DATA-DRIVEN
DECISION-MAKING IN
COVID-19 PANDEMIC
MANAGEMENT
CraftCon, 26.3.2021
Sami Laine, Competence Lead
Siili Solutions Oyj
1. Data-Driven Decision-Making in Covid-19
Pandemic Management
2. Data Management Best Practices
26.3.2021
SIILI CraftCon / Embrace Change 2
AGENDA (what is important)
(how to be successful)
Introduction
News start to fill with
figures about diagnosed
cases and patient
deaths…
A new virus
emerges –
Covid-19
Data-Driven Decision-Making in Covid-19 Pandemic Management
These decisions are
based on facts and
simulations…
Data-driven
decisions shut
down societies!
Data-Driven Decision-Making in Covid-19 Pandemic Management
Lockdowns strain
dramatically global
economic outlook for
unknown time period...
This has dramatic
impact to
economy and daily
life!
Data-Driven Decision-Making in Covid-19 Pandemic Management
• Lockdown does not make sense!
— Lillrank Professori lyö tiskiin rankan
laskelman: Koronatoimien kustannus on yli
0,5 miljoonaa euroa per säästynyt
elinvuosi – "Lockdownissa ei ole mitään
järkeä”, MTV Uutiset
But are our efforts
proportional to their
costs, side-effects
and alternatives?
• Pandemic controls cause other
problems!
— WHO, Levels of domestic violence
increase globally, including in the Region,
as COVID-19 pandemic escalates
Are we saving 10K lives from Covid-19 in Finland but killing more people
elsewhere and ruining even more lives, careers and happiness?
1. Data-Driven Decision-Making in Covid-19 Pandemic Management
• How many die to influenza?
— Taskinen, Pajunen, Kuinka monen
kuoleman syy on influenssa – kertovatko
luvut kaiken?, Tilastokeskus
But are our efforts
proportional to their
costs, side-effects
and alternatives?
• Malaria causes 400k deaths per year
down from 1M yearly deaths about a
decade ago.
— WHO, The "World malaria report
2019" at a glance
Are we using thousands billions to save 1M western elderly from covid-
19 but not a few more billions to save 1M developing world children
dying to other diseases?
1. Data-Driven Decision-Making in Covid-19 Pandemic Management
• Davenport, Godfrey, Redman, To Fight
Pandemics, We Need Better Data,
MIT Sloan Management Review
— We still don’t know how many people
have the virus, how many are hospitalized,
how many are in intensive care units, and
how many are on ventilators.
Hello world, we
have a data
problem!
Data-Driven Decision-Making in Covid-19 Pandemic Management
• Nevalainen, Huono data voi johtaa
harhaan taistelussa koronaa vastaan,
Unit Magazine
— Luvuissa, tilastoissa ja niihin pohjautuvissa
malleissa on tärkeää kyky paitsi tulkita
data-analyysien tuloksia, myös arvioida
niiden luotettavuutta.
Data is incomplete, inconsistent, incorrect, manipulated, biased, and so
on...
Did we just throw away 1/2 euro-area GDP
because of incorrect data?
Did we just trade the lives of
1M western world elderly to
10M developing country
children and parents due to
erroneous data?
Data-Driven Decision-Making in Covid-19 Pandemic Management
WHY EXACTLY IS
HIGH QUALITY DATA
CRITICAL FOR
DECISION-MAKING?
THEY ARE WINNING THE VIRUS!
Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis
Which country is doing better?
Data-Driven Decision-Making in Covid-19 Pandemic Management
THEY ARE WINNING THE VIRUS!
Because they are not testing people...
In reality, almost everyone already has it!
Because virus actually spreads slower...
At the same time, we do not know WHY!
Persons with Covid-19 diagnosis
Persons with Covid-19 condition (asymptomatic)
Persons with Covid-19 diagnosis
Persons with Covid-19 condition
(asymptomatic)
Detecting difference between highly and mildly contagious virus
requires testing asymptomatic population!
Data-Driven Decision-Making in Covid-19 Pandemic Management
THEY ARE WINNING THE VIRUS!
Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis
Which country is doing better?
Data-Driven Decision-Making in Covid-19 Pandemic Management
THEY ARE WINNING THE VIRUS!
Because everyone has already had it!
Therefore, controls waste resources.
Because virus spreads slower...
Therefore, controls surpress virus...
Persons with Covid-19 diagnoses
Persons with Covid-19 antibodies
Persons with Covid-19 diagnoses
Persons with Covid-19 antibodies
Detecting difference between common and rare virus requires
measuring antibodies!
Data-Driven Decision-Making in Covid-19 Pandemic Management
THEY ARE WINNING THE VIRUS!
Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis
Which country is doing better?
Data-Driven Decision-Making in Covid-19 Pandemic Management
THEY ARE WINNING THE VIRUS!
Because they do not test dead people!
In reality, people have died in masses
Because people are not getting sick!
In reality, they are controlling it.
Persons with Covid-19 deaths
Persons died after Covid-19 (all deaths)
Persons with Covid-19 deaths
Persons died after Covid-19 (all deaths)
Detecting difference between controlling and losing to the virus
requires measuring total death rate!
Data-Driven Decision-Making in Covid-19 Pandemic Management
4/23/2021 17
The difference
between official
statistics and
reality
— Official statistics
— Infections < 60K
— Deaths = 426
— Sample study
— Antibodies = 23%
— Population > 24M
— Conclusions
— Survivors > 5,5M
— Deaths = ?
23M (2019 estimate)
In Lombardy, excess deaths to all causes was +111% in comparison to earlier
years! Almost half of those are not associated (officially) to Covid-19.
In Lagos, Covid-19 is 100 times more common than official statistics present!
To make valid conclusions we need to know all the variables!
CONTAGIOUSNESS
- Infections (Unknown!)
- Symptoms
- Antibodies
(Unknown!)
IMPACT
- Hospitalization
- Mortality (Unknown!)
- QUALY (Unknown!)
EPISODE
- Duration
- Complications
(Unknown!)
- Reinfections (Unknown!)
Public media is following very partial picture of the disease and even that
is heavily distorted across countries.
Data-Driven Decision-Making in Covid-19 Pandemic Management
We still do not know what kind of virus covid-19 really is
due to lack of high-quality data and understanding!
Mostly harmless virus Sometimes dangerous
virus
Always deadly virus
Very fast to infect
Mediocre to infect
What is Covid-19?
Very slow to infect
Are deadly cases
outliers or common
cases?
How fast it
spreads across
populations?
Data-Driven Decision-Making in Covid-19 Pandemic Management
Oh, and the
virus also keeps
mutating!
Understanding the problem context is important when
selecting optimal reaction to you needs in your situation!
Prevention
• Vaccine
• Distancing
• Protection
Treatment
• Medicine
• Hospitalization
• Intensive care
Controls
• Diagnose
• Track
• Isolate
Mode
• Inform
• Recommend
• Command
Globally, there is huge differences in abilities to invest in interventions.
Data-Driven Decision-Making in Covid-19 Pandemic Management
Depending on the characteristics of the virus the same
effort can be waste of trillions euros or savior of societies!
Mostly harmless virus Sometimes dangerous
virus
Always deadly virus
Very fast to infect
Mediocre to infect
What is Covid-19?
Very slow to infect
Isolate fast but only
infected individuals is
enough
Isolate fast and
massively even
healthy populations.
Data-Driven Decision-Making in Covid-19 Pandemic Management
It’s already gone when
testing. Focus on
antibodies.
Even a small improvement
in understanding can save
billions euros and help
millions people!
You need to understand the problem and its context to
priorize the investments to data assets!
Data-Driven Decision-Making in Covid-19 Pandemic Management
Understand and frame the
problem context
• Learn all the
variables
• Learn their
relations
Focus on
measurement of
relevant questions
• Think business
objectives
• Prioritize
measurements
Productize the
information
production process
• Manage data as
an asset.
• Manage data as
a product.
Build ICT
systems
DATA MANAGEMENT
BEST PRACTICES
Sami Laine, Competence Lead
Siili Solutions Oyj
Manage Business Value Chain across Business
Contexts
Data Management Best Practices
Physicians
Managers
Citizens
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Researchers
Outpatient
visits
Inpatient
episodes
Homes
DATA MANAGEMENT & GOVERNANCE
Manage Business Value Chain across Business
Contexts
The Three Grand Challenges in Data Quality Management
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Outpatient
visits
Inpatient
episodes
Home
activities
DATA MANAGEMENT & GOVERNANCE
You need to collect high quality data and also
link them to meaningful value chains!
