Risk Adjustment Methods Research Articles

Performance and quality measurement in occupational health services: current status and agenda for further research.

There is growing interest in evaluating the impact and effectiveness of occupational healthcare services (OHS). Managed care networks, case and disease management, worker health promotion, and utilization review services have proliferated in OHS, usually without scientifically based, objective measures of their impact on the quality of care, outcomes, or costs. Performance and quality improvement (QI) efforts in OHS are different from those in general health care because the population, range of conditions, setting of care, and primary outcomes of interest differ considerably. However, scientific knowledge resulting from the application of QI in general health care can provide a useful framework for the development of measures in OHS. We describe the development, selection, and evaluation of OHS performance and quality measures, and propose a research agenda. Quality and performance measures in OHS should address common work-related conditions with high costs, suboptimal outcomes, or unclear parameters for ideal practices. They should reflect stakeholder priorities. Measures should be precise and quantifiable, clearly specified, reliable, valid, and sensitive to important changes. They should be based upon data that can be collected and analyzed at a reasonable cost and effort. The focus of measurement should be on situations where improvement can be achieved by those directly involved in OHS delivery, through defined approaches. Appropriate severity and case-mix adjustment may be especially problematic in OHS. Confidentiality, regulatory concerns, and appropriate use of results must be considered in the development and use of OHS measures. Research is needed to identify, develop, and validate OHS-specific measures, to establish risk adjustment methods, and to test the utility of this information in improving OHS.

Risk adjustment of Florida mental health outcomes data: concepts, methods, and results.

This article discusses outcome evaluation systems for mental health programs. It reviews and critically evaluates design and analysis methods for strengthening the validity of such uncontrolled comparisons. The article examines methods for statistically adjusting preexisting groups, now referred to as risk adjustment or case-mix adjustment, and offers guidelines for determining when this procedure is appropriate. Then, analyses on two dependent variables--a global rating of functioning and a consumer satisfaction measure--available from an outcomes evaluation system currently underway in Florida are used to demonstrate the proposed method of risk adjustment. Results for 24 providers of mental health services showed that while risk adjustment only made a small difference in the overall provider rankings, the ranking of some specific providers changed considerably. The article concludes with a discussion of the implications of this research.

Comparison of Risk Adjusters for Medicaid-Enrolled Children With and Without Chronic Health Conditions

Several capitation payment systems have been developed and implemented recently by public and private insurers as well as by individual managed care organizations. Many pediatricians have expressed concern that methods for establishing capitation rates may not adequately account for the higher expected expenditures for children with chronic health conditions. In this study, we evaluate a demographic- and 4 diagnosis-based models, paying particular attention to their performance for children with chronic health conditions. We selected children 18 years of age and under who were enrolled in the Maryland Medicaid Program in 1995 and 1996. We defined the population of children with chronic health conditions using ICD-9 codes. Individual and group-level analyses were utilized to measure the ability of the different risk adjustment models to predict expenditures in 1996 based upon information available in 1995. All 4 diagnosis-based models significantly outperformed the demographic model for children overall and for children with chronic health conditions. Differences between diagnosis-based models were small, especially as the size of test populations increased. Risk adjustment methods that account directly for health status promise to reduce incentives to exclude children with chronic illnesses from managed care plans and to provide a foundation for more appropriate payments to pediatricians who care for these children.

An examination of methods for risk-adjustment of rehospitalization rates.

This study examines three methods of conducting risk-adjustment to determine if the choice of method results in different conclusions about comparative mental health center performance. The three methods of risk-adjustment are stratification-weighting, logistic regression without interaction effects, and logistic regression with interaction effects. The dependent variable of interest is psychiatric rehospitalization within 14-60 days of discharge to a community mental health center. Subjects are adults discharged in fiscal year 1998 from inpatient psychiatric care to a designated community mental health center in Oklahoma. Using each method, we examine the mental health centers to determine whether their rehospitalization rates are significantly greater than, less than, or not different from, expected. Results show that, for some agencies, method of risk adjustment leads to different conclusions about center performance. Results are discussed with respect to identifying the preferred method of risk-adjustment, study limitations, and next steps in developing risk-adjustment technology and applications.

