Medical Transcription: Identifying Errors, Protecting Patients
In April, we began a project at both the MT Tools Online and MT Inner Circle websites called “Identifying Errors, Protecting Patients.” This came about as we began to have discussions about the impact of the emerging technologies in the healthcare industry and some of the things that medical transcriptionists were seeing that caused concern. Today, let’s talk about some of those results so far.
It’s important to note that this was never intended to be a scientific study, but more of an anecdotal collection of information. What we hoped to capture was information about what is really happening “behind the scenes” in healthcare documentation. With the constant push to either have physicians enter their own data or to have speech recognition technology be “good enough,” it was time to take a look at what could happen if that were the case. And while we know there are things that come up every single day that MTs fix and/or flag so that the patient’s record will be correct, we really didn’t have any real life examples or data to share. A form was created and placed on the websites where MTs could enter data that showed these kind of potential errors, without providing any identifying information on the patients impacted.
In an industry that is mostly paid on production-based pay, I frankly wasn’t sure how many MTs would participate. It meant taking extra time in the day to add to the list, time that might otherwise be spent making money. The response has been great. MTs from 19 different states have participated in adding information to this list so that the story can be told. And we do have a compelling story.
In a period of about three months, over 400 entries were added to the collection. The choices for error types were: dictated left/right inconsistency, discrepancies/inconsistencies, lab value errors, medication errors in dosage, medication errors in name of drug, patient demographics, and speech-recognition (SRT) errors.
What We Found
In looking at the types of errors, here’s what we see:
Dictated left/right inconsistencies: 1.92% of total
Discrepancies/inconsistencies: 7.93% of total
Lab value errors: 2.64% of total
Medication Error in dosage: 2.88% of total
Medication Error in name of drug: 3.13% of total
Incorrect Information: 11.78% of total
SRT Errors 69.71% of total
While SRT errors represent a large percentage of what was reported, one thing that stood out as I look at the information is that within that category, 24% of those errors are medication errors, many that could cause serious problems for the patient.
Here are some examples from the errors that were reported:
Dictated left/right inconsistency: This is the type of error where a physician may start a report saying it’s the patient’s left leg and switch in the middle of the document to say it’s the right leg.
Discrepancies/Inconsistencies: Some of the things listed in this category include:
“The hospital course consisted of advancement of the diet. Diet has been a combination of diarrhea and cramps.”
“SPECT myocardial infarction” instead of “SPECT myocardial perfusion.”
“Final diagnosis: Sleep apnea and so on.” (Yes, that’s what was actually dictated.)
“She developed statin therapy and was discontinued off her Lipitor.” (Should have been statin myopathy)
A report where in one place it said the patient’s father was deceased, and in another said the father was living.
Penicillin listed as an allergy and then prescribed as a discharge medication.
Lab Value Errors included such things as “creatinine 138” instead of “creatinine 1.38.” In another example, the same report contained two conflicting values for the CPK, 3.42 and 172.
Medication Error in Dosage: These are as dictated:
Fosamax 70 mg q. day (while 70 mg is a dosage, that’s a weekly dosage, not daily).
Lasix 400 mg
Metoprolol 500 mg (the dosages I see for this one are 12.5, 25, and 50).
Medication Error in Drug Name: These errors included things like the physician dictating simvastatin instead of Synthroid, Neutron (which the physician spelled) for peripheral neuropathy instead of Neurontin, and even this sentence “probably the gout could be stepped up” (instead of colchicine).
Patient Demographics/Incorrect Information:
This category shows some of the things we often see where the patient was not identified at all, no medical record number given, no date of service dictated, or even the wrong medical record number keyed into the system but the patient’s name dictated which means someone has to catch that. This type of error runs the risk of the wrong information being placed in the wrong patient’s record.
In the category of incorrect information, we see some of the following examples:
“Patient lives in the same house with her daughter in a separate small apartment. Her daughter is married; as well, her daughter is 10 years younger than her.” (The last part made me realize something was wrong, called dictator, he clarified all references should be “sister.”)
