PubMed Health⌕ Search

Biomedical subjects

W J Long

Publications and source records attributed to W J Long.

At least 19 recordsLinked to original sources
PubMed2000

New approaches to measuring the performance of programs that generate differential diagnoses using ROC curves and other metrics.

INTRODUCTION: Evaluation of computer programs which generate multiple diagnoses can be hampered by a lack of effective, well recognized performance metrics. We have developed a method to calculate mean sensitivity and specificity for multiple diagnoses and generate ROC curves. METHODS: Data came from a clinical evaluation of the Heart Disease Program (HDP). Sensitivity, specificity, positive and negative predictive value (PPV, NPV) were calculated for each diagnosis type in the study. A weighted mean of overall sensitivity and specificity was derived and used to create an ROC curve. Alternative metrics Comprehensiveness and Relevance were calculated for each case and compared to the other measures. RESULTS: Weighted mean sensitivity closely matched Comprehensiveness and mean PPV matched Relevance. Plotting the Physician's sensitivity and specificity on the ROC curve showed that their discrimination was similar to the HDP but sensitivity was significantly lower. CONCLUSIONS: These metrics give a clear picture of a program's diagnostic performance and allow straightforward comparison between different programs and different studies.

Diagnosis, Computer-Assisted↗
PubMed1998

A government purchaser perspective: TennCare--strengthening the safety net.

BACKGROUND: On January 1, 1994, in response to escalating Medicaid costs, the State of Tennessee implemented TennCare, a statewide managed health care system for three population groups: Medicaid eligibles, uninsurable individuals, and people who lacked access to insurance through their place of employment. METHODS: Under the TennCare program, the State contracts with ten managed care organizations to provide a comprehensive benefit package to each enrollee. Preventive services are exempt from copayment or deductible requirements. Each patient is assigned to a primary care provider who is responsible for assuring that the enrollee receives preventive services. RESULTS: More than 400,000 previously uninsured and uninsurable persons now receive health care coverage through the TennCare program. An extensive quality assurance program monitors managed care organization service delivery, with particular emphasis on prevention and access. Effective case management, combined with an enrollment cap, have enabled the program to grow at a predictable rate, without exceeding the annual rate of growth of state revenues. CONCLUSIONS: The TennCare program has been successful in using the savings experienced during the shift from a fee-for-service to a managed care Medicaid program to expand coverage to previously uninsured populations.

Humans↗
PubMed1998

Using classification tree and logistic regression methods to diagnose myocardial infarction.

Early and accurate diagnosis of myocardial infarction (MI) in patients who present to the Emergency Room (ER) complaining of chest pain is an important problem in emergency medicine. A number of decision aids have been developed to assist with this problem but have not achieved general use. Machine learning techniques, including classification tree and logistic regression (LR) methods, have the potential to create simple but accurate decision aids. Both a classification tree (FT Tree) and an LR model (FT LR) have been developed to predict the probability that a patient with chest pain is having an MI based solely upon data available at time of presentation to the ER. Training data came from a data set collected in Edinburgh, Scotland. Each model was then tested on a separate Edinburgh data set, as well as on a data set from a different hospital in Sheffield, England. Previously published models, the Goldman classification tree[1] and Kennedy LR equation[2], were evaluated on the same test data sets. On the Edinburgh test set, results showed that the FT Tree, FT LR, and Kennedy LR performed equally well, with ROC curve areas of 94.04%, 94.28%, and 94.30%, respectively, while the Goldman Tree's performance was significantly poorer, with an area of 84.03%. The difference in ROC areas between the first three models and the Goldman model is significant beyond the 0.0001 level. On the Sheffield test set, results showed that the FT Tree, FT LR, and Kennedy LR ROC areas were not significantly different (p > = 0.17), while the FT Tree again outperformed the Goldman Tree (p = 0.006). Unlike previous work[3], this study indicates that classification trees, which have certain advantages over LR models, may perform as well as LR models in the diagnosis of patients with MI.

Algorithms↗
PubMed1998

Differential diagnoses of the heart disease program have better sensitivity than resident physicians.

