Estimation Of Pretest Probability Research Articles

Could our pretest probabilities become evidence based? A prospective survey of hospital practice.

We sought to measure the proportion of patients on our clinical service who presented with clinical problems for which research evidence was available to inform estimates of pretest probability. We also aimed to discern whether any of this evidence was of sufficient quality that we would want to use it for clinical decision making. Prospective, consecutive case series and literature survey. Inpatient medical service of a university-affiliated Veterans' Affairs hospital in south Texas. Patients admitted during the 3 study months for diagnostic evaluation. Patients' active clinical problems were identified prospectively and recorded at the time of discharge, transfer, or death. We electronically searched medline and hand-searched bibliographies to find citations that reported research evidence about the frequency of underlying diseases that cause these clinical problems. We critically appraised selected citations and ranked them on a hierarchy of evidence. We admitted 122 patients for diagnostic evaluation, in whom we identified 45 different principal clinical problems. For 35 of the 45 problems (78%; 95% confidence interval [95% CI], 66% to 90%), we found citations that qualified as disease probability evidence. Thus, 111 of our 122 patients (91%; 95% CI, 86% to 96%) had clinical problems for which evidence was available in the medical literature. During 3 months on our hospital medicine service, almost all of the patients admitted for diagnostic evaluation had clinical problems for which evidence is available to guide our estimates of pretest probability. If confirmed by others, these data suggest that clinicians' pretest probabilities could become evidence based.

More Data on Spiral CT and Diagnosis of Pulmonary Embolism

Increasingly, spiral CT is incorporated into diagnostic strategies for pulmonary embolism (PE). In this multicenter French study, researchers examined outcomes with the following algorithm: All patients with suspected PE underwent clinical estimation of pretest probability of PE (using a published scheme), spiral CT, and leg ultrasound (US). Patients with negative CT and US and …

Quantitative evaluation of the diagnostic thinking process in medical students.

To explore the diagnostic thinking process of medical students. Two hundred twenty-four medical students were presented with 3 clinical scenarios corresponding to high, low, and intermediate pre-test probability of coronary artery disease. Estimates of test characteristics of the exercise stress test, and pre-test and post-test probability for each scenario were elicited from the students (intuitive estimates) and from the literature (reference estimates). Post-test probabilities were calculated using Bayes' theorem based upon the intuitive estimates (Bayesian estimates of post-test probability) and upon the reference estimates (reference estimates of post-test probability). The differences between the reference estimates and the intuitive estimates, and between Bayesian estimates and the intuitive estimates were used for assessing knowledge of test characteristics, and ability of estimating pre-test and post-test probability of disease. Medical students could not rule out disease in low or intermediate pre-test probability settings, mainly because of poor pre-test estimates of disease probability. They were also easily confused by test results that differed from their anticipated results, probably because of their inaptitude in applying Bayes' theorem to real clinical situations. These diagnostic thinking patterns account for medical students or novice physicians repeating unnecessary examinations. Medical students' diagnostic ability may be enhanced by the following educational strategies: 1) emphasizing the importance of ruling out disease in clinical practice, 2) training in the estimation of pre-test disease probability based upon history and physical examination, and 3) incorporation of the Bayesian probabilistic thinking and its application to real clinical situations.

Clinical Diagnosis of Urinary Tract Infection

To the Editor: In their Rational Clinical Examination article, Dr Bent and colleagues evaluated the accuracy and precision of the clinical history and examination for the diagnosis of urinary tract infection (UTI). Clinicians often rely on a single symptom or sign to establish the diagnosis, so it is important that they be aware of the pretest probability. However, the pretest probability of infection that Bent et al cite (48%) seems very high to us. One of us (T.F.) has calculated a 25% to 29% pretest probability of UTI in patients attending 8 primary care practices in the United Kingdom. We are also aware of another study that found a pretest probability of 35%. We therefore reexamined the studies that Bent et al used, added the study by Dobbs and Fleming, and performed a meta-analysis on only those articles that exclusively studied women with symptoms suggestive of UTI. We excluded studies if they included men or women with suspected vaginal infection. There was no heterogeneity (P=.12), so a fixed-effects model was used. This analysis yielded a combined pretest probability estimate of 35.6% (95% confidence interval, 32.1%-39.0%). We then repeated the posttest calculations by Bent et al based on our pretest probability of 35.6%. The results are shown in the TABLE. These data suggest that the most useful symptoms are the absence of vaginal symptoms, while dysuria and frequency do not markedly increase the posttest probability of UTI. We did not calculate the product of likelihood ratios as Bent et al did because this method assumes the likelihood ratios are independent and can give rise to inflated overall figures. In practice, patients with dysuria are also likely to have urinary frequency; therefore, if likelihood ratios are to be combined, regression methods should be used to establish the independent value of symptoms. In summary, we find little evidence on which to base a reliable pretest probability of UTI. More evidence is needed in unselected populations in primary care. Clinicians frequently prescribe antibiotics on the basis of single symptoms, which do not appear to alter the posttest probabilities sufficiently to rule in or rule out UTI.

