John Torous, MD

In today’s global society, smartphones are ubiquitous, used by >2.5 billion people.¹ They provide limitless availability of on-demand services and resources, unparalleled computing power by size, and the ability to connect with anyone in the world.

Digital applications and new mobile technologies can be used to change the nature of the psychiatrist–patient relationship. The future of clinical practice is changing with the help of smartphones and apps. Diagnosis, follow-up, and treatment will never look the same as we come to better understand and apply emerging technologies.²

Both Android and iOS—the 2 largest mobile operating systems by market share³—provide outlets for the dissemination of mobile applications. There are currently >10,000 mental health–related apps available for download.⁴ One particular use case of mental health–related apps is digital phenotyping.

In this article, we aim to:

• define digital phenotyping
• explore the potential advances in patient care afforded by emerging technology
• discuss the ethical dilemmas and future of mental health apps.

The possibilities of digital phenotyping

Digital phenotyping is capturing a patient’s real-time clinical state using digital technology to better understand the patient’s state outside of the clinic. While digital phenotyping may seem new, the concepts behind it are grounded in good clinical care.

For example, it is important to assess sleep and physical activity for nearly all patients, regardless of diagnosis. However, the patient’s retrospective recollection of sleep, mood, and other clinically relevant metrics is often unreliable, especially when visits are months apart. With smartphones, it is possible to automatically collect metrics for sleep, activity, mood, and much more in real time from the convenience of our patients’ personal devices (Figure 1).

Smartphones can capture a seemingly endless number of data streams, from patient-interfacing active data, such as journal entries, messaging, and games, to data that is captured passively, such as screen time, Global Positioning System information, and step count. Clinicians can work with patients to customize which digital phenotyping data they would like to capture. In one study, researchers worked with 17 patients with schizophrenia by capturing self-reported surveys, anonymized phone call logs, and location data to see if they could predict relapse by observing variations in how patients interact with their smartphones.⁵ They observed that the rate of behavioral anomalies was 71% higher in the 2 weeks prior to relapse than during other periods. The data captured by the smartphone will depend on the patient and the clinical needs. Some clinicians may only want to collect data on step count and screen time to learn if a patient is overusing his or her smartphone, which might be related to becoming less physically active.

Continue to: One novel data stream...

One novel data stream offered by smartphone digital phenotyping is cognition. While we know that impaired cognition is a core symptom of schizophrenia, and that cognition is affected by depression and anxiety, cognitive symptoms are clinically challenging to quantify. Thus, the cognitive burden of mental illness and the cognitive effects of treatment are often overlooked. However, smartphones are beginning to offer a novel means of capturing a patient’s cognitive state through the use of common clinical tests. For example, the Trail Making Test measures visual attention and executive function by having participants connect dots that differ in number, color, or shape in an ascending pattern.⁶ By having patients perform this test on a smartphone, clinicians can utilize the touchscreen to capture the user’s discrete actions, such as time to completion and misclicks. These data can be used to build novel measures of cognitive performance that can account for learning bias and other confounding variables.⁷ While these digital cognitive biomarkers are still in active research, it is likely that they will quickly be developed for broad clinical use.

In addition to the novel data offered by digital phenotyping, another benefit is the low cost and ease of use. Unlike wearable devices such as smartwatches, which can also offer data on steps and sleep, smartphone-based digital phenotyping does not require patients to purchase or use additional devices. Running on patients’ smartphones, digital phenotyping offers the ability to capture rich and continuous health data without added effort or cost. Given that the average person interacts with their phone more than 2,600 times per day,⁸ smartphones are well suited for capturing large amounts of information that may provide insights into patients’ mental health.

For illnesses such as depression and anxiety, the clinical relevance of digital phenotyping is in the ability to capture symptoms as they occur in context. Figure 2 provides a simplified example of how we can learn that for this fictitious patient, exercise greatly improves anxiety, whereas being in a certain environment worsens it. Other insights about sleep and social settings could also provide further information about the context of the patient’s symptoms. While these correlations alone will not lead to better clinical outcomes, it is easy to imagine how such data could help a patient and clinician start a conversation about making impactful changes.

Continue to: Case report...

Case report: Digital phenotyping

To illustrate how digital phenotyping could be put to clinical use, we created the following case report of a fictional patient who agrees to be monitored via her smartphone.

Consider a hypothetical patient we will call Ms. T who is in her mid-20s and has been diagnosed with schizophrenia. On a follow-up visit, she says she has insomnia. She also reports having a recent loss of appetite and higher levels of anxiety. After reviewing her smartphone data (Figure 3), the clinician sees an inversely proportional relationship between her sleep quality and symptoms of anxiety, psychosis, and depression, which suggests that these symptoms might be due to poor sleep. Her step count has been fairly stable, indicating that there is no significant correlation between physical activity and her other symptoms.

Continue to: The clinician shows...

The clinician shows Ms. T the data to help her understand why a trial of cognitive-behavioral therapy for insomnia, or at least improving sleep hygiene, may offer several benefits. The clinician advises her to continue to use the app to help assess her response to these interventions and monitor her progress in real time.

Dilemma: The ethics of continuous observation

The rich data captured by digital phenotyping afford many clinical opportunities, but also raise concerns. Among these are 3 significant ethical implications.

Firstly, the same data that may help a clinician learn about what environments are associated with less anxiety for the patient may also reveal personal details about where that patient has been or with whom they have interacted. In the wrong hands, such personal data could cause harm. And even in the hands of a trusted clinician, a breach in the patient’s privacy begs the question: “Should such information be anyone’s business at all?”

