Bio


Shachter's interests include:

Influence diagram knowledge representation and solution

Intelligent decision systems

Medical decision analysis

Decision analysis fundamentals

Planning under uncertainty

Prof. Shachter joined Stanford's faculty directly after receiving his Ph.D. degree. His doctoral dissertation developed a method for purchasing an expert's forecast that encourages accurate revelation of the expert's beliefs as probabilities. Since then his research has focused on the representation, manipulation, and analysis of uncertainty and probabilistic reasoning in decision systems. As part of this work, he developed the DAVID influence diagram processing system for the Macintosh. He has developed models scheduling patients for cancer follow-up, and analyzing vaccination strategies for HIV and Helobacter pylori.

Academic Appointments


Professional Education


  • BS, Massachusetts Institute of Technolog, Management (1976)
  • PhD, UC Berkeley, Operations Research (1982)

Current Research and Scholarly Interests


Prof. Shachter joined Stanford's faculty directly after receiving his Ph.D. degree. His doctoral dissertation developed a method for purchasing an expert's forecast that encourages accurate revelation of the expert's beliefs as probabilities. Since then his research has focused on the representation, manipulation, and analysis of uncertainty and probabilistic reasoning in decision systems. As part of this work, he developed the DAVID influence diagram processing system for the Macintosh. He has developed models scheduling patients for cancer follow-up, and analyzing vaccination strategies for HIV and Helobacter pylori.

He has worked closely with many students in Bioinformatics, where he holds a courtesy appointment. He has been active in the Conference on Uncertainty in Artificial Intelligence, is a full member of INFORMS and its Decision Analysis Society. He has held memberships in the American Association for Artificial Intelligence, the Society for Medical Decision Making, and the Society for Decision Professionals.

2013-14 Courses


Journal Articles


  • Formulating Asymmetric Decision Problems as Decision Circuits DECISION ANALYSIS Bhattacharjya, D., Shachter, R. D. 2012; 9 (2): 138-145
  • Asymmetric Decision Problems as Decision Circuits.  Decision Analysis Bhattacharjya, D., Shachter, R. 2012; 2 (9): 138-145
  • Cost-Effectiveness of a Potential Prophylactic Helicobacter pylori Vaccine in the United States JOURNAL OF INFECTIOUS DISEASES Rupnow, M. F., Chang, A. H., Shachter, R. D., Owens, D. K., Parsonnet, J. 2009; 200 (8): 1311-1317

    Abstract

    Helicobacter pylori vaccines are under development to prevent infection. We quantified the cost-effectiveness of such a vaccine in the United States, using a dynamic transmission model.We compartmentalized the population by age, infection status, and clinical disease state and measured effectiveness in quality-adjusted life years (QALYs). We simulated no intervention, vaccination of infants, and vaccination of school-age children. Variables included costs of vaccine, vaccine administration, and gastric cancer treatment (in 2007 US dollars), vaccine efficacy, quality adjustment due to gastric cancer, and discount rate. We evaluated possible outcomes for periods of 10-75 years.H. pylori vaccination of infants would cost $2.9 billion over 10 years; savings from cancer prevention would be realized decades later. Over a long time horizon (75 years), incremental costs of H. pylori vaccination would be $1.8 billion, and incremental QALYs would be 0.5 million, yielding a cost-effectiveness ratio of $3871/QALY. With school-age vaccination, the cost-effectiveness ratio would be $22,137/QALY. With time limited to <40 years, the cost-effectiveness ratio exceeded $50,000/QALY.When evaluated with a time horizon beyond 40 years, the use of a prophylactic H. pylori vaccine was cost-effective in the United States, especially with infant vaccination.

    View details for DOI 10.1086/605845

    View details for Web of Science ID 000270089400018

    View details for PubMedID 19751153

  • Are Patients Getting the oGisto in Risk Communication? Patient Understanding of Prognosis in Breast Cancer Treatment JOURNAL OF CANCER EDUCATION Hutton, D. W., Belkora, J. K., Shachter, R. D., Moore, D. H. 2009; 24 (3): 194-199

    Abstract

    Many oncologists consult the Adjuvant! prognostic model to communicate risk with breast cancer patients; however, little is known about how effective that communication is.The authors analyzed this small data set featuring 20 breast cancer patients' risk estimates, focusing on rankings or gist of the estimates.Overall, there was no gain in the accuracy of patient rankings. The number of patients with more accurate estimates was matched by the number of patients with less accurate estimates after consultation.The current methods used by oncologists to present Adjuvant! risks were not effective in helping patients to get the gist of their risks.

    View details for DOI 10.1080/08858190902876452

    View details for Web of Science ID 000266978800007

    View details for PubMedID 19526406

  • How can economic schemes curtail the increasing sex ratio at birth in China? DEMOGRAPHIC RESEARCH Bhattacharjya, D., Sudarshan, A., Tuljapurkar, S., Shachter, R., Feldman, M. 2008; 19: 1831-1850

    Abstract

    Fertility decline, driven by the one-child policy, and son preference have contributed to an alarming difference in the number of live male and female births in China. We present a quantitative model where people choose to sex-select because they perceive that married sons are more valuable than married daughters. Due to the predominant patrilocal kinship system in China, daughters-in-law provide valuable emotional and financial support, enhancing the perceived present value of married sons. We argue that inter-generational transfer data will help ascertain the extent to which economic schemes (such as pension plans for families with no sons) can curtail the increasing sex ratio at birth.

