Clinical code usage in UK general practice: a cohort study exploring 18 conditions over 14 years.

Area of Science:

• Primary care informatics
• Clinical coding analysis
• Health data quality

Background:

• Diagnostic Read codes are crucial for primary care data. Understanding their usage patterns and diversity is essential for accurate health record analysis.
• Previous research has not extensively explored the frequency and diversity of Read code usage for a wide range of conditions over extended periods.

Purpose of the Study:

• To assess the frequency and diversity of diagnostic Read code usage for 18 common conditions in UK primary care.
• To examine changes in Read code usage patterns between 2000 and 2013.

Main Methods:

• A population-based cohort study utilizing data from 684 UK general practices in the Clinical Practice Research Datalink (CPRD GOLD).
• Analysis included canonical correlation analysis for frequency ranking and Shannon entropy index, richness, and evenness for diversity quantification.
• Examined incident and total clinical codes for conditions including asthma, diabetes, hypertension, cancer, and severe mental illness (SMI).

Main Results:

• Read code usage showed a positive monotonic correlation for most conditions, with high 5-year correlations for hypertension (HT), hypothyroidism, osteoarthritis, and SMI.
• Code usage diversity remained stable throughout the study period.
• Cancer, diabetes, and SMI exhibited the highest richness, while AF, hypothyroidism, and LD had lower richness; persistence varied, with SMI and hypothyroidism codes lasting up to 5 years, and cancer, diabetes, and LD codes lasting up to 2 years.

Conclusions:

• The diversity of Read code usage for most conditions remained stable, indicating an under-reported research area.
• Generated mental health code lists demonstrate longer persistence compared to cardiometabolic conditions and cancer.
• Improving primary care data quality necessitates more consistent and less diverse coding practices; future research should consider the transition to SNOMED CT.