Discover more from Zhao
[Summary] Digital Technologies in Psychiatry: Present and Future (Hirschtritt, Insel 2018)
TLDR; Tech has the ability to address 5 major issues: measurement, access, care delay, fragmented care, and stigma. Digital phenotyping and digital interventions show the most promise.
TLDR; Tech has the ability to address 5 major issues: measurement, access, care delay, fragmented care, and stigma. Digital phenotyping is showing promise for objective measurement and digital interventions are showing promise for greater access; one main challenge for both is real world adherence and usage is low so far.
Other promising uses of tech include chat bots, combining tech to support clinician or lay-clinician work, and anonymous peer support.
The are a growing number of mental health technology solutions and people are open to using them.
“As just one reflection of the growth of health technology, a recent report identified over 1,000 companies launched in the past five years, with a total investment of over $20 billion (4). A 2017 IQVIA Institute for Human Data Science report found 318,000 health apps and 340 consumer wearable devices available worldwide; apps that addressed mental health and behavioral disorders constituted 28% of all disease-specific apps (5). Furthermore, 45% of the respondents to an anonymous, online depression screening tool offered by Mental Health America (MHA) in 2016 stated that they would be interested in receiving online- or mobile-based support from MHA for mental health issues (only 18% would accept referral for a face-to-face interview) (6)”
Problems That Digital Health May Address
5 broad problems that it may address: measurement, access, delays in care, fragmented care, stigma.
(1) Absence of object and standardized measurement, especially compared to most other areas of healthcare.
“First is the absence of objective and standardized measurement. Although other medical fields, such as oncology (7), have integrated structured, standardized, and validated tools based on biological markers and objective signs to diagnose and guide treatment, analogous tools in psychiatry are either absent or, when available (e.g., semistructured interviews for diagnosis), rarely used outside of research settings. Objective, continuous, ecological measurements of emotion, cognition, and behavior are not available either for diagnosis or for assessing outcomes. Recently, measurement-based care has been suggested as a foundation for improving the quality of mental health care (8).”
(2) Access: most people in the world and US do not have access to evidence-based mental health tools or therapists.
“The second issue is related to access to mental health services. In many parts of the developed world and most parts of the developing world, evidence-based mental health care is not available. The Substance Abuse and Mental Health Services Administration reports that 55% of U.S. counties lack a mental health professional (9). A recent interagency report on the state of treatment for serious mental illness reported that about 2% of patients receive evidence-based treatments such as assertive community treatment, supported employment, and family psychoeducation (10).”
(3) Extensive delay in getting care, usually out of reluctance to seek care.
“Third, for those who find treatment, there is usually an extensive delay that is due not only to lack of access but also to reluctance to seek care. Among individuals seeking treatment for first-episode psychosis in community clinics in the United States, the median duration of untreated psychosis was 74 weeks (11). This delay has serious consequences for outcomes, as patients often first receive treatment after a long period of psychosocial disability, impairing their ability to recover”
(4) Care is highly fragmented and difficult to navigate for clients.
“Fourth, mental health care is highly fragmented, with different professionals, different treatments, and even different perspectives creating a maze that patients are challenged to navigate. Despite significant comorbidity among psychiatric disorders (including substance use disorders) and physical medical diseases, in practice, their care is siloed, with different medical records and different reimbursement plans (12). Even mental health care itself frequently lacks continuity”
(5) Stigma that prevents seeking or adhering to care.
“Finally, and connected to the other four problems, are the attitudinal barriers, often referred to as stigma and consisting of a range of negative attitudes—especially negative views of psychiatrists, diagnostic labels, and treatments—that limit some people from seeking or adhering to care. For instance, results of the National Comorbidity Survey Replication revealed that 55% of people who were diagnosed as having a mental illness are not in care (14). Although many assumed that this low rate of care was due to lack of access, a revealing follow-up to this report found that the reason was more often “attitudinal barriers” rather than “structural barriers””
Assessment: Digital Phenotyping Addressing Lack of Measurement
Digital phenotyping would allow continuous, precise, and objective measurement whereas current tools are not very objective, hardly used, or used only briefly (once-a-week).
“Even when validated assessment tools are used, clinicians are limited to episodic self-report measures collected in a clinic, which are subject to recall bias. In addition, many of these clinically oriented tools are necessarily brief and therefore capture only a narrow spectrum of a patient’s overall state (e.g., general depression symptoms) and are administered infrequently, leading to a collection of one-time, or “snapshot,” impressions of a patient’s mental health. Ideally, clinicians would have access to continuous, precise, objective, and passively collected data about their patients, which would be synthesized to yield clinically meaningful information.”
Digital phenotyping will allow for continuous and passive collection of data.