    View details for Web of Science ID 000259977900001

    View details for PubMedID 21113272

  • Bayesian network to predict breast cancer risk of mammographic microcalcifications and reduce number of benign biopsy results: Initial experience RADIOLOGY Burnside, E. S., Rubin, D. L., Fine, J. P., Shachter, R. D., Sisney, G. A., Leung, W. K. 2006; 240 (3): 666-673

    Abstract

    To retrospectively determine whether a Bayesian network (BN) computer model can accurately predict the probability of breast cancer on the basis of risk factors and mammographic appearance of microcalcifications, to improve the positive predictive value (PPV) of biopsy, with pathologic examination and follow-up as reference standards.The institutional review board approved this HIPAA-compliant study; informed consent was not required. Results of 111 consecutive image-guided breast biopsies performed for microcalcifications deemed suspicious by radiologists were analyzed. Mammograms obtained before biopsy were analyzed in a blinded manner by a breast imager who recorded Breast Imaging Reporting and Data System (BI-RADS) descriptors and provided a probability of malignancy. The BN uses probabilistic relationships between breast disease and mammography findings to estimate the risk of malignancy. Probability estimates from the radiologist and the BN were used to create receiver operating characteristic (ROC) curves, and area under the ROC curve (A(z)) values were compared. PPV of biopsy was also evaluated on the basis of these probability estimates.The BN and the radiologist achieved A(z) values of 0.919 and 0.916, respectively, which were not significantly different. If the 34 patients estimated by the BN to have less than a 10% probability of malignancy had not undergone biopsy, the PPV of biopsy would have increased from 21.6% to 31.2% without missing a breast cancer (P < .001). At this level, the radiologist's probability estimation improved the PPV to 30.0% (P < .001).A probabilistic model that includes BI-RADS descriptors for microcalcifications can distinguish between benign and malignant abnormalities at mammography as well as a breast imaging specialist can and may be able to improve the PPV of image-guided breast biopsy.

    View details for DOI 10.1148/radiol.2403051096

    View details for Web of Science ID 000240098200007

    View details for PubMedID 16926323

  • Value of quantitative D-dimer assays in identifying pulmonary embolism: Implications from a sequential decision model ACADEMIC EMERGENCY MEDICINE Duriseti, R. S., Shachter, R. D., Brandeau, M. L. 2006; 13 (7): 755-766

    Abstract

    To examine the cost-effectiveness of a quantitative D-dimer assay for the evaluation of patients with suspected pulmonary embolism (PE) in an urban emergency department (ED).The authors analyzed different diagnostic strategies over pretest risk categories on the basis of Wells criteria by using the performance profile of the ELISA D-dimer assay (over five cutoff values) and imaging strategies used in the ED for PE: compression ultrasound (CUS), ventilation-perfusion (VQ) scan (over three cutoff values), CUS with VQ (over three cutoff values), computed tomography (CT) angiogram (CTA) with pulmonary portion (CTP) and lower-extremity venous portion, and CUS with CTP. Data used in the analysis were based on literature review. Incremental costs and quality-adjusted-life-years were the outcomes measured.Computed tomography angiogram with pulmonary portion and lower-extremity venous portion without D-dimer was the preferred strategy. CUS-VQ scanning always was dominated by CT-based strategies. When CTA was infeasible, the dominant strategy was D-dimer with CUS-VQ in moderate- and high-Wells patients and was D-dimer with CUS for low-Wells patients. When CTP specificity falls below 80%, or if its overall performance is markedly degraded, preferred strategies include D-dimer testing. Sensitivity analyses suggest that pessimistic assessments of CTP accuracy alter the results only at extremes of parameter settings.In patients in whom PE is suspected, when CTA is available, even the most sensitive quantitative D-dimer assay is not likely to be cost-effective. When CTA is not available or if its performance is markedly degraded, use of the D-dimer assay has value in combination with CUS and a pulmonary imaging study. These conclusions may not hold for the larger domain of patients presenting to the ED with chest pain or shortness of breath in whom PE is one of many competing diagnoses.

    View details for DOI 10.1197/j.aem.2006.02.011

    View details for Web of Science ID 000239051800008

    View details for PubMedID 16723725

  • Individualizing generic decision models using assessments as evidence JOURNAL OF BIOMEDICAL INFORMATICS Scott, G. C., Shachter, R. D. 2005; 38 (4): 281-297

    Abstract

    Complex decision models in expert systems often depend upon a number of utilities and subjective probabilities for an individual. Although these values can be estimated for entire populations or demographic subgroups, a model should be customized to the individual's specific parameter values. This process can be onerous and inefficient for practical decisions. We propose an interactive approach for incrementally improving our knowledge about a specific individual's parameter values, including utilities and probabilities, given a decision model and a prior joint probability distribution over the parameter values. We define the concept of value of elicitation and use it to determine dynamically the next most informative elicitation for a given individual. We evaluated the approach using an example model and demonstrate that we can improve the decision quality by focusing on those parameter values most material to the decision.

    View details for DOI 10.1016/j.jbi.2004.12.003

    View details for Web of Science ID 000231514200007

    View details for PubMedID 16084471

  • Influence Diagrams for Team Decision Analysis. Decision Analysis Detwarasiti, A., Shachter, R., D. 2005; 2 (4): 207-228
  • Individualizing Generic Decision Models Using Assessments as Evidence. Journal of Biomedical Informatics Scott, G., Shachter, R. 2005; 4 (38): 281-29
  • A probabilistic expert system that provides automated mammographic-histologic correlation: Initial experience AMERICAN JOURNAL OF ROENTGENOLOGY Burnside, E. S., Rubin, D. L., Shachter, R. D., Sohlich, R. E., Sickles, E. A. 2004; 182 (2): 481-488

    Abstract

    We sought to determine whether a probabilistic expert system can provide accurate automated imaging-histologic correlations to aid radiologists in assessing the concordance of mammographic findings with the results of imaging-guided breast biopsies.We created a Bayesian network in which Breast Imaging Reporting and Data System (BI-RADS) descriptors are used to convey the level of suspicion of mammographic abnormalities. Our system is a computer model that links BI-RADS descriptors with diseases of the breast using probabilities derived from the literature. Mammographic findings are used to update pretest probabilities (prevalence of disease) into posttest probabilities applying Bayes' theorem. We evaluated the histologic results of 92 consecutive imaging-guided breast biopsies for concordance with the mammographic findings during radiology-pathology review sessions. First, radiologists with no knowledge of the biopsy results chose BI-RADS descriptors for the mammographic findings. After the histologic diagnosis was revealed, the radiologists assessed concordance between the pathologic results and the mammographic findings. We then input the information gathered from these sessions into the Bayesian network to produce an automated mammographic-histologic correlation.We had a sampling error rate of 1.1% (1/92 biopsies). Our expert system was able to integrate pathologic diagnoses and mammographic findings to obtain probabilities of sampling error, thereby enabling us to identify the incorrect pathologic diagnosis with 100% sensitivity while maintaining a specificity of 91%.Our probabilistic expert system has the potential to help radiologists in identifying breast biopsy results that are discordant with mammographic findings and discovering cases in which biopsy sampling errors may have occurred.

