Our medical patients are becoming increasingly complex. Patients come in with multiple preexisting problems and are taking many more medications than ever before. Electronic medical records (EMR) systems are being adopted across the country, with hopes that they will be able to help us organize and sort through the information in the patient’s charts.
The Journal of the American Medical Association published a study today titled Automated Identification of Postoperative Complications Within an Electronic Medical Record Using Natural Language Processing.
The study compared using ICD-9 codes, (a system where each diagnosis is given a specific code [e.g. 714.0 for rheumatoid arthritis]) to the use of free text search to identify adverse events that occurred while the patient was in the hospital. The study showed that free-text search was able to identify more cases of acute renal failure than ICD-9 codes, and that the two methods were similar in identifying the other diagnoses included in their study (venous thromboembolism, pneumonia, sepsis, and postoperative myocardial infarction).
Many of the EMR systems that I have used personally have had limited functionality for searching within a patient chart. The EMR used by the VA in the above study, Computerized Patient Record System (CPRS), has had a search feature for some time.
While free text search will certainly be helpful in identifying adverse events after patients have left the hospital (and thus, help us learn to prevent further complications), it would also be helpful for clinicians currently working with patients in the hospital and clinics.
Why would advanced search function be helpful to clinicians? Because good patient care lies in the details.
We would be able to quickly and easily find (in a chart that might be a few inches thick if printed out):
- The details of how a specific diagnosis was treated over time by going specifically to each note where that diagnosis was mentioned (e.g. making it easier to see that my patient with lupus nephritis may have never really gotten a good trial of MMF, before abandoning it as a treatment option) .
- Specific details of side effects to medications (e.g. is this antibiotic listed as an allergy because of a rash and shortness of breath, or because of stomach upset?)
- Finding the specific details of how a diagnosis may have evolved over time (e.g. such as how a patient with a scleroderma actually presented as a polyarticular inflammatory arthritis).
- Details of why certain treatments were chosen at specific times (e.g. finding out that Dr. X chose a certain medication over another because of abnormal liver function tests at the time).
- Fix our misspellings, or better yet, improve search accuracy by finding those misspellings within the patient record.
- Suggest similar or related searches.
- Search only records from a certain timeframe, or from certain physicians.
What about the problem of searching outside medical records — the records from another clinic that are scanned into the system? I see no reason that these cannot be searched equally as well as note that are written directly into the EMR sitting in front of you. Evernote, a free online service that I use to scan and organize documents, receipts, etc., offers the capability to search documents that I have scanned, and works extremely well.