Methodologic Challenges in Developing and Implementing Measures of Quality for Child Health Care

To review the major building blocks in measurement of quality for child health care, with recommendations for future research. We describe a framework of building blocks for quality measurement and discuss how an investigator's choices for each component are constrained by the special features of child health care. Methodologic challenges for children's health care include developmental change and dependency on others, fragmentary care and inadequate health care data, unusual care settings, potential for long-term consequences, proxy reporting of outcomes and patient experience, small sample sizes, and lack of evidence that links processes and outcomes of care and of methods for risk adjustment. We cite examples of child-specific measures of quality that illustrate solutions to these challenges. Children are different from adults, and measures of health care quality for children must differ from those for adults. We suggest future research on measures of quality directed toward overcoming the methodologic problems specific to child health care.

Comorbidity-based payment methodology for Medicaid enrollees with HIV/AIDS.

Managed care organizations (MCOs) may be incurring financial losses from persons with human immunodeficiency virus (HIV) or acquired immunodeficiency syndrome (AIDS) (PWHA). This study developed a statistical model to examine which specific comorbidities are important contributors to the variations in health care costs of PWHA. Individuals were classified into cost groups to simulate biased selection in MCOs. Capitation payments for various cost groups under different methodologies were compared. The statistical model closely matched payments with the actual costs of care. Capitation payments for HIV/AIDS enrollees based on this model will better protect MCOs than the traditional risk-adjustment methods.

Challenges in comparing risk-adjusted bypass surgery mortality results: Results from the Cooperative Cardiovascular Project

OBJECTIVESWe sought to evaluate the predictive accuracy of four bypass surgery mortality clinical risk models and to examine the extent to which hospitals’ risk-adjusted surgical outcomes vary depending on which risk-adjustment method is applied.BACKGROUNDCardiovascular “report cards” often compare risk-adjusted surgical outcomes; however, it is unclear to what extent the risk-adjustment process itself may affect these metrics.METHODSAs part of the Cooperative Cardiovascular Project’s Pilot Revascularization Study, we compared the predictive accuracy of four bypass clinical risk models among 3,654 Medicare patients undergoing surgery at 28 hospitals in Alabama and Iowa. We also compared the agreement in hospital-level risk-adjusted bypass outcome performance ratings depending on which of the four risk models was applied.RESULTSAlthough the four risk models had similar discriminatory abilities (C-index, 0.71 to 0.74), certain models tended to overpredict mortality in higher-risk patients. There was high correlation between a hospital’s risk-adjusted mortality rates regardless of which of the four models was used (correlation between risk-adjusted rating, 0.93 to 0.97). In contrast, there was limited agreement in which hospitals were identified as “performance outliers” depending on which risk-adjustment model was used and how outlier status was defined.CONCLUSIONSA hospital’s risk-adjusted bypass surgery mortality rating, relative to its peers, was consistent regardless of the risk-adjustment model applied, supporting their use as a means of provider performance feedback. Designation of performance outliers, however, can vary significantly depending on the benchmark and methods used for this determination.

Risk adjustment for high utilizers of public mental health care.