Secondly, many apps that offer digital phenotyping could also store patient data—something that currently pervades social media and causes reasonable discomfort for many people. You might have personally encountered this with social media platforms such as Facebook. When it comes to mobile mental health apps, clinicians should carefully understand the data usage agreement of any digital phenotyping app they wish to use and then share this information with their patients.

Finally, while it is possible to collect the types of data outlined in this article, less is known about how to use it directly in clinical care. Understanding for each patient which data streams are most meaningful and which data streams are noise that should be ignored is an area of ongoing research. A good first step may be to begin with data streams that are known to be clinically relevant and valuable, such as sleep and physical activity.^9-11

Continue to: Discussion...

Although smartphones can gather a wide range of active and passive data, other data streams hold potential for predicting relapse and performing other clinically relevant actions. One data stream that could be of clinical use is genomic sequencing.¹² The genotyping of patients provides a wealth of information about the underlying biology, and genomic sequencing has never been cheaper.¹³

Combining the data gathered via digital phenotyping with that of genotyping could help elucidate the mechanisms by which specific diseases and symptoms occur. This could be very promising to better understand and treat our patients. However, as is the case with genomics, digital phenotyping has important ethical implications. If used in the proper way to benefit our patients, the future for this new method is bright.

In today’s global society, smartphones are ubiquitous, used by >2.5 billion people.¹ They provide limitless availability of on-demand services and resources, unparalleled computing power by size, and the ability to connect with anyone in the world.

Digital applications and new mobile technologies can be used to change the nature of the psychiatrist–patient relationship. The future of clinical practice is changing with the help of smartphones and apps. Diagnosis, follow-up, and treatment will never look the same as we come to better understand and apply emerging technologies.²

Both Android and iOS—the 2 largest mobile operating systems by market share³—provide outlets for the dissemination of mobile applications. There are currently >10,000 mental health–related apps available for download.⁴ One particular use case of mental health–related apps is digital phenotyping.

In this article, we aim to:

• define digital phenotyping
• explore the potential advances in patient care afforded by emerging technology
• discuss the ethical dilemmas and future of mental health apps.

The possibilities of digital phenotyping

Digital phenotyping is capturing a patient’s real-time clinical state using digital technology to better understand the patient’s state outside of the clinic. While digital phenotyping may seem new, the concepts behind it are grounded in good clinical care.

For example, it is important to assess sleep and physical activity for nearly all patients, regardless of diagnosis. However, the patient’s retrospective recollection of sleep, mood, and other clinically relevant metrics is often unreliable, especially when visits are months apart. With smartphones, it is possible to automatically collect metrics for sleep, activity, mood, and much more in real time from the convenience of our patients’ personal devices (Figure 1).

Smartphones can capture a seemingly endless number of data streams, from patient-interfacing active data, such as journal entries, messaging, and games, to data that is captured passively, such as screen time, Global Positioning System information, and step count. Clinicians can work with patients to customize which digital phenotyping data they would like to capture. In one study, researchers worked with 17 patients with schizophrenia by capturing self-reported surveys, anonymized phone call logs, and location data to see if they could predict relapse by observing variations in how patients interact with their smartphones.⁵ They observed that the rate of behavioral anomalies was 71% higher in the 2 weeks prior to relapse than during other periods. The data captured by the smartphone will depend on the patient and the clinical needs. Some clinicians may only want to collect data on step count and screen time to learn if a patient is overusing his or her smartphone, which might be related to becoming less physically active.

Continue to: One novel data stream...

One novel data stream offered by smartphone digital phenotyping is cognition. While we know that impaired cognition is a core symptom of schizophrenia, and that cognition is affected by depression and anxiety, cognitive symptoms are clinically challenging to quantify. Thus, the cognitive burden of mental illness and the cognitive effects of treatment are often overlooked. However, smartphones are beginning to offer a novel means of capturing a patient’s cognitive state through the use of common clinical tests. For example, the Trail Making Test measures visual attention and executive function by having participants connect dots that differ in number, color, or shape in an ascending pattern.⁶ By having patients perform this test on a smartphone, clinicians can utilize the touchscreen to capture the user’s discrete actions, such as time to completion and misclicks. These data can be used to build novel measures of cognitive performance that can account for learning bias and other confounding variables.⁷ While these digital cognitive biomarkers are still in active research, it is likely that they will quickly be developed for broad clinical use.

In addition to the novel data offered by digital phenotyping, another benefit is the low cost and ease of use. Unlike wearable devices such as smartwatches, which can also offer data on steps and sleep, smartphone-based digital phenotyping does not require patients to purchase or use additional devices. Running on patients’ smartphones, digital phenotyping offers the ability to capture rich and continuous health data without added effort or cost. Given that the average person interacts with their phone more than 2,600 times per day,⁸ smartphones are well suited for capturing large amounts of information that may provide insights into patients’ mental health.

For illnesses such as depression and anxiety, the clinical relevance of digital phenotyping is in the ability to capture symptoms as they occur in context. Figure 2 provides a simplified example of how we can learn that for this fictitious patient, exercise greatly improves anxiety, whereas being in a certain environment worsens it. Other insights about sleep and social settings could also provide further information about the context of the patient’s symptoms. While these correlations alone will not lead to better clinical outcomes, it is easy to imagine how such data could help a patient and clinician start a conversation about making impactful changes.

Continue to: Case report...

Case report: Digital phenotyping

To illustrate how digital phenotyping could be put to clinical use, we created the following case report of a fictional patient who agrees to be monitored via her smartphone.

