Baseline NIH stroke scale responses estimate the probability of each particular stroke subtype

BACKGROUND

Emergency treatment of ischemic stroke should ideally be mechanism specific, but acute subtype diagnosis is problematic. Since different subtypes often are associated with specific patterns of neurological deficits, we hypothesize that scores on baseline NIH stroke scale (NIHSS) items may help emergently stratify patients by their probability of having a particular stroke subtype.

METHODS

We performed multivariate polytomous logistic regression analyses on 1,281 patients enrolled in the Trial of ORG 10172 in Acute Stroke Treatment (TOAST). We tested the predictive value of individual items to the baseline NIHSS exam, and syndromic combinations of those items, in anticipating the TOAST stroke subtype at 3 months adjusting for atrial fibrillation. We then used the most significant NIHSS items to construct a predictive model.

RESULTS

The NIHSS items that discriminate between stroke subtypes are language, neglect, visual field and brachial predominance of weakness. Among patients without atrial fibrillation, a normal score for these 4 variables conveys a 46% chance of lacunar stroke, 12% of atherothrombotic stroke and 10% of cardioembolism. This pattern gradually reverses with increased numbers of abnormal responses. Those with abnormalities in all 4 items have a 0.1% chance of lacunar stroke, 50% of atherothrombotic stroke and 39% of cardioembolism.

CONCLUSIONS

Language, neglect, visual fields and brachial predominance of weakness in the baseline NIHSS help discriminate between subtypes, particularly between lacunar and nonlacunar strokes. Clinical trials testing interventions aimed to particular stroke mechanisms may use these NIHSS items to emergently stratify patients based on their probability of having a particular stroke subtype.

Do faculty and resident physicians discuss their medical errors?

BACKGROUND

Discussions about medical errors facilitate professional learning for physicians and may provide emotional support after an error, but little is known about physicians' attitudes and practices regarding error discussions with colleagues.

METHODS

Survey of faculty and resident physicians in generalist specialties in Midwest, Mid-Atlantic and Northeast regions of the US to investigate attitudes and practices regarding error discussions, likelihood of discussing hypothetical errors, experience role-modelling error discussions and demographic variables.

RESULTS

Responses were received from 338 participants (response rate = 74%). In all, 73% of respondents indicated they usually discuss their mistakes with colleagues, 70% believed discussing mistakes strengthens professional relationships and 89% knew at least one colleague who would be a supportive listener. Motivations for error discussions included wanting to learn whether a colleague would have made the same decision (91%), wanting colleagues to learn from the mistake (80%) and wanting to receive support (79%). Given hypothetical scenarios, most respondents indicated they would likely discuss an error resulting in no harm (77%), minor harm (87%) or major harm (94%). Fifty-seven percent of physicians had tried to serve as a role model by discussing an error and role-modelling was more likely among those who had previously observed an error discussion (OR 4.17, CI 2.34 to 7.42).

CONCLUSIONS

Most generalist physicians in teaching hospitals report that they usually discuss their errors with colleagues, and more than half have tried to role-model discussions. However, a significant number of these physicians report that they do not usually discuss their errors and some do not know colleagues who would be supportive listeners.

Internists' attitudes towards terminal sedation in end of life care

OBJECTIVE

To describe the frequency of support for terminal sedation among internists, determine whether support for terminal sedation is accompanied by support for physician assisted suicide (PAS), and explore characteristics of internists who support terminal sedation but not assisted suicide.

DESIGN

A statewide, anonymous postal survey.

SETTING

Connecticut, USA.

PARTICIPANTS

677 Connecticut members of the American College of Physicians.

MEASUREMENTS

Attitudes toward terminal sedation and assisted suicide; experience providing primary care to terminally ill patients; demographic and religious characteristics.

RESULTS

78% of respondents believed that if a terminally ill patient has intractable pain despite aggressive analgesia, it is ethically appropriate to provide terminal sedation (diminish consciousness to halt the experience of pain). Of those who favoured terminal sedation, 38% also agreed that PAS is ethically appropriate in some circumstances. Along a three point spectrum of aggressiveness in end of life care, the plurality of respondents (47%) were in the middle, agreeing with terminal sedation but not with PAS. Compared with respondents who were less aggressive or more aggressive, physicians in this middle group were more likely to report having more experience providing primary care to terminally ill patients (p = 0.02) and attending religious services more frequently (p<0.001).

CONCLUSIONS

Support for terminal sedation was widespread in this population of physicians, and most who agreed with terminal sedation did not support PAS. Most internists who support aggressive palliation appear likely to draw an ethical line between terminal sedation and assisted suicide.

Risk adjusting cesarean delivery rates: a comparison of hospital profiles based on medical record and birth certificate data

OBJECTIVES

Compare the discrimination of risk-adjustment models for primary cesarean delivery derived from medical record data and birth certificate data and determine if the two types of models yield similar hospital profiles of risk-adjusted cesarean delivery rates.

DATA SOURCES/STUDY SETTING

The study involved 29,234 women without prior cesarean delivery admitted for labor and delivery in 1993-95 to 20 hospitals in northeast Ohio for whom data abstracted from patient medical records and data from birth certificates could be linked.

STUDY DESIGN

Three pairs of multivariate models of the risk of cesarean delivery were developed using (1) the full complement of variables in medical records or birth certificates; (2) variables that were common to the two sources; and (3) variables for which agreement between the two data sources was high. Using each of the six models, predicted rates of cesarean delivery were determined for each hospital. Hospitals were classified as outliers if observed and predicted rates of cesarean delivery differed (p < .05).

PRINCIPAL FINDINGS

Discrimination of the full medical record and birth certificate models was higher (p < .001) than the discrimination of the more limited common and reliable variable models. Based on the full medical record model, six hospitals were classified as statistical (p < .01) outliers (three high and three low). In contrast, the full birth certificate model identified five low and four high outliers, and classifications differed for seven of the 20 hospitals. Even so, the correlation between adjusted hospital rates was substantial (r = .71). Interestingly, correlations between the full medical record model and the more limited common (r = .84) and reliable (r = .88) variable birth certificate models were higher, and differences in classification of hospital outlier status were fewer.

CONCLUSION

Birth certificates can be used to develop cesarean delivery risk-adjustment models that have excellent discrimination. However, using the full complement of birth certificate variables may lead to biased hospital comparisons. In contrast, limiting models to data elements with known reliability may yield rankings that are more similar to rankings based on medical record data.

Length of stay as a source of bias in comparing performance in VA and private sector facilities: lessons learned from a regional evaluation of intensive care outcomes

OBJECTIVES

Compare intensive care unit (ICU) mortality and length of stay (LOS) in a VA hospital and private sector hospitals and examine the impact of hospital utilization on mortality comparisons.

RESEARCH DESIGN

Retrospective cohort study.

SUBJECTS

Consecutive ICU admissions to a VA hospital (n = 1,142) and 27 private sector hospitals (n = 51,249) serving the same health care market in 1994 to 1995.

MEASURES

Mortality and ICU LOS were adjusted for severity of illness using a validated method that considers physiologic data from the first 24 hours of ICU admission. Mortality comparisons were made using two different multivariable techniques.

RESULTS

Unadjusted in-hospital mortality was higher in VA patients (14.5% vs. 12.0%; P = 0.01), as was hospital (28.3 vs. 11.3 days; P <0.001) and ICU (4.3 vs. 3.9 days; P <0.001) LOS. Using logistic regression to adjust for severity, the odds of death was similar in VA patients, relative to private sector patients (OR 1.16, 95% CI 0.93-1.44; P = 0.18). However, a higher proportion of VA deaths occurred after 21 hospital days (33% vs. 13%; P <0.001). Using proportional hazards regression and censoring patients at hospital discharge, the risk for death was lower in VA patients (hazard ratio 0.70; 95% CI 0.59-0.82; P <0.001). After adjusting for severity, differences in ICU LOS were no longer significant (P = 0.19).

CONCLUSIONS

Severity-adjusted mortality in ICU patients was lower in a VA hospital than in private sector hospitals in the same health care market, based on proportional hazards regression. This finding differed from logistic regression analysis, in which mortality was similar, suggesting that comparisons of hospital mortality between systems with different hospital utilization patterns may be biased if LOS is not considered. If generalizable to other markets, our findings further suggest that ICU outcomes are at least similar in VA hospitals.

Use of complementary and alternative medicine by older patients with arthritis: a population-based study

OBJECTIVE

To determine the prevalence of complementary and alternative medicine (CAM) use and to identify factors associated with its use in older patients with arthritis.

METHODS

A population-based telephone survey of 480 elderly patients with arthritis was conducted to determine demographics, comorbidities, health status, arthritis symptoms, and the use of CAM and traditional providers and treatments for arthritis.

RESULTS

CAM provider use was reported by 28% of respondents, and 66% reported using one or more CAM treatments. Factors independently related to CAM provider use (P < 0.05) included podiatrist or orthotist use, physician visits for arthritis, and fair or poor self-reported health. For CAM treatments, independent associations were found with physical or occupational therapist use, physician visits for arthritis, chronic obstructive pulmonary disease, and alcohol abstinence. Rural residence, age, income, education, and health insurance type were unrelated to CAM use.

CONCLUSION

Many older patients with arthritis reported seeing CAM providers, and most used CAM treatments. The use of CAM for arthritis was most common among those with poorer self-assessed health and higher use of traditional health care resources.