“Digital phenotyping is the term used to describe this new form of assessment (17). Digital phenotyping is based on sensor data, such as activity and geolocation; human–computer interaction reflected in keyboard and scrolling patterns; and voice and speech patterns. In addition, “digital exhaust,” like the wake behind a boat, including social media posts and search histories, can be part of the digital phenotype. Although smartphones can be used to collect rating scales and momentary assessments throughout the day, the appeal of digital phenotyping is that the assessment is passive, not requiring patients to do anything other than charge and use their phones.”
Studies in digital phenotyping are already showing promising signs detecting depression, schizophrenia, PTSD.
“It may not be surprising that severe depression or mania would be reflected in activity or sleep data or that psychosis would influence human–computer interaction. How would these data be used in practice? Current studies are exploring several applications. In clinical trials, digital phenotyping is being used to detect early changes in depression, schizophrenia, and posttraumatic stress disorder (PTSD), validated by gold-standard clinical ratings. The hope is that signals from the phone or a wearable will ultimately augment, or even replace, clinical ratings, potentially providing earlier signals of relapse or recovery. Other trials are using the rich data from smartphones to search for cognitive or behavioral subtypes that might predict treatment response. Still others are using digital phenotyping to monitor patients with serious mental illness, creating a “smoke alarm” of early relapse for care managers. The concept behind all of these applications is that monitoring behavior ecologically by using a device that has already been adopted by patients will improve diagnosis and treatment, specifically targeting the lack of measurement and the delay in care noted previously.”
Machine learning algorithms are already detecting digital bio markers of cognitive function based off keystroke and scrolling patterns.
The search for valid, actionable signals is ongoing. Sleep and activity are relatively straightforward, but digital phenotyping also includes digital biomarkers extracted from subtle aspects of smartphone use, such as keystroke and scrolling patterns. For consenting individuals who had their smartphone keystroke and scrolling patterns collected and who also completed extensive gold-standard neuropsychological tasks, machine learning algorithms identified digital features—essentially, biomarkers—associated with processing speed, verbal memory, and measures of executive function with high correlations (approximately 0.8 ).
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6550173/ Digital biomarkers of cognitive function
There are still many challenges facing digital phenotyping such as adherence to wearables, noisy data, reproducibility of 20-year-old usage to 70-yr-old populations.
“The smartphone may seem the ideal solution to the measurement problem in psychiatry. In fact, rigorous studies of digital phenotyping demonstrate that this is feasible but also challenging. Depressed people stop charging and using their phones; sensor data are noisy; and collecting speech, text, or search data introduces serious privacy issues. Wearable devices provide valid measures of sleep and activity; but adoption and adherence are significant problems, with fewer than 10% still in use after 6 months (19). Additional challenges include problems with reproducibility in diverse samples; it is not yet clear that the digital biomarkers collected in 20-year-olds who use their smartphones more than 4 hr/day will generalize to 70-year-olds who use their phones much less frequently. Furthermore, validation remains a challenge: Unlike the development of digital tools for diabetes or heart disease, the lack of “ground truth” physiologic biomarkers for mental disorders may limit the interpretation of digital data (20)”
Treatments: Digital Interventions Addressing Access and Delay
Efficacy and effectiveness are different: effectiveness requires performance under real-world conditions.
“In recent years, multiple studies have addressed the efficacy—a measure of how well an intervention leads to the expected results in an ideal or controlled circumstance—of digital mental health interventions. However, fewer studies have directly measured the effectiveness—the degree to which an intervention performs under real-world conditions—of these rapidly evolving technologies. Nevertheless, available evidence suggests the value of digital interventions relative to face-to-face interventions as a means of improving access to evidence-based treatments.”
_Evidence for efficacy_
Digital solutions are showing relatively good efficacy in RCTs.
“A recent meta-analysis of 22 smartphone-based apps for depression among 18 randomized controlled trials (RCTs) and 3,414 participants found a significant decrease in depressive symptoms from baseline (Hedges’ g [a measure of standardized mean difference between study groups]=0.38; 95% confidence interval [CI]: 0.24–0.52; p<0.001), including all 18 studies, and a larger effect when smartphone-based interventions were compared with inactive controls (g=0.56; 95% CI=0.38–0.74, p<.001) than when compared with active controls (g=0.22; 95% CI=0.10–0.33, p<.001; 21).”
But we still need to see if digital interventions will be truly effective at scale and with sustained engagement.
“This promising field remains challenged by the relatively few published RCTs across all of psychiatry (and even fewer that include youths), significant heterogeneity among studies (e.g., inclusion criteria, use of in-person supplementation of digital interventions, control conditions, and outcome measures), and the relatively short study durations (usually one to three months). If the target is access, then the litmus test will be the shift from efficacy to effectiveness, which will require adoption of digital interventions at scale and sustained engagement.”