Consider a hypothetical patient we will call Ms. T who is in her mid-20s and has been diagnosed with schizophrenia. On a follow-up visit, she says she has insomnia. She also reports having a recent loss of appetite and higher levels of anxiety. After reviewing her smartphone data (Figure 3), the clinician sees an inversely proportional relationship between her sleep quality and symptoms of anxiety, psychosis, and depression, which suggests that these symptoms might be due to poor sleep. Her step count has been fairly stable, indicating that there is no significant correlation between physical activity and her other symptoms.

Continue to: The clinician shows...

The clinician shows Ms. T the data to help her understand why a trial of cognitive-behavioral therapy for insomnia, or at least improving sleep hygiene, may offer several benefits. The clinician advises her to continue to use the app to help assess her response to these interventions and monitor her progress in real time.

Dilemma: The ethics of continuous observation

The rich data captured by digital phenotyping afford many clinical opportunities, but also raise concerns. Among these are 3 significant ethical implications.

Firstly, the same data that may help a clinician learn about what environments are associated with less anxiety for the patient may also reveal personal details about where that patient has been or with whom they have interacted. In the wrong hands, such personal data could cause harm. And even in the hands of a trusted clinician, a breach in the patient’s privacy begs the question: “Should such information be anyone’s business at all?”

Secondly, many apps that offer digital phenotyping could also store patient data—something that currently pervades social media and causes reasonable discomfort for many people. You might have personally encountered this with social media platforms such as Facebook. When it comes to mobile mental health apps, clinicians should carefully understand the data usage agreement of any digital phenotyping app they wish to use and then share this information with their patients.

Finally, while it is possible to collect the types of data outlined in this article, less is known about how to use it directly in clinical care. Understanding for each patient which data streams are most meaningful and which data streams are noise that should be ignored is an area of ongoing research. A good first step may be to begin with data streams that are known to be clinically relevant and valuable, such as sleep and physical activity.^9-11

Continue to: Discussion...

Although smartphones can gather a wide range of active and passive data, other data streams hold potential for predicting relapse and performing other clinically relevant actions. One data stream that could be of clinical use is genomic sequencing.¹² The genotyping of patients provides a wealth of information about the underlying biology, and genomic sequencing has never been cheaper.¹³

Combining the data gathered via digital phenotyping with that of genotyping could help elucidate the mechanisms by which specific diseases and symptoms occur. This could be very promising to better understand and treat our patients. However, as is the case with genomics, digital phenotyping has important ethical implications. If used in the proper way to benefit our patients, the future for this new method is bright.

In today’s global society, smartphones are ubiquitous, used by >2.5 billion people.¹ They provide limitless availability of on-demand services and resources, unparalleled computing power by size, and the ability to connect with anyone in the world.

Digital applications and new mobile technologies can be used to change the nature of the psychiatrist–patient relationship. The future of clinical practice is changing with the help of smartphones and apps. Diagnosis, follow-up, and treatment will never look the same as we come to better understand and apply emerging technologies.²

Both Android and iOS—the 2 largest mobile operating systems by market share³—provide outlets for the dissemination of mobile applications. There are currently >10,000 mental health–related apps available for download.⁴ One particular use case of mental health–related apps is digital phenotyping.

In this article, we aim to:

• define digital phenotyping
• explore the potential advances in patient care afforded by emerging technology
• discuss the ethical dilemmas and future of mental health apps.

The possibilities of digital phenotyping

Digital phenotyping is capturing a patient’s real-time clinical state using digital technology to better understand the patient’s state outside of the clinic. While digital phenotyping may seem new, the concepts behind it are grounded in good clinical care.

For example, it is important to assess sleep and physical activity for nearly all patients, regardless of diagnosis. However, the patient’s retrospective recollection of sleep, mood, and other clinically relevant metrics is often unreliable, especially when visits are months apart. With smartphones, it is possible to automatically collect metrics for sleep, activity, mood, and much more in real time from the convenience of our patients’ personal devices (Figure 1).

Smartphones can capture a seemingly endless number of data streams, from patient-interfacing active data, such as journal entries, messaging, and games, to data that is captured passively, such as screen time, Global Positioning System information, and step count. Clinicians can work with patients to customize which digital phenotyping data they would like to capture. In one study, researchers worked with 17 patients with schizophrenia by capturing self-reported surveys, anonymized phone call logs, and location data to see if they could predict relapse by observing variations in how patients interact with their smartphones.⁵ They observed that the rate of behavioral anomalies was 71% higher in the 2 weeks prior to relapse than during other periods. The data captured by the smartphone will depend on the patient and the clinical needs. Some clinicians may only want to collect data on step count and screen time to learn if a patient is overusing his or her smartphone, which might be related to becoming less physically active.

Continue to: One novel data stream...

One novel data stream offered by smartphone digital phenotyping is cognition. While we know that impaired cognition is a core symptom of schizophrenia, and that cognition is affected by depression and anxiety, cognitive symptoms are clinically challenging to quantify. Thus, the cognitive burden of mental illness and the cognitive effects of treatment are often overlooked. However, smartphones are beginning to offer a novel means of capturing a patient’s cognitive state through the use of common clinical tests. For example, the Trail Making Test measures visual attention and executive function by having participants connect dots that differ in number, color, or shape in an ascending pattern.⁶ By having patients perform this test on a smartphone, clinicians can utilize the touchscreen to capture the user’s discrete actions, such as time to completion and misclicks. These data can be used to build novel measures of cognitive performance that can account for learning bias and other confounding variables.⁷ While these digital cognitive biomarkers are still in active research, it is likely that they will quickly be developed for broad clinical use.

