Dave Childers

Simulated analysis evaluating whether post-discharge care coordination reduces 30-day hospital readmission rates in a value-based care setting. Uses synthetic patient- and practice-level data with a staggered adoption design and a generalized linear mixed model to estimate the effect of care management programs.

End-to-end workflow transforming synthetic EHR records (Synthea) into an analysis-ready dataset, then applying survival analysis to estimate time from lung cancer diagnosis to death. Covers data ingestion, parquet-based storage, and patient-level time-to-event modeling.

Heap Illuminate is a data science layer within Heap's digital analytics platform that automatically surfaces user friction points and behavioral insights for product managers and designers. Contributed statistical modeling for core Illuminate product features during tenure as Senior Data Scientist at Heap.

Authored Duo Security's 2021 State of the Auth Report, a survey-based study of user experiences and perceptions of multi-factor authentication. Research was conducted via Qualtrics with analysis in SQL and R, synthesizing findings into a published industry report on MFA adoption trends and attitudes.

Published in the Journal of Adolescent Health (2014). Used interrupted time series with ARMA errors to evaluate the causal effect of Michigan's 2010 universal text messaging restriction on motor vehicle crash rates across driver age groups and crash severities. Found that the restriction was not associated with reductions in severe crashes, with small increases observed in the most severe crash types. Lead analyst and co-author.

Published in JAMA Ophthalmology. Longitudinal cohort study using medical claims data from a large nationwide US managed care network to identify risk factors for thyroid-associated ophthalmopathy (TAO) among newly diagnosed Graves disease patients. Applied multivariable Cox regression to estimate hazard ratios across treatments and comorbidities, finding that surgical thyroidectomy and statin use were each associated with substantially reduced hazard for developing TAO. Lead analyst.

Co-inventor on a US patent (2017) for a method of disaggregating household energy usage signals into device-level consumption estimates. Developed during tenure as a data scientist at Powerley, a residential smart-grid startup. The method combines a gateway device capturing aggregate energy signals with user inputs to identify individual appliance-level usage.