We describe a prospective clinical evaluation of a computer program to assist with the diagnosis of heart disease. The Heart Disease Program (HDP) is a large diagnostic program covering most areas of heart disease and some related areas of general medicine. The program's output is a set of differential diagnoses with explanations and it can be deployed in a clinical setting using a web interface. A framework for assessing the complex diagnostic summaries generated by the HDP was developed and the program's diagnostic accuracy in a clinical setting was assessed. The diagnoses used for comparison came from the physician entering the case, a "gold standard" assigned by review of patient charts and investigations, and the opinions of expert cardiologists. The data collection, methods of comparison, example analyses and results on 114 cases are presented here. The HDP had a significantly higher sensitivity for both the gold standard (60%) and the cardiologist's diagnoses (58%) than the physicians did (39%, 34%). These findings were consistent in the 2 collection cohorts and for the more serious diagnoses alone. The significance of these findings and the many challenges in comparing these different diagnoses and minimizing bias are discussed.

Cardiology↗
PubMed1997

Reasoning requirements for diagnosis of heart disease.

Over the past dozen years, the Heart Disease Program (HDP) has been developed to assist physicians in reasoning about cardiovascular disorders. Driven by several evaluations, the inference mechanism has progressed from a logic based model, to a Bayesian Probability Network (BPN) and finally a pseudo-Bayesian network with temporal and severity reasoning. Though aspects of cardiovascular reasoning are handled well by BPNs, temporal reasoning, homeostatic feedback mechanisms and effects of disease severities require additional inference strategies. This article discusses how these reasoning problems are handled, and deals with closely linked issues in building the user interface to collect detailed cardiovascular data and provide clear explanations of diagnoses.

Artificial Intelligence↗
PubMed1996

Characterization of a soluble stable human cytomegalovirus protease and inhibition by M-site peptide mimics.

The human cytomegalovirus (HCMV) protease is a potential target for antiviral chemotherapeutics; however, autoprocessing at internal sites, particularly at positions 143 and 209, hinders the production of large quantities of stable enzyme for either screening or structural studies. Using peptides encompassing the sequence of the natural M-site substrate (P5-P5', GVVNA/SCRLA), we previously demonstrated that substitution of glycine for valine at the P3 position in the substrate abrogates processing by the recombinant protease in vitro. We now demonstrate that introduction of the V-to-G substitution in the P3 positions of the two major internal processing sites, positions 143 and 209, in the mature HCMV protease renders the enzyme stable to autoprocessing. When expressed in Escherichia coli, the doubly substituted protease was produced almost exclusively as the 30-kDa full-length protein. The full-length V141G, V207G (V-to-G changes at positions 141 and 207) protease was purified as a soluble protein by a simple two-step procedure, ammonium sulfate precipitation followed by DEAE ion-exchange chromatography, resulting in 10 to 15 mg of greater than 95% pure enzyme per liter. The stabilized enzyme was characterized kinetically and was indistinguishable from the wild-type recombinant protease, exhibiting Km and catalytic constant values of 0.578 mM and 13.18/min, respectively, for the maturation site (M-site) peptide substrate, GVVNASCRLARR (underlined residues indicate additions to or substitutions from peptides derived from the wild-type substrate). This enzyme was also used to perform inhibition studies with a series of truncated and/or substituted maturation site peptides. Short nonsubstrate M-site-derived peptides were demonstrated to be competitive inhibitors of cleavage in vitro, and these analyses defined amino acids VVNA, P4 through P1 in the substrate, as the minimal substrate binding and recognition sequence for the HCMV protease.

Amino Acid Sequence↗
PubMed1996

Web interface for the Heart Disease Program.

The task of making a large complex diagnostic program available to a broad audience of physicians has become more feasible with the ubiquitous accessibility of the client-server architecture of the World Wide Web. This paper describes the design and implementation of a Web interface for the Heart Disease Program (HDP). The client-server architecture imposes a number of requirements on the program. The graphical capabilities of the Web enable a number of enhancements to the program but also cause some limitations. Our initial experience with physicians using the HDP through the Web interface has been positive and we are now conducting an evaluation of the HDP using this form of access.

Computer Communication Networks↗
PubMed1995

A comparison of performance of mathematical predictive methods for medical diagnosis: identifying acute cardiac ischemia among emergency department patients.

