Lower levels of free T3 (FT3) occur during acute illness, as part of "euthyroid sick syndrome." A chronic form of this syndrome also exists. To determine whether thyroid hormone levels predict long-term survival. This was a "big-data" study of thyroid function tests from samples taken between 2008 and 2014. Data were crossed with electronic health records for morbidity and mortality. Test results were converted to age- and gender-adjusted percentiles (AGAPs). The hazard ratio for death was crossed with ranges of initial AGAPs and change in AGAPs for 2 subgroups: "not healthy" (subjects with at least 1 of 5 chronic conditions registered in their electronic health chart) and "healthy" (all others). 2 453 091 sets of thyroid function tests from 365 965 distinct patients were evaluated. 258 695 sets remained after excluding patients registered as taking thyroid preparations or anti-thyroid drugs. Hazard ratio for death, planned before data collection. The cohort included 151 868 not healthy and 106 827 healthy people. After a median of 6.8 years, 5865/151 868 (10.4%) of the not healthy had died and 2504/106 827 (2.3%) of healthy participants. Low initial FT3 AGAPs were predictive of poor survival. The hazard ratio for survival compared between the lowest 5 and highest 50 percentiles of initial FT3 AGAPs for not healthy participants was 5.71 [confidence interval (CI) 5.23-6.26, P < .001] and for healthy was 3.92 (CI 3.06-5.02, P < .001). Low FT3 AGAPs predicted poor survival, most strongly among not healthy people.