FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2066295280> ?p ?o ?g. }

• W2066295280 endingPage "237" @default.
• W2066295280 startingPage "233" @default.
• W2066295280 abstract "In George Orwell's 1936 essay “Shooting an Elephant,” he describes the daunting task, required of him as a young police officer stationed in colonial Burma, of bringing down a rogue bull elephant with a small-caliber Winchester. In this issue of Annals, Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar have taken aim at a no less formidable target, one that conceptually approximates the magnitude and scale of Orwell's: the derivation of a clinical decision rule predicting short-term outcome among patients presenting to the emergency department (ED) with undifferentiated syncope.Stiell and Wells2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar have proposed rigorous methodologic standards for the development of clinical decision rules in emergency medicine. Extending and modifying earlier work by Wasson et al3Wasson JH Sox HC Neff RK et al.Clinical prediction rules: application and methodological standards.N Engl J Med. 1985; 313: 793-799Crossref PubMed Scopus (952) Google Scholar and Feinstein,4Feinstein AR Clinimetrics. Yale University Press, New Haven, CT1987Crossref Google Scholar these authors have suggested 8 criteria against which the methodologic quality of the “derivation phase” in the development of a clinical decision rule can be measured. These include: (1) definition of the main target outcome variable; (2) definition of each predictor variable; (3) assessment of the reliability of each predictor variable; (4) description of patient selection; (5) justification of sample size; (6) reporting of mathematical techniques; (7) assessment of the “sensibility” of the rule; and (8) analysis of the accuracy of the rule's performance in the derivation set.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar By holding the article by Quinn et al2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar up to the bright light of the exacting specifications outlined by Stiell and Wells,2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar we can develop a systematic, point-by-point critical appraisal of the methodology used in the derivation of the San Francisco Syncope Rule.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar1. Definition of outcomeThe outcome of a decision rule should be clinically important, clearly defined, and blindly assessed.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar define “serious outcomes” among ED patients with syncope or near syncope as: (1) death; (2) myocardial infarction (elevated troponin or ECG changes, and confirmation of diagnosis by cardiology); (3) arrhythmia (documented on monitoring and thought to be temporally linked to the syncopal episode); (4) pulmonary embolism (requiring treatment or confirmation at autopsy); (5) stroke; (6) subarachnoid hemorrhage; (7) significant hemorrhage (requiring transfusion); (8) other conditions for which an acute intervention was performed (surgery for valvular heart disease, abdominal aortic aneurysm, ruptured spleen, or ectopic pregnancy; or insertion of pacemaker or balloon pump; or use of pressors; or endoscopic treatment of esophageal varices); or (9) any condition among patients discharged from the ED requiring a return ED visit and subsequent hospitalization. Two independent reviewers, blinded to the presence or absence of potential predictor variables, classified 1-week outcomes dichotomously as serious or nonserious on the basis of these definitions.In contrast to earlier5Kapoor WN Karpf M Wieand S et al.A prospective evaluation and follow-up of patients with syncope.N Engl J Med. 1983; 309: 197-204Crossref PubMed Scopus (703) Google Scholar and more recent6Colivicchi F Ammirati F Melina D et al.Development and prospective validation of a risk stratification system for patients with syncope in the emergency department: the OESIL risk score.Eur Heart J. 2003; 24: 811-819Crossref PubMed Scopus (297) Google Scholar work identifying predictors of adverse events within the year after a syncopal episode, Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar have focused on short-term (7-day) outcomes. In support of this choice, they argue persuasively that such a time frame is more relevant to emergency practice because immediate hospitalization may offer an opportunity to prevent those unfavorable outcomes destined to occur shortly after the sentinel event.Although the definition of the target outcome at which the decision rule is aimed appears to meet Stiell and Well's2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar criteria of clinical importance, clarity, and blinded review, readers may object that the list of outcomes, although extensive, is not exhaustive. Because syncope is associated with such a panoply of adverse events, triggered by diverse etiologies, many of which are exceedingly rare (and therefore will be under-represented even in a large dataset), it seems unreasonable to expect that any working definition of serious outcomes could encompass all conceivable possibilities without becoming clinically unwieldy.2. Definition of predictor variablesThe predictor variables for a decision rule should be clearly defined and gathered prospectively with appropriate blinding.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Combining a review of the literature with expert consensus, these authors identified 50 well-defined candidate predictor variables, composed of historical features, physical findings, laboratory results, and ECG abnormalities.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google ScholarInformation on each predictor variable was gathered prospectively, using a standardized data collection instrument designed specifically for this purpose, as recommended by Stiell and Wells.