BACKGROUND: Publicly funded mental health systems are increasingly implementing managed care systems, such as capitation, to control costs. Capitated contracts may increase the risk for disenrollment or adverse outcomes among high cost clients with severe mental illness. Risk-adjusted payments to providers are likely to reduce providers' incentives to avoid or under-treat these people. However, most research has focused on Medicare and private populations, and risk adjustment for individuals who are publicly funded and severely mentally ill has received far less attention. AIMS OF THE STUDY: Risk adjustment models for this population can be used to improve contracting for mental health care. Our objective is to develop risk adjustment models for individuals with severe mental illness and assess their performance in predicting future costs. We apply the risk adjustment model to predict costs for the first year of a pilot capitation program for the severely mentally ill that was not risk adjusted. We assess whether risk adjustment could have reduced disenrollment from this program. METHODS: This analysis uses longitudinal administrative data from the County of Los Angeles Department of Mental Health for the fiscal years 1991 to 1994. The sample consists of 1956 clients who have high costs and are severely mentally ill. We estimate several modified two part models of 1993 cost that use 1992 client-based variables such as demographics, living conditions, diagnoses and mental health costs (for 1992 and 1991) to explain the variation in mental health and substance abuse costs. RESULTS: We find that the model that incorporates demographic characteristics, diagnostic information and cost data from two previous years explains about 16 percent of the in-sample variation and 10 percent of the out-of-sample variation in costs. A model that excludes prior cost covariates explains only 5 percent of the variation in costs. Despite the relatively low predictive power, we find some evidence that the disenrollment from the pilot capitation initiative input have been reduced if risk adjustment had been used to set capitation rates. DISCUSSION: The evidence suggests that even though risk adjustment techniques have room to improve, they are still likely to be useful for reducing risk selection in capitation programs. Blended payment schemes that combine risk adjustment with risk corridors or partial fee-for-service payments should be explored. IMPLICATIONS FOR HEALTH CARE PROVISION, USE, AND POLICY: Our results suggest that risk adjustment methods, as developed to data, do not have the requisite predictive power to be used as the sole approach to adjusting capitation rates. Risk adjustment is informative and useful; however, payments to providers should not be fully capitated, and may need to involve some degree of risk sharing between providers and public mental health agencies. A blended contract design may further reduce incentives for risk selection by incorporating a partly risk-adjusted capitation payment, without relying completely on the accuracy of risk adjustment models. IMPLICATIONS FOR FURTHER RESEARCH: Risk adjustment models estimated using data sets containing better predictors of rehospitalization and more precise clinical information are likely to have higher predictive power. Further research should also focus on the effect of combination contract designs.

Performance status of health care facilities changes with risk adjustment of HbA1c.

To develop a risk adjustment method for HbA1c, based solely on administrative data and to determine the extent to which risk-adjusted HbA1c changes the identification of high- or low-performing medical facilities. Through use of pharmacy records, 204,472 diabetic patients were identified for federal fiscal year 1996 (FY96). Complete information (HbA1c levels, demographic data, inpatient records, outpatient pharmacy utilization records) was available on 38,173 predominantly male patients from 48 Veterans Health Administration (VHA) medical facilities. Hierarchical mixed-effects models were used to estimate risk-adjusted unique facility-level HbA1c. Predicted HbA1c demonstrated expected patterns for major factors known to influence glycemic control. Poorer glycemic control was seen in minorities and patients with greater disease severity, longer duration of disease (using treatment type or presence of amputation as surrogates), and more extensive comorbidity (measured by an adapted Charlson index). Better glycemic control was seen in Caucasians, older diabetic patients, and patients with higher outpatient utilization. The number of performance outliers was reduced as a result of risk adjustment. For mean HbA1c levels, 7 facilities that were initially identified as statistically significant outliers were no longer outliers after risk adjustment. For high-risk HbA1c (>9.5%) rates, 12 facilities that were initially identified as statistically significant outliers were no longer outliers after risk adjustment. Risk adjustment using only administrative data resulted in substantial changes in identification of high or low performers compared with non-risk-adjusted HbA1c. Although our findings are exploratory, risk adjustment using administrative data may be a necessary and achievable step in quality assessment of diabetes care measured by rates of high-risk HbA1c (>9.5%).

