Skip to content

Predictive Readmission Modelling with Fairness Auditing in Medicaid Hospitals

Algorithmic readmission risk stratification in safety-net hospital environments exhibits severe vulnerabilities to historical documentation bias, class imbalance, and structural policy shifts. Standard mathematical fairness interventions often degrade clinical discriminative capacity and redistribute critical diagnostic errors across vulnerable patient groups rather than resolving underlying disparities. Establishing domain-aware sensitivity auditing and calibrated post-processing strategies provides a robust pathway to achieve equitable resource allocation without compromising patient safety.

Goal of work

Examine the structural drivers of algorithmic bias in Medicaid readmission models to establish clinical fairness auditing standards.

Methodology

Comparative secondary synthesis and sensitivity analysis of clinical fairness frameworks across safety-net hospital predictive models.

Scientific novelty

Integrates domain-aware clinical utilities with global sensitivity auditing to resolve error redistribution traps in safety-net readmission modeling.

Document Preview

Review the formatting and introduction. The full version will refine the structure for the selected document standard.

PhD Dissertation

Degree:
Predictive Readmission Modelling with Fairness Auditing in Medicaid Hospitals

Author:

Group

First M. Last

Advisor:

Dr. First Last

City, 2026

Contents

Introduction
Chapter 1. Conceptual Frameworks of Algorithmic Equity in Public Healthcare Systems
1.1 Structural Disparities and Safety-Net Hospital Demographics
1.2 Predictive Readmission Architectures and Public Payer Dynamics
1.3 Definitions of Algorithmic Harm and Sociotechnical Alignment
1.4 Regulatory Policies and Value-Based Care Penalties
Chapter 2. Methodological Standards for Auditing Clinical Prediction Models
2.1 Comparative Analysis of Group Fairness Criteria in Medicine
2.2 Sensitivity Formulations and Longitudinal Measurement Shifts
2.3 Evaluating Post-Processing and In-Processing Mitigation Strategies
2.4 Domain-Aware Evaluation Protocols for Highly Imbalanced Outcomes
Chapter 3. Diagnostic Assessment of Readmission Risk Inequities
3.1 Feature Sensitivity and Systemic Bias Propagation Pathways
3.2 Subgroup Error Redistribution and the Illusion of Statistical Parity
3.3 Policy Shock Transmission in Public Reporting Systems
3.4 Intersectionality and Payer-Level Stratification Vulnerabilities
Chapter 4. Comparative Evaluation of Bias Mitigation Frameworks
4.1 Threshold Calibration Across Vulnerable Demographic Strata
4.2 Deep Reinforcement Learning for Multicentric Bias Regularization
4.3 Trade-Offs Between Global Discrimination and Clinical Safety
4.4 Reject Option Classification in Resource-Constrained Environments
Chapter 5. Institutional Implementation and Governance Strategies
5.1 Operationalizing Low-Resource Bias Mitigation Playbooks
5.2 Algorithmic Auditing Integration in Electronic Health Records
5.3 Continuous Monitoring and Post-Deployment Risk Controls
5.4 Ethical Accountability and Value-Driven Healthcare Policy
Chapter 6. Theoretical Framework
Conclusion
Bibliography

Introduction

Predictive readmission modeling serves as a critical mechanism for resource allocation, care coordination, and financial penalty mitigation within safety-net hospitals serving low-income populations. Modern hospital decision support systems increasingly leverage statistical learning algorithms to identify high-risk beneficiaries, yet structural inequities encoded within clinical documentation frequently propagate through automated scoring pipelines [1]. When applied indiscriminately in safety-net settings, standard risk stratification models risk systematically misestimating clinical vulnerability, reinforcing institutional disparities and distorting resource distribution across historically marginalized cohorts [7].