In addition to the novel data offered by digital phenotyping, another benefit is the low cost and ease of use. Unlike wearable devices such as smartwatches, which can also offer data on steps and sleep, smartphone-based digital phenotyping does not require patients to purchase or use additional devices. Running on patients’ smartphones, digital phenotyping offers the ability to capture rich and continuous health data without added effort or cost. Given that the average person interacts with their phone more than 2,600 times per day,⁸ smartphones are well suited for capturing large amounts of information that may provide insights into patients’ mental health.

For illnesses such as depression and anxiety, the clinical relevance of digital phenotyping is in the ability to capture symptoms as they occur in context. Figure 2 provides a simplified example of how we can learn that for this fictitious patient, exercise greatly improves anxiety, whereas being in a certain environment worsens it. Other insights about sleep and social settings could also provide further information about the context of the patient’s symptoms. While these correlations alone will not lead to better clinical outcomes, it is easy to imagine how such data could help a patient and clinician start a conversation about making impactful changes.

Continue to: Case report...

Case report: Digital phenotyping

To illustrate how digital phenotyping could be put to clinical use, we created the following case report of a fictional patient who agrees to be monitored via her smartphone.

Consider a hypothetical patient we will call Ms. T who is in her mid-20s and has been diagnosed with schizophrenia. On a follow-up visit, she says she has insomnia. She also reports having a recent loss of appetite and higher levels of anxiety. After reviewing her smartphone data (Figure 3), the clinician sees an inversely proportional relationship between her sleep quality and symptoms of anxiety, psychosis, and depression, which suggests that these symptoms might be due to poor sleep. Her step count has been fairly stable, indicating that there is no significant correlation between physical activity and her other symptoms.

Continue to: The clinician shows...

The clinician shows Ms. T the data to help her understand why a trial of cognitive-behavioral therapy for insomnia, or at least improving sleep hygiene, may offer several benefits. The clinician advises her to continue to use the app to help assess her response to these interventions and monitor her progress in real time.

Dilemma: The ethics of continuous observation

The rich data captured by digital phenotyping afford many clinical opportunities, but also raise concerns. Among these are 3 significant ethical implications.

Firstly, the same data that may help a clinician learn about what environments are associated with less anxiety for the patient may also reveal personal details about where that patient has been or with whom they have interacted. In the wrong hands, such personal data could cause harm. And even in the hands of a trusted clinician, a breach in the patient’s privacy begs the question: “Should such information be anyone’s business at all?”

Secondly, many apps that offer digital phenotyping could also store patient data—something that currently pervades social media and causes reasonable discomfort for many people. You might have personally encountered this with social media platforms such as Facebook. When it comes to mobile mental health apps, clinicians should carefully understand the data usage agreement of any digital phenotyping app they wish to use and then share this information with their patients.

Finally, while it is possible to collect the types of data outlined in this article, less is known about how to use it directly in clinical care. Understanding for each patient which data streams are most meaningful and which data streams are noise that should be ignored is an area of ongoing research. A good first step may be to begin with data streams that are known to be clinically relevant and valuable, such as sleep and physical activity.^9-11

Continue to: Discussion...

Although smartphones can gather a wide range of active and passive data, other data streams hold potential for predicting relapse and performing other clinically relevant actions. One data stream that could be of clinical use is genomic sequencing.¹² The genotyping of patients provides a wealth of information about the underlying biology, and genomic sequencing has never been cheaper.¹³

Combining the data gathered via digital phenotyping with that of genotyping could help elucidate the mechanisms by which specific diseases and symptoms occur. This could be very promising to better understand and treat our patients. However, as is the case with genomics, digital phenotyping has important ethical implications. If used in the proper way to benefit our patients, the future for this new method is bright.

Have your patients asked you about smartphone apps? If they haven’t yet, they may soon, as interest in apps for mental health continues to expand. There are now >10,000 mental health–related smartphone apps.¹ The rapid rise of these apps is partly due to their potential to transform a patient’s smartphone into a monitoring and therapeutic platform, capable of capturing mental health symptoms in real time and delivering on-the-go therapy. Setting aside questions about the potential of mobile health, 2 urgent questions remain for the busy psychiatrist in clinical practice: What is the current evidence base for mental health apps, and what should you tell your patients about them?

For most apps, evidence of efficacy is limited

While the evidence base for mental health smartphone apps continues to expand, for many of these apps, there is no evidence of effectiveness. The growing consensus is that most commercially available apps are not evidence-based and some are even dangerous. For example, researchers who examined >700 mindfulness apps on the iTunes and Google Play stores found that only 4% provided acceptable mindfulness training and education.² Another study of 58 apps that claimed to offer sobriety assessments found that none had ever been formally evaluated.³ Evidence-based reviews of suicide prevention apps have identified potentially harmful apps,⁴ and studies evaluating apps for bipolar disorder⁵ and depression⁶ have yielded similar results—few have any evidence supporting their use, and some offer dangerous and harmful advice. For example, researchers found that one app for bipolar disorder advised patients who are experiencing a manic episode to drink alcohol.⁵ Currently, the vast majority of commercially available apps are not appropriate for clinical care. This finding is not unique to mental health; similar findings have been reported for apps for cancer.⁷ The bottom line is that the apps that your patients are finding, and perhaps already using, may not be useful or effective.