Use of complementary and alternative medicine by older patients with arthritis: a population-based study

OBJECTIVE

To determine the prevalence of complementary and alternative medicine (CAM) use and to identify factors associated with its use in older patients with arthritis.

METHODS

A population-based telephone survey of 480 elderly patients with arthritis was conducted to determine demographics, comorbidities, health status, arthritis symptoms, and the use of CAM and traditional providers and treatments for arthritis.

RESULTS

CAM provider use was reported by 28% of respondents, and 66% reported using one or more CAM treatments. Factors independently related to CAM provider use (P < 0.05) included podiatrist or orthotist use, physician visits for arthritis, and fair or poor self-reported health. For CAM treatments, independent associations were found with physical or occupational therapist use, physician visits for arthritis, chronic obstructive pulmonary disease, and alcohol abstinence. Rural residence, age, income, education, and health insurance type were unrelated to CAM use.

CONCLUSION

Many older patients with arthritis reported seeing CAM providers, and most used CAM treatments. The use of CAM for arthritis was most common among those with poorer self-assessed health and higher use of traditional health care resources.

Outcomes-based trial of an inpatient nurse practitioner service for general medical patients

Although teaching hospitals are increasingly using nurse practitioners (NPs) to provide inpatient care, few studies have compared care delivered by NPs and housestaff or the ability of NPs to admit and manage unselected general medical patients. In a Midwest academic teaching hospital 381 patients were randomized to general medical wards staffed either by NPs and a medical director or medical housestaff. Data were obtained from medical records, interviews and hospital databases. Outcomes were compared on both an intention to treat (i.e. wards to which patients were randomized) and actual treatment (i.e. wards to which patients were admitted) basis. At admission, patients assigned randomly to NP-based care (n = 193) and housestaff care (n= 188) were similar with respect to demographics, comorbidity, severity of illness and functional parameters. Outcomes at discharge and at 6 weeks after discharge were similar (P>0.10) in the two groups, including: length of stay; charges; costs; consultations; complications; transfers to intensive care; 30-day mortality; patient assessments of care; and changes in activities of daily living, SF-36 scores and symptom severity. However, after randomization, 90 of 193 patients (47%) assigned to the NP ward were actually admitted to housestaff wards, largely because of attending physicians and NP requests. None the less, outcomes of patients admitted to NP and housestaff wards were similar (P>0.1). NP-based care can be implemented successfully in teaching hospitals and, compared to housestaff care, may be associated with similar costs and clinical and functional outcomes. However, there may be important obstacles to increasing the number of patients cared for by NPs, including physician concerns about NPs' capabilities and NPs' limited flexibility in managing varying numbers of patients and accepting off-hours admissions.

Effectiveness of ERCP in cholangitis: a community-based study

BACKGROUND

Although experts have demonstrated the efficacy of endoscopic retrograde cholangiopancreatography (ERCP) in cholangitis, the effectiveness of ERCP in unselected patients has not been measured. The aim was to investigate the clinical impact of ERCP performed at any time and of early ERCP (within 24 hours of admission) in patients with a primary discharge diagnosis of cholangitis.

METHODS

A retrospective record review of patients admitted to eight area hospitals with an International Classification of Diseases (ICD)-9 diagnosis consistent with cholangitis was performed. Extracted data included clinical characteristics, ERCP findings, and patient outcome. The associations of ERCP overall and early ERCP with length of stay were examined. Confounding factors including severity of illness, etiology of cholangitis, and hospital type were adjusted for in a multivariate analysis.

RESULTS

A total of 116 patients were studied. ERCP was performed in 71 patients with endoscopic therapy administered in 57 (80%). ERCP overall was not associated with any change in length of hospital stay. However, compared with other invasive biliary procedures, ERCP was associated with a shorter hospital stay (median 5 vs. 9.5 days, p = 0.01) and a 36% (95% CI [5%, 57%]) reduction in severity-adjusted length of stay. Patients who had early ERCP had a significantly shorter hospital stay than those who had delayed ERCP (median 4 vs. 7 days, p < 0.005) and early ERCP was associated with a 34% (95% CI [11%, 48%]) reduction in severity-adjusted length of stay.

CONCLUSION

Early ERCP may be an effective strategy for shortening the length of stay in patients hospitalized with cholangitis.

Using severity-adjusted mortality to compare performance in a Veterans Affairs hospital and in private-sector hospitals

The objective of this study was to compare hospital mortality in Veterans Affairs (VA) and private-sector patients. The study included 5016 patients admitted to 1 VA hospital. Admission severity of illness was measured using a commercial methodology that was developed in a nationwide database of 850,000 patients from 111 private-sector hospitals. The method uses data abstracted from patients' medical records to predict the risk of death in individual patients, based on the normative database. Analyses compared actual and predicted mortality rates in VA patients. VA patients had higher (P < .05) severity of illness than private-sector patients. The observed mortality rate in VA patients was 4.0% and was similar (P = .09) to the predicted risk of death (4.4%; 95% confidence interval 4.0-4.9%). In subgroup analyses, actual and predicted mortality rates were similar in medical and surgical patients and in groups stratified according to severity of illness, except in the highest severity stratum, in which actual mortality was lower than predicted mortality (57% vs 73%; P < .001). We found that in-hospital mortality in 1 VA hospital and a nationwide sample of private-sector hospitals were similar, after adjusting for severity of illness. Although not directly generalizable to other VA hospitals, our findings nonetheless suggest that the quality of VA and private-sector care may be similar with respect to one important and widely used measure.

Use of intensive care-specific interventions in major teaching and other hospitals: a regional comparison

OBJECTIVE

To compare the use of 40 specific medical interventions in intensive care units (ICUs) of major teaching and other hospitals

DESIGN

Retrospective cohort study.

SETTING

Thirty-eight ICUs in 28 hospitals in a large metropolitan region.

PATIENTS

A total of 12,929 consecutive eligible admissions to medical, surgical, neurologic, or mixed medical/surgical ICUs between January 1, and June 30, 1994.

MEASUREMENTS

The use of 40 diagnostic and therapeutic interventions during the first 24 hrs of ICU admission were obtained from patient medical records and a weighted intervention score was determined for each patient. Admission severity of illness was measured by using the Acute Physiology and Chronic Health Evaluation III methodology.

MAIN RESULTS

Patients at the five teaching hospitals had a greater severity of illness (mean predicted risk of in-hospital death, 15.1%+/-21.9% vs. 11.2%+/-19.0%; p < .01) than patients at the 23 other hospitals. Patients at major teaching hospitals also had higher mean intervention scores (3.5+/-4.9 vs. 2.3+/-3.7; p < .01). Differences in intervention scores persisted after controlling for severity of illness, admission diagnosis, and admission source. However, scores varied among the major teaching hospitals. When examined individually, only three of the five major teaching hospitals had higher (p < .05) interventions scores, compared with other hospitals, whereas one had a lower (p < .05) intervention score.

CONCLUSIONS

Patients in ICUs at major teaching hospitals were, in aggregate, more likely to receive diagnostic and therapeutic interventions than patients at other hospitals. Variation among major teaching hospitals suggests that factors other than teaching status also affect the use of these interventions.

The accuracy of diagnosis and procedural codes for patients with upper GI hemorrhage

BACKGROUND

Little is known about the accuracy of diagnostic and procedural codes for common gastrointestinal (GI) conditions and endoscopic procedures.

METHODS

Eight hundred eighty-two patients with upper GI hemorrhage admitted in 1994 to 1 of 13 regional hospitals were studied. Based on endoscopy reports, the source of hemorrhage, performance of upper endoscopy and use of endoscopic therapy were determined, and we assessed the sensitivity and positive predictive value of discharge codes for measuring the source of hemorrhage and use of upper endoscopy.

RESULTS

The sensitivity and positive predictive value of principal diagnosis coding for source of hemorrhage were typically 85% to 95%. The sensitivity and predictive value of coding for upper endoscopy were 97.7% and 99.9%, respectively, and were 72.3% and 99.4%, respectively, for endoscopic therapy. Accuracy did not differ between the 4 major teaching and 9 other hospitals.

CONCLUSIONS

Hospital-based diagnostic and procedural codes are a reasonably accurate source of data for clinical and outcomes analyses of upper GI hemorrhage. In particular, it is possible to discern from these data the source of hemorrhage and the overall use of upper endoscopy.

Differences in patient-reported processes and outcomes between men and women with myocardial infarction

OBJECTIVE

Previous research measuring differences in the care between men and women with myocardial infarction has focused on differences in procedure use and mortality. However, little is known about differences in processes and outcomes that are reported by patients, such as interpersonal processes of care and health status. Our goal was to measure differences in patient-reported measures for men and women who recently were hospitalized with myocardial infarction.

PARTICIPANTS AND SETTING

We surveyed by mail patients with myocardial infarction discharged to home from one of 27 Cleveland area hospitals 3 months following discharge; 502 (64%) of 783 patients responded. The mean age of subjects was 65 years and 40% were women.

MEASUREMENTS

Process measures included the quality of communication during the hospitalization and at time of discharge and reports of health education discussions during hospitalization. Outcome measures included physical and mental health component scores of the Medical Outcomes Study 36-Item Short-Form Health Survey, change in work status, and days spent in bed because of ill health. We compared processes and outcomes in men and women using multivariate analyses that adjusted for age, other demographic characteristics, comorbid conditions, severity of the myocardial infarction, and premorbid global health status.

