The problem of missing data for learning health systems focused on first-episode psychosis.

BACKGROUND

A Learning Health System (LHS) requires data to improve care.

DESIGN

Data are from the ESPRITO LHS that includes 13 US clinics providing coordinated specialty care (CSC) for first-episode psychosis. Causes of missing data examined were: clinic patients not enrolling in ESPRITO, participants prematurely disengaging from treatment and missing patient-reported outcomes.

RESULTS

ESPRITO informed consent used a verbal opt-out format. This resulted in a high participant agreement rate (83.5 %) but limitations on data sharing within ESPRITO. During a 6-month period, 15.4 % of ESPRITO participants prematurely terminated treatment. An exploratory analysis revealed factors associated with increased premature termination likelihood: being homeless or having unstable housing, not being prescribed a long-acting injectable antipsychotic and factors associated with decreased premature termination likelihood: having commercial insurance, longer duration of CSC treatment, better scores on the Global Functioning: Social Scale and reporting higher likelihood to attend on the Intent to Attend scale. Examining patient-reported outcomes, rates of missing data with participants still in treatment on the Questionnaire about the Process of Recovery were 26.5 % at first major assessment rising up to 59.8 % on later assessments.

CONCLUSIONS

Missing data are a substantial problem for first-episode psychosis-focused LHS. LHS designs should consider factors that may influence LHS data participation and a LHS research priority should be developing interventions to decrease missing data. LHS data analyses should also consider potential differential characteristics of individuals who are versus who are not included in LHS data sets.