METHODOLOGICAL ISSUES AND RESULTS OF THE TRANSITION TO ICD10 IN WEST GERMANY

1
METHODOLOGICAL ISSUES AND RESULTS OF THE
TRANSITION TO ICD10 IN WEST GERMANY

• Marketa Pechholdova
• Charles University (Prague), INED (Paris)
• ? pechholdova_at_gmail.com

FRONTIERS AND CHALLENGES IN THE STUDY OF HUMAN
LONGEVITY THE 2nd HMD SYMPOSIUM, 13-14th June
2008 ROSTOCK
2
OUTLINE

• ICD system in general
• ICD10 and its potential impact on cause-of-death
  data comparability
• Case study transition to ICD10 in absence of
  bridge-coding
• Data
• Applied method
• Encountered problems (and solutions)

3
ICD SYSTEM IN GENERAL

• Exists as of 1893
• Regular (cca 10 year) revisions
• ICD classification itself
• coding rules
• definitions
• tabulations lists
• format of medical death certificate

4
ICD10 SPECIFICS

• Largest change of mortality statistics in 50
  years
• Main changes are
• Coding scheme 4-digit alphanumeric
• Number of categories from 5000 (ICD9) to 8000
• Chapters added and rearranged
• Cause of death titles changed and regrouped
  (some between ICD chapters)
• Coding rules new rules for direct sequels,
  improbable sequences, primary cancer sites,
  ill-defined causes

5
ICD10 TRANSITION HELPERS I

• Transition helpers ICD Translator
• bridge coding studies
• WHO ICD Translator (1997)
• Very first official WHO list of symmetrical
  correspondences between two ICD revisions
• Full 4-digit ICD9/ICD10 list
• DIMDI Crosswalks for ICD9 and ICD10 (1995)
• German correspondence list between ICD9/10
• After update, gt90 identical to ICD Translator

6
ICD10 TRANSITION HELPERS II

• Transition helpers ICD Translator
• bridge coding studies
• Bridge coding studies
• Double coding exercise on a sample/totality of
  deaths of one year
• The only actual guideline for identifying the
  impact of the coding changes

7
BRIDGE CODING STUDIES
8
COMPARABILITY RATIOS (Ci)BY ICD10 CHAPTERS
9
CASE STUDY WEST GERMANY

• ICD10 adopted in 1998
• Transition to ICD10 performed on a 3-digit level
• ICD9 data continuous series of mortality by
  cause of death between 1968-1997 (previous work)
• ICD10 data for 1998-2006, but problem with
  territorial coverage

10
CLASSIFICATIONS TERRITORY
ICD8
HMD females
HMD males
11
DATA TO WORK ON
Reconstruction source data females
Reconstruction source data males
12
FIRST ATTEMPT ICD TRANSLATOR

• First idea direct use of the ICD Translator to
  reclassify ICD9 into ICD10 by
• Transforming ICD Translator into 3-digit level
• Adapting it to the German ICD lists
• Automatically detecting mutual correspondences
• NO SATISFACTORY RESULTS
• 292 associations First association gathered 702
  (53) out of 1333 items, second gathered 100
  (13) - need of manual approach anyway

13
TRANSITION METHOD

• A posteriori double reclassification, based on
  ICD Translator and the DIMDI Crosswalks
• STEP 1 Creating the smallest possible clusters
  of deaths with similar medical content
  elementary associations
• STEP 2 Calculate the transition coefficients
• STEP 3 Inspect the series, make additional
  corrections to solve unusual issues (i.e.
  irregularities of the age structure, coding
  changes, classification changes etc.

14
ELEMENTARY ASSOCIATIONS
15
SELECTED DISEASES
16
SEPTICAEMIA expected increase
BEFORE REDISTRIBUTION
AFTER REDISTRIBUTION
ICD9
ICD10
A40
038
A41
17
PNEUMONIA expected decrease
BEFORE REDISTRIBUTION
AFTER REDISTRIBUTION
ICD9
ICD10
J18
486
485
18
HEPATITIS interchapter exchange
BEFORE REDISTRIBUTION
AFTER REDISTRIBUTION
ICD9
ICD10
571 Chronic liver disease and cirrhosis
B18 Chronic viral hepatitis
570,572-573
19
NEW CATEGORY MESOTHELIOMA
BEFORE REDISTRIBUTION
AFTER REDISTRIBUTION
ICD9
ICD10
C45
C45
20
OTHER NON-STANDARD CASES

• Newly added category C97 (Malignant neoplasms of
  independent (primary) multiple sites)
• no correspondence in ICD9
• according to bridge coding proportional
  redistribution into cancer of well defined site
• Transport accidents change of the classification
  logic
• ICD10 by the type of injured person X ICD9 by
  nature of accident
• Reclassification not possible

21
DISCUSSION

• Presented results, focused on the problematic
  ICD10 categories, have not shown significant
  fluctuations for West Germany - artifact of the
  inertia of the coding practices?
• Bridge coding can serve as a reference for some
  cases (ex. new items)
• Analysis on the 4-digit ICD level would improve
  the exactitude of the method