9/14/2018 1 SDM v.2018: Evolution of service delivery models in genetic counseling to meet the increasing demands Stephanie A Cohen, MS, LCGC Disclosures • None
9/14/2018
1
SDM v.2018: Evolution of service delivery models in genetic counseling to
meet the increasing demands
Stephanie A Cohen, MS, LCGC
Disclosures
• None
9/14/2018
2
• Recognize service delivery models and how they have changed over the past 7 years
• Explore motivations for implementing an alternate service delivery model
• Describe outcomes measures to determine the effectiveness of SDMs
Learning Objectives
Trends in genetic testing
• Increased demand
– Media attention
– DTC marketing
– Increased public awareness
– Guidelines and recommendations
– Insurance requirements
– Test utilization management
• Strain on resources
– Access to trained genetics professionals
– Cost
– Time
9/14/2018
3
Genetic Counseling: Why do we need alternate service delivery models?
• Improve access
– Geographic barriers
– SES, linguistic and cultural
• Keep up with increased demand and availability of genetic tests
– Example: pharmacogenomics
• Genetics professionals do not have time or resources to see every patient
Guttmacher, Jenkins and Uhlmann. “Genomic Medicine: Who Will Practice It? A Call to Open Arms”. American Journal of Medical Genetics (2001) 106:216-222.
New models• Efficient
• Financially viable
• Establish competencies
• Integrated into clinical practice
• Interdisciplinary
• Outcomes‐based
• “Genetically literate” workforceGuttmacher, Jenkins and Uhlmann. “Genomic Medicine: Who Will Practice It? A Call to Open Arms”. American Journal of Medical Genetics (2001) 106:216-222.
9/14/2018
4
Cohen, S., Gustafson, S., Marvin, M., Riley, B., Uhlmann, W., Liebers, S., et al. (2012). Report from the National Society of Genetic Counselors Service Delivery Model Task Force: A Proposal to Define Models, Components, and Modes of Referral. Journal of Genetic Counseling, 21(5), 645-651.
Service Delivery Models
2010 SDM survey
2316 NSGC members invited
820 responses (35.4%)
715 clinical care
701 useable responses
Cohen, S. A., Marvin, M. L., Riley, B. D., Vig, H. S., Rousseau, J. A., & Gustafson, S. L. (2013). Identification of genetic counseling service delivery models in practice: a report from the NSGC Service Delivery Model Task Force. J Genet Couns, 22(4), 411‐421. doi:10.1007/s10897‐013‐9588‐0
9/14/2018
5
2010 SDM use
45.3% use more than 1 SDM
2010 SDM study overview• Those using an in‐person SDM reported the ability to see the
highest number of patients per week (p<0.0001) and were the most likely to bill in some manner (p<0.0001).
• Those using telephone GC had the shortest wait time (77% <1 week, p<0.0001) and shortest visit time (53%<30 minutes, p<0.0001), but least likely to bill (67% do not bill, p<0.0001)
• Those using telegenetic and telephone GC served patients who lived the furthest away, with 48.3 % and 35.8 %% respectively providing GC to patients who live >4 h away.
9/14/2018
6
2017 SDM survey responses
3616 NSGC members invited
590 responses (16.3%)
553 clinical care
517 useable responses
Respondent demographics
29
36
14
7
3
11
0
5
10
15
20
25
30
35
40
45
Cancer Prenatal Pediatrics General Cardiac Mulitple/noone particular
Other
Specialty (%)
2010 (N=556)
2017 (N=395)
9/14/2018
7
Work setting (%)
0
5
10
15
20
25
30
35
40
45
50
2010 2017
Years experience (%)
0
5
10
15
20
25
30
35
40
45
<5 years 5‐10 years 11‐15 years >15 years
2010 2017
9/14/2018
8
SDM “always” or “often” (%)
0
10
20
30
40
50
60
70
80
90
100
In‐person Telegenetics Telephone Group
2010 2017
P<0.01
P<0.05
NSP<0.05
In‐person wait times (%)
0
5
10
15
20
25
30
35
40
<1 week 1‐2 weeks 2‐4 weeks 1‐2 months 2‐4 months >4 months
2010 2017
p<0.01
9/14/2018
9
Telephone wait times (%)
0
10
20
30
40
50
60
70
80
90
<1 week 1‐2 weeks 2‐4 weeks 1‐2 months >2 months
2010 2017
p< 0.01
Telegenetics wait times (%)
0
5
10
15
20
25
30
<1 week 1‐2 weeks 2‐4 weeks 1‐2 months >2 months
2010 2017
p=0.57
9/14/2018
10
Number of patients seen in‐person
0
5
10
15
20
25
30
35
1 to 5 6 to 10 11 to 15 16‐20 21‐30 >30
2010 2017
p‐value = 0.145
No SDM reported ability to see more patients in 2017 than in 2010
Telegenetic patient distance (% most or all)
0
5
10
15
20
25
30
35
40
45
50
<30 min 30‐60 min 1‐2 hrs 2‐4 hrs > 4hrs
2010 2017
9/14/2018
11
2017 SDM study overview• Those using an In‐person SDM still see the most patients per
week (30% 6‐10 patients, 27.5% see 11‐15 patients/week) (p<0.01) and are still the most likely to bill in some manner (p<0.01).
• Those using telephone GC still have the shortest wait time (62.4% <1 week, p<0.01) and the shortest visit time (49%<30 minutes, p<0.01) and were still the least likely to bill (67% do not bill, p<0.01)
• Those using telegenetic and telephone GC served patients who lived the furthest away, with 45.3% and 39.4 % respectively providing GC to some/most/all patients who live >4 h away
Have we improved access??
• Overall seeing more use of TG and telephone GC, but no increase in the number of patients seen or billing practices
• Wait times for in‐person and telephone GC have increased
• TG now appears to be used for patients who live near and far
9/14/2018
12
Implementation of new SDMs
• 54.4% of respondents indicated their current model of service delivery was not adequate to address the need in their area.
• 64.8% indicated they were either in the process of or planning to make changes to their SDM
• 74% identified barriers to implementation
Qualitative analysis
Why is your current model inadequate? (N=173)
• Lack of geographic
access
• Lack of physician availability
Barriers to implementation new SDM (N=151)
• Lack of support
• Quality concerns
• Funding
• Technology issues
• Lack of physical space
•Lack of staffing
•Billing & licensure
9/14/2018
13
Outcomes measures in SMDs
• Access
– Geographic distance
– Timeliness (3rd next available)
– Number of patients
Other quantitative measures
• Referral rate
• Efficiency (time spent by HCP, eg)
• Average cost/visit
• Adverse events
• Knowledge/understanding
• Sustainability (billing)
9/14/2018
14
Qualitative measures
• Satisfaction (Patient, Provider, Referring HCP)
• Adherence
• Anxiety/comfort
• Patient empowerment
Innovative SDMs
• Driven by need to improve access
• Which is more important – SDM or efficiency?
• What outcomes are we trying to achieve?
9/14/2018
15
Thank you!!
Exploring the Alternatives:Evolving Genetic Counseling Service Delivery Models and
Impact on Our Practice
Tara Schmidlen, MS, LGC
Clinical Investigator/Genetic CounselorGeisinger
9/14/2018
16
Disclosures
• I am a paid employee of Geisinger, a non-profit integrated health system that serves more than 3 million residents in Pennsylvania and in southern New Jersey.
• Geisinger pays Clear Genetics, Inc, a healthcare technology company based in San Francisco, CA to develop chatbots.
• Geisinger staff members work with Clear Genetics staff members to collaboratively develop chatbots for use with patients enrolled in our MyCode® Community Health Initiative.
• MyCode® is funded by Regeneron Pharmaceuticals
• I am not paid by Clear Genetics. (Occasionally they do re-tweet me)
Learning Objectives• Recognize routine patient communication
opportunities that lend themselves to the use of innovative alternative service delivery models.
• Illustrate the use of chatbots to address common ancillary tasks typically conducted by genetic counselors and genetic counseling and/or research assistants
9/14/2018
17
We’ve got a problem…
Alternative Service Delivery Models• Phone counseling
• Telemedicine
• Group counseling
• Web portals, online educational materials
• Videos for pre-test education
• Post-test only genetic counseling
• Training other providers to provide basic counseling
• Genetic counseling assistants
• Chatbots!
9/14/2018
18
What is a chatbot?• Chatbots are a technology-based simulated conversation
tool used in scaling communications.
• Chatbots can:
• Answer simple questions
• Increase and maintain consumer engagement
• Promote products and services
• Provide convenient, easy access between consumers and service providers
• Chatbots are used in many settings
• Banking and insurance industry
• Retail and service
• Travel-airlines and hotels
• Health care providers
Why Chatbots?• Deploy by link, no app
needed
• Phones, tablets, desktop PCs
• Personalized to the patient
• Back end analytics allow for seeing what, when and how patients interact with the bots
• EPIC integration/interfacing ability
• Scheduling visits, sending kits
• Chat transcript in EHR encounter
• Many potential use cases!
• Free up GCs for higher level, and billable patient care!
9/14/2018
19
• Clear Genetics, Inc. is a healthcare technology company based in San Francisco, CA
• Clear Genetics combines the knowledge of experts in the field of genetics with artificial intelligence to create data-driven, HIPAA compliant chatbots that:
• Goal: Enroll and sequence 250,000 Geisinger patients into the MyCode® Community Health Initiative
• Research Objective: Improve our ability to predict and prevent disease using genetic information
• Requirements: Geisinger patient, blood samples and EHR access
• Returning Results for Actionable Genes:
• Genes known to be associated with increased risks for disease (ACMGv2 +CF, HFE, heritable cancers, heritable heart disease)
• Pathogenic or likely pathogenic variants
• Diseases with established methods for prevention or early detection
• ~3.5-4% will receive a result
9/14/2018
20
Top 3 Conditions Reported
Patient Communication Workflow• Patients consent to participate in-person or online
• Patient-participant notified of results by phone
• Packet containing result report, educational materials, result sharing family letter is mailed
• Free genetic counseling is offered
• Patient-participants are connected to healthcare providers for screening and risk management
• 1 month follow up and 6 month follow up calls• Receive results? Gather family history? Meet with provider?
Share with relatives? Initiate screening/management?
9/14/2018
21
Time Study-Current Approach• Average time for 1 month follow up call
– Patient doesn’t answer: 5 mins
– Patient does answer: 10 mins
• 3 attempts made per patient
• Minimum effort required: 877 patients with a result x 10 mins=8,770 mins=146 hours of labor=3.65 weeks of full time effort (assuming reach everyone on 1st call!)
• Average time for Cascade preparation
– 18 minutes per patient to generate results packet mailing
– 5 minutes Cascade letter generation
– Minimum effort required: 877 patients with a result x 5 mins-4,385 mins=73 hours of labor=1.82 weeks of full time effort
Chatbots for Scaling Genomic Counseling
GIAGenetic Information Assistant
Introducing:
9/14/2018
22
Consent Chatbot
Consent Chatbot
9/14/2018
23
Patient Follow-Up Chatbot
Family Sharing Tool
9/14/2018
24
Family Sharing Tool
Cascade Chatbot
9/14/2018
25
Smart FAQ
9/14/2018
26
Pilot Testing• Initial feedback on high-fidelity prototypes collected
via email
• Pilot testing conducted via usertesting.com
• Verbal response to questions assessing functionality, acceptability and understanding were collected
• Participants were prompted with open-ended questions on:
• Purpose of the chat
• Intended recipient
• Chat partner (person/computer)
• Personality
• Clarity and intuitiveness
• Preferences
• Suggested improvements
Pilot Results• Participants describing
correct purpose and target audience
• 12 participants (n=8 cascade, n=4 follow-up)
• Participants rating chatbots “easy and intuitive” to use
• n=12
• Participants rating the chatbot as a positive interaction
• n=12
• Most common descriptors of chatbot personality
• “friendly” (n=4) “professional” (n=4)
“comfortable”
“This sort of news is better delivered by humans.”
“Where did (relative) give them my personal information? Trust is major
for divulging medical history. I want to be sure I'm talking to someone
legitimate.”
