Ready-made, recalibrated, or Remodeled? Issues in the use of risk indexes for assessing mortality after coronary artery bypass graft surgery.

BACKGROUND: Risk indexes for operative mortality after cardiac surgery are used for comparative profiling of surgeons or centers. We examined whether clinicians and managers should use an existing index without modification, recalibrate it for their populations, or derive a new model altogether. METHODS AND RESULTS: Drawing on 7491 consecutive patients who underwent isolated CABG at 2 Toronto teaching hospitals between 1993 and 1996, we compared 3 strategies: (1) using a ready-made model originally derived and validated in our jurisdiction; (2) recalibrating the ready-made model to better fit the population; and (3) deriving a new model with additional risk factors. We assessed statistical accuracy, ie, area under a receiver-operator characteristic curve (ROC); precision, ie, statistical goodness-of-fit; and actual impact on both risk-adjusted operative mortalities (RAOM) and performance rankings for 14 surgeons. The new model was slightly more accurate than the ready-made model (ROC, 0.78 versus 0.76; P<0.05), albeit not different from the recalibrated model (ROC, 0.77). The ready-made model showed poor fit between the predicted and observed results (P<0.001), leading to significant underestimation of RAOM (1.6+/-0. 2%) compared with the other strategies (2.5+/-0.2%; P=0.048). Remodeling also changed the performance rankings among half the surgeons with higher RAOM. CONCLUSIONS: Poorly calibrated risk algorithms can bias the calculation of RAOM and alter the results of surgeon-specific profiles. Any existing index used for risk assessment in cardiac surgery should be episodically recalibrated or compared with new models derived from local subjects to ensure that its performance remains optimal.

Breast cancer patients' attitudes about rationing postlumpectomy radiation therapy: applicability of trade-off methods to policy-making.

PURPOSE: Along with evidence, clinical policies must take patients' values into account. Particularly where evidence is limited and where assumptions of utility-maximizing behavior may not be valid, new methods such as trade-off techniques (TOTs), which allow elicitation of patients' treatment alternatives, might be useful in policy formulation. We used TOTs to assess breast cancer patients' attitudes toward two clinical policies designed to ration adjuvant postlumpectomy breast radiation therapy. METHODS: Cross-sectional interviews were performed in a tertiary cancer center. A total of 102 patients were presented with information about the side effects and benefits associated with two hypothetical decisions: (1) willingness to receive treatment elsewhere to shorten the wait for radiation therapy, and (2) foregoing radiation therapy in the face of small marginal benefits. For each scenario, a TOT was used to identify the maximal acceptable wait time (MAWT) for therapy and the benefit threshold at which the patient would forego therapy. Associations of clinical and demographic factors with these decisions were determined by regression analysis. RESULTS: Patients would be willing to wait, on average, 7 weeks before wanting to leave their city for radiation therapy, less than the 13-week delay our patients actually faced. Older patients were less willing to wait (P = .013); 46% of patients would not give up radiation therapy, even in the face of no stated benefit. Willingness to give up radiation therapy was predicted by willingness to accept delay (odds ratio [OR], 1.84; 95% confidence interval [CI], 1.05 to 3.37) and being employed (OR, 2.61; 95% CI, 1.08 to 6.54). Patients with larger tumors were less willing to give up radiation therapy (OR, 0.57; 95% CI, 0.31 to 0.97). CONCLUSION: Even in difficult decisions such as rationing postlumpectomy breast cancer radiation therapy, TOTs can inform policy formulation by indicating the distributions of patients' preferences.

Population-wide mortality trends among patients hospitalized for acute myocardial infarction: the Ontario experience, 1981 to 1991.