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Whether or not the physicians who completed the data form were blinded to the investigators' definition of serious outcomes is unclear.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar However, at the data acquisition stage of a derivation study, before identification of those predictor variables destined to become “risk factors,” it seems unlikely that a lack of blinding to outcome would interject significant recording bias into the dataset.3. Reliability of predictor variablesThe predictor variables used in a decision rule should possess good interobserver agreement.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar In this study, interobserver agreement measured the reliability (reproducibility) of historical information, physical findings, and ECG interpretations obtained by 2 different clinicians on the same patient. Consistent with Stiell and Wells' recommendations,2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar chose to use the κ coefficient, which corrects for agreement attributable to chance.7Kramer MS Feinstein AR Clinical biostatistics: LIV. The biostatistics of concordance.Clin Pharmacol Ther. 1982; 29: 111-123Crossref Scopus (951) Google Scholar A subset of 265 patients (39% of the derivation cohort) was independently evaluated by 2 physicians. Those predictors with κ of 0.5 or less were deemed too unreliable for retention in the final model and were discarded.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google ScholarTo place this in context, a κ greater than 0.80 indicate “almost perfect” interobserver agreement, 0.60 to 0.80 is consistent with “substantial” interobserver agreement, and 0.40 to 0.60 is considered to reflect “moderate” interobserver agreement.8Landis JR Koch GG The measurement of observer agreement for categorical data.Biometrics. 1977; 33: 159-174Crossref PubMed Scopus (49765) Google Scholar A κ equal to 0 indicates agreement no better than that expected by chance alone. Although the cut point of 0.5 requires slightly less interobserver agreement than the κ of greater than 0.6 used in the derivation of the Ottawa Ankle Rule,9Stiell IG Greenberg GH McKnight RD et al.A study to develop clinical decision rules for the use of radiography in acute ankle injuries.Ann Emerg Med. 1992; 21: 384-390Abstract Full Text PDF PubMed Scopus (494) Google Scholar the relatively small number of patients with serious outcomes in this dataset1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar is likely to produce unbalanced marginal totals in the 2×2 table from which the κ coefficient is calculated. This in turn may lead to a falsely depressed κ resulting from overcorrection for chance.10Feinstein AR Cicchetti DV High agreement but low kappa: I. The problems of two paradoxes.J Clin Epidemiol. 1990; 43: 543-549Abstract Full Text PDF PubMed Scopus (2031) Google Scholar, 11Cicchetti DV Feinstein AR High agreement but low kappa: II. Resolving the paradoxes.J Clin Epidemiol. 1990; 43: 551-558Abstract Full Text PDF PubMed Scopus (1293) Google Scholar For this reason, it is helpful when investigators report simple agreement (as a percentage, uncorrected for chance), along with the κ coefficient, thus allowing interested readers to determine whether the statistical quirks of κ may be painting an unduly harsh picture of interobserver concordance.4. Selection of patientsStudy setting, inclusion/exclusion criteria, clinical characteristics, and demographic characteristics of the derivation cohort should be clearly described.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Quinn et al,1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar working in a university teaching hospital, employed research assistants to identify patients eligible for study entry. Those with head trauma, definite seizure, loss of consciousness associated with substance abuse (including alcohol), or altered mental status were excluded. Relevant demographic and some clinical characteristics of the derivation cohort are reported in Table 1.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar Tables 2 and 3 provide more detailed information on historical features, physical findings, laboratory results, and ECG abnormalities.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google ScholarThe attending physician made the final decision to enroll eligible patients. To the extent that the same physician would then be required to complete the rather extensive data collection instrument, this circumstance might provide an opportunity for selection bias. Although use of the ED attending physician as the final arbiter of study entry may have been necessary for any number of practical reasons, this nonetheless runs the theoretical risk of systematically excluding sicker syncope patients. This might occur simply because the physician caring for such a patient may have less time to devote to completing the requisite paperwork and thus choose not to enroll an otherwise eligible patient.5. Sample sizeThe sample size should take into account the degree of precision with which the rule's anticipated performance will be reported, and should be appropriate to the type of multivariate analysis chosen to develop the rule.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar Because the precision of a confidence interval (CI) is inversely related to its width, the narrower the interval surrounding any point estimate, the greater the precision of that estimate. Assuming a serious outcome rate of 10%, the authors estimated they would need 62 patients with serious outcomes in order to have a 95% CI with a width of 15%.