Using Mobile Technology To Detect Early Warning Signs of Mental Health Challenges and Enhance Treatment Delivery For Youth
Using Mobile Technology To Detect Early Warning Signs of Mental Health Challenges and Enhance Treatment Delivery For Youth
This project tests the feasibility of implementing a smartphone application - Ginger.io - in the UC Davis Early Psychosis Program, and investigates whether mobile health technology can improve treatment delivery and outcomes in individuals with early psychosis. Ginger.io is a smartphone application that utilizes methods of passive data collection (i.e. data gathered without active interaction/contribution from the user) to gather communication, movement, and interaction data from smartphone devices to model individuals' social, physical, and mental health. These models are used to infer health-related outcomes and could inform treatment. By implementing the Ginger.io application in the UC Davis Early Psychosis Program with an integrated clinical and research infrastructure, the investigators will be able to quickly determine its feasibility for use in early psychosis populations, while simultaneously developing its ability to systematically capture aspects of relapse and recovery that are unique to this patient population.
Objectives: This project has three principle objectives related to early psychosis care: 1) improve treatment delivery, 2) improve patient outcomes, and 3) lower treatment costs. The project will target individuals in the early stages of psychotic illness, including individuals at high risk for developing a psychotic illness (termed "clinical high risk" or CHR) and individuals within three years of their first psychotic episode (termed "first episode psychosis" or FEP). The early stages of psychotic illness represent a critical period for intervention; early identification of clinical deterioration and subsequent targeted intervention is crucial for rapid remission of symptoms and reduced relapse rates. However, without the information necessary to identify patients in need of such intervention, providers are limited in their ability to respond rapidly. Within the UCD Early Psychosis Program, a mobile health application such as Ginger.io has the potential to equip the providers and caregivers with valuable insight into a patient's status in real-time without the burden of increased appointments and intrusive monitoring, allowing the identification of early psychosis patients most in need of outreach, and routing of treatment resources to the right patients at the right time.
The project comprises two phases. In both phases, patients will use the Ginger.io app on their mobile device to complete daily surveys assessing mood and social interactions, and weekly surveys assessing symptoms. Patients will also complete monthly gold-standard clinical interviews at the UC Davis Early Psychosis Program. Once enrolled, the investigators will ask patients to participate for the duration of the study, a total of 14 months across Phase I and II.
Phase I involves the collection of pilot data to assess feasibility and conduct exploratory analyses. Participants will include individuals in the earliest stages of psychosis (CHR and FEP). The investigators will recruit 60 participants to allow for 15% dropout, aiming to obtain complete data on 50 participants. Given the emphasis on feasibility and acceptability in Phase I, research staff will monitor patient Dashboards daily and provide clinicians with weekly summaries for each patient. Clinicians and patients will provide ratings on satisfaction and effect of Ginger.io on treatment at the end of Phase I.
Specific aims for Phase I are as follows:
Specific Aim 1: Determine the feasibility and acceptability of Ginger.io in an early psychosis population.
Feasibility will be measured by utilization of the application (i.e. number of enrolled and active participants versus total patient population invited). Acceptability will be measured with satisfaction survey ratings of Ginger.io by patients and clinicians. Analysis will include summary statistics reporting utilization rates and satisfaction ratings. The investigators hypothesize high utilization of Ginger.io and low dropout, as well as high satisfaction and endorsement of continued use of Ginger.io.
Specific Aim 2: Identify passive behavior data associated with gold-standard measures of clinical symptoms and social functioning.
The investigators will examine associations between passive and active data collected via Ginger.io. The investigators will then validate both the active and passive behavior data by examining associations with symptom and social functioning ratings obtained via gold-standard clinical interviews conducted at monthly research evaluations. The investigators hypothesize patients' passive interaction data will positively associate with active survey data of symptoms and social interactions (e.g. increase reports of mood will relate to more texts/calls), and that both active and passive data regarding symptoms and social behavior will relate to gold-standard clinical measures of symptoms and social functioning.
Specific Aim 3: Identify Ginger.io data related to key patient outcomes to inform calibration of patient status alerts.
Exploratory analyses using regression models and repeated measure GLMs will examine whether the passive and active data from Ginger.io statistically predict key patient outcomes, including psychotic symptom flares, increased clinic contact, psychiatric ER visits or hospitalizations, and progression from a high risk state to first episode psychosis. The investigators hypothesize reductions in phone calls/text messages and/or increases in self-reported ratings of basic symptoms will predict psychotic symptom flares, increased clinic utilization, psychiatric ER visits and hospitalizations. This analysis will be used to calibrate patient status alerts for Phase II that will notify clinicians which patients are at risk of symptom exacerbation and relapse.
Specific Aim 4: Evaluate effect of Ginger.io on key treatment variables. Key variables of interest are medication adherence, therapeutic alliance, and insight. All three variables will be measured using self-report questionnaires (see Procedures Involved) administered to patients at the beginning and end of the project, and as part of the monthly clinical interviews, enabling the evaluation of change over time. The investigators hypothesize that, following three months of Ginger.io use, patients will report improved medication adherence, therapeutic alliance, and insight into their illness.
Phase II will extend the findings of Phase I through the implementation of Ginger.io in the UC Davis Early Psychosis Programs (EDAPT and SacEDAPT) to allow clinicians to directly interact with the Dashboard and integrate its information directly into patient care. Patients will continue to complete daily and weekly surveys via Ginger.io, and monthly clinical assessments with research staff. Additionally, clinicians will monitor patient Dashboards on a daily basis and will respond to patient status alerts as they arise (see Aim 2 below). Phase I participants will be asked to remain in the study through Phase II, yielding a more comprehensive longitudinal dataset. Phase II will involve a total of 120 participants (combined Phase I and newly recruited participants) to account for 15% dropout, aiming to obtain data for at least 100 participants.
Phase II has four specific aims.
Specific Aim 1: Continue to validate predictive value of Ginger.io data related to key patient outcomes.
This is a direct extension of Aim 3 in Phase I and will be addressed through continued linear modeling of passive and active data from Ginger.io to identify and test statistical predictors of key patient outcomes. This will facilitate continued calibration of patient status alerts used in the Dashboard to notify clinicians when patients are at risk of clinical deterioration.
Specific Aim 2: Evaluate effect of patient status alerts on treatment delivery. Key variables of interest include clinician-rated utility of the patient status alerts, clinician choice of treatment delivery following an alert, and the action response time following notification of an alert. In response to each alert, clinicians will indicate the type of clinical intervention they pursued (e.g. "called patient to check in", "scheduled appointment with psychiatrist"). All treatment will be clinician-driven and provided in a naturalistic manner, and data will be collected on the types of treatments chosen for a given alert. Clinicians will also rate the utility (e.g. not useful/useful) of the alert based on information gathered during follow-up. Action response time will be recorded via Dashboard interaction. Preliminary analyses will be exploratory and will inform continued improvements to the Dashboard and calibration of patient status alerts. Summary statistics regarding clinician responses, average response time, and utility ratings will be included. The investigators hypothesize that clinicians will rate patient status alerts as useful and informative to their treatment planning and intervention.
Specific Aim 3: Analysis of the impact of Ginger.io on cost of care. Key variables of interest include cost of outpatient care, and psychiatric ER visits and hospitalizations. To evaluate the impact of Ginger.io, the historical average monthly and annual cost of care per patient will be compared to the average monthly and annual cost of care during and following completion of the project. The investigators hypothesize that implementation of Ginger.io will reduce the number and duration of psychiatric ER visits and hospitalizations, thereby reducing costs.
Specific Aim 4: Continued evaluation of feasibility and acceptability of Ginger.io in an early psychosis population.
The investigators will continue to evaluate feasibility and acceptability of Ginger.io in both patients and clinicians using satisfaction surveys. The investigators hypothesize that patients will continue to endorse high satisfaction rates and continued use of Ginger.io. Additionally, the investigators hypothesize that clinicians will report high satisfaction rates with the Dashboard and endorse its continued use in clinical care.
Inclusion Criteria:
Exclusion Criteria: