AIMS: To examine the associations between diastolic dysfunction indices and long-term risk of all-cause mortality in adults over 23-year follow-up. METHODS AND RESULTS: Participants (n = 2734) of the population-based Tromsø Study of Norway had echocardiography in 1994-1995. Of these 67% were repeated in 2001 and/or 2007-2008. Mortality between 1994 and 2016 was determined by linkage to the national death registry. Cox regression was used to model the hazard of all-cause mortality in relation to left atrial parameters (treated as time-dependent using repeated measurements) adjusted for traditional risk factors and cardiovascular disease. During the follow-up, 1399 participants died. Indexed left atrial diameter, mitral peak E deceleration time, and mitral peak E to peak A ratio showed an U-shaped association with all-cause mortality. Combining left atrial diameter with mitral peak E deceleration time increased the prognostic accuracy for all-cause mortality whereas adding mitral peak E to peak A ratio did not increase prognostic value. We estimated new optimal cutoff values of left atrial diameter, mitral peak E deceleration time, and mitral peak E to peak A ratio for all-cause mortality outcome. E/e' had a cubic relation to mortality. CONCLUSION: Both enlarged and small left atrial diameters were associated with increased all-cause mortality risk. A combination of Doppler-based left ventricle filling parameters had an incremental effect on all-cause mortality risk. The cutoff values of diastolic dysfunction indices we determined had similar all-cause mortality prediction ability as those recommended by American Association of Echocardiography and European Association of Cardiovascular Imaging.