Differential diagnosis of jaundice: applicability of the Copenhagen Pocket Chart proved in Stockholm patients

Transfer of hypercalcemia discriminant functions between local hospitals

Transferability of discriminant functions is potentially useful both from an economical point of view and because, in general, medical knowledge, in this case discriminant functions, should be transferable. In the present study we have evaluated the transferability of discriminant functions, estimated from routine laboratory analysis, age and sex in two consecutively recorded populations with hypercalcemia including 162 and 257 patients with hypercalcemia. Discriminant functions were developed for each sex to distinguish between hypercalcemia associated with malignancy and hypercalcemia associated with other medical diseases. The total diagnostic accuracy in Herlev was 82 and 78%, in women and men, and increased to 87 and 86% in both sexes considering cases classified with posterior probability levels of 60%. In Hvidovre the total diagnostic accuracy was 81 and 84% in women and men, and increased to 83 and 89% at posterior probability levels higher than 60%. Transfer of the discrimination functions between the hospitals was followed by a decrease in diagnostic accuracy of 6-16%. At a posterior probability of 60% the diagnostic accuracies were 79% or more in the receiving hospital, in both sexes, except for men in Hvidovre. In relation to these results the concept of genuine and non-genuine transfer factors is introduced and discussed.

Computer-assisted diagnosis in gastroenterology

During the last 20 years, Danish gastroenterologists and biostatisticians have been involved in practical and theoretical research aimed at providing computer-assisted decision support in clinical practice. On the practical side, we summarize two clinical applications, one in the acute abdomen, the other in jaundice. The former project is part of a worldwide effort which appears able to reduce considerably the perforated appendix rate and the negative laparotomy rate. In the latter project, also entwined in international activities, three methods of early statistical discrimination of jaundice causes have been tested, and it has been shown how ultrasonographic data can be incorporated into the diagnostic assessment. On the theoretical side, a Dutch-Danish group has been looking into how one best designs a trial and analyses its data in the context of measuring the quality and impact of machine-made diagnostic advice. Having pointed out the international ramifications of these activities, we outline what may lie ahead: Gastroenterologists must be prepared to join large-scale field trials of clinical advice-giving software. In the future, however, such software will often become embedded in general hospital recordkeeping systems.

Evaluation of a computer-based diagnostic score system in the diagnosis of jaundice and cholestasis

A prospective series of 220 patients with jaundice and/or cholestasis was analysed by means of discriminant analysis to evaluate the diagnostic value of various symptoms and signs and basic hepatopancreatobiliary laboratory tests in the differential diagnosis of these patients. In addition, a computer-based diagnostic score (DS) system was developed and compared with the diagnostic value of clinical evaluation (CE), ultrasound (US), computed tomography (CT), and endoscopic retrograde cholangiopancreatography (ERCP). A multivariate stepwise discriminant analysis showed five independent diagnostic factors in distinguishing extrahepatic and intrahepatic diseases: duration of jaundice (p = 0.002), serum protein concentrations (p = 0.002), Normotest (p = 0.04), fever during the past 3 months (p = 0.54), and age (p = 0.58). To sum up the contributions of independent factors, a DS was developed. The discrimination function was as follows: duration of jaundice x 0.97 + age x -0.40 + fever during the past 3 months x 0.33 + serum protein concentration x 0.35 + Normotest x -0.27. The diagnostic sensitivity of DS in the detection of extrahepatic disease was 96%, with a specificity of 80% and an efficiency of 93%. The sensitivities of the imaging methods (62-85%) were inferior to that of DS, whereas the specificities were better (94-98%). The sensitivity of CE was only slightly lower (86%) than that of DS, but the specificity was lowest (57%). In conclusion, relatively few (5 of 40) of the classical symptoms, signs, and biochemical tests proved to be independent diagnostic factors in distinguishing extrahepatic and intrahepatic diseases.(ABSTRACT TRUNCATED AT 250 WORDS)

Converting a rule-based expert system into a belief network

The theory of belief networks offers a relatively new approach for dealing with uncertain information in knowledge-based (expert) systems. In contrast with the heuristic techniques for reasoning with uncertainty employed in many rule-based expert systems, the theory of belief networks is mathematically sound, based on techniques from probability theory. It therefore seems attractive to convert existing rule-based expert systems into belief networks. In this article we discuss the design of a belief network reformulation of the diagnostic rule-based expert system HEPAR. For the purpose of this experiment we have studied several typical pieces of medical knowledge represented in the HEPAR system. It turned out that, due to the differences in the type of knowledge represented and in the formalism used to represent uncertainty, much of the medical knowledge required for building the belief network concerned could not be extracted from HEPAR. As a consequence, significant additional knowledge acquisition was required. However, the objects and attributes defined in the HEPAR system, as well as the conditions in production rules mentioning these objects and attributes, were useful for guiding the selection of the statistical variables for building the belief network. The mapping of objects and attributes in HEPAR to statistical variables is discussed in detail.