    View details for Web of Science ID 000188590800046

    View details for PubMedID 14736686

  • Improving a Bayesian network's ability to predict the probability of malignancy of microcalcifications on mammography CARS 2004: COMPUTER ASSISTED RADIOLOGY AND SURGERY, PROCEEDINGS Burnside, E. S., Rubin, D. L., Shachter, R. D. 2004; 1268: 1021-1026
  • Using a Bayesian network to predict the probability and type of breast cancer represented by microcalcifications on mammography MEDINFO 2004: PROCEEDINGS OF THE 11TH WORLD CONGRESS ON MEDICAL INFORMATICS, PT 1 AND 2 Burnside, E. S., Rubin, D. L., Shachter, R. D. 2004; 107: 13-17

    Abstract

    Since the widespread adoption of mammographic screening in the 1980's there has been a significant increase in the detection and biopsy of both benign and malignant microcalcifications. Though current practice standards recommend that the positive predictive value (PPV) of breast biopsy should be in the range of 25-40%, there exists significant variability in practice. Microcalcifications, if malignant, can represent either a non-invasive or an invasive form of breast cancer. The distinction is critical because distinct surgical therapies are indicated. Unfortunately, this information is not always available at the time of surgery due to limited sampling at image-guided biopsy. For these reasons we conducted an experiment to determine whether a previously created Bayesian network for mammography could predict the significance of microcalcifications. In this experiment we aim to test whether the system is able to perform two related tasks in this domain: 1) to predict the likelihood that microcalcifications are malignant and 2) to predict the likelihood that a malignancy is invasive to help guide the choice of appropriate surgical therapy.

    View details for Web of Science ID 000226723300003

    View details for PubMedID 15360765

  • Statistics and causal inference: A review - Discussion TEST Fienberg, S. E., Haviland, A. M., Heckerman, D., Shachter, R., Kadane, J. B., Moral, S., Pearl, J. 2003; 12 (2): 319-345
  • Costs and benefits of imperfect HIV vaccines: Implications for vaccine development and use QUANTIATIVE EVALUATION OF HIV PREVENTION PROGRAMS Owens, D. K., Edwards, D. M., Shachter, R. D. 2002: 143-171
  • An approach to automate & individualize interactive decision support for patients AMIA 2002 SYMPOSIUM, PROCEEDINGS Scott, G. C., Shachter, R. D., Lenert, L. A. 2002: 1158-1158
  • Quantifying the population impact of a prophylactic Helicobacter pylori vaccine VACCINE Rupnow, M. F., Shachter, R. D., Owens, D. K., Parsonnet, J. 2001; 20 (5-6): 879-885

    Abstract

    Helicobacter pylori vaccines, which have been suggested as promising interventions to control infection, are under development. We sought to quantify the potential population impact of a prophylactic H. pylori vaccine.We developed a mathematical model that compartmentalized the population according to age, infection status and clinical state. A proportion of individuals was assumed to acquire infection and develop gastritis, duodenal ulcer (DU), chronic atrophic gastritis and gastric cancer (GC). We first simulated the model without vaccine intervention, to obtain estimates of H. pylori prevalence, and GC and DU incidences based on intrinsic dynamics. We then incorporated a prophylactic vaccine (80% efficacy, lifetime protection, 80% coverage) targeting all infants. We tested vaccination programs over unlimited as well as limited time spans. Analyses were performed for the US, Japan and a prototypical developing country.In the US, our model predicted a decrease in H. pylori prevalence from 12.0% in 2010 to 4.2% in 2100 without intervention. With 10 years of vaccination beginning in 2010, prevalence would decrease to 0.7% by year 2100. In the same period, incidence of H. pylori-attributable GC would decrease from 4.5 to 0.4 per 100,000 with vaccine (compared to 1.3 per 100,000 without vaccine). Incidence of H. pylori-attributable DU would decrease from 33.3 to 2.5 per 100,000 with vaccine (compared to 12.2 per 100,000 without vaccine). In Japan, incidence of H. pylori-attributable GC would decrease from 17.6 to 1.0 per 100,000 after 10 years of vaccination (compared to 3.0 per 100,000 without vaccine). In a prototypical developing country, after 10 years of vaccination, H. pylori-attributable GC would decrease from 31.8 to 22.5 per 100,000 by 2090, returning to the original level by mid-2100s. Under continuous vaccination, it would decrease to 5.8 per 100,000 by 2100.In the US and Japan, a 10-year vaccination program would confer almost the same reduction in H. pylori and associated diseases as a vaccination effort that extends beyond 10 years. In developing countries, a continuous vaccination effort would be required to eliminate the pathogen and its associated diseases.

    View details for Web of Science ID 000172871700028

    View details for PubMedID 11738753

  • A dynamic transmission model for predicting trends in Helicobacter pylori and associated diseases in the United States EMERGING INFECTIOUS DISEASES Rupnow, M. F., Shachter, R. D., Owens, D. K., Parsonnet, J. 2000; 6 (3): 228-237

    Abstract

    To assess the benefits of intervention programs against Helicobacter pylori infection, we estimated the baseline curves of its incidence and prevalence. We developed a mathematical (compartmental) model of the intrinsic dynamics of H. pylori, which represents the natural history of infection and disease progression. Our model divided the population according to age, infection status, and clinical state. Case-patients were followed from birth to death. A proportion of the population acquired H. pylori infection and became ill with gastritis, duodenal ulcer, chronic atrophic gastritis, or gastric cancer. We simulated the change in transmissibility consistent with the incidence of gastric cancer and duodenal ulcer over time, as well as current H. pylori prevalence. In the United States, transmissibility of H. pylori has decreased to values so low that, should this trend continue, the organism will disappear from the population without targeted intervention; this process, however, will take more than a century.

