Carotid endarterectomy for asymptomatic patients. Assessing results of a quantitative synthesis.

Carotid endarterectomy (CE) surgery for asymptomatic patients remains controversial despite hundreds of published studies and recent randomized trials. Safety and efficacy are assessed using a quantitative synthesis method derived from meta-analysis and a "critical multiplist" inference approach. In addition, multivariate analyses reveal that use of a surgical shunt could further improve CE outcomes. Methods are examined for both their "confirmatory" and "exploratory" value.

Validating risk-adjusted mortality as an indicator for quality of care.

Risk-adjusted mortality is perhaps the most commonly referenced outcome indicator for assessments of hospital quality. While mortality rate possesses considerable intuitive appeal as a quality indicator, scientific evidence concerning its appropriateness for this purpose is mixed. In this paper, we use a computerized discharge database to model mortality risks for patients hospitalized for three different conditions: cardiac disease (ischemic heart disease, coronary artery disease, angina, and left ventricular aneurysm), acute myocardial infarction, and septicemia. We then use a database of peer review quality findings to determine whether the ratio of observed to expected deaths in each of these conditions relates validly to quality. The results of our analyses provide strong support for the validity of one of our mortality indicators, weak support for another, and no support for the third. We conclude that before inferences about hospital quality are made using any risk-adjusted mortality indicator, the validity of the quality/outcome relationship must be established explicitly for that measure.

Precision of blood cholesterol measurement and high blood cholesterol case-finding and treatment.

Imprecise blood cholesterol measurement can be expected to adversely affect large scale efforts to detect and treat high blood cholesterol. Using protocols specified by the National Cholesterol Education Program (NCEP), we used computer simulation techniques to quantify the effects of blood cholesterol measurement variability on misclassification rates, costs, effectiveness, and cost-effectiveness of high blood cholesterol case-finding and treatment. At the time of initial case-finding, increased measurement variability was associated with a moderate decrease in the proportions assigned to a treatment state and in the positive predictive value of such an assignment. After 10 years of continual case-finding and treatment, measurement variability dramatically affected proportions assigned to drug treatment and diminished the percent on drugs with blood cholesterol levels truly above NCEP cutpoints. Extreme variability in blood cholesterol measurement increased per capita costs by 14-18% and diminished cost-effectiveness by at least 11-12%. The adverse effects of measurement variability on cost-effectiveness were much more pronounced if adjustments to life-expectancy were made to recognize the lower quality of life associated with drug treatment. Misclassification rates can be decreased and cost-effectiveness improved by performing repeated measurements of blood cholesterol before increasing intensity of treatment. Improvement in the precision of measurement are especially beneficial for low-risk individuals.

Investigating early readmission as an indicator for quality of care studies.

Readmission to a hospital shortly following a previous discharge may be viewed as an adverse outcome of care. Consequently, early readmission represents a potentially useful indicator for monitoring quality. While a number of recent research studies have focused on this issue, several important questions concerning appropriate use of early readmission as a quality of care indicator remain to be addressed. In this article, using data on all discharges for 1 year from 18 hospitals, several of these questions are investigated. Specifically, whether the significant predictors of readmission risk are different for different types of cases (defined using DRGs), whether case severity is an important predictor of readmission risk, whether readmission risks differ systematically with hospital size and other characteristics, whether readmission risk is a function of patients' lengths-of-stay, and whether readmission risk is influenced by whether or not patients are discharged home or into organized care environments are explored. For this study, the focus is on patients who experienced unplanned readmissions to acute care hospitals within 31 days of a prior discharge. The Patient Management Category classification system and ICD-9-CM diagnosis and procedure codes are used to identify, and then exclude from consideration, those readmissions that occurred as part of an appropriately planned sequence of care. In each of 22 sets of related DRGs, analysis of unplanned readmissions indicates that severity/complexity is an important risk factor for early readmission and that clinical and other risk factors differ for different DRG groups. Thus, in future studies of early readmissions, researchers will need to control for both the type (e.g., DRG) and severity/complexity of individual cases. In examining relationships between early readmission and hospital characteristics, no consistent patterns suggestive of quality of care problems were detected.

Risk factors for early readmission among veterans.

This study was undertaken to identify demographic, clinical, and social risk factors for early readmission in the veteran population. Readmissions within 30 days of discharge were considered "early." A randomly selected 50 percent sample of 6,317 veterans discharged consecutively from one Department of Veterans Affairs medical center (VAMC) was used to build a logistic regression model for early readmission. Of these patients, 22 percent had early readmissions. The adjusted odds ratios (OR) of greater magnitude for early readmission (p less than .05) were associated with discharge from a geriatrics/intermediate care bed (OR = 2.75 relative to medical ward), discharge diagnosis of a chronic disease (OR = 2.03-2.67 relative to acute or self-limiting disorders), and two or more surgical procedures performed during the index admission (OR = 1.87 relative to no surgery). Increasing distance from the VA hospital and increasing age also added readmission risk (OR = 1.18 and 1.10, respectively). Length of stay and the social risk factors of marital status and place of disposition were not sufficiently predictive to enter the model. The model was validated successfully on the second 50 percent sample of patients. We conclude that clinical and demographic factors are more predictive of early readmission than are social factors. Early readmission models could be used to improve VA discharge planning and to focus quality assurance and utilization review efforts on providers whose early readmission rates exceed those predicted by the models.