“History of hysterectomy for breast cancer.” (Should be history of fibroids)
“9-month-old” (should be 9-year-old)
Speech Recognition Technology
When we started this project, I was contacted by someone who discouraged including SRT errors in this report, with the rationale that “those are edited and the engines learn from it and get better.” I made the decision to include SRT errors for a couple of reasons:
- This is a huge percentage of the errors we see in clinical documentation.
- There is a push in the healthcare industry to accept what is generated through SRT as “good enough” for clinical documentation. In some presentations, it has been suggested that using SRT and Natural Language Processing (NLP) will allow information to be automatically data tagged and entered into the EHR to meet meaningful use criteria.
- Vendors are still attempting to sell the use of SRT with claims that it can do away with transcription costs and therefore pay for itself.
If this happens in health care, I worry about what our documentation will become. With the first phrase being what was actually dictated and the second being what SRT produced, take a look at a few examples:
“Informed consent was obtained” became “informed consent for suicide was obtained.”
“Piriformis” became “para 4 minutes.”
“Indications for procedure” became “indications for seizures.”
“Lexiscan stress test” became “Mexican stress test.”
“Procedure explained in detail to the patient” became “procedure explained in detail to the uterus.”
“She had copious purulent secretions” became “she had alcohol abuse.”
Medication Errors in SRT
These are pretty serious in my opinion. As I think about the potential for technology to create tags that automatically enter things into the electronic record, what I see is the potential for these things to then become orders that create prescriptions or medication orders for patients. And yes, we can all hope that the clinical decision support systems will catch things like this, however, I see many physicians talking about not using those systems or becoming “fatigued” with all of the pop ups such that they ignore them. So, let’s take a look at a few medication errors created by SRT:
Lovaza became Flonase. This patient will still have high cholesterol, but maybe their allergies will improve!
Tramadol became Trileptal. Clearly a medication for epilepsy isn’t going to help this patient’s rheumatoid arthritis very much.
Glyburide became Namenda. These two aren’t even close in sound, and so the patient with diabetes ends up getting treatment for Alzheimer’s.
Ertapenem became metoprolol. This patient will end up with controlled blood pressure, but it sure won’t cure that bacterial infection.
Aricept became Percocet. In this case, the patient may still have dementia symptoms, but they won’t be in pain.
Haloperidol became Allopurinol. For this patient, they probably won’t have gout symptoms, but their psychosis sure won’t improve.
This list goes on and on and is concerning when it comes to patients receiving proper medications to treat their illnesses.
Then There’s the Funny and Absurd
We’ve all seen those bloopers that come from SRT errors. Here are just a few that I thought might give you a chuckle. Fortunately I’m not an artist, but I can imagine some fun graphics for some of these.
“A bolus of meat was present in the esophagus” came out “A bolus of stool was present within the esophagus.”
“Biceps tenosynovitis” became “biceps penis and synovitis.”
“Getting iron on hemodialysis” became “getting high on hemodialysis.”
“He is noting an interest in trying Viagra” became “He is noting an interest in trying vagina.”
“Intrauterine pregnancy” became “in the urine pregnancy.”
“Left labia minora” became “Left labia menorah.”
“Oligospermia” became “Olympic sperm.”
“Processed meats” became “prostatectomy.”
“The patient is stable from a neurological standpoint” became “The patient is stable from a tickle standpoint.”
“The patient should not operate heavy machinery” became “The patient should not operate heavy missionary.”
“The patient was encouraged to continue dieting” became “The patient was encouraged to continue dying.”
What’s The Answer?
The electronic health record is here to stay. We are dealing with it as MTs as well as patients. It’s critically important that the information contained in healthcare records be accurate. The technology just doesn’t exist to make that happen today. Here are a few things I believe we need to do to be proactive about this:
That’s a few ideas from me. Now let’s chat about this. What ideas do you have to spread the word about this important issue and where do we start?
Filed under: Clinical Medicine
Like this post? Subscribe to my RSS feed and get loads more!