PROJECT SUMMARY Light chain (AL) amyloidosis is a recalcitrant and deadly hematologic disease characterized by organ dysfunction from insoluble fibril deposition derived from clonal free light chains arising from a monoclonal gammopathy. The disease has a high early mortality of 40-45% at two years due to heart failure. Patients with advanced AL amyloidosis have high morbidity and mortality in the initial period after diagnosis owing to cardiac dysfunction. Despite experiencing multiple symptoms and demonstrating signs of the disease, many patients are diagnosed late, sometimes by years, because these ‘precursor diagnoses’ are often non-specific. Observational data also suggest that Black individuals are more likely to be underdiagnosed with cardiac amyloidosis. Monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (MGUS+) are more common in Black individuals, as is the prevalence of hypertrophic cardiomyopathy and chronic kidney disease, both of which also occur in AL amyloidosis. We hypothesize that patients can be diagnosed early by assessing the patterns of precursor diagnoses that predate AL amyloidosis diagnosis. Our application seeks to create an algorithm using Bayesian machine learning statistical methodology to create an alert system that can help guide physicians toward early consideration of an AL amyloidosis diagnosis. We will execute the following specific aims using nationally representative Medicare data: 1) Identify patterns of precursor diagnoses associated with the occurrence of AL amyloidosis and develop a predictive algorithm using Bayesian machine learning techniques in Medicare beneficiaries with MGUS+. Patterns will be examined longitudinally at one-year timepoints over a five-year period preceding the AL amyloidosis diagnosis contrasting between MGUS+ with known AL and MGUS+ with no known AL to identify patterns that might best predict disease. 2) Study the performance of the predictive MGUS-AL algorithm. This will be assessed internally in the Medicare data set, overall and by race groups (Aim 2A) and external validation using TriNetX multicenter EHR data for MGUS+ patients of all ages, races, and insurance coverage (Aim 2B). 3) Estimate the number of potentially undiagnosed AL amyloidosis patients with MGUS+. Based on the patterns identified in Aim 1 and validated in Aim 2, we will identify subjects at high risk for undiagnosed AL amyloidosis (Aim 3A) and estimate the excess 2-year mortality and number of potential lives saved by our early warning system, overall and by racial group (Aim 3B). This study provides an unprecedented opportunity to identify patterns of precursor diagnoses to diagnose AL amyloidosis early. An important anticipated outcome is to improve health disparities by increasing AL amyloidosis diagnosis in Black individuals who are already at higher risk for MGUS and other end-organ damage associated with AL amyloidosis. The novel, rigorous and easy-to-implement early...