Identifying periods of acceptable computer usage in primary care research databases

Area of Science:

• Primary care research
• Health informatics
• Epidemiology

Background:

• Primary care databases are crucial for epidemiological research.
• Data quality, particularly defining accurate patient follow-up periods, significantly impacts research findings.
• Existing methods for defining data start dates may introduce bias.

Purpose of the Study:

• To evaluate the impact of different data quality filters on disease incidence and prescribing trends.
• To validate a novel method for determining adequate computer usage periods.
• To improve the reliability of primary care research databases.

Main Methods:

• Defined Acceptable Computer Usage (ACU) based on record entry frequency.
• Compared ACU and Acceptable Mortality Recording (AMR) with registration and computerization dates.
• Calculated incidence rates for various outcomes (e.g., antibiotic prescriptions, myocardial infarction, cancers) using different start date definitions in The Health Improvement Network (THIN) database.

Main Results:

• Acceptable Computer Usage (ACU) dates were, on average, 3.3 years after Acceptable Mortality Recording (AMR) dates.
• Using AMR or ACU dates resulted in incidence trends more aligned with external data compared to registration or computerization dates.
• Early time period incidence rates showed high sensitivity to the chosen start date definition.

Conclusions:

• The latest of Acceptable Mortality Recording (AMR) and Acceptable Computer Usage (ACU) dates enhance the integrity of The Health Improvement Network (THIN) data.
• This validated method improves the reliability of longitudinal data analysis in primary care research.