    View details for Web of Science ID 000087321300002

    View details for PubMedID 10827112

  • Helicobacter pylori vaccine development and use: A cost-effectiveness analysis using the institute of medicine methodology HELICOBACTER Rupnow, M. F., Owens, D. K., Shachter, R., Parsonnet, J. 1999; 4 (4): 272-280

    Abstract

    Prophylactic vaccination has been suggested as a better strategy than antibiotics to control Helicobacter pylori infection. We evaluated the cost-effectiveness (CE) of H. pylori vaccine development and use in the United States and developing countries, using a method developed by the Institute of Medicine (IOM).The IOM model includes costs of vaccine development, vaccination program, and averted medical treatments; morbidity and mortality prevented; expected efficacy and use; and proportion of disease that is vaccine-preventable. The model employs infant mortality equivalence (IME) to estimate disease burden; with IME, the societal cost of infection-related morbidity is expressed as equivalent to a specific rate of infant deaths. We tested model assumptions by univariate sensitivity analyses.In the United States, H. pylori vaccine would save 1,176 IME and would cost $58.71 million (1997 dollars) annually, yielding a CE ratio of $49,932 per IME; the health benefits would exceed all IOM-studied vaccines, even when efficacy dropped to 55%. H. pylori vaccine could be cost-saving if priced at less than $60 per course. In developing countries, H. pylori vaccine would rank unfavorably both in terms of health benefits (33,518 IME) and costs ($5,254 million). None of the changes in assumptions improved significantly the H. pylori vaccine's ranking relative to other IOM-studied vaccines.Compared to other vaccines evaluated in the IOM study, H. pylori vaccine warrants public resource allocation for accelerated development and use in the United States but not for use in developing countries.

    View details for Web of Science ID 000083763500011

    View details for PubMedID 10597398

  • Efficient value of information computation UNCERTAINTY IN ARTIFICIAL INTELLIGENCE, PROCEEDINGS Shachter, R. D. 1999: 594-601
  • Population effects of preventive and therapeutic HIV vaccines in early- and late-stage epidemics AIDS Owens, D. K., Edwards, D. M., Shachter, R. D. 1998; 12 (9): 1057-1066

    Abstract

    To evaluate the population effects of potential preventive and therapeutic vaccines in early- and late-stage epidemics in a population of homosexual men.An epidemic model was used that simulated the course of the epidemic for a population of homosexual men in San Francisco, California. Vaccine programs were evaluated by the number of cases of HIV averted, the effect on the prevalence of HIV, and by the gain in quality-adjusted life years (QALY) for the total population.In the model, a preventive vaccine prevented 3877 cases of HIV infection during a 20-year period, reduced the projected prevalence of HIV infection from 12 to 7% in a late-stage epidemic, and gained 15,908 QALY. A therapeutic vaccine that did not affect the infectivity of vaccine recipients increased the number of cases of HIV infection by 210, resulted in a slight increase in the prevalence of HIV infection from 12 to 15% in a late-stage epidemic, and gained 8854 QALY. If therapeutic vaccines reduced infectivity, their use could produce net gains of QALY in the population that were similar to gains from the use of preventive vaccines. In an early-stage epidemic, the advantage of a preventive vaccine program relative to a therapeutic vaccine program was markedly enhanced.Both preventive and therapeutic vaccine programs provided substantial benefit, but their relative merit depended on which outcome measures were assessed. Evaluation of HIV vaccine programs based solely on cases averted or on prevalence of HIV in the population underestimates the benefit associated with therapeutic vaccine programs. The effect of a therapeutic HIV vaccine on the epidemic outcomes depended markedly on whether the therapeutic vaccine reduced the infectivity of the vaccine recipient. The relative merits of preventive and therapeutic vaccines depend on the stage of the epidemic. Field vaccine trials should evaluate correlates of infectivity, such as HIV viral load. HIV vaccine implementation strategies should be tailored to the dynamics of the epidemic in specific populations.

    View details for Web of Science ID 000074918500014

    View details for PubMedID 9662203

  • A dynamic HIV-transmission model for evaluating the costs and benefits of vaccine programs INTERFACES Edwards, D. M., Shachter, R. D., Owens, D. K. 1998; 28 (3): 144-166
  • Representation and analysis of medical decision problems with influence diagrams MEDICAL DECISION MAKING Owens, D. K., Shachter, R. D., Nease, R. F. 1997; 17 (3): 241-262

    Abstract

    Influence diagrams are a powerful graphic representation for decision models, complementary to decision trees. Influence diagrams and decision trees are different graphic representations for the same underlying mathematical model and operations. This article describes the elements of an influence diagram, and shows several familiar decision problems represented as decision trees and as influence diagrams. The authors also contrast the information highlighted in each graphic representation, demonstrate how to calculate the expected utilities of decision alternatives modeled with an influence diagram, provide an overview of the conceptual basis of the solution algorithms that have been developed for influence diagrams, discuss the strengths and limitations of influence diagrams relative to decision trees, and describe the mathematical operations that are used to evaluate both decision trees and influence diagrams. They use clinical examples to illustrate the mathematical operations of the influence-diagram-evaluation algorithm; these operations are arc reversal, chance node removal by averaging, and decision node removal by policy determination. Influence diagrams may be helpful when problems have a high degree of conditional independence, when large models are needed, when communication of the probabilistic relationships is important, or when the analysis requires extensive Bayesian updating. The choice of graphic representation should be governed by convenience, and will depend on the problem being analyzed, on the experience of the analyst, and on the background of the consumers of the analysis.

