Increasingly multiple outcomes are collected in order to characterize treatment effectiveness or to evaluate the impact of large policy initiatives. Often the multiple outcomes are non-commensurate, e.g. measured on different scales. The common approach to inference is to model each outcome separately ignoring the potential correlation among the responses. We describe and contrast several full likelihood and quasi-likelihood multivariate methods for non-commensurate outcomes. We present a new multivariate model to analyze binary and continuous correlated outcomes using a latent variable. We study the efficiency gains of the multivariate methods relative to the univariate approach. For complete data, all approaches yield consistent parameter estimates. When the mean structure of all outcomes depends on the same set of covariates, efficiency gains by adopting a multivariate approach are negligible. In contrast, when the mean outcomes depend on different covariate sets, large efficiency gains are realized. Three real examples illustrate the different approaches.
There is increasing interest in utilizing novel markers of cardiovascular disease risk, and consequently, there is a need to assess the value of their use. This scientific statement reviews current concepts of risk evaluation and proposes standards for the critical appraisal of risk assessment methods. An adequate evaluation of a novel risk marker requires a sound research design, a representative at-risk population, and an adequate number of outcome events. Studies of a novel marker should report the degree to which it adds to the prognostic information provided by standard risk markers. No single statistical measure provides all the information needed to assess a novel marker, so measures of both discrimination and accuracy should be reported. The clinical value of a marker should be assessed by its effect on patient management and outcomes. In general, a novel risk marker should be evaluated in several phases, including initial proof of concept, prospective validation in independent populations, documentation of incremental information when added to standard risk markers, assessment of effects on patient management and outcomes, and ultimately, cost-effectiveness.
BACKGROUND: In patients with a major cardiac event, the first priority is to minimize time to treatment. For many patients, first contact with the health system is through emergency medical services (EMS). We set out to identify patient-level and neighborhood-level factors that were associated with elapsed time in EMS.
METHODS AND RESULTS: A retrospective cohort study was conducted in 10 municipalities in Dallas County, Tex, from January 1 through December 31, 2004. The data set included 5887 patients with suspected cardiac-related symptoms. The region was served by 29 hospitals and 98 EMS depots. Multivariate models included measures of distance traveled, time of day, day of week, and patient and neighborhood characteristics. The main outcomes were elapsed time in EMS (continuous; in minutes) and delay in EMS (dichotomous; >15 minutes beyond median elapsed time). We found positive associations between patient characteristics and both average elapsed time and delay in EMS care. Variation in average elapsed time was not large enough to be clinically meaningful. However, approximately 11% (n=647) of patients were delayed >or=15 minutes. Women were more likely to be delayed (adjusted odds ratio, 1.52; 95% confidence interval, 1.32 to 1.74), and this association did not change after adjusting for other characteristics, including neighborhood socioeconomic composition.
CONCLUSIONS: Compared with otherwise similar men, women have 50% greater odds of being delayed in the EMS setting. The determinants of delay should be a special focus of EMS studies in which time to treatment is a priority.
Disparities in access to and quality of health care along racial and ethnic lines are an important national problem. Health care purchasers and payers have a potentially important role to play in alleviating this problem. Using national surveys of 609 employers and 252 health plans with HMO products in 41 U.S. markets, we examined awareness of racial and ethnic disparities in health care access and quality, perceptions of employer and health plan role in addressing disparities, and reported efforts to measure and reduce disparities. Our findings suggest that most health plans and many employers are aware of the existence of substantial disparities and that health plans, but not employers, have taken steps to examine and influence patterns of care by race and ethnicity among their members.
The slow spread of treatments supported by empirical evidence and the rapid diffusion of treatments lacking such support play major roles in the lower quality of mental health care received by people with severe mental illnesses compared with the care of less severely ill people. Further, the rapid spread of treatments that are of low cost-effectiveness limits the system's ability to provide the full gamut of high-value treatments available to treat this vulnerable population. Using the case of schizophrenia, we review the context in which these paradoxical patterns of diffusion have occurred, and we propose policy solutions.
