overall process performance and comparison of the process performance for cohorts of interest to the ND Context analysis – Supervised learning techniques were applied to determine context variables that can be used to distinguish between short & long duration and low & high cost claims
RESULTS ACHIEVED
Claimant has Legal Representation Comparative Performance Time for claims to reach key process milestones Claims Involving Unregistered vs. Claims Involving Unidentified Vehicles 14 process milestones identified where there was significant performance variation Claims involving Unregistered proceeded fast to the point where a liability decision was made However, from this point on, claims involving Unidentified vehicles proceeded faster to finalization Key Activity Differences - Unregistered and Unidentified Vehicles Claimant has Legal Representation vs. Direct Claims Clear differences in the average time taken to reach key milestone events Cohort of direct-claimant claims take less time on average to reach the milestones than does the cohort of legally represented claims Indicates that the process changes that related to direct claimants had a positive effect on claims-management and process outcomes Key Activity Differences - Legally Represented & Direct Claimants
Context Analysis Identifying factors that impact claim duration and costs Key indicative factors include ND engages its own legal counsel Claimant’s employment status Whether claimant engages legal representative Injury severity (particularly whiplash) Was an Independent Medical Examination required Claim Costs Key indicative factors include Was an Independent Medical Examination required Duration – Liability to Settlement Claimant’s age at accident Injury severity Claimant’s pre-accident employment type Claimant’s legal representative has “No win, no fee” policy
LESSONS LEARNED
Take home lessons … Process mining techniques and findings (legally represented vs direct claimant) validated the ND’s process improvement initiatives relating to handling direct claims Key gaps exist in process mining toolset for performing process-performance comparisons across multiple cohorts (of process instances) Currently requires cohort-at-a-time data preparation and analysis followed by a manual comparison Consideration of context factors broadens the scope of process modelling beyond simply uncovering sequences and durations of events Facilitates reasoning about process specifics and provides explanatory (and predictive) capability for process behaviour