The Predictive Suite
Predictive Outcomes Suite
This is a modular probability of success methodology. The underlying predictive pipelines are agnostic—you can integrate or deploy any of these phases independently. Build custom empirical models from your historical data and calculate real-time success probabilities for your active clinical cohort without your data ever leaving the clinic.
ACTIVE TOOL
Phase 1: Predictive Blueprint
Upload historical patient data to calculate the coefficients needed to predict success, session-by-session, tailored precisely to your population.
ACTIVE TOOL
Phase 2: Trajectory Tracker
Score your active patients. This independent tool ingests raw outcome data alongside statistical coefficients (from Phase 1 or any external model) to compute live success probabilities.
IN DEVELOPMENT
Phase 3: Live Interaction Demo
A standalone, live interactive dashboard for clinicians. Input raw item scores during a session to instantly visualize the patient’s current progress against expected success curves. This tool can securely connect to Phase 2, or it can operate independently by ingesting probabilities from any external clinical API.
Available Variable Selection Methodologies
A Priori (Logistic Regression)
Forces the model to retain only the specific predictors you define, regardless of statistical significance. Perfect for testing specific clinical hypotheses or maintaining strict theoretical compliance.
Exploratory Machine Learning
Utilizes 10-fold Cross-Validated Elastic Net Regularization to automatically prune statistically weak predictors. It empirically identifies the strongest predictors of success from massive datasets.
Hybrid ML (The Best of Both)
Allows you to mandate that certain "core" variables (like treatment attendance) are never removed, while letting the Machine Learning engine dynamically optimize and prune all secondary exploratory variables.
Core System Capabilities
Local & Secure
This builder runs entirely offline. Transform your clinic's EMR exports into predictive models locally, guaranteeing zero risk of HIPAA violations or data leaks.
In-Process Research
Certain aspects of this methodology are derived from the forthcoming peer-reviewed publication:
Coleman, J. J., Miller, S. D., Del Re, A. C., & Wampold, B. E. (in press). Improving Feedback in Measurement Based Care: The Success Probability Index.
Handling Messy EMR Data
Generic predictive models fail when confronted with the noise and sparsity of clinical EMR data. This system relies on transparent, defensible mathematics to isolate systematic treatment effects.
Predictive Modeling
Calculates the underlying trends needed to predict a patient's likelihood of success in real-time.
Session-by-Session Focus
We don't force patients into a single global model. The system evaluates their progress independently at every single session.
Dashboard Integration
The models you export here are designed to integrate directly into active clinical dashboards to support your care team.
Custom Treatment Success
Flexible mapping allows you to define clinical success exactly how you want it, with multiple customizable pathways and definitions.
Custom Predictors & Outcomes
Full control over your model configuration. You can customize exactly which psychological outcomes and process predictors you want to include.
Flexible Temporal Inputs
Simply map in your raw session numbers or specific calendar dates, and the system will automatically process and standardize the timeline for you.
Automated Data Quality Checks
If there are weird anomalies in your data, the system instantly catches them and gives you precise feedback on why things aren't running.
EMR Data Transformation
Seamlessly ingests your clinic's raw CSV exports and automatically structures the data for advanced statistical analysis.
Secure Local Processing (No Data Leaves Your Clinic)
All data processing and model training happens entirely on your local machine. Your Protected Health Information never leaves your clinic.
Trend Nuance Detection
Captures sudden leaps, stalls, and therapeutic swings that standard linear reporting tools completely miss.
Ready-to-Deploy Exports
Instantly compiles your newly trained models into a clean, downloadable .zip file specifically calibrated for your engineers.
Comprehensive Documentation
Automatically generates polished, easy-to-read PDF guides for both your developers and program managers alongside your models.
Calculate session-by-session success probabilities for your active clinical cohort.
Phase 2 applies statistical coefficients to your current, active patients. You can upload the model parameters generated from Phase 1, or ingest your own custom models.
This application securely processes your raw clinical data locally in your browser, generating session-by-session probabilities of success for each individual patient. This provides your clinical team with actionable insight to identify who is on target and whose trajectory requires intervention.
Model Ingestion
Supports JSON model ingestion directly from Phase 1 (or any custom statistical model), combined with robust, flexible mapping for any clinical variable schema.
Local Processing
No PHI is transferred externally. All predictions are generated locally using statistical processing.
Interactive Dashboard in Development
This module is currently being finalized. Once deployed, it will allow clinicians to input raw item scores during a live session and instantly visualize the patient's current progress against expected success curves.
Request Feasibility Review
Our team will review your structural data requirements and confirm the correct tier within 24 hours.
Contact Request
If your default mail client did not open, you can copy my direct email details below: