2. 2
What is a Virtual Assistant?
A Virtual Assistant is designed to aid clinician’s manage the wellness
of patients.
The Virtual Assistant is a natural language ‘chat bot’ recognising and processing structure voice input.
The aim is to eliminate routine tasks to save time and increase efficiency.
The Virtual Assistant takes on manual repetitive tasks. Through structured conversations with patients it collects data clinical teams require
for signing off on discharges or providing updates around medication compliance. Uses include checking the health of patients who’s routine
care is ongoing without the need for clinical or administrative time or collecting PROM scores for all surgical patients; whilst at the same time
highlighting through the clinical dashboard patients who require support.
It extends care into homes by complementing existing tele-health initiatives, offering healthcare contact that otherwise wouldn’t be available to
patients with the data collected through the normal telephone network.
Where has it worked before?
This system has been implemented and utilised in South East Asia specifically in China.
For the past 2 years, across all of parts of China, the virtual assistant has led to :
- Over 12,000 calls per week in one hospital alone
- 98% of all calls have resulted in Patients completing the Virtual Assistant questionnaires
The UK case studies are to include
• Capturing wellness post operation to spot early infections which could lead to SEPSIS and prevent readmissions
• Preventing mortality from secondary Heart failure
• Reminding people with long term conditions about the need for good medication compliance.
3. 3Deloitte Confidential – For Internal Use Only
Data Exchange –
from ePR to VA
Soft Switch &
Symatic Language
Dialogue flow
Virtual
Agent
Agent
Assist
Knowledge Base
(Prodigy)
Result Stored in
Virtual Assistant
Nurse
Desktop
PSTN Network
Virtual Assistant
(VA)
4. Nurse listens
to transcript,
arranges
followup care
for patient.
acknowledges
response in VA
Normal
Responses
Intervention
Identified
Hospital ePR
System
Patient
Discharge
Information
Data Exchange –
from VA to GP
6. Txt file of patient
transcript sent to GP
system
HL7 Data
Input
1. Patient agrees to a
call when
discharged home
from Hospital
Explict consent given by
Patient for enrolment in
Virtual Assistant.
2. Patient
recieves phone
call at home
from Virtual
Assistant
5. Struggling Patients recieve
followup phone call at home
from Nurse
Solution overview
The Virtual Assistant uses NVIDA hardware for voice recognition
allowing the solution to be hosted within the customers own secure
network.
Deloitte
developed
interface
Deloitte
developed
interface
3. Patient
authentication &
question/answers
processed
SIP relay
4. 4
Why use this solution?
The Virtual Assistant is driven by the US NVIDIA DRIVE platform to create an natural voice recognition system with more than a million
words in natural language embedded. This library is swelled by Prodigy, a clinical advice and guidance NICE database unique to this
product. The NVIDIA GPU processors are available through a white label license and are very fast and scalable allowing the Virtual
Assistant to be available and made secure on the edge of the organisations secure network.
The white-label licensed service allows organisations to personalise the voice assistants using their own voice, enabling clinicians to build
their voice assistant brand and tap into other use case opportunities that can emerge.
As the data is retained by the organisation a clinical dashboard is available for any patient interaction providing clinicians with the ability to
listen and review patients conversations in the Virtual Assistant at the same time extracting the text of the conversations for integration into
local patient record system providing a further layer checks and validation.
Amazon, by contrast, licenses Alexa and customers must call it “Alexa” in queries and owns the customer data which cannot be extracted.
Apple doesn’t license its Siri voice assistant and Google doesn’t allow people to customise its Google Assistant name or own the data
created by their customers. These platforms use synthetic voices whereas NVIDIA natural language processing allows creation of the GPs
natural voice.
NVIDIA Collective AI alliance includes Yelp, AccuWeather.com, Sportstrader, Xignite, FlightStats, Onkyo, Sharp, Uber and Samsung
ARTIK.
This is a mature product built on US technology that is now calling
over 12000 patients per week in one Chinese hospital with a 98%
call completion rate.
5. 5
The technology and associated processes with Natural Language APIs will enable a more
personalized, intuitive, and effective patient care experience.
Average Patient Interaction Time
Readmission and Infection Rates
Self Service
STAFFING COST REDUCTIONPATIENT CARE IMPROVEMENT
Hospital Resource shortage
Patient Care Quality
Post-hospital follow-up care
Case Study:
Post hospital follow-up calls are current run by GPs who have resource contraints. Because of this a
large number of patients do not recieve followup care post discharge, resulting being readmitted to
hospital
Increase the calls, for all discharged patients, use the bots to intentify where support is needed
Opportunity:
Virtual Assistant
PREVENTING HOSPITAL READMISSIONS CAN
IMPROVE
- HEALTH SYSTEM BED CAPACITY
- PATIENT RECOVERY TIMES
- ON SET OF DEMENTIA IN FRAIL AND
ELDERLY DUE TO PROLONGED
HOSPITALISATION
- EARLY IDENTIFICATION OF SEPSIS AND
OTHER LIFE THREATENING POST OP
INFECTIONS.
6. 6
Types of questions can be organised to match the patient or
condition
http://112.17.124.29:6038/#/login
Each call can be scheduled to fit around the
pathway and questions added or removed for
each patient
7. 7
Follow up screen after a call with the Patient
The initial question
can be replayed here
A Patients response can be
heard and reviewed here
along with txt of the reply
8. 8
Abnormal responses are flagged
Follow-on questions
can be agreed and
can depend on the
previous response
Abnormal responses
flagged and listed on
a worklist
9. 9
Explore process, risks and issues
Some considerations …
Patient data
confidentiality
restrictions
Pressure on budgets, unable fund set up
Disparate data
sets to
integrate
Majority of patient demographic not familiar
with chat bots
What is the discharge
process
Technology infrastructure varies between
NHS Organisations
The Challenges