We live in the information age. More importantly, we live in the data age. We are wasting tons of electrcity to gather as much data as possible in order to fuel an industry which hasn’t even matured. This has created a ton of data labeling jobs such as clinical data labeling jobs for physicians.
Data Highways & Highways
A long time ago we started building roads and highways, years before we had enough cars on the road. The assumption was that we’ll eventually build enough cars to justify them.
So we built and built and connected a lot of cities, but also changed the landscape around us, not always for the better.
Now we’re collecting data, stashashing it in massive warehouses packed with hardware servers. To justfiy this we’re collecting data anywhere we can. But that’s just the beginning.
We need to label the data – that’s the hard part. That requires human input and that’s labor. We are using tons of people from overseas but you’re not going to be able to label clinical data in the US from physicians in India.
Complexity of Clinical Data Labeling
You can have all sorts of people label clinical data, from nurses to billing staff to scribes to medical students. But how good will the output be?
Imagine clinical images, such as plain film x-rays or MRI images. You want these labeled by top-notch radiologist. And then you have to do data validation of the labeled data.
Then there is audio labeling and chart labeling. That kind of clinical data labeling work has to be done by someone who really understands the clinical visits in the USA.
Data labeling is complex. When you listen to an audio conversation between a patient and a doctor the engineers must first piece the audio apart. The engineers design an interface for an ontology which represents all the clinical data extracted from the audio.
“Are you having any pain there as well?”
“Not much – it’s mostly below the belly button.”
The above example is far more complex than it seems. What does ‘there’ refer to? We need context from previous parts of the conversation. ‘below the belly button’ is what anatomical or symptom complaint? What’s ‘not much’ – is it non-existent, mild, or moderate in intensity?
Lack of Understanding of Medicine
I have done some consulting in this space of clinical data, building ontologies, data labeling, and creating logical relationships for machine learning processes. I would say that proper clinical data labeling is hard and the engineers who work on thse workflows often don’t understand the complexity.
If you don’t get the way medicine works, the way the patient-doctor visit progresses, and what the doctor is thinking then you’ll derive very little from the data labeling process.
This is what is happening right now. A lot of companies are hiring data labelers who aren’t physicians. Or they are physicians from other countries.
The data is getting labeled alright but the labels aren’t actually representative of the data. Imagine I give you a stack of color swatches to label but I only give you 10 colors to choose from. When you see a white swatch you’ll label it white – same with a darker white, a lighter white, a matte white, and so on.
Clinical Data Labeling Jobs
So where can you find clinical data labeling jobs?
How much can you get paid?
What’s the work like?
The work is tedious and it can seem repetitive. But, it’ll be remote work and you’ll only need a decent internet connection and a laptop.
The pay may be anywhere from $50 per hour to $150. Expect it to be far lower if you’re just labeling and aren’t offering any feedback to the data design team (the engineers).
You’ll have to search for health informatics or health information specialist. You can use keywords like clinical data labeling or data labeling healthcare in order to find such gigs.
There is no mainstream job site for this kind of work, yet. Just like how telemedicine blew up in 2016, clinical data labeling jobs for physicians will become abundant.
The Actual Work
You’ll likely sit in front of a computer screen and go to a website where you can connect the dots. As in, you’ll get a snippet of text, an image, or an audio portion, and you’ll be asked to link it to some already existing concept.
This is how most data labeling companies interact with their labelers. The work is repetitive in that sense. But what’s interesting about it is that it forces you to think in a whole different way about the patient-doctor relationship.
It’s also a reprieve from clinical medicine. It’s a good way to use your skills outside of the exam room or the OR.