Presentation: Reconstructing Cause-of-Death Series

Slide 1

  • Applying a Reconstruction Method to Cause-of-Death Series in Different Countries Svitlana Poniakina, France Meslé, Denys Dukhovnov, Magali Barbieri, Yui Ohtsu, Jacques Vallin
  • XXX International Population Conference (IPC 2025)
  • 13-18 July 2025, Brisbane, Australia

Slide 2

  • Plan: 1. Introduction (2 mins) 2. Method (4 min) 3. Example - case study (4 mins) 4. Analysis perspectives (2 mins) 5. HCD and team work (1 minute)

Slide 3

  • Variations in mortality are the result of the interaction between pathologies and human efforts to combat them
  • To understand the evolution of mortality, we need to follow the evolution of its profile by medical cause of death
  • As early as the mid-19th century - an international classification of causes of death.
  • 1.Introduction

Slide 4

  • Bertillon’s nomenclature in use for Parisian statistics for causes of death, was adopted as the 1st International Classification of Diseases (1893: ICD-0)
  • Periodic revisions of ICD ensure that the classification reflects new practices and knowledge
  • International Classification of Diseases

Slide 5

  • Classification Revisions
  • The nomenclature is rapidly used by many cities and countries
  • Since its adoption, it was revised 11 times

Slide 6

  • The same structure but with considerable internal disturbance
  • 1893
  • CIM-0
  • 14 divisions
  • 203 items
  • 1989
  • CIM-10
  • 21 chapitres

  • 10 000 items

  • at each revision items contents were deeply updated
  • new items were created for taking into the account the medical progress or the extension of the fields of utilisation of the Classification
  • Classification Revisions

Slide 7

  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • Classification Revisions: Heart Diseases

Slide 8

  • Answer 3: Certain infectious and parasitic diseases?

Slide 9

  • There is no single principle dictating how diseases subdivide
  • For example:
    1. Neoplasms – according to behaviour (malignant, benign)
  • then primary, secondary, ill-defined
  • then by anatomic location
    1. Endocrine disorders – according to anatomy
    1. Infectious diseases - according to infectious agent
  • Most suitable medical premise: anatomy, aetiology, pathology

Slide 10

  • Revisions make impossible to follow trends in specific causes of death
  • Revisions if ICD complicate data linking across time
  • Exchanges between the items are very complex
  • After World War 2, no explanation is given by WHO about the reasons of the changes
  • What to do ?

Slide 11

  • Use the results of double classifications carried out at the time of the classification change
  • Yes, for the few countries where they exist
  • But cannot be used for other countries
  • Use double classification a posteriori :
  • Methode Ined (France Meslé et Jacques Vallin)
  • Solutions:

Slide 12

    1. Method
  • Step 1: Constructing Correspondence Tables
  • Step 2: Identifying Fundamental Associations
  • Step 3: Double Coding and Transition Coefficients
  • A-posteriori corrections

Slide 13

  • Step 1: Constructing Correspondence Tables

Slide 14

  • 1:1 association
  • N:N association
  • Correspondance table
  • From correspondence matrix to elementary associations
  • Correspondances matrix
  • ICD9
  • ICD10

Slide 15

  • Types des associations :
  • 1:1 - direct (tc=1)
  • N:1 - grouping (tc=1)
  • 1:N - split (tc proportional)
  • N:N - complex (tc ?)
  • Step 2: Identifying Fundamental Associations

Slide 16

  • Example of associaton type N:N

Slide 17

  • Step 3: Calculating transition coefficients (by age group and sex)
  • Cross-tables
  • Coefficients

Slide 18

    1. Case Study - Alzheimer’s Disease

Slide 19

    1. Case Study - Alzheimer’s Disease
  • Year of transition to ICD-10:
  • France: 2000
  • Spain: 1999
  • Japan: 1995

Slide 20

  • Fundamental association - France

Slide 21

  • Fundamental association - Spain
  • F03 - Unspecified dementia

Slide 22

  • Fundamental association - Japan

Slide 23

  • Fundamental associations
  • France - direct transition, G30 corresponding neatly to 331.0
  • Spain - Alzheimer’s disease was intertwined with mental and behavioral disorders. The associations include both neurological and psychiatric diagnoses (e.g., F01, F03, and F051)
  • Japan - associations spanned multiple ICD chapters, including respiratory conditions (J690), senility (R54), and accidental inhalation (W78–W79)

Slide 24

  • Result of Reconstruction - France

Slide 25

  • Result of Reconstruction - Spain

Slide 26

  • Result of Reconstruction - Japan
  • ICD9 ICD10

Slide 27

  • Some Principles of Reconstruction:
  • IPC 2025
  • Country-specific reconstruction is essential
  • Reconstruct into the most recent version of ICD
  • Reconstruct at the highest level of detail possible.
  • Disaggregate by sex and age group

Slide 28

  • Conclusions
  • IPC 2025
  • The observed differences in Alzheimer’s mortality trends are not solely driven by epidemiology.
  • They also reflect changes in diagnostic preference, reporting norms, and ICD structure.
  • Without reconstruction, comparisons across countries or over time would be biased or misleading.

Slide 29

  • Historical mortality trends: France

Slide 30

  • Joining HCD
  • Cause-of-death data have been reconstructed for 19 countries -> Human Cause-of-Death Data Series (HCD) / now fully integrated into the Human Mortality Database (HMD)
  • We are happy to guide all the researchers who would like to work on reconstruction of cause-of-death data with future integration of these data into HCD
  • If you are interested in collaborating, please feel free to contact me at: svitlana.poniakina@ined.fr