OBJECTIVES: This study attempted to determine population-based trends in in-hospital patient fatality from acute myocardial infarction. BACKGROUND: The in-hospital prognosis for patients with acute myocardial infarction should be improving as a result of adoption of treatments proved in randomized trials (e.g., thrombolytic, beta-adrenergic blocking and anticoagulant agents and aspirin). However, all trials are subject to selection biases, eligibility is limited for some therapies, and proved therapies may be underused even among eligible patients. METHODS: Using administrative data from all general hospitals in Ontario, Canada, we analyzed 17,489, 17,839, 18,393, 18,794, 18,716 and 19,748 records of patients with a primary discharge diagnosis of myocardial infarction for fiscal years 1981, 1983, 1985, 1987, 1989 and 1991, respectively. RESULTS: After age and gender adjustment, the overall relative reduction in in-hospital case fatality rates for the 10-year period was 26.9% (99% confidence interval [CI] 26.8% to 26.9%), corresponding to an absolute reduction of 6% (99% CI 5.6% to 6.4%). Age- and gender-standardized case fatality rate decreased from 22.3% in 1981 to 21.4% in 1985, followed by a highly significant decline to 16.3% in 1991. On the basis of the relation of comparative mortality to days of hospital stay, declining mortality was not an artifact of decreasing length of stay. CONCLUSIONS: There have been encouraging improvements in survival after acute myocardial infarction over the past 6 years. Further improvements may require development of new therapies that can be more widely applied to this patient population.

Assessing the outcomes of coronary artery bypass graft surgery: how many risk factors are enough? Steering Committee of the Cardiac Care Network of Ontario.

OBJECTIVES: We sought to determine whether more comprehensive risk-adjustment models have a significant impact on hospital risk-adjusted mortality rates after coronary artery bypass graft surgery (CABG) in Ontario, Canada. BACKGROUND: The Working Group Panel on the Collaborative CABG Database Project has categorized 44 clinical variables into 7 core, 13 level 1 and 24 level 2 variables, to reflect their relative importance in determining short-term mortality after CABG. METHODS: Using clinical data for all 5,517 patients undergoing isolated CABG in Ontario in 1993, we developed 12 increasingly comprehensive risk-adjustment models using logistic regression analysis of 6 of the Panel's core variables and 6 of the Panel's level 1 variables. We studied how the risk-adjusted mortality rates of the nine cardiac surgery hospitals in Ontario changed as more variables were included in these models. RESULTS: Incorporating six of the core variables in a risk-adjustment model led to a model with an area under the receiver operating characteristic (ROC) curve of 0.77. The ROC curve area slightly improved to 0.79 with the inclusion of six additional level 1 variables (p = 0.063). Hospital risk-adjusted mortality rates and relative rankings stabilized after adjusting for six core variables. Adding an additional six level 1 variables to a risk-adjustment model had minimal impact on overall results. CONCLUSIONS: A small number of core variables appear to be sufficient for fairly comparing risk-adjusted mortality rates after CABG across hospitals in Ontario. For efficient interprovider comparisons, risk-adjustment models for CABG could be simplified so that only essential variables are included in these models.

Biology or bias: practice patterns and long-term outcomes for men and women with acute myocardial infarction.

OBJECTIVES: The goal of our study was to examine how age and gender affect the use of coronary angiography and the intensity of cardiac follow-up care within the first year after acute myocardial infarction (AMI). Another objective was to evaluate the association of age, gender and treatment intensity with five-year survival after AMI. BACKGROUND: Utilization rates of specialized cardiac services inversely correlate with age. Gender-specific practice patterns may also vary with age in a manner similar to known age-gender survival differences after AMI. METHODS: Using linked population-based administrative data, we examined the association of age and gender with treatment intensity and long-term survival among 25,697 patients hospitalized with AMI in Ontario between April 1, 1992, and December 31, 1993. A Cox proportional hazards model was used to adjust for socioeconomic status, illness severity, attending physician specialty and admitting hospital characteristics. RESULTS: After adjusting for baseline differences, the relative rates of angiography and follow-up specialist care for women relative to men, respectively, fell 17.5% (95% confidence interval [CI], 13.6 to 21.3, p < 0.001) and 10.2% (95% CI, 7.1 to 13.2, p < 0.001) for every 10-year increase in age. Conversely, long-term AMI survival rates in women relative to men improved with increasing age, such that the relative survival in women rose 14.2% (95% CI, 10.1 to 17.5, p < 0.001) for every 10-year age increase. CONCLUSIONS: Gender differences in the intensity of invasive testing and follow-up care are strongly age-specific. While care becomes progressively less aggressive among older women relative to men, survival advantages track in the opposite direction, with older women clearly favored. These findings suggest that biology is likely to remain the main determinant of long-term survival after AMI for women.

