Project Summary
"Using Meta-level Smartphone Data to Promote Early Intervention in Schizophrenia”
Schizophrenia is one of the most debilitating disorders in the world today. It affects over 2.4 million
adult Americans each year. NIMH director Dr. Thomas Insel has declared “The best chance for preventing
serious functional disability among people with schizophrenia may be to intervene at the earliest stages of the
disorder, at the first episode of psychosis or even before symptoms appear. However, to act before symptoms
appear requires improved predictive capacity” (NIMH 2011 Budget). Creating tools to identify high-risk,
`prodromal' individuals may be the single most important step towards developing effective interventions to
reduce the duration of untreated psychosis (DUP), and thereby also reduce the morbidity and mortality
associated with schizophrenia. Recent studies have shown that over 54% of individuals with schizophrenia are
re-hospitalized within the first 12 months following their initial hospitalization. Even after the first
hospitalization, preventing relapse and re-hospitalization may lessen the long-term severity of the illness. In
this SBIR Phase I study, we propose to determine the feasibility of screening for prodromal individuals and
individuals at high risk of relapse by applying interpretive algorithms to Passively Gathered Meta-level
Smartphone Data (PGMSD). We hypothesize that PGMSD can effectively assist in screening for prodromal
individuals who are progressing toward psychosis as well as for remotely assessing individuals at risk for
relapse during the critical 12-month period following their first episode of psychosis (FEP).
In Phase I, we plan to recruit 70 individuals who have been or are being evaluated at the Prodromal
clinics at Columbia, UCSD, and UCLA, where an estimated 70 to 90% of clients already own Smartphones.
Data gathered may include: the frequency of telephone calls, emails, and texts, to assess within person
changes in social connectedness; GPS, accelerometer data, to assess physical activity, isolation, and sleep
patterns. In the past, several IRB-approved studies have used smartphones for gathering similar data from
patients. Algorithms will be developed using several techniques including machine learning to convert the
meta-level data into measures of social functioning, physical isolation, physical activity, and sleep/wake
reversals. In addition to achieving technological success, our goal in Phase I is to provide evidence of our
ability to use PGMSD algorithms to differentiating group means of participants who are controls, prodromal, or
experiencing their FEP (SIPS 1 or 2; 3, 4, or 5; or 6).
In Phase II, we will further develop and validate these algorithms. If successful, the Phase II project will
have a large and sustained impact as our algorithms will help (1) identify at-risk individuals who `are' or `are
not' progressing toward conversion, (2) serve as an objective measure of treatment effectiveness; (3) give rise
to clinical reports delivered to EHR systems that hold promise for preventing relapse during the critical 12
months after initial diagnosis, potentially reducing hospitalization and re-hospitalization rates.