Trajectory-based models trained on longitudinal claims data surface the patients, markets, and risks your conventional analytics won't see for another 6 to 18 months. One platform. Three commercial functions. Continuously updated.
Request a DemoSHAARPEC Foresight identifies physicians whose patient panels are already moving toward treatment eligibility — before any script is written. Updated as new claims arrive, deployable directly into field force planning tools.
Prescribers ranked by predicted future patient eligibility in their panels, refreshed with each data cycle and deployable to field force planning tools.
Demand mapped by geography, health system, and site of care — where volume is going, not where it's been.
Infusion centers and specialty clinics ranked by concentration of pre-treatment patients — critical for site-administered therapies.
Three layers of demand sit below the waterline of prescription analytics: patients approaching eligibility, diagnosed but untreated, and approaching competitive switch. This is where Foresight operates.
SHAARPEC Foresight builds a pre-launch demand map from patient trajectory signals in the current population — using analogous therapy data as the model foundation. Available from launch day, updated continuously as new claims arrive.
A demand model built before your drug has a single prescription, trained on trajectories from the most analogous approved therapy.
Patient demand concentration by geography, health system, and site of care — available on launch day, not six months after first NRx.
Signal density mapped by territory and prescriber segment so field force reaches the right intersections first, before prescription data shows them.
SHAARPEC Foresight generates continuously updating real-world evidence from the same trajectory infrastructure that drives commercial intelligence — refreshed with each data cycle rather than frozen at a study completion date.
Living healthcare resource utilization updated each data period, designed around AMCP Format 5.0 and 2025 AMCP RWE Standards.
Analysis that extends as competitors launch — answering the formulary question being asked today, not the one from 18 months ago.
Evidence accumulates over the treatment period on a fixed CMS statutory clock — not commissioned after selection.
Trajectory intelligence validated in randomized trials, population-scale registries, and Nature portfolio journals.
Longitudinal analysis of 5,987 heart failure patients linking claims data to real-world HCRU outcomes. Demonstrates that trajectory-informed care — connecting medication adherence to utilization patterns over time — translates to measurable cost differences at population scale.
€1,130 lower cost per patient associated with guideline-adherent care (P<0.001, 95% CI 574–1,687)
Peer-reviewed validation of the synthetic patient trajectory generation methodology underlying SHAARPEC's model training infrastructure. Variational graph autoencoders generate realistic longitudinal EHR sequences while preserving statistical properties and patient privacy.
Synthetic trajectories validated to match real longitudinal EHR distributions — enabling privacy-safe model training and data sharing across organizations
Comparative ML study evaluating explainable (CatBoost + SHAP) against deep learning (LSTM) models for 100-day HF readmission prediction across 15,612 admissions. Validates that explainable trajectory models match deep learning performance — a key finding for clinical deployment and regulatory acceptance.
CatBoost AUPRC 68% vs LSTM 66% — explainable model matches deep learning; both substantially outperform LACE index (AUPRC 51%)
Prospective, randomized study design co-authored by SHAARPEC (Hallandia V), establishing the methodology for evaluating trajectory-based risk enrichment in a regional population of 330,000. SHAARPEC is named as an organizational contributor to the trial design and risk prediction model.
SHAARPEC (Hallandia V) listed as organizational co-author — prospective validation of trajectory-based risk stratification methodology
Randomized controlled trial testing trajectory-based ML risk stratification for AF screening across 2,960 patients. ML-enriched identification of high-risk individuals, combined with 14-day continuous ECG monitoring, detected substantially more new AF cases than standard care across a catchment population of 330,000.
5.6× improvement in AF detection (RR 5.6, 95% CI 2.2–14.4) — ML+invitation vs. standard care; number needed to invite: 32
Transformer-based model predicting adult ADHD diagnosis from routine EHR clinical codes at 6, 12, and 18 months before clinical identification. Demonstrates trajectory prediction generalizability and consistent performance across indications beyond cardiology.
AUC 0.79 (95% CI 0.76–0.81) at 6-month horizon; stable at 12 months (0.76) and 18 months (0.75)
We start with your data question and show you what the signal looks like for your therapy.
Request a Demo