-
[show abstract]
[hide abstract]
ABSTRACT: Auditing has received much attention recently as a method for radiologists to use to evaluate their interpretation of screening mammograms. U.S. Food and Drug Administration regulations require that some sort of audit be in place before a mammography screening facility can receive accreditation. Auditing presents a unique opportunity to monitor accuracy continually and identify problems early. Audit data present unique challenges, however, and appropriate methods must be used to control the risk of errors.
This article introduces a simple method for the task of deciding if a radiologist yields an acceptable positive predictive value based on audit. The method is based on "sequential" decision-making techniques that have found wide application in quality control problems. These techniques are developed for diagnostic radiology and embodied in an easy-to-use decision-making chart.
Several examples, based on audit data from actual mammography facilities, provide insights into the use of these charts and the influence of (a) the selection of standards, (b) the selection of error risks, and (c) radiologist variability. The examples also serve to demonstrate another important property of this method--that is, it specifies the minimum amount of data that has to be collected before any decision can reliably be made.
The chart presented in this article provides a method by which audit data can be used objectively to evaluate the accuracy of screening mammogram interpretation. The method controls the risk of either falsely accepting an unqualified radiologist or falsely rejecting a qualified radiologist. It should be a useful tool to radiologists who must evaluate their own practices.
Academic Radiology 05/1999; 6(4):216-23. · 1.69 Impact Factor
-
C A Beam
Academic Radiology 02/1999; 6 Suppl 1:S81-2. · 1.69 Impact Factor
-
C A Beam
[show abstract]
[hide abstract]
ABSTRACT: Clustered data is not simply correlated data, but has its own unique aspects. In this paper, various methods for correlated receiver operating characteristic (ROC) curve data that have been extended specifically to clustered data are reviewed. For those methods that have not yet been extended, suggestions for their application to clustered ROC studies are provided. Various methods with respect to their ability to meet either of two objectives of the analysis of clustered ROC data are compared to consider a variety of ROC indices and their accessibility to researchers. The available statistical methods for clustered data vary in the range of indices that can be considered and in their accessibility to researchers. Parametric models permit all indices to be considered but, owing to computational complexity, are the least accessible of available methods. Nonparametric methods are much more accessible, but only permit estimation and inference about ROC curve area. The jackknife method is the most accessible and permits any index to be considered. Future development of methods for clustered ROC studies should consider the continuation ratio model, which will permit the application of widely available software for the analysis of mixed generalized linear models. Another area of development should be in the adoption of bootstrapping methods to clustered ROC data.
Statistical Methods in Medical Research 01/1999; 7(4):324-36. · 2.44 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: To demonstrate the range of gains and losses that radiologists might experience from independent double reading in screening mammography.
From a national random sample of radiologists, the authors formed 131 pairs. For each radiologist, the authors analyzed the increase relative to his or her individual true-positive rate (TPR) or false-positive rate (FPR), number of additional cancers detected, and change to negative biopsy rate that would result from independent double reading after pairing.
The average radiologist can expect an 8%-14% gain in TPR and a 4%-10% increase in FPR with pairing. For some radiologists, double reading increased the TPR with a small concomitant increase in FPR. Other radiologists, however, realized small gains in TPR with large increases in FPR. Adding the reading from a more experienced radiologist did not necessarily improve the TPR of a radiologist with less experience.
Radiologists can form complementary and noncomplementary pairs for double reading. Use of this procedure must be decided on an individual basis.
Academic Radiology 12/1996; 3(11):891-7. · 1.69 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: To evaluate the effectiveness of screening mammography by estimating the variability in radiologists' ability to detect breast cancer within the US population of radiologists at mammography centers accredited by the American College of Radiology.
A two-way sample survey design was used as follows. Fifty mammography centers having an American College of Radiology-accredited unit were randomly sampled from across the United States. One hundred eight radiologists from these centers gave blinded interpretation to the same set of 79 randomly selected screening mammograms. The mammograms were from women who had been screened at a large screening center. Before their sampling, these women had been stratified by their breast disease status, established either by biopsy or by 2-year follow-up. Rates of biopsy recommendations were summarized by the mean, median, minimum, maximum, and range of sensitivity and specificity. Overall cancer detection ability was summarized by similar statistics for receiver operating characteristic curve areas. Ninety-five percent lower confidence bounds on the ranges in accuracy measures were established by boo-strapping.