Characteristics and mortality outcomes of thrombolysis trial participants and nonparticipants: a population-based comparison.

OBJECTIVES: This study was done to compare characteristics and outcomes of patients with acute myocardial infarction participating in two thrombolysis trials with those of nontrial patients at study hospitals and external hospitals. BACKGROUND: Preferential recruitment of lower risk patients into randomized trials of thrombolysis has been suggested by earlier studies. However, to date there has not been a definitive population-based comparison of characteristics and outcomes for thrombolysis trial participants and nonparticipants. METHODS: Population-based data on hospital admissions and mortality from acute myocardial infarction for all hospitals in Ontario from 1989 to 1992 were linked to data on trial participants in two distinct thrombolysis studies (GUSTO I and LATE). Included were 1,304 patients entered into GUSTO, 12,657 nonparticipants at GUSTO hospitals, 249 patients entered into LATE, 5,997 nonparticipants at LATE hospitals and 12,299 patients at external hospitals. The main outcomes were differences in age, gender, comorbidity scores, coronary revascularization and survival to hospital discharge. RESULTS: Patients in both GUSTO and LATE were significantly more likely to be <70 years old (odds ratio [OR] 2.8 and 3.2, respectively), to be male (OR 2.0 and 2.1, respectively), to have low comorbidity scores (OR 2.0 and 2.3, respectively) and, for GUSTO alone, to undergo coronary revascularization (OR 2.4). Nontrial patients were similar between trial hospitals and external hospitals. In-hospital mortality rates for GUSTO and LATE patients were lower (6.9% and 6.6%, respectively) than for nonparticipants at study hospitals (16.8% and 19.7%, respectively; p<0.001 for both comparisons). Survival to hospital discharge remained higher among GUSTO (OR 1.9) and LATE patients (OR 2.0) than nonparticipants at study hospitals even after adjustment for age, gender, revascularization and comorbidity scores. CONCLUSIONS: Compared with nontrial patients, thrombolysis trial participants are younger, more often male, undergo more revascularization and have less comorbid disease. Even after adjustment for these factors, participants have a survival advantage over nonparticipants that is larger than expected from thrombolysis alone. These findings are not attributable to inferior care or skewed populations at hospitals that did not join these major trials. Further study of these selection biases may guide future trial design and deepen our understanding of why thrombolytics have been underused for high risk patients in routine practice.

Atenolol use and clinical outcomes after thrombolysis for acute myocardial infarction: the GUSTO-I experience. Global Utilization of Streptokinase and TPA (alteplase) for Occluded Coronary Arteries.

OBJECTIVES: We assessed the use and effects of acute intravenous and later oral atenolol treatment in a prospectively planned post hoc analysis of the GUSTO-I dataset. BACKGROUND: Early intravenous beta blockade is generally recommended after myocardial infarction, especially for patients with tachycardia and/or hypertension and those without heart failure. METHODS: Besides one of four thrombolytic strategies, patients without hypotension, bradycardia or signs of heart failure were to receive atenolol 5 mg intravenously as soon as possible, another 5 mg intravenously 10 min later and 50 to 100 mg orally daily during hospitalization. We compared the 30-day mortality of patients given no atenolol (n=10,073), any atenolol (n=30,771), any intravenous atenolol (n=18,200), only oral atenolol (n=12,545) and both intravenous and oral drug (n=16,406), after controlling for baseline differences and for early deaths (before oral atenolol could be given). RESULTS: Patients given any atenolol had a lower baseline risk than those not given atenolol. Adjusted 30-day mortality was significantly lower in atenolol-treated patients, but patients treated with intravenous and oral atenolol treatment vs. oral treatment alone were more likely to die (odds ratio, 1.3; 95% confidence interval, 1.0 to 1.5; p=0.02). Subgroups had similar rates of stroke, intracranial hemorrhage and reinfarction, but intravenous atenolol use was associated with more heart failure, shock, recurrent ischemia and pacemaker use than oral atenolol use. CONCLUSIONS: Although atenolol appears to improve outcomes after thrombolysis for myocardial infarction, early intravenous atenolol seems of limited value. The best approach for most patients may be to begin oral atenolol once stable.

Triiodothyronine augmentation in the treatment of refractory depression. A meta-analysis.

BACKGROUND: Several trials have addressed the efficacy of liothyronine sodium therapy in euthyroid, nonpsychotic depressed patients refractory to tricyclic antidepressant therapy. We undertook a meta-analysis of these trials. METHODS: The MEDLINE database (1966 to May 1995) and published reference lists were examined for controlled clinical trials of triiodothyronine augmentation in euthyroid patients with refractory depression. Quality assessment and data abstraction were performed independently by two reviewers. Results were aggregated three ways: the relative response rate compared with controls, accepting each trial's definition of clinical response; absolute improvement in response rates; and improvements in depression scores, analyzed as continuous variables without a prespecified threshold for clinical response. RESULTS: Aggregating eight studies with a total of 292 patients, patients treated with triiodothyronine augmentation were twice as likely to respond as controls (relative response, 2.09; 95% confidence interval [CI], 1.31 to 3.32; P = .002). This corresponded to a 23.2% absolute improvement in response rates (95% CI, 4.5% to 41.9%; P = .02). Improvements in depression scores were moderately large (standardized effect size, 0.62; P < .001). However, study quality was uneven, and results were statistically heterogeneous. Among the four randomized double-blind studies, pooled effects were not significant (relative response, 1.53; 95% CI, 0.70 to 3.35; P = .29), but one study with negative results accounted for most of the intertrial heterogeneity in results. CONCLUSIONS: Triiodothyronine augmentation may be an effective empirical method of increasing response rates and decreasing depression severity scores in a subgroup of patients with depression refractory to tricyclic antidepressant therapy, but the total number of patients randomized was small, and additional placebo-controlled data are required for a definitive verdict. Since therapeutic trends now favor other drugs, future trials might usefully examine triiodothyronine augmentation with selective serotonin reuptake inhibitors or compare potentiation strategies, eg, lithium vs triiodothyronine, for managing refractory depression. Such trials would benefit from much larger sample sizes than those reviewed here.

Diagnosing gestational diabetes mellitus. Is the gold standard valid?

In North America, gestational diabetes mellitus (GDM) is diagnosed from a 100-g oral glucose tolerance test (OGTT) with criteria proposed by the National Diabetes Data Group (NDDG). These criteria were derived in the 1950s from an unrepresentative sample of women tested predominantly in the latter stages of pregnancy. The original studies did not check for reproducibility of OGTT results. Measurements were made with the Somogyi-Nelson whole-blood glucose technique, and test translation errors are present in the threshold values proposed for modern plasma glucose oxidase methods. Whereas GDM is now diagnosed with a view to adverse maternal-fetal outcomes, the criteria were chosen to reflect maternal risk of developing glucose intolerance as shown by a 75-g OGTT in the nonpregnant state over the ensuing 8 yr. For that outcome, the positive predictive value of the criteria was only 36.1%, and this is a marked overestimate, because the study cohort was a highly selected group with an increased incidence of glucose intolerance both during and after pregnancy. The criteria are also conceptually flawed in that they impose a dichotomous definition of normal and abnormal on gestational glucose tolerance, when the risk of adverse maternal-fetal outcomes and later diabetes mellitus should logically be graded upward with higher values on the gestational OGTT and with the degree of fasting hyperglycemia. Although NDDG criteria merit continued use for lack of a better alternative, new diagnostic criteria for GDM should be derived and validated.

The Toronto Tri-Hospital Gestational Diabetes Project. A preliminary review.

In this study, we assessed maternal-fetal outcomes in untreated patients with increasing carbohydrate intolerance not meeting the current criteria for the diagnosis of gestational diabetes mellitus (GDM), examined the relationship between birth weight and mode of delivery among women with untreated borderline GDM, treated overt GDM, and normoglycemia, and established more efficient screening strategies for detection of GDM. This was a prospective analytic cohort study in which nondiabetic women aged > or = 24 years were eligible for enrollment. A 50-g glucose challenge test (GCT) and a 100-g oral glucose tolerance test (OGTT) were administered at 26 and 28 weeks gestational age, respectively. Risk factors for unfavorable maternal-fetal outcomes were recorded. Time since the last meal prior to the screening test was recorded, as well. Caregivers and patients were blinded to glucose values except when test results met the National Diabetes Data Group criteria for GDM. Maternal and fetal outcomes, including the mode of the delivery, were recorded in the postpartum period. Of 4,274 patients screened, 3,836 (90%) continued to the diagnostic oral glucose tolerance test. GDM was seen in 145 women. Increasing carbohydrate intolerance in women without overt gestational diabetes was associated with a significantly increased incidence of cesarean section, preeclampsia, macrosomia, and need for phototherapy, as well as an increased length of maternal and neonatal hospital stay. Multivariate analysis showed that increasing carbohydrate intolerance remained an independent predictor for various unfavorable outcomes, but the strength of the associations was diminished. Compared with normoglycemic control subjects, the untreated borderline GDM group had increased rates of macrosomia (28.7 vs. 13.7%, P < 0.001) and cesarean delivery (29.6 vs. 20.2%, P = 0.03). Usual care of known GDM patients normalized birth weights, but the cesarean delivery rate was about 33%, whether macrosomia was present or absent. An increased risk of cesarean delivery among treated patients compared with normoglycemic control subjects persisted after adjustment for multiple maternal risk factors. As for the screening tests, time since the last meal had a marked effect on mean plasma glucose. Receiver operating characteristic curve analysis allowed the selection of the most efficient cut points for the GCT based on the time since the last meal. These cut points were 8.2, 7.9, and 8.3 mmol/l (1 mmol/l = 18.015 mg/dl) for elapsed postprandial time of < 2, 2-3, and > 3 h, respectively. With this change from the current threshold of 7.8 mmol/l, the number of patients with a positive screening test dropped from 18.5 to 13.7%. There was an increase in positive predictive value from 14.4 to 18.7%. The overall rate of patient misclassification fell from 18.0 to 13.1%. In conclusion, increasing maternal carbohydrate intolerance in pregnant women without GDM is associated with a graded increase in adverse maternal and fetal outcomes. Infant macrosomia is an important factor in high cesarean delivery rates for women with untreated borderline GDM. Although detection and treatment of GDM normalizes birth weights, rates of cesarean delivery remain inexplicably high. Recognition of GDM may lead to a lower threshold for surgical delivery. The efficiency of screening for GDM can be enhanced by adjusting the current GCT threshold of 7.8 mmol/l to new values related to time since the last meal before screening. Further analyses are underway to elucidate whether maternal risk factors can be used to achieve additional efficiency gains in screening.

Parkinsonism in Ontario: increased mortality compared with controls in a large cohort study.

BACKGROUND: PD was associated with increased mortality before levodopa therapy became available. There have been conflicting reports of PD mortality in the modern era. OBJECTIVE: To assess current mortality rates in a large unselected population receiving treatment for parkinsonism (PKM) followed for up to 6 years. METHODS: Cases were identified using linked administrative databases, including physician service and prescription drug claims, generated in Ontario's universal health insurance system. Control subjects were identified from the provincial registry of citizens and age and sex matched to cases. Comparative mortality was evaluated over the 6-year period of the study (1993/94 to 1998/99). The sensitivity of the findings was tested with differing case definitions. RESULTS: In 1993, 15,304 patients with PKM were identified and were age and sex matched to 30,608 control subjects (1:2 ratio). Over the study period, 50.8% (7,779) of the cases with PKM died compared with 29.1% (8,899) of the control subjects. The cases with PKM had an overall mortality odds ratio of 2.5 (95% CI: 2.4, 2.6) compared with the control group. Results were consistent whether cases were defined by physician diagnosis, use of anti-PD drugs, or both criteria. CONCLUSION: Despite modern drug therapy, PKM continues to confer a sharply increased mortality on unselected patients followed for several years.

Prognostic significance of diabetes as a predictor of survival after cardiac catheterization. APPROACH Investigators. Alberta Provincial Program for Outcome Assessment in Coronary Heart Disease.