A statistically rule-based decision support system for the management of patients with suspected liver disease

The paper describes how a decision support system in liver diseases, mostly oriented to prediction of the necessity for liver biopsy, has been developed. The system designed is a hybrid one and consists of two parts: logical and statistical. The logical part contains rules, formulated on the basis of current medical knowledge, which enables recognition of clear cases; diseased or non-diseased. The unclear cases are classified on the basis of rules statistically extracted from databases. These rules have been reached after a comprehensive exploratory analysis of the sample of 165 patients with slightly to moderately raised levels of routine liver tests but without signs or symptoms of liver diseases. The extracted decision diagrams which simulate traditional medical diagnosis conduct have been found to be superior to discriminant analysis and probabilistic inductive learning. They use only a limited number of laboratory tests to detect the necessity for biopsy.

Computer-aided decision support in hepatology

The aim of this study was to describe and to evaluate the publications of the last 30 years devoted to computer-aided decision support in clinical hepatology. The search used Medlars and references of articles. Computer-aided decision support (CADS) was classified in two categories: statistical systems and knowledge-based systems. Two specific questionnaires were used for methodologic evaluation, one for statistical systems and one for knowledge-based systems. They were filled out independently by two observers. A total of 31 papers were selected among 55 identified between 1960 and 1991. The maximum possible for the two scores was 24. The methodologic quality ranged from 4 to 22 (median, 12) for statistical systems and from 8 to 12 (median, 9) for knowledge-based systems. The poor level of methodology could explain in part the lack of utilization of computer-aided decision support in the daily clinical practice of hepatologists.

Second evaluation of HEPAR, an expert system for the diagnosis of disorders of the liver and biliary tract

HEPAR is an expert system which can be used as a supportive tool in the diagnosis of disorders of the liver and biliary tract. When consulted for a particular patient, the system assesses the patient's disorder in two stages. In the first stage, data of medical history, physical examination and routine laboratory tests are used to determine whether the disorder is hepatocellular or biliary obstructive, and whether benign or malignant features are present. In the second stage, the system produces a differential diagnosis out of a set of 80 disease categories, using the conclusions of the first stage of the assessment and the results of a restricted number of supplementary investigations, i.e. serological tests and ultrasonography. The conclusions of the two stages are ordered by the amount of evidence computed by HEPAR. The system is not yet suitable for clinical use. The system has been evaluated using data of 181 patients selected from a population of 214 consecutively admitted patients with jaundice. The patients were classified by the system as having a hepatocellular or biliary obstructive, and a benign or malignant disorder in 96% and 100% of the cases, respectively. When comparing the conclusions with strongest evidence with the final clinical diagnoses, the patients were classified correctly in 86% and 83% of the classified cases, respectively. A differential diagnosis was produced in 87% of the patients. A correspondence between the diagnosis with strongest evidence and the final clinical diagnosis was found in 78% of the classified patients.

Development and validation of HEPAR, an expert system for the diagnosis of disorders of the liver and biliary tract

HEPAR is an expert system in the field of hepatology which can be used as a supportive tool in the diagnosis of disorders of the liver and biliary tract. The representation language employed in HEPAR consists of production rules with object-attribute-value triples. In the system, medical diagnosis is modelled according to the strategy which is clinically used in this field. The patient's disorder is assessed in two stages. In the first stage, available data from medical interview, physical examination and simple laboratory tests are used to determine whether the disorder is hepatocellular or biliary obstructive in nature, and whether benign or malignant features are present. In the second stage the system produces a subset of possible diagnoses out of a set of more than 80 disease categories, using additional data from supplementary tests. The results of the first stage classification cause considerable pruning of the search space. In this paper the structure, development process and approach followed in the validation of the HEPAR system are described. The results of a recent performance validation study using data from 181 consecutive patient cases are discussed.

Expert systems and the pancreatic cancer problem: decision support in the pre-operative diagnosis

In this paper, after reviewing the main issue in artificial intelligence, decision support systems, medical decision-making, expert systems and some of their applications in medicine, we focus on the diagnostic aspect of pancreatic cancer. We briefly examine the most significant applications both from the oncological and from the diagnostic point of view. We discuss the medical problems mentioning incidence and mortality, aetiological factors and diagnosis, considering the roles of surgery and adjuvant therapies. Finally we justify the decision to develop an expert system in such a medical domain and discuss the SPES (Surgical Pancreatic Expert System) project, its parts dealing with the different medical phases of pancreatic cancer diagnosis and therapy: pre-operative, intra-operative and adjuvant therapies. In particular we discuss diagnostic aspects of pancreatic cancer disease, pointing out the aims of the project, methodologies, tools used and future developments.

