Improving practitioner decision making capabilities with data and analytics v1

Improving practitioner
decision making capabilities
with data and analytics
Ali Khan
Data & Analytics Director
Auckland District Health Board
17-03-2021

Welcome Haere Mai | Respect Manaaki | Together Tūhono | Aim High Angamua
About Auckland District Health Board
Introduction 2
• Auckland District Health Board (ADHB) serves around 10 per
cent of the country's population
• Provider of primary, secondary, tertiary and quaternary
services for around 1.6 million people in the northern region
• Regional and national centre of excellence
Auckland
Hospital
Greenlane Clinical
Centre
Starship Childrens
Hospital

Professional Background DHB Responsibilities Personal
Flat : 40 km/hr
Downhill : ??
Introduction
About me
• 20+ years in Technology and Data
• Enterprise transformation and data
migration projects
• Data governance implementations
• Industry data models
• Pervious life - Design and development
of n-tier web applications and solutions
• Intelligence, Analytics, Data
governance, Data quality & Data
architecture
• Data Strategy for ADHB
• Regional - Architecture group, Data
design authority, HIP Steerco, DSI
Working Group
• National – Integration Steering
Committee
• Sci-fi junkie
• Avid e-scooter rider
3

Health systems have recognised connected data across care settings is key to this transformation
Advances in Data and Analytics are creating new opportunities to
change how healthcare services are provided
4
The public health system will be shifting, from a
sometimes fragmented health and disability system with
a siloed service models, to a more connected and
whānau-centric approach (Health and Disablity System
Review March 2020).
Data and analytics are essential to making the
community, patient and whānau experience transparent
and to informing system redesign and that will also
improve performance.

At ADHB we are starting to use previously inaccessible data by applying new technologies
Using data in new ways through new capabilities is allowing ADHB to
gain better clinical and operational insights
5
Radiology data analysis
Radiology – Chest Nodule Detection
Care Navigation – Digital Workflow Tool
Workflow data as a patient story
Equity Focused Planned Care Response
AED hourly presentations forecasting

Analysis of our backlog of business requests shows that most business demand was focussed on the
left hand side of the analytics spectrum
However, one of the biggest growth areas is self-service analytics, the provision
of data and analytics capabilities for business users to develop their own
intelligence - when they need them
6
Analytics Spectrum
20-30% of demand
High Cost, High Repeatability, Gold
Quality
70-80% of total demand
Low Cost, High Variability,
Many One-off Insights
Management /
Performance
Strategic /
Compliance
Data Science
Ad-hoc Data
Analysis,
experimentation,
Research
Operational
Reporting
Operational
Analysis
Traditional Business
Intelligence Sweet Spot
Sweet Spot for Self-Service
Analytics

Business teams safely and securely use their data and the latest tools to build new insights and drive
innovation.
The Opportunity: By enabling business users to safely access and use their data in secure
dedicated environments, with supporting tools and environments, we can significant
increase the value business users derive from their data
7
PICU
Peri-
Operative
Labs Raw Data
Curated Data
Operational Data
Store
Enterprise Data
Warehouse
Data Marts
Security
&
Auditing
Data
Governance
(Data
Catalogue)
Advanced Analytics Models
OLAP / Semantic Models
Sandboxes /
Playpens
Data
Data
Data
Business
Intelligence
Tools
Data
Integration
Tools
Enterprise
Data
Catalogue
Each sandbox area is equipped with:
• Data storage for the temporary persistence
of data.
• Data integration tools to transform data.
Business users can still use database
constructs such as stored procedures to
write code.
• Business intelligence tools to develop
reports if needed.
• A data catalogue to explore and request
access to data.
• Utilities to automatically dispatch data to
sandboxes.
• Built-in security and audit capabilities to
track what data is being used and by whom.
A “Citizen Data Science”
example – where the
CLABs (central line
associated bacteraemias)
in the quarter a cluster or
not

The most difficult part is the change in data culture which needs to be supported by an operating model
that feeds and grows the self-service eco-system
ADHB’s self-service model adopts a holistic approach to foster the growth and
value of this capability
8
2. Operating
Model
Modern
Intelligence
Paradigm
3. Tools &
Technology
1. Data
Culture
Data Culture
Operating Model
Tools & Technology
Enable the army of citizen analysts, scientists and engineers to quickly discover the
data they need to make decisions. Data & intelligence teams will transition to a
model of co-delivery where they work closely with business teams, enabling them to
use the tools of their choice to get the insights they need.
Reconfiguration of our data and intelligence teams, with new capabilities added
to lift productivity, improve the quality of interactions with business teams and
speed of delivery.
We also need to make changes to our existing technology, continuing the rollout
of PowerBI but also add new capabilities to accelerate data delivery.