Existing algorithmic fairness paradigms frequently focus on generic mathematical parity metrics that fail to reflect complex clinical realities and demographic heterogeneity. Imposing unconstrained statistical parity constraints on highly imbalanced clinical data can paradoxically destabilize decision boundaries, elevate false negative rates in vulnerable sub-populations, and degrade overall clinical utility [4]. Furthermore, external policy changes and structural reporting transitions induce measurement shocks that undermine model calibration over time, making superficial fairness fixes insufficient for sustained equity [6].

This dissertation investigates the structural mechanisms of algorithmic disparity in hospital readmission models and establishes a multidimensional auditing and mitigation framework tailored to safety-net healthcare systems. By synthesizing global sensitivity analysis, clinical risk control, and domain-aware threshold calibration, the research formulates rigorous diagnostic criteria for public payer datasets [6], [7]. The resulting framework bridges the gap between theoretical machine learning fairness and practical clinical governance in high-stakes Medicaid environments.

Methodological Standards for Auditing Clinical Prediction Models

Evaluating clinical machine learning systems within safety-net environments requires auditing methodologies that look beyond aggregate demographic parity. Traditional mathematical fairness interventions frequently assume that equalizing error rates across demographic cohorts resolves algorithmic harm. However, secondary evaluations of clinical risk prediction reveal that forcing strict parity constraints under severe class imbalance can destabilize decision thresholds and inadvertently elevate false negative classifications among high-vulnerability populations [4]. When models fail to detect impending readmissions among marginalized patients, the resulting misclassification denies critical transitional support to those most dependent on targeted safety-net interventions [7]. To overcome these limitations, methodological auditing must synthesize global sensitivity analysis with clinical utility boundaries. Structural diagnostic frameworks demonstrate that algorithmic bias propagates through correlated feature families, where temporal policy transitions and measurement shocks dramatically alter feature importance across operating eras [6]. Auditing protocols in public payer settings must therefore evaluate model robustness along multiple dimensions, including structural asymmetry, temporal stability, and clinical cost sensitivity [6]. Rather than relying on generic algorithmic toolkits, safety-net informatics systems require domain-aware evaluation standards that balance statistical parity with the imperative to preserve overall predictive discrimination and patient safety [4], [7].

References

  1. Fairness and Bias Mitigation in Student Success Prediction Models
    Godfrey Perfectson Oise
    DOI Link
  2. Fairness as a Prudential Risk: Integrating Algorithmic Bias Constraints into Solvency Capital Modeling
    Samuel Foster
    DOI Link
  3. Bias Mitigation with Fairness Guarantees Using Risk Control
    Alberto García-Galindo, Marcos López-De-Castro, Rubén Armañanzas
    DOI Link
  4. The Illusion of Fairness: When Bias Mitigation Harms Clinical Performance in Imbalanced Healthcare Prediction
    Pooya Zeinali, Negar Zeinali
  5. Algorithmic Fairness and Bias Mitigation for Clinical Machine Learning: A New Utility for Deep Reinforcement Learning
    Jenny Yang, Andrew A. S. Soltan, David A. Clifton
  6. Beyond Outcome Disparities: A Sensitivity Framework for Auditing Algorithmic Fairness
    Samuele Lo Piano
  7. Identifying and mitigating algorithmic bias in the safety net
    Shaina Mackin, Vincent J. Major, Rumi Chunara et al.
  8. clinicalfair: Algorithmic Fairness Assessment for Clinical Prediction Models
    Cuiwei Gao

Bibliography

Verified SourcesFormatting StandardsHigh UniquenessPro Models
Launch Offer -25%

Dissertation

APA 7th Edition (Publication Manual)

$26$34
  • 120+ pages
  • High originality drafting
  • Export to Word
  • Correct formatting
  • Public Preview
    A preview by another author cannot be made private. Your work will be private and completely unique.
  • Bibliography (150+, APA 7th Edition)
    +$1
  • Add alternative sources (News, .gov, .edu)

Dissertation

APA 7th Edition (Publication Manual)