HEPAR: an expert system for the diagnosis of disorders of the liver and biliary tract

The HEPAR system is a medical expert system intended as a supportive tool for the diagnosis of disorders of the liver and biliary tract. In the system, the role of information from the medical interview, physical examination, and simple non-invasive diagnostic methods, such as ultrasonography, is emphasized in order to reduce the number of diagnostic procedures to be applied. Medical diagnosis is modelled in the system using the formalism of if-then rules. Based on patient data available, the system determines whether the data indicate a hepatocellular or a biliary obstructive disorder, and whether or not benign or malignant features are present. Moreover, the system produces a subset of possible diagnoses out of a set of more than 70 diagnostic categories. The system is still under development. In a preliminary study, the HEPAR system has been evaluated using data from 94 jaundiced patients. Conclusions with respect to the hepatocellular or biliary obstructive nature of the disorder, the benign or malignant nature of the disorder, and the final diagnosis, were reached in 95%, 65%, and 80% of the cases, respectively. These conclusions were correct in 85%, 92%, and 80% of the classified cases, respectively. The system was shown not to be sensitive to incompleteness of information.

A model-based system to determine the relative value of different variables in a diagnostic system using Bayes theorem

A model system is described which simulates the presenting features of cases of vaginal discharge. This system was used to examine the effects of removing individual clinical features on the overall efficiency of diagnosis by Bayes theorem. The diagnostic efficiency was significantly reduced by elimination of inflammation, of a frothy discharge, or of a curdy discharge. Elimination of more than one significant factor further reduced the number of correct diagnoses, but elimination of more than one non-significant factor made no obvious difference. The most significant clinical feature was presence of inflammation; elimination of this feature had a substantial effect on the diagnosis of gardnerella, viral, gonorrhoea and foreign body. Elimination of two of the variables (bloodstaining and odour) which did not influence overall diagnostic efficiency nevertheless had a substantial effect on the diagnosis of neoplasms and foreign bodies. It is proposed that a simulation of this type is of potential practical value in determining a minimum subset of clinical features for diagnostic systems involving Bayes theorem.

Assessing diagnostic tests by a strictly proper scoring rule

Evaluation of univariate quantitative diagnostic tests by strictly proper scoring rules is considered as an alternative to the traditional error rate measures. In principle, the posterior probability of disease as a function of the test value is estimated from training observations, and subsequently the score is assessed on a set of test samples. The same subjects may serve as training and test samples when the bootstrap procedure is applied for estimation of standard errors and correction of bias. The method is demonstrated using serum bile acids and bilirubin in patients with liver disease. The power for comparison of scores from two tests is compared with that from error rate measures for some typical situations.

Use of computers in diagnosis of jaundice

The development of computer-assisted diagnostic systems for the differential diagnosis of jaundice has been attempted in several studies during the last 25-30 years. All working systems have depended on numerical methods whereas expert systems have not yet become operational. The first step in the construction of a system is the collection of a data base from a series of jaundiced patients of clinical information (indicants) i.e. symptoms, signs, and the results from laboratory tests. The best discriminating indicants are selected and processed into a mathematical rule. The performance of this rule must be tested on an independent test sample of relevant patients. The performance may also be compared to that of clinicians. So far the computerized diagnoses have not been more accurate than those of clinicians. However, computer-assisted diagnostics may form an important first step in clinical decision making regarding the selection of a confirmatory diagnostic test (direct cholangiography, ultrasonography, liver biopsy, etc.) in the evaluation of the jaundiced patient.

Viral hepatitis. The alphabet game

Differential diagnosis of viral hepatitis begins with a check for darkened urine and bile in the urine. These hallmarks of conjugated hyperbilirubinemia immediately rule out prehepatic liver disease. Next, studies are done for the elevated transaminase levels that are characteristic of hepatitis infection, and a thorough history is taken to rule out drug- and toxin-induced hepatitis that may mimic acute viral hepatitis. Elevated alkaline phosphatase is a good marker of cholestasis. Ultrasonography can clarify this diagnosis. The classic presenting symptoms of viral hepatitis are jaundice, nausea, vomiting, malaise, anorexia, and dull right upper quadrant pain. However, serologic studies are needed to detect the presence of specific viral agents.

Differential diagnosis of jaundice: junior staff experience with the Copenhagen pocket chart

Originally published in 1984, the Copenhagen Pocket Chart for early differentiation between causes of jaundice has been tested with success in centres outside Denmark. Using a logistic discrimination model, it estimates probabilities of obstruction and non-obstruction in each case (and provides a further subdivision if desired). Here we evaluate its performance in the hands of young clinicians on a consecutive series of 173 jaundiced patients from two Danish hospitals. The chart performed as well as in the original series: confident diagnoses (probability greater than or equal to 0.80) were assigned to 124 patients; of these 115 proved correct (93%). In 46 patients diagnostic probabilities were less than 0.80, and 3 patients had an unknown cause of jaundice. There were 108 cases in which physician and chart were in agreement, both with a confident diagnosis, and only one of these cases was wrong. In one hospital, contributing 107 cases, each patient was independently examined by a medical student in addition to the physician's examination. Student performance was equally good, practically speaking, in particular when taking the scores on the chart into consideration. As to observer disagreement, the student and the physician typically differed on 0-2 of the chart's 21 items. In no case, however, did this lead to a confident obstructive diagnosis being changed into a confident diagnosis of non-obstruction, or vice versa.