Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
The goodness of data: The Optum approach to cultivating high value data for oncology research Elenee argentinisADVANCED DATA AND ANALYTICS, LIFE SCIENCESOPTUM
A N A L YT I C S
Agenda
1. The needs of oncology researchers
2. Why Optum is positioned to drive transformation
3. Oncology enrichment at scale — going from breadth to depth
4. Questions
© 2019 Optum, Inc. All rights reserved. 2
5449
3734
29 28 26 2620
1714
9 72
47 45
3336 37
31 32
18
11
19
1013 12
5
192
177
62
165 201
89
179
55 34 7929
42 4020
563
206
10792
174
83 87 62 3571
30 21 19 3 0
100
200
300
400
500
600
0
10
20
30
40
50
60
Onc
olog
y
Neu
rolo
gy
Infla
mm
ator
y an
dau
toim
mun
e
Met
abol
ic a
ndG
I dis
orde
rs
Imm
unol
ogy
and
infe
ctio
us/v
acci
ne
Oth
ers
Car
diov
ascu
lar
Hem
atol
ogy
Opt
halm
olog
y
Res
pira
tory
Der
mat
olog
y
Neo
hrol
ogy
Rep
rodu
ctiv
e
Crit
ical
Car
e NO
. O
F M
AR
KE
TE
D /
PIP
EL
INE
DR
UG
S
NO
. O
F C
OM
PA
NIE
S
Companies with Pipeline DrugsCompanies with Marketed DrugsMarketed drugs
Oncology dominates pipeline of top 100 companiesThe movement to specialty areas continues
Biomed Tracker 2018 – top 107 biopharmaceutical companies by revenue
© 2019 Optum, Inc. All rights reserved. 3
How many of you are engaged in the care of oncology patients or in oncology research?(select all that apply)
a) Patient care
b) Oncology research
c) Both
d) Neither
CARE REIMBURSEMENT
Optum is positioned to support life sciences oncology research needs…We sit at the intersection of care and reimbursement
Deterministically linked
Pathology
Radiology
Care plans
Clinical assessments
Vital signs, labs
Signs, symptoms
Filled Rx
Reimbursed
Care utilizationSubmitted
claims
Optum clinical EHR and medical and pharmacy claims datasets
© 2019 Optum, Inc. All rights reserved. 5
How many of you are using real world oncology data?(select all that apply)
a) Using EHR data
b) Using claims data
c) Using genomics/molecular data
d) Using imaging data
e) Not using any
We have the building blocks for high quality research across cancers
Data breadthPopulation size
Continuum of care
Care outside the oncology clinic
Pre- and post-cancer encounters
(signs, symptoms, assessments, tests, procedures, labs)
Data quality(provenance and
traceability)
Good data for oncology research
Data depthCompleteness of cancer
specific variables
© 2019 Optum, Inc. All rights reserved. 7
How much oncology data does your organization license?a) We license one file for a specific cancer at a time (i.e., PDL-1)
b) We license a few data sets at a time (two to four mutation files)
c) We license the entirety of one or many data vendors datasets
d) We don’t license any data
BreadthThese populations include data from the entire patient journey
PRE- AND POST-CANCER CARE
Labs, procedures, biomarkers
Radiation therapy
Follow up care
BREADTH OF SPECIALTIES(e.g., for breast or ovarian cancer)
Signs, symptoms
Diagnosis/staging
Chemo-therapy
Clinical assessments
Surgery
Oral medication
Cardiology
Osteopathic/ orthopediconcology
Radiology oncology
Gynecology
Primary care
Surgical oncology
Medical oncology
Psychology
A significant portion of cancer patients get their pre-diagnosis screening and intervention outside the oncology clinic
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3752900/
© 2019 Optum, Inc. All rights reserved. 9
Example of the importance of continuum of careProstate cancer
Urology conducts more PSA tests and prescribes nearly as much as oncologists because they often treat early stages of prostate cancer.
