Causality in the Heart Domain

There are a small number of patterns of causation that the program needs to capture. These can be characterized as follows:

Immediate

This takes place immediately if it happens at all. For example, most valvular lesions cause a murmur and if so, it is there when the lesion is present.

Event-like

This happens as a result of pathophysiological states that predispose the patient to the entity. For example, a patient with coronary artery disease can have an MI any time with no particular time relationship. Perhaps some additional stress actually triggers the MI, but in practical terms the immediate cause is rarely known and not important.

Delayed

This is like immediate causation except that there is some definable time interval between cause and effect. For example, a patient with an MI may develop pericarditis within the first two weeks after the MI, usually more than two days after the event.

Progressive

This is causation that, once it takes place, continues. Progressive states persist chronically and just get worse unless radical correction takes place. For example, rheumatic heart disease often causes mitral stenosis (restriction). The mitral stenosis will gradually get worse until valve repair or replacement is done.

Accumulative

This requires the cause to exist over a period of time. For example, the fluid accumulation that produces pedal edema (swelling of the feet) is caused (indirectly) by low cardiac output. The cardiac output does not have to be consistently low, but if it is low on average for long enough there will be enough water retention to produce the pedal edema.

Intermittent

When a cause is present some effects and findings occur intermittently. For example, paroxysmal atrial fibrillation is a rhythm disturbance that may persist for seconds to days and may happen again at random intervals. Often it gradually increases in frequency and eventually becomes chronic atrial fibrillation.

Corrective

Some states return other states to normal. These are usually therapies, but in some cases they are other pathophysiologic states that have influences in opposition to the effect, such as dehydration `correcting' high blood volume.

Besides these characteristics of causation, there are some characteristics of the states that have implications for temporal reasoning. States produced by accumulative causation and sometimes by delayed causation tend to remain after the cause has ended since the same mechanism tends to delay the return to a normal state. The progressive and accumulative descriptions could also be viewed characteristics of the state, since once they start, their continuation is independent of considerations of what caused them. Some states have a maximum time period. For example, stress of many kinds including an MI cause an increased sympathetic nervous state with sweating, rapid heart rate, and so forth, but this only lasts for a few days at most, even though the cause may continue. Other states only last for a period of time, not because they are self-limiting but because they are either corrected or the patient does not survive. These characteristics constrain the time bounds of causes and effects and what conclusions can be drawn from findings.

It should be noted that these characterizations are not mutually exclusive. Delayed states can be progressive or intermittent and so forth. Probably the most complicated causal pattern in the domain is that of an MI. An MI is usually caused by coronary artery disease but the time is random, without any triggering event in the majority of cases, and having an acute onset. Thus, it is the archetype for event-like causation. Over the first few hours, the MI causes chest pain, signs of acute ventricular dysfunction, and signs of sympathetic nervous response. Over the first two days, the enzyme changes are evident. Over the first two weeks the electrocardiographic patterns change into ones characterized as evolving and the patient may experience pericarditis. Often the damage to the heart is permanent, producing chronic ventricular dysfunction and the electrocardiographic pattern of an old MI.

This temporal characterization of the heart domain may not be adequate for other medical domains, but it will probably be nearly so. One obvious temporal pattern that has been left out is cyclic phenomena. Those do happen in this domain. For example, Cheyne-Stokes respiration is an oscillation between slow and rapid breathing that sometimes happens with older patients in heart failure. However, all such phenomena in this domain are recognized as entities in themselves and are treated in the same way as other findings. That is, the cyclic nature of the phenomena is not something the program needs to reason about.