There is a range of at least 40% among US radiologists in their screening sensitivity. There is a range of at least 45% in the rates at which women without breast cancer are recommended for biopsy. As indicated by receiver operating characteristic curve areas, the ability of radiologists to detect cancer mammograms varies by as much as 11%.
Our findings indicate that there is wide variability in the accuracy of mammogram interpretation in the population of US radiologists. Current accreditation programs that certify the technical quality of radiographic equipment and images but not the accuracy of the interpretation given to mammograms may not be sufficient to help mammography fully realize its potential to reduce breast cancer mortality.
Archives of Internal Medicine 02/1996; 156(2):209-13. · 11.46 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Seriously or terminally ill patients are frequently incapacitated and unable to express their preferences regarding cardiopulmonary resuscitation (CPR). In this situation, family members or other surrogate decision makers are often asked whether they believe the patient would want to be resuscitated. We evaluated the concordance of patient CPR preferences and surrogate perceptions of the patient preferences in the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT), a large, multicenter study of seriously ill hospitalized patients.
We compared patient preferences and surrogate perceptions in 1226 pairings in which both patient and surrogate responded to CPR decision questions. We also examined factors that might influence patient-surrogate concordance.
Twenty-nine percent of patients with paired data did not want to be resuscitated; 26% of surrogates did not believe the patient they represented would want to be resuscitated. Within pairs, the overall exact agreement with respect to CPR decisions was 74%. For patients favoring CPR, only 16% of the surrogates misconstrued the patient's wishes. For patients who did not want to be resuscitated, however, 50% of the surrogates did not reflect the patient's wishes. If patients reported telling surrogates their CPR preference, concordance was significantly improved if the surrogate believed the patient did not want to be resuscitated and was significantly worsened if the surrogate believed the patient wanted CPR. This finding is likely an artifact of patients being more likely to report their preference to surrogates if that preference was not to be resuscitated.
Surrogates' perceptions of patient CPR preferences are often inaccurate, particularly for those patients who do not want to be resuscitated. Methods to improve communication between patients and surrogates on CPR preferences should be developed and evaluated.
Archives of Family Medicine 07/1995; 4(6):518-23.
-
C A Beam
Academic Radiology 04/1995; 2 Suppl 1:S4-13; discussion S57-60, S68-9 pas. · 1.69 Impact Factor
-
Radiology 12/1994; 193(2):582. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Preliminary reports have described the use of MR imaging for the detection of deep venous thrombosis. However, no prospective study comparing MR imaging with contrast venography (the gold standard) has been reported. Accordingly, we performed a prospective, blinded study of the efficacy of MR imaging in 61 consecutive patients with clinically suspected deep venous thrombosis. In cases of disagreement, additional testing was performed to determine the diagnosis.
From June 1991 to February 1992, 61 patients with clinically suspected deep venous thrombosis were examined with venography and MR imaging. The average time between studies was 3 hr. In 21 of the 61 patients, the final diagnosis was deep venous thrombosis.
For detection of deep venous thrombosis in the pelvis, the sensitivity of MR imaging was 100% (9/9) with a 95% confidence interval of 72-100% and the specificity was 95% (52/55) with a 95% confidence interval of 85-99%. In the thigh, the sensitivity (16/16) and specificity (43/43) were both 100% with 95% confidence intervals of 83-100% and 93-100%, respectively. In the calf, the sensitivity was 87% (13/15) with a 95% confidence interval of 60-98% and the specificity was 97% (36/37) with a 95% confidence interval of 86-100%.
We found no statistically significant difference between MR imaging and contrast venography in the detection of deep venous thrombosis. This result suggests that MR imaging is at least as sensitive and specific as contrast venography in the detection of deep venous thrombosis.
American Journal of Roentgenology 08/1993; 161(1):131-9. · 2.78 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: One promising approach to flow quantification uses the velocity-dependent phase change of moving protons. A velocity-encoding phase subtraction technique was used to measure the velocity and flow rate of fluid flow in a phantom and blood flow in volunteers.
In a model, the authors measured constant flow velocities from 0.1 to 270.0 cm/second with an accuracy (95% confidence intervals) of +/- 12.5 cm/second. There was a linear relationship between the magnetic resonance imaging (MRI) measurement and the actual value (r2 = .99; P = .0001).
