Data Science, Crisis Counseling, and the Mental Health Industry
While Data Science is not currently widely used in the mental health industry, there are encouraging signs showing how data science could make a significant impact in the not too distant future. This guide will outline the scope of the problem; cover limitations with current use, as well as present several case studies giving optimism for a much bigger impact in the future.
Scope of the Problem
Mental illness continues to affect tens of millions of people each year. The Substance Abuse and Mental Health Services Administration (SAMHSA), in its 2019 Survey on Drug Use and Mental Health, found that one in five adults suffers from a mental illness, with a higher proportion of women (22%) suffering from it relative to men (15%). However, mental illness is not limited to adults. In fact, the National Alliance on Mental Illness (NAMI) reports that half of all chronic mental illnesses begin by the age of 14. Unfortunately, it is estimated by the NIMH that over half of those affected by mental illness do not receive treatment for various reasons, whether it is due to cost, the negative stigma associated with it or the lack of access. For those who are treated, NIMH has learned an average of eight to ten years passes between the onset of symptoms and initial treatment.
The lack or delay of treatment can have many negative effects on individuals suffering from mental illnesses. Without effective treatment, mental illness can have a devastating impact on the lives of many teenagers. Over a third (37%) of teenagers diagnosed with a mental illness drop out of school as reported by NIMH. Without the opportunities an education provides, it is likely these young adults find their futures limited. Many may find themselves making decisions adversely affecting their futures. The majority (70%) of teens in a juvenile justice system have a mental illness (NIMH). This illustrates how mental illness, when left untreated, can affect not only the individual, but society overall. Most sobering is the correlation between mental illness and the suicide rate among the younger population. The NIMH study found nine out of ten youths who died from suicide suffered from a mental illness.
Unable to find the help they need, adults with mental illness find it difficult to get their lives on the right path. The National Institute of Mental Health (NIMH) has found that a quarter of homeless adults in shelters suffer from a serious mental illness (SMI). The organization defines a SMI as “a mental, behavioral, or emotional disorder resulting in serious functional impairment, which substantially interferes with or limits one or more major life activities”. Without treatment, SMI’s can have debilitating effects on a person’s life. It may begin with strained relationships, decreases in social interactions, physical ailments and the inability to work. This is illustrated by the estimated $193.2 billion in lost wages resulting from serious mental illnesses (NIMH). The correlation between mental illness and suicide is a sobering statistic. Similar to the teenage population, mental illness is the main cause of adults committing suicide, one of the leading causes of death in the United States.
Globally, mental illness is under-diagnosed, underreported and not well understood, particularly in lower income areas of the world. If researchers report the number of mental illness diagnoses, countries with more healthcare resources available, and subsequently higher income levels, would likely have higher incidences of mental illness within their population. In addition, mental illness is broadly defined, with worldwide organizations having varying mental health conditions included in the statistics they report. The Global Burden of Disease (2016) study, conducted by the Institute of Health Metrics & Evaluation (IHME), estimates that one in six people worldwide suffers from at least one mental illness or substance abuse disorder. This statistic is equal to approximately one billion people globally. The World Health Organization (WHO) found the United States to have the highest prevalence of mental health disorders relative to other countries. Analysis of the WHO data found that 27% of adults in the United States will suffer from some form of mental disorder within a 12-month period. However, this statistic was calculated based on the limited data available and variance in reporting worldwide.
Data Science’s Current Role in the Diagnosis and Treatment of Mental Illnesses
Relative to other fields of medicine, the use of Data Science in the area of mental health has been limited. At a basic level, Data Science allows researchers to estimate the effects that mental illnesses have had and will have on the United States and global populations. When the effects are quantified and supported by data, it provides a stronger case for much needed resources to be allocated to mental health research and treatment accessibility.
There have been movements to better understand the value Data Science can provide the field of mental health. Data Scientists have begun using machine learning to process large data sets in the hopes of revealing patterns that would help identify causes, signs and symptoms of mental illnesses and aid in selecting the most effective treatments. While identifying the causes, researchers would also be able to discover preventative measures to combat mental illnesses. However, the algorithms used to accomplish these tasks continues to be refined and are not being widely used in a consistent manner.
As Data Scientists work to perfect these algorithms, data availability will determine their success. Data Science’s impact on the mental health industry continues to be limited by the availability of reliable data sources. Many organizations worldwide collect mental health data but do so without a coordinated effort to consolidate. Larry Pickett, CEO and co-founder of RxDataScience, believes that the commercial interests of health systems, pharmaceutical companies and data providers are preventing the coordination of mental health data resources. Medical scientists and professionals researching mental health illnesses find themselves limited by small populations sizes when trying to identify the correct diagnosis or treatment plan.