OBJECTIVE: The authors summarize points for consideration generated in a National Institute of Mental Health (NIMH) workshop convened to provide an opportunity for reviewers from different disciplines-specifically clinical researchers and statisticians-to discuss how their differing and complementary expertise can be well integrated in the review of intervention-related grant applications.
METHODS: A 1-day workshop was convened in October, 2004. The workshop featured panel presentations on key topics followed by interactive discussion. This article summarizes the workshop and subsequent discussions, which centered on topics including weighting the statistics/data analysis elements of an application in the assessment of the application's overall merit; the level of statistical sophistication appropriate to different stages of research and for different funding mechanisms; some key considerations in the design and analysis portions of applications; appropriate statistical methods for addressing essential questions posed by an application; and the role of the statistician in the application's development, study conduct, and interpretation and dissemination of results.
RESULTS: A number of key elements crucial to the construction and review of grant applications were identified. It was acknowledged that intervention-related studies unavoidably involve trade-offs. Reviewers are helped when applications acknowledge such trade-offs and provide good rationale for their choices. Clear linkage among the design, aims, hypotheses, and data analysis plan and avoidance of disconnections among these elements also strengthens applications.
CONCLUSION: The authors identify multiple points to consider when constructing intervention-related grant applications. The points are presented here as questions and do not reflect institute policy or comprise a list of best practices, but rather represent points for consideration.
OBJECTIVE: This study examined whether there are service disparities among homeless adults with severe mental illnesses, a vulnerable population with a high level of unmet need.
METHODS: Data were collected at baseline for 6,829 black, Latino, and non-Latino white participants in the Access to Community Care and Effective Services and Support study. Outcome variables were measures of utilization of psychiatric outpatient, housing, and case management services in the previous 60 days. The sample was divided into white-black and white-Latino cohorts. Within each cohort, participants were stratified into comparable groups by propensity scores that estimated log-odds of being black or Latino as a function of several confounding variables. White-black and white-Latino differences in mean number of visits (a measure of intensity) and in the mean probability of at least one visit (a measure of access) were subsequently estimated for each of the three services.
RESULTS: The composition of the sample was 50% black, 6% Latino, and 44% white. Service utilization was low for the three services regardless of race-ethnicity. On multivariate analyses of service utilization in the previous 60 days, blacks made fewer psychiatric outpatient visits than whites (mean difference=.46, 95% confidence interval [CI]=.10 to .81]), yet Latinos had more case management visits than whites (mean difference=-.51, CI=-1.03 to -.05]). Analyses of access did not reveal racial-ethnic disparities.
CONCLUSIONS: Whereas blacks used psychiatric outpatient services less frequently than whites, hence experiencing a service disparity, Latinos used case management services more than whites did. Possible contributors and clinical and methodological implications of these results are discussed.
OBJECTIVE: To examine racial/ethnic longitudinal disparities in antimanic medication use among adults with bipolar-I disorder.
METHODS: Observational study using administrative data from Florida's Medicaid program, July 1997 to June 2005, for enrollees diagnosed with bipolar-I disorder (N = 13,497 persons; 126,413 person-quarters). We examined the likelihood of receiving one of the following during a given quarter: (1) any antimanic agent (antipsychotic or mood stabilizer) or none, and (2) mood stabilizers, antipsychotic monotherapy, or neither. Binary and multinomial logistic regression models predicted the association between race/ethnicity and prescription fills, adjusting for clinical and demographic characteristics. Cohort indicators for year that the enrollee met study criteria were included to account for cohort effects.
RESULTS: Averaging over all cohorts and quarters, compared with whites, blacks had lower odds of filling any antimanic and mood stabilizer prescriptions specifically (40%-49% and 47%-63%, respectively), but similar odds of filling prescriptions for antipsychotic monotherapy. After Bonferroni adjustment, compared with whites, there were no statistically significant disparities for Hispanics in filling prescriptions for any, or specific antimanic medications.