PROSTATE CANCER DIAGNOSIS2016‒2018 | N=424493
Comparison of specialty in prostate cancer urology is highly engaged in both diagnosis and treatment
PHARMACOTHERAPYN=69,568
TESTSPSA, N=141,528
PROVIDER SPECIALTYUrology (N=26,814) | Oncology (N=7,106)
Distribution of pharmacotherapyUrology (N=6,589) | Oncology (N=8,475)
Distribution of testsUrology (N=27,309) | Oncology
(N=2,934)
Source: Optum EHR data (2007 – 2017)
© 2019 Optum, Inc. All rights reserved. 10
Melanoma Bladder cancer
Cancer patients care outside of oncology clinicPatients with melanoma and bladder cancer get up to 50% of their care from other specialties
Reference: Counts based on ICD 10 diagnostic codes for specific cancer and provider specialty, Optum EHR data 2018
© 2019 Optum, Inc. All rights reserved. 11
51%
21%
13%
8%
7%
Oncology and related
Dermatology
General phsycian
Surgery (misc)
Other specialties
49%
29%
17%
3%
2%
Oncology and related
Dermatology
General phsycian
Surgery (misc)
Other specialties
Top five other specialties who deliver supportive are• Plastic surgery• Otolaryngologist• Cardiology• Hematology• Ophthalmology
Oncology and related• Oncology 11.57%• Pathology 21.65%• Radiology 12.01%• Anesthesia 5%
Top five other specialties who deliver supportive care• Cardiology• Hematology• Rehabilitation• Psychiatry• Gynecology
Oncology and related• Oncology 5.69%• Pathology 9.26%• Radiology 14.84%• Anesthesia 11.18%
Optum oncology data quality supports publicationsOptum data has been used in many types publications
© 2019 Optum, Inc. All rights reserved. 12
Cancer types addressed (most common)• Breast, lung, pancreatic, prostate, glioblastoma, ovarian, thyroid,
renal, colon, multiple myeloma
Types of studies• Comparative effectiveness research• Outcomes research (i.e., overall survival studies have been done)• Algorithms to identify cancer types (i.e., pancreatic cancer patients)• Therapeutic choices in specific cancer types• Surveillance of patients specific gene mutations to assess
risk of developing prostate cancer
> 30 articles and abstractsOver the past two years
Optum oncology dataUsed for studies published in peer reviewed journals
© 2019 Optum, Inc. All rights reserved. 13
1 U.S. prevalence of endocrine therapy-naive locally advanced or metastatic breast cancer Breast cancer 2019 Apr
2 Analysis of Real-World Data on Overall Survival in Multiple Myeloma Patients With >/=3 Prior Lines of Therapy Including a Proteasome Inhibitor (PI) and an Immunomodulatory Drug (IMiD), or Double Refractory to a PI and an IMiD Multiple myeloma 2016 Nov
3 U.S. prevalence of endocrine therapy–naïve locally advanced or metastatic breast cancer Breast cancer 2019 Apr
4 Treatment regimens, treatment duration, and between-treatment duration in patients with small cell lung cancer across lines of therapy: An analysis of electronic health records data Small cell lung cancer 2018 Jul
5 Validation of algorithms to identify pancreatic cancer and thyroid neoplasms from health insurance claims data in a 10‐year follow‐up study Thyroid neoplasm 2018 Aug
6 Comorbidities and concomitant medication use in small cell lung cancer patients treated with third and later-lines of therapy: An analysis of electronic health records data Small cell lung cancer 2018 Jul
7 Incidence of pericardial and pleural effusions in patients with small cell lung cancer treated with third-line or later-line therapy: An analysis of electronic health records data Small cell lung cancer 2018 Jul
8 Incidence of pre-treatment and off-treatment pericardial and pleural effusions by line of therapy in patients with small cell lung cancer: An analysis of electronic health records data Small cell lung cancer 2018 May
9 A real-world evidence study to define the prevalence of endocrine therapy-naïve hormone receptor-positive locally advanced or metastatic breast cancer in the US Breast cancer 2017
10 Analysis of Real-World Data on Overall Survival in Multiple Myeloma Patients With ≥3 Prior Lines of Therapy Including a Proteasome Inhibitor (PI) and an Immunomodulatory Drug (IMiD), or Double Refractory to a PI and an IMiD Multiple myeloma 2016
11 Active Surveillance Use Among a Low-risk Prostate Cancer Population in a Large US Payer System: 17-Gene Genomic Prostate Score Versus Other Risk Stratification Methods Prostate cancer 2017
12 Comparison of Overall Survival Associated with Lenalidomide+Dexamethasone and Bortezomib+Dexamethasone Among Relapsed/Refractory Multiple Myeloma Patients: A Matched Analysis of Real World and Clinical Trial Populations Multiple myeloma 2017
What oncology enrichment at
SCALElook like
Optum oncology data — Going from breadth to depth
B R E A D T H P R E C I S I O N
Marketing insights
Health economics
Post market safety
Med affairs
Clinical trial planning and execution
Regulatory grade evidence
We can give life sciences the benefit of our breadth and tailor our data for depth
© 2019 Optum, Inc. All rights reserved. 15
Our approachEnrichment starts with tuning the data to answer specific questions
Which intervention is better?
ComparativeEffectiveness
& Quality
Who is the patient?
Study inclusion criteria:population health
What is the cost?
TCOCBurden of illness
Cost effectiveness
What is the risk?
Risk estimationRisk stratification
© 2019 Optum, Inc. All rights reserved. 16
Cancers with similar features allow us to scale so the staging construct and metastatic concepts shared by solid tumors will allow us to offer a suite of pan-oncology files using similar techniques adapted to each cancer.
