To use data mining techniques on the electronic medical record (EMR) in the Emergency Department (ED) with the intent of improving care while reducing costs.
All patient records for a 6-month period were examined. All records with an initial complaint related to shortness of air or related were extracted for a total of 1329 patient records. There were 53,000+ charges ordered for these patients in the ED. The data mining techniques of text analysis, transactional time series, and association rules were used to examine the data.
A total of 187 patients were discharged to ICU after the ED, with 25 of the 187 originally triaged as non-urgent. In contrast, only 2 out of 626 discharged home were triaged as emergent. Although there is no standardization in the language of initial patient complaints, there is a general grouping of ten different complaints as listed in the table. Bronchitis and asthma require a similar amount of time in ED treatment; COPD requires an extra 100 minutes of treatment time on average compared to the other two conditions. Patients complaining of shortness of air were tracked into a heart protocol, receiving tests of Troponin and CBC; or were tracked into a respiratory protocol, receiving oxygen and mini-nebulizer treatment. There was 87% confidence that glucose testing was associated with EKG and CBC, a 70% confidence that it was associated with Troponin test, but almost no association with SOA treatments.
Data mining the EMR can provide useful information that can be used to improve care while reducing costs. There is sufficient variability in treatment for patients with similar complaints that the variability can be examined and reduced while developing optimal treatment patterns.
Once the variability in treatment is documented, it can be examined to develop standardized protocols for patient complaints. Optimal paths of care can be developed. Specific protocols can also reduce the time needed to treat patients in the ED.
Patricia Cerrito, Grant monies (from sources other than industry) NIH grant #1UC1HS014897-01, ED Information Systems-Kentucky & Indiana Hospitals, Dave Pecoraro, PI, Patricia Cerrito, co-PI.