Swiss Re AG · SIX:SREN
Swiss Re — Transforming Healthcare (February 2017)
35 slides from Swiss Re's conference presentation, Feb 9, 2017
Slide 1: Predictive analytics and wearables in insurance
Predictive analytics and wearables in insurance 9th February 2017 Transforming Healthcare Daniel Thyssen, Ermir Qeli
Slide 2: Trends transforming the insurance industry
Setting the scene Trends transforming the insurance industry Explosion of data Rapid change in volume consumer behaviour Significant improvements of Proliferation of digital sensing and analytical platforms/market places capabilities Substantial increase in Emergence of agile digital computational p
Slide 3: Technology with three major implications for the insurance industry
Setting the scene Technology with three major implications for the insurance industry Technological advancements (e.g. internet of things, cognitive computing, blockchain) Implications for the industry 1 2 3 Disruption of Change Transformation industry structure in risk pools of value chain (e.g. ne
Slide 4: Impact falls into two main categories
Setting the scene Impact falls into two main categories Evolutionary – Leaner, Better, Faster Existing business processes can be better supported and automation levels of «mechanical» tasks can be increased, using… Text Analytics & Data Capture Advanced Search Robotics Automation Predictive
Slide 5: Smart Analytics @ Swiss Re
Smart Analytics @ Swiss Re Evolution and revolution in the digital world
Slide 6: Predictive Analytics and Machine Learning is extensively used at Swiss Re
Predictive Analytics and Machine Learning is extensively used at Swiss Re in various areas – P&C, Life & Health and Special Lines Applications Global Motor Risk Use Predictive Modelling to predict accident level (frequency and severity) on a Map regional level for motor insurance to enable insights
Slide 7: Accident & Health Cost Claims Forecasting
IMAGERY Accident & Health Cost Claims Forecasting Informing underwriting and reserving decisions with descriptive and predictive analytics using big set of medical treatment data Opportunity Approach Benefit The profitability of the ESL portfolio • Acquisition of five years medical The solution allo
Slide 8: A Case of Big Data Analytics: Context
A Case of Big Data Analytics Context Corso A&H ESL Portfolio Medical market in the US Corso Accident & Health (A&H) provides Strong medical inflation employer stop-loss (ESL) solutions in the US In particular, increase in the frequency of This product is designed to protect catastrophic clai
Slide 9: Large External Claims Database to Support the Analysis
A Case of Big Data Analytics Large External Claims Database to Support the Analysis 5,000,000,000 external consultations 1,700,000,000 drug payments 40,000,000 insureds 5 years of data 2 TB Physician consultations BIG Drug prescriptions DATA Hospital admissions Demographics and employment data Predi
Slide 10: From a Large Dataset to Business Answers
A Case of Big Data Analytics From a Large Dataset to Business Answers Massive Practical data set Data Descriptive Predictive answers for of 2 Preparation Analytics Analytics business terabytes What is the What is the Which key features optimal IT setup to medical inflation have the highest ach
Slide 11: Underwriting Support Solutions
Descriptive Analytics Underwriting Support Solutions Inflation on the Average Total Cost per Patient How does the medical expenses evolve 14% over time? What is the dynamicity in 12% medical inflation? 10% Drugs Various inflation measurements to validate UW 8% Inpatient assumptions 6% Outpatient 4
Slide 12: Treatments Inflation: Low/Mid/High Cost comparison
Trends Analysis Treatments Inflation: Low/Mid/High Cost comparison Clear positive relationship between the 2009 mean treatment cost and the mean inflation observed from 2009 to 2013 High cost treatments do not have the same inflation as low or middle cost treatments. End stage renal disease Aver
Slide 13: Understanding the Medical Risk
Descriptive Analytics Understanding the Medical Risk What are the critical medical complications that can lead to Normal Diabetic Patient High Claim Diabetic Patient Cardiac Arrest very large claims? Anti-Diabetic Insulin Agents Vasodilation A network-based analysis is used to Skin Agents Diabetes
Slide 14: Predictive Analytics: 50% Notifications Definition
Predictive Analytics: 50% Notifications Definition A notification is sent by the prime insurer/employer to Corso whenever the total claims have reached half of the deductible for a given individual (a.k.a. census point) This notification is embedded in a form containing Demographic information
Slide 15: Problem Formulation
Predictive Analytics: 50% Notifications Problem Formulation The 50% notification is translated into a policy year machine-learning problem, where the features match the information of the time line per 50% notification form enrolee notification 31/12 t 01/01 time The time window considered is as
Slide 16: Forecasting Results
Predictive Analytics: 50% Notifications Forecasting Results Achieved results Precision: 66% Recall: 67% F-score: 0.67 Results interpretation: Out of all the 50% notifications coming in, we can successfully flag 67% of all those that will end up exceeding USD 125k at the end of the policy