BACKGROUND: There is increasing interest in mathematical methods for the prediction of medical outcomes. Three methods have attracted particular attention: logistic regression, classification trees (such as ID3 and CART), and neural networks. To compare their relative performance, we used a large clinical database to develop and compare models using these methods. METHODS: Each modeling method was used to generate predictive instruments for acute cardiac ischemia (which includes acute myocardial infarction and unstable angina pectoris), using prospectivel-collected clinical data on 5773 patients, who presented over a two year period to six hospitals' emergency departments with chest pain or symptoms suggesting acute ischemia. This data set was then split into training (n = 3453) and test (n = 2320) sets. Of 200 available variables, modeling was restricted to those available within the first 10 minutes of emergency department care (history, physical exam, and electrocardiogram). RESULTS: When the number of variables was limited to eight, representing a practical number for input in the real-time clinical setting, the logistic regression's receiver-operating characteristic (ROC) curve area, as a measure of diagnostic performance, was 0.887; the classification tree model's ROC curve area was 0.858, and the neural network's ROC curve area was 0.902. When the number of variables used by a model was not limited, the logistic regression's ROC area was 0.905, the classification tree model's 0.861, and the neural network's 0.923. Among these models the neural networks had noticeably poorer calibration. When the outputs from each of these unrestricted models were presented to each of the other methods as an additional independent variable, the ROC areas of the new "hybrid" models were not significantly better than the original unlimited models (ROC areas 0.858 to 0.920). CONCLUSIONS: Logistic regression, classification tree, and neural network models all can provide excellent predictive performance of medical outcomes for clinical decision aids and policy models. Their ultimate limitations seem due to the availability of the information in data (a "data barrier") rather than their respective intrinsic properties. Choices between these methods would seem to be most appropriately based on the needs of the specific application, rather than on the premise that any one of these methods is intrinsically more powerful.

Adult↗
PubMed1994

Peptide substrate cleavage specificity of the human cytomegalovirus protease.

The human cytomegalovirus UL80 gene encodes an 80-kDa precursor polyprotein whose N-terminal 256-amino acid domain is a protease. This enzyme cleaves a specific peptide bond that results in its own release from the precursor, as well as a peptide bond near the C terminus of the viral assembly protein. The latter cleavage is apparently required for encapsidation of the viral genomic DNA and maturation of the viral capsid. A series of peptide substrates, representing the assembly protein cleavage site, was used to study the enzyme's substrate requirements and specificity. It was found that efficient cleavage minimally required the amino acid residues spanning the P4 to P4' positions. Substitution at any of these residues adversely affected the reaction. Conservation of the hydrophobic residues at P3 and P4 was essential. In addition, cleavage of a peptide representing the protease domain release site was reduced almost 100-fold relative to cleavage of the assembly protein maturation site peptide substrate.

Amino Acid Sequence↗
PubMed1994

Susceptibilities of human immunodeficiency virus type 1 enzyme and viral variants expressing multiple resistance-engendering amino acid substitutions to reserve transcriptase inhibitors.

To evaluate the potential that multiply resistant human immunodeficiency virus type 1 variants may arise during combination nucleoside and nonnucleoside reverse transcriptase inhibitor therapy, we constructed a series of mutant reverse transcriptase enzymes and viruses that coexpressed various combinations of resistance-associated amino acid substitutions. Substitutions at residues 100 (Leu-->Ile) and 181 (Tyr-->Cys), which mediate resistance to the nonnucleosides, suppressed resistance to 3'-azido-3'-deoxythymidine (AZT) when coexpressed with AZT-specific substitutions. However, a number of viral variants that exhibited significantly reduced susceptibilities to both classes of inhibitors were constructed.

Drug Resistance, Microbial↗
PubMed1994

Summarization of complex causal diagnostic hypotheses.

The Heart Disease Program produces detailed causal diagnostic hypotheses for patients with cardiovascular diseases. This poster discusses our experience with summarizing these hypotheses for the physician. The basic approach is to merge the nodes of the structure indicating causal mechanism into the more important nodes. Analysis of the results shows that to generate effective summaries the identification of syndromes is very important, the definitions of the labels need to be carefully enforced, the causality of diseases should be supported by evidence and not just probability, and the sense of causal order must be carefully preserved.

Heart Diseases↗
PubMed1993

A comparison of logistic regression to decision-tree induction in a medical domain.

