Abstract
The authors describe their experiences creating technology to automatically capture facial images from patients during triage and registration for integration into the electronic medical record (EMR) to reduce data retrieval and data entry errors. The prototype system was tested across a variety of ethnicities with facial images captured successfully in 100% of cases with a median time to capture of 0.75 seconds.
Objective
Despite the use of electronic medical record (EMR) technology and computerized physician order entry (CPOE) systems, medical error remains pervasive. In particular, patient-identity errors are particularly persistent (i.e. technically accurate orders written for the wrong patient). Patient identity errors can account for as many as 35% of all medical errors [1,2,3,4,5,6]. Augmenting existing EMR textual identifiers with patient facial images may decrease patient identification errors. Facial images could be captured through manual, volitional processes - however, automated methods have several potential advantages including: faster throughput during triage or registration, consistency of image resolution and capture quality, and enhanced compliance by triage personnel. The authors share their experience building a prototype to automatically capture patient facial images directly into an EMR during triage or registration.
Methods
A system to automatically capture patient facial images into EMRs was created. The images are captured at approximately 15 Hz using the open-source Blepo computer vision library[7]. Facial detection is performed in real time on the images using the Viola-Jones algorithm[8]. Detected facial images are automatically placed into the triage clinician’s data entry screen in the EMR for selection.
Results
The prototype system was tested on a convenience sample of twenty-two adult, hospital employees representing a range of ages and ethnicities. Ethnicity of the sample population was 59% Caucasian, 14% African American, 14% Indian/Middle Eastern, and 14% Hispanic or mixed. Age ranged from 22 years to 57 years. Median age was 36 +/− 9. 100% of the facial images (22/22) were successfully captured. Median time for capture was less than 1 second (0.75 seconds).
Discussion
The prototype system successfully captures facial images in adult patients. Based upon these positive results, the system is currently being integrated into the enterprise-wide EMR system for testing in ED triage environments.
References
- 1.Rolland P. Occurrence of Dispensing Errors and Efforts to Reduce Medication Errors at the Central Arkansas Veteran’s Healthcare System. Drug Saf. 2004;27(4):271–82. doi: 10.2165/00002018-200427040-00004. [DOI] [PubMed] [Google Scholar]
- 2.Callum JL, Kaplan HS, Merkley LL, Pinkerton PH, Rabin Fastman B, Romans RA, Coovadia AS, Reis MD. Reporting of near-miss events for transfusion medicine: improving transfusion safety. Transfusion. 2001 Oct;41(10):1204–11. doi: 10.1046/j.1537-2995.2001.41101204.x. [DOI] [PubMed] [Google Scholar]
- 3.Ballesteros S, Ramon F, Martinez-Arrieta R, Larrotcha C, Cabrera J. Ten years of latrogenic intoxications from the Spanish poison control center. Vet Hum Toxicol. 2003 Mar;45(2):93–4. [PubMed] [Google Scholar]
- 4.Morishita J, Katsuragawa S, Kondo K, Doi K. An automated patient recognition method based on an image-matching technique using previous chest radiographs in the picture archiving and communication system environment. Med Phys. 2001 Jun;28(6):1093–7. doi: 10.1118/1.1373403. [DOI] [PubMed] [Google Scholar]
- 5.Blado ME, Tomlinson A. Monitoring the accuracy of a PACS image database. J Digit Imaging. 2002;15(Suppl 1):87–95. doi: 10.1007/s10278-002-5029-6. Epub 2002 Mar 21. [DOI] [PubMed] [Google Scholar]
- 6.Ito H, Yamazumi S. Common types of medication errors on long-term psychiatric care units. Int J Qual Health Care. 2003 Jun;15(3):207–12. doi: 10.1093/intqhc/mzg038. [DOI] [PubMed] [Google Scholar]
- 7.Blepo computer vision library. http://www.ces.clemson.edu/~stb/blepo/
- 8.Viola P, Jones MJ. Robust real-time face detection. International Journal of Computer Vision. 2004;57(2):137–154. [Google Scholar]