Biases in Electronic Health Record Research
Author Information
Author(s): Hripcsak George, Knirsch Charles, Zhou Li, Wilcox Adam, Melton Genevieve B
Primary Institution: Columbia University
Hypothesis
Large-scale electronic health record research introduces biases compared to traditional manually curated retrospective research.
Conclusion
The study found that while a naïve approach to using electronic health records can approximate results, significant errors can lead to distorted outcomes.
Supporting Evidence
- The electronic health record data can provide useful information despite its challenges.
- Errors in cohort selection and characterization can lead to significant outcome changes.
- Manual review of cases is crucial for improving data accuracy.
- Narrowing the cohort can yield better results but may introduce bias.
Takeaway
This study shows that using computer records to study pneumonia can be tricky because mistakes in the data can change the results a lot.
Methodology
The study reused data from a previous pneumonia study, applying a classification system to define a cohort and calculate mortality rates.
Potential Biases
Errors in selecting and characterizing the cohort can significantly affect outcomes.
Limitations
The study's results may be biased due to the narrowing of the cohort, which could exclude important cases.
Participant Demographics
Patients with community-acquired pneumonia, including various age groups and health conditions.
Statistical Information
Confidence Interval
95%
Want to read the original?
Access the complete publication on the publisher's website