Sleep Temporal Entropy (STE), a novel metric analyzing sleep stage fragmentation patterns, demonstrated superior predictive power for cardiometabolic disease and mortality compared to conventional sleep measures across 8,081 adults in two independent cohorts. The metric revealed a surprising U-shaped mortality relationship, where both extremely consolidated and highly fragmented sleep patterns increased death risk. Most striking was the finding for REM sleep entropy: individuals in the lowest quintile faced 58% higher all-cause mortality risk and nearly triple the cardiovascular death risk (HR 2.83) compared to moderate fragmentation levels. This challenges the common assumption that more consolidated sleep is always better, suggesting optimal sleep health exists in a moderate fragmentation range. The entropy-based approach captures stage-specific disruptions that traditional wake-episode counting misses, potentially explaining why some sleep studies show conflicting results. As a preprint awaiting peer review, these findings require validation, but the large sample size and consistent patterns across cohorts are encouraging. The metric's scalability from standard sleep study data could transform sleep medicine by providing more precise risk stratification tools, moving beyond simple sleep duration toward sophisticated architectural analysis.