The lack of reasoning about severity and chronicity of diseases is a problem that caused many of the diagnostic failures. In addition, there are many unreasonable alternate hypotheses that should be eliminated for reasons of severity and chronicity. The program should not use an acute cause to account for a chronic effect or (in most situations) a mild cause to account for a severe effect. For example, an acute MI can cause pulmonary congestion and pulmonary congestion in general can produce findings such as nocturnal dyspnea (PND) over days, but the pulmonary congestion caused by an acute MI a few hours previous has not been present long enough to account for PND. Acute and chronic versions of the same pathophysiologic state can have different findings. As mentioned above, the acute pulmonary congestion in case 128 had only tachypnea and none of the X-ray or chest findings that come usually within a few hours. In cases 125 and 213, the opposite problem existed. When high LAP has existed for years, the lungs adapt and there are again few findings indicating pulmonary congestion, but the program took this lack of findings as evidence against high LAP.
The severity of a disease also affects how it presents, with milder forms having fewer findings. In cases 79, 133, and 214 there was known congestive cardiomyopathy and mild pedal edema with no other findings consistent with right heart failure. Since the causal pathway from the disease to pedal edema goes through several nodes, each of which may have negative evidence, the program estimates the probability to be higher if pedal edema is considered as having an external cause or was a therapy side effect than to include the additional nodes in the hypothesis. In reality, if the right heart failure is mild, the abnormal state of the intermediate nodes may be undetectable, so the lack of evidence should not count against the hypothesis. Whenever there is a physiological state that represents the cumulative effect of a dynamic process, there may be a difference between the process currently and the state. This makes the appropriate handling of chronicity and severity a necessity.
One approach we used to avoid the problem of severity is to make some of the causal probabilities between nodes conditional on other nodes being in the hypothesis. For example, only chronically very high pulmonary artery pressure can cause pulmonic regurgitation. That limits the ultimate causes to primary pulmonary hypertension or mitral stenosis, two states that produce very high pulmonary artery pressure over years. Making the immediate link conditional can interfere with the generation of hypotheses (as it did in cases 121 and 209) since the probability of a pathway may change when other nodes are added to the hypothesis, increasing the risk of missing good hypotheses. It only covers up the real problem of reasoning about severity and chronicity.
The KB does not currently differentiate between different kinds of renal diseases. This has proved to be a problem in several cases, because some renal disease can cause pedal edema. When pedal edema is the only manifestation of right heart failure and there is known renal disease, the program often hypothesizes that the pedal edema is due to the renal disease rather than hypothesizing the right heart failure mechanism, which may have considerable negative evidence along the causal chain. The problem is that only nephrotic syndrome or renal disease severe enough to cause oliguria produce pedal edema, and these constitute less than 10%of renal disease. In case 141 there was proteinuria, which is characteristic of nephrotic syndrome, but not knowing the severity of the proteinuria it is still more likely that the pedal edema is caused by right heart failure than by renal disease.