_3 new trends in digital interventions_
(1) Using digital interventions for populations with serious mental illness.
“First is the use of digital interventions for patients with serious mental illness (23–25). Apps for peer support, medication adherence, and cognitive remediation are showing high rates of adoption among this heretofore hard-to-reach population (26). “
(2) Combining digital and traditional approaches (eg prescription digital therapeutic to augment traditional CBT).
“Second is the potential for combining digital and traditional approaches. Following a pivotal multisite RCT (24), the U.S. Food and Drug Administration (FDA) approved the first mobile app for substance use disorders among adults as “a prescription digital therapeutic” (essentially, digital cognitive–behavioral therapy [CBT]) to accompany standard outpatient treatment (27). This approval now allows clinicians to prescribe the app, called reSET (Pear Pharmaceutics), to patients as part of a comprehensive care plan that includes conventional, in-person substance use treatment.”
(3) Chatbots supported by AI (but they have yet to undergo rigorous testing).
“A third, more innovative trend is the use of “chatbots” supported by sophisticated AI. A recent RCT studying the effects of a fully automated conversational agent that reinforced CBT techniques in a text-message format, compared with an information-only control, found a significantly greater decrease in depressive symptoms among individuals in the intervention group (28). Though promising as an approach to the problems of access and delay, interventions such as chatbots with sophisticated AI have not yet undergone rigorous testing at scale. While we might assume that our patients would prefer face-to-face interaction to interaction with a bot, that assumption may not hold for the majority of individuals with a mental illness who are not in treatment—either because of lack of access or negative attitudes about psychiatry.”
_Challenges in digital interventions_
The main challenge for digital interventions is translating efficacy to real world effectiveness. Most people do not adopt or adhere to online treatments.
“The gap between efficacy and effectiveness is nontrivial. In fact, efforts to move digital interventions from research into practice have generally failed (33). Outside of research protocols, patients and providers either do not adopt or do not adhere to online treatments. Mohr et al. (34) remind us that, going forward, we need to consider digital interventions as technology-enabled services rather than stand-alone products, integrating them into the fabric of a patient’s life and the workflow of a provider’s practice. To improve adoption and adherence, the next generation of digital interventions will likely look less like Internet-based manualized treatment and more like the video games for ADHD or immersive VR.”
A Solution-Focused Research Approach to Achieve an Implementable Revolution in Digital Mental Health (Mohr 2018 Feb) https://pubmed.ncbi.nlm.nih.gov/29238805/
Three Problems With Current Digital Mental Health Research … and Three Things We Can Do About Them(Mohr 2017 May) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6903906/
Care Integration: Addressing Fragmentation
Companies have built software and dashboards to facilitate collaborate care. But one major challenge is a lack of standardized data in mental health care.
“Collaborative care has been shown repeatedly to improve outcomes for people with depression (35). Several companies have created software to facilitate collaborative care, with innovations like an AI nurse to extend the care manager’s reach; dashboards for data flow to help triage patient care; and designing matching algorithms to link primary care, specialty care, and patients. In contrast to the most of the intervention software that has been built for patients, digital tools for integration have been built for providers with the goal of improving the efficiency as well as the quality of care.”
“One additional challenge for mental health integration, relative to collaborative care for diabetes or cardiovascular disease, is the relative lack of standardized data. Electronic health records have not been widely adopted by outpatient mental health providers (36), and in truth are not especially popular with patients and providers in general medicine (37). Mental health data are both more sensitive and less portable than general medical data. Can digital tools overcome this challenge? One approach, borrowed from the do-it-yourself culture of the digital era, builds on efforts to empower patients and families with their own data by giving patients agency and ensuring transparency in the clinical process. The OpenNotes project, which provides this transparency for medical records, is associated with improved treatment adherence and greater patient satisfaction with their provider, suggesting that this kind of an approach might address not only the fragmentation of care but also the negative attitudes toward mental health conditions (38). Unexpectedly, sharing medical records with patients did not generate a new burden for clinicians. To our knowledge, no one has integrated this kind of transparency with mobile diagnostics or other digital interventions.”
Emerging Tools for Mental Health
_Using Digital Technologies to Augment Clinical Capacity_
Tech enables “task shifting” which can help clinicians and non professionals make better decisions with data, much like how Waze helps drivers pick better routes.