MAIN RESULTS

In multivariate analyses, women were as likely as men to report at least one problem with communication during the hospitalization (odds ratio [OR] 0.86; 95% confidence interval [95% CI] 0.56 to 1. 33) or at time of discharge (OR 1.24; 95% CI, 0.82 to 1.89) and to report that they were given dietary advice before discharge (OR 0. 60; 95% CI, 0.36 to 1.01), were told what to do if they developed chest pain (OR 1.21; 95% CI, 0.66 to 2.23), or, if they smoked cigarettes, given advice about how to stop smoking (OR 0.64; 95% CI, 0.26 to 1.58). However, 3 months after discharge, women reported worse physical health (P <.05) and mental health (P <.05), were more likely to report spending time in bed because of ill health (OR 1. 80; 95% CI, 1.06, 3.05), and were more likely to report working less than before their myocardial infarction (OR 4.02; 95% CI, 1.58 to 10. 20).

CONCLUSIONS

In terms of processes of care measured with patient reports, women with myocardial infarction reported their quality of care to be similar to that of men. However, 3 months following myocardial infarction, women reported worse health status and were less likely to return to work than men.

Relationships between in-hospital and 30-day standardized hospital mortality: implications for profiling hospitals

OBJECTIVE

To examine the relationship of in-hospital and 30-day mortality rates and the association between in-hospital mortality and hospital discharge practices.

DATA SOURCES/STUDY SETTING

A secondary analysis of data for 13,834 patients with congestive heart failure who were admitted to 30 hospitals in northeast Ohio in 1992-1994.

DESIGN

A retrospective cohort study was conducted.

DATA COLLECTION

Demographic and clinical data were collected from patients' medical records and were used to develop multivariable models that estimated the risk of in-hospital and 30-day (post-admission) mortality. Standardized mortality ratios (SMRs) for in-hospital and 30-day mortality were determined by dividing observed death rates by predicted death rates.

PRINCIPAL FINDINGS

In-hospital SMRs ranged from 0.54 to 1.42, and six hospitals were classified as statistical outliers (p <.05); 30-day SMRs ranged from 0.63 to 1.73, and seven hospitals were outliers. Although the correlation between in-hospital SMRs and 30-day SMRs was substantial (R = 0.78, p < .001), outlier status changed for seven of the 30 hospitals. Nonetheless, changes in outlier status reflected relatively small differences between in-hospital and 30-day SMRs. Rates of discharge to nursing homes or other inpatient facilities varied from 5.4 percent to 34.2 percent across hospitals. However, relationships between discharge rates to such facilities and in-hospital SMRs (R = 0.08; p = .65) and early post-discharge mortality rates (R = 0.23; p = .21) were not significant.

CONCLUSIONS

SMRs based on in-hospital and 30-day mortality were relatively similar, although classification of hospitals as statistical outliers often differed. However, there was no evidence that in-hospital SMRs were biased by differences in post-discharge mortality or discharge practices.

Variations in risk-adjusted cesarean delivery rates according to race and health insurance

OBJECTIVE

To assess the association between race and insurance and Cesarean delivery rates after adjusting for clinical risk factors that increase the likelihood of cesarean delivery.

DESIGN

Retrospective cohort study in 21 hospitals in northeast Ohio.

SUBJECTS

25,697 women without prior cesarean deliveries admitted for labor and delivery January 1993 through June 1995.

METHODS

Demographic and clinical data were abstracted from patients' medical records. The risk of cesarean delivery was adjusted for 39 maternal and neonatal risk factors that were included in a previously developed risk-adjustment model using nested logistic regression analysis.

MAIN OUTCOME MEASURES

Odds ratios for cesarean delivery in nonwhite patients relative to whites and for patients with government insurance or who were uninsured relative to patients with commercial insurance.

RESULTS

The overall rate of cesarean delivery was similar in white and nonwhite patients (15.8% and 16.1%, respectively), but rates varied (P < 0.001) according to insurance (17.0%, 14.2%, 10.7% in patients with commercial insurance, government insurance, and without insurance, respectively). However, after adjusting for clinical factors, the adjusted odds ratio (OR) of cesarean delivery was higher in nonwhite patients (OR = 1.34; 95% CI: 1.14-1.57; P < 0.001), but similar for patients with government insurance (OR = 1.01; 95% CI: 0.90-1.14; P = 0.84) and lower for uninsured patients (OR = 0.65; 95% CI, 0.41, 1.03; P = 0.067), albeit not statistically significant. In analyses stratified according to quintiles of predicted risk of cesarean delivery, racial differences were largely limited to patients in the lower risk quintiles. However, differences in odds ratios for uninsured patients were seen across the risk quintiles, although odds ratios were not statistically significant.

CONCLUSION

After adjusting for clinical factors, race and insurance status may independently influence the use of cesarean delivery. The higher rates in nonwhites and lower rates in the uninsured may reflect differences in patient preferences or expectations, differences in physician practice, or unmeasured risk factors. The lower odds of cesarean delivery in uninsured women, particularly women at high risk, may raise the issue of underutilization of services and warrants further study.

Increased risk of death in patients with do-not-resuscitate orders

BACKGROUND

Whereas studies have shown higher mortality rates in patients with do-not-resuscitate (DNR) orders, most have not accounted for confounding factors related to the use of DNR orders and/or factors related to the risk of death.

OBJECTIVE

To determine the relationship between the use of DNR orders and in-hospital mortality, adjusting for severity of illness and other covariates.

DESIGN

Retrospective cohort study.

PATIENTS

There were 13,337 consecutive stroke admissions to 30 hospitals in 1991 to 1994.

MEASURES

To decrease selection bias, propensity scores reflecting the likelihood of a DNR order were developed. Scores were based on nine demographic and clinical variables independently related to use of DNR orders. The odds of death in patients with DNR orders were then determined using logistic regression, adjustment for propensity scores, severity of illness, and other factors.

RESULTS

DNR orders were used in 22% (n = 2,898) of patients. In analyses examining DNR orders written at any time during hospitalization, unadjusted in-hospital mortality rates were higher in patients with DNR orders than in patients without orders (40% vs. 2%, P<0.001); the adjusted odds of death was 33.9 (95% CI, 27.4-42.0). The adjusted odds of death remained higher in analyses that only considered orders written during the first 2 days (OR 3.7; 95% CI, 3.2-4.4) or the first day (OR 2.4; 95% CI, 2.0-2.9). In stratified analyses, adjusted odds of death tended to be higher in patients with lower propensity scores.

CONCLUSION

The risk of death was substantially higher in patients with DNR orders after adjusting for propensity scores and other covariates. Whereas the increased risk may reflect patient preferences for less intensive care or unmeasured prognostic factors, the current findings highlight the need for more direct evaluations of the quality and appropriateness of care of patients with DNR orders.

Patient assessments of hospital maternity care: a useful tool for consumers?

OBJECTIVE

To examine three issues related to using patient assessments of care as a means to select hospitals and foster consumer choice-specifically, whether patient assessments (1) vary across hospitals, (2) are reproducible over time, and (3) are biased by case-mix differences.

DATA SOURCES/STUDY SETTING

Surveys that were mailed to 27,674 randomly selected patients admitted to 18 hospitals in a large metropolitan region (Northeast Ohio) for labor and delivery in 1992-1994. We received completed surveys from 16,051 patients (58 percent response rate).

STUDY DESIGN

Design was a repeated cross-sectional study.

DATA COLLECTION

Surveys were mailed approximately 8 to 12 weeks after discharge. We used three previously validated scales evaluating patients' global assessments of care (three items)as well as assessments of physician (six items) and nursing (five items) care. Each scale had a possible range of 0 (poor care) to 100 (excellent care).

PRINCIPAL FINDINGS

Patient assessments varied (p<.001) across hospitals for each scale. Mean hospital scores were higher or lower (p<.01) than the sample mean for seven or more hospitals during each year of data collection. However, within individual hospitals, mean scores were reproducible over the three years. In addition, relative hospital rankings were stable; Spearman correlation coefficients ranged from 0.85 to 0.96 when rankings during individual years were compared. Patient characteristics (age, race, education, insurance status, health status, type of delivery) explained only 2-3 percent of the variance in patient assessments, and adjusting scores for these factors had little effect on hospitals' scores.

CONCLUSIONS

The findings indicate that patient assessments of care may be a sensitive measure for discriminating among hospitals. In addition, hospital scores are reproducible and not substantially affected by case-mix differences. If our findings regarding patient assessments are generalizable to other patient populations and delivery settings, these measures may be a useful tool for consumers in selecting hospitals or other healthcare providers.

Are readmissions to the intensive care unit a useful measure of hospital performance?

BACKGROUND

Although patients readmitted to intensive care units (ICUs) typically have poor outcomes, ICU readmission rates have not been studied as a measure of hospital performance.

OBJECTIVES

To determine variation in ICU readmission rates across hospitals and associations of readmission rates with other ICU-based measures of hospital performance.

RESEARCH DESIGN

Observational cohort study.

SUBJECTS

One hundred three thousand nine hundred eighty four consecutive ICU patients who were admitted to twenty eight hospitals who were then transferred to a hospital ward in those 28 hospitals.

MEASURES

Predicted risk of in-hospital death and ICU length of stay (LOS) were determined by a validated method based on age, ICU admission source, diagnosis, comorbidity, and physiologic abnormalities. Severity-adjusted mortality rates, LOS, and readmission rates were determined for each hospital.