Standardized infection ratio: the next generation of risk-adjusted wound infection measures

Abstract Background The scientific literature overwhelmingly supports the value of surgical site infection (SSI) surveillance with feedback of infection rates to surgeons. Recognizing that feedback has little value without some form of adjustment for patient factors, surgeons long ago developed the four-level wound classification system. In the 1980s the SENIC multivariate risk index was introduced, followed by the National Nosocomial Infection Surveillance multivariate risk index which is the currently recommended US standard. Methods To address recent advances in surgical technology, ambulatory surgery, and differences between the USA and other countries, newer risk-adjustment methods were explored. The ideal approach would overcome the following common practical problems encountered in generating SSI rates such as: (1) new patient-level risk factors, (2) technology-specific factors, (3) small monthly denominators and (4) postdischarge surveillance. Despite increased mathematical complexity, an acceptable risk-adjustment technique must still provide easily interpretable results. Results The standardized infection ratio (SIR) is proposed as the next generation in epidemiological approaches for monitoring and feeding back risk-adjusted SSI rates to surgeons. The SIR is a single monthly number for each surgeon, calculated by dividing the observed number of SSIs by the expected number predicted from a complex statistical model of patient-level and technology-specific risk factors. Like the well known standardized mortality ratio, the SIR offers the advantage of providing one generalizable quality index regardless of the mix of operations performed. In addition, the SIR is well suited to control chart techniques and for providing valid interhospital comparisons. Conclusion The SIR model offers simple data input and easily interpretable output suitable for application in all healthcare settings worldwide.

Investors' Response to the Performance of Professional Fund Managers: Evidence from the Australian Wholesale Funds Market

This study investigates the influence of past performance on investors' choice of professionally managed investment fund and provides evidence of a link between recent performance and the movement of assets-under-management. The sensitivity of funds flow to recent performance is estimated using data for 124 Australian wholesale funds during the 1980-to-1995 period. The results provide unequivocal evidence of the influence of prior period performance (annual and three-year average) on the movement of assets-under-management. Tests are conducted using gross return and three methods of risk adjustment with index, group average, CAPM as benchmarks. The results provide little evidence of differences in investor response to various measures of performance.

Predicting functional status outcomes in hospitalized patients aged 80 years and older.

To develop a model estimating the probability of a patient aged 80 years or older having functional limitations 2 months and 12 months after being hospitalized. A prospective cohort study. Four teaching hospitals in the US. Enrolled patients were nonelective hospital admissions aged 80 years or older who stayed in hospital at least 48 hours. The 804 patients who survived and completed an interview at 2 months and the 450 who completed an interview at 12 months were from the 1266 patients in the Hospitalized Elderly Longitudinal Project (HELP) (76% and 47% of survivors, respectively). Median age of the 2-month survivors was 84.7 years. Patient function 2 and 12 months after enrollment was defined by the number of dependencies in Activities of Daily Living (ADLs). Ordinal logistic regression models were constructed to predict functional status. Predictors included demographic characteristics, disease category, geriatric conditions, severity of physiologic imbalance, current quality of life, and exercise capacity and ADLs 2 weeks before study admission. Before admission, 39% of patients were functionally independent in ADLs. Of patients who survived and were interviewed at 2 months, 32% were functionally independent, and at 12 months, 36% were independent. Among patients with no baseline dependencies, 42% had developed one or more limitations 2 months later, and 41 % had limitations 12 months later. The patient's ability to perform activities of daily living at baseline was the most important predictor of functional status at both 2 and 12 months. In a multivariable predictive model, independent predictors of poorer functional status at 2 months included: worse baseline functional status and quality of life; depth of coma, if any; lower serum albumin level; presence of dementia, depression, or incontinence; being bedridden; medical record documentation of need for nursing home; and older age. Model performance, assessed using Somers' D, was 0.61 for 2 months and 0.57 for 12 months (Receiver Operating Characteristic (ROC) area = 0.81 and .79, respectively.) Bootstrap validation of the month 2 model also yielded a Somers' D = 0.60. The models were well calibrated over the entire risk range. The ROC area for prediction of the loss of independence was 0.76 for 2 months and 0.68 for 12 months. Many older patients are functionally impaired at the time of hospitalization, and many develop new functional limitations. A limited amount of readily available clinical information can yield satisfactory predictions of functional status 2 months after hospitalization. Models like this may prove to be useful in clinical care. This work illuminates a potential method for risk adjustment in research studies and for monitoring quality of care.

Does the Fed beat the foreign-exchange market?

This paper’s estimates and tests of Fed intervention profits are the first that explicitly adjust for foreign-exchange risk premia; failure to adjust may grossly affect estimated profits. Profits appear economically and statistically significant, whether risk premia are modeled as time-constant or as appreciation’s market beta depending on Fed intervention. The estimates are sensitive to the method of risk adjustment and to the periods used. Because a key variable, cumulative intervention, is I(1), test statistics may have non-standard distributions, a problem affecting past tests; this paper’s tests account for non-standard distributions. Possible explanations of these profits have mixed empirical support in the literature.

Target, application, and interpretation of scores and alternative methods for risk assessment in cardiac surgery.

The more popular the use of different methods for risk adjustment becomes, the more often data are applied without any regard about the primary target and/or about important assumptions. Furthermore, risk adjustment is no longer restricted for quality assurance purposes, but became a "tool" of health policy. Few working groups currently use risk adjustment for the development of new therapeutic concepts. The aim of our study is to clarify possibilities and limitations of popular risk adjustment methods. 4985 Patients underwent isolated CABG. Statistics was performed by calculating descriptive statistics, Parsonnet, and Higginsscores. Furthermore, the parametric, time-adjusted hazard function by Blackstone was used. Descriptive statistics allows intra-, and interinstitutional comparisons of single items to identify "outlying" results. Risk scores aim to predict preoperatively the risk category of the patient who undergoes cardiac surgery. However, since different scores are based on a score-specific combination of variables, and different definitions of the investigation interval, different results may occur, when different scores are calculated for a single patient. However, the use for example, of scores in patient groups allows description of changing risk structures. Most of the scores derive from univariate analyses and monophasic functions. However, survival curves are predominantly multiphasic and require a consideration of the time-dependency of "risk factors". An increasing number of patients with severe comorbidity undergoes cardiac surgery. To evaluate reliably present and futurous therapeutic options, risk adjustment is necessary. Since various tools for risk-adjustment are available, a serious discussion about reliability and application is necessary.

Risk Adjustment for Complications of Hysterectomy: Limitations of Routinely Collected Administrative Data

Abstract Objective: We sought to determine the utility of routinely collected administrative data for risk adjustment for complications of hysterectomy. Study Design: Using abstracted discharge data on 107,648 women undergoing hysterectomy in North Carolina from 1988 through 1994, we constructed logistic regression models for the prediction of medical and surgical complications incorporating coded demographic, diagnostic, and procedural data. Results: The overall complication rate was 16%, with surgical complications (11.8%) more common than medical complications (6.7%). Hysterectomy type, teaching hospital status, patient age ≥65 years, and insurance status of Medicaid or no insurance were significantly associated with both medical and surgical complication risk, as were procedures performed for cancer or pregnancy complications. Models that incorporated coded comorbidity were better predictors of medical complications (C = 0.714) than surgical complications (C = 0.630). Conclusion: Although surgical complications of hysterectomy are more common than medical complications, risk adjustment methods that use routinely collected administrative data are better at predicting medical complications. Ambiguities in coding, misclassification, and uncoded factors such as disease severity limit the utility of administrative data for risk adjustment for hysterectomy complications. (Am J Obstet Gynecol 1999;181:567-75.)

Impact of risk adjusted clinical outcomes methodology--quality measures on hospital mortality data: a statistical and case study approach.