CONCLUSIONS: Rates of antimanic medication use were low regardless of race/ethnicity. However, we found disparities in antimanic medication use for blacks compared with whites and these disparities persisted over time. We found no Hispanic-white disparities. Quality improvement efforts should focus on all individuals with bipolar disorder, but particular attention should be paid to understanding disparities in medication use for blacks.
BACKGROUND: The rankings of "America's Best Hospitals" by U.S. News & World Report are influential, but the performance of ranked hospitals in caring for patients with routine cardiac conditions such as heart failure is not known.
METHODS AND RESULTS: Using hierarchical regression models based on medical administrative data from the period July 1, 2005, to June 30, 2006, we calculated risk-standardized mortality rates and risk-standardized readmission rates for ranked and nonranked hospitals in the treatment of heart failure. The mortality analysis examined 14 813 patients in 50 ranked hospitals and 409 806 patients in 4761 nonranked hospitals. The readmission analysis included 16 641 patients in 50 ranked hospitals and 458 473 patients in 4627 nonranked hospitals. Mean 30-day risk-standardized mortality rates were lower in ranked versus nonranked hospitals (10.1% versus 11.2%, P<0.01), whereas mean 30-day risk-standardized readmission rates were no different between ranked and nonranked hospitals (23.6% versus 23.8%, P=0.40). The 30-day risk-standardized mortality rates varied widely for both ranked and nonranked hospitals, ranging from 7.9% to 12.4% for ranked hospitals and from 7.1% to 17.5% for nonranked hospitals. The 30-day risk-standardized readmission rates also spanned a large range, from 18.7% to 29.3% for ranked hospitals and from 19.2% to 29.8% for nonranked hospitals.
CONCLUSIONS: Hospitals ranked by U.S. News & World Report as "America's Best Hospitals" in "Heart & Heart Surgery" are more likely than nonranked hospitals to have a significantly lower than expected 30-day mortality rate, but there was much overlap in performance. For readmission, the rates were similar in ranked and nonranked hospitals.
BACKGROUND: In 2009, the Centers for Medicare & Medicaid Services is publicly reporting hospital-level risk-standardized 30-day mortality and readmission rates after acute myocardial infarction (AMI) and heart failure (HF). We provide patterns of hospital performance, based on these measures.
METHODS AND RESULTS: We calculated the 30-day mortality and readmission rates for all Medicare fee-for-service beneficiaries ages 65 years or older with a primary diagnosis of AMI or HF, discharged between July 2005 and June 2008. We compared weighted risk-standardized mortality and readmission rates across Hospital Referral Regions and hospital structural characteristics. The median 30-day mortality rate was 16.6% for AMI (range, 10.9% to 24.9%; 25th to 75th percentile, 15.8% to 17.4%; 10th to 90th percentile, 14.7% to 18.4%) and 11.1% for HF (range, 6.6% to 19.8%; 25th to 75th percentile, 10.3% to 12.0%; 10th to 90th percentile, 9.4% to 13.1%). The median 30-day readmission rate was 19.9% for AMI (range, 15.3% to 29.4%; 25th to 75th percentile, 19.5% to 20.4%; 10th to 90th percentile, 18.8% to 21.1%) and 24.4% for HF (range, 15.9% to 34.4%; 25th to 75th percentile, 23.4% to 25.6%; 10th to 90th percentile, 22.3% to 27.0%). We observed geographic differences in performance across the country. Although there were some differences in average performance by hospital characteristics, there were high and low hospital performers among all types of hospitals.
CONCLUSIONS: In a recent 3-year period, 30-day risk-standardized mortality rates for AMI and HF varied among hospitals and across the country. The readmission rates were particularly high.
BACKGROUND: Exposure to adverse family environments in childhood can influence the risk trajectory for developing substance use disorders in adolescence. Evidence for this is largely based on cross-sectional studies which have been unable to establish the temporality of this association and investigate underlying pathways.
