Reliability of Diagnostic Coding in ICU Patients
Author Information
Author(s): Misset Benoît, Nakache Didier, Vesin Aurélien, Darmon Mickael, Garrouste-Orgeas Maïté, Mourvillier Bruno, Adrie Christophe, Pease Sébastian, de Beauregard Marie-Aliette Costa, Goldgran-Toledano Dany, Métais Elisabeth, Timsit Jean-François
Primary Institution: Fondation Hôpital Saint-Joseph, Université Paris-Descartes
Hypothesis
The poor reproducibility of medical diagnoses observed in administrative databases is also found in research databases.
Conclusion
The coding of medical diagnoses varied between different observers, which could limit the interpretation and validity of research and epidemiological programs using diagnoses as inclusion criteria.
Supporting Evidence
- The average number of diagnoses coded per patient was 4.6.
- Only 34% of primary diagnoses were matched by both external coders.
- 18% of all codes were selected by all three coders.
Takeaway
Doctors in the ICU often don't agree on the same diagnosis for patients, which can make it hard to trust the data used for research.
Methodology
The study analyzed 100 randomly selected medical records from the Outcomerea ICU database, comparing diagnoses coded by the treating physician with those recoded by two independent senior physicians.
Potential Biases
The external coders worked in different hospitals and may have had varying expertise, potentially influencing their coding methods.
Limitations
The study had a small sample size and coding was performed by physicians rather than trained administrative coders, which may affect reliability.
Participant Demographics
Patients admitted to 12 French ICUs between 1998 and 2004.
Statistical Information
P-Value
p<0.0001
Confidence Interval
95% CI 25% to 43%
Statistical Significance
p<0.0001
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website