RESULTS

One or more ICU readmissions occurred in 5.8% patients who were initially classified as postoperative and in 6.4% patients who were initially classified as nonoperative. In-hospital mortality rate was 24.7% in patients who were readmitted as compared with 4.0% in other patients (P < 0.001). After adjusting for predicted risk of death, the odds of death remained 7.5 times higher (OR 7.5, 95% CI, 6.8-8.3). Readmitted patients also had longer (P < 0.001) ICU LOS (5.2 vs. 3.7 days) and hospital LOS (29.3 vs. 11.7 days). Severity-adjusted readmission rates varied across hospitals from 4.2% to 7.6%. Readmission rates were not correlated with severity-adjusted hospital mortality, ICU LOS, or hospital LOS.

CONCLUSIONS

ICU patients who were subsequently readmitted have a higher risk of death and longer LOS after adjusting for severity of illness. However, readmission rates were not associated with severity-adjusted mortality or LOS. Those data indicate that ICU readmission may capture other aspects of hospital performance and may be complementary to these measures.

Community-wide assessment of intensive care outcomes using a physiologically based prognostic measure: implications for critical care delivery from Cleveland Health Quality Choice

STUDY OBJECTIVES

To examine the applicability of a previously developed intensive care prognostic measure to a community-based sample of hospitals, and assess variations in severity-adjusted mortality across a major metropolitan region.

DESIGN

Retrospective cohort study.

SETTING

Twenty-eight hospitals with 38 ICUs participating in a community-wide initiative to measure performance supported by the business community, hospitals, and physicians.

PATIENTS

Included in the study were 116,340 consecutive eligible patients admitted to medical, surgical, neurologic, and mixed medical/surgical ICUs between March 1, 1991, and March 31, 1995.

MAIN OUTCOME MEASURES

The risk of hospital mortality was assessed using a previous risk prediction equation that was developed in a national sample, and a reestimated logistic regression model fit to the current sample. The standardized mortality ratio (SMR) (actual/predicted mortality) was used to describe hospital performance.

RESULTS

Although discrimination of the previous national risk equation in the current sample was high (receiver operating characteristic [ROC] curve area = 0.90), the equation systematically overestimated the risk of death and was not as well calibrated (Hosmer-Lemeshow statistic, 2407.6, 8 df, p < 0.001). The locally derived equation had similar discrimination (ROC curve area = 0.91), but had improved calibration across all ranges of severity (Hosmer-Lemeshow statistic = 13.5, 8 df, p = 0.10). Hospital SMRs ranged from 0.85 to 1.21, and four hospitals had SMRs that were higher or lower (p < 0.01) than 1.0. Variation in SMRs tended to be greatest during the first year of data collection. SMRs also tended to decline over the 4 years (1.06, 1.02, 0.98, and 0.94 in years 1 to 4, respectively), as did mean hospital length of stay (13.0, 12.4, 11.6, and 11.1 days in years 1 to 4; p < 0.001). However, excluding the increasing (p < 0.001) number of patients discharged to skilled nursing facilities attenuated much of the decline in standardized mortality over time.

CONCLUSIONS

A previously validated physiologically based prognostic measure successfully stratified patients in a large community-based sample by their risk of death. However, such methods may require recalibration when applied to new samples and to reflect changes in practice over time. Moreover, although significant variations in hospital standardized mortality were observed, changing hospital discharge practices suggest that in-hospital mortality may no longer be an adequate measure of ICU performance. Community-wide efforts with broad-based support from business, hospitals, and physicians can be sustained over time to assess outcomes associated with ICU care. Such efforts may provide important information about variations in patient outcomes and changes in practice patterns over time. Future efforts should assess the impact of such community-wide initiatives on health-care purchasing and institutional quality improvement programs.

The relationship of gender and in-hospital death: increased risk of death in men

OBJECTIVES

The prognostic importance of gender in hospitalized patients has been poorly studied. The current study compared in-hospital death rates between men and women after adjusting for severity of illness.

DESIGN

Retrospective cohort study.

PATIENTS

89,793 eligible patients with 6 common nonsurgical diagnoses who were discharged from 30 hospitals in Northeast Ohio in 1991 to 1993.

METHODS

Admission severity of illness (ie, predicted risk of death) was calculated using multivariable models that were based on data abstracted from patients' clinical records (ROC curve areas, 0.83-0.90). In hospital death rates were then adjusted for predicted risks of death and other covariates using logistic regression analysis.

RESULTS

Adjusted odds of death were higher (P < 0.05) in men, compared with women, for 4 diagnoses (stroke [OR, 1.60]; obstructive airway disease [OR, 1.38]; gastrointestinal hemorrhage [OR 1.32]; pneumonia [OR, 1.18]) and similar for two diagnoses (congestive heart failure [OR, 1.12]; and acute myocardial infarction [OR, 0.97]). These differences were somewhat attenuated by excluding patients discharged to skilled nursing facilities or other hospitals from analysis; nonetheless, the odds of death in men remained higher for 3 diagnoses.

CONCLUSIONS

The findings indicate that inhospital death rates are generally higher in men than in women, after adjusting for severity of illness. In addition, the risk of in-hospital death in men and women was influenced by diagnosis. These differences may reflect gender-related variation in the utilization of hospital services, the effectiveness of care, over- or underestimation of severity of illness, or biological differences in men and women.

Early endoscopy in upper gastrointestinal hemorrhage: associations with recurrent bleeding, surgery, and length of hospital stay

BACKGROUND

The impact of upper endoscopy in patients with upper gastrointestinal hemorrhage treated in community practice is unknown. Thus we examined the effectiveness of endoscopy performed within 24 hours of admission (early endoscopy).

METHODS

Medical records of 909 consecutive hospitalized patients with upper gastrointestinal hemorrhage who underwent endoscopy at 13 hospitals in a large metropolitan area were reviewed. We evaluated unadjusted and severity-adjusted associations of early endoscopy with recurrent bleeding or surgery to control hemorrhage, length of hospital stay, and associations of endoscopic therapy in patients with bleeding ulcers or varices.

RESULTS

Early endoscopy was performed in 64% of patients and compared with delayed endoscopy and was associated with clinically significant reductions in adjusted risk of recurrent bleeding or surgery (odds ratio [OR] 0.70: 95% CI [0.44, 1.13]) and a 31% decrease in adjusted length of stay (95% CI: [24%, 37%]). In patients at high risk for recurrent bleeding, the use of early endoscopic therapy to control hemorrhage was associated with reductions in recurrent bleeding or surgery (OR 0.21: 95% CI [0.10, 0.47]) and length of stay (-31%: 95% CI [-44%, -14%).

CONCLUSION

In this study of community-based practice, the routine use of endoscopy, and in selected cases endoscopic therapy, performed early in the clinical course of patients with upper gastrointestinal hemorrhage was associated with reductions in length of stay and, possibly, the risk of recurrent bleeding and surgery.

Racial variation in the use of do-not-resuscitate orders

OBJECTIVE

To compare the use of do-not-resuscitate (DNR) orders in African-American and white patients using a large, multisite, community-based sample.

MEASUREMENTS

Our sample included 90,821 consecutive admissions to 30 hospitals in a large metropolitan region with six nonsurgical conditions from 1993 through 1995. Demographic and clinical data were abstracted from medical records. Admission severity of illness was measured using multivariate risk-adjustment models with excellent discrimination (receiver-operating characteristic curve areas, 0.82-0.88). Multiple logistic regression analysis was used to determine the independent association between race and use of DNR orders, adjusting for age, admission severity, and other covariates.

MAIN RESULTS

In all patients, the rate of DNR orders was lower in African Americans than whites (9% vs 18%; p <.001). Rates of orders were also lower ( p <. 001) among African Americans in analyses stratified by age, gender, diagnosis, severity of illness, and in-hospital death. After adjusting for severity and other important covariates, the odds of a DNR order remained lower ( p <.001) for African-Americans relative to whites for all diagnoses, ranging from 0.38 for obstructive airway disease to 0.71 for gastrointestinal hemorrhage. Results were similar in analyses limited to orders written by the first, second, or seventh hospital day. Finally, among patients with DNR orders, African Americans were less likely to have orders written on the first hospital day and more likely to have orders written on subsequent days.

CONCLUSIONS

The use of DNR orders was substantially lower in African Americans than in whites, even after adjusting for severity of illness and other covariates. Identification of factors underlying such differences will improve our understanding of the degree to which expectations for care differ in African American and white patients.

Does length of hospital stay during labor and delivery influence patient satisfaction? Results from a regional study

OBJECTIVE

To examine the relationship between patients' satisfaction with hospital obstetric care, length of stay, and patients' perceived appropriateness of the length of stay.

STUDY DESIGN

A cross-sectional study.

PATIENTS AND METHODS

We surveyed 27,789 women (a 58% response rate) discharged after labor and delivery from 18 hospitals in a large metropolitan region from 1992 through 1994. Patient satisfaction was assessed using the Patient Judgment System, a previously validated instrument. Our analysis focused on four scales evaluating specific aspects of care (physician care, nursing care, provision of information, and preparation for discharge) and two single-item indicators of satisfaction (overall quality and willingness to return to the hospital).

RESULTS

Patients with shorter lengths of stay were more likely (P < 0.001) to perceive their stays as "too short." In addition, the six measures of satisfaction were lower (P < 0.001) in patients who perceived their stays as too short. However, the hypothesized lower satisfaction in patients with shorter stays was not observed; differences in satisfaction according to length of stay were small and of questionable practical significance.