However, early studies have demonstrated efficacy of some apps for several conditions, including schizophrenia,⁸ depression,⁹ anxiety disorders,¹⁰ and suicidal ideation.¹¹ Although many of the apps evaluated in these studies are not available to the public, or still require large-scale assessment before they are ready for mainstream clinical care, this research demonstrates that mental health apps can help improve treatment outcomes. As this research develops, a wave of evidence-based and effective mental health apps may be available in the near future.

Although it is unknown how many patients are presently using mental health apps, there is strong anecdotal evidence that an increasing number of patients who use these apps and other forms of digital technology are finding some benefits. In many cases, patients may actually be ahead of the research. For example, one study that conducted an online survey of patients with schizophrenia noted that some patients are using their smartphones to play music to help block auditory hallucinations.¹²

Why online reviews are of limited use

As this evidence continues to mature, and with an ever-growing number of mental health apps available on commercial marketplaces, busy psychiatrists need to navigate this complex space. Even psychiatrists who decide to not use apps as part of care still need to be knowledgeable about them, because patients are likely to ask about the benefits of using apps, and they will expect an informed response. How would you reply if your patient asked you about a new mood-tracking app he or she recently heard about? On what would you base your recommendation and opinion?

Reading online app reviews for guidance is not a good solution. A recent study found little relationship between the star ratings of health apps and the quality of those apps,¹³ which suggests that a 5-star rating on the app store is of limited use.

Unlike medications whose ingredients do not change over time, or manualized psychotherapies that use specific protocols, mental health apps are dynamic and constantly changing.¹⁴ Think of how often the apps on your smartphone update. Thus, the version of a mental health app that your patient downloads today may be very different from the version that received a favorable user review last month. And just as there is no single medication or therapy that is ideal for every patient, neither is there a single “best” app for all patients with the same disorder. Picking an app is a personal decision that cannot be made based on a single score or numeric rating. Furthermore, the validity of app rating systems is unclear. One study found a wide variation in the interrater reliability of measures used to evaluate apps from sources that included PsyberGuide, the Anxiety and Depression Association of America, and the research literature. Quality measures such as effectiveness, ease of use, and performance had relatively poor interrater reliability.¹⁵ This means that, for example, an app that one patient finds “easy to use” may be difficult to use for another. Thus, providing patients with suggestions based on an app’s ratings may result in providing information that sounds useful, but often is misleading.

One possible solution is a risk-based and personalized assessment approach to evaluating mental health apps. Although it does not offer scoring or recommendations of specific apps, the American Psychiatric Association (APA) App Evaluation Model (Figure) provides a framework to guide discussion and informed decision-making about apps. (The authors of this article helped create this model, but receive no compensation for that volunteer work.) The pyramid shape reflects the hierarchical nature of the model. To begin the process, start at the base of the pyramid and work upward.

Ground. First, consider the context of the app by determining basic facts, such as who made it, how much it costs, and its technology requirements. This ground layer establishes the credibility of the app’s creator by questioning his or her reputation, ability to update the app, and funding sources. Understanding the app’s business model also will help you determine whether the app will stand the test of time: Will it continue to exist next month or next year, or will a lack of reliable funding lead the vendor to abandon it?

Risk. The next layer assesses the risk, privacy, and security features of the app. Many mental health apps actively aim to avoid falling under the jurisdiction of U.S. federal health care privacy rules, such as the Health Insurance Portability and Accountability Act of 1996, so there is no guarantee that sensitive data supplied to an app will be protected. The true cost of a “free” app often is your patient’s personal mental health information, which the app’s developer may accumulate and sell for profit. Thus, it is wise to check the privacy policy to learn where your patient’s data goes. Furthermore, patients and psychiatrists must be vigilant that malware-infected apps can be uploaded to the app store, which can further compromise privacy.¹⁶ You may be surprised to learn that many apps lack a privacy policy, which means there are no protections for personal information or safeguards against the misuse of mental health data.¹⁷ Checking that an app at least promises to digitally protect mental health data through encryption and secure storage also is a good step.

The goal of considering these factors is not to create a score, but rather to be aware of them and consider them in the context of the specific app, patient, and clinical situation. Doing so helps determine whether the app meets the appropriate risk, privacy, and security standards for your patient.

Evidence. The next layer of the evaluation framework is evidence. The goal is to seek an app with clinical evidence of effectiveness. Simply put, if a patient is going to use an app, he should use one that works. An app without formal evidence may be effective, but it is important to make sure the patient is aware that these claims have not been verified. Many apps claim that they offer cognitive-behavioral therapy or mindfulness therapy, but few deliver on such claims.¹⁸ It is wise to try an app before recommending it to a patient to ensure that it does what it claims it does, and does not offer dangerous or harmful recommendations.

Ease of use. Across all health apps, there is growing recognition that most downloaded apps are never used. Patient engagement with mental health apps appears to rapidly decline over the first week of use.¹⁹ There also is emerging evidence that many apps are not user-friendly. A recent study of several common mood-tracking apps found that patients with depression had difficulty entering and accessing their data.²⁰ Because many psychiatric disorders are chronic or last at least several months, it is especially important to consider how engaging and usable the app will be for your patient. Usability varies from patient to patient, so it is best to check directly with your patient regarding his comfort with apps and mobile technology. Offering check-ins and support to help patients keep on track with apps may be critical for successful outcomes.

Interoperability. The final layer of the model is data sharing and interoperability. It is important to determine if the data collected or generated by the app are available to you, the patient, the treatment team, and others involved in the patient’s care. As mental health treatment moves toward integrated care, apps that fragment care (by not sharing information) impede care. Check if the app can share data with an electronic medical record, or if there is a plan to review and act on data from the app as part of your patient’s treatment plan.