METHODS: The sample consisted of 1421 adolescents from the Project on Human Development in Chicago Neighborhoods, a three wave longitudinal study conducted between 1994 and 2001 that followed children from ages 10 to 22. Logistic regression analyses with multiple imputation were conducted to examine the relation between familial conflict in childhood and substance use disorders in late adolescence and emerging adulthood. We conducted mediational analyses to determine if internalizing and externalizing problems explain this relationship, and we investigated whether external social support mitigates the adverse effects of familial conflict on the development of substance use disorders.
RESULTS: Familial conflict was significantly associated with the risk of substance use disorders during adolescence (odds ratio: 1.23; 95% CI: 1.02-1.47), and 30% of this effect was due to higher levels of externalizing problems (but not internalizing problems). External social support in childhood did not buffer the effects of familial conflict on substance use disorders during adolescence.
CONCLUSION: Exposure to familial conflict early in life increases the risk of substance use disorders during late adolescence and emerging adulthood, due partly to higher levels of externalizing problems, but not internalizing problems. Future research is needed to identify additional pathways underlying this association, and the extent to which these pathways are modifiable.
To determine whether Hispanic and African-American patients are treated by cardiac surgeons with better or worse risk-standardized outcomes than surgeons of white patients, clinical data from the Massachusetts Data Analysis Center Registry were analyzed on all patients who underwent isolated coronary artery bypass grafting (CABG) from 2002 to 2004 by surgeons who performed >or=10 operations. Surgeons were divided into 4 groups based on their risk-standardized 30-day all-cause mortality incidence rates (top decile, top quartile, bottom quartile, and bottom decile). A total of 12,973 isolated CABGs were performed by 56 surgeons for 11,800 whites (91%), 413 Hispanics (3.2%), and 251 African-Americans (1.9%). White patients were more likely to be treated by surgeons in the top decile than by surgeons in the bottom decile (odds ratio [OR] 1.37, 95% confidence interval [CI] 1.07 to 1.76). In contrast, Hispanic patients were almost 3 times more likely to be treated by surgeons in the bottom decile compared with the top decile (OR 2.85, 95% CI 1.82 to 4.47). Compared with whites, Hispanic patients were about 1/2 as less likely to be treated by surgeons in the top decile (OR 0.51, 95% CI 0.35 to 0.75). African-American and white patients were similarly likely to be treated by surgeons in the top- and bottom-quality performance groups. In conclusion, Hispanics undergoing isolated CABG in Massachusetts were more likely to be operated on by cardiac surgeons with higher risk-standardized mortality rates than by surgeons with lower rates.
CONTEXT: During the last 2 decades, health care professional, consumer, and payer organizations have sought to improve outcomes for patients hospitalized with acute myocardial infarction (AMI). However, little has been reported about improvements in hospital short-term mortality rates or reductions in between-hospital variation in short-term mortality rates.
OBJECTIVE: To estimate hospital-level 30-day risk-standardized mortality rates (RSMRs) for patients discharged with AMI.
DESIGN, SETTING, AND PATIENTS: Observational study using administrative data and a validated risk model to evaluate 3,195,672 discharges in 2,755,370 patients discharged from nonfederal acute care hospitals in the United States between January 1, 1995, and December 31, 2006. Patients were 65 years or older (mean, 78 years) and had at least a 12-month history of fee-for-service enrollment prior to the index hospitalization. Patients discharged alive within 1 day of an admission not against medical advice were excluded, because it is unlikely that these patients had sustained an AMI.
MAIN OUTCOME MEASURE: Hospital-specific 30-day all-cause RSMR.
RESULTS: At the patient level, the odds of dying within 30 days of admission if treated at a hospital 1 SD above the national average relative to that if treated at a hospital 1 SD below the national average were 1.63 (95% CI, 1.60-1.65) in 1995 and 1.56 (95% CI, 1.53-1.60) in 2006. In terms of hospital-specific RSMRs, a decrease from 18.8% in 1995 to 15.8% in 2006 was observed (odds ratio, 0.76; 95% CI, 0.75-0.77). A reduction in between-hospital heterogeneity in the RSMRs was also observed: the coefficient of variation decreased from 11.2% in 1995 to 10.8%, the interquartile range from 2.8% to 2.1%, and the between-hospital variance from 4.4% to 2.9%.
