Descripción de diversas herramientas para el análisis de datos masivos (Real World Data) de las Comorbilidades en la Historia Clínica Electrónica (HCE). Ponencia invitada en el Summer School de la Universidad de Barcelona (UB), 5 julio 2019. Hospital Clínico de Barcelona.
Herramientas de Análisis de Comorbilidades en la HCE (SummerSchool UB)
1. Comorbidity is associated
with worse health outcomes,
more complex clinical
management, and increased
health care costs
Valderas 2009
2. Miguel-Angel Mayer MD MPH PhD
Research Programme on Biomedical Informatics (GRIB)
Institut Hospital del Mar d’Investigacions Mèdiques (IMIM)
Universitat Pompeu Fabra (UPF)
Analytical tools for discovering
comorbidity patterns in
patients with chronic diseases
3. Co-multi-
morbidity
definitions
‘The presence of co-existing or
additional diseases with reference
to an initial diagnosis or with
reference to the index condition
that is the subject of study’
‘Comorbidity may affect the ability
of affected individuals to function
and also their survival; it may be
used as a prognostic indicator for
length of hospital stay, cost
factors, and outcome or survival’
Medical Subject Headings (MESH), 1989
Multimorbidity 2018 (complex interactions)
4. Co-multi-
morbidity
definitions
‘The presence of co-existing or
additional diseases with reference
to an initial diagnosis or with
reference to the index condition
that is the subject of study’
‘Comorbidity may affect the ability
of affected individuals to function
and also their survival; it may be
used as a prognostic indicator for
length of hospital stay, cost
factors, and outcome or survival’
Medical Subject Headings (MESH), 1989
Multimorbidity 2018 (complex interactions)
7. The availability of Electronic Health Records
(EHRs) for data mining offers the opportunity
to discover disease associations and
Comorbidity Patterns from the clinical
history of patients gathered during routine
medical care
9. ü gain a better understanding of
the aetiology of the diseases
ü identify new disease subtypes
& patient stratification
ü determine more effective and
safer treatments
ü enable preventive medicine
Because we can...
10. 2 General Hospitals
2 Mental Health Care Centres
1 Social-Healthcare Centre
Barcelona Area
Population: 325.000 inhabitants
Parc Salut Mar Barcelona
11. IMASIS and IMASIS-2
EHR
PSMAR
(IMASIS)
IMASIS-2
Parc de Salut Mar Research
Reports,
Aggregated
Data &
Analysis
Patient identifiable
information
Data anonymization process
(pseudo) Anonymized data
Pacurariu A. et al. Electronic healthcare databases in Europe: descriptive analysis of characteristics and potential
for use in medicines regulation. BMJ Open 2018; vol 8(9)
12. Parc Salut Mar
EMIF Catalogue
Platform Information
Framework
EHR
EMIF
European Medical
Information Framework
EHR4CR
IMASIS
EHR of Parc de Salut Mar
Integrative Biomedical Informatics - GRIB (IMIM-UPF)
13.
14. What if we have some
specific tools to analyse
comorbidities?
15. Tools to identify comorbidity
from RWD
comoRbidity R package
Comorbidity4j and Comorbidity4web
Temporal patterns on disease
trajectories
Tools to investigate molecular
causes of disease comorbidity
comoRbidity R package
Guildify web server
Tools to enable
comorbidity
identification and
analysis
19. • Local computer execution, guaranteeing data
protection standards
The comoRbidity package enables us:
20. • Selecting the dataset to use for the
comorbidity study under consideration
Hospital medical records Cohorts/registries
The comoRbidity package enables us:
21. comoRbidity R package: depression & cancer
Female patients in IMASIS: 367,041 (47.67%)
Patients suffering from the index disease(s): 36,389
Number of comorbidities: 44
Male patients in IMASIS: 402,947 (52.33%)
Patients suffering from the index disease(s): 26,742
Number of comorbidities: 36
Mayer MA. et al. Using Electronic Health Records to Assess Depression and Cancer Comorbidities.
Stud Health Technol Inform. 2017;235:236-240
23. Pair of diseases (depression and cancer) with
significant p-value (Fisher exact test)
Mayer MA. et al. Using Electronic Health Records to Assess Depression and Cancer Comorbidities.
Stud Health Technol Inform. 2017;235:236-240
25. Temporal
directionality
• In some cases, as in stomach, lung, and bowel
cancers in women, there is a preferred order in
the disease onset where depression occurs
previously to these cancers
• In other cases, the diagnosis of depression occurs
following the diagnosis of cancer such as breast
cancer in women and prostate cancer in men.
26. Open-source Java tool to perform systematic
analyses of comorbidities
Support for Observational Medical Outcomes
Partnership Common Data Model (OMOP CDM)
data format
Interactive browser-based validation of input data
Interactive web visualizations to explore and
refine results
27.
28.
29.
30.
31.
32. The inclusion of the time dimension into disease
comorbidity studies has been crucial for achieving a better
understanding of the disease progression and studying its
temporal characteristics
Temporal patterns in patient disease
trajectories
Giannoula A, Gutierrez-Sacristán A, Bravo Á, Sanz F, Furlong LI. Identifying
temporal patterns in patient disease trajectories using dynamic time warping:
A population-based study. Scientific Reports. 2018 Mar 9;8(1):4216.
33. Time ordered patient disease trajectory
t1:2004 Coronary atherosclerosis
t2:2005 Pers hist malign neopl of bladder
t3:2007 Diabetes with ketoacidosis
t4:2010 Presenile dementia with delirium
t5:2011 Pneumonia, unsp.
Common Disease Trajectories
d1 d2 d3
time
N. Patients = 120
N. Patients = 230
N. Patients = 1500
d2 d3
d4 d3
Disease trajectory: an
ordered series of diagnoses
where the diagnoses were
observed in the patients in a
specific order
34. Common disease trajectories
clustering
Cluster 1 Cluster 2
Group disease
trajectories based on:
• Similar time patterns
• Similar diseases
- They can be used as a basis for predicting the most
probable next steps in diseases progression
- To reveal complex time-dependent disease patterns
35. Lessons learned
• It is essential to have a thorough understanding of
the EHR (types of data available or how the data
were collected)
• The interpretation of the analysis requires a
comprehensive knowledge of the scientific basis
of the diseases studied
• The limitations of these approaches need to be
taken into account
Integrative Biomedical Informatics - GRIB (IMIM-UPF)
36. Conclusions
§ The benefits of aggregating
larger sets of routinely collected
clinical data are well
documented and of great
societal benefit
§ These data sets probably are
not going to answer all possible
clinical questions but will
contribute to optimize research,
processes and timelines
Integrative Biomedical Informatics - GRIB (IMIM-UPF)
37. Integrative Biomedical Informatics Group
Prof. Ferran Sanz (GRIB director)
Dr. Laura I. Furlong (IBI Head)
Dr. Miguel-Angel Mayer
miguelangel.mayer@upf.edu
@mmayerp