Measuring mean pulsatile flow from 125 to 1,900 mL/minute, the accuracy of the MRI pulsatile flow measurements (95% confidence intervals) was +/- 70 mL/minute. There was a linear relationship between the MRI pulsatile flow measurement and the actual value (r2 = .99; P = .0001). In 10 normal volunteers, the authors tested the technique in vivo, quantitating flow rates in the pulmonary artery and the aorta. The average difference between the two measurements was 5%. In vivo carotid flow waveforms obtained with MRI agreed well with the shape of corresponding ultrasound Doppler waveforms.
Velocity-encoding phase subtraction MRI bears potential clinical use for the evaluation of blood flow. Potential applications would be in the determination of arterial blood flow to parenchymal organs, the detection and quantification of intra- and extra-cardiac shunts, and the rapid determination of cardiac output and stroke volume.
Investigative Radiology 03/1993; 28(2):109-15. · 4.59 Impact Factor
-
C A Beam
[show abstract]
[hide abstract]
ABSTRACT: Research studies in diagnostic radiology often compare the diagnostic abilities of two imaging techniques. The "power" of such studies is the probability that they will detect a difference in abilities of a certain amount when, indeed, such a difference does exist. This article outlines several strategies that can be used to assess and improve the power of radiologic diagnostic studies. These strategies include selection of cases and controls, matching, use of one-tailed tests, selection of significance level, and choice of sample size.
American Journal of Roentgenology 10/1992; 159(3):631-7. · 2.78 Impact Factor
-
Radiology 07/1992; 183(3):619-20. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Retrospective analysis of 36 embolization procedures in 29 patients with gastrointestinal bleeding was undertaken, and the presence or absence of coagulopathy was identified as a major factor affecting embolization outcome. Embolization was successful in 18 of 29 (62%) patients and unsuccessful in 11 (38%). Eight of 11 failures (73%) occurred in patients with a coagulopathy, whereas three patients (27%) in whom embolization was successful also had a coagulopathy. Embolization was 2.9 times more likely to be unsuccessful (P = .0463) and death from bleeding after embolization was 9.6 times more likely to occur (P = .0065) in patients with a coagulopathy than in those without. Because embolization was successful in six of 14 (43%) coagulopathy patients, the authors advocate embolization in patients with gastrointestinal bleeding and coagulopathy, while all efforts to correct the coagulopathy would be made as early as possible.
Radiology 06/1992; 183(2):505-8. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Images of a phantom obtained with computed radiography and standard screen-film imaging were compared to evaluate observer threshold perception performance with a modified contrast-detail technique. Optimum exposure necessary for performance with the imaging plate technique to match that with screen-film techniques was determined, as was comparative performance with variation in kilovoltages, plate type, spatial enhancement, and hard-copy interpolation method. It was found that computed radiography necessitates about 75%-100% more exposure than screen-film radiography to optimally match performance with Ortho-C film with Lanex regular or medium screens (Eastman Kodak, Rochester, NY) for detection of objects 0.05-2.0 cm in diameter. However, only minimal loss of detection performance (approximately 10% overall) was experienced if standard screen-film exposures were used with computed radiography. Little change in observer performance was found with variation in plate type, spatial enhancement, or method of hard-copy interpolation. However, perception performance with computed radiographic images was better at lower kilovoltages.
Radiology 05/1992; 183(1):179-87. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: In a retrospective study, the authors compared results in 81 patients who underwent 117 percutaneous transluminal angioplasty (PTA) procedures in the iliac arteries and 21 patients who underwent iliac artery atherectomy. Both groups were similar with respect to symptoms and risk factors. The overall technical success rates were 97.4% for PTA and 100% for atherectomy. The initial clinical success was 92.8% for PTA and 100% for atherectomy. Complications requiring surgery occurred in four of 117 (3.4%) PTA and one of 21 (4.8%) atherectomy procedures. The 2-year patency rate for PTA was similar to that achieved with atherectomy in a smaller patient population. These results indicate that atherectomy does not offer any definite advantage over conventional PTA for the routine management of iliac lesions and make it difficult to justify the additional expense and complexity of this technique.