PURPOSE: Diabetes is a recognized risk factor for the development of cardiac disease, but its importance as a prognostic factor among patients with known cardiovascular disease is less clear. We evaluated survival in patients with and without diabetes who underwent cardiac catheterization for presumed coronary artery disease. SUBJECTS AND METHODS: We analyzed data from a prospective cohort study that captures detailed clinical information and longitudinal outcomes for all patients who undergo cardiac catheterization in Alberta, Canada. We studied 11,468 patients, 1959 (17%) of whom had diabetes. Logistic regression was used to model predictors of 1-year mortality, and proportional hazards analysis was used to model predictors of survival up to 3 years after cardiac catheterization. RESULTS: One-year mortality was 7.6% for patients with diabetes versus 4.1% for those without diabetes (odds ratio = 1.9, 95% confidence interval [CI]: 1.6 to 2.3). After adjusting for other characteristics of the patients, including comorbid conditions, previous cardiac history, coronary anatomy, and renal function, the odds ratio for 1-year mortality was 1.1 (95% CI: 0.8 to 1.3). Similarly, the adjusted hazard ratio for longer term mortality was 1. 2 (95% CI: 1.0 to 1.4, mean follow-up of 702 days). CONCLUSIONS: These results suggest that there is little or no independent association between diabetes and mortality for up to 3 years after cardiac catheterization. Estimates of short- to intermediate-term prognosis for diabetic patients with coronary artery disease should be based on the presence of other prognostic factors associated with diabetes.

Aspirin and fibrinolysis in acute myocardial infarction: meta-analytic evidence for synergy.

A meta-analysis of randomized clinical trials of fibrinolysis was performed, examining the interaction between aspirin and fibrinolysis in treating patients with acute myocardial infarction. Reductions in the odds of death up to 35 days were assessed for patients receiving tissue plasminogen activator or streptokinase up to 6 hours after the onset of symptoms. No significant difference in effectiveness between tissue plasminogen activator and streptokinase was demonstrated. The overall reduction in odds of death due to fibrinolytic therapy was 28%. However, there was a significant difference between the odds reduction of 24% when fibrinolysis is compared to placebo, and 40% when fibrinolysis and aspirin combined are compared to aspirin alone (p = 0.02). This difference indicates that there exists a synergistic interaction between coronary fibrinolysis and aspirin rather than independence of their beneficial effects, as is generally believed. These results illustrate the perils of assessing drug efficacy, even in an overview of all relevant trials, without consideration of identifiable sources of heterogeneity such as the interaction between the treatment of interest and co-interventions. They also demonstrate the potential application of logistic regression diagnostic techniques to meta-analyses.

Can there be a more patient-centred approach to determining clinically important effect sizes for randomized treatment trials?

Sample sizes for treatment trials with categorical outcomes are conventionally derived by balancing three elements: a difference between alternative treatments in the event rates for the outcomes of interest (commonly termed the clinically important difference), the alpha error tolerance (false positive risk) and the beta error tolerance (false negative risk). Clinically important differences used to plan trials are chosen in part based on earlier experience with similar interventions (i.e. biological or clinical plausibility). Methodological conventions and clinicians' perceptions will also affect choices. Lastly, practical concerns about the feasibility of accruing large numbers of subjects may drive trialists to specify bigger differences as clinically important, with a view to containing sample size requirements. We suggest that patients or other members of the public be given an active role in determining the magnitude of the clinically important treatment effect for trial planning. Probability trade-offs could be constructed to enable patients and/or healthy volunteers to indicate the degree of benefit they would want from a "new" treatment, given the potential side-effects of the same treatment. This method has the advantage of respecting patient autonomy and principles of informed consent. It provides an additional consideration when plausible effect sizes and error tolerances on hypothesis tests are balanced against feasibility of accruing various sample sizes. Its primary disadvantage is inconvenience, as it adds another step to trial design. On the other hand, if patient-based clinically important differences are generated for a variety of disease states and types of treatments, specific trade-off exercises may be needed only for unusual trials.(ABSTRACT TRUNCATED AT 250 WORDS)

Incorporating variations in the quality of individual randomized trials into meta-analysis.

Meta-analysis is a method of synthesizing evidence from multiple sources. It has been increasingly applied to combine results from randomized trials of therapeutic strategies. Unfortunately there is often variation in the quality of the trials that are included in meta-analyses, limiting the value of combining the results in an overview. This variation in quality can lead to both bias and reduction in precision of the estimate of the therapy's effectiveness. There are a number of methods for quantifying the quality of trials including the detailed Chalmers system and simple scales. The nature of the relationship between these quality scores and the true estimate of effectiveness is unknown at this time. We discuss four methods of incorporating quality into meta-analysis: threshold score as inclusion/exclusion criterion, use of quality score as a weight in statistical pooling, visual plot of effect size against quality score and sequential combination of trial results based on quality score. The last method permits an examination of the relation between quality and both bias and precision on the pooled estimates. We conclude that while it is possible to incorporate the effect of variation of quality of individual trials into overviews, this issue requires more study.

Dealing with missing data in observational health care outcome analyses.

Observational outcome analyses appear frequently in the health research literature. For such analyses, clinical registries are preferred to administrative databases. Missing data are a common problem in any clinical registry, and pose a threat to the validity of observational outcomes analyses. Faced with missing data in a new clinical registry, we compared three possible responses: exclude cases with missing data; assume that the missing data indicated absence of risk; or merge the clinical database with an existing administrative database. The predictive model derived using the merged data showed a higher C statistic (C = 0.770), better model goodness-of-fit as measured in a decile-of-risk analysis, the largest gradient of risk across deciles (46.3), and the largest decrease in deviance (-2 log likelihood = 406.2). The superior performance of the enhanced data model supports the use of this "enhancement" methodology and bears consideration when researchers are faced with nonrandom missing data.

Adapting to waiting lists for coronary revascularization. Do Canadian specialists agree on which patients come first?

STUDY OBJECTIVES: To assess specialists' adaptation to long waiting lists for coronary revascularization, and their acceptance of a formal queue-ordering schema proposed by an expert panel. DESIGN: Mail survey of practitioners in referral centers using 49 hypothetical case scenarios. Scenarios were rated for maximum acceptable delay prior to coronary surgery, on a scale with seven interventional time frames graded from emergency to three to six months' permissible delay. The survey included the proposed schema and rating system; respondents were invited to differ as they saw fit. HYPOTHETICAL PATIENTS: Assumed uniformly to be middle aged with typical angina, but clinical factors varied, eg, severity and stability of angina, response to medical therapy, coronary anatomy, and noninvasive test results. PHYSICIAN SUBJECTS: There were 122 respondents, for a 60 percent response rate, including a majority of cardiac surgeons and invasive cardiologists on staff in Ontario teaching hospitals. MEASUREMENTS AND RESULTS: Fifty-seven percent rated some scenarios for acceptable waiting times of three to six months; another 39 percent rated their least urgent scenarios to wait six weeks to three months. Interpractitioner agreement was high: for 48/49 scenarios, at least 75 percent of urgency ratings fell within two contiguous points on the scale. Symptom status was the dominant determinant of waiting time, with mean maximum acceptable wait of 74 days for patients with mild-moderate stable angina but three days for those receiving parenteral nitroglycerin (p less than 0.00001). About half the ratings matched those predicted based on the original panel's consensus criteria; 90 percent were within one scale point. CONCLUSIONS: Specialist practitioners in Ontario have adapted to waiting lists for coronary artery bypass surgery/percutaneous transluminal coronary angioplasty, and assess the priority of hypothetical patients in similar ways and in reasonable accord with formal queue-ordering criteria. This behavior may help mitigate the impact of resource constraints, allowing delay of services for those with less acute need--a potential contrast to delayed access in America based on low income or lack of insurance.

Predictive accuracy study: comparing a statistical model to clinicians' estimates of outcomes after coronary bypass surgery.