ServiceType PatientNumber Diagnosis
Covid19Symptomat
ic SourceSystem
Outpatient visit 21321321 U07.2; R05 TRUE EPR
Outpatient visit 13123541 U07.2 FALSE EPR
Outpatient visit 31436516 U07.2 FALSE EPR
Outpatient visit 72368981 U07.2; R05 TRUE EPR
ServiceType PatientNumber Diagnosis
Covid19Infecte
d SourceSystem
Laboratory 21321321U07.1; R05 TRUE LIMS
Laboratory 13123541 U07.1 TRUE LIMS
Laboratory 31436516 FALSE LIMS
Laboratory 72368981 FALSE LIMS
ServiceType PatientNumber Diagnosis
Covid19Infecte
d Hospitalized IntensiveCare SourceSystem
Inpatient episode 21321321 U07.1; R05 TRUE TRUE FALSE EPR
Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS
Inpatient episode 9916516 TRUE TRUE TRUE EPR
Inpatient episode 53319181 FALSE TRUE FALSE EPR
ServiceType PatientNumber Isolated Covid19Infected SourceSystem
Home 21321321 TRUE TRUE EPR
Home 13123541 TRUE TRUE ERP
A pathway of severily ill covid-
positive patient with obesity
hospitalized
Simple diagnosis counts are meaningless!
Remember the need to know all the variables and
to understand context!
Improve care pathway and maximize business value i.e prevent expensive
treatment
The Three Grand Challenges in Data Quality Management
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Outpatient
visits
Inpatient
episodes
Home
activities
A pathway of severily ill covid-
positive patient with obesity
hospitalized
DATA MANAGEMENT & GOVERNANCE
Meaningful linked data helps to recognize when to
fix process like prevent expensive treatment!
ServiceType PatientNumber Diagnosis
Covid19Symptomat
ic SourceSystem
Outpatient visit 21321321 U07.2; R05 TRUE EPR
Outpatient visit 13123541 U07.2 FALSE EPR
Outpatient visit 31436516 U07.2 FALSE EPR
Outpatient visit 72368981 U07.2; R05 TRUE EPR
ServiceType PatientNumber Diagnosis
Covid19Infecte
d SourceSystem
Laboratory 21321321U07.1; R05 TRUE LIMS
Laboratory 13123541 U07.1 TRUE LIMS
Laboratory 31436516 FALSE LIMS
Laboratory 72368981 FALSE LIMS
ServiceType PatientNumber Diagnosis
Covid19Infecte
d Hospitalized IntensiveCare SourceSystem
Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE FALSE EPR
Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS
Inpatient episode 9916516 TRUE TRUE TRUE EPR
Inpatient episode 53319181 FALSE TRUE FALSE EPR
ServiceType PatientNumber Isolated Covid19Infected SourceSystem
Home 21321321 TRUE TRUE EPR
Home 13123541 TRUE TRUE ERP
ServiceType PatientNumber Diagnosis
Covid19Infecte
d Hospitalized IntensiveCare SourceSystem
Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE TRUE ICU
E66 High risk comorbidity increases
risk for intensive care!
Improve care pathway and maximize business value i.e stop infection chain
before you know it exists!
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Outpatient
visits
Inpatient
episodes
Home
activities
Severily ill covid-positive patient with
obesity hospitalized.
DATA MANAGEMENT & GOVERNANCE
Meaningful contextual data helps to act even
before you have actual data!
ServiceType PatientNumber Diagnosis
Covid19Symptomat
ic SourceSystem
Outpatient visit 21321321 U07.2; R05 TRUE EPR
Outpatient visit 13123541 U07.2; Z20.8 FALSE EPR
Outpatient visit 31436516 U07.2 FALSE EPR
Outpatient visit 72368981 U07.2; R05 TRUE EPR
ServiceType PatientNumber Diagnosis
Covid19Infecte
d SourceSystem
Laboratory 21321321U07.1; R05 TRUE LIMS
Laboratory 13123541 U07.1 TRUE LIMS
Laboratory 31436516 FALSE LIMS
Laboratory 72368981 FALSE LIMS
ServiceType PatientNumber Diagnosis
Covid19Infecte
d Hospitalized IntensiveCare SourceSystem
Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE FALSE EPR
Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS
Inpatient episode 9916516 TRUE TRUE TRUE EPR
Inpatient episode 53319181 FALSE TRUE FALSE EPR
ServiceType PatientNumber Isolated Covid19Infected SourceSystem
Home 21321321 TRUE TRUE EPR
Home 13123541 TRUE TRUE ERP
Z20.8 = Contact with Covid positive!
