Computer-based coding of free-text job descriptions to efficiently identify occupations in epidemiological studies

Area of Science:

• Occupational health
• Epidemiology
• Bioinformatics

Background:

• Accurate occupational classification is crucial for identifying health risks in epidemiological research.
• Manual coding of job titles to Standardized Occupation Classification (SOC) codes is labor-intensive and prone to errors.
• Automated methods are needed to streamline this process for large-scale studies.

Purpose of the Study:

• To develop and validate SOCcer, a novel algorithm for computer-assisted assignment of SOC-2010 codes.
• To improve the efficiency and reliability of occupational coding in epidemiological research.
• To assess the performance of SOCcer against expert-assigned codes and exposure estimates.

Main Methods:

• Developed job title, task-based, and industry-based classifiers.
• Trained a logistic model using 14,983 expert-assigned jobs to derive empirical weights.
• Validated SOCcer in two datasets by comparing assigned SOC codes and lead exposure estimates.

Main Results:

• SOCcer achieved 44.5% agreement at the 6-digit and 76.3% at the 2-digit SOC level for 11,991 case-control study jobs.
• Agreement improved with higher algorithm scores, enabling identification of assignments needing review.
• Lead exposure estimates showed good agreement (κ 0.6-0.8) between SOCcer and manual assignments, though performance varied for specialized job descriptions.

Conclusions:

• SOCcer offers a significant improvement in the efficiency of occupational data coding for large epidemiological studies.
• While some manual review may still be necessary, automated coding substantially aids research.
• This tool facilitates better integration of occupational exposures into health research.