CONCLUSION

The findings suggest that patients' satisfaction with obstetric care may not depend on the absolute duration of stay but rather on whether patients perceive the length of stay to be adequate. The results are timely because of recent legislation that mandates minimum hospital stays for labor and delivery.

Endoscopic practice for upper gastrointestinal hemorrhage: differences between major teaching and community-based hospitals

BACKGROUND

Differences in endoscopic practice in major teaching and community hospitals are not known.

METHODS

A total of 1031 consecutive patients discharged from 13 hospitals (4 major teaching, 9 others) in 1994 with upper gastrointestinal hemorrhage were studied. Data obtained from chart abstraction included endoscopic findings and therapy and selected outcomes. Multivariable analyses adjusted for admission severity of illness and endoscopic findings.

RESULTS

Rates of endoscopy were similar between patients admitted to major teaching and other hospitals, although procedures to control hemorrhage were used more often in major teaching hospitals (35% vs. 19%, p < 0.001). Use of endoscopic therapy was higher in major teaching hospitals for lesions in which therapy is recommended, as well as other lesions. Recurrent bleeding was also more common in major teaching hospitals (14.3% vs. 7.8%, p = 0.001), and the difference persisted in multivariable analysis (odds ratio 1.69: 95% CI [1.09 to 2.64], p = 0.02). Unadjusted and adjusted length of stay were somewhat shorter in major teaching hospitals.

CONCLUSIONS

There was large variation in the use of endoscopic therapy, with higher rates observed in major teaching hospitals for lesions in which therapy is recommended, as well as other stigmata. Further studies are needed to better define the reasons for the practice variation and to assess the impact on other outcomes such as readmission and costs.

Pediatric bone marrow cellularity: are we expecting too much?

PURPOSE

Accurate assessment of marrow cellularity is necessary for establishing diagnoses and monitoring the effects of treatment in a large number of malignant and nonmalignant pediatric illnesses, and for evaluating sibling donors for transplantation. However, normal values for age-related bone marrow cellularity in pediatric patients have not been well established. This study was designed to better define pediatric normal values for bone marrow cellularity.

PATIENTS AND METHODS

A retrospective review of 448 bone marrow core biopsy or clot specimens, including 45 samples from healthy donors, were taken from the posterior iliac crest of patients aged from younger than 1 to 18 years (55% male). All samples were collected and fixed in a standardized fashion. Patients with hematopoietic malignancies and other systemic conditions known to impact marrow cellularity were excluded.

RESULTS

The mean cellularity of the entire sample was 65.4%. Cellularity was similar in boys and girls, but varied (p < 0.001) with age. Cellularity was highest in patients younger than 2 years (79.8%), and declined in patients 2 to 4 years old (68.6%) and 5 to 9 years old (59.1%). Cellularity remained stable in older patients (60.1% and 61.1%, respectively, in patients 10 to 14 and 15 to 18 years of age). Adjusting for age and gender, mean cellularity was similar in patients with an underlying nonhematologic malignancy compared to health donors but was roughly 6% higher in patients with hematopoietic disorders.

CONCLUSIONS

This study demonstrates that average cellularity during the first two decades of life, using current techniques of marrow collection and standardized analysis, is lower than previously estimated. In addition, cellularity declined with age until the age of 5 years, but was similar thereafter. After adjusting for age, differences according to diagnosis were relatively small.

What patients should ask of consumers' guides to health care quality

Consumers' guides that profile the quality of care of individual health care providers may be influential in shaping health care markets. We propose four simple questions that can be used to evaluate such guides: (a) Does the guide measure distinct and important domains of health care quality? (b) Are the individual measures of quality described simply and precisely? (c) Do the measures take into account relevant differences between patients? (d) Are the ratings of quality presented fairly? Using these four questions, we examine the validity of one prominent guide that annually identifies America's best hospitals and present a set of recommendations for the design of future guides. Although the evaluation of health care quality is undoubtedly complex, the four questions that we pose provide a basis for developing a more rational approach to informing the public about health care quality.

Associations between the use of do-not-resuscitate orders and length of stay in patients with stroke

OBJECTIVES

The study sought to describe the association between do-not-resuscitate (DNR) orders and length of hospital stay (LOS), and how the association varies according to in-hospital mortality, timing of the DNR order, and admission severity of illness.

METHODS

The authors conducted a retrospective cohort analysis involving standardized review of patients' medical records. The study was performed at 30 acute care hospitals in a large metropolitan area. The authors studied the data of 13,337 consecutive patients with a primary diagnosis of stroke discharged in 1991 through 1994.

RESULTS

Do-not-resuscitate orders were written for 22% (n = 2,898) of the sample. In all patients, mean LOS was longer in patients with DNR orders than in patients without orders (12.0 versus 9.5 days; P < 0.001). A series of Cox regression analyses were performed to adjust LOS for admission severity of illness and other covariates. In analyses of patients discharged alive (n = 12,011), LOS was similar in patients with DNR orders written on days 1 to 2 compared with patients without DNR orders. However, LOS was longer in patients with DNR orders written on days 3 to 7 (Hazard Ratio [HR], 1.59; 95% CI, 1.43-1.77) and on day 8 or later (HR, 2.72; 95% CI, 2.34-3.16). In analyses of patients who died (n = 1,326), LOS was shorter for patients with DNR orders written on days 1 and 2 (HR, 0.59; 95% CI, 0.49-0.71) than for patients without DNR orders but was longer among patients with DNR orders written on day 8 or later (HR, 2.58; 95% CI, 2.06-3.22). In analyses stratified by admission severity, the relative effect of a DNR order tended to be less in patients with higher severity.

CONCLUSIONS

The relationship between DNR orders and LOS is complex and varies according to in-hospital mortality, the timing of the DNR order, and admission severity of illness. These findings highlight the importance of explicitly accounting for such factors in studies evaluating the implications of DNR orders on the costs of hospital care.

Patient and hospital characteristics associated with patient assessments of hospital obstetrical care

OBJECTIVES

The goals of this study were to examine the relationship of patient assessments of hospital care with patient and hospital characteristics. In addition, the authors sought to assess relationships between patient assessments and other patient-derived measures of care (eg, how much they were helped by the hospitalization and amount of pain experienced).

METHODS

The authors surveyed 16,051 women (response rate, 58%) discharged after labor and delivery from 18 hospitals during the study period of 1992 to 1994. Patient assessments were obtained using a previously validated survey instrument, Patient Judgment of Hospital Quality, that includes eight scales assessing different aspects of the process of care (eg, physician care, discharge procedures) and other single item assessments (eg, overall quality). For this study, we utilized five of the scales (physician care, nursing care, information, discharge preparation, global assessments [willingness to brag, recommend or return to the hospital]). For analysis, items were rated on a five-point ordinal scale from poor to excellent. For scoring purposes, responses were transformed to linear ratings, ranging from 0 to 100 (eg, 0 = poor care, 100 = excellent care).

RESULTS

In multivariable analyses, the authors found that patients who were older, white, not married, uninsured or had commercial insurance, and in better health status were significantly more likely to give higher assessments (P < 0.01), although very little of the variance in assessment scores was explained by these characteristics (2%-3%). In bivariate analyses, patient assessments were higher in nonteaching hospitals and those with fewer beds, fewer deliveries, lower cesarean-section (C-section) rates, fewer patients with Medicaid, and higher rates of vaginal births after C-section deliveries. When these variables were utilized as independent predictors in multivariable analyses using adjusted nested linear regression (to account for clustering of patients), few of the hospital characteristics reached a level of statistical significance. Finally, correlations between the five scales and other patient assessments of quality, such as how much they were helped by the hospitalization, were statistically significant (P < 0.01) and high in magnitude, ranging from 0.47 to 0.61.

CONCLUSIONS

Although hospital scores differed according to several patient and hospital characteristics, the magnitude of the associations was relatively small. The findings suggest that, with respect to obstetric care, patient assessments may represent a robust measure that can be applied to diverse hospitals and patient casemix.

Variations in standardized hospital mortality rates for six common medical diagnoses: implications for profiling hospital quality

OBJECTIVES

The authors determined whether standardized hospital mortality rates varied for six common medical diagnoses.

METHODS

The retrospective cohort study included 89,851 patients aged 18 years and older discharged from 30 hospitals in a large metropolitan area in 1991 to 1993 with a principal diagnosis of acute myocardial infarction, congestive heart failure, pneumonia, stroke, obstructive lung disease, or gastrointestinal hemorrhage. For each hospital, standardized mortality ratios (observed/predicted mortality) were determined using validated risk-adjustment models that were based on clinical data elements abstracted from patients' hospital records. Hospitals also were categorized into quintiles on the basis of standardized mortality ratios. Correlations between standardized mortality ratios and agreement between quintile rankings were determined for each pair of diagnoses.

RESULTS

Correlations between hospital-standardized mortality ratios for individual diagnoses were generally weak. For the 15 possible pairs of diagnoses, Pearson coefficients ranged from -0.10 to 0.43; only six were 0.30 or greater. Agreement between hospital quintile rankings was also generally low, with weighted kappa values ranging from -0.12 to 0.42. Three of 15 kappa values were less than 0 (ie, agreement lower than chance), and only four exceeded 0.20, the threshold for "fair" agreement. Although simulated analyses found that random variation and relatively low hospital volumes accounted for some of the difference in standardized mortality ratios for diagnoses, a large proportion of the difference remained unexplained.

