Transcript Register research approaches

Register-based research in
the Nordic countries
Mika Gissler
Nordic School of Public Health, Gothenburg, Sweden &
THL National Institute for Health and Welfare, Helsinki, Finland
Why good possibilities
to register-based studies?
• Traditions: population statistics have been collected
more than 250 years and health statistics more than
150 years in the Nordic countries.
• First real registers were started in the 1940-1950s,
when improved computers were available: health
care personnel, cancer register.
• Personal identification numbers since 1960s.
• Several data quality studies have shown the high
quality of routinely collected registers.
• Data protection allows research use of register data.
Important registers
in the Nordic countries
•
•
•
•
•
•
•
•
•
Cancer register
Registers on infectious diseases
Hospital discharge registers
Cause-of-death registers
Birth and malformation registers
Register-based Census
Health care quality registers
Prescription registers
Hospital outpatient registers
1940s
1950s
1960s
1960s
1960s
1990s
1990s
1990s
1990s
Unique registers and data
in the Nordic countries
• IVF (in vitro fertilization) register, Denmark
• Register on induced abortions and sterilisations,
Finland
• Register on visual impairments, Finland
• Register on breast and cervical cancer screening,
Finland
• Multiple generation register, Sweden
• Multiple generation studies in the Norwegian
Medical Birth Register
• Biobanks in all Nordic countries + possibilities to
link them to other registers.
Important registers for studies in
psychiatry and mental health
•
•
•
•
•
•
Hospital discharge registers
Cause-of-death registers
Pension Registers
Register-based Census
Prescription registers
Hospital outpatient registers
1960s
1960s
1960s
1990s
1990s
1990s
Examples of register-based studies in
psychiatry and mental health
• Register-based studies:
– Cross-sectional studies
– Trends
– Longitudinal studies
• Combination of data from different sources:
– Medical records
– Questionnaires
– Biobank material
Example 1: Life expectancy
among psychiatric patients
• Registers:
– THL: Hospital Discharge Register 1980-2003
– Finnish Centre of Pension: Pension Register 1980-2003
– Statistics Finland: Cause-of-Death Register 1981-2003
• Data:
– The data included 361 898 persons aged 15 years or more
– 17 638 persons with dementia and 2 630 with intellectual
disability were excluded
• Life expectancy at 15 years and for ages 15-64 years
were calculated separately by using Wiesler's method.
Life expectancy increased
80
70
60
50
40
30
Finland
Psychatric patients
20
10
Men
Women
2001-03
1996-00
1991-95
1986-90
1981-85
2001-03
1996-00
1991-95
1986-90
1981-85
0
Conclusions
• Life expectancy at 15 years has increased among
Finnish population with hospital discharge or pension
due to mental disorders between 1981 and 2003:
– Finland: +3.5 years, psychiatric patients +5.8 years
• F30-39: +10 years, F40-49: +8 years, F20-29: +6 years, but
• F10-19: -0.6 years
• Risk for death
– diseases and medical conditions
– external causes and poisoning
2-fold
6-fold
• Similar results from other Nordic countries.
Example 2: Maternal smoking
and children’s F-diagnoses
• Registers:
– THL: Medical Birth Register 1987-1989
– THL: Hospital Discharge Register 1987-2007
– Social Insurance Institute: Reimbursed psychotropic
medicine 1994-2007
– Statistics Finland: Cause-of-Death registers 1987-2008
• Data:
– Children born in 1987-1989, excluding perinatal deaths,
multiples, and children with major congenital anomalies
– Final study population: 175 869 children (94.4%)
Risk for adverse psychiatric
outcomes by maternal smoking
2.50
2.00
1.50
Inpatient care
Outpatient care
1.00
Death
Psychotropic medication
0.50
0.00
No
smoking
< 10 cig
> 10 cig
Crude
No
smoking
< 10 cig
> 10 cig
Adjusted
Adjusted by maternal age, parity, sex, gestational age, birth weight, 5 minute Apgar score and
maternal psychiatric diagnosis before birth.
Conclusions
• Children exposed to maternal smoking has an
increased risk for receiving a F-diagnosis in inpatient
or outpatient care in childhood and adolescent.
• The increased risk can be observed for all diagnosis
excluding schizophrenia and anorexia.
• Register studies cannot confirm the real effect of
smoking.
– However, a recent local study in Turku has shown that
prenatal smoking exposure is associated with smaller
regional brain volumes in preterm infants (Ekblad et al., J
Pediatrics 2009).
Example 3: Use of psychotropic
drugs and pregnancy outcomes
• Registers:
– The ‘Drug and Pregnancy’ -database 1996-2006, to be
annually completed 2007 onwards
• Data:
– All births in the Medical Birth Register
– All induced abortions in the Abortion Register
– All congenital anomalies in the Malformation Register
• Use of prescribed & reimbursed drugs (Social Insurance Institution)
– 3 months before pregnancy
– during pregnancy
– 3 months after pregnancy
The use of psychotropic medicine before
the pregnancy starts
• The Drug and Pregnancy -database 1996-2006:
– Total 622 671 births and 117 229 induced abortions
– Excluded: induced abortions due to fetal reasons
– Separate analysis: first pregnancies
• All drug purchases 3 months before pregnancy
were used as a proxy measure of mental health
disorders.
Conclusions
•
Measured by the use of psychotropic medicine,
women’s pre-existing mental health status is worse
for women having an induced abortion than for
women giving a birth.
– All pregnancies: Adjusted OR 1.94 (95% CI 1.87-2.02)
– First pregnancies: Adjusted OR 1.56 (95% CI 1.44-1.68)
– Highest risk for women using hypnotics and sedatives,
antipsychotics and antidepressants.
•
This essential confounding factor should not be
neglected when investigating the occurrence of
pregnancy-related mental health problems.
Example 4: Mothers’ and children’s longterm follow-up after substance abuse
during pregnancy
• Basic data:
– 524 women followed-up prenatally at special out-patient
clinics and a control group of 1792 women matched for
maternal age, parity, time and place of delivery.
• Registers:
– THL: Medical Birth Register, Hospital Discharge Register,
Child Welfare Register
– Statistics Finland: Cause-of-Death Register
– Social Insurance Institution: Information on prescribed
medicine, social benefits, pensions and rehabilitations
Mothers’ outcome, %
•
•
•
•
•
•
•
Death
F-diagnosis, inpatient
F-diagnosis, outpatient
Intoxication care
Pensions, any cause
Rehabilitation, any cause
Special reimbursement
– Psychosis
• Drug reimbursement N05
• Drug reimbursement N06
Cases
Controls
8.0
46.0
47.1
41.3
16.8
9.5
27.0
0.2
3.6
8.3
1.8
2.2
5.6
18.4
***
***
***
***
***
***
***
10.9
1.4 ***
71.4
68.1
20.9 ***
26.5 ***
Children’s outcome, %
•
•
•
•
•
•
•
•
•
Death
F-diagnosis, inpatient
F-diagnosis, outpatient
Care benefit for sick child
Rehabilitation, any cause
Special reimbursement
Drug reimbursement N05
Drug reimbursement N06
Child taken into custody
Cases
Controls
1.4
7.1
8.1
25.0
5.1
12.1
9.1
4.6
46.0
1.0
2.8
2.6
13.9
2.4
11.1
5.6
1.4
2.4
NS
***
***
***
***
NS
**
***
***
Conclusions
•
•
•
Combination of medical records and
registers was feasible, even though it was
difficult to get all the necessary permissions.
Women with substance abuse displayed
significant long-term abuse-related
morbidity and mortality, rehabilitation, early
retirement, and use of prescribed medicine.
Also their children had increased morbidity,
rehabilitation, and use of prescribed
medicine, and almost half of them were
taken into custody.
Why register research?
• Easy to form data:
– cross-sectional studies
– longitudinal studies (history, follow-up)
•
•
•
•
•
•
•
Easy to repeat the same study.
No limitations for sample size (rare cases --- total population).
Population-based studies feasible.
No need to contact patients.
Follow-up relatively easy.
No participation bias nor research bias.
No reporting bias.
Problems related to register research
• The data is unavailable
– primary health care, diseases and conditions not requiring a contact to
health care system, self-rated health, opinions, experiences,...
• Data protection: are such studies possible in general?
• Ethically controversial topics:
– abortion, miscarriage, infertility, malformations, psychiatric disorders,
family studies, contact to relatives of a death patient, genetics…
• High data costs: Statistical offices, Central Population Register
• Data overload syndrome
– Too much data, too little time…?
• Fishing:
– Easy to find statistically significant results, if the data is large.
Finally
• Register-based studies seems to be feasible, e.g. for
cross-sectional, longitudinal and trend studies
• Combination of data from other registers and from
other sources, such as medical records, questionnaires
and even biobank material is possible.
• Data protection questions have not been an issue, at
least until now.
• The lack of information from primary health care will
be solved after the national electronic patient journal
system is in use.
Promotion of register research
• Denmark: National Centre for Register-based
Research, Århus Universitet http://www.ncrr.dk/
• Finland: Finnish Information Centre for Register
Research http://www.rekisteritutkimus.fi/
• Norway: Special issue on register-based research in
Norsk Epidemiologi 14 (1): 2004.
• Sweden: Grants for register-based research by the
National Board of Welfare and Health (Socialstyrelsen)