Carbon monoxide-related injury estimation using ICD-coded data: methodologic implications for public health...

Area of Science:

• Public Health
• Epidemiology
• Medical Informatics

Background:

• National mortality data is used to estimate unintentional deaths from carbon monoxide (CO) poisoning.
• Accurate estimation of CO-related fatalities from International Classification of Diseases, 9th Revision (ICD-9) coded data presents challenges.

Purpose of the Study:

• To explore accurate methods for estimating CO-related deaths using ICD-9 coded U.S. mortality data.
• To evaluate different case ascertainment criteria and coding analyses for CO poisoning deaths.

Main Methods:

• Identified and classified CO-related ICD-9 codes into five certainty levels.
• Conducted single and multiple-cause-of-death analyses on 20 years of mortality data (1979-1998).
• Calculated sensitivity and positive predictive values for various case ascertainment criteria.

Main Results:

• Single-cause analysis using CO-exclusive E-codes was accurate but missed about one-third of CO deaths.
• Single-cause analysis with non-exclusive E-codes overestimated CO-related fatalities.
• Multiple-cause-of-death analysis, using the N986 nature-of-injury code, achieved 99.5% sensitivity for identifying true CO deaths.

Conclusions:

• Estimating CO deaths from ICD-coded data necessitates a deep understanding of coding rules and limitations.
• Multiple-cause-of-death analysis and specific injury codes (like N986) are crucial for accurate CO poisoning death ascertainment.
• Single-cause analysis alone is insufficient for comprehensive CO-related mortality estimation.