Back to All Events

Protocol for Improving Smartphone Data Quality in Accelerated Medicines Partnership (AMP) Schizophrenia (SCZ)

Protocol for Improving Smartphone Data Quality in Accelerated Medicines Partnership (AMP) Schizophrenia (SCZ)

Erlend Lane, Harvard Medical School/ BIDMC Matt Flathers, Harvard Medical School/ BIDMC John Torous, Harvard Medical School/ BIDMC Scott Woods, Yale

Background: Digital Phenotyping data, particularly data gathered from smartphones, has become increasingly utilized in psychiatric research for its capacity to collect high volume, ecological data that is low participant burden and sensitive to change. Digital phenotyping data collection brings a variety of challenges distinct from cross sectional research, including specific smartphone configurations subject to change and ongoing engagement from participants, something difficult to sustain over the course of 12 months. Accelerated Medicines Partnership (AMP) Schizophrenia (SCZ) represents one of the largest studies currently collecting smartphone data, gathered across more than 40 sites, in eight different languages for up to one year.

Methods: Our team aims to outline and implement a protocol for support staff to improve digital phenotyping data quality across sites, based on evidence of data quality improvement efforts initiated within the project. Measurements will be comparisons of pre and post intervention data quality, measured in the volume of passive data points across participants across sites. This will help assess and establish best practices for the collection of smartphone data in psychiatric samples, particularly for large, multi-month studies.

Results: The results of our QI efforts will be utilizable and distributed as a model of data quality management for digital data in multisite, longitudinal studies of participants with psychosis. We have achieved prior results in this domain, in which our team produced an average data quality of .82 for 33 participants across 6 weeks in a digitally supported clinical program, and we seek to replicate this output in a much larger, longitudinal research environment.

Conclusion: This work integrates prior experience in improving data quality in single-site short term collection of digital phenotyping data to larger scale efforts. Understanding best practices for data quality maintenance in digital phenotyping study design is critical for this burgeoning field.