“With family it’s sometimes very emotional. With this (tool) you
immediately get right information from the experts. You have a nice
knowledge base and potentially calm nerves.”
9/14/2018
27
• Suggested improvements:
• Introduce chatbots by email
• Add more encouragement to see a doctor
• Use less jargon
• Add links/information on genes
• Add links/information on cancer risks without a variant.
• Cascade and follow-up chatbots might be an acceptable, user-friendly mechanism to perform common tasks needed in the delivery of scalable genomic counseling
Pilot Results
Chatbot User Testing• Patient Focus Groups (n=62)
• 3 representative areas of patient population
• Consent (n=33), Patient Follow Up and Cascade Chatbots (n=29)
• Mechanical Turk (n=203)
• Consent chatbot
• Comprehension
• Validately Unmoderated Online User Testing (n=DISASTER)
• Audio & video capture of participants using chatbots and talking aloud in response to prompts
• Comprehension
• Implementation Preference Data
9/14/2018
28
Focus Group Testing• To gauge patient interest and improve chatbots prior to further
testing/deployment
• 3 In-Person Focus Groups for Consent Chatbot
• 3 In-Person Focus Groups for Patient Follow-Up and Cascade
• Participants viewed/interacted directly with chatbot & provided general feedback
• Moderator asked targeted questions to assess:
• Usability
• Functionality
• Acceptability
• Understanding
Participant Demographics
Consent GSCGender
Male 10 (30%) 10 (34%)Female 23 (70%) 19 (66%)
Age18-35 1 (3%) 1 (3%)36-55 6 (18%) 2 (7%)56-75 21 (64%) 20 (70%)76+ 5 (15%) 6 (20%)
RaceCaucasian 31 (94%) 27 (93%)Mixed/Non-C 2 (6%) 2 (7%)
Consent GSCEducation
Grades 9 -11 0 1 (3%)Grade 12/GED 6 (18.2%) 7 (24%)1-3 yrs after HS 7 (21.2%) 8 (28%)College 4+ years 11 (33.3%) 5 (17%)Advanced degree 9 (27.3%) 8 (28%)
Employment StatusEmployed 10 (30%) 8 (28%)Unemployed 0 1 (3%)Retired 22 (67%) 16 (55%)Unable to Work 1 (3%) 4 (14%)
9/14/2018
29
Participant DemographicsConsent GSC
Do you know what a chatbot is?Yes 5 (15%) 5 (17%)No 15 (45%) 18 (62%)I'm not sure 13 (40%) 6 (21%)
Have you ever used a chatbot?Yes 2 (6%) 3 (10%)No 23 (70%) 17 (59%)I'm not sure 8 (24%) 9 (31%)
Typical reaction to new technology:I am typically eager to use new technology 17 (52%) 12 (41%)I am typically overwhelmed by new technology 4 (12%) 6 (21%)It depends… 12 (36%) 10 (35%)
Focus Group Results-Consent
“…if you're handed a piece of paper, and you can ignore it,
and you just read it and sign it, and you're not even sure what you signed, then later on are you going to come back and
say, "I didn't realize this is what I was signing up for." Where
this at least makes you look at the really pertinent information
so that you are providing informed consent instead of
just consent.”
Key Takeaways:
• Chatbot much more informative than the in-person consent experience
• Liked the customizable length of the chat
• Liked the ability to type in questions
• Keep in-person consent an option
• Have consenters hand out cards with a link to the chatbot for people in a hurry
9/14/2018
30
Focus Group Results-Cascade“I think this is very good
because of the information and the
answers to the questions that they're going to ask. When you make the call, you're not going to have all that at the tip of your tongue. It's good to tell them, but it's also good
to have answers because everybody's
going to start asking you questions”
Key Takeaways:• Most would talk to relative before sending
chatbot link• Different approaches for different
relatives-chat vs. letter vs. call• Liked the pre-view function in FST to see
what their relatives would see if sent the chatbot
• Send the links to proband through MyGeisinger
“I think this is good because, it gives factual information,
instead of people being out on the internet, researching it themselves, asking their
neighbor. Here is someone who can answer the questions
accurately for them. I think it's a great tool.”
Post-Focus Group Iterations
Consent
• Add summary re-cap
• Add gratitude expression
• Add more info about blood draws needed
Cascade/Follow Up
• Emphasize HIPAA secure link at beginning
• Add logo/hospital icon on landing page
• Add more risk info-LRE and non-genetic risk
• Edit function on as default
9/14/2018
31
Consent Mechanical Turk Data• Comprehension of key concepts: 14-item multiple choice
quiz
• Link to chatbot and quiz on Amazon’s Mechanical Turk
• Restricted demographics: age, income, education of Geisinger patients.
• Round 1: 101 users (Group A), results were reviewed and the chatbot was updated to address questions with average correct scores of ≤75%.
• Round 2: 102 users (Group B) tested the updated chatbot.
• Total average comprehension score for all 14 items was high (80% Group A, 85% Group B).
• Group A scored <75% correct on 6 questions:
– type and timing of results returned
– purpose of the chatbot
– access to health records.
• A summary recap was added at the end of the chatbot.
• Group B reviewed summary recap chatbot and per-question average scores improved to >75% for all but one question.
• The summary recap resulted in significantly improved comprehension (p=0.003) of key elements of the consent.
Consent Mechanical Turk Data
9/14/2018
32
Chatbot Implementation• Preferences for Patient Follow-up and Family Sharing Tool/Cascade
Chatbot were collected at time of result disclosure calls beginning in August 2018:
– 64 completed disclosure calls
• Among those who were reached, 40/64 (63%) were asked about preferences for receiving electronic communications:
– 25/40 (63%) accepted, 15/40 (37%) declined
– 22/40 agreed to cascade (55%)
• 22/25 (88%)-agreed to 1 month and cascade
• 3/25 (12%) agreed to 1 month and declined cascade
– 15/25 (60%) MyG, 8/25 (32%)-email, 2/25 (8%)-text
• Consent-pending EPIC integration
Implementation Workflow
9/14/2018
33
Next Steps• EPIC EHR integration
• Tracking Cascade uptake
– Can we meet/exceed 10% rate with letter?
• Tracking Patient Follow Up and Outcomes
– Bot interactions vs. engaged in recommended behaviors
• Interview patients/relatives who have used chatbots
• Further Iteration & Development of Other Chatbots
– 6 month follow up
9/14/2018
34
Tracking OutcomesData Collected in Clinic Hub
• Whether the patient opens FST
• How patient interacts with the tool
• Whether patient shares & how (email, text, FB, copy link)
• How many relatives open the chat and how many times they open it
• How the relatives interact with the chat – what content they see and how they respond to each message
• Whether the relatives plan to get counseling, testing, etc.
Not Collected in Clinic Hub
• How many relatives the patient shares with
• Whether the relatives actually go on to get testing
Scaling Genomic Medicine With Chatbots
a bot
9/14/2018
35
• Streamline healthy population screening• Increase identification of at-risk relatives• Triage of patients at point-of-care (GI, OB) based on NCCN or
custom criteria• Provide pretest education for patients with family history of cancer• Consent patients to genetic testing• Guide patients using widget on website
Other Use Cases for Chatbots
Acknowledgements•Geisinger Genomic Medicine Team
• Amy C. Sturm, M.S.
• Marci L.B. Schwartz, Sc.M.
• Adam H. Buchanan, M.S., M.P.H.
• Janet L. Williams, M.S.
• Alanna K. Rahm, Ph.D., M.S.
• W. Andy Faucett, M.S.
• Amanda Lazzeri, B.S.
• Lauren Frisbie, B.S.
• Cara McCormick, M.P.H.
• Christa L. Martin, Ph.D.
• Marc S. Williams, M.D.
• David H. Ledbetter, Ph.D.
Clear Genetics Team
• Moran Snir, M.Sc., M.B.A.
• Guy Snir, B.S.,M.B.A.
• Emilie Simmons, M.S.
• GIA
Geisinger Bioethics Team
• Jennifer K. Wagner, J.D., Ph.D.
• Michelle N. Meyer, Ph.D., J.D.