“One solution to the global problem of access to mental health providers is task sharing, an approach that uses nonprofessionals with targeted training to extend what an expert clinician can provide. The challenge with task shifting is ensuring that village health workers or suicide hotline volunteers have the skills to provide high-quality care. Is there a technology solution? In the world of managing traffic, many drivers rely on smartphone-based navigation systems to reach their destinations in the most time-efficient manner. Apps like Waze rely on both passive and active data from other users to generate the best route and provide continuously updated, turn-by-turn guidance. In this way, such apps augment a driver’s ability to navigate the roads safely and efficiently. An analogous app for nonprofessional mental health providers could offer patient-specific guidance based on passively and actively imputed data (for instance, information gathered from the electronic health record, or patient-reported symptom-severity scale results) to generate more accurate diagnostic and prognostic data and more effective treatment recommendations. In the context of global mental health or crisis intervention, where we have to rely on task shifting, these tools can give volunteers the ability to provide higher quality care than might be available today. As just one example, data scientists are combining text and social context data to help a volunteer on a suicide text line or phone line assess risk and recommend local interventions (40)”
_Online Peer Support: Will There Be an Airbnb for Mental Health Care?_
As younger generations are more comfortable with anonymous online support, peer-to-peer support may become more feasible and accessible; for instance with 7 Cups.
“This provider-centered approach has a long tradition and defines our comfort zone but appears less appealing to consumers, especially to millennials, many of whom are accustomed to peer-generated digital content. The future of mental health care might take the shape of various digital mental health services built around peer support within trusted, online communities. These services, some of which are already available, offer free, anonymous, global support within minutes without any brick-and-mortar location. Users can engage with such a service for peer support and, if interested, they can pursue stepped-up options, ranging from text-based therapies to individual psychotherapy with a licensed provider.”
“These and similar interventions may not be the sole answer for people with serious mental illnesses or for children with developmental disorders, but this general approach offers anonymity, social connections, and skill building for a large population of people in need of mental health care—and it is popular. 7 Cups, for instance, claims over 21 million conversations in 140 languages, as well as 19% month-on-month growth, with two million monthly users (41)”
Limitations and Future Directions
Tech will face the challenge of data privacy.
“First and foremost, protection of mental health information among these technologies is a pressing issue given the complex regulatory structure of the digital landscape. Specifically, at the time of this writing, it seems that data collected on mental health digital platforms are subject to the Health Insurance Portability and Accountability Act Security Rule and Privacy Rule only if they collect personal health information (42). Likewise, there is a risk that developers may apply user-derived information to generate marketing data. “
Mental health disparity may widen between those who can afford smart phones and those who have greater digital fluency.
“Second, although digital technologies hold the potential to reduce disparities in mental health care, evolving technologies may, in fact, widen gaps between those who can access digital care and those who cannot. Cost is becoming less of an issue, with increasing market competition among software developers and near-ubiquitous smartphone ownership in developed countries; however, individuals with lower digital fluency (43) may be intimidated and feel alienated by these digital resources and may find it difficult to engage with them.”
Most digital interventions have evidence for adults, but we are lacking tools for adolescents and youth (where 1 in 3 may already meet criteria for disorders)
“Third, a majority of digital interventions are targeted toward and have evidence for efficacy among adults; there are comparatively fewer resources and data to support digital interventions for youth (46). This age gap can likely be attributed to both market forces (e.g., a more robust user base exists for mobile apps) and regulatory concerns (e.g., human subject protection standards being more stringent for youths compared with noninstitutionalized adults). However, given that as many as one in every three to five individuals will meet criteria for a psychiatric disorder during their childhood or adolescence (47, 48), and psychiatric treatment is most effective when delivered early in development (49), future research and development should focus on the mental health needs of this vulnerable population.”
There will also be a challenge in integrating insights from studies into the real world of tech.
“Likewise, app and device developers will need to integrate insights from carefully controlled, laboratory-based studies to assess the efficacy of these tools. Communication and collaboration between these two disparate fields—one focused on rapid, iterative development, user development, and scalability; and the other focused on scientific rigor and publication of results will continue to be challenging, but the strengths of both are necessary to develop and refine new technologies to help those afflicted by mental illness (51).”
Integration of Digital Technologies Into Practice
To integrate new technology, clinicians could use a scale to determine what would work best for a patient or consult online rating guides.
“The authors identify three major categories: app capabilities and functions (e.g., psychoeducational materials versus symptom tracking, availability of conversational agents, or chatbots), workflow issues (e.g., accommodation of patients’ daily routines, appeal of the app design), and cultural or access issues (e.g., accounting for technological literacy and access issues). To help structure this evaluation, clinicians can use standardized scales such as the publicly available, 23-item Mobile App Rating Scale (53). The American Psychiatric Association has established a five-step “app evaluation model” to help clinicians assess the quality and usability of an app (54). These rating tools are useful when considering a single app—because either the patient asks about a particular app or the clinician has come across an app through his or her own search—but are less useful when considering a problem (such as depression) and then searching for the “best” app to address the clinical concern. In these, perhaps more common, scenarios, clinicians can consult regularly updated rating guides, such as those maintained by the Anxiety and Depression Association of America (55) and the PsyberGuide (56), part of the nonprofit One Mind Institute.”