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar Because the target sensitivity is not stated, it is not possible to replicate their sample size calculation with certainty. If we assume that the target sensitivity was 100% (ie, all 62 patients with serious outcomes were expected to be correctly identified by the decision rule), then the 95% CI would range from 94% to 100%, a width of only 6%. Alternatively, if we assume the target sensitivity to be 95%, which seems a more realistic expectation for an entity as heterogeneous as syncope, the 95% CI then ranges from about 86% to 99%, a width much closer to the target of 15%.In retrospect, none of this is critically important, because the actual sensitivity of the derived rule turned out to be 96%, bounded by a 95% CI with a precision better than the goal the authors set for themselves. This is true whether the width of the CI is 8%, as reported by Quinn et al,1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar using “bootstrap estimates,” or 10%, calculated using the exact method for proportions.12Gardner MJ Altman DG Statistics With Confidence: Confidence Intervals and Statistical Guidelines. British Medical Journal, London, United Kingdom1989Google ScholarOf potential concern is the problem of “overfitting” of the dataset. Overfitting may prevent a decision rule that performs well in the derivation stage from successful validation when prospectively applied to an independent dataset.13Charlson ME Ales KL Simon R et al.Why predictive indexes perform less well in validation studies.Arch Intern Med. 1987; 147: 2155-2161Crossref PubMed Scopus (179) Google Scholar It tends to occur when the chosen mathematical model is freighted with an excess of predictor variables relative to the number of target outcome events. In this instance, 79 serious outcome variables may be insufficient to support 26 candidate predictor variables without risking overfitting.14Harrell Jr, FE Lee KL Matchar DB et al.Regression models for prognostic prediction: advantages, problems, and suggested solutions.Cancer Treatment Reports. 1985; 69: 1071-1077PubMed Google Scholar Stiell and Wells2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar recommend that the multivariate model from which the rule is to be derived contain at least 10 outcome events of interest for each predictor variable entered into the model. This recommendation is based on findings drawn from early14Harrell Jr, FE Lee KL Matchar DB et al.Regression models for prognostic prediction: advantages, problems, and suggested solutions.Cancer Treatment Reports. 1985; 69: 1071-1077PubMed Google Scholar and recently confirmed15Peduzzi P Concato J Kemper E et al.A simulation study of the number of events per variable in logistic regression analysis.J Clin Epidemiol. 1996; 49: 1373-1379Abstract Full Text PDF PubMed Scopus (4883) Google Scholar empiric simulation with logistic regression. The “rule of 10” is thought to offer a rough approximation of the amount of information contained in a given dataset. This should then logically place a ceiling on the degrees of freedom one can expend in analytic exploration of that dataset. Whether the concept of overfitting can be extrapolated from the algebraic modeling of logistic regression (whose output is probability prediction) to the targeted-cluster methods of recursive partitioning (whose output is classification) is unknown. Although Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar appeared to have “violated” the rule of 10, because overfitting has not been empirically shown to cause the same sorts of problems with recursive partitioning that it does in logistic modeling, this can only be noted at this time as a hypothetically important methodologic issue. Fortunately, this hypothesis will be tested as the rule moves forward to validation.6. Mathematical techniquesThe mathematical model chosen to derive the decision rule should be clearly described and appropriate to the intended clinical use of the rule.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar The authors reduced 50 potential predictor variables to 26 through a combination of univariate screening and application of a reliability threshold of a κ coefficient greater than 0.5.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar These 26 variables were entered into a χ2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar recursive partitioning analysis, using an excellent program known as KnowledgeSEEKER (KnowledgeStudio, Angoss Software International, Toronto, Ontario, Canada), which allows interaction between the investigators and the software at each step of the analytic process. The use of interactive manual override is appropriate to ensure that the recursive partitioning is not blindly driven by statistical considerations alone, without regard for the clinical “sensibility” of the predictor variables selected for inclusion in the final model.As shown in the Figure,1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar this multivariate method produces a tree-like structure, consisting of predictor variables displayed as branches or nodes, selected because they provide the best sequential classification of patients with serious outcomes. As the partitioning proceeds down the tree, each branch represents the next predictor variable that best identifies serious outcomes among the remaining subset of patients occupying the branch immediately above.The use of univariate selection to identify a reduced number of predictor variables eligible for entry into a multivariate model has been challenged.16Sun GW Shook TL Kay GL Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis.J Clin Epidemiol. 