Diagnostic Strategies for Common Medical Problems. Edgar R. Black, Donald R. Bordely, Thomas G. Tape, and Robert J. Panzer, eds. Philadelphia: American College of Physicians–American Society of Internal Medicine, 1999, 674 pp., $45.00, Paperback. ISBN 0-943126-74-6.

This is the second edition of a text that seeks to blend information on the operating characteristics of laboratory (and other) tests with clinical estimates of pretest probability to provide a rational basis for the use of technology in clinical decision making. The first edition, published in 1991, grew out of an article entitled “Selection and Interpretation of Diagnostic Tests and Procedures: Principles and Applications”, published as a supplement to the Annals of Internal Medicine (1981;94:553–600) by members of the General Medicine Unit at the University of Rochester. This followed the publication of Beyond Normality: The Predictive Value and Efficiency of Medical Diagnosis by Galen and Gambino (New York: Wiley, 1975), the monograph largely responsible for introducing the notion of “predictive value” to the laboratory community. …

Disease prediction models aim to guide medical decision making.

Current Clinical Issues16 October 2001Disease Prediction Models Aim To Guide Medical Decision MakingTom ReynoldsTom ReynoldsSearch for more papers by this authorAuthor, Article, and Disclosure Informationhttps://doi.org/10.7326/0003-4819-135-8_Part_1-200110160-00023 SectionsAboutFull TextPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinkedInRedditEmail When performing a history and physical examination on a patient with a disorder such as chest pain, physicians try to gauge the likelihood that each differential diagnosis explains the chief complaint. Determining the likelihood of disease as an estimate of probability has important advantages, and this will become increasingly evident as medical information systems, such as those developed for hand-held personal digital assistants (PDAs), help physicians to estimate probability at the moment of decision making.Clinical prediction rules provide a structured approach to calculating a probability estimate by incorporating the key variables that predict whether a disease is present or ... Author, Article, and Disclosure InformationAffiliations: PreviousarticleNextarticle Advertisement FiguresReferencesRelatedDetails Metrics Cited byPEDIS Classification in Diabetic Foot Ulcers PatientsPredicting gastrointestinal infection morbidity based on environmental pollutants: Deep learning versus traditional modelsAn evolutionary deep neural network for predicting morbidity of gastrointestinal infections by food contaminationThe comorbidity of ADHD in children diagnosed with autism spectrum disorderReliability and Validity of the Perfusion, Extent, Depth, Infection and Sensation (PEDIS) Classification System and Score in Patients with Diabetic Foot UlcerValidation and comparison of currently available stratification systems for patients with diabetes by risk of foot ulcer developmentRisk stratification systems for diabetic foot ulcers: a systematic reviewDo Physicians Attend to Base Rates? Prevalence Data and Statistical Discrimination in the Diagnosis of Coronary Heart DiseaseClinical experience did not reduce the variance in physicians' estimates of pretest probability in a cross-sectional surveyProbabilistic reasoning and clinical decision-making: do doctors overestimate diagnostic probabilities? 16 October 2001Volume 135, Issue 8_Part_1Page: 637-640KeywordsBreast cancerCancer preventionCardiac surgeryDrugsIschemiaMedical risk factorsMuscular dystrophiesMyocardial infarctionStable coronary artery diseaseVascular surgery ePublished: 16 October 2001 Issue Published: 16 October 2001 Copyright & PermissionsCopyright © 2001 by American College of Physicians. All Rights Reserved.PDF downloadLoading ...

Effect of a patient's psychiatric history on physicians' estimation of probability of disease.

A questionnaire was mailed to 300 Iowa family physicians to determine the influence of a prior psychiatric history on decision making. The response rate was 77%. Respondents were less likely to believe that a patient had serious illness when presenting with a severe headache or abdominal pain if the patient had a prior history of depression ( P <.05) or prior history of somatic complaints ( P <. 05), compared with a patient with no past history. Respondents were less likely to report that they would order testing for a patient with headache or abdominal pain if the patient had a history of depression ( P <.05, P =.08, respectively) or somatic complaints ( P <.01). Differences in likelihood of ordering tests were not significant after adjusting for differences in estimated probability of disease. We conclude that physicians respond differently to patients with psychiatric illness because of their estimation of pretest probability of disease rather than bias. We conclude that past psychiatric history influences physicians' estimation of disease presence and willingness to order tests.