He is getting positive result in two days
even if we do not know it yet!
COUNTS DON’T COUNT WITHOUT CONTEXT!
Succesfull analytics focuses always
on business value chain!
Manage Information Production Process from data Creation to Usage
Data Management Best Practices
Physicians
Managers
Citizens
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Researchers
Outpatient
visits
Inpatient
episodes
Home
activities
DATA MANAGEMENT & GOVERNANCE
Improve the quality of data and provide audit trail to results to support decision-
makers
Data Management Best Practices
20% error
rate
20% error
rate
Y=100 cases = safe country
X=200 cases = problematic
country
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Positive
diagnosis
=200
Positive
diagnosis
=100
DATA MANAGEMENT & GOVERNANCE
You need to understand how data was
created and results calculated to use them
for valid conclusions!
No! It’s safer to be
in country with
more recognized
Covid-cases!
Total tests
=20000 per 1M
Total tests
=500 per 1M
1% positive
rate
20% positive
rate
COUNTS ARE MISLEADING WITHOUT
COMPLETE AUDIT TRAIL!
Succesfull analytics maintains
always full audit trail!
Manage Data Lifecycle from data Creation to Deletion
Data Management Best Practices
Physicians
Managers
Citizens
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Researchers
Outpatient
visits
Inpatient
episodes
Home
activities
DATA MANAGEMENT & GOVERNANCE
Manage states and transitions to provide high quality data for decision-makers
Data Management Best Practices
Created Active Inactived
DATA MANAGEMENT & GOVERNANCE
State
Event State
Event State
Event State
Event
Event
Event
State
Manage Data Lifecycle of the whole patient registry from data Creation to Deletion
Manage states and transitions to provide high quality data for decision-makers
Data Management Best Practices
DATA MANAGEMENT & GOVERNANCE
Manage Data Lifecycle of the whole patient registry from data Creation to Deletion
Fetus
Baby was
born Alive person
Comorbidity
was detected Risk patient
Tested positive
for pregnancy Dead person
Comorbidity
was cured
15 years
passed from
death
Person died
Deleted person
PatientNumber Diagnosis
13123541
31436516
53319181
21321321
72368981
9916516
Pre-birth date Lifetime data Archived data
Manage states and transitions to provide high quality data for decision-makers
Data Management Best Practices
Fetus
Baby was
born Alive person
Comorbidity
was detected
Pre-birth date Lifetime data Archived data
Risk patient
Tested positive
for pregnancy Dead person
DATA MANAGEMENT & GOVERNANCE
The challenge is that data can be scattered across numerous systems and countries!
Comorbidity
was cured
15 years
passed from
death
Person died
Deleted person
PatientNumber Diagnosis
13123541
31436516
53319181
PatientNumber Diagnosis
21321321 E66
72368981 I25.1
9916516 I25.1
Detect and track E66 and I25.1
comorbidities since they increase risk
for intensive care and death.
COUNTS ARE RISKY WITHOUT LIFECYCLE !
Succesfull analytics always
manages how data changes across
time and contexts
There is three grand challenges that need to be
managed to enable trustworthy data for valid decision-
making
Data Management Best Practices
Mobile Applications
for field workers
Business Intelligence
for managers
Public Internet sites
for customers
DATA SUPPLY DATA MANUFACTURING
DATA
CONSUMPTION
Statistical Tools
for data scientists
Marketing
Sales
Delivery
DATA MANAGEMENT & GOVERNANCE
Manage Information Production Process from data Creation to Usage
26.3.2021
SIILI CraftCon / Embrace Change 38
Two things to
remember!
Thanks for your attention!
QUESTIONS?
Sami Laine
Competence Lead, Siili Solutions Oyj
President, DAMA Finland ry
Program Committee Co-Chair, MIT CDOIQ Symposium
https://www.linkedin.com/pub/sami-laine/2/a61/970

Siili craft con-20210326-sami-laine-ddd_mincpm-final

  • 1.
    DATA-DRIVEN DECISION-MAKING IN COVID-19 PANDEMIC MANAGEMENT CraftCon,26.3.2021 Sami Laine, Competence Lead Siili Solutions Oyj
  • 2.
    1. Data-Driven Decision-Makingin Covid-19 Pandemic Management 2. Data Management Best Practices 26.3.2021 SIILI CraftCon / Embrace Change 2 AGENDA (what is important) (how to be successful) Introduction
  • 3.