1996; 49: 907-916Abstract Full Text PDF PubMed Scopus (610) Google Scholar However, the technique used by Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar has been used previously to screen variables for selection in several well-validated decision rules.17Stiell IG Greenberg GH McKnight RD et al.Decision rules for the use of radiography in acute ankle injuries: refinement and prospective validation.JAMA. 1993; 269: 1127-1132Crossref PubMed Scopus (294) Google Scholar, 18Stiell IG Greenberg GH Wells GA et al.Prospective validation of a decision rule for the use of radiography in acute knee injuries.JAMA. 1996; 275: 611-615Crossref PubMed Google ScholarBecause the authors were interested in the derivation of a decision rule that optimized sensitivity rather than one targeted at overall accuracy, they correctly chose recursive partitioning as a mathematical strategy in preference to logistic regression. Logistic models identify a group of independent predictor variables that minimize misclassification of patient outcomes, thus maximizing accuracy. However, recursive partitioning has the advantage of aiming toward the correct classification, not of all patients (maximal accuracy), but of either all patients with serious outcomes (maximal sensitivity) or all patients with nonserious outcomes (maximal specificity). For a decision rule on syncope to be clinically useful, it is more important that it maximize sensitivity than accuracy. This is because the ultimate goal of the rule is to identify those patients at risk of serious short-term outcomes, on the theory that impending adverse events might be averted or mitigated by prompt hospitalization.7. SensibilityThe rule should be clinically reasonable, easy to use, and suggest a course of action.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar The 5 variables that Quinn et al1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar derived to predict short-term serious outcomes possess face and content validity, are unambiguous, and should be readily obtainable within a time frame consistent with the pace of emergency practice. These predictors are relatively easy to remember with the help of the authors' suggested mnemonic of CHESS (Congestive heart failure, Hematocrit <30%, ECG abnormal [non-sinus rhythm or new changes compared to previous ECG], Shortness of breath, or Systolic blood pressure <90 mm Hg).1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar By extrapolation, the rule suggests that hospitalization should be considered for any patient with at least 1 of these 5 “risk factors.”8. AccuracyAuthors should report the performance characteristics of their decision rule in the derivation dataset (sensitivity, specificity, and positive and negative likelihood ratios), and provide an estimate of its impact.2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar The test properties of the rule are clearly displayed in Table 4, accompanied by the 2×2 matrix from which they were derived: sensitivity equal to 96% (95% CI 89% to 99%); specificity equal to 62% (95% CI 58% to 66%); negative likelihood ratio equal to 0.06 (95% CI 0.02 to 0.19); and positive likelihood ratio equal to 2.5 (95% CI 2.3 to 2.8).1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar With the exception of the precision of the rule's sensitivity, these estimates are identical to those reported by Quinn et al.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar Using the exact method,12Gardner MJ Altman DG Statistics With Confidence: Confidence Intervals and Statistical Guidelines. British Medical Journal, London, United Kingdom1989Google Scholar the lower limit of the CI for sensitivity is 89%, in contrast to the 92% reported by the authors using bootstrap estimates.1Quinn JV Stiell IG McDermott D et al.Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes.Ann Emerg Med. 2003; 43: 224-232Abstract Full Text Full Text PDF Scopus (316) Google Scholar Although this difference of 3% is probably not clinically significant, it is noteworthy that the sensitivity of the rule has a lower limit to its CI somewhere in the neighborhood of 90%.By applying the decision rule to the derivation cohort, the authors determined that 45% of the patients had at least 1 of the 5 predictors for short-term adverse outcome and should therefore have been considered for admission. Because the actual admission rate in this cohort was 55%, the authors estimate that application of the rule to these same patients would have reduced the proportion of those hospitalized by approximately 10%.Summary critical appraisalOn the basis of this appraisal, the derivation of the San Francisco Syncope Rule appears to have met virtually all of Stiell and Wells' exacting standards,2Stiell IG Wells GA Methodologic standards for the development of clinical decision rules in emergency medicine.Ann Emerg Med. 1999; 33: 437-447Abstract Full Text Full Text PDF PubMed Scopus (423) Google Scholar with 2 possible exceptions: (1) questionable selection bias, which may have resulted in failure to enroll some of the sicker, otherwise eligible syncope patients (see “Selection of Patients,” above); and (2) possible “overfitting” of the dataset due to an insufficient number of serious outcomes necessary to support the number of candidate predictor variables entered into the multivariate model (see “Sample Size,” above). The first of these concerns is relatively minor, and the second is largely theoretical. Neither seems sufficiently critical to preclude moving forward to determine whether either of these methodologic considerations has any practical importance. This would be accomplished through prospective validation of this newly derived decision rule on an independent dataset.Implications for clinical practice: none at the present timeBecause this decision rule is imperfectly sensitive (96%), even if it validates with similar test properties (rules rarely validate with better performance characteristics than those reported at derivation13Charlson ME Ales KL Simon R et al.Why predictive indexes perform less well in validation studies.Arch Intern Med. 1987; 147: 2155-2161Crossref PubMed" @default.
