JOUGLA PraguesDer

European study on causes of death quality and comparability : 

European study on causes of death quality and comparability Automating coding systems for mortality data June 3-5, 2004 – Prague, Czech Republic E Jougla, G Pavillon INSERM – CépiDc - France Epidemiological Center on Medical Causes of Death

Slide2: 

Summary 1. European Union data on causes of death statistics 2. Investigations on data quality and comparability 3. Differences due to the coding process

Slide3: 

1. European Union data on causes of death statistics Summary

Slide4: 

* numbers by 5 years age group * raw death rate and standardised death rate (total & 0-64 yrs) Dissemination at national and regional level (Eurostat – New Cronos) Causes of death data

Slide10: 

”AVOIDABLE MORTALITY“ (Males <65 yrs 1994-1996)

Slide12: 

FOR FRANCE, PRIMARY PREVENTION AS A PRIORITY IN HEALTH PROGRAMS

Slide13: 

BUT NECESSITY TO ASSESS THE INTERNATIONAL COMPARABILITY AND QUALITY OF THE DATA

Slide14: 

2. Investigations on data quality and comparability Summary

Slide15: 

Eurostat Task Force on Causes of Death Statistics (TF/COD) since 1997

Slide18: 

Recommendations - General causes of death certification - Coverage - Confidentiality - Query practices - Training practices

Slide20: 

Summary

Slide21: 

Specific international coding studies * Set of death certificates from different countries * Certificates coded by national offices and by a WHO reference centre

Slide22: 

Specific international coding studies Investigation rare - First study (1965 – 6 countries - 1000 certificates) - cancer (1978, 1986) - diabetes (1989)

Slide23: 

International coding studies - Cancer (1978,1986) * Random set of 1243 US death certificates mentioning cancer * Nine countries coded the 1243 death certificates Percy et al. Am J Epidemiol 1989, 129 : 934-946.

Slide24: 

International coding studies - Cancer (1986) Differences in the underlying cause of death with US coders at the 3-digit level Differences % W Germany 16 France 15 USSR 15 Netherlands 12 England 10 Brazil 9 New Zealand 8 Canada 6 Percy

Slide25: 

Country International coding studies - Cancer (1986) France 96 3 1 England 92 6 3 W Germany 92 6 3 New Zealand 91 8 8 Brazil 91 6 2 Netherlands 91 6 3 USSR 90 8 2 Canada 89 7 4 Circulatory diseases (%) Other diseases (%) Cancer (%) Underlying cause coded (certificates mentioning cancer) United States 88 7 5

Slide26: 

Country International coding studies - Cancer (1986) England 151 145 -4% Netherlands 146 142 -3% France 140 128 -9% New Zealand 140 137 -2% W Germany 137 131 -4% Canada 134 133 -1% Percy Corrected rates Variation Published rates Comparison of published death rates with “corrected” rates

Slide27: 

International coding studies - Diabetes (1989) * Random set of 200 death certificates mentioning diabetes from 9 EC countries * The 1800 certificates recoded by a WHO reference centre Jougla et al. Int J Epidem 1992, 21:343-351.

Slide28: 

Possibility to apply the ICD General Rule (%) International coding studies - Diabetes (1989) Country > 4 conditions reported (%) 96 N Ireland 14 92 Scotland 10 86 Netherlands 28 84 Rep Ireland 18 64 France 32 60 W Germany 54 52 Switzerland 36 36 Belgium 8

Slide29: 

Country Belgium 68 72 0.9 France 43 37 0.9 Rep. Ireland 41 34 0.8 Netherlands 35 22 0.6 Switzerland 28 30 1.1 Scotland 24 22 0.9 W Germany 24 30 1.2 N Ireland 7 18 2.5 Jougla Reference coding (b) (%) Ratio (b/a) National Coding (a) (%) Proportion of certificates with diabetes coded as the underlying cause International coding studies - Diabetes (1989)

Slide30: 

Country Diabetes mortality rates per 100 000 (1989) Netherlands 26 31 Switzerland 21 21 Rep Ireland 20 17 Belgium 17 24 W Germany 16 17 France 12 12 Jougla Male Scotland 11 9 N Ireland 5 3 Female

Slide31: 

* Impact of use of different ICD revisions Bridge coding “ICD9 / ICD 10”

Slide32: 

ICD 9 - ICD 10 ICD 10 Date

Slide33: 

Objective  To document the changes Example ”Bridge Coding - France“ - ICD9 + Manual coding - ICD 10 + Automatic coding (Styx) * Comparison based on the short 65 Eurostat list (compatible ICD9 - ICD10) Methods  * Sample of 53 000 deaths in 1999 (1 out 10)

Slide34: 

Bridge coding - ICD9 / ICD10 by chapters Sample 1/10 – 1999 - France

Slide35: 

Bridge coding - ICD9 / ICD10 by chapters Sample 1/10 – 1999 - France

Slide36: 

Cancer – Death rates - France 1990-1999 (ICD9) - 2000 (ICD10)

Slide37: 

Pneumonia - Death rates - France 1990-1999 (ICD9) - 2000 (ICD10)

Slide38: 

* To improve between-country comparability of coding Automating Coding Systems * To document changes (coding, ICD revision or updates)

Slide39: 

Automatic coding systems in EU

Slide40: 

ACS Eurostat reports (Pavillon, Johansson et al.)

Slide41: 

IMPROVEMENT OF THE INTERNATIONAL COMPARABILITY AND QUALITY OF THE CERTIFICATION PROCESS

European study on causes of death quality and comparability : 

European study on causes of death quality and comparability Automating coding systems for mortality data June 3-5, 2004 – Prague, Czech Republic E Jougla, G Pavillon INSERM – CépiDc - France Epidemiological Center on Medical Causes of Death

Slide43: 

Differences France / UK (Males <65 yrs – 1999)