    News start tofill with figures about diagnosed cases and patient deaths… A new virus emerges – Covid-19 Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 4.
    These decisions are basedon facts and simulations… Data-driven decisions shut down societies! Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 5.
    Lockdowns strain dramatically global economicoutlook for unknown time period... This has dramatic impact to economy and daily life! Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 6.
    • Lockdown doesnot make sense! — Lillrank Professori lyö tiskiin rankan laskelman: Koronatoimien kustannus on yli 0,5 miljoonaa euroa per säästynyt elinvuosi – "Lockdownissa ei ole mitään järkeä”, MTV Uutiset But are our efforts proportional to their costs, side-effects and alternatives? • Pandemic controls cause other problems! — WHO, Levels of domestic violence increase globally, including in the Region, as COVID-19 pandemic escalates Are we saving 10K lives from Covid-19 in Finland but killing more people elsewhere and ruining even more lives, careers and happiness? 1. Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 7.
    • How manydie to influenza? — Taskinen, Pajunen, Kuinka monen kuoleman syy on influenssa – kertovatko luvut kaiken?, Tilastokeskus But are our efforts proportional to their costs, side-effects and alternatives? • Malaria causes 400k deaths per year down from 1M yearly deaths about a decade ago. — WHO, The "World malaria report 2019" at a glance Are we using thousands billions to save 1M western elderly from covid- 19 but not a few more billions to save 1M developing world children dying to other diseases? 1. Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 8.
    • Davenport, Godfrey,Redman, To Fight Pandemics, We Need Better Data, MIT Sloan Management Review — We still don’t know how many people have the virus, how many are hospitalized, how many are in intensive care units, and how many are on ventilators. Hello world, we have a data problem! Data-Driven Decision-Making in Covid-19 Pandemic Management • Nevalainen, Huono data voi johtaa harhaan taistelussa koronaa vastaan, Unit Magazine — Luvuissa, tilastoissa ja niihin pohjautuvissa malleissa on tärkeää kyky paitsi tulkita data-analyysien tuloksia, myös arvioida niiden luotettavuutta. Data is incomplete, inconsistent, incorrect, manipulated, biased, and so on...
  • 9.
    Did we justthrow away 1/2 euro-area GDP because of incorrect data? Did we just trade the lives of 1M western world elderly to 10M developing country children and parents due to erroneous data? Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 10.
    WHY EXACTLY IS HIGHQUALITY DATA CRITICAL FOR DECISION-MAKING?
  • 11.
    THEY ARE WINNINGTHE VIRUS! Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis Which country is doing better? Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 12.
    THEY ARE WINNINGTHE VIRUS! Because they are not testing people... In reality, almost everyone already has it! Because virus actually spreads slower... At the same time, we do not know WHY! Persons with Covid-19 diagnosis Persons with Covid-19 condition (asymptomatic) Persons with Covid-19 diagnosis Persons with Covid-19 condition (asymptomatic) Detecting difference between highly and mildly contagious virus requires testing asymptomatic population! Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 13.
    THEY ARE WINNINGTHE VIRUS! Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis Which country is doing better? Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 14.
    THEY ARE WINNINGTHE VIRUS! Because everyone has already had it! Therefore, controls waste resources. Because virus spreads slower... Therefore, controls surpress virus... Persons with Covid-19 diagnoses Persons with Covid-19 antibodies Persons with Covid-19 diagnoses Persons with Covid-19 antibodies Detecting difference between common and rare virus requires measuring antibodies! Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 15.
    THEY ARE WINNINGTHE VIRUS! Persons with Covid-19 diagnosis Persons with Covid-19 diagnosis Which country is doing better? Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 16.
    THEY ARE WINNINGTHE VIRUS! Because they do not test dead people! In reality, people have died in masses Because people are not getting sick! In reality, they are controlling it. Persons with Covid-19 deaths Persons died after Covid-19 (all deaths) Persons with Covid-19 deaths Persons died after Covid-19 (all deaths) Detecting difference between controlling and losing to the virus requires measuring total death rate! Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 17.
    4/23/2021 17 The difference betweenofficial statistics and reality — Official statistics — Infections < 60K — Deaths = 426 — Sample study — Antibodies = 23% — Population > 24M — Conclusions — Survivors > 5,5M — Deaths = ? 23M (2019 estimate) In Lombardy, excess deaths to all causes was +111% in comparison to earlier years! Almost half of those are not associated (officially) to Covid-19. In Lagos, Covid-19 is 100 times more common than official statistics present!