0
10
20
30
40
Lung
Can
cer
Beas
tC
ance
r
Col
onC
ance
r
Pros
tate
Can
cer
Mel
anom
a
Blad
der/
Uro
thel
ial
Car
cino
ma
Ren
al C
ell
Car
cino
ma
Cen
tral
Ner
vous
Syst
em
Gas
tric
Can
cer
Panc
reat
icC
ance
r
Adva
nced
Solid
Tum
ors
Thyr
oid
Can
cer
Ova
rian
Can
cer
Hea
d an
dN
eck
SCC
Soft
Tiss
ueSa
rcom
a
Solid Tumors
78
147159
250
0
50
100
150
200
250
300
HematologicMalignancies
Solid Tumors
0
10
20
30
Mul
tiple
Mye
lom
a
Acut
e M
yelo
idLe
ukem
ia
Acut
eLy
mph
obla
stic
Leuk
emia
Chr
onic
Lym
phoc
ytic
Leuk
emia
Chr
onic
Mye
loid
Leuk
emia
B C
ell
Lym
phom
a/H
odgk
ins…
Mye
lody
spla
sti
c Sy
ndro
me
Non
-Hod
gkin
sLy
mph
oma
Follic
ular
Lym
phom
a
T C
ell
Lym
phom
a
Man
tle C
ell
Lym
phom
a
Smal
lLy
mph
ocyt
icLy
mph
oma
Diff
use
Larg
eB-
cell
Lym
phom
a
Hematologic Malignancies
The enrichment plan Sequenced starting with solid tumors
MarketedPipeline
Biomed Tracker 2018 – top 107 biopharmaceutical companies by revenue
© 2019 Optum, Inc. All rights reserved. 17
We can use clinical notes to take our data from broad to deep
140K providers More than 700 hospitals
and 7,000 clinics
4.5 billion free-text medical notes
74 million patientsMore than 45% have activity spanning more than 3 years
More than 30% have activity spanning more than 5 years
Optum Electronic Health Record Data 2007 - 2018
© 2019 Optum, Inc. All rights reserved. 18
Do you have an internal natural language processing team?
a) No we outsource it
b) We have a few dedicated employees
c) We have a fully staffed department
Architectural flexibilityFlexibility lets us extract and build models on top of the extracted entities
Structured dataDeliver new standardized variables that researchers can use
Modeling dataAdd analytics to create derived variables from combinations of extracted structured and unstructured fields
Extraction dataNLP pipeline extract the terms that define concepts and relationships between them: TNM, stage, grade, biomarkers, tumor characterization, metastatic references
Annotations• Annotators are subject matter experts
with clinical background• Annotations followed by a curation
(reconciliation) phase
Models• Models are constantly evaluated during
development process• Standard metrics such as PPV (precision),
Sensitivity (recall) and F-score are reported on a held-out test set
Output• Output is normalized to reduce variability and
as an additional step to prevent PHI leaks• Quality check are performed (volumetrics,
outliers, distributions, comparisons with available structured data, extrinsic evaluation, …)
© 2019 Optum, Inc. All rights reserved. 20
We’ve realized the power that linked data can bring to patient care
EHR/EMR
Medical claims
Devices
Storefront pharmacy
Genomics Imaging
Groceries, CPG
Media
SDOHsocio-econ
Diagnostics
Biomarkers
Pharmacy claims
Genetic panel reveals HER2 mutation
Buying cigarettes (told doc she quit)
Patient reports a lumpin her breast
Radiology confirms tumor size, lymph
nodes
Using apps for depression management
Jane SmithAge 57, BMI 22
Stage III Breast Cancer | HER2 positive ER/PR neg
Doesn’t owna car
© 2019 Optum, Inc. All rights reserved. 21
Now we can support our clients with enterprise linking
© 2019 Optum, Inc. All rights reserved. 22
Assemble a more holistic view of the patient to expand the set of questions that can be answered in health care
1. ProtectDe-identify datasets to protect patient privacy and reduce risk
2. LinkConnect matching patient records across datasets to increase data completeness and dimensionality
3. ExchangeSecurely exchange data in our open ecosystem as either a data owner or data user
Trial dataProvider populationsDigital apps
Genomics and molecular data
Imaging data
What can we explore linking for our clients?
© 2019 Optum, Inc. All rights reserved. 23
EHR and claims
EHR and claims
EHR and claims
How are we doing?We want your experience at Optum Forum to be exceptional. We value your feedback.
Please select this session in the mobile app and click on the survey button to complete a short review.
To provide feedback on working with Optum or our solutions, please visit Your Optum Experience on Level 1.
#OptumForum2019
Upcoming PresentationsJuly 31, Ironwood 32:15 pmLeveraging EHR and Claims Data to Predict Changing Drug Utilization TrendsConor Wyand, Director of Real World Evidence and Value, Optum Life Sciences