About Total Life
Total Life is a behavioral health telehealth practice serving older adults, with 200+ licensed clinicians across 49 states. Most of our patients reach us through Medicare and commercial insurance, and a large share are aging adults who fall outside traditional VA and specialty mental-health infrastructure. Because we deliver care remotely and repeatedly over time, we hold something unusual: dense, longitudinal behavioral-health measurement on an older-adult population that is rarely captured at this scale or cadence.

The research
We are building a research program around a simple but underexplored idea: most suicide-risk detection treats risk as a fixed state, measured once. We want to test whether the dynamics already present in routine care — how symptom scores move over time, and how patients engage with care as those symptoms change — carry information about near-term risk that standard one-time screening misses. The initial project is a retrospective study on fully de-identified data (no new data collection, no patient contact), developing and validating a multimodal risk-detection model and quantifying its incremental value over standard screening. This work is being submitted for federal research funding, and the person in this role would be a named co-investigator on that application.

The role
You would lead the statistical thinking behind this work — design and analysis, not just execution. Specifically:

• Serve as co-investigator on a federally funded research application (biosketch and defined effort commitment required)
• Design the statistical and validation plan: power / sample-size analysis for a rare-event outcome, model evaluation, discrimination and calibration, temporal validation
• Advise on analytic methods for de-identified behavioral-health and claims data, including de-identification standards
• Partner with our clinical and engineering teams to guide model development and interpret results responsibly

What we're looking for

• PhD (or equivalent) in biostatistics, statistics, or a closely related field
• Demonstrated experience building risk-prediction models, ideally with EHR, claims, or longitudinal healthcare data
• Working knowledge of rare-event / class-imbalance methods and prediction-model reporting standards (e.g., TRIPOD)
• Comfort with de-identified clinical data and methods such as HIPAA Expert Determination
• US-based and able to be named on a US federal (DoD/NIH) research application
• Prior experience on federally funded research (NIH/DoD) is a strong plus

Engagement
Part-time / contract, remote. Compensation commensurate with experience and scope. Immediate start preferred given active research timelines, with potential to grow alongside our research program.