BACKGROUND: The purpose of this study was to compare clinicians' prior probability estimates of operative mortality (OM) and prolonged intensive care unit stay (ICU) length of stay greater than 48 hours after coronary artery bypass graft surgery (CABG) with estimates derived from statistical models alone. METHODS: Nine clinicians estimated the predicted probability of OM and ICU stay greater than 48 hours from an abstract of information for each of 100 patients selected from the 1996 to 1997 database of 1,904 patients who underwent isolated CABG. Logistic regression models were used to calculate the predicted probability of OM and ICU stay greater than 48 hours for each patient. The study sample was split into two parts; clinicians were randomly given access to a predictive rule to guide their judgements for one part of the study. RESULTS: Clinicians' estimates were similar with or without access to the rule, and both parts of the study were therefore pooled. Clinicians significantly overestimated the probability of OM (model 6.3% +/- 1%, clinicians 7.6% +/- 3%, p = 0.0001) and ICU stay greater than 48 hours (model 25% +/- 2%, clinicians 28% +/- 1%, p = 0.0012). Clinicians' estimates of OM were not significantly higher than the model's for nonsurvivors (0.8% +/- 0.7%, p = 0.2), but were significantly higher for survivors (1.4% +/- 0.3%, p = 0.039). CONCLUSIONS: Clinicians trusted their own empiric estimates rather than a predictive rule and overestimated the probability of OM and ICU stay greater than 48 hours.

Coronary bypass and carotid endarterectomy: does a combined approach increase risk? A metaanalysis.

BACKGROUND: Patients with concomitant carotid and coronary artery disease present a surgical dilemma. We compared the stroke and mortality rates for combined coronary artery bypass grafting and carotid endarterectomy in which both procedures were performed under a single anesthetic, versus a staged approach, in which coronary artery bypass grafting and carotid endarterectomy were performed separately. METHODS: A computerized MEDLINE search supplemented with a manual bibliographic review was performed for all peer-reviewed English language publications that contained both combined and staged coronary artery bypass grafting/carotid endarterectomy patient cohorts. Outcomes of interest were stroke, death, and stroke or death; aggregation of outcome rates was performed with the Mantel-Haenszel method. RESULTS: Sixteen studies were identified with a total of 844 combined patients and 920 staged patients. None of the studies was completely randomized. The combined surgical group had a higher prevalence of unstable angina; the two groups had a similar prevalence of symptomatic carotid disease and severe carotid stenosis. Meta-analysis revealed a significantly increased risk of the composite end point, stroke or death, for patients undergoing combined procedures (relative risk 1.49; 95% confidence interval 1.03-2.15; p = 0.034). There was also a trend toward increased risk during combined procedures for the end points of stroke (relative risk 1.50; 95% confidence interval 0.97-2.32; p = 0.068) and death (relative risk 1.55; 95% confidence interval 0.94-2.53; p = 0.084) considered separately. The crude event rates for stroke were 6.0% versus 3.2% for combined versus staged procedure, 4.7% versus 2.9% for death, and 9.5% versus 5.7% for stroke or death. Two of the 16 individual studies showed a statistically significant increase in the risk of stroke or death for combined procedure (p < 0.05). CONCLUSIONS: Combined coronary artery bypass grafting and carotid endarterectomy may be associated with a higher risk of stroke or death than staged procedures. A randomized trial needs to be performed to determine the optimal management of patients with concomitant carotid and coronary artery disease.

Cardiopulmonary bypass, rewarming, and central nervous system dysfunction.

BACKGROUND: During cardiopulmonary bypass a nasopharyngeal temperature greater than 38 degrees C at the end of rewarming may indicate cerebral hyperthermia. This could exacerbate an ischemic brain injury incurred during cardiopulmonary bypass. METHODS: In a cohort of 150 aortocoronary bypass patients neuropsychologic test scores of 66 patients whose rewarming temperature exceeded 38 degrees C were compared with those who did not. There were no differences between groups with respect to demographic and intraoperative variables. RESULTS: A trend was seen for hyperthermic patients to do worse on all neuropsychologic tests in the early postoperative period but not at 3-month follow-up. By analysis of covariance hyperthermic patients did worse on the visual reproduction subtest of the Weschler memory scale at 3 months (p = 0.02), but this difference was not found by linear regression (p = 0.10). CONCLUSIONS: We were unable to demonstrate any significant deterioration in patients rewarmed to greater than 38 degrees C in the early postoperative period. The poorer performance in the visual reproduction subtest of the Wechsler memory scale at 3 months in the group rewarmed to more than 38 degrees C is interesting but far from conclusive. Caution with rewarming is still advised pending more in-depth study of this issue.