Research map / Learning theory and algorithms / Prediction and decision losses
Unified Screening for Multiple Diseases
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Casts unified multi-disease screening as a referral (activation) problem solved via a linear program and Lagrangian duality, characterizing risk-dependent optimal decision boundaries and comparing against independent per-disease screening in simulations.
Problem
Designing screening programs for multiple diseases simultaneously under competing risks, limited budgets and diagnostic error constraints, where per-disease programs ignore interactions between diseases.
Approach
Formulate unified screening as a referral LP choosing among four joint activation actions per risk vector, use Lagrange duality to show the optimal policy is a threshold on κ-gaps, and derive curved, risk-dependent activation boundaries for two diseases, contrasted with constant thresholds for independent screening.
Key results
- Unified model mean survival time 37.70 vs 37.47 for independent screening across the risk grid.
- Optimal activation threshold for one disease is a curved function of the other disease's risk, matching numerical LP solutions.
- Joint screenings increased to 2363 from 1847 under unified referral.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use unified referral decisions that account for both disease risks instead of independent disease-specific thresholds | Two-disease screening · In-silico two-disease screening | mean survival time (years): 37.47 -> 37.70 | Helped |
| Use unified referral decisions that account for both disease risks instead of independent disease-specific thresholds | Two-disease screening · In-silico two-disease screening | number of joint screenings: 1847 -> 2363 | Helped |
| Use unified referral decisions instead of independent disease-specific thresholds | Two-disease screening · In-silico two-disease screening | survival time | Mixed |
Surprising findings
- Independent screening performs better than unified screening at moderate, similar risk levels for both diseases.
Limitations
- Theoretical characterization only for two diseases and uniform risk distribution.
- Some structural assumptions (monotonicity, unique intersection) are only empirically supported.
- Model assumes screening is beneficial; overdiagnosis and harms are not modeled.
Experimental setup
Datasets: synthetic simulated patient data
Benchmarks: in-silico simulations
Metrics: expected survival time (years over 50) screening cost budget joint screening counts
Baselines: independent screening programs
Builds on: sequential hypothesis testing competing risks analysis
Closest papers by meaning
- Domain constraints improve risk prediction when outcome data is missing (ICLR 2024 · similarity 0.81)
- ScreenShot: A Foundation Model for Few-Shot Combination Drug Screening (NeurIPS 2026 · similarity 0.80)
- Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agents (ICML 2025 · similarity 0.79)
- Deep Reinforcement Learning for Early Diagnosis of Lung Cancer (AAAI 2024 · similarity 0.79)
- Safe Exploration in Dose Finding Clinical Trials with Heterogeneous Participants (ICML 2024 · similarity 0.78)
- Data-driven Design of Randomized Control Trials with Guaranteed Treatment Effects (ICML 2025 · similarity 0.78)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.