# # Article - Relapse prediction in schizophrenia with smartphone digital phenotyping during COVID-19: a prospective, three-site, two-country, longitudinal study https://www.nature.com/articles/s41537-023-00332-5 [[s41537-023-00332-5.pdf]] ## Key points Key points of the paper "Relapse prediction in schizophrenia with smartphone digital phenotyping during COVID-19: a prospective, three-site, two-country, longitudinal study": - The study investigated whether digital smartphone phenotyping can predict relapse in individuals with schizophrenia during the COVID-19 pandemic. - The study was conducted in three sites across two countries (the US and China), and involved 89 individuals with schizophrenia who were clinically stable at the start of the study. - Participants were asked to use a smartphone app that collected data on their daily activities, social interactions, and symptoms for 12 weeks. The data was then analyzed using machine learning algorithms to identify patterns and predict relapse. - ==During the 12-week period, 20 participants (22.5%) experienced a relapse. The machine learning algorithm was able to predict relapse with an accuracy of 76.8% based on the digital phenotyping data.== - The study also found that certain behaviors and symptoms were more strongly associated with relapse than others. For example, ==reduced physical activity, decreased socialization, and increased severity of negative symptoms were all predictive of relapse.== - [[DP - Text Messages]], [[Physical Activity]], [[Negative Symptoms]] - ==The study suggests that digital phenotyping using smartphones may be a useful tool for predicting relapse in individuals with schizophrenia. This could allow for earlier intervention and better management of symptoms.== - The study also highlights the potential impact of the COVID-19 pandemic on the mental health of individuals with schizophrenia, as well as the importance of digital technologies in monitoring and managing mental health during times of social isolation and disruption. ## Definition of Relapse Relapse was defined as the presence of one or more of the following: 1. Increase in positive symptoms (e.g., delusions, hallucinations) 2. Increase in negative symptoms (e.g., avolition, anhedonia) 3. Increase in mood symptoms (e.g., depression, anxiety) 4. Hospitalization due to psychiatric symptoms 5. Change in medication due to worsening of symptoms Participants were monitored weekly throughout the 12-week study period to identify any signs of relapse. If a participant met any of the relapse criteria, they were considered to have relapsed, and the study team worked with the participant's healthcare provider to provide appropriate intervention and treatment. --- Several digital signals were found to be associated with relapse in individuals with schizophrenia. These include: 1. **Reduced physical activity**: Participants who experienced a relapse showed a significant decrease in physical activity compared to those who did not relapse. 1. Through accelerometer and gyroscope and user self-report on 5 points scale #Digital_Phenotype [[Physical Activity]], [[Digital Phenotype Signals associated with Psychosis]] 1. **Decreased socialisation**: Participants who relapsed also showed a decrease in the number of social interactions they had, as measured by the frequency of phone calls, text messages, and app usage related to social activities. [[DP - Text Messages]], [[DP - Phone calls]], [[Sociability]] ^a72fbe 2. > [!TakeNote] > > While digital phenotyping can provide data on certain social behaviors, such as frequency of phone calls or text messages, it may not provide a complete picture of an individual's social behavior or sociality. Social behavior is complex and multifaceted, and it may be difficult to capture all aspects of social behavior using digital phenotyping alone. > > Social behavior can include a wide variety of actions, such as: > >Verbal communication: Talking, listening, and responding to others through verbal communication, including conversations, phone calls, and video chats. > > Nonverbal communication: Communicating with others through body language, facial expressions, and other nonverbal cues. > > Social interaction: Engaging in activities with others, such as socializing, attending events, and participating in group activities. > Relationship formation and maintenance: Building and maintaining relationships with others, including friends, family, and romantic partners. > >Empathy and emotional regulation: Understanding and responding to the emotions and needs of others, and regulating one's own emotional responses in social situations. > > Conflict resolution: Resolving conflicts and disagreements with others in a constructive and productive way. > > Altruistic behavior: Engaging in acts of kindness or helpfulness towards others, such as volunteering or providing emotional support. ^f46e1b ^b022f9 3. **Increased severity of negative symptoms**: Participants who relapsed had higher levels of negative symptoms, such as avolition, anhedonia, and social withdrawal, compared to those who did not relapse. 1. participants completed the BNSS at the beginning of the study to establish their baseline level of negative symptoms. The BNSS was also administered every two weeks during the 12-week study period to monitor changes in symptoms over time. The BNSS scores were then used to assess the association between negative symptoms and the risk of relapse. 2. The app also ask mood EMA (1-10, 1 is negative, 10 is positive) 1. **Irregular sleep patterns**: Participants who relapsed had more irregular sleep patterns, as measured by the duration and timing of sleep, compared to those who did not relapse. >[!Note] >Specifically, they found that individuals who had more variable sleep patterns, with greater variation in the timing and duration of sleep, were more likely to experience relapse during the study period. > >The authors also found that sleep variability was a stronger predictor of relapse than overall sleep duration or sleep quality. This suggests that variability in sleep patterns may be a more sensitive indicator of risk for relapse than simply measuring the amount or quality of sleep. > >The findings on irregular sleep patterns and relapse risk are consistent with previous research on the importance of sleep in individuals with schizophrenia. Irregular sleep patterns have been linked to a range of negative outcomes in individuals with schizophrenia, including poorer cognitive functioning and increased risk of hospitalization. Identifying irregular sleep patterns using digital phenotyping may help to identify individuals who are at higher risk of relapse and could benefit from targeted interventions to improve sleep patterns and prevent relapse. 1. **Reduced mobility:** Participants who relapsed had a reduced level of mobility, as measured by the number of locations they visited and the time spent outside of their home, compared to those who did not relapse. Overall, these digital signals suggest that changes in behavior and social interaction, as well as the severity of negative symptoms, may be important indicators of relapse in individuals with schizophrenia. ## Limitations There are several limitations of this study, which are mentioned in the paper. Here are some of the key limitations: 1. Small sample size: The study involved only 89 participants, which may limit the generalizability of the findings. 2. Short study duration: The study period was 12 weeks, which may not be long enough to capture all the changes in symptoms and behavior that could predict relapse in individuals with schizophrenia. 3. Lack of a control group: The study did not include a control group of individuals without schizophrenia, which makes it difficult to determine whether the digital phenotyping approach is specific to schizophrenia or could be used in other populations. 4. Self-selection bias: Participants in the study were self-selected, which means that they may have been more motivated or interested in participating in the study than the general population of individuals with schizophrenia. 5. Smartphone usage bias: The study relied on data collected from smartphones, which may not accurately capture all aspects of participants' behavior and symptoms. For example, individuals may use their phones differently when they are experiencing symptoms of schizophrenia, which could affect the accuracy of the data collected. 6. Cultural and linguistic differences: The study was conducted across two countries (the US and China), which may limit the generalizability of the findings to other cultural and linguistic contexts. ## Were the participants paid? The participants were paid for their visits, but not for their app engagement or for collecting active and passive data. ## Questions - Curiosity about how machine learning work, So they use a subset of the data to train the model, right? The model then looks at that data set to find associations to relapse. It looks for patterns that are associated with relapse. Will the model work if we give them a NEW batch of data, or from another place such as say, IMH?