Bayesian Modeling of Muscle Biopsy Contracture Testing for Malignant Hyperthermia Susceptibility

Background: Phenotyping malignant hyperthermia (MH) by contracture testing has a low but quantifiable degree of inaccuracy, measured by its sensitivity and specificity. Quantifying the limitations inherent in diagnostic testing for MH can help resolve issues in clinical practice, such as the interpretation of a negative test and the apparent lack of complete genetic linkage to RYR1. Methods : Bayesian models, mathematical descriptions of the outcome of diagnostic testing, were constructed. The inputs to the model include patient factors, summarized in a single number called pretest probability (PTP), and sensitivity and specificity that specify the accuracy of the entire test process. The outputs of the model include positive predictive value (PPV) and negative predictive value (NPV), which are numeric expressions of diagnostic certainty of positive and negative test results. A special case was constructed for equivocal results. Results: The PPV, NPV, and efficiency of contracture testing for MH are functions of PTP, sensitivity, and specificity. The NPV is high for all clinical PTP, whereas PPV is clinically useful for moderate to high PTP. Conclusions: Diagnostic contracture testing for MH is clinically useful because of high NPV and can exclude MH with near certainty. For MH probands, the clinical grading scale for MH may guide PTP estimation, whereas for relatives of probands, PTP is a function of kinship to a known MH-susceptible relative. A sequential testing strategy optimizes diagnostic information by maximizing PTP within a pedigree. Incomplete testing of parents of an MH susceptible child can pose a significant risk of false-negative results for the untested parent. Even with optimal pedigree testing strategies, the PPV drift effect results in a considerable source of phenotypic uncertainty for genetic linkage studies.

Rational diagnosis of pulmonary embolism (RADIA PE) in symptomatic outpatients with suspected PE: an improved strategy to exclude or diagnose venous thromboembolism by the sequential use of a clinical model, rapid ELISA D-dimer test, perfusion lung scan, ultrasonography, spiral CT, and pulmonary angiography.

A prospective management decision analysis for the exclusion and diagnosis of pulmonary embolism (PE) based on pre-test clinical probability (PCP) estimation for PE, a rapid ELISA D-dimer test, perfusion lungscan (P-scan), CUS, spiral CT, and pulmonary angiography is proposed. The modified PCP model for PE of Wells et al. allows reasonably accurate classification of patients with no, low, moderate, and high probability for PE. The combined rational use of the evidence-based noninvasive imaging techniques P-scan, CUS, and spiral CT with the rapid ELISA D-dimer test and PCP will reduce the need for invasive pulmonary angiography to perhaps 10 to 15% of patients, who initially presented with suspected PE. A Rational Diagnosis of Pulmonary Embolism (RADIA PE) model is proposed for testing in a large multicenter study of patients with suspected PE.

Rational diagnosis of deep vein thrombosis (RADIA DVT) in symptomatic outpatients with suspected DVT: simplification and improvement of decision rule analysis for the exclusion and diagnosis of DVT by the combined use of a simple clinical model, a rapid sensitive D-dimer test and compression ultrasonography (CUS).

The diagnosis of deep vein thrombosis (DVT) in outpatients is difficult to establish. The classical clinical signs of DVT are unspecific and found in several other conditions besides DVT. Therefore, these patients are subjected to elaborate, expensive invasive or noninvasive procedures that actually confirm DVT in only 20-30% of patients in this setting. A prospective management decision study, based on a pretest clinical probability (PCP) estimation, compression ultrasonography (CUS), and a rapid enzyme-linked immunoadsorbent assay (ELISA) D-dimer test, is proposed. The PCP model of Wells allows for reasonably accurate classification of patients into low, moderate and high probability for suffering DVT. Combined with CUS, high predictability is achieved. The D-dimer assays presently available can be rapidly performed and are accurate to a high degree, especially in ruling out DVT. It is proposed to establish a "Rational Diagnosis of Deep Vein Thrombosis (RADIA DVT)" model to be tested in a large number of patients with suspected DVT. This model would be less expensive to perform and likely yield a highly accurate diagnosis on the basis of which anticoagulant therapy could be initiated or withheld.

Bayesian modeling of muscle biopsy contracture testing for malignant hyperthermia susceptibility.

Phenotyping malignant hyperthermia (MH) by contracture testing has a low but quantifiable degree of inaccuracy, measured by its sensitivity and specificity. Quantifying the limitations inherent in diagnostic testing for MH can help resolve issues in clinical practice, such as the interpretation of a negative test and the apparent lack of complete genetic linkage to RYR1. Bayesian models, mathematical descriptions of the outcome of diagnostic testing, were constructed. The inputs to the model include patient factors, summarized in a single number called pretest probability (PTP), and sensitivity and specificity that specify the accuracy of the entire test process. The outputs of the model include positive predictive value (PPV) and negative predictive value (NPV), which are numeric expressions of diagnostic certainty of positive and negative test results. A special case was constructed for equivocal results. The PPV, NPV, and efficiency of contracture testing for MH are functions of PTP, sensitivity, and specificity. The NPV is high for all clinical PTP, whereas PPV is clinically useful for moderate to high PTP. Diagnostic contracture testing for MH is clinically useful because of high NPV and can exclude MH with near certainty. For MH probands, the clinical grading scale for MH may guide PTP estimation, whereas for relatives of probands, PTP is a function of kinship to a known MH-susceptible relative. A sequential testing strategy optimizes diagnostic information by maximizing PTP within a pedigree. Incomplete testing of parents of an MH susceptible child can pose a significant risk of false-negative results for the untested parent. Even with optimal pedigree testing strategies, the PPV drift effect results in a considerable source of phenotypic uncertainty for genetic linkage studies.

Radiologists' Interpretations: The Problem of Context Bias

To the Editor. —Drs Egglin and Feinstein 1 provided us with a new and commendable insight with regard to the way the diagnostic judgments of radiologists are modified by the perceptual context in which test results are presented to them. The method and arguments they presented in the article convinced me that our thought process in front of an imaging study follows Bayesian rules. The radiologists in their study appeared to use the early cases in the series they had to read as a pilot group: their concrete perceptions and influences were used to calibrate their readings for the rest of the cases in the series. We would like to believe that in our daily practice we also use pretest probability estimates. However, we infer the pretest probability from risk factors and the results of tests (clinical or otherwise) that are presented to us by a narrative of the clinical

Post-Test Probability of Asthma Following Methacholine Challenge

Using recently published data, a nomogram was constructed to estimate the likelihood of asthma following methacholine challenge. Based on Bayes' theorem, the nomogram makes use of the sensitivity and specificity of methacholine challenge to calculate the post-test probability of asthma once the physician makes a determination of the pretest probability, that is, the likelihood of asthma before the test results are considered. A family of curves is presented to cover several levels of cumulative breath units at which the test could become positive, and a single curve is presented for a negative test after 224 cumulative breath units. Separate curves are presented for smokers and nonsmokers. The estimate of pretest probability is most crucial in negative tests where likelihood of asthma is considered high, and in positive tests in patients in whom asthma is considered unlikely. Although these curves will not apply precisely to a different data base, the concept of the relationship between pretest and post-test probability helps in the interpretation of the test results and stresses the importance of using all available information in making a diagnosis.

Physician tolerance for uncertainty. Use of liver-spleen scans to detect metastases.

To gain insight into diagnostic test use, we interviewed 42 physicians who ordered 62 liver-spleen scans to detect possible metastases. Before receiving the actual scan results, physicians gave (1) pretest probability estimates for liver metastasis; (2) probability estimates and management plans given the most common scan results; and (3) probabilities above and below which they would stop the diagnostic workup (their decision thresholds). After learning the actual scan results, physicians gave their posttest probabilities and management plans. Thirteen physicians (31%) desired absolute certainty to rule in metastasis. Fifty percent planned further evaluations even when probability estimates of metastasis reached one of their probability targets for ending the workup. Greater attention to decision thresholds may improve physicians' understanding of their tolerance for diagnostic uncertainty and the ability of diagnostic tests to reduce it.

Physician Tolerance for Uncertainty

To gain insight into diagnostic test use, we interviewed 42 physicians who ordered 62 liver-spleen scans to detect possible metastases. Before receiving the actual scan results, physicians gave (1) pretest probability estimates for liver metastasis; (2) probability estimates and management plans given the most common scan results; and (3) probabilities above and below which they would stop the diagnostic workup (their decision thresholds). After learning the actual scan results, physicians gave their posttest probabilities and management plans. Thirteen physicians (31%) desired absolute certainty to rule in metastasis. Fifty percent planned further evaluations even when probability estimates of metastasis reached one of their probability targets for ending the workup. Greater attention to decision thresholds may improve physicians' understanding of their tolerance for diagnostic uncertainty and the ability of diagnostic tests to reduce it.