More information about the APA App Evaluation Model, including additional factors to consider within each layer, is available from the APA for free at https://www.psychiatry.org/psychiatrists/practice/mental-health-apps/app-evaluation-model. For a sample of factors to consider when evaluating a mental health app, see the Table.

A reasonable strategy

Although the APA App Evaluation Model does not endorse any particular app, it can help guide more informed decision-making. As the evidence on mental health apps continues to evolve, it will become easier to make definitive statements on what constitutes a useful app. For now, the best strategy when discussing mental health apps with patients is to combine the use of this model with your clinical judgment.

Have your patients asked you about smartphone apps? If they haven’t yet, they may soon, as interest in apps for mental health continues to expand. There are now >10,000 mental health–related smartphone apps.¹ The rapid rise of these apps is partly due to their potential to transform a patient’s smartphone into a monitoring and therapeutic platform, capable of capturing mental health symptoms in real time and delivering on-the-go therapy. Setting aside questions about the potential of mobile health, 2 urgent questions remain for the busy psychiatrist in clinical practice: What is the current evidence base for mental health apps, and what should you tell your patients about them?

For most apps, evidence of efficacy is limited

While the evidence base for mental health smartphone apps continues to expand, for many of these apps, there is no evidence of effectiveness. The growing consensus is that most commercially available apps are not evidence-based and some are even dangerous. For example, researchers who examined >700 mindfulness apps on the iTunes and Google Play stores found that only 4% provided acceptable mindfulness training and education.² Another study of 58 apps that claimed to offer sobriety assessments found that none had ever been formally evaluated.³ Evidence-based reviews of suicide prevention apps have identified potentially harmful apps,⁴ and studies evaluating apps for bipolar disorder⁵ and depression⁶ have yielded similar results—few have any evidence supporting their use, and some offer dangerous and harmful advice. For example, researchers found that one app for bipolar disorder advised patients who are experiencing a manic episode to drink alcohol.⁵ Currently, the vast majority of commercially available apps are not appropriate for clinical care. This finding is not unique to mental health; similar findings have been reported for apps for cancer.⁷ The bottom line is that the apps that your patients are finding, and perhaps already using, may not be useful or effective.

However, early studies have demonstrated efficacy of some apps for several conditions, including schizophrenia,⁸ depression,⁹ anxiety disorders,¹⁰ and suicidal ideation.¹¹ Although many of the apps evaluated in these studies are not available to the public, or still require large-scale assessment before they are ready for mainstream clinical care, this research demonstrates that mental health apps can help improve treatment outcomes. As this research develops, a wave of evidence-based and effective mental health apps may be available in the near future.

Although it is unknown how many patients are presently using mental health apps, there is strong anecdotal evidence that an increasing number of patients who use these apps and other forms of digital technology are finding some benefits. In many cases, patients may actually be ahead of the research. For example, one study that conducted an online survey of patients with schizophrenia noted that some patients are using their smartphones to play music to help block auditory hallucinations.¹²

Why online reviews are of limited use

As this evidence continues to mature, and with an ever-growing number of mental health apps available on commercial marketplaces, busy psychiatrists need to navigate this complex space. Even psychiatrists who decide to not use apps as part of care still need to be knowledgeable about them, because patients are likely to ask about the benefits of using apps, and they will expect an informed response. How would you reply if your patient asked you about a new mood-tracking app he or she recently heard about? On what would you base your recommendation and opinion?

Reading online app reviews for guidance is not a good solution. A recent study found little relationship between the star ratings of health apps and the quality of those apps,¹³ which suggests that a 5-star rating on the app store is of limited use.

Unlike medications whose ingredients do not change over time, or manualized psychotherapies that use specific protocols, mental health apps are dynamic and constantly changing.¹⁴ Think of how often the apps on your smartphone update. Thus, the version of a mental health app that your patient downloads today may be very different from the version that received a favorable user review last month. And just as there is no single medication or therapy that is ideal for every patient, neither is there a single “best” app for all patients with the same disorder. Picking an app is a personal decision that cannot be made based on a single score or numeric rating. Furthermore, the validity of app rating systems is unclear. One study found a wide variation in the interrater reliability of measures used to evaluate apps from sources that included PsyberGuide, the Anxiety and Depression Association of America, and the research literature. Quality measures such as effectiveness, ease of use, and performance had relatively poor interrater reliability.¹⁵ This means that, for example, an app that one patient finds “easy to use” may be difficult to use for another. Thus, providing patients with suggestions based on an app’s ratings may result in providing information that sounds useful, but often is misleading.

One possible solution is a risk-based and personalized assessment approach to evaluating mental health apps. Although it does not offer scoring or recommendations of specific apps, the American Psychiatric Association (APA) App Evaluation Model (Figure) provides a framework to guide discussion and informed decision-making about apps. (The authors of this article helped create this model, but receive no compensation for that volunteer work.) The pyramid shape reflects the hierarchical nature of the model. To begin the process, start at the base of the pyramid and work upward.

Ground. First, consider the context of the app by determining basic facts, such as who made it, how much it costs, and its technology requirements. This ground layer establishes the credibility of the app’s creator by questioning his or her reputation, ability to update the app, and funding sources. Understanding the app’s business model also will help you determine whether the app will stand the test of time: Will it continue to exist next month or next year, or will a lack of reliable funding lead the vendor to abandon it?

