Mental Health Treatment: What's wrong with it currently, and what does the future look like?
Mental Health Treatment
- Or why was it so hard to get my own diagnosis, why do I find my own treatment so confusing (medication, exercise, everything!), and why is ML in advertising but not in psychology.*
Firstly, are mental health diagnosis's from psychiatrists and psychologists accurate generally.
- They can be. But fundamentally there are two things doctors do diagnostically (and sometimes a third).
- Real life quizzes.
- Prod you with stuff and investigate.
MODERN PROBLEMS? MODERN SOLUTION TIME :D
Is the gold standard for diagnosis(DSM) actually a credible place to base ground-truth of mental health diagnosises.
- It's probably accurate for a large proportion of cases but over-generalizes.
Are medical treatments, like ritalin or SSRI's, effectively measured for there efficacy?
- To some extent research can determine where medical treatments are very effective.
- There is definitely a lot of unknowns around efficacy of different medical treatments.
Is there a better way of developing a model of an individuals neurotype using technology?
- Yes. Moodmap.
Is there research showing technological data sources can create accurate diagnosises
- Yes there is.
- Potential issues can be with the labelling of the data sets. i.e., diagnosis of radiological images via machine learning, but the labellers are radiologists, which injects a lot of error from the outset into the model.*
Specificity matters. It's better to be 100% certain the model has a diagnosis correct, rather than getting false positive. It's very rare to achieve 100% specifity though and if you were using a screen tool, you would sacrifice specifity because you would want greater sensitivity, so e.g., specifity may only be 20%.
To implement a tool like this requires a huge invasion of a persons privacy. Solid, essentially a new way of using personal data and sharing it, seems like the obvious way around this.
https://solid.inrupt.com/ Surely looking at an individual when diagnosing someone with a mental illness is the wrong approach altogether. A graph of the patient, with the people around the patient, is likely a far better diagnosis avenue. Harder to achieve obviously.