[CACDDST: An Easy-to-Use Diagnostic Decision-Support Tool on the Web]

< Introduction >

The basic job of primary physicians is primarily to diagnose and treat common diseases. However, there are sometimes cases of rare diseases that are difficult to diagnose. In order to diagnose these rare diseases without delay, it is indispensable for primary physicians to be able to recall themselves of all possible diseases including rare ones. For the purpose of assisting the physicians aiming for an early diagnosis of these diseases, we have developed a CACDDST (computer-aided clinical diagnostic decision-support tool.)

< Disease Knowledge Database (DKDB) >

Each disease presents distinct and characteristic clinical manifestations including signs, symptoms and laboratory results (CMs). We accumulated the CMs of approximately 1500 diseases.

Subsequently, we classified CMs into some grades according to their importance in acquiring an accurate diagnosis. Finally, we divided these grades using weight points, which are calculated in some unique way.

The distinguishing feature of our CACDDST is the unique database. From now on, we will refer to this unique database as the "Disease Knowledge Database (DKDB)." There are two kinds of DKDBs, namely the main DKDB and sub-DKDB. The main DKDB is the union of all Disease Units. The sub-DKDB was designed to assist the main DKDB. The CMs in the sub-DKDB contain less important data which we could not house in the main DKDB. However, the sub-DKDB is also the union of other Disease Units.

< Outline of the CACDDST >

Next, we give an outline of the CACDDST. In order to accomplish recalling/diagnostic efficiency, we adapted a unique scoring system for CMs in the main DKDB. There are two principles for point assignment: (1) Higher point values are given for CMs that are most important for diagnosing the disease. (2) Overall score of the CMs of one Disease Unit in the main DKDB and/or the average score of each CM of one Disease Unit in the main DKDB should be within a certain range. Especially, the meaning of the second principle is to guarantee fairness of the Disease Unit in the main DKDB. The sub-DKDB has no such constraint, yet we have prepared it to improve the accuracy of the CACDDST.

As of April 2006, these DKDBs hold approximately 1500 diseases and 630 CMs on file. It would be prudent at this juncture to highlight that, on the DKDBs, we have not considered factors such as the frequency and pattern of the onset of diseases, and symptomatic changes in the clinical course; gender; age; race or region. We would suggest that these factors are considered at the actual differential diagnosis by each individual clinician.

< One verification of the capability of the CACDDST >

In this part, we would like to demonstrate the validity of CACDDST and simultaneously comment on the usage of the CACDDST by using one example.

We have calculated that CACDDST is a useful clinical problem-solving tool, with an ideal and practical design, using an actual paper, "A Jaundiced Eye," which appears in "The New England Journal of Medicine(*)." Tables 1 and 2 convey each CM input for this search and displays the consequent results. (Please access this website for the details regarding Tables 1 and 2.)

Table 1. The first search and its results, CMs for the search (all information was collected during the first consultation):

  • Sore throat,
  • Fatigue ("run down" feeling),
  • Cough,
  • Fever,
  • Abdominal pain,
  • Conjunctivitis (red-eye, itchy eye and thick discharge),
  • Dark-colored urine
  • Cytomegalovirus infection scored 8.95 points and ranked 61st on the list.
  • After inputting all CMs as keywords and narrowing down the number of diseases, cytomegalovirus infection was listed as the 5th out of all 16 suspected diseases.

Table 2. The second search and its results, CMs for the search (the next 11 items are added to the original search items):

  • Jaundice,
  • Lymph node swelling,
  • Hepatosplenomegaly,
  • Nausea,
  • Vomiting,
  • Elevation of serum alanine aminotransferase level,
  • Tachycardia,
  • Skin lesion,
  • Dyspnea,
  • Leukocytosis,
  • Pulmonary infiltration (abnormal lung shadow in the chest x-ray)
  • Cytomegalovirus infection scored 9.72 points and ranked 20th on the list.
  • Among the CMs and inputs used in the primary search, some CMs were excluded as keywords because "Dark-colored urine" has the same meaning as "Jaundice", "Nausea" is almost the same meaning as "Vomiting", and "Tachycardia" is a nonspecific symptom. After inputting another 15 CMs as keywords and narrowing down the diseases, cytomegalovirus infection was listed as 4th on the ranking scale out of all 15 diseases.

The ultimate diagnosis for this case is cytomegalo-virus infection, and we can summarize the verified results from Tables 1 and 2:

  • The more CMs used during the primary search, the higher the efficiency of the CACDDST will be.
  • The efficiency of the CACDDST will be dramatically improved by conducting the secondary search. However, it is crucial that the keywords should be carefully selected.
  • The diagnostic decision-support tool has proved to be useful if the CACDDST is prepared and used appropriately.

< Conclusion >

Reaching a foolproof diagnosis is never an easy job for a clinician. Often, a simple diagnostic procedure or test is overlooked and the disese eludes diagnosis. This is the reason we need such a system to be realized.

Since the beginning of the 1980s, we have sought new methods to analyze a patient's clinical manifestations by using computers that can store a multitude of informations. Nowadays, we can utilize the originally prepared disease knowledge database (DKDB) as a CACDDST, which can be used in the clinical field of our daily works.

This system is already available to the public at this site. We eagerly look forward to comments with respect to this system's performance.

(*) John AK, Henry R, Chadd S, Findlay W, Sanjay S.: A Jaundiced Eye. N Engl J Med 2006;354:1516-1520.