How can it be that humanity has achieved the monumental technical feat of putting a man on the moon, yet we still have not found a surefire way to make him happy? To me, the latter task seems far more important and far easier. These are the reasons why mental illness continues to ravage our society despite the miraculous rate of technical innovation in almost every other field of human endeavor.
Barrier 1: Professional Treatment is Unaffordable
Due to the high price of professional help, 28 million Americans are forced to bear the burden of their illnesses alone and without treatment. The cost to see a psychiatrist generally exceeds $100 per hour.
Solution: Prescriptive Algorithms and Therapeutic A.I.
Computers do not charge $100 per hour. We could provide empirical data regarding the effectiveness and dangers of specific treatments for individuals with similar symptomatological profiles. The accuracy and detail of this information would grow over time as prescriptive algorithms would constantly be refined through recursive analysis of new statistical data. Artificially intelligent psychotherapist psychotherapists could engage in machine learning, continually refining their user interactions. This would enable it to respond in a way that maximizes user happiness guided by recursive feedback algorithms.
Barrier 2: Clinical Research is Prohibitively Expensive
It costs no less than $1 billion dollars to bring an individual drug from the extremely well-functioning brain of a pharmaceutical innovator to the tortured brain of a victim of mental illness. Phase 3 clinical trials account for the majority of this cost. The per patient cost of these trials generally exceeds $26,000.
Solution: Crowdsourcing Research is Free
Millions of Americans are currently engaged in their own informal self-experimentation in an attempt to solve personal health problems. This experimentation involves small lifestyle changes related to physical activity, social interaction, diet, sleep, medications, supplements, meditation, and cognitive behavioral therapy. Unfortunately, almost all of the data from these experiments is lost to the ages. We need to make it easy to capture all of this precious data and facilitate observational studies. This data could reveal the currently unknown benefits of and adverse effects of countless treatments for free.
Barrier 3: Subjects in FDA Trials Are Not Representative of the Actual People Receiving Treatment
External validity is the extent to which the results can be generalized to a population of interest. The population of interest is usually defined as the people the intervention is intended to help.
Phase III clinical trials are designed to exclude a vast majority of the population of interest. In other words, the subjects of the drug trials are not representative of the prescribed recipients, once said drugs are approved. One investigation found that only 14.5% of patients with major depressive disorder fulfilled eligibility requirements for enrollment in an antidepressant efficacy trial.
As a result, the results of these trials are not necessarily generalizable to patients matching any of these criteria:
- Suffer from multiple mental health conditions (e.g. post-traumatic stress disorder, generalized anxiety disorder, bipolar disorder, etc.)
- Engage in drug or alcohol abuse
- Suffer from mild depression (Hamilton Rating Scale for Depression (HAM-D) score below the specified minimum)
- Use other psychotropic medications
These facts call into question the external validity of standard efficacy trials.
Furthermore patient sample sizes are very small. Number of subjects per trial on average:
- 275 patients are sought per cardiovascular trial
- 20 patients per cancer trial
- 70 patients per depression trial
- 100 per diabetes trial
Solution: Collect Quantified Self Data on Actual Patients
In the real world, no patient can be excluded. Even people with a history of drug or alcohol abuse, people on multiple medications, and people with multiple conditions must be treated. Only through the crowdsourcing of this research, would physicians have access to the true effectiveness rates and risks for their real world patients.
The results of crowdsourced studies would exhibit complete and utter external validity since the test subjects are identical to the population of interest.
Furthermore, self-trackers represent a massive pool of potential subjects dwarfing any traditional trial cohort. Diet tracking is the most arduous form of self tracking. Yet, just one of the many available diet tracking apps, MyFitnessPal, has 30 million users.
Tracking any variable in isolation is nearly useless in that it cannot provide the causal which can be derived form combining data streams. Hence, this 30 million user cohort is a small fraction of the total possible stratifiable base.
Barrier 4: Shortage of Mental Health Professionals
There are simply not enough psychologists and psychiatrists to meet societal demand. Almost 90 million Americans live in federally designated “Mental Health Professional Shortage Areas”.
Solution: Psycho-Therapeutic Software
Apps don’t have a three-month waiting list. Anyone with a computer could have immediate access to mental healthcare.
Barrier 5: Publication Bias in Industry-Funded Clinical Trials
Pharmaceutical companies that sponsor research often report only “positive” results, leaving out the non-findings or negative findings where a new drug or procedure may have proved more harmful than helpful. Selective publishing can prevent the rapid spread of beneficial treatments or interventions, but more commonly it means that bad news and failure of medical interventions go unpublished. A past analysis of clinical trials supporting new drugs approved by the FDA showed that just 43 percent of more than 900 trials on 90 new drugs ended up being published. In other words, about 60 percent of the related studies remained unpublished even five years after the FDA had approved the drugs for market. That meant physicians were prescribing the drugs and patients were taking them without full knowledge of how well the treatments worked.
Solution: We Must Make ALL Research Findings Publicly Available
The effectiveness rates for any and all drugs or other medical interventions would be made immediately available to everyone, regardless of outcome.
