Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Member Risk Model: A
method to reduce over
utilisation by predicting
high cost claimants
Valentina Rapuano
Analytics Manager
Bupa Healthcare Analytics
Tuesday 20 September, 2016
Bupa Private and Confidential
Introduction
2
Within the private medical insurance industry,
there is growing concern that private
healthcare provisions may encourage over use
of medical services.
Is the efficiency and ease of access
associated with private care responsible for
over-diagnosis that results in excessive or
inappropriate medical treatment?
Bupa Private and Confidential
Care management initiatives
3
Case Management
Condition Management &
SDM
Supported Self-care Management
Prevention & Wellness Promotion
Our interest in a member risk model is related to suitably identifying members
most likely to incur high cost claims, who would benefit from targeted care
management.
Bupa Private and Confidential
Case study: Complex Case Management
4
Key features
• Complex care nurse delivers face-to-face member outreach and tailored support.
• Care assessment provides overall view of healthcare needs, coordinating care with healthcare professionals.
• Members identified in the top 0.5% of predicted claims spend, with two or more long-term conditions.
Desired outcomes
• Improvement of health outcomes and experience of members with complex care needs.
• Preventing unnecessary healthcare use and avoiding adverse affects which might otherwise occur due to uncoordinated healthcare delivery.
Bupa Private and Confidential
Exploratory analyses
5
Testing the existing model’s performance highlighted poor prediction of 12-month
claims costs for our high cost claimants.
Ave
rage
an
nu
al c
laim
s sp
end
9,000
8,000
7,000
6,000
5,000
4,000
3,000
2,000
1,000
0
Predicted spend
Actual spend
Bupa Private and Confidential
Review of existing Member Risk Model
6
Main objectives
• Recalibrate model inputs to represent current member population and coding
structures.
• Address model discrepancies particularly for those members with no claiming
history.
Desired outcomes
• Provide updated prediction for members’ 12-month
claims plus view on short-term risk via 3-month
claims prediction.
• Being able to identify members with a short-term
risk means that interventions can be tailored to the
member and implemented with immediate affect.
Bupa Private and Confidential
Challenges and approaches
7
Learning from previous work
Improved data inputs
Flexibility in predictions
Ensemble modelling
Performance analyses
Resource allocation
Understanding the data
Bupa Private and Confidential
Data structuring
8
Pre-authorisation Data Variables For a subset of treatment areas, a measure of “time since last occurrence” are used and are supplemented by an indicator of expected claims for members with no claims history.
Specialty Variables Building on MSK specific classifications incorporating body part and procedures, details relating to the medical specialty of the lead consultant are used as an indicator of members’ medical profile.
Demographic Based Variables Basic demographics such as age and gender extended to include details of membership, relationship status and a London treatment indicator, to provide richer understanding of member base.
Clinical Based Variables Clinical profile including chronic conditions and types of cancer enhanced with indicators against specified pre-existing conditions, to provide insight into potential claim areas for new members.
Bupa Private and Confidential
Project summary: Phases of development
9
Exploratory analyses
Data structuring
Methodology Model
validation
1 month 5 months 14 weeks 2 months
QUESTIONS?
Bupa Private and Confidential
Contact Details
11
If you have any further questions, please feel free to contact me:
Valentina Rapuano
+44(0)20 7656 3675
THANK YOU