The cases used for testing and refining the program were gathered from discharge summaries at The New England Medical Center Hospital of patients hospitalized in 1988 and early 1989. These were all patients for whom discharge summaries were available, in Diagnosis Related Groups (DRG) 121, 123, 127, and 135. These DRGs include patients with acute myocardial infarctions (MIs) with cardiovascular complications (both discharged alive and expired), heart failure and shock, and valvular disorders with complications. Since these are general categories with many specific diseases included and additional diseases present as complications, these DRGs contained all of the available types of cases that are relevant to the domain of the Heart Failure Program. There are other DRGs that the program could handle, such as uncomplicated MIs, but most of these cases are relatively simple. The DRG for complicated MIs included enough variety to test the program more thoroughly for acute ischemia cases than for most other diseases. By choosing these DRGs the case set was assured of including the most complicated cases, providing the greatest challenge for the program.
The discharge summaries, usually two to three page long accounts of the patient admission, include the history of the present illness, medications, physical exam, and laboratory data on admission, short description of the hospital course, final diagnosis, and discharge information. These summaries are dictated by the house officers from the medical record after the patient has been discharged. Because these are usually highly condensed, they do not provide all of the data available in the patient record. However, they proved to be useful summaries of the patient data and not atypical of a case description that might be given by a house officer.
A total of 246 cases were collected. Of these four were eliminated because of insufficient data. We were very liberal in accepting cases, assuming that if the cardiologists examining the information could come to some reasonable conclusion from the data, the program should also be able to. Thus we included cases from which a considerable amount of data was missing. The average age of the patients was 67.5, including 8 from 24 to 40 years old with 120 females and 122 males.
The New England Medical Center Hospital is a tertiary care, teaching
hospital, so the patient population includes a large number of complex
cases. Of the 30 disease entities included in the model knowledge base,
19 are definitely true of patients in the sample according to the expert
diagnoses and three more are possibly true. These 22 diseases are
listed in table 1. The columns are the number of cases in which the
disease was known to be present prior to admission, the number in which
it was a definite part of the expert's diagnosis but not previously
known, and the number in which it was mentioned as a possibility in the
diagnosis.
The number of patients labeled as congestive cardiomyopathy is high because we considered this to be a physiologic diagnosis including a number of etiologies, such as ischemic, hypertensive, and other causes of a dilated heart. The high number of patients labeled as renal insufficiency results from a liberal definition of renal insufficiency on the discharge summaries and does not mean that the renal function tests were abnormal at the time of admission. The program distinguishes between acute and chronic variants of mitral regurgitation and renal insufficiency, but the expert diagnoses often did not, so they are included together here.
These cases were primarily ones in which multiple diseases were present. There were an average of 1.7 diseases known prior to admission and an additional average of 1.5 diseases definitely present as part of the expert diagnosis. There were only 5 cases in which the experts could not identify any diseases as definitely true. (If we had included more simple cases, there would have been many since all acute MIs or new unstable angina with no other diseases would have been in this category.) There were 6 cases in which there were 6 or more known or definite diseases.