Electronic health records can be a more efficient way to pinpoint cases for epidemiological studies, according to a study published this month in BMC Medical Informatics and Decision Making.
The information in EHRs is often reused for large research studies, such as investigating the association between drugs and possible adverse events. However, the first step in conducting such research is to identify which patients have the illness or event to be studied, which usually is accomplished through time-consuming manual review. Moreover, it has been hard to use EHRs for such case identification, as typically that occurs through free text narratives, which is more difficult to data mine for information.
The researchers hypothesized that by using a random sampling strategy and particular keywords, they could enable EHRs to automatically identify cases from the free text. They tested their theory--and found success--using a large Dutch EHR database and two data sets, one for acute renal failure and one on hepatobiliary disease.
"An automatic case-identification system with high sensitivity and reasonable specificity can be used as a pre-filter to significantly reduce the workload by reducing the amount of records that needs to be manually validated," the authors said. "Using manual validation on the reduced set instead of the set retrieved by the broad query could save weeks of manual work in each epidemiological study."
EHR data mining holds great promise for many large-scale projects, such as studying population health, identifying patients in need of follow up care, and reducing patient flow problems in hospitals.
To learn more:
- read the study