Despite many advancements in the technology, transcription is still an imperfect science. Computer software cannot discern accents and unique speech patterns, and transcriptions done by individuals are subject to human error. While this would be a cause for annoyance and lost business in the legal or computing world, the stakes are raised considerably higher in medicine, where a transcription mistake could result in a career-ending malpractice lawsuit and even a loss of life. Here are five high profile medical transcription errors.
1. Sharron Juno
Many hospitals outsource their medical contracting duties, for reasons of cost-efficiency and expediency. But a contractor in India made a critical mistake in transcribing the notes of the doctor treating Sharron Juno, a diabetic. Instead of 8 units of insulin, Juno received a lethal dose of 80. Juno died less than two weeks later, and the hospital was successfully sued for $140 million.
The investigation into the cause of death took Juno’s family’s lawyer from Alabama to cities all across the United States, before the trail took him to Mumbai, India. The hospital saved two cents per line by contracting their work to a transcripting agency, who themselves sub-contracted their work out to India.
2. Natural Language Processing
As a way of reducing the likelihood of mistakes as in the Sharron Juno case (and liability in the event that something goes wrong), many hospitals are turning to electronic health records to automate patient documentation, and keep it in-house.
EHR uses natural language processing, software that attempts to automatically transcribe physician dictation to a patients’ record, eliminating the need for a human transcriber. However, there are still errors as the software struggles to decipher the nuances and subtleties of human speech. The problem is even further exacerbated when two doctors provide conflicting diagnoses. A physician’s assistant relates the example of a natural language processor in an EHR interpreting a doctor’s diagnosis of a patient’s condition as being one type of heart attack. However, within the same record, a cardiologist felt the condition was the result of a different type of heart failure. The PA cited this discrepancy as one example of the challenges of turning medical transcription over to electronic means.
The problem of using speech recognition software for medical transcribing was starkly made evident at Princess Margaret Hospital in Ontario, where 23% of automated breast imaging reports contained at least one significant error. By comparison, reports created by human transcribers contained at least one critical mistake in only 4% of reports. A doctor who wrote the study that investigated the errors pointed out that speech recognition software can sometimes fail to distinguish between such words as ‘centimeter’ and ‘millimeter’. In breast imaging, where the errors were detected, the difference in measurement is critical to a patient’s health.
The Sharron Juno case is just one example of the inherent risks that come from outsourcing medical transcriptions. Both hospitals and transcription agencies do it because it saves them money, and the agencies’ overseas arms (usually in India), promise a 24-hour turnaround and a low charge, with the quality of the actual transcript a negotiable commodity.
A blogger who once edited and administered a medical transcription agency writes of one transcriber reporting “phlebitis in his leg” as “flea bite his last leg”.
5. Catching Mistakes A Little Too Late
When a woman came into the ER at a particular hospital, the doctor seeing her felt that she was simply drug-seeking, and dictated that she was “addicted” and the medicine was “making her sicker”. The recording was sent to India, where it was transcribed and sent back to the United States.
The finished transcript read that the patient was not “addicted”, but “dick-chick”, and “making her sicker” was rendered as “sleep with a girl”. The errors were not caught until the patient complained to the hospital about her treatment, at which point she was made aware of the errors.