Each component has a synergy and dependency on the other making this a compelling model that can
work for all different types of users
Data Culture: We have defined data culture as 3 key components 9
Data
Literacy &
Capability
Inquisitive
Workforce
Data &
Tools
 Ability to discover any data in our
systems (on premise or cloud)
 Getting approvals quickly to use data
 The tools and data environments
needed to safely and securely use any
type and size of data
 Curious about observed patterns and willing to
dig a bit deeper
 The desire to use detailed insights before
making decisions
 Willing to learn new skills
 Understanding the data and what
it means
 Ability to decipher insights and
their business context
 Sufficient technical skills to
manipulate data to create relevant
insights
enables the use of

Our data use framework combines fine grain access controls and audit & monitoring capabilities to
maintain the safety of our valuable health data. Our end-users sign up to a data use contract and are
trained to use data safely and understand their obligations.
Tools & Technology: Safe use of data is the biggest challenge and potential
roadblock for self-service
14
Data,
Technology
& Security
Controls
Auditing &
Monitoring
Privacy & Security,
Ethics &
Consent
Delivery, Support
& Training
ADHB
Self-Service
Framework
1. Security Roles & Membership (who is
in what role)
2. Monthly Access sign-off (deltas)
3. Data Governance Forum (Business
Owners, Data Stewards)
1. ADHB Privacy and Security Policy
2. PowerBI Admin User terms and
conditions
3. Privacy Impact Assessment
4. Data Classification (applied to data
and reporting)
5. Staff Confidentiality Agreement
6. Self-Service User Agreement
7. Data Access Request Form
1. Auditing and Monitoring of access rights
2. User access history (reports, datasets, tables)
3. Data Access Reporting
1. Training / On-boarding

These personas have been tested with end users and continue to be refined and optimised as our
technical environments evolve
Operating Model: We have aligned our service offerings to a set of business
personas (Gartner)
12
INFORMATION PANEL
01
70%
SS users
Interaction
with content
Data literacy
Technical Skill 5% 50%
Info Consumer Info Explorer Citizen Analysts Citizen Data Scientist
ANALYTICS WORKBENCH
02
DATA SCIENCE LAB
03
Duplícate & modify
Able to nominate their
content
Unable to self validate their
work
Literate
Added ability to bring in new data to
“Sandbox”.
Able to promote their content
Able to self-validate and validate
work of others
Fluent - Multilingual
50% 75%
25% 5%
Content query
usage
Consume & Interact
No formal content
creation rights
Unable to validate or
nominate any new
content.
Conversational
Can build prototypes. Using “Sandbox”. Brings in domain-
specific data, flat files or third party data to enhance
analysis.
Able to nominate their content
Able to validate work of others
Competent
Validation of
work
Citizen Developer
Consumer Explorer Innovator Expert
25%

These capabilities allow a wider range of staff within directorates to
develop deeper insights - quickly
12
Decision Maker /
Information
Consumer
Information
Explorer
Citizen Analyst
Citizen Data
Developer
Citizen Data
Scientist
Key Activities
Technical Skills
Required
Tools Used
• View prebuilt reports • View prebuilt reports
• Slice and dice intelligence
reports
• Create and run excel pivot
tables to explore data
• Build new SQL queries to
interrogate data
• Build PowerBI Reports
from datasets
• Publish new reports
• Write complex code to
transform data
• Develop deterministic
models
• Prototype new
intelligence outputs
• Build analytical models
using statistical and
programmatic
approaches
• None • Excel pivot tables and
pivot charts
• PowerBI Development
• SQL Code
• PowerBI development
• SQL Scripting &
procedures
• Lightweight programming
(e.g. HTML, JavaScript)
• R / Python
• Advanced SQL coding
• Lightweight programming
(e.g. HTML, JavaScript)
• Web browser • Excel
• PowerBI Web (optional)
• PowerBI Desktop
• PowerBI Web
• SQL Management Studio
(or similar)
• PowerBI Desktop
(or similar)
• R Studio
• PowerBI Desktop or
similar
(or similar)