The effectiveness of risk adjustment in improving mortality as a performance measure for hospitals remains uncertain. New techniques of risk adjustment should be empirically tested, and health care professionals, using the data derived from such measures, should be queried before final acceptance of these technologies of measurement is warranted. The Risk Adjusted Clinical Outcomes Methodology-Quality Measures (RACOM-QM), a relatively new risk-adjustment methodology developed by the QuadraMed Corporation, was used by Maryland hospitals for risk adjustment for the first time in 1997. A research study was undertaken by the Maryland Hospital Association to determine the impact of RACOM-QM on mortality rates, its empirical validity, and its acceptance in the field. The relationship between RACOM-QM mean risk scores and mortality rates was examined using inpatient hospital mortality data for Maryland in 1996. Using these same data, the empirical relationship between risk-adjusted and unadjusted mortality by diagnosis-related group (DRG) was also investigated. Case studies were undertaken to glean information about the use and acceptability of this new methodology in 2 hospital settings in Maryland. There was a strong relationship between mean mortality risk scores and mortality rates. The analysis of the empirical relationship between risk-adjusted and unadjusted mortality by DRG yielded support for the impact of RACOM-QM in adjusting inpatient mortality rates. The case studies supported the utility of this method of risk adjustment in increasing the interpretation of mortality data and in helping to identify areas in which to investigate quality in more depth in 2 hospital settings. This study provides overall support for the usefulness of risk adjustment and, specifically, the RACOM-QM, in increasing the interpretation of inpatient mortality rates in Maryland's acute care hospitals. This study also suggests that use of the RACOM-QM improved comparative analysis of inpatient mortality rates among Maryland hospitals. Finally, the results of the case study analysis suggest that improved internal review of mortality rates and increased clinician acceptance of these rates as indicators of performance were enhanced by the use of a risk adjustment methodology.

Risk adjustment for complications of hysterectomy: Limitations of routinely collected administrative data

Objective: We sought to determine the utility of routinely collected administrative data for risk adjustment for complications of hysterectomy. Study Design: Using abstracted discharge data on 107,648 women undergoing hysterectomy in North Carolina from 1988 through 1994, we constructed logistic regression models for the prediction of medical and surgical complications incorporating coded demographic, diagnostic, and procedural data. Results: The overall complication rate was 16%, with surgical complications (11.8%) more common than medical complications (6.7%). Hysterectomy type, teaching hospital status, patient age ≥65 years, and insurance status of Medicaid or no insurance were significantly associated with both medical and surgical complication risk, as were procedures performed for cancer or pregnancy complications. Models that incorporated coded comorbidity were better predictors of medical complications ( C = 0.714) than surgical complications ( C = 0.630). Conclusion: Although surgical complications of hysterectomy are more common than medical complications, risk adjustment methods that use routinely collected administrative data are better at predicting medical complications. Ambiguities in coding, misclassification, and uncoded factors such as disease severity limit the utility of administrative data for risk adjustment for hysterectomy complications. (Am J Obstet Gynecol 1999;181:567-75.)

Definition and adjustment of Cesarean section rates and assessments of hospital performance

Demand is growing for comparative data such as Cesarean section rates, but little effort has been made to develop either standardized definitions or risk adjustment approaches. To determine to what extent a seemingly straightforward indicator like Cesarean section rate will vary when calculated according to differing definitions used by various performance measurement systems. Retrospective data abstraction of 200 deliveries per hospital. Fifteen acute care hospitals including two from outside the USA. Four widely-used performance measurement systems provided specifications for their Cesarean section indicators. Indicator specifications varied on inclusion criteria (whether the population was defined using Diagnostic Related Groups or ICD-9-CM procedure codes or ICD-9-CM diagnosis codes) and risk-adjustment methods and factors. Rates and rankings were compared across hospitals using different Cesarean section indicator definitions and indicators with and without risk adjustment. Calculated Cesarean section rates changed substantially depending on how the numerator and denominator cases were identified. Relative performance based on Cesarean section rankings is affected less by differing indicator definitions than by whether and how risk adjustment is performed. Judgments about organizational performance should only be made when the comparisons are based upon identical indicators. Research leading to a uniform indicator definition and standard risk adjustment methodology is needed.