CONCLUSION: Between 1995 and 2006, the risk-standardized hospital mortality rate for Medicare patients discharged with AMI showed a significant decrease, as did between-hospital variation.
BACKGROUND: The first version of The Society of Thoracic Surgeons National Adult Cardiac Surgery Database (STS NCD) was developed nearly 2 decades ago. Since its inception, the number of participants has grown dramatically, patient acuity has increased, and overall outcomes have consistently improved. To adjust for these and other changes, all STS risk models have undergone periodic revisions. This report provides a detailed description of the 2008 STS risk model for coronary artery bypass grafting surgery (CABG).
METHODS: The study population consisted of 774,881 isolated CABG procedures performed on adult patients aged 20 to 100 years between January 1, 2002, and December 31, 2006, at 819 STS NCD participating centers. This cohort was randomly divided into a 60% training (development) sample and a 40% test (validation) sample. The development sample was used to identify predictor variables and estimate model coefficients. The validation sample was used to assess model calibration and discrimination. Model outcomes included operative mortality, renal failure, stroke, reoperation for any cause, prolonged ventilation, deep sternal wound infection, composite major morbidity or mortality, prolonged length of stay (> 14 days), and short length of stay (< 6 days and alive). Candidate predictor variables were selected based on their availability in versions 2.35, 2.41, and 2.52.1 of the STS NCD and their presence in (or ability to be mapped to) version 2.61. Potential predictor variables were screened for overall prevalence in the study population, missing data frequency, coding concerns, bivariate relationships with outcomes, and their presence in previous STS or other CABG risk models. Supervised backwards selection was then performed with input from an expert panel of cardiac surgeons and biostatisticians. After successfully validating the fit of the models, the development and validation samples were subsequently combined, and the final regression coefficients were estimated using the overall combined (development plus validation) sample.
RESULTS: The c-index for the mortality model was 0.812, and the c-indices for other endpoints ranged from 0.653 for reoperation to 0.793 for renal failure in the validation sample. Plots of observed versus predicted event rates revealed acceptable calibration in the overall population and in numerous subgroups. When patients were grouped into categories of predicted risk, the absolute difference between the observed and expected event rates was less than 1.5% for each endpoint. The final model intercept and coefficients are provided.
CONCLUSIONS: New STS risk models have been developed for CABG mortality and eight other endpoints. Detailed descriptions of model development and testing are provided, together with the final algorithm. Overall model performance is excellent.
BACKGROUND: Adjustment for case-mix is essential when using observational data to compare surgical techniques or providers. That is most often accomplished through the use of risk models that account for preoperative patient factors that may impact outcomes. The Society of Thoracic Surgeons (STS) uses such risk models to create risk-adjusted performance reports for participants in the STS National Adult Cardiac Surgery Database (NCD). Although risk models were initially developed for coronary artery bypass surgery, similar models have now been developed for use with heart valve surgery, particularly as the proportion of such procedures has increased. The last published STS model for isolated valve surgery was based on data from 1994 to 1997 and did not include patients undergoing mitral valve repair. STS has developed new valve surgery models using contemporary data that include both valve repair as well as replacement. Expanding upon existing valve models, the new STS models include several nonfatal complications in addition to mortality.
METHODS: Using STS data from 2002 to 2006, isolated valve surgery risk models were developed for operative mortality, permanent stroke, renal failure, prolonged ventilation (> 24 hours), deep sternal wound infection, reoperation for any reason, a major morbidity or mortality composite endpoint, prolonged postoperative length of stay, and short postoperative length of stay. The study population consisted of adult patients who underwent one of three types of valve surgery: isolated aortic valve replacement (n = 67,292), isolated mitral valve replacement (n = 21,229), or isolated mitral valve repair (n = 21,238). The population was divided into a 60% development sample and a 40% validation sample. After an initial empirical investigation, the three surgery groups were combined into a single logistic regression model with numerous interactions to allow the covariate effects to differ across these groups. Variables were selected based on a combination of automated stepwise selection and expert panel review.
RESULTS: Unadjusted operative mortality (in-hospital regardless of timing, and 30-day regardless of venue) for all isolated valve procedures was 3.4%, and unadjusted in-hospital morbidity rates ranged from 0.3% for deep sternal wound infection to 11.8% for prolonged ventilation. The number of predictors in each model ranged from 10 covariates in the sternal infection model to 24 covariates in the composite mortality plus morbidity model. Discrimination as measured by the c-index ranged from 0.639 for reoperation to 0.799 for mortality. When patients in the validation sample were grouped into 10 categories based on deciles of predicted risk, the average absolute difference between observed versus predicted events within these groups ranged from 0.06% for deep sternal wound infection to 1.06% for prolonged postoperative stay.
CONCLUSIONS: The new STS risk models for valve surgery include mitral valve repair as well as multiple endpoints other than mortality. Model coefficients are provided and an online risk calculator is publicly available from The Society of Thoracic Surgeons website.
BACKGROUND: Since 1999, The Society of Thoracic Surgeons (STS) has published two risk models that can be used to adjust the results of valve surgery combined with coronary artery bypass graft surgery (CABG). The most recent was developed from data for patients who had surgery between 1994 and 1997 using operative mortality as the only endpoint. Furthermore, this model did not specifically consider mitral valve repair plus CABG, an increasingly common procedure. Consistent with STS policy of periodically updating and improving its risk models, new models for valve surgery combined with CABG have been developed. These models specifically address both perioperative morbidity and mitral valve repair, and they are based on contemporary data.
METHODS: The final study population consisted of 101,661 procedures, including aortic valve replacement (AVR) plus CABG, mitral valve replacement (MVR) plus CABG, or mitral valve repair (MVRepair) plus CABG between January 1, 2002, and December 31, 2006. Model outcomes included operative mortality, stroke, deep sternal wound infection, reoperation, prolonged ventilation, renal failure, composite major morbidity or mortality, prolonged postoperative length of stay, and short postoperative length of stay. Candidate variables were screened for frequency of missing data, and imputation techniques were used where appropriate. Stepwise variable selection was employed, supplemented by advice from an expert panel of cardiac surgeons and biostatisticians. Several variables were forced into models to insure face validity (eg, atrial fibrillation for the permanent stroke model, sex for all models). Based on preliminary analyses of the data, a single model was employed for valve plus CABG, with indicator variables for the specific type of procedure. Interaction terms were included to allow for differential impact of predictor variables depending on procedure type. After validating the model in the 40% validation sample, the development and validation samples were then combined, and the final model coefficients were estimated using the overall 100% combined sample. The final logistic regression model was estimated using generalized estimating equations to account for clustering of patients within institutions.
RESULTS: The c-index for mortality prediction for the overall valve plus CABG population was 0.75. Morbidity model c-indices for specific complications (permanent stroke, renal failure, prolonged ventilation > 24 hours, deep sternal wound infection, reoperation for any reason, major morbidity or mortality composite, and prolonged postoperative length of stay) for the overall group of valve plus CABG procedures ranged from 0.622 to 0.724, and calibration was excellent.
CONCLUSIONS: New STS risk models have been developed for heart valve surgery combined with CABG. These are the first valve plus CABG models that also include risk prediction for individual major morbidities, composite major morbidity or mortality, and short and prolonged length of stay.
Increasingly, multiple outcomes are collected in order to characterize treatment effectiveness or to evaluate risk factors. These outcomes tend to be correlated because they are measuring related quantities in the same individuals. While the analysis of outcomes measured in the same scale (commensurate outcomes) can be undertaken with standard statistical methods, outcomes measured in different scales (non-commensurate outcomes), such as mixed binary and continuous outcomes, present more difficult challenges.In this paper we contrast some statistical approaches to analyze non-commensurate multiple outcomes. We discuss the advantages of a multivariate method for the analysis of non-commensurate outcomes including situations of missing data. A real data example from a clinical trial, comparing different treatments for depression in low-income women, is used to illustrate the differences between the statistical approaches.
