Conventional echocardiographic characterization of diastolic function requires manual analysis of Doppler E-and A-wave amplitudes, deceleration times, isovolumic relaxation times, and pulmonary venous flow patterns. Mathematic modeling of the suction pump activity of the heart permits characterization of diastolic function through model-based image processing, which relies solely on transmitral Doppler images. This automated method uniquely specifies the entire E-wave contour using 3 parameters (x(o), k, and c) that determine E-wave amplitude, width, and rate of decay. Moreover, the index β =c2 - 4k, reflecting the balance between chamber viscosity and stiffness/recoil, represents a novel parameter for characterizing diastolic function. We analyzed Doppler E waves from 39 patients (mean age 79 years, 61% women, mean ejection fraction 47%) using the model-based image processing technique. A value of β < -900 was selected as indicative of severe diastolic dysfunction. Of 17 subjects with β < -900, 8 (47%) were no longer alive at 1 year. Of 22 subjects with β > -900, all were alive (p = 0.001). The index β, dichotomized at < -900, had a predictive accuracy of 0.769 (30 of 39), a negative predictive value of 1.0 (22 of 22 alive), and a positive predictive value of 0.471 (8 of 17 deceased) for 1-year vital status. Of 14 subjects with deceleration time ≤160 ms, 5 (36%) were deceased at 1 year, whereas for deceleration time > 160 ms, 22 of 25 patients were alive (p = NS). Of 16 subjects with ejection fraction <45%, 6 (38%) were deceased at 1 year. Of 23 subjects with ejection fraction >45%, 21 were alive at 1 year (p = 0,074). On multivariate analysis, β dichotomized at -900 was the strongest independent predictor of 1-year mortality. We conclude that evaluation of diastolic function using model-based image processing provides valuable prognostic information in elderly patients with heart failure.