Analysts in directorate teams can prototype new datasets to build deeper insights.
If required, these can be optimised by HIT and released into production.
13
How does it work?
1) HIT uses our new Self-Service data
generation engine to create business
and analyst data views (simple excel
like views). These views remove all
complexities of joining data from
different systems
2) The business data views are
published as PowerBI data sets
3) PowerBI reports and dashboards are
generated using 1 or more PowerBI
data sets
4) Business Intelligence analysts (or
similar) in the business can generate
new insights by :-
a) Creating new reports using
PowerBI Datasets
b) Adding data to existing business
data views and publish new
PowerBI datasets (manual)
c) Create brand new business data
views using analyst data views,
business data views and/or raw
source system data views
Analyst
Playpen
• Business data views
• Analyst data Views
• Raw source system data
Source Systems
e.g. CMS,
Safer sleep
Data Warehouse
(Titan)
new
Import
PowerBI
Data Sets
Self-
Service
Engine
Enterprise Datasets –
Generated by
HIT SS Engine
Prototype +
Ad0hoc insights
Convert Prototype
insights into Enterprise
Directorate Staff
Use playpen to build new insights using :-
• Additional data from Titan
and source systems
• Import custom data from
spreadsheets & external sources
• Write SQL to build new views
HIT Data & Analytics team-
• Creates new libraries of simplified
views joining data from systems as
directed by customers
• Implements processes to publish
data to PowerBI data sets on a
regular basis (daily, hourly etc)
• Implement security and monitoring
to ensure data usage is governed
PowerBI
Visualisations
Microsoft Excel
Pivot tables and
charts
new
new

To create a self-serve “Plug & play” model of consuming data and analytics, we are pivoting our internal
teams into a different set of capabilities. This is enabling us to unlock value that sits within business
teams that an IT centric delivery model cannot realise
Operating Model: Our hybrid operating model wraps the core capabilities into
directorate specific capabilities
14
Supported by a hybrid operating Model
Current – Centralised
With Informal federated delivery
Future – Managed
hybrid delivery model
An intelligence capability model
A modern Intelligence paradigm
enabling pillars

A traditional capability vertical ensures our team develop high levels of competence in their area of
expertise. The focus on outcomes is provided by delivery squads which we recalibrate on a regular basis
Operating Model: Within the data & analytics, we have organised our delivery
functions into 2 layers
15
Maintaining structures aligned with our core
capabilities to increase competence and capability
Organised into self-organising teams (relatively) with a focus
a clear set of objectives and outcomes (Porter’s model)
Self-Service
Delivery
Squads
Projects Automation
Clinical
Databases
Data
Management

We have some of the building blocks already but are slowly investing in missing capabilities
16
Tools & Technology: We are acquiring and implementing a new range of tools to
build the foundations of frictionless self-service environments
Data
Virtualisation
Data
Visualisation
Data Movement
& Transformation
Data Quality
Management
Data
Catalog
CDC /
Replication
Workflow
Management
Data Visualisation
ADHB have PowerBI, which is a leading data visualisation and analysis tool. Additionally, existing
toolsets like SAP Business Objects, SAP Lumira and Excel offer a range of options for end users.
Workflow Management
With the acquisition of ServiceNow, Microsoft Office 365 and Azure,
we have all the tools we need to implement workflow management.
Data Movement & Transformation
ETL tools are normally used to extract, transform and load data. Our investment in WhereScape RED will need to
be reviewed overtime given the recent vendor acquisition and unmet demand for a higher capability toolset to
support data access from/to cloud based applications.
CDC / Data Replication
The ability to replicate all or part of any database within our eco-system is badly missing at ADHB and
required immediately to decrease the cost and time taken to provision new data to end-users. This
capability will also minimize time and effort needed to integrate data from cloud based systems.
Data Virtualisation
Data virtualization enables all data to come together without having to use complex and expensive ETL
procedures to physically move the data into one place. In other words, the data stays in its original location, but
you can query it just as if it was local. Some of this capability exists within newer versions of existing tools.
Data Catalog
The ability to find data located in any system regardless of where it sits. This is a big gap within our eco-
system and needs to be addressed urgently.
Data Quality Management
As the quality of data is better understood through the use of a data catalog, tools to monitor
and fix data quality can be purchased. The HARP transformation program would benefit
significant from a data quality and profiling tool.
Tools to needed enable self-service

Tools & Technology: Within the self-service environments, data is organised
into several layers allowing the different personas to operate
17
Analyst
Playpen
Business
Views &
Analyst Views
Raw Data
Subsets
 Data store for end users to transform data
 Users can also create and run analytical models
 New data can be uploaded by end-users directly
 Read-only access for end-users
 Simplified business views organised as a cube
 Business Views are organised by subject areas within data
 Maintained and governed by HIT
LABS
Mental
Health
Periop Others
 Read-only access for end-users
 Approved subsets of raw copy of source system data
 Data refreshed on agreed calendar and reconciled
 Maintained and governed by HIT
 Data sourced from source systems for multiple uses
 Uses include EDW provisioning, data Science, data quality
management, digital systems Consumption etc
 Dedicated and secure PowerBI workspaces
 Users can refresh report data from all 3 layers below
 Users can build and share new reports
Visualisation /
Presentation
Layer
Source System Data
Raw Copy
Working space
for end-users
Workspace/Directorate
specific data subsets
(dedicated for
self-service)
Shared Raw Copy
(also used for other
purposes)
End-user managed
supported by Self-Service Squad
Managed by D&A
Self-Service Squad
Managed by D&A
Engineering team in partnership
with healthAlliance
LABS
Mental
Health
Periop Others
LABS
Mental
Health
Periop Others
LABS
Mental
Health
Periop Others