• W2066295280 created "2016-06-24" @default.
• W2066295280 creator A5045866941 @default.
• W2066295280 date "2004-02-01" @default.
• W2066295280 modified "2023-09-27" @default.
• W2066295280 title "Shooting an Elephant" @default.
• W2066295280 cites W1983897914 @default.
• W2066295280 cites W2012301002 @default.
• W2066295280 cites W2014659929 @default.
• W2066295280 cites W2015416871 @default.
• W2066295280 cites W2037668591 @default.
• W2066295280 cites W2043080042 @default.
• W2066295280 cites W2056699838 @default.
• W2066295280 cites W2060847368 @default.
• W2066295280 cites W2060932845 @default.
• W2066295280 cites W2076061140 @default.
• W2066295280 cites W2086481317 @default.
• W2066295280 cites W2087524811 @default.
• W2066295280 cites W2098512198 @default.
• W2066295280 cites W2114763047 @default.
• W2066295280 cites W2164777277 @default.
• W2066295280 cites W2338589661 @default.
• W2066295280 cites W96751186 @default.
• W2066295280 doi "https://doi.org/10.1016/j.annemergmed.2003.09.014" @default.
• W2066295280 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/14747813" @default.
• W2066295280 hasPublicationYear "2004" @default.
• W2066295280 type Work @default.
• W2066295280 sameAs 2066295280 @default.
• W2066295280 citedByCount "5" @default.
• W2066295280 countsByYear W20662952802020 @default.
• W2066295280 crossrefType "journal-article" @default.
• W2066295280 hasAuthorship W2066295280A5045866941 @default.
• W2066295280 hasBestOaLocation W20662952801 @default.
• W2066295280 hasConcept C545542383 @default.
• W2066295280 hasConcept C71924100 @default.
• W2066295280 hasConceptScore W2066295280C545542383 @default.
• W2066295280 hasConceptScore W2066295280C71924100 @default.
• W2066295280 hasIssue "2" @default.
• W2066295280 hasLocation W20662952801 @default.
• W2066295280 hasLocation W20662952802 @default.
• W2066295280 hasOpenAccess W2066295280 @default.
• W2066295280 hasPrimaryLocation W20662952801 @default.
• W2066295280 hasRelatedWork W1506200166 @default.
• W2066295280 hasRelatedWork W1995515455 @default.
• W2066295280 hasRelatedWork W2039318446 @default.
• W2066295280 hasRelatedWork W2048182022 @default.
• W2066295280 hasRelatedWork W2080531066 @default.
• W2066295280 hasRelatedWork W2604872355 @default.
• W2066295280 hasRelatedWork W2748952813 @default.
• W2066295280 hasRelatedWork W2899084033 @default.
• W2066295280 hasRelatedWork W3032375762 @default.
• W2066295280 hasRelatedWork W3108674512 @default.
• W2066295280 hasVolume "43" @default.
• W2066295280 isParatext "false" @default.
• W2066295280 isRetracted "false" @default.
• W2066295280 magId "2066295280" @default.
• W2066295280 workType "article" @default.