Risk. The next layer assesses the risk, privacy, and security features of the app. Many mental health apps actively aim to avoid falling under the jurisdiction of U.S. federal health care privacy rules, such as the Health Insurance Portability and Accountability Act of 1996, so there is no guarantee that sensitive data supplied to an app will be protected. The true cost of a “free” app often is your patient’s personal mental health information, which the app’s developer may accumulate and sell for profit. Thus, it is wise to check the privacy policy to learn where your patient’s data goes. Furthermore, patients and psychiatrists must be vigilant that malware-infected apps can be uploaded to the app store, which can further compromise privacy.¹⁶ You may be surprised to learn that many apps lack a privacy policy, which means there are no protections for personal information or safeguards against the misuse of mental health data.¹⁷ Checking that an app at least promises to digitally protect mental health data through encryption and secure storage also is a good step.

The goal of considering these factors is not to create a score, but rather to be aware of them and consider them in the context of the specific app, patient, and clinical situation. Doing so helps determine whether the app meets the appropriate risk, privacy, and security standards for your patient.

Evidence. The next layer of the evaluation framework is evidence. The goal is to seek an app with clinical evidence of effectiveness. Simply put, if a patient is going to use an app, he should use one that works. An app without formal evidence may be effective, but it is important to make sure the patient is aware that these claims have not been verified. Many apps claim that they offer cognitive-behavioral therapy or mindfulness therapy, but few deliver on such claims.¹⁸ It is wise to try an app before recommending it to a patient to ensure that it does what it claims it does, and does not offer dangerous or harmful recommendations.

Ease of use. Across all health apps, there is growing recognition that most downloaded apps are never used. Patient engagement with mental health apps appears to rapidly decline over the first week of use.¹⁹ There also is emerging evidence that many apps are not user-friendly. A recent study of several common mood-tracking apps found that patients with depression had difficulty entering and accessing their data.²⁰ Because many psychiatric disorders are chronic or last at least several months, it is especially important to consider how engaging and usable the app will be for your patient. Usability varies from patient to patient, so it is best to check directly with your patient regarding his comfort with apps and mobile technology. Offering check-ins and support to help patients keep on track with apps may be critical for successful outcomes.

Interoperability. The final layer of the model is data sharing and interoperability. It is important to determine if the data collected or generated by the app are available to you, the patient, the treatment team, and others involved in the patient’s care. As mental health treatment moves toward integrated care, apps that fragment care (by not sharing information) impede care. Check if the app can share data with an electronic medical record, or if there is a plan to review and act on data from the app as part of your patient’s treatment plan.

More information about the APA App Evaluation Model, including additional factors to consider within each layer, is available from the APA for free at https://www.psychiatry.org/psychiatrists/practice/mental-health-apps/app-evaluation-model. For a sample of factors to consider when evaluating a mental health app, see the Table.

A reasonable strategy

Although the APA App Evaluation Model does not endorse any particular app, it can help guide more informed decision-making. As the evidence on mental health apps continues to evolve, it will become easier to make definitive statements on what constitutes a useful app. For now, the best strategy when discussing mental health apps with patients is to combine the use of this model with your clinical judgment.

Have your patients asked you about smartphone apps? If they haven’t yet, they may soon, as interest in apps for mental health continues to expand. There are now >10,000 mental health–related smartphone apps.¹ The rapid rise of these apps is partly due to their potential to transform a patient’s smartphone into a monitoring and therapeutic platform, capable of capturing mental health symptoms in real time and delivering on-the-go therapy. Setting aside questions about the potential of mobile health, 2 urgent questions remain for the busy psychiatrist in clinical practice: What is the current evidence base for mental health apps, and what should you tell your patients about them?

For most apps, evidence of efficacy is limited

While the evidence base for mental health smartphone apps continues to expand, for many of these apps, there is no evidence of effectiveness. The growing consensus is that most commercially available apps are not evidence-based and some are even dangerous. For example, researchers who examined >700 mindfulness apps on the iTunes and Google Play stores found that only 4% provided acceptable mindfulness training and education.² Another study of 58 apps that claimed to offer sobriety assessments found that none had ever been formally evaluated.³ Evidence-based reviews of suicide prevention apps have identified potentially harmful apps,⁴ and studies evaluating apps for bipolar disorder⁵ and depression⁶ have yielded similar results—few have any evidence supporting their use, and some offer dangerous and harmful advice. For example, researchers found that one app for bipolar disorder advised patients who are experiencing a manic episode to drink alcohol.⁵ Currently, the vast majority of commercially available apps are not appropriate for clinical care. This finding is not unique to mental health; similar findings have been reported for apps for cancer.⁷ The bottom line is that the apps that your patients are finding, and perhaps already using, may not be useful or effective.

However, early studies have demonstrated efficacy of some apps for several conditions, including schizophrenia,⁸ depression,⁹ anxiety disorders,¹⁰ and suicidal ideation.¹¹ Although many of the apps evaluated in these studies are not available to the public, or still require large-scale assessment before they are ready for mainstream clinical care, this research demonstrates that mental health apps can help improve treatment outcomes. As this research develops, a wave of evidence-based and effective mental health apps may be available in the near future.