CONCLUSIONS

Standardized hospital mortality rates varied for six diagnoses that likely are managed by similar practitioners. Although variability may be decreased by restricting analyses to hospitals with large volumes, the findings indicate that for many hospitals, diagnosis-specific mortality rates may be an inconsistent measure of hospital quality, even when data are aggregated for multiple years.

Using hospital performance data in quality improvement: the Cleveland Health Quality Choice experience

BACKGROUND

Cleveland Health Quality Choice is a regional initiative to assess hospital performance which was implemented in 1989. The project developed and validated CHOICE, a severity adjustment system that includes diagnosis-specific models for medical, surgical, and obstetrical patients which are based on clinical data abstracted from patients' medical records.

METHODOLOGY

Since 1992 Cleveland Health Quality Choice has disseminated semi-annual reports that profile hospital mortality rates, lengths of stay, and cesarean section rates using the CHOICE severity adjustment models. Hospitals receive tabular and graphical representations of hospital outcomes and electronic patient-level data files that can be used to further examine outcomes in clinical subgroups.

RESULTS

Four case studies illustrate how outcomes data derived from the CHOICE models led to the development of successful hospital programs to decrease lengths of stay, cesarean section rates, and hospital mortality rates. Although each case study reflected a unique approach to process improvement, several common characteristics were observed: (1) establishment of interdisciplinary process improvement teams with senior physician and nursing leadership; (2) detailed review of the process of care to identify modifiable clinical practices likely to affect outcomes; (3) development of practice guidelines based on group consensus or published recommendations that were designed to affect modifiable practices; and (4) aggressive sharing of serial data with individual practitioners.

CONCLUSIONS

Although outcomes data can provide powerful insight on where to target quality improvement efforts, hospitals must identify influential and modifiable clinical practices. Such efforts are most likely to be successful if driven by interdisciplinary work groups, supported by senior clinicians and administrators, and based on locally accepted practice standards.

Impact of risk-adjusting cesarean delivery rates when reporting hospital performance

CONTEXT

Hospitals and health plans are often ranked on rates of cesarean delivery, under the assumption that lower rates reflect more appropriate, more efficient care. However, most rankings do not account for patient factors that affect the likelihood of cesarean delivery.

OBJECTIVE

To compare hospital cesarean delivery rates before and after adjusting for clinical risk factors that increase the likelihood of cesarean delivery.

DESIGN

Retrospective cohort study.

SETTING

Twenty-one hospitals in northeast Ohio.

PATIENTS

A total of 26127 women without prior cesarean deliveries admitted for labor and delivery from January 1993 through June 1995.

MAIN OUTCOME MEASURES

Hospital rankings based on observed and risk-adjusted cesarean delivery rates.

RESULTS

The overall cesarean delivery rate was 15.9% and varied (P<.001) from 6.3% to 26.5% in individual hospitals. Adjusted rates varied from 8.4% to 22.0%. The correlation between unadjusted and adjusted hospital rankings (ie, 1-21) was only modest (R=0.35, P=.12). Whereas 7 hospitals were classified as outliers (ie, had rates higher or lower [P<.05] than overall rate) on the basis of both unadjusted and adjusted rates, outlier status changed for 5 hospitals (24%), including 2 that changed from outliers to nonoutliers, 2 that changed from nonoutliers to outliers, and 1 that changed from a high outlier to a low outlier.

CONCLUSIONS

Cesarean delivery rates varied across hospitals in a single metropolitan region. However, rankings that fail to account for clinical factors that increase the risk of cesarean delivery may be methodologically biased and misleading to the public.

Use of intensive care units for patients with low severity of illness

OBJECTIVE

To determine variations among hospitals in use of intensive care units (ICUs) for patients with low severity of illness.

DESIGN

Retrospective cohort study.

SETTING

Twenty-eight hospitals with 44 ICUs in a large metropolitan region.

PATIENTS

Consecutive eligible patients (N=104,487) admitted to medical, surgical, neurological, or mixed medical-surgical ICUs from March 1, 1991, to March 31, 1995.

OUTCOME MEASURES

The predicted risk of in-hospital death for each patient was assessed using a validated method that is based on age, ICU admission source, diagnosis, severe comorbid conditions, and abnormalities in 17 physiologic variables. Admissions were classified as low severity if the patient's predicted risk of death was less than 1%. In a subset of 12,929 consecutive patients, use of 19 specific interventions typically delivered in ICUs was examined.

RESULTS

Twenty thousand four hundred fifty-one admissions (19.6%) were categorized as low severity, including 23.6% of postoperative and 16.9% of nonoperative admissions. Alcohol and other drug overdoses accounted for 40.2% of nonoperative low-severity admissions; laminectomy and carotid endarterectomy accounted for 52.3% of postoperative low-severity admissions. Mortality among patients with low-severity illness was 0.3%, and only 28.6% received an ICU-specific intervention during the first ICU day. Although mean ICU length of stay was shorter (P<.001) in low-severity admissions (2.2 vs 4.7 days in nonoperative and 2.4 vs 4.2 days in postoperative admissions), low-severity admissions accounted for 11.1% of total ICU bed days. Rates of low-severity admissions varied (P<.001) across hospitals, ranging from 5% to 27% for nonoperative and 9% to 68% for postoperative admissions.

CONCLUSIONS

A large proportion of patients admitted to the ICU have a low probability of death and do not receive ICU-specific interventions. Rates of low-severity admissions varied among hospitals. The development and implementation of protocols to target ICU care to patients most likely to benefit may decrease the number of low-severity ICU admissions and improve the cost-effectiveness of ICU care.

The effectiveness of early endoscopy for upper gastrointestinal hemorrhage: a community-based analysis

OBJECTIVES

The effectiveness of upper endoscopy in unselected patients with upper gastrointestinal hemorrhage has not been well studied. This study was undertaken to identify factors associated with the performance of early endoscopy (ie, within 1 day of hospitalization) and, after adjusting for these factors, to determine associations between early endoscopy and in-hospital mortality, length of stay, and performance of surgery.

METHODS

Subjects in this observational cohort study were 3,801 consecutive admissions with upper gastrointestinal hemorrhage to 30 hospitals in a large metropolitan region. Demographic and clinical data were abstracted from hospital records. A multivariable model based on factors that potentially could relate to the decision to perform endoscopy was developed to determine the propensity (0 to 100%) for early endoscopy in each patient.

RESULTS

Early endoscopy was performed in 2,240 patients (59%), and although it was not associated with mortality after adjusting for severity of illness among all patients, it was associated with a higher risk of death for patients in the lowest propensity group. Early endoscopy was associated with a lower likelihood of upper gastrointestinal surgery in all patients and in the two highest propensity groups and with a shorter length of stay in the entire cohort and in all subgroups.

CONCLUSIONS

In the absence of specific contraindications, early endoscopy should be considered because of associated reductions in length of stay and surgical intervention. Further studies are needed to identify subgroups in whom the procedure may be associated with adverse effects on survival.

The relation between health status changes and patient satisfaction in older hospitalized medical patients

OBJECTIVE

To examine the relation between two patient outcome measures that can be used to assess the quality of hospital care: changes in health status between admission and discharge, and patient satisfaction.

DESIGN

Prospective cohort study.

SETTING AND PATIENTS

Subjects were 445 older medical patients (aged > or =70 years) hospitalized on the medical service of a teaching hospital.

MEASUREMENTS AND MAIN RESULTS

We interviewed patients at admission and discharge to obtain two measures of health status: global health and independence in five activities of daily living (ADLs). At discharge, we also administered a 5-item patient satisfaction questionnaire. We assessed the relation between changes in health status and patient satisfaction in two sets of analyses, that controlled for either admission or discharge health status. When controlling for admission health status, changes in health status between admission and discharge were positively associated with patient satisfaction (p values ranging from .01 to .08). However, when controlling for discharge health status, changes in health status were no longer associated with patient satisfaction. For example, among patients independent in ADLs at discharge, mean satisfaction scores were similar regardless of whether patients were dependent at admission (i.e., had improved) or independent at admission (i.e., remained stable) (79.6 vs 81.2, p = .46). Among patients dependent in ADLs at discharge, mean satisfaction scores were similar regardless of whether they were dependent at admission (i.e., remained stable) or independent at admission (i.e., had worsened) (74.0 vs 75.7, p = .63). These findings were similar using the measure of global health and in multivariate analyses.

CONCLUSIONS

Patients with similar discharge health status have similar satisfaction regardless of whether that discharge health status represents stable health, improvement, or a decline in health status. The previously described positive association between patient satisfaction and health status more likely represents a tendency of healthier patients to report greater satisfaction with health care, rather than a tendency of patients who improve following an interaction with the health system to report greater satisfaction. This suggests that changes in health status and patient satisfaction are measuring different domains of hospital outcomes and quality. Comprehensive efforts to measure the outcomes and quality of hospital care will need to consider both patient satisfaction and changes in health status during hospitalization.

The use of patient perceptions in the evaluation of health-care delivery systems

OBJECTIVES

Patient perceptions are increasingly used to measure quality of care in a diversity of health-care delivery settings. The goals of this article are to review the current use of patient perceptions and to review what is known about the sensitivity of patient perceptions for discerning variations in care across delivery systems.

