Advanced Medical Treatment Outcome Predictive Model

Create a PostgreSQL analytical query that builds a predictive model for treatment outcomes using multiple patient cohorts. Develop a complex query that integrates patient demographics, treatment history, genetic markers, and longitudinal health records. Implement window functions to calculate rolling health metrics, use machine learning extensions like MADlib to perform predictive analysis, and generate confidence intervals for treatment success probabilities. Include error handling for incomplete datasets and demonstrate how to export results for visualization in Excel or Google Sheets.