Although it is unknown how many patients are presently using mental health apps, there is strong anecdotal evidence that an increasing number of patients who use these apps and other forms of digital technology are finding some benefits. In many cases, patients may actually be ahead of the research. For example, one study that conducted an online survey of patients with schizophrenia noted that some patients are using their smartphones to play music to help block auditory hallucinations.¹²

Why online reviews are of limited use

As this evidence continues to mature, and with an ever-growing number of mental health apps available on commercial marketplaces, busy psychiatrists need to navigate this complex space. Even psychiatrists who decide to not use apps as part of care still need to be knowledgeable about them, because patients are likely to ask about the benefits of using apps, and they will expect an informed response. How would you reply if your patient asked you about a new mood-tracking app he or she recently heard about? On what would you base your recommendation and opinion?

Reading online app reviews for guidance is not a good solution. A recent study found little relationship between the star ratings of health apps and the quality of those apps,¹³ which suggests that a 5-star rating on the app store is of limited use.

Unlike medications whose ingredients do not change over time, or manualized psychotherapies that use specific protocols, mental health apps are dynamic and constantly changing.¹⁴ Think of how often the apps on your smartphone update. Thus, the version of a mental health app that your patient downloads today may be very different from the version that received a favorable user review last month. And just as there is no single medication or therapy that is ideal for every patient, neither is there a single “best” app for all patients with the same disorder. Picking an app is a personal decision that cannot be made based on a single score or numeric rating. Furthermore, the validity of app rating systems is unclear. One study found a wide variation in the interrater reliability of measures used to evaluate apps from sources that included PsyberGuide, the Anxiety and Depression Association of America, and the research literature. Quality measures such as effectiveness, ease of use, and performance had relatively poor interrater reliability.¹⁵ This means that, for example, an app that one patient finds “easy to use” may be difficult to use for another. Thus, providing patients with suggestions based on an app’s ratings may result in providing information that sounds useful, but often is misleading.

One possible solution is a risk-based and personalized assessment approach to evaluating mental health apps. Although it does not offer scoring or recommendations of specific apps, the American Psychiatric Association (APA) App Evaluation Model (Figure) provides a framework to guide discussion and informed decision-making about apps. (The authors of this article helped create this model, but receive no compensation for that volunteer work.) The pyramid shape reflects the hierarchical nature of the model. To begin the process, start at the base of the pyramid and work upward.

Ground. First, consider the context of the app by determining basic facts, such as who made it, how much it costs, and its technology requirements. This ground layer establishes the credibility of the app’s creator by questioning his or her reputation, ability to update the app, and funding sources. Understanding the app’s business model also will help you determine whether the app will stand the test of time: Will it continue to exist next month or next year, or will a lack of reliable funding lead the vendor to abandon it?

Risk. The next layer assesses the risk, privacy, and security features of the app. Many mental health apps actively aim to avoid falling under the jurisdiction of U.S. federal health care privacy rules, such as the Health Insurance Portability and Accountability Act of 1996, so there is no guarantee that sensitive data supplied to an app will be protected. The true cost of a “free” app often is your patient’s personal mental health information, which the app’s developer may accumulate and sell for profit. Thus, it is wise to check the privacy policy to learn where your patient’s data goes. Furthermore, patients and psychiatrists must be vigilant that malware-infected apps can be uploaded to the app store, which can further compromise privacy.¹⁶ You may be surprised to learn that many apps lack a privacy policy, which means there are no protections for personal information or safeguards against the misuse of mental health data.¹⁷ Checking that an app at least promises to digitally protect mental health data through encryption and secure storage also is a good step.

The goal of considering these factors is not to create a score, but rather to be aware of them and consider them in the context of the specific app, patient, and clinical situation. Doing so helps determine whether the app meets the appropriate risk, privacy, and security standards for your patient.

Evidence. The next layer of the evaluation framework is evidence. The goal is to seek an app with clinical evidence of effectiveness. Simply put, if a patient is going to use an app, he should use one that works. An app without formal evidence may be effective, but it is important to make sure the patient is aware that these claims have not been verified. Many apps claim that they offer cognitive-behavioral therapy or mindfulness therapy, but few deliver on such claims.¹⁸ It is wise to try an app before recommending it to a patient to ensure that it does what it claims it does, and does not offer dangerous or harmful recommendations.

Ease of use. Across all health apps, there is growing recognition that most downloaded apps are never used. Patient engagement with mental health apps appears to rapidly decline over the first week of use.¹⁹ There also is emerging evidence that many apps are not user-friendly. A recent study of several common mood-tracking apps found that patients with depression had difficulty entering and accessing their data.²⁰ Because many psychiatric disorders are chronic or last at least several months, it is especially important to consider how engaging and usable the app will be for your patient. Usability varies from patient to patient, so it is best to check directly with your patient regarding his comfort with apps and mobile technology. Offering check-ins and support to help patients keep on track with apps may be critical for successful outcomes.

Interoperability. The final layer of the model is data sharing and interoperability. It is important to determine if the data collected or generated by the app are available to you, the patient, the treatment team, and others involved in the patient’s care. As mental health treatment moves toward integrated care, apps that fragment care (by not sharing information) impede care. Check if the app can share data with an electronic medical record, or if there is a plan to review and act on data from the app as part of your patient’s treatment plan.

More information about the APA App Evaluation Model, including additional factors to consider within each layer, is available from the APA for free at https://www.psychiatry.org/psychiatrists/practice/mental-health-apps/app-evaluation-model. For a sample of factors to consider when evaluating a mental health app, see the Table.

A reasonable strategy

Although the APA App Evaluation Model does not endorse any particular app, it can help guide more informed decision-making. As the evidence on mental health apps continues to evolve, it will become easier to make definitive statements on what constitutes a useful app. For now, the best strategy when discussing mental health apps with patients is to combine the use of this model with your clinical judgment.