METHODS

This article first provides a rationale for using patient perceptions to evaluate delivery systems and reviews proposed frameworks for measuring perceptions. It then reviews illustrative studies that have used patient perceptions to compare delivery systems or that have examined associations between patient perceptions and other health-care indicators.

RESULTS

Although the results of these studies suggest some general relations between patient perceptions and characteristics of delivery systems, findings are often inconsistent across individual studies. These inconsistencies may be related to several potential methodological limitations, including failure to account for the impact of patient mix, ceiling effects of patient responses, nonresponse bias, differences in data collection methods and timing of surveys, use of proxy respondents, and differences in survey instruments.

CONCLUSIONS

The discussion concludes with five conceptual challenges and recommendations for further research: (1) to establish the sensitivity of patient perceptions for discerning differences across delivery systems; (2) to establish relations between alternative frameworks for measuring patient perceptions; (3) to standardize the measurement of patient perceptions; (4) to define optimal ways of presenting patient perceptions data to users; and (5) to broaden the "patient" populations in which perceptions of care have been measured.

Variation in the use of do-not-resuscitate orders in patients with stroke

OBJECTIVES

To identify sociodemographic and clinical characteristics associated with the use of do-not-resuscitate (DNR) orders in hospitalized patients with stroke. To examine whether the use of DNR orders varies across hospitals.

METHODS

This observational cohort study used data collected for 13337 consecutive eligible patients with a primary diagnosis of stroke. These patients were discharged in 1991 through 1994 from 30 hospitals in a large metropolitan area. Study data were abstracted from patients' hospital records using standard forms. Admission severity of illness was measured using a validated multivariable model. Sociodemographic and clinical factors independently associated with the use of DNR orders were identified using stepwise logistic regression.

RESULTS

Do-not-resuscitate orders were written for 2898 patients (22%). Patient characteristics independently (P < .01) associated with increased use of DNR orders included increasing age (odds ratio [OR], 1.06 per year); admission from a skilled nursing facility (OR, 2.44) or through the emergency department (OR, 1.49); cancer (OR, 2.73), intracerebral hemorrhage (OR, 2.12), coma (OR, 7.47), or lethargy or stupor on admission neurological assessment (OR, 3.38); and increasing admission severity (OR; 1.29 per decile). In contrast, African American race was associated with lower use of DNR orders (OR, 0.54). Although substantial variation in the use of DNR orders was observed across hospitals, with rates ranging from 12% to 32%, adjusting for the above patient characteristics eliminated much of this variation, including differences between major teaching and other hospitals and between hospitals with and without religious affiliations.

CONCLUSIONS

In our community-based analysis of patients with stroke, the use of DNR orders was common and was strongly related to several patient characteristics. These factors explained much of the variation across hospitals. While our analysis did not account for differences in patient preferences for treatment, the differences we observed in the use of DNR orders across sociodemographic groups are suggestive of variations in care and may have important implications for the cost and quality of hospital care.

Severity-adjusted mortality and length of stay in teaching and nonteaching hospitals. Results of a regional study

CONTEXT

Major teaching hospitals are perceived as being more expensive than other hospitals and, thus, unattractive to managed care. However, little empirical data exist about their relative quality and efficiency. The current study compared severity-adjusted mortality and length of stay (LOS) in teaching and nonteaching hospitals.

DESIGN

Retrospective cohort study.

SETTING

Thirty hospitals in northeast Ohio.

PATIENTS

A total of 89851 consecutive eligible patients discharged in 1991 through 1993 with myocardial infarction, congestive heart failure, obstructive airway disease, gastrointestinal hemorrhage, pneumonia, or stroke.

MAIN OUTCOME MEASURES

In-hospital mortality and LOS of patients in major teaching (n=5), minor teaching (n=6), and nonteaching (n=19) hospitals were adjusted for admission severity of illness using multivariable models based on demographic and clinical data abstracted from patients' medical records.

RESULTS

The adjusted odds of death was 19% lower (95% confidence interval [CI], 2%-34%; P=.03) for patients in major teaching hospitals compared with non-teaching hospitals but was similar (95% CI, 7% lower to 28% higher; P=.28) for patients in minor teaching hospitals. The findings were generally consistent in analyses stratified according to diagnosis, age, race, predicted risk of death, and other covariates. In addition, risk-adjusted LOS was 9% lower (95% CI, 8%-10%; P<.001) among patients in major teaching hospitals relative to nonteaching hospitals but was similar (95% CI, 2% lower to 11% higher; P=.17) in minor teaching hospitals. Major teaching hospitals also cared for higher proportions of nonwhite and poorly insured patients.

CONCLUSIONS

Risk-adjusted mortality and LOS were lower for patients in major teaching hospitals than for patients in minor teaching and nonteaching hospitals. If generalizable to other regions, the results provide evidence that hospital performance, as assessed by 2 commonly used indicators, may be higher in major teaching hospitals. These findings are noteworthy at a time when the viability of many major teaching hospitals is threatened by powerful health care market forces and by potential changes in federal financing of graduate medical education.

Lack of gender and racial differences in surgery and mortality in hospitalized Medicare beneficiaries with bleeding peptic ulcer

OBJECTIVE

Determine the relation of race and gender to outcome from bleeding peptic ulcer.

DESIGN

Retrospective cohort study.

SETTING

All acute care hospitals in the United States.

PATIENTS

A 100% sample of hospitalized Medicare beneficiaries older than 64 years (n = 82,868) with a primary discharge diagnosis of peptic ulcer with hemorrhage.

MEASUREMENTS AND MAIN RESULTS

Surgical treatment was performed in 6.9% of patients, 30-day mortality was 8.5%, and average length of stay was 9.4 days. Surgery was somewhat more common in men than women (7.3% vs 6.5%, p < .001), and in whites than African Americans (6.9% vs 6.3%, p < .001), but neither race nor gender was associated with surgery in multivariable analysis adjusting for potentially confounding factors. Mortality rates were similar in African Americans and whites (8.5%), and somewhat higher in men than women (10.7% vs 9.3%, p < .001). In multivariable analysis, there was no difference in mortality across gender and racial groups. Although unadjusted and adjusted lengths of stay were longer for African Americans and shorter for men, the differences were modest (i.e., 16% increase and 6% decrease in multivariable analysis, respectively, p < .0001).

CONCLUSIONS

In this national sample, there is no significant gender or racial difference in therapy and outcome for patients with hemorrhagic peptic ulcer. The findings raise the possibility that studies that have shown race and gender differences in management of coronary artery disease and cancer may not be generalizable to other common diagnoses.

Declines in hospital mortality associated with a regional initiative to measure hospital performance

To determine changes in hospital mortality that occurred in association with the dissemination of data by a regional initiative to profile hospital performance, we conducted a retrospective cohort study of patients admitted before and subsequent to dissemination of comparative data in 1992. The analysis included 101,060 consecutive eligible discharges from 30 hospitals in Northeast Ohio with eight diagnoses: acute myocardial infarction, congestive heart failure (CHF), obstructive airway disease, gastrointestinal hemorrhage, pneumonia, stroke, coronary artery bypass surgery, and lower bowel resection. Baseline (1991, N = 35,629) mortality rates were compared to rates during three subsequent periods (July-December 1992, N = 20,392; January-June 1993, N = 23,070; and July-December 1993, N = 21,969). Mortality rates were risk-adjusted using validated multivariable models based on data abstracted from patient's medical records. For all conditions, risk-adjusted mortality declined from a baseline rate of 7.5% to rates of 6.8%, 6.8%, and 6.5%, respectively, during the three subsequent periods. Using weighted linear regression analysis to estimate trends across periods, declines in mortality rates were significant for CHF (0.50% per period; P = 0.002) and pneumonia (0.38% per period; P = 0.03). We conclude that hospital mortality declined in association with the dissemination of comparative data. Although changes in hospital care were not directly examined, the results suggest that initiatives to examine provider performance may have a beneficial impact on quality of care.

A regional evaluation of variation in low-severity hospital admissions

OBJECTIVE

Determine patient and hospital-level variation in proportions of low-severity admissions.

DESIGN

Retrospective cohort study.

SETTING

Thirty hospitals in a large metropolitan region.

PATIENTS

A total of 43,209 consecutive eligible patients discharged in 1991 through 1993 with congestive heart failure (n = 25,213) or pneumonia (n = 17,995).

MEASUREMENTS AND MAIN RESULTS

Admission severity of illness was measured from validated multivariable models that estimated the risk of in-hospital death; models were based on clinical data abstracted from patients' medical records. Admissions were categorized as "low severity" if the predicted risk of death was less than 1%. Nearly 15% of patients (n = 6,382) were categorized as low-severity admissions. Compared with other patients, low-severity admissions were more likely (p < .001) to be nonwhite and to have Medicaid or be uninsured. Low-severity admissions had shorter median length of stay (4 vs 7 days; p < .001), but accounted for 10% of the total number of hospital days. For congestive heart failure, proportions of low-severity admissions across hospitals ranged from 10% to 25%; 12 hospitals had rates that were significantly different (p < .01) than the overall rate of 17%. For pneumonia, proportions ranged from 3% to 22%; 12 hospitals had rates different from the overall rate of 12%. Variation across hospitals remained after adjusting for patient sociodemographic factors.