Risk adjustment for complications of hysterectomy: utility of routinely collected administrative data

Objective: To determine the ability of risk adjustment methods that use routinely collected administrative data to explain variability in complication rates after hysterectomy. Methods: Discharge data on all non-radical hysterectomies performed in North Carolina between January 1988 and September 1994 were obtained from the North Carolina Medical Database Commission. Complications were categorized as 1) surgical (eg, hemorrhage, organ injury) and 2) medical (eg, myocardial infarction, pneumonia). Comorbidities included peritoneal adhesions and chronic medical problems. Hospital charges were adjusted for inflation. Univariate analyses were performed using the Kruskall-Wallis test for skewed continuous variables and χ 2 tests for categorical variables. Multivariate analysis was performed using unconditional logistic regression, with complication rate as the dependent variable. Results: There were 107,648 cases performed at 134 hospitals, with an overall complication rate of 9.5%. When cancer and pregnancy cases were excluded, the surgical complication rate was 5.0%, medical 3.2%. Patients with surgical complications were significantly younger (median age 42 vs 46) and had significantly higher total hospital charges (median $8,127 vs $7,496) than patients with medical complications. Complication rates for individual hospitals varied from 1.5% to 29.3%, with rates highest for academic medical centers (24.3% vs 7.2% for non-teaching hospitals). Significant predictors of complications in univariate analyses included type of hysterectomy, indication, age ≥65 years, insurance status, and teaching hospital status. Coded comorbid conditions were variable in their association with complications. Adjusted odds ratios, controlled for indication and type of procedure, for age ≥65, Medicaid or no insurance, and teaching hospital status were all greater for medical complications than for surgical complications in multivariate analysis (table). The predictive ability of multivariate analysis was better for medical complications than for surgical complications (C-statistic for medical complication model 0.763 vs 0.644 for surgical complications). Risk Factor Surgical Complications Medical Complications Adjusted odds ratio 95% CI Adjusted odds ratio 95% CI Age ≥65 years 1.73 1.58, 1.90 4.54 4.16, 4.96 Medicaid/ no insurance (compared to private/ Medicare/HMO 1.36 1.26, 1.47 1.62 1.48, 1.77 Academic medical center (compared to non-teaching hospital) 1.86 1.74, 1.99 3.28 3.06, 3.52 Conclusions: Surgical complications of hysterectomy are more common, occur in younger women, and are associated with higher hospital charges than medical complications. Complication rates vary widely between hospitals, with teaching hospitals having the highest rates. This difference persists after adjustment for coded comorbidities. Possible explanations for the inability of multivariate analysis to explain the wide range in observed complication rates include 1) shortcomings in the available data or models, such as variability in coding practices between hospitals or variability in surgical difficulty between hospitals that is not captured with ICD-9-CM codes (eg, uterine size, cancer stage) and 2) variability in the quality of care between hospitals. Further research is needed to help determine the role of each of these explanations. Risk adjustment methods that use administrative data based on currently available coding standards are poor predictors of surgical complication rates after hysterectomy and should not be used to reach conclusions about quality of care.

Remaking Medicare: the voucher myth.

There is growing enthusiasm for transforming Medicare into a voucher system. Advocates claim vouchers would increase the health care choices available to Medicare beneficiaries, reduce the regulatory burden on the federal government, and promote the benefits of fair market competition. In addition, some analysts contend vouchers are the only feasible solution to Medicare's short-term financing problems and the long-term "crisis" of the retirement of the baby-boom generation. The author argues against these claims. Vouchers would not work as advertised by proponents because of the limitations of risk-adjustment methods and unrealistic assumptions about consumer choice. Moreover, the elderly and disabled Medicare population is ill-suited to cope in a competitive insurance system. Implementation of vouchers would therefore pose a threat to both the health of beneficiaries and the stability of the Medicare program. The implications of this analysis for Medicare reform are discussed.