Barrier 6: Lack of Motivation to Carry Out Treatment is a Primary Symptom of Depressive Illness
It’s a rather tragic catch 22 that the only people who have a true appreciation of the pain of mental illness usually also have a fundamental impairment in areas of the brain that support motivation, energy, and hope.
The people who have been most passionate in their desire to help me have generally suffered from depression themselves. The fatalistic attitude inherent to depression often leaves sufferers in a state of analysis paralysis. Instead of taking action to solve their problems, they tend to just create long lists of possible reasons why everything will go wrong.
The fatalistic pessimism preventing the suffering from performing the systematic self-experimentation required to optimize their neurotransmitter levels is itself a result of their sub-optimal neurotransmitter levels.
I was extremely fortunate that I had just enough motivation and resources (intellectual and fiscal) to take the necessary actions to solve my problem. However, in general all the things you must do to recover from depression are made difficult by the symptoms of depression. For example, you should eat well, sleep well, be active, and think realistically. Yet the typical symptoms of depression include poor appetite, insomnia, lethargy, and negative thinking.
Solution: Ever-Present Electronic Reminders
It will take the gentle, yet eternal, persistence that can only be provided by an electronic companion to overcome the extreme proclivity towards non-compliance that characterizes those suffering from mental illness. Software can do this. She could always be there to inform you in a caring way when you are engaging in behaviors that exacerbate your condition (i.e. improper diet, insufficient sleep, lack of exercise, or a low degree of social interaction).
Barrier 7: Mental Illness is Invisible
Those most capable of getting things done in this world are those who have won the genetic lottery and naturally possess near-optimal neurotransmitter levels. These are the people who have the ability to get through 8 years of medical school. It is relatively easy for healthy people to see and empathize with the pain of physical illnesses. However, for those with mental illness, there is no language in their lungs, bridge of thought, or mental ink to convey the invisible pain they feel.
Solution: Use Data to Visualize the Pain
The Positive and Negative Affect Schedule (PANAS) developed by David Watson, Lee A. Clark, and Auke Tellegen, is a scale which uses the measure of 20 different feelings and emotions to quantify one’s overall mood.
Barrier 8: Data Gathering is Currently Very Arduous
At the present time, it requires a great deal of effort and diligence on the part of the self-tracker to gather all of the data required to identify the triggers of mental illness and quantify the effectiveness of different treatments. Tracking ones mood, diet, sleep, activity, and medication intake can be extremely time-consuming.
Solution: Automate Data Acquisition with New Technology
We need applications that could be a friend to you, asking you questions about your day in a caring way. Although she will be collecting data, it would just feel like a therapist letting you vent. Voice recognition may be used to quantify emotion through conscious verbal expression, spectral analysis of the magnitudes of different frequencies of speech would probably be a better means of quantifying unconscious human affect and thus providing more accurate data for the machine learning process. The user can place the burden of self-tracking on the software rather than carrying the burden themselves.
Furthermore, our application could automatically retrieve information coming from a variety of sources. The data sources would include:
- Bio-metric Devices: that could measure vital signs and biomarkers
- Purchase Records: Data regarding consumption of foods and supplements could be automatically collected by and inferred from receipts or other financial aggregation services like mint.com or Slice.
- Auditory Records: Voice recognition may be used to quantify emotion through conscious verbal expression, spectral analysis of the magnitudes of different frequencies of speech would probably be a better means of quantifying unconscious human affect and thus providing more accurate data for the machine learning process. CommonSense is a cloud-based platform for sensor data.
- Visual Affect Data via Web-Cameras: By tracking hundreds of points on the subjects’ face, InSight can accurately capture emotional states.
- Prescription Records: Microsoft Healthvault can automatically collect lab results, prescription history, and visit records from a growing list of labs, pharmacies, hospitals, and clinics.
Barrier 9: Human Bias
To err is human. Many different biases may influence decision-making and diagnosis in medical practice. The two primary types are:
- Affective biases (intrusion of the physician’s feelings and emotions)
- Cognitive biases (distortions of thinking)
Almost always, physician biases are unconscious. Recent research in medicine has indicated a profound impact of cognitive biases on reasoning, decision-making, and diagnosis. Common cognitive biases include, but are not limited to:
- Anchoring – focusing on 1 symptom or diagnosis and failing to consider other possibilities
- Premature Closure – uncritical acceptance of an initial diagnosis
- Search Satisfaction – calling off the search when just 1 abnormality has been found
- Availability – recent or vivid patient diagnoses are more easily brought to mind (i.e. are more available) and overemphasized in assessing the probability of a current diagnosis.
All of these biases can interfere with reaching a correct diagnosis or recommending appropriate treatment for patients.
Solution: Algorithms Do Not Lie
Patients deserve real quantitative data regarding the effectiveness rates and adverse effects of all potential treatments for individuals who share their extremely specific symptomatological profile. Their diagnosis should be algorithmic and data driven. The Research Domain Criteria Project (RDoC) provides using new ways of classifying mental disorders based on dimensions of observable behavior and neurobiological measures which could be used to detect individual predictors of response. This is the only way to ensure that each patient receives the most appropriate treatment available.
