LLM tools for dirty death data

I know there are quite a few LLM tools being showcased in our community. However, I wonder if any are specific to death data which sometimes has more than one cause of death as its source value.

A few examples:

  1. Heart failure, ventricular fibrillation, cardiac arrest
  2. Cardiac arrest, tobacco use disorder, coronary artery disease
  3. DKA, cardiac arrhythmia

As you can see from the examples, this isn’t a straight mapping. If there is >1 cause, a decision on which one to map would need to be made. For examples 1 and 2, I would pick cardiac arrest as the cause of death. The other conditions listed would be contributing factors. For example #3, diabetic ketoacidosis can contribute to your death. And cardiac arrhythmia can contribute or cause death, but people can live long lives with cardiac arrhythmias.

Is there a tool which ranks the conditions within a source_value field and then maps the condition to a standard concept_id? If yes, how does the ranking logic work?