This paper compares the performance of logistic regression to decision-tree induction in classifying patients as having acute cardiac ischemia. This comparison was performed using the database of 5773 patients originally used to develop the logistic-regression tool and test it prospectively. Both the ability to classify cases and the ability to estimate the probability of ischemia were compared on the default tree generated by the C4 version of ID3. They were also compared on a tree optimized on the learning set by increased pruning of overspecified branches, and on a tree incorporating clinical considerations. Both the LR tool and the improved trees performed at a level fairly close to that of the physicians, although the LR tool definitely performed better than the decision tree. There were a number of differences in the performance of the two methods, shedding light on their strengths and weaknesses.

Algorithms↗
PubMed1993

Comprehensive mutant enzyme and viral variant assessment of human immunodeficiency virus type 1 reverse transcriptase resistance to nonnucleoside inhibitors.

The nonnucleoside reverse transcriptase (RT) inhibitors comprise a class of structurally diverse compounds that are functionally related and specific for the human immunodeficiency virus type 1 RT. Viral variants resistant to these compounds arise readily in cell culture and in treated, infected human. Therefore, the eventual clinical usefulness of the nonnucleoside inhibitors will rely on a thorough understanding of the genetic and biochemical bases for resistance. A study was performed to assess the effects of substitutions at each RT amino acid residue that influences the enzyme's susceptibility to the various nonnucleoside compounds. Single substitutions were introduced into both purified enzyme and virus. The resulting patterns of resistance were markedly distinct for each of the tested inhibitors. For instance, a > 50-fold loss of enzyme susceptibility to BI-RG-587 was engendered by any of four individual substitutions, while the same level of relative resistance to the pyridinone derivatives was mediated only by substitution at residue 181. Similarly, substitution at residue 181. Similarly, substitution at residue 106 had a noted effect on virus resistance to BI-RG-587 but not to the pyridinones. The opposite effect was mediated by a substitution at residue 179. Such knowledge of nonucleoside inhibitor resistance profiles may help in understanding the basis for resistant virus selection during clinical studies of these compounds.

Acquired Immunodeficiency Syndrome↗
PubMed1992

Functional analysis of HIV-1 reverse transcriptase amino acids involved in resistance to multiple nonnucleoside inhibitors.

Several novel, structurally distinct classes of specific human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) nonnucleoside inhibitors have been described recently. These include the pyridinone derivatives L-697,639, L-697,661, and L-696,229 as well as BI-RG-587 and the tetrahydroimidazo[4,5,1-j,k]-benzodiazepin-2(1H)-one and -thione compounds. Previous studies have implicated involvement of the RT amino acid residues at positions 103, 181, and 188 in the activity of the compounds. Accordingly, HIV-1 RT mutants containing a series of amino acid substitutions at these positions were constructed. The relative resistance of purified mutant enzymes to each of the inhibitors was assessed. This analysis established the functional equivalence of the three inhibitor classes and provided evidence for the interaction of the 103 site with the 181/188 region. Amino acid substitutions at these positions were also found to influence RT sensitivity to inhibition by phosphonoformate, thereby suggesting a close association between this pyrophosphate analog's binding site in RT and the binding site of the nonnucleoside inhibitors. In addition, aromatic stacking of the amino acid side groups at residues 181 and 188 was suggested to be required for inhibitor activity.

Amino Acid Sequence↗
PubMed1992

Development of a knowledge base for diagnostic reasoning in cardiology.

This paper reports on a formative evaluation of the diagnostic capabilities of the Heart Failure Program, which uses a probability network and a heuristic hypothesis generator. Using 242 cardiac cases collected from discharge summaries at a tertiary care hospital, we compared the diagnoses of the program to diagnoses collected from cardiologists using the same information as was available to the program. With some adjustments to the knowledge base, the Heart Failure Program produces appropriate diagnoses about 90% of the time on this training set. The main reasons for the inappropriate diagnoses of the remaining 10% include inadequate reasoning with temporal relations between cause and effect, severity relations, and independence of acute and chronic diseases.

Artificial Intelligence↗
PubMed1992

Hospital positioning: a strategic tool for the 1990s.

The authors extend the process of market positioning in the health care sector by focusing on the simultaneous utilization of traditional research methods and emerging new computer-based adaptive perceptual mapping technologies and techniques.

Computer Graphics↗