Despite these limitations, Data Science is currently making a positive impact in areas such as suicide prevention. Crisis counseling hotlines are using machine learning to help identify callers, at a high risk of committing suicide or self-harm through the words they communicate. These crisis counseling centers can collect valuable information on a regular basis, such as time of day and day of week when the most calls are received, which allows centers to ensure adequate resources are available at any given time. The data collected also aids in counselor training by helping counselors identify the appropriate questions to ask callers and building profiles of those who may need additional or long-term therapy beyond the scope of the center’s services or capabilities. Using Data Science, crisis counselors can direct at-risk people to the professional help they may need.
Future Role of Data Science in the Mental Health Industry
Through data analysis and machine learning, Data Scientists will continue to strive towards a better understanding of mental illnesses and expand access to treatment. Theories and hypotheses will continue to be tested and proven or invalidated as additional data is collected, consolidated and analyzed and algorithms are perfected. Large investments continue to be made into initiatives and studies including neuroscience research aimed at understanding the genetics behind mental illnesses. With this research, the hope is that medical professionals will be able to better diagnose, treat and possibly prevent mental health disorders. As funding continues and increases, improved diagnoses will lead to more appropriate treatments for those suffering with mental illness.
Data Scientists and mental health professionals are beginning to understand the value that artificial intelligence (AI) can provide to the mental health field. Tools powered by AI continue to be developed to assess and diagnose mental health illnesses. AI tools provide an attractive option for the mental health industry as it is cost effective, shortens the time between diagnosis and treatment and has been more likely to create positive results. Research has found that patients are more open to sharing personal information with an AI-powered therapist (or app) relative to a human, as it provides a level of anonymity and removes the perception of being judged. However, the effectiveness of AI-powered tools is again limited by the data used to build the algorithms on which it is based. Data Scientists must ensure that the data used is diverse, reliable and truly reflective of the worldwide population of people suffering from mental illnesses.
Finding new and improved data collection methods will be vital to Data Science’s future impact on mental healthcare. Researchers and companies are investigating technologically advanced tools that will improve the quality and amount of data being collected. The technology will collect data as patients go about their daily routines, which researchers believe will eliminate recall bias. Developers of these technologies will need to continue to work closely with researchers and mental health professionals to determine what types of data will be most useful when diagnosing, treating and preventing mental illnesses.
Technological advancements will be influential in the role Data Science plays in mental health research. Researchers at Verily, a company focused on developing technology to collect health related data, are calling for mental health professionals to support the use of Data Science in diagnosis, treatment and prevention. This would require them to rely less on subjective observations and more on evidence-based research using Data Science. The main advantage would be less variability in diagnoses and treatments for mental illnesses. New technology in wearable monitoring devices will allow mental health professionals to gather data outside of the clinical office by measuring and tracking activity, behavior and other health variables throughout the day. These types of technological developments will increase the volume and variety of data available and ultimately allow researchers to use Data Science to reliably test and identify learnings that will benefit mental healthcare.
Case Study – Mental Health Data Science at Columbia University
Mental Health Data Science at Columbia University (2019) provides support to researchers through statistical collaborations, data analytics methodology development and data management. Each year the department works with an average of 50 researchers to advance the study of mental health. More specifically, the goals of the department are to support mental health research projects using traditional proven biostatistical data analysis methodologies, to develop and utilize data analytics methodologies as it relates to technological advances in mental health data collection and developing secure systems that are helpful in the various areas of mental health research projects. Through the two groups, Biostatistics and Data Management, the department hopes to promote collaboration amongst researchers in the mental health field by sharing expertise and methodologies.
Mental Health Data Science is currently providing support to several research projects funded by grants from organizations such as the National Institute of Health (NIH), which includes agencies such as the National Institute of Mental Health (NIMH) and the National Institute of Neurological Disorders and Strokes (NINDS), and the Substance Abuse and Mental Health Services Administration (SAMHSA). Areas of research include Healthcare Policy and Treatment Access, Suicide Prevention and Early Disease Prediction and Treatment Strategy.
Case Study – Canadian Military and Veteran Mental Health Initiative
In 2018, $12M was dedicated to Data Science research focused on the mental health issues faced by Canadian military personnel, veterans and their families. The research project, known as the Advanced Analytics Initiative, is a partnership between the Canadian Institute for Military and Veteran Health Research (CIMVHR) and corporate partners, IBM Canada and Babcock Canada. The initiative looks to identify Data Science methodologies and processes that will help treat mental health issues common among military personnel and veterans, including post traumatic stress disorder (PTSD) and post-concussive syndrome (PCS). The hope is that the partnership will connect much needed resources, such as technology and data science expertise, to the health data being collected. This will allow researchers to gain insight into diagnosis, treatment and prevention of mental illnesses within Canada’s military population and eventually the greater global population suffering from mental illnesses.