  • 18.
    To make validconclusions we need to know all the variables! CONTAGIOUSNESS - Infections (Unknown!) - Symptoms - Antibodies (Unknown!) IMPACT - Hospitalization - Mortality (Unknown!) - QUALY (Unknown!) EPISODE - Duration - Complications (Unknown!) - Reinfections (Unknown!) Public media is following very partial picture of the disease and even that is heavily distorted across countries. Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 19.
    We still donot know what kind of virus covid-19 really is due to lack of high-quality data and understanding! Mostly harmless virus Sometimes dangerous virus Always deadly virus Very fast to infect Mediocre to infect What is Covid-19? Very slow to infect Are deadly cases outliers or common cases? How fast it spreads across populations? Data-Driven Decision-Making in Covid-19 Pandemic Management Oh, and the virus also keeps mutating!
  • 20.
    Understanding the problemcontext is important when selecting optimal reaction to you needs in your situation! Prevention • Vaccine • Distancing • Protection Treatment • Medicine • Hospitalization • Intensive care Controls • Diagnose • Track • Isolate Mode • Inform • Recommend • Command Globally, there is huge differences in abilities to invest in interventions. Data-Driven Decision-Making in Covid-19 Pandemic Management
  • 21.
    Depending on thecharacteristics of the virus the same effort can be waste of trillions euros or savior of societies! Mostly harmless virus Sometimes dangerous virus Always deadly virus Very fast to infect Mediocre to infect What is Covid-19? Very slow to infect Isolate fast but only infected individuals is enough Isolate fast and massively even healthy populations. Data-Driven Decision-Making in Covid-19 Pandemic Management It’s already gone when testing. Focus on antibodies. Even a small improvement in understanding can save billions euros and help millions people!
  • 22.
    You need tounderstand the problem and its context to priorize the investments to data assets! Data-Driven Decision-Making in Covid-19 Pandemic Management Understand and frame the problem context • Learn all the variables • Learn their relations Focus on measurement of relevant questions • Think business objectives • Prioritize measurements Productize the information production process • Manage data as an asset. • Manage data as a product. Build ICT systems
  • 23.
    DATA MANAGEMENT BEST PRACTICES SamiLaine, Competence Lead Siili Solutions Oyj
  • 24.
    Manage Business ValueChain across Business Contexts Data Management Best Practices Physicians Managers Citizens DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Researchers Outpatient visits Inpatient episodes Homes DATA MANAGEMENT & GOVERNANCE
  • 25.
    Manage Business ValueChain across Business Contexts The Three Grand Challenges in Data Quality Management DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Outpatient visits Inpatient episodes Home activities DATA MANAGEMENT & GOVERNANCE You need to collect high quality data and also link them to meaningful value chains! ServiceType PatientNumber Diagnosis Covid19Symptomat ic SourceSystem Outpatient visit 21321321 U07.2; R05 TRUE EPR Outpatient visit 13123541 U07.2 FALSE EPR Outpatient visit 31436516 U07.2 FALSE EPR Outpatient visit 72368981 U07.2; R05 TRUE EPR ServiceType PatientNumber Diagnosis Covid19Infecte d SourceSystem Laboratory 21321321U07.1; R05 TRUE LIMS Laboratory 13123541 U07.1 TRUE LIMS Laboratory 31436516 FALSE LIMS Laboratory 72368981 FALSE LIMS ServiceType PatientNumber Diagnosis Covid19Infecte d Hospitalized IntensiveCare SourceSystem Inpatient episode 21321321 U07.1; R05 TRUE TRUE FALSE EPR Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS Inpatient episode 9916516 TRUE TRUE TRUE EPR Inpatient episode 53319181 FALSE TRUE FALSE EPR ServiceType PatientNumber Isolated Covid19Infected SourceSystem Home 21321321 TRUE TRUE EPR Home 13123541 TRUE TRUE ERP A pathway of severily ill covid- positive patient with obesity hospitalized Simple diagnosis counts are meaningless! Remember the need to know all the variables and to understand context!
  • 26.