    View details for Web of Science ID A1997XG99200001

    View details for PubMedID 9219185

  • A measure of decision flexibility UNCERTAINTY IN ARTIFICIAL INTELLIGENCE Shachter, R. D., Mandelbaum, M. 1996: 485-491
  • Decision theoretic foundations for causal reasoning JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH Heckerman, D., Shachter, R. 1995; 3: 405-430
  • Decision-Theoretic Foundations for Causal Reasoning. Journal of Artificial Intelligence Research Heckerman, D., Shachter, R., D. 1995; 3: 405-430
  • A PHYSICIAN-BASED ARCHITECTURE FOR THE CONSTRUCTION AND USE OF STATISTICAL-MODELS METHODS OF INFORMATION IN MEDICINE Lehmann, H. P., Shachter, R. D. 1994; 33 (4): 423-432

    Abstract

    Physicians need specially tailored computer tools to take advantage of published research results. We present a knowledge-based computer framework--the physician-based (PB) architecture--for constructing such tools, and we use the problem of physicians' interpretation of two-arm parallel randomized clinical trials (TAPRCT) as a working example. Statistical models are represented by influence diagrams. The interpretation of influence-diagram elements are mapped into users' language in a domain-specific, physician-based user interface, called a patient-flow diagram. Statistical-model transformations that maintain the semantic relationships of the model and that embody clinical-epidemiological knowledge are encoded in a mediating structure called the cohort-state diagram. The algorithm that coordinates the interactions among the knowledge representations uses modular actions called construction steps. This architecture has been implemented in a Bayesian system, called THOMAS, that supports physician decision making in light of TAPRCT data. This support entails assessing clinical significance, prior beliefs, and methodological concerns. We suggest that the PB architecture applies to a wide range of statistical tools and users.

    View details for Web of Science ID A1994PN89900014

    View details for PubMedID 7799819

  • REPRESENTATION OF PREFERENCES IN DECISION-SUPPORT SYSTEMS COMPUTERS AND BIOMEDICAL RESEARCH FARR, B. R., Shachter, R. D. 1992; 25 (4): 324-335

    Abstract

    The recommendations of computer-based decision-support systems depend on the preferences of the expert who is responsible for the decisions. Often, these preferences are only represented implicitly, rather than explicitly, in the system. Decision-theoretic preference models that explicitly represent the preferences of the decision maker provide numerous advantages for decision-support systems. In this paper, we describe these advantages. The creation and refinement of decision-theoretic preference models, however, is a difficult task. We describe an accurate and efficient method for determining the preferences of domain experts and refining the model that captures those preferences. In this preference assessment method, we simulate familiar decisions in the expert's area of expertise. We then infer the preferences of the expert from the choices that the expert makes on the simulated decisions, and use the preference information to refine the model automatically.

    View details for Web of Science ID A1992JH89400002

    View details for PubMedID 1511594

  • Representation of Preferences in Decision Support Systems. Comput Biomed Res Farr, B., R., Shachter, R., D. 1992; 25 (4): 324-335
  • Patient-Specific Explanation in Models of Chronic Disease. AI in Medicine Jimison, H., B., Fagan, L., M., Shachter, R., D., Shortliffe, E., H. 1992; 3 (4): 191-205
  • EVALUATION OF NONLINEAR OPTIMIZATION FOR SCHEDULING OF FOLLOW-UP CYSTOSCOPIES TO DETECT RECURRENT BLADDER-CANCER MEDICAL DECISION MAKING Kent, D. L., NEASE, R. A., Sox, H. C., Shortliffe, L. D., Shachter, R. 1991; 11 (4): 240-248

    Abstract

    Standard recommendations for patients who have had superficial bladder cancer are inspection by cystoscopy quarterly for a year or two after tumor removal, then half-yearly and yearly. The authors assessed the potential for improvement in scheduling cystoscopies according to probabilistic optimization techniques. Eight hypothetical practices were created, based on retrospective analysis of 918 bladder-cancer-patient charts. Standard and alternative recommendations for the interval to next cystoscopy were compared. The alternatives were derived from patient-specific predictions of future tumor risks (based on the patient's prior recurrence rate and tumor stage and grade) and a nonlinear optimization approach to allocation of the same number of cystoscopies as were available for standard follow-up. The optimization proposed longer intervals between visits for low-risk patients and shorter intervals for high-risk patients. Overall, optimization reduced expected tumor detection delays by 30%, from 12.6 to 8.7 weeks. When optimization intervals were shorter than standard, cancer was found more often at subsequent cystoscopies (34% vs 27%, p less than 0.05), suggesting that the optimization was a better predictor of cancer recurrence. If reduction in tumor-detection delay is the goal of follow-up for recurrent cancers, then urologists can improve monitoring by using probabilistic optimization methods for scheduling cystoscopies. Further understanding of the accuracy of predictive models for bladder-cancer recurrence rates is desirable. Subsequently, the optimization method developed here may be tested prospectively.

    View details for Web of Science ID A1991GJ22700002

    View details for PubMedID 1662739

  • FUSION AND PROPAGATION WITH MULTIPLE OBSERVATIONS IN BELIEF NETWORKS ARTIFICIAL INTELLIGENCE Peot, M. A., Shachter, R. D. 1991; 48 (3): 299-318
  • A GRAPH-BASED INFERENCE METHOD FOR CONDITIONAL-INDEPENDENCE UNCERTAINTY IN ARTIFICIAL INTELLIGENCE Shachter, R. D. 1991: 353-360
  • Representation of preferences in decision-support systems. Proceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care FARR, B. R., Schachter, R. D. 1991: 1018-1024

    Abstract

    The recommendations of computer-based decision-support systems depend on the preferences of an expert on which the model is based. Often, these preferences are represented only implicitly, rather than explicitly, in the system. Decision-theoretic preference models that explicitly represent the preferences of the decision maker provide numerous advantages for decision-support systems. In this paper, we describe these advantages. The creation and refinement of decision-theoretic preference models, however, remains a difficult task. We describe an accurate and efficient method for determining the preferences of domain experts and for refining the model that captures those preferences. In this preference-assessment method, we simulate decisions common in the expert's area. We then infer the preferences of the expert from the choices that she makes on the simulated decisions, and use the preference information to refine the model automatically.

