Diapositiva 1

Report
Building a framework for integrated use of
health information system and clinical
data in epidemiology
R. Di Domenicantonio
Workshop
Challenges for epidemiology in the context of National Health Service
Rome, 15-16 October 2012
Case definition / identification
Classification of acute miocardical infarction from clinical findings
AHA Scientific Statement. Case Definitions for Acute Coronary Heart Disease in Epidemiology and Clinical
Research Studies. Circulation. 2003;108:2543-2549
Case definition / identification
through Health Information System
IN HOSPITAL DEATH WITHOUT:
DEATH
2006-2009
410 AMI
411-414 CHD
411-414 CHD
HOSPITAL ADMISSION IN PAST 3 YEARS
410-AMI
412-OLD MYOCARDIAL INFARCTION
FIRST
CORONARY
EVENT
HOSPITAL
DISCHARGE
2006-2009
410 AMI
DISCHARGED ALIVE WHITHIN 3 DAYS
SECONDARY DIAGNOSIS:
412-OLD MYOCARDIAL INFARCTION
HOSPITAL ADMISSION IN PAST 3 YEARS :
410-AMI
412-OLD MYOCARDIAL INFARCTION
Case-definition validation
A samples of patients admitted to hospital in 2003 validated according to
the American Heart Association criteria
Barchielli et al. Hospital discharge data for assessing myocardial infarction events and trends, and effects of diagnosis
validation according to MONICA and AHA criteria. J Epidemiol Community Health. 2012 May;66(5):462-7.
Source of case identification
Incidence of Hypertension by deprivation
Hospitalization database
Physician billing database
Aubé-Maurice J et al. Divergent associations between incident hypertension and deprivation based on different
sources of case identification. Chronic Dis Inj Can. 2012 Jun;32(3):121-30.
Population registries
INSURED
55 health district in 12 LHA collects
access, changes,
delection in
regional health service
insured population
• Anagraphic data
• General practitioner data
• Residence (municipality - updated ?)
Lazio Region
population
(from 2006)
• Address1
• Census trait
3 mounth
update
DEP
Georeference
SES indicator
by census
trait
Mortality
information
system
death
RESIDENT
20 administrative
district collect
residence
and
anagraphic
data
Municipality of
ROME registry
office
6 mounth
update
• Anagraphic data
• Residence / address
(actual, stand. coded)
• Census trait
• Life status
Municipality
of ROME
population
(from 1997)
Health / Population Information System
Population
cohorts
EMERGENCY
Anagraphic
data
HOSPITAL
ADMISSION
OUTPATIENT
population
registries
MORTALITY
DRUG PRESC
HIS Code
(Anonimous)
Limits due to privacy legislation
EXEMPTIONS
Tracking of
health
care
contacts
Occurence:Incidence of first coronay event in Lazio
Region, Years 2006-2009
Standardized Rate X 100.000 residents
Incidence rate ratio (ref. Regional)
Lazio Region
ROME
Quintile of rate
Rome Municipality
Relative risk
Adherence to evidence-based drug therapy
Cohort of AMI patients followed 12 months after hospitalization in Rome
Kirchmayer U et Al. Socio-demographic differences in adherence to evidence-based drug therapy after hospital discharge
from acute myocardial infarction: a population-based cohort study in Rome, Italy. J Clin Pharm Ther. 2012 Feb;37(1):37-44.
Primary care outcome indicator
Among population with diabetes:
Rate of diabetes related complication by area of residence
Age, gender standardized Rate (X 1,000)
Composite measure:
• Diabetes Short-Term Complications
(ketoacidosis, hyperosmolarity, coma)
• Diabetes Long-Term Complications
(renal, eye, neurological, circulatory)
• Lower-extremity amputation
Health Information Systems (H.I.S.)
Pros
Cons
Extensive coverage
Inaccuracy of information
Time
Inconsistency of coding over time
Cost
Exclusion of people with no contact
with health services
Integration of all patient’s data
Homogenity (?)
Information reflects reimbursement
policies from NHS
Integration of HIS and clinical data
TWO APPROACHES:
A priori case definition, based on literature review
– Validation on
Clinical data (only sensitivity)
Population data (sensitivity & specificity)
Panel approach
– Building of predictive models based on clinical data of subject
with ascertained diagnosis
– Validation (sensitivity & specificity)
Inflammatory bowel disease (IBD)
Sensitivity of the case finding algorithm in respect to a panel of Crohn’s
disease patients by year of diagnosis and source
Case def. / identification:
•Discharge (ICD9 555)
•Exemption register
•Years 2000-2009
Panel data:
•Patients from
gastroenterology wards
of five hospitals
Sensitivity
Year of
diagnosis
Patients
Before 2006
1,980
82.1
76.2
39.0
2006
109
88.1
76.2
61.5
2007
97
88.7
78.4
63.9
2008
102
81.4
72.6
57.8
2009
70
68.6
51.4
45.7
Total
2,358
82.2
75.4
42.1
Overall Hospital Exemptions
Chronic obstructive pulmunary disease (COPD)
Lazio Region prevalence comparison between different case definition
A-priori case definion: from
literature review, knowledge
Survey (question.)
Predictive model
Prevalence %
Predictive model: from health
care consumption patterns of
panels (2) of clincal
ascertained patients
General Practitioner
A-priori
Case identification:
•Hospitalization
•Drug prescription
•year 2007
Age class
Belleudi V. et al : Definition and validation of a predictive model to identify patients with Chronic Obstructive
Pulmonary Disease (COPD) from administrative databases. Epidemiol Prev. 2012 May;36(3-4):162-71
Comparison between GP and HIS for Diabetes
and COPD
Prevalent patient identified in HIS and in 49 general practitioner data.
Lazio Region, years 2006-2008
Case definition:
• Diabetes
– Discharges
– Exemptions
– Drug prescriptions
• COPD
– predictive model
COPD
DIABETES
GP
HIS
Is it possible to identify heart failure patients
using administrative database?
Development and validation of a predictive model using information
from HIS and clinical data from population with and without disease
(PREDICTOR study)
Bootstrap
PREDICTOR
patients
(HF +) (HF -)
Administrative
data
Ambulatory
setting
patients
(HF +)
Control group
subjects
(HF -)
Predictive
model
HF
associated
Cut off
factors
Internal
validation
External
validation
Results
• Age, consumption of several drug (furosemide, diuretics,
antianemic preparations), previous hospitalizations
(HF, myocardial infarction, other heart diseases) are factor
associated with COPD
• Model showed high specificity (96.8%) and low sensitivity
(34.2%) (although higher than only hospitalization)
• More appropriate to select cohorts of severe patient than to
estimate occurrence
Added value of clinical data
• Integration of clinical and HIS data is a useful approach to asses
validity of case-defining and to develop innovative model to
identify population affected by chronic disease
• Uncertainty / lack of information in administrative data invoke
their utilization to:
– Check accuracy of coding in claims to guarantee comparable
case-definition in different epidemiologic studies
– Gather addictional information to develop and validate severity of
disease measure
Added value of clinical data
Future challenge: MOH project ongoing !
– Objective: validate some commonly used case definitions,
including clinical stage classification, with respect to
information gathered by general practitioners
•
•
•
•
Diabetes
Coronary heart disease
Hearth failure
Hypertension
– unselected, highly representative population of control
Building a framework for integrated use of
health information system and clinical
data in epidemiology
R. Di Domenicantonio
Workshop
Challenges for epidemiology in the context of National Health Service
Rome, 15-16 October 2012

similar documents