A mathematical representation of the expert panel's guidelines for high blood cholesterol case-finding and treatment.

The National Cholesterol Education Program (NCEP) for high blood cholesterol case-finding and treatment recommended discrete treatments according to the results of sequential measurements of continuous variables (total and low-density lipoprotein cholesterol [LDL-C]). These measurements are subject to intra- and interindividual variability. The authors describe a computer simulation of the NCEP that acknowledges these complexities. The simulation reduces the NCEP into steps, which are represented in a decision tree. The calculation of probabilities at chance nodes takes into account the conditional nature of sequential measurements of blood cholesterol. The simulation tracks medical resource use and estimates LDL-C reductions within 20 strata, each defining 5% of population distribution for LDL-C. This approach enables a detailed representation of the case-finding process--the sequence of blood cholesterol tests and associated cut-off values that identify individuals needing more intensive evaluation and treatment.

Factors influencing readmission risk: implications for quality monitoring.

By applying multiple logistic regression to data from the 1980 National Medical Care Utilization and Expenditure Survey, independent risk factors for readmission to an acute care hospital within 31 days of the preceding discharge were identified. Subjects who were initially admitted for a high-risk condition, those with poor perceived health status, and those who had no surgical procedures performed were most likely to be readmitted. Sex, race, marital status, insurance coverage, and access to outpatient care did not independently influence readmission risk. Readmission risk models used to monitor quality of care need not adjust for these nonmedical factors.

Effects of deceptive self-reports of quitting on the results of treatment trials for smoking: a quantitative assessment.

Problems with self-report measures for smoking motivate the use of biochemical tests in treatment trials for smoking. These biochemical tests, unfortunately, are not perfect. In this paper, we present an algebraic model of bias in treatment trials for smoking. Bias is expressed in terms of the deception rate among continued smokers in a control group, the relative deception rate among continued smokers in an experimental group, and the sensitivity and specificity of a biochemical test which may be used either to confirm self-reports of quitting or to replace self-report entirely. For given test specificity and sensitivity, the model defines deception rates for which different biochemical testing strategies are preferred. The model is presented in the context of current knowledge on the phenomenon of deception among adult smokers. The paper concludes that better judgements regarding the role of biochemical tests in treatment trials for smoking require more precise information regarding the magnitude and determinants of deception.

Clinical and sociodemographic risk factors for readmission of Medicare beneficiaries.

In a random sample of Medicare beneficiaries, multiple logistic regression was used to examine clinical, sociodemographic, and insurance coverage risk factors for readmission within 60 days of discharge. The patients most likely to be readmitted were those with poor health status or with chronic diseases and those who had not had surgery. Age, marital status, living situation, and having insurance to supplement Medicare were not independently predictive of readmission risk. The dominant roles of health status, diagnosis, and surgery as predictors of readmission provide evidence that risk-adjusted readmission rates can be equitably used for quality of care studies.

Cefotaxime vs nafcillin and tobramycin for the treatment of serious infection. Comparative cost-effectiveness.

To evaluate the cost-effectiveness of cefotaxime sodium at a dosage of 12 g/day vs nafcillin sodium and tobramycin sulfate for the treatment of serious infection, the hospital and physician charges of patients enrolled in a prospective, randomized, clinical trial were analyzed. For 187 patients receiving therapy empirically, mean hospital charges for the interval in which the trial antibiotics were used were $3,550 +/- $1,740 for cefotaxime and $3,160 +/- $1,990 for nafcillin and tobramycin. After adjusting for cost-generating factors, charges for cefotaxime were greater than for nafcillin and tobramycin, but the difference was not significant. For 107 patients with clinically or bacteriologically documented infection, mean charges were $3,980 +/- $1,800 for cefotaxime and $4,170 +/- $1,780 for nafcillin and tobramycin. Adjusted charges did not differ. Incremental charges for cefotaxime per additional response were $1,630 in all patients and -$820 in patients with clinically or bacteriologically documented infections.

Comparative cost effectiveness of gentamicin and tobramycin.

Gentamicin and tobramycin were compared for cost effectiveness in the treatment of adult patients with serious infections in a general medical service. We used data from a randomized double-blind trial in which the only observed difference between the clinical effects of these aminoglycosides was the incidence of nephrotoxicity (26% with gentamicin and 12% with tobramycin). According to 1984 cost data, the combined average drug and nephrotoxicity costs per patient were $127 for tobramycin and $72 for gentamicin. An extensive sensitivity analysis--varying frequency and cost of nephrotoxicity, dialysis requirements, aminoglycoside acquisition costs, and length of hospitalization--showed gentamicin to be more cost effective than tobramycin, unless hospitalization is prolonged by an average of at least 15 days for patients with severe nephrotoxicity or at least 3 days for all patients with moderate or severe nephrotoxicity.