    View details for PubMedID 1807565

  • AN ORDERED EXAMINATION OF INFLUENCE DIAGRAMS NETWORKS Shachter, R. D. 1990; 20 (5): 535-563
  • DYNAMIC-PROGRAMMING AND INFLUENCE DIAGRAMS IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS TATMAN, J. A., Shachter, R. D. 1990; 20 (2): 365-379
  • A Bayesian method for synthesizing evidence. The Confidence Profile Method. International journal of technology assessment in health care Eddy, D. M., Hasselblad, V., Shachter, R. 1990; 6 (1): 31-55

    Abstract

    This article describes a collection of meta-analysis techniques based on Bayesian statistics for interpreting, adjusting, and combining evidence to estimate parameters and outcomes important to the assessment of health technologies. The result of an analysis by the Confidence Profile Method is a joint posterior probability distribution for the parameters of interest, from which marginal distributions for any particular parameter can be calculated. The method can be used to analyze problems involving a variety of types of outcomes, a variety of measures of effect, and a variety of experimental designs. This article presents the elements necessary for analysis, including prior distributions, likelihood functions, and specific models for experimental designs that include adjustment for biases.

    View details for PubMedID 2361818

  • AN INTRODUCTION TO A BAYESIAN METHOD FOR META-ANALYSIS - THE CONFIDENCE PROFILE METHOD MEDICAL DECISION MAKING Eddy, D. M., Hasselblad, V., Shachter, R. 1990; 10 (1): 15-23

    Abstract

    The Confidence Profile Method is a new Bayesian method that can be used to assess technologies where the available evidence involves a variety of experimental designs, types of outcomes, and effect measures; a variety of biases; combinations of biases and nested bases; uncertainty about biases; an underlying variability in the parameter of interest; indirect evidence; and technology families. The result of an analysis with the Confidence Profile Method is a posterior distribution for the parameter of interest, posterior distributions for other parameters, and a covariance matrix for all the parameters in the model. The posterior distributions incorporate all the uncertainty the assessor chooses to describe about any of the parameters used in the analysis.

    View details for Web of Science ID A1990CH04300004

    View details for PubMedID 2182960

  • Dynamic Programming and Influence Diagrams. IEEE Transactions on Systems, Man and Cybernetics Tatman, J., A., Shachter, R., D. 1990; 2 (20): 365-379
  • An Ordered Examination of Influence Diagrams. Networks Shachter, R., D. 1990; 20: 535-563
  • GAUSSIAN INFLUENCE DIAGRAMS MANAGEMENT SCIENCE Shachter, R. D., Kenley, C. R. 1989; 35 (5): 527-550
  • EFFICIENT SCHEDULING OF CYSTOSCOPIES IN MONITORING FOR RECURRENT BLADDER-CANCER MEDICAL DECISION MAKING Kent, D. L., Shachter, R., Sox, H. C., HUI, N. S., Shortliffe, L. D., Moynihan, S., Torti, F. M. 1989; 9 (1): 26-37

    Abstract

    Proper timing for repeated evaluations is difficult to assess. The authors analyzed scheduling of cystoscopy to monitor patients for detection of recurrent bladder cancer assuming that 1) minimizing tumor detection delay helps prevent cancer morbidities; 2) only limited numbers of cystoscopies are available; 3) prediction of recurrence or progression to invasive cancer is uncertain; 4) future tumors recur according to a Poisson process. Assumptions 3 and 4 permit estimation of each patient's recurrence rate. Thus, patients may be compared according to their relative risks of future tumors. With these assumptions, nonlinear optimization theory was used to calculate monitoring schedules for a model practice. Given 5.4 available visits per week per 100 patients, cystoscopy was recommended in 9-11 weeks for high-risk patients and in 30-40 weeks for low-risk patients, depending on stages, grades, and numbers of previous tumors. In contrast, standard cystoscopy was recommended in 13, 26, or 52 weeks, depending only on time elapsed since last recurrence. The calculated schedule implied an average detection delay for potentially invasive tumors of eight weeks, while standard practice led to detection delays of 11 weeks (38% worse). These results suggest that inclusion of each patient's tumor history in an optimization approach may improve follow-up care for patients who have superficial bladder cancers. This approach is being evaluated in a larger clinical setting.

    View details for Web of Science ID A1989R671600005

    View details for PubMedID 2643017

  • PROBABILISTIC INFERENCE AND INFLUENCE DIAGRAMS OPERATIONS RESEARCH Shachter, R. D. 1988; 36 (4): 589-604
  • LOCAL COMPUTATIONS WITH PROBABILITIES ON GRAPHICAL STRUCTURES AND THEIR APPLICATION TO EXPERT SYSTEMS - DISCUSSION JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY Kelly, F. P., Hand, D. J., Thatcher, A. R., Smith, J. Q., Critchley, F., Smith, A. F., Hilden, J., Kendall, W. S., Cooper, G. F., Dawid, A. P., Dempster, A. P., Almond, R. G., Dubois, D., Prade, H., Fienberg, S. E., Meyer, M. M., Olesen, K. G., Andersen, S. K., Gammerman, A., Aitken, C. G., Andersen, J. D., BARLOW, R. E., Berzuini, C., Stefanelli, M., Cheeseman, P., Good, I. J., HAVRANEK, T., HENRION, M., Jensen, F. V., Kong, A., Mardia, K. V., McLeish, M., Pearl, J., Phillips, L. D., PORTEOUS, B. T., Pratt, I., Rector, A. L., Shachter, R. D., Shafer, G., Shenoy, P., Smets, P., Thomas, A., Tritchler, D. L., Wermuth, N., Whittaker, J., Lauritzen, S. L., Spiegelhalter, D. J. 1988; 50 (2): 194-224
  • THINKING BACKWARD FOR KNOWLEDGE ACQUISITION AI MAGAZINE Shachter, R. D., HECKERMAN, D. E. 1987; 8 (3): 55-61
  • EVALUATING INFLUENCE DIAGRAMS OPERATIONS RESEARCH Shachter, R. D. 1986; 34 (6): 871-882
  • Evaluating Influence Diagrams. Operations Research Shachter, R., D. 1986; 11-12 (34): 871-882

Books and Book Chapters


  • Pearl Causality and the Value of Control. Heuristics, Probability, and Causality: A Tribute to Judea Pearl Shachter, R., D., Heckerman, D., E. edited by Dechter, R., Geffner, H., Halpern, J., Y. College Publications.. 2010: 431-447
  • Solving influence diagrams: exact algorithms. Wiley Encyclopedia of Operations Research and Management Science. Wiley. Shachter, R., Bhattacharjya, D. edited by Cochran, J., J. 2010
  • Model Building with Belief Networks and Influence Diagrams. Advances in Decision Analysis: From Foundations to Applications Shachter, R., D. edited by Edwards, W., Ralph, J., Miles, F. Cambridge University Press.. 2007: 177-201
  • A Bayesian Network to Assist Mammography Interpretation. Operations Research and Health Care Rubin, D., Burnsie, E., Shachter, R. edited by Brandeau, M., Sainfort, F., Pierskalla, W. Kluwer.. 2004: 695-702
  • The Cost Effectiveness of Partially Effective HIV Vaccines. Operations Research and Health Care Owens, D., Edwards, D., Cavallaro, J., Shachter, R. edited by Brandeau, M., Sainfort, F., Pierskalla, W. Kluwer.. 2004: 403-418
  • A method for the dynamic selection of models under time constraints. Selecting Models from Data: Artificial Intelligence and Statistics Rutledge, G., Shachter, R., D. edited by Cheeseman, P., Oldford, R., W. New York: Springer-Verlag.. 1994: 79-88
  • Meta-Analysis by the Confidence Profile Method: The Statistical Synthesis of Evidence. Eddy, D., M., Hasselblad, V., Shachter, R. Boston: Academic Press.. 1992
  • Directed Reduction Algorithms and Decomposable Graphs. Uncertainty in Artificial Intelligence Shachter, R., D., Andersen, S., K., Poh, K., L. edited by Bonnisone, P., Henrion, M., Kanal, L., N. Amsterdam: North-Holland.. 1991: 197-208
  • Uncertainty in Artificial Intelligence 4. Shachter, R., D., Levitt, T., S., Lemmer, J., F., Kanal, L., N. Amsterdam: North-Holland. 1990
  • Simulation Approaches to General Probabilistic Inference on Belief Networks. Uncertainty in Artificial Intelligence Shachter, R., D., Peot, M. edited by Henrion, M., Shachter, R., D., Lemmer, J., F. Amsterdam: North-Holland.. 1990: 221-230
  • Evidence Absorption and Propagation through Evidence Reversals. Uncertainty in Artificial Intelligence Shachter, R., D. edited by Henrion, M., Shachter, R., D., Lemmer, J., F. Amsterdam: North-Holland.. 1990: 173-190
  • An Influence Diagram Approach to Medical Technology Assessment. Influence Diagrams, Belief Nets, and Decision Analysis Shachter, R., D., Eddy, D., M., Hasselblad, V. edited by Oliver, R., M., Smith, J., Q. Chichester: Wiley. 1990: 321-350
  • A Linear Approximation Method for Probabilistic Inference. Uncertainty in Artificial Intelligence Shachter, R., D. edited by Shachter, R., D., Levitt, T., S., Lemmer, J., F. Amsterdam: North-Holland.. 1990: 93-103
  • Uncertainty in Artificial Intelligence 5. Henrion, M., Shachter, R., D., Lemmer, J., F., Kanal, L., N. Amsterdam: North-Holland. 1990
  • A Heuristic Bayesian Approach to Knowledge Acquisition: Application to the Analysis of Tissue-Type Plasminogen Activator. Uncertainty in Artificial Intelligence Shachter, R., D., Eddy, D., M., Hasselblad, V., Wolpert, R. edited by Kanal, L., N., Levitt, T., S., Lemmer, J., F. Amsterdam: North-Holland.. 1989: 183-190
  • Efficient Inference on Generalized Fault Diagrams. Uncertainty in Artificial Intelligence Shachter, R., D., Bertrand, L., J. edited by Kanal, L., N., Levitt, T., S., Lemmer, J., F. Amsterdam: North-Holland.. 1989: 325-332
  • DAVID: Influence Diagram Processing System for the Macintosh. Uncertainty in Artificial Intelligence Shachter, R., D. edited by Lemmer, J., F., Kanal, L., N. Amsterdam: North-Holland.. 1988: 191-196
  • A Backwards View for Assessment. Uncertainty in Artificial Intelligence Shachter, R., D., Heckerman, D., E. edited by Lemmer, J., F., Kanal, L., N. Amsterdam: North-Holland.. 1988: 317-324
  • Evaluating Influence Diagrams. Reliability and Quality Control Shachter, R., D. edited by Basu, A. Amsterdam: North-Holland.. 1986: 321-344
  • Intelligent Probabilistic Inference. Uncertainty in Artificial Intelligence Shachter, R., D. edited by Kanal, L., N., Lemmer, J., F. Amsterdam: North-Holland.. 1986: 371-382
  • An Incentive Approach to Eliciting Probabilities. Low Probability/High Consequence Risk Analysis Shachter, R., D. New York: Plenum Press. 1983: 137-152

Conference Proceedings


  • Approximate Kalman Filter QLearning for Continuous StateSpace MDPs. Tripp, C., Shachter, R. edited by Nicholson, A., Smyth, P. 2013
  • Backtracking for More Efficient Large Scale Dynamic Programming. Tripp, C., Shachter, R. 2012
  • Strictly Proper Mechanisms with Cooperating Players. Chun, S., Shachter, R. edited by Cozman, F., Pfeffer, A. 2011
  • Dynamic programming in influence diagrams with decision circuits. Shachter, R., Bhattacharjya, D., Grunwald, P., Spirtes, P. 2010
  • Three new sensitivity analysis methods for influence diagrams. Bhattacharjya, D., Shachter, R. edited by Grunwald, P., Spirtes, P. 2010
  • Sensitivity analysis in decision circuits. Bhattacharjya, D., Shachter, R. edited by McAllester, D., Myullymki, P. 2008
  • Evaluating influence diagrams with decision circuits. Bhattacharjya, D., Shachter, R. edited by Parr, R., Gaag, L., van der 2007
  • User-Agent Value Alignment. Shapiro, D., Shachter, R. 2002
  • Using background knowledge to speed reinforcement learning in physical agents. Shapiro, D., Langley, P., Shachter, R. 2001
  • Using decision models to automate & individualize interactive patient-oriented decision support aids Scott, G. C., Shachter, R., Lenert, L. A. BMJ PUBLISHING GROUP. 2001: 1024-1024
  • Second Opinion: A framework for using decision models to automate & individualize interactive patient decision support aids Scott, G. C., Shachter, R., Lenert, L. A. BMJ PUBLISHING GROUP. 2001: 833-833
  • A Bayesian network for mammography Burnside, E., Rubin, D., Shachter, R. HANLEY & BELFUS INC. 2000: 106-110

