Techniques are provided for use by a pacemaker or other implantable medical device for detecting and tracking trends in cardiopulmonary fluid transfer rates—such as heart-to-lung fluid perfusion rates and lung-to-lymphatic system fluid excretion rates—and for detecting heart failure, dyspnea or other cardiopulmonary conditions. In one example, the device periodically measures transthoracic admittance values. A first exponential time-constant (k1) is determined using curve-fitting from admittance values obtained while the patient is in a sleep posture. Time-constant k1 is representative of the fluid perfusion rate. A second exponential time-constant (k2) is determined based on admittance values obtained while the patient is standing/walking/sitting. The second exponential time-constant (k2) is representative of the fluid excretion rate from the lungs. The device then detects trends, if any, in the time-constants (or in “DC” baseline values) to detect or predict medical conditions such as an imminent heart failure exacerbation.