Journal of Vascular and Interventional Radiology 03/1992; 3(1):67-71. · 2.08 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Using a 1.5 T magnetic resonance imaging (MRI) system, cine phase-contrast and magnitude images were obtained in three phantoms that simulated different anatomic configurations of aortic dissection. The dissection phantoms were made of compliant materials, and pulsatile flow was used in all experiments. Phantoms differed only in the location of the fenestration between the true and false lumens (I: an upstream "entry" only, II: both upstream "entry" and downstream "re-entry," and III: a downstream "entry" only). Flow jets, flap motion, and wave propagation were clearly visualized in cine MR images of each phantom, and quantitatively analyzed with reference to the stimulated cardiac cycle of the pump. Flow in the false lumen was always bidirectional. Upstream and downstream flow waves collided and dispersed within the false lumen. Flow through the false lumen was the same in phantoms I and II, and least in phantom III. The average area of the true lumen was largest in phantom III and smallest in I. Phantom I had the highest overall flow rate in the false lumen and greatest change in false lumen size during the cardiac cycle, while the downstream "entry" phantom had the lowest of both parameters. Flow phenomena in aortic dissections can be studied by cine phase-contrast MRI.
Investigative Radiology 01/1992; 26(12):1071-8. · 4.59 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: The discomfort that patients experience during mammography is probably related to a variety of factors, one of which might be the amount of compression used. The authors measured the amount of force applied to the breasts during mammography and the resulting breast thickness in 560 women and correlated these measurements with the patient's subjective impression of the examination. The amount of force applied ranged from 49 to 186.2 N (median, 122.5 N). Breast thickness ranged from 10 to 88 mm (median, 46.5 mm). Forty-seven women (8%) rated the examination as painful (ie, mammography was either "very uncomfortable" or "intolerable"). Logistic regression analysis revealed a highly significant relationship between probability of a painful response and ratio of force to thickness (P = .007). Current guidelines suggest that maximum available force be at least 160 N but not more than 200 N. However, because increasing force is associated with increasing likelihood of pain, technologists should be aware that these recommended maximum limits are not intended to be used routinely for all women. In this study, high-quality mammograms were obtained in a majority of women with forces less than the maximum available level.
Radiology 12/1991; 181(2):355-7. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: In this paper we study a statistic that is suitable for comparing a discrete diagnostic marker to one or more continuous diagnostic markers. Test procedures and confidence intervals are based on asymptotic normality. The statistic is applicable for correlated data in which all the markers are obtained for each subject. The statistic was studied for use in comparing two markers for rectal bleeding. Examples for this application and two more general applications are presented.
Biometrics 10/1991; 47(3):907-19. · 1.83 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: The authors modified the methods used by Cooper et al (JAMA 1988; 259:3277-3280) to describe magnetic resonance (MR) clinical research during 1985-1988 in an objective manner. To avoid other shortcomings faced by Cooper et al, the authors randomly sampled 500 articles and obtained estimates of probable errors. Excluding case reports, all 40 of the studies with human subjects provided some evidence of prior research planning, which compares favorably with the 22% found by Cooper et al. Use of a standard for truth was similar between the data sets. Observer blinding, random ordering of procedures, and assessment of interobserver variability continued to be uncommon. Most studies provided only minimal statistical information. The work of these authors provides a study design and the baseline needed for future assessments of MR imaging evaluations.
Radiology 08/1991; 180(1):265-70. · 5.73 Impact Factor
-
[show abstract]
[hide abstract]
ABSTRACT: Detection of abnormal catecholamine levels and localization of tumor mass are important factors in the diagnosis and treatment of pheochromocytoma. Iodine 131-labeled metaiodobenzylguanidine scintigraphy was performed in 64 patients with suspected pheochromocytoma if their urinary catecholamine levels were borderline or elevated, or if the clinical suspicion for pheochromocytoma was high in spite of normal urinary catecholamine determinations. The 131I-metaiodobenzylguanidine scans were evaluated for abnormal localization of tracer. Twenty-four-hour urine collections were analyzed for vanillylmandelic acid, homovanillic acid, dopamine, epinephrine, and norepinephrine. Thirty of the 64 patients had pheochromocytomas. The 131I-metaiodobenzylguanidine scan had a sensitivity and a specificity of 88%. The 24-hour urine vanillylmandelic acid and norepinephrine measurements had the best sensitivity (97%), while the vanillylmandelic acid and homovanillic acid measurements had the best specificity (91%). In patients in whom the vanillylmandelic acid measurement and the 131I-metaiodobenzylguanidine scan were normal, no pheochromocytomas were found. In patients in whom the vanillylmandelic acid measurement and 131I-metaiodobenzylguanidine scan were abnormal, a pheochromocytoma was always present. The 131I-metaiodobenzylguanidine scan often documents the presence or absence of a pheochromocytoma and provides localization of the tumor in the preoperative evaluation of these patients.
Archives of Internal Medicine 08/1991; 151(7):1397-402. · 11.46 Impact Factor