    Abstract

    The interpretation of a mammogram and decisions based on it involve reasoning and management of uncertainty. The wide variation of training and practice among radiologists results in significant variability in screening performance with attendant cost and efficacy consequences. We have created a Bayesian belief network to integrate the findings on a mammogram, based on the standardized lexicon developed for mammography, the Breast Imaging Reporting And Data System (BI-RADS). Our goal in creating this network is to explore the probabilistic underpinnings of this lexicon as well as standardize mammographic decision-making to the level of expert knowledge.

    View details for Web of Science ID 000170207500023

    View details for PubMedID 11079854

  • Learning about H-pylori transmission dynamics from the distinct patterns of duodenal ulcer and gastric cancer Tsugawa, M. F., Shachter, R., Owens, D. K., Parsonnet, J. W B SAUNDERS CO-ELSEVIER INC. 1999: A338-A338
  • Predicting the next century of H-pylori prevalence and associated diseases in the United States Tsugawa, M. F., Shachter, R., Owens, D. K., Parsonnet, J. W B SAUNDERS CO-ELSEVIER INC. 1999: A339-A339
  • Bayes-Ball: The Rational Pastime (for Determining Irrelevance and Requisite Information in Belief Networks and Influence Diagrams). Shachter, R., D. 1998
  • Learning from What You Don't Observe. Peot, M., A., Shachter, R., D. 1998
  • Decision Flexibility. Chavez, T., Shachter, R., D. 1995
  • A Definition and Graphical Representation for Causality. Heckerman, D., E., Shachter, R., D. 1995
  • Three Approaches to Probability Model Selection. Poland, W., B., Shachter, R., D. 1994
  • A Decision-Based View of Causality. Heckerman, D., E., Shachter, R., D. 1994
  • Laplace's Method Approximations for Probabilistic Inference in Belief Networks with Continuous Variables. Azevedo-Filho, A., Shachter, R., D. 1994
  • Global Conditioning for Probabilistic Inference in Belief Networks. Shachter, R., D., Andersen, S., K., Szolovits, P. 1994
  • End-User Construction of Influence Diagrams for Bayesian Statistics: Lehmann, H., P., Shachter, R., D. 1993
  • A Method for the Dynamic Selection of Models Under Time Constraints Rutledge, G., Shachter, R., D. edited by Cheeseman, P. 1993
  • Using Potential Influence Diagrams for Probabilistic Inference and Decision Making: Shachter, R., D., Ndilikilikesha, P., M. 1993
  • Mixtures of Gaussians and Minimum Relative Entropy Techniques for Modeling Continuous Uncertainties. Poland, W., B., Shachter, R., D. 1993
  • DECISION-MAKING USING PROBABILISTIC INFERENCE METHODS Shachter, R. D., Peot, M. A. MORGAN KAUFMANN PUB INC. 1992: 276-283
  • Structural Controllability and Observability in Influence Diagrams. Chan, B., Y., Shachter, R., D. 1992
  • AN INTRODUCTION TO A BAYESIAN METHOD FOR METAANALYSIS - THE CONFIDENCE PROFILE METHOD Eddy, D. M., Hasselblad, V., Shachter, R. NATL ACADEMY PRESS. 1990: 101-116
  • Directed Reduction Algorithms and Decomposable Graphs. Shachter, R., D., Andersen, S., K., Poh, K., L. 1990
  • Symbolic Probabilistic Inference in Belief Networks. Shachter, R., D., D'Ambrosio, B., Del Favero, B., A. 1990
  • Evidence Absorption and Propagation through Evidence Reversals. Shachter, R., D. 1989
  • Simulation Approaches to General Probabilistic Inference on Belief Networks Shachter, R., D., Peot, M. 1989
  • AN INFLUENCE DIAGRAM APPROACH TO HEALTH TECHNOLOGY-ASSESSMENT WITHIN THE CONFIDENCE PROFILE METHOD Shachter, R. D., Eddy, D. M., Hasselblad, V. HANLEY & BELFUS INC. 1988: 346-346
  • DECISION-THEORY FOR RETROSPECTIVE JUDGEMENTS OF DECISIONS Yu, A., Kent, D. L., HIGGINS, M. C., Mazur, D. J., Sox, H. C., Evans, P. A., Shachter, R. D., Fujimura, I., Howard, R. A. HANLEY & BELFUS INC. 1988: 332-332
  • OPTIMAL SCHEDULING FOR PATIENTS WITH SUPERFICIAL BLADDER-CANCER Kent, D., Nease, R., Sox, H., Shachter, R. SLACK INC. 1988: A341-A341
  • An Influence Diagram Approach to the Confidence Profile Method for Health Technology Assessment. Shachter, R., D., Eddy, D., M., Hasselblad, V. 1988
  • A Linear Approximation Method for Probabilistic Inference. Shachter, R., D. 1988
  • A Heuristic Bayesian Approach to Knowledge Acquisition: Application to Analysis of Tissue-Type Plasminogen Activator. Shachter, R., D., Eddy, D., M., Hasselblad, V. 1987
  • Efficient Inference on Generalized Fault Diagrams. Shachter, R., D., Bertrand, L., J. 1987
  • DAVID: Influence Diagram Processing System for the Macintosh. Shachter, R., D. 1986
  • A Backwards View for Assessment. Shachter, R., D., Heckerman, D., E. 1986
  • Intelligent Probabilistic Inference. Shachter, R., D. 1985