BACKGROUND: Cholinesterase inhibitors are commonly prescribed to treat dementia, but their adverse effect profile has received little attention. These drugs can provoke symptomatic bradycardia and syncope, which may lead to permanent pacemaker insertion. Drug-induced syncope may also precipitate fall-related injuries, including hip fracture.
METHODS: In a population-based cohort study, we investigated the relationship between cholinesterase inhibitor use and syncope-related outcomes using health care databases from Ontario, Canada, with accrual from April 1, 2002, to March 31, 2004. We identified 19 803 community-dwelling older adults with dementia who were prescribed cholinesterase inhibitors and 61 499 controls who were not.
RESULTS: Hospital visits for syncope were more frequent in people receiving cholinesterase inhibitors than in controls (31.5 vs 18.6 events per 1000 person-years; adjusted hazard ratio [HR], 1.76; 95% confidence interval [CI], 1.57-1.98). Other syncope-related events were also more common among people receiving cholinesterase inhibitors compared with controls: hospital visits for bradycardia (6.9 vs 4.4 events per 1000 person-years; HR, 1.69; 95% CI, 1.32-2.15), permanent pacemaker insertion (4.7 vs 3.3 events per 1000 person-years; HR, 1.49; 95% CI, 1.12-2.00), and hip fracture (22.4 vs 19.8 events per 1000 person-years; HR, 1.18; 95% CI, 1.04-1.34). Results were consistent in additional analyses in which subjects were either matched on their baseline comorbidity status or matched using propensity scores.
CONCLUSIONS: Use of cholinesterase inhibitors is associated with increased rates of syncope, bradycardia, pacemaker insertion, and hip fracture in older adults with dementia. The risk of these previously underrecognized serious adverse events must be weighed carefully against the drugs' generally modest benefits.
While much psychiatric research is based on randomized controlled trials (RCTs), where patients are randomly assigned to treatments, sometimes RCTs are not feasible. This paper describes propensity score approaches, which are increasingly used for estimating treatment effects in non-experimental settings. The primary goal of propensity score methods is to create sets of treated and comparison subjects who look as similar as possible, in essence replicating a randomized experiment, at least with respect to observed patient characteristics. A study to estimate the metabolic effects of antipsychotic medication in a sample of Florida Medicaid beneficiaries with schizophrenia illustrates methods.
BACKGROUND: Readmission soon after hospital discharge is an expensive and often preventable event for patients with heart failure. We present a model approved by the National Quality Forum for the purpose of public reporting of hospital-level readmission rates by the Centers for Medicare & Medicaid Services.
METHODS AND RESULTS: We developed a hierarchical logistic regression model to calculate hospital risk-standardized 30-day all-cause readmission rates for patients hospitalized with heart failure. The model was derived with the use of Medicare claims data for a 2004 cohort and validated with the use of claims and medical record data. The unadjusted readmission rate was 23.6%. The final model included 37 variables, had discrimination ranging from 15% observed 30-day readmission rate in the lowest predictive decile to 37% in the upper decile, and had a c statistic of 0.60. The 25th and 75th percentiles of the risk-standardized readmission rates across 4669 hospitals were 23.1% and 24.0%, with 5th and 95th percentiles of 22.2% and 25.1%, respectively. The odds of all-cause readmission for a hospital 1 standard deviation above average was 1.30 times that of a hospital 1 standard deviation below average. State-level adjusted readmission rates developed with the use of the claims model are similar to rates produced for the same cohort with the use of a medical record model (correlation, 0.97; median difference, 0.06 percentage points).
CONCLUSIONS: This claims-based model of hospital risk-standardized readmission rates for heart failure patients produces estimates that may serve as surrogates for those derived from a medical record model.