    Improve care pathwayand maximize business value i.e prevent expensive treatment The Three Grand Challenges in Data Quality Management DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Outpatient visits Inpatient episodes Home activities A pathway of severily ill covid- positive patient with obesity hospitalized DATA MANAGEMENT & GOVERNANCE Meaningful linked data helps to recognize when to fix process like prevent expensive treatment! ServiceType PatientNumber Diagnosis Covid19Symptomat ic SourceSystem Outpatient visit 21321321 U07.2; R05 TRUE EPR Outpatient visit 13123541 U07.2 FALSE EPR Outpatient visit 31436516 U07.2 FALSE EPR Outpatient visit 72368981 U07.2; R05 TRUE EPR ServiceType PatientNumber Diagnosis Covid19Infecte d SourceSystem Laboratory 21321321U07.1; R05 TRUE LIMS Laboratory 13123541 U07.1 TRUE LIMS Laboratory 31436516 FALSE LIMS Laboratory 72368981 FALSE LIMS ServiceType PatientNumber Diagnosis Covid19Infecte d Hospitalized IntensiveCare SourceSystem Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE FALSE EPR Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS Inpatient episode 9916516 TRUE TRUE TRUE EPR Inpatient episode 53319181 FALSE TRUE FALSE EPR ServiceType PatientNumber Isolated Covid19Infected SourceSystem Home 21321321 TRUE TRUE EPR Home 13123541 TRUE TRUE ERP ServiceType PatientNumber Diagnosis Covid19Infecte d Hospitalized IntensiveCare SourceSystem Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE TRUE ICU E66 High risk comorbidity increases risk for intensive care!
  • 27.
    Improve care pathwayand maximize business value i.e stop infection chain before you know it exists! DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Outpatient visits Inpatient episodes Home activities Severily ill covid-positive patient with obesity hospitalized. DATA MANAGEMENT & GOVERNANCE Meaningful contextual data helps to act even before you have actual data! ServiceType PatientNumber Diagnosis Covid19Symptomat ic SourceSystem Outpatient visit 21321321 U07.2; R05 TRUE EPR Outpatient visit 13123541 U07.2; Z20.8 FALSE EPR Outpatient visit 31436516 U07.2 FALSE EPR Outpatient visit 72368981 U07.2; R05 TRUE EPR ServiceType PatientNumber Diagnosis Covid19Infecte d SourceSystem Laboratory 21321321U07.1; R05 TRUE LIMS Laboratory 13123541 U07.1 TRUE LIMS Laboratory 31436516 FALSE LIMS Laboratory 72368981 FALSE LIMS ServiceType PatientNumber Diagnosis Covid19Infecte d Hospitalized IntensiveCare SourceSystem Inpatient episode 21321321 U07.1; R05; E66 TRUE TRUE FALSE EPR Inpatient episode 73221511 U07.1 TRUE TRUE FALSE LIMS Inpatient episode 9916516 TRUE TRUE TRUE EPR Inpatient episode 53319181 FALSE TRUE FALSE EPR ServiceType PatientNumber Isolated Covid19Infected SourceSystem Home 21321321 TRUE TRUE EPR Home 13123541 TRUE TRUE ERP Z20.8 = Contact with Covid positive! He is getting positive result in two days even if we do not know it yet!
  • 28.
    COUNTS DON’T COUNTWITHOUT CONTEXT! Succesfull analytics focuses always on business value chain!
  • 29.
    Manage Information ProductionProcess from data Creation to Usage Data Management Best Practices Physicians Managers Citizens DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Researchers Outpatient visits Inpatient episodes Home activities DATA MANAGEMENT & GOVERNANCE
  • 30.
    Improve the qualityof data and provide audit trail to results to support decision- makers Data Management Best Practices 20% error rate 20% error rate Y=100 cases = safe country X=200 cases = problematic country DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Positive diagnosis =200 Positive diagnosis =100 DATA MANAGEMENT & GOVERNANCE You need to understand how data was created and results calculated to use them for valid conclusions! No! It’s safer to be in country with more recognized Covid-cases! Total tests =20000 per 1M Total tests =500 per 1M 1% positive rate 20% positive rate
  • 31.
    COUNTS ARE MISLEADINGWITHOUT COMPLETE AUDIT TRAIL! Succesfull analytics maintains always full audit trail!
  • 32.
    Manage Data Lifecyclefrom data Creation to Deletion Data Management Best Practices Physicians Managers Citizens DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Researchers Outpatient visits Inpatient episodes Home activities DATA MANAGEMENT & GOVERNANCE
  • 33.
    Manage states andtransitions to provide high quality data for decision-makers Data Management Best Practices Created Active Inactived DATA MANAGEMENT & GOVERNANCE State Event State Event State Event State Event Event Event State Manage Data Lifecycle of the whole patient registry from data Creation to Deletion
  • 34.