The Advanced Analytics Initiative has awarded funding to studies addressing diagnosis and treatment, including treatment access. Researchers from Western University and the University of Alberta are developing and testing machine algorithms that could help predict mental illnesses and the efficacy of specific treatment based on the symptoms a patient presents. Another research group at Queens University is addressing mental health treatment and accessibility. Using Artificial Intelligence powered by IBM’s Watson platform, the group is developing a real-time chat bot app that will help those suffering from mental health issues to be connected to resources for treatment.
Case Study – Virtual Therapist, University of Southern California
To make mental health care more accessible, the Institute of Creative Technologies at the University of Southern California has developed a virtual therapist, a 3-D chatbot viewed on a television screen. The virtual therapist, Ellie, has the capability to show facial expressions and detect non-verbal cues using algorithms. Data, in the form of patients’ verbal input and facial expressions, run through these algorithms which determine Ellie’s response, both visual and verbal.
In a study with military personnel who recently returned from a tour in Afghanistan, Ellie was found to be more effective at detecting Post Traumatic Stress Disorder (PTSD) symptoms than the military’s routine health assessment. Researchers believe the ability to effectively build rapport and trust with interviewees is the key to Ellie’s success. The algorithms used by the virtual therapist are continually being improved with additional data from human therapists, making the assessment process more productive.
The goal of virtual therapists is to provide more mental health resources for those in need. The intention is not to replace human therapists, but to divert more of mental health professionals’ time to the treatment of mental illness. The shortage of mental health professionals in the United States has made it difficult for those suffering from a mental illness to receive the treatment they need. Approximately 40% of the U.S. population live in areas with a shortage of mental health professionals (as determined by the federal government), and “more than 60% of U.S. counties are without a single psychiatrist within their borders”. Ellie is an example of how Data Science in the mental health industry can have a significant impact on mental healthcare accessibility.
Case Study – Artificial Intelligence and Suicide Prevention, Crisis Text Line
Crisis Text Line (CTL) is a crisis counseling center fielding text messages from those who may be suffering from a crisis or contemplating self-harm or suicide. The service gives these people the ability to communicate with a counselor through text messages, a method they may feel more comfortable with than a live phone conversation.
Since its start in 2013, Crisis Text Line (CTL) has exchanged 30 million texts with users, allowing for a wealth of data to be collected. Presently, the data being collected has helped people messaging CTL to get the help they need, but the hope is the data can be used for earlier detection of self-harm or suicidal tendencies outside CTL. By analyzing the collected data, data scientists have been able to find patterns in the language used by those messaging with counselors at CTL. The goal is to share these patterns with the public so a larger population of people suffering crises can be helped.
CTL collects three types of data – the actual text conversation with the counselor, and after the interaction, feedback from both the counselor and the texter. Data scientists have been able to analyze this data to identify invaluable patterns. For example, they have been able to identify key words such as “Advil” and “Ibuprofen” that were more indicative of high risk of suicide than words previously thought to be (“cut”, “die”, “kill”, etc.). This illustrates how machine learning can be more effective than the human intuition traditionally used by mental health professionals.
Case Study – Analyzing EHR for Predicting Suicide Risk, Mental Health Research Network
Using Electronic Health Records (EHR) and results from a depression questionnaire, researchers from Kaiser Permanente were able to build an analytics model that predicts suicide risk within 90 days of a visit to a mental health professional. This model allowed researchers to identify those patients at the highest risk of suicide with predictors including “prior suicide attempts, mental health and substance use diagnoses, medical diagnoses, psychiatric medications prescribed, inpatient or emergency room care, and scores on a standardized depression questionnaire.”
The use of Electronic Health Records (EHR) allows for data scientists to develop a more accurate model than what has previously been used. Models developed in the past by others have employed fewer data elements, while long-time traditional methods of using questionnaires and clinical assessments have been found to be even less effective at assessing suicide risk. In the study, executed by Kaiser Permanente researchers, the model identified the 5% of the test population with the highest suicide risk. Of those in the overall test population who attempted suicide, 43% were identified as high risk, while 48% of suicide deaths were from the same test group. This is a significant improvement from previous models that predicted one-half to one-third of attempted suicides and suicide deaths. This study illustrates how data science can be used in identifying high risk patients, but the long-term hope is to use the same approach to use analytics to develop treatment plans, protocols for patient follow-ups and assessments for those who may require long-term treatment.