They also help us govern environments, ensuring we continue to meeting our obligations to patient data
Our data stewards are our self-service champions in the business and
are supporting adoption within their directorates
Patient
Day / Inpatient
S
O
Patient
Outpatient
S
O
Patient
Theatres
S
O
Clinical Coding
S
O
Clinical Records
S
O
Radiology
S
O
Laboratories
S
O
Child Health
S
O
Ophthalmology
S
O
Community
S
O
Mental Health
S
O
Pharmacy
S
O
Human Resources
S
O
Finance
S
O
Women’s Health
S
O
17

Our data governance function has been instrumental in gaining
support and approval to proceed
19
Digital Steering
Committee
Data Governance Support
Office
Data Stewardship
Council
Working
Group(s)
Information Governance
& Privacy Group

We have also opened up a range of new opportunities for our teams to focus on in 2021
Progress: With self-service, we have increased the value being delivered by the
D&A team across the DHB
20
Self-Service Implementation
(Stage 1)
In-Progress
• Labs Self-Service
• PICU
• Perioperative
• HR (PowerBI Only)
• Cardiology (PBI Only)
Operational Data Store (temporary)
• Lab demand optimisation
• Women’s Health IOC Dashboard
(initial)
Stage 1 – Extended (information
explorer, consumer only)
• Production Planning (initial)
• Performance Improvement (Planned
Care scorecard)
• Outpatient (DNA Reporting)
• Mental Health (Patrick)
• Allied Health (Joe Monkhouse)
• Adult Long Term & Community (BAU
Backlog)
• Clinical Coding (diagnoses)
• Child Health Excellence
Cloud Data Environment &
Tools
Stage 2 – Self Service
• Women’s Health
• Quality & Safety (Jenny)
• Radiology (Nicola)
• Finance (Nicki Hill)
• Cancer & Blood (Ben Lawrence)
• Pharmacy (tbc)
• Ophthalmology (tbc)
• Community (tbc)
• ACC (tbc)
Data Warehouse / ODS (Cloud)
• Lab demand optimisation
• Quality data mart (tbc)
• Production Planning Reporting
• Healthcare Logic / SFN (tbc)
• IOC – Women’s Health
• Hospital IOC extensions
Digital
• Service Now (ODBC)
• Single View
Production Data Analytics
Implementation
POC’s
• Sleep Studies
• Radiology AI
• Others
Single View
• Single view of patient
• Production Planning Analytical
models
• Integrated pathways data (prod
planning)
Clinical Database
Design & Implementation
Confirmed Demand
• Enterprise Dendrite
• Dendrite Multi-site Registries
• Redcap Implementation
• Edge Data Integration
• Edge BI Reporting
Data API’s / Micro-Services
Confirmed Demand
• Service Now Data API’s
• Clinical Trial Mgmt API’s
• Clinical Audit Data API’s
• Clinical Research API’s
HIT Portfolio Demand
• SNOW Digital Apps requiring Data
e.g. planned care projects
• Data supply to key apps e.g. e-orders,
SMT, CarePathways etc where
appropriate

Cloud Data Staging Layer
Our citizen workforce will lead the development valuable new insights which can be deployed across
various parts of our data environments
The Future: As our self-service environments are mature, business led data projects will
provide relevant content for the rest of our data environments
21
Older/delicate DHB
Source Systems
Raw Data Staging Layer
On-Premise
Cloud
Newer DHB
Source Systems
Cloud Raw Data (Structured)
PowerBI
Workspace
Operational
Data Stores
DevOps
CI/CD
GitHub
Management
Console
Fast Data
Cache
(COSMOS /
Redis/Mem
cache)
API Gateway & Catalog
Batch Load,
Micro Batch
Streaming
Batch,
Micro batch
Cloud Raw Data (Unstructured)
Playpen
(for self-
service)
Reconciliation
MDM
(single views
of …)
Container
Management
for Analytics &
Microservices
EDW
Analytical
Datasets
Micro Batch
Streaming
Medical Devices,
Data Lake
Batch, Micro Batch, Streaming
Data Catalog
Data Profiling

Improving practitioner decision making capabilities with data and analytics v1

More Related Content

What's hot (20)

Similar to Improving practitioner decision making capabilities with data and analytics v1 (20)

Recently uploaded (20)

Improving practitioner decision making capabilities with data and analytics v1