Using Electronic Patient Records to Discover Disease Correlations
Author Information
Author(s): Roque Francisco S., Jensen Peter B., Schmock Henriette, Dalgaard Marlene, Andreatta Massimo, Hansen Thomas, Søeby Karen, Bredkjær Søren, Juul Anders, Werge Thomas, Jensen Lars J., Brunak Søren
Primary Institution: Technical University of Denmark
Hypothesis
Can electronic patient records be used to discover disease correlations and stratify patient cohorts?
Conclusion
The study demonstrates that text mining of electronic patient records can enhance disease correlation discovery and improve patient stratification.
Supporting Evidence
- The study analyzed 5,543 psychiatric patient records from a Danish hospital.
- Text mining matched 218,963 text strings to ICD10 terms.
- Combining mined and assigned codes resulted in an average of 12.3 codes per patient.
- Manual curation flagged 93 interesting disease pairs.
- High precision rates of 87.78% for incidence and 84.03% for association were achieved.
Takeaway
Researchers looked at patient records to find connections between different diseases and group patients better based on their health information.
Methodology
The study used text mining techniques to extract clinical information from free-text notes in electronic patient records and combined it with structured data.
Potential Biases
Potential biases exist in the structured data due to reimbursement coding practices.
Limitations
The findings may be specific to the patient population studied and may not generalize to broader populations.
Participant Demographics
The patient population included 70% from the Copenhagen area, with 61% being males and an average age of 30 years.
Statistical Information
P-Value
1.17×10−3
Statistical Significance
p<0.01
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website