    Manage states andtransitions to provide high quality data for decision-makers Data Management Best Practices DATA MANAGEMENT & GOVERNANCE Manage Data Lifecycle of the whole patient registry from data Creation to Deletion Fetus Baby was born Alive person Comorbidity was detected Risk patient Tested positive for pregnancy Dead person Comorbidity was cured 15 years passed from death Person died Deleted person PatientNumber Diagnosis 13123541 31436516 53319181 21321321 72368981 9916516 Pre-birth date Lifetime data Archived data
  • 35.
    Manage states andtransitions to provide high quality data for decision-makers Data Management Best Practices Fetus Baby was born Alive person Comorbidity was detected Pre-birth date Lifetime data Archived data Risk patient Tested positive for pregnancy Dead person DATA MANAGEMENT & GOVERNANCE The challenge is that data can be scattered across numerous systems and countries! Comorbidity was cured 15 years passed from death Person died Deleted person PatientNumber Diagnosis 13123541 31436516 53319181 PatientNumber Diagnosis 21321321 E66 72368981 I25.1 9916516 I25.1 Detect and track E66 and I25.1 comorbidities since they increase risk for intensive care and death.
  • 36.
    COUNTS ARE RISKYWITHOUT LIFECYCLE ! Succesfull analytics always manages how data changes across time and contexts
  • 37.
    There is threegrand challenges that need to be managed to enable trustworthy data for valid decision- making Data Management Best Practices Mobile Applications for field workers Business Intelligence for managers Public Internet sites for customers DATA SUPPLY DATA MANUFACTURING DATA CONSUMPTION Statistical Tools for data scientists Marketing Sales Delivery DATA MANAGEMENT & GOVERNANCE Manage Information Production Process from data Creation to Usage
  • 38.
    26.3.2021 SIILI CraftCon /Embrace Change 38 Two things to remember!
  • 39.
    Thanks for yourattention! QUESTIONS? Sami Laine Competence Lead, Siili Solutions Oyj President, DAMA Finland ry Program Committee Co-Chair, MIT CDOIQ Symposium https://www.linkedin.com/pub/sami-laine/2/a61/970

Editor's Notes

  • #9 We still don’t know how many people have the virus, how many are hospitalized, how many are in intensive care units, and how many are on ventilators. There is poor data on testing availability, and testing results are too often incorrect, delayed, or not counted. Contact tracing, necessary for avoiding community spread of coronavirus, lacks both the needed data and the human or technological resources to use it. Kasvava numerotieto ei läheskään aina lisää ymmärrystämme. Etenkään jos emme tiedä, mistä se on peräisin, jolloin on vaikeaa arvioida tietojen laatua, Nevalainen tiivistää. Luvuissa, tilastoissa ja niihin pohjautuvissa malleissa on tärkeää kyky paitsi tulkita data-analyysien tuloksia, myös arvioida niiden luotettavuutta.
  • #13 In the beginning, we did not know which country we are! We do not know the virus...
  • #15 Across the globe, we do not know yet which country we are! We do not know the virus...
  • #17 Currently, most of countries do not know which country they are! We do not know the virus...
  • #20 In reality, there is much more dimensions that should be considered than those above. We learn more about Covid-19 when we study more each variable and compare them to each others.
  • #22 In reality, there is much more dimensions that should be considered than those above. We learn more about Covid-19 when we study more each variable and compare them to each others.
  • #25 Collected data is relevant and valuable for a purpose and objective. This aims to increase VALUE FOR BUSINESS CUSTOMER.
  • #29 Meaningful contextual data helps TO MAKE VALUABLE CONCLUSIONS! YOU CAN act even before you have actual data!
  • #31 Collected and Transformed correctly to a specific use case. This aims to CUSTOMER NEEDS.
  • #32 You need to understand how data was created and how results were calculated to use them for valid conclusions!
  • #34 Data lifecycle is a series of phases that data passes through from creation, through use, to its destruction. Collected and stored correctly to represent reality. This aims to improve match with ORIGINAL REALITY.
  • #35 Data lifecycle is a series of phases that data passes through from creation, through use, to its destruction. Collected and stored correctly to represent reality. This aims to improve match with ORIGINAL REALITY.
  • #36 Master data, like patient number, combine fragmented data together. Transition events, such as laboratory tests, can be scattered across numerous systems and countries!