Slide 17: The Evolution of Health Tech – Wearables
The Evolution of Health Tech – Wearables
Slide 18: Wearable Devices - Examples
Wearable Devices - Examples Resound builds smart hearing aids and accessories for people with mild to severe hearing loss. The ReSound LiNX2, is its Omron's wristband Project second-gen aid that enables Zero is a clinically accurate wearers to control their hearing blood pressure monitor (BPM) via a
Slide 19: We distinguish two basic types of wearables
We distinguish two basic types of wearables Fitness-Grade Medical-Grade Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 21
Slide 20: Nice, but - what for? Biomedical data in the insurance chain
Nice, but - what for? Why should we be interested? Biomedical data in the insurance chain Underwriting In Force Claims • Easier (pre) selection • Ongoing underwriting, • Reduce claims for Life and Illness • Preferred classes possible shifting of class claims (DI, CI) • Selection effect (cf. no claim
Slide 21: e-Health: Swiss Re is pursuing wearable opportunities through partnerships
e-Health: Swiss Re is pursuing wearable opportunities through partnerships with clients and providers Biovotion GOQii Swiss Re’s engagement: Investment Innovation: Medical grade wearable Swiss Re’s engagement: Partnership Innovation: Fitness system of engagement MLC – On Track Swiss Re’s engagement:
Slide 22: The Cases: MLC On-Track – Engaging Customers
The Cases: MLC On-Track – Engaging Customers MOT Project Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 24
Slide 23: The Cases: Biovotion – Medical-Grade Data
The Cases: Biovotion – Medical-Grade Data A Quiet A Working A Day with Sunday Day Challenges Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 25
Slide 24: The Cases: GoQii – Your Personal Health Coach
The Cases: GoQii – Your Personal Health Coach GOQii Ecosystem Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 26
Slide 25: The e2e View – It's Not Just About Data Analytics
The e2e View – It’s Not Just About Data Analytics Monitor & store Smart Analytics The Markets -Prescriptive, -Predictive, -Clustering Engagement The Data The Models Data-driven business Ethics Challenges and Opportunities Insurer/ Reinsurer Glucose Engagement Glucose The Products Ethics Ethics The M
Slide 26: Pitfalls and Successes
Pitfalls • Basis Peak device overheating • Pickup rate @ MOT • “useless data” without personal parameters • Regulatory change sets GoQii project back by 6m Successes • Everbest Challenge: Wearables-based Behavioural Engineering in 2013 at Swiss Re • Pickup rates @ MLC going up following revised enga
Slide 27: Swiss Re leads the way
Vision • Biovotion: 19 sensor signals including blood glucose by 2018 manage diabetes predict / prevent heart attacks • Fluidless Underwriting close protection gaps • Predictive models based on standard fitness data • Prescriptive … insurance becomes lifestyle (and oh, we pay your claims, to
Slide 28: Appendix
Appendix Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 31
Slide 29: The Technology Clusters are elements of broader Technology Themes
The Technology Clusters are elements of broader Technology Themes Technology Technology Cluster Characteristics Engagement Space & Data Exchange Partner driven / external exchange Exchange Platforms Portal / Client Services SR driven / internal exchange Platforms & Clouds SR driven / internal base W
Slide 30: Challenges and Opportunities Described (1/2)
Challenges and Opportunities Described (1/2) Engagement Glucose Ethics The Data The Markets The Models The Products Engagement: One of the key successfactors is getting and keeping individuals interested in interacting with their insurance provider on a regular basis. This is an art of its own and h
Slide 31: Challenges and Opportunities Described (2/2)
Challenges and Opportunities Described (2/2) Engagement Glucose Ethics The Data The Markets The Models The Products The Products: This is what we bring to the market – translating our risk models into something that is of value to the primary insurers and end customers and can be bought and sold. Th
Slide 32: Biovotion VSM-1 Pilot: Circadian Spirals (1/2)
Biovotion VSM-1 Pilot: Circadian Spirals (1/2) A Quiet Sunday A Working Day A Day with Challenges what happened here? heavy lunch and here? a stressful pre- concert practice Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 35
Slide 33: Biovotion VSM-1 Pilot: Circadian Spirals (2/2)
Biovotion VSM-1 Pilot: Circadian Spirals (2/2) 9th November 2016 An Active Saturday What the … ?? news on the US elections the concert a good post-run practice dehydration session + caffeine drank >1L peaceful sleep of water Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healt
Slide 34: Predictive Analytics
Predictive Analytics & Wearables | D. Thyssen, E. Qeli | Transforming Healthcare 37
Slide 35: Legal notice
Legal notice ©2016 Swiss Re. All rights reserved. You are not permitted to create any modifications or derivative works of this presentation or to use it for commercial or other public purposes without the prior written permission of Swiss Re. The information and opinions contained in the presentati