CONCLUSIONS

Rates of low-severity admissions for congestive heart failure and pneumonia varied across hospitals and were higher among nonwhite and poorly insured patients. Although the current study does not identify causes of this variability, possible explanations include differences in access to ambulatory services, decisions to admit patients for clinical indications unrelated to the risk of hospital mortality, and variability in admission practices of individual physicians and hospitals. The development of protocols for ambulatory management of low-severity patients and improvement of access to outpatient care would most likely decrease the utilization of more costly hospital services.

Measuring prognosis and case mix in hospitalized elders. The importance of functional status

OBJECTIVE

Although physical function is believed to be an important predictor of outcomes in older people, it has seldom been used to adjust for prognosis or case mix in evaluating mortality rates or resource use. The goal of this study was to determine whether patients' activity of daily living (ADL) function on admission provided information useful in adjusting for prognosis and case mix after accounting for routine physiologic measures and comorbid diagnoses.

SETTING

The general medical service of a teaching hospital.

PARTICIPANTS

Medical inpatients (n = 823) over age 70 (mean age 80.7, 68% women).

MEASUREMENTS

Independence in ADL function on admission was assessed by interviewing each patient's primary nurse. We determined the APACHE II Acute Physiology Score (APS) and the Charlson comorbidity score from chart review. Outcome measures were hospital and 1-year mortality, nursing home use in the 90 days following discharge, and cost of hospitalization. Patients were divided into four quartiles according to the number of ADLs in which they were dependent.

MAIN RESULTS

ADL category stratified patients into groups that were at markedly different risks of mortality and higher resource use. For example, hospital mortality varied from 0.9% in patients dependent in no ADL on admission, to 17.4% in patients dependent in all ADLs. One-year mortality ranged from 17.5% to 54.9%, nursing home use from 3% to 33%, and hospital costs varied by 53%. In multivariate analyses controlling for APS. Charlson scores, and demographic characteristics, compared with patients dependent in no ADL, patients dependent in all ADLs were at greater risk of hospital mortality (odds ratio [OR] 13.7; 95% confidence interval [CI] 3.1-58.8), 1-year mortality (OR 4.4; 2.7-7.4), and 90-day nursing home use (OR 14.9; 6.0-37.0). The DRG-adjusted hospital cost was 50% higher for patients dependent in all ADLs. ADL function also improved the discrimination of hospital and 1-year mortality models that considered APS, or Charlson scores, or both.

CONCLUSIONS

ADL function contains important information about prognosis and case mix beyond that provided by routine physiologic data and comorbidities in hospitalized elders. Prognostic and case-mix adjustment methods may be improved if they include measures of function, as well as routine physiologic measures and comorbidity.

Weak associations between hospital mortality rates for individual diagnoses: implications for profiling hospital quality

OBJECTIVES

This study examined the consistency of hospital mortality rates across different diagnoses.

METHODS

Standardized mortality ratios for patients discharged in 1991 from US hospitals were determined via the Medicare Hospital Information Report.

RESULTS

Correlations between standardized mortality ratios for different diagnoses were relatively weak, ranging from .03 to .34. Agreement between hospital rankings (based on standardized mortality ratios), as measured by the weighted kappa statistic, was also weak.

CONCLUSIONS

The present results indicate that hospital mortality rates for individual diagnoses are weakly associated. Thus, it may not be valid to generalize conclusions about hospital performance from a single diagnosis.

Predictions of hospital mortality rates: a comparison of data sources

BACKGROUND

Comparing hospital mortality rates requires accurate adjustment for patients' intrinsic differences. Commercial severity systems require either administrative data that omit vital clinical facts about patients' conditions at hospital admission or costly, time-consuming abstraction of medical records. The validity of supplementing administrative data with laboratory data has not been assessed.

OBJECTIVE

To compare risk-adjusted mortality predictions using administrative data alone; administrative data plus laboratory values; and the combination of administrative, laboratory, and clinical data.

DESIGN

Retrospective cohort study.

SETTING

30 acute care hospitals.

PATIENTS

46,769 patients hospitalized with acute myocardial infarction, cerebrovascular accident, congestive heart failure, or pneumonia.

MEASUREMENTS

Each patient's probability of dying was estimated by using administrative data only (unrestricted administrative models), administrative data restricted to secondary diagnoses that are unlikely to be hospital-acquired complications (restricted administrative models), restricted administrative data plus laboratory data (laboratory models), and restricted administrative data plus laboratory and abstracted clinical data (clinical models).

RESULTS

The unrestricted administrative models predicted death better than the restricted administrative models (average areas under the receiver-operating characteristic [ROC] curves, 0.87 and 0.75, respectively) and as well as the laboratory models and the clinical models (average areas under the ROC curves, 0.86 and 0.87, respectively). The good mortality predictions obtained by using the unrestricted administrative models result from inclusion of hospital-acquired complications that commonly precede death. The laboratory models ranked 93% of patients and 95% of hospitals in a manner similar to the clinical models; in comparison, rankings provided by the laboratory models were similar to those provided for 75% of patients and 69% of hospitals by the unrestricted administrative models and for 72% of patients and 77% of hospitals by the restricted administrative models.

CONCLUSIONS

Adding laboratory data (often available electronically) to restricted administrative data sets can provide accurate predictions of inpatient death from acute myocardial infarction, cerebrovascular accident, congestive heart failure, or pneumonia. This alternative avoids the cost of data abstraction and the serious errors associated with using administrative data alone.

Racial variation in predicted and observed in-hospital death. A regional analysis

OBJECTIVE

To compare observed, predicted, and risk-adjusted hospital mortality rates in white and African-American patients and to determine whether, as prior studies suggest, African-American patients would have higher predicted risks of death and similar or higher risk-adjusted mortality.

DESIGN

Retrospective cohort study.

SETTING

Thirty hospitals in northeast Ohio.

PATIENTS

A total of 88205 eligible patients consecutively discharged in the years 1991 through 1993 with the following 6 diagnoses: acute myocardial infarction, congestive heart failure, obstructive airways disease, gastrointestinal hemorrhage, pneumonia, and stroke.

METHODS

We measured predicted risks of death at admission for each diagnosis using validated multivariable models based on standard clinical data abstracted from patients' medical records. We then adjusted in-hospital mortality rates in white and African-American patients for predicted risk of death and other covariates using logistic regression analysis.

MAIN OUTCOME MEASURES

Predicted risk of death at admission and observed hospital mortality in white and African-American patients.

RESULTS

Predicted risks of death were lower (P<.001) in African Americans for 4 of the 6 diagnoses. Adjusted odds of hospital death were lower (P<.01) in African Americans for 2 of the 6 diagnoses (congestive heart failure and obstructive airways disease) and similar for the other 4 diagnoses. For all diagnoses, in aggregate, the adjusted odds of hospital death were 13% lower in African-American compared with white patients (multivariable odds ratio, 0.87; 95% confidence interval, 0.80-0.94). Findings were similar if further adjustments were made for differences in length of stay, site of hospitalization, or discharge triage practices.

CONCLUSION

Contrary to our a priori hypotheses, predicted risks of death and risk-adjusted mortality rates were generally lower in African-American patients. Our finding of lower predicted risk may reflect racial differences in hospital admission practices or in access to outpatient care. However, our findings suggest that, once hospitalized, African-American patients attained similar or better outcomes, as measured by an important measure--hospital mortality.

Care of patients with upper gastrointestinal hemorrhage in academic medical centers: a community-based comparison

BACKGROUND & AIMS

A common perception among purchasers is that academic medical centers are inefficient and overutilize technology; however, little empirical information exists. The aim of this study was to compare treatment and outcomes of patients with upper gastrointestinal hemorrhage admitted to major teaching hospitals and other hospitals in a large metropolitan area.

METHODS

Data on 3801 consecutive eligible patients admitted to five major teaching hospitals and 25 other hospitals from 1991 to 1993 were obtained by review of medical records. Admission severity of illness was measured using validated multivariable models.

RESULTS

Rates of upper endoscopy were somewhat lower among the 1004 patients discharged from fellowship hospitals, compared with the other 2797 patients (82.9% vs. 85.6%; P < 0.05), and the use of other procedures was similar. Although patients admitted to fellowship hospitals tended to have a higher severity of illness, both unadjusted (6.3 +/- 9.0 vs. 7.1 +/- 7.5 days; P < 0.01) and risk-adjusted length of stay were somewhat shorter. Mortality rates were similar between hospitals, and patients admitted to fellowship hospitals were somewhat less likely to be transfused.

CONCLUSIONS

In patients with upper gastrointestinal hemorrhage, teaching hospitals do not appear to provide inefficient care or overutilize expensive treatments when compared with community facilities. These findings are noteworthy at a time when viability of academic centers and fellowship training is threatened.

The annual guide to "America's best hospitals". Evidence of influence among health care leaders

To determine health care leaders' opinions about a prominent guide to hospital quality, we surveyed 82 chief executive officers (CEOs) and 80 chiefs of staff of hospitals listed in the 1994 edition of the guide and 50 directors of employer based coalitions. Most of the CEOs (87%) and chiefs of staff (86%) said the guide was used in advertising. More than three quarters of the CEOs, chiefs of staff, and coalition directors who were familiar with the guide thought it was accurate, and most indicated that key constituencies (e.g., physicians, corporate managers) were aware of the guide. Our results demonstrate the likely influence of one prominent guide to health care quality and highlight the need for formal independent assessment of such guides.

Impact of interhospital transfers on outcomes in an academic medical center. Implications for profiling hospital quality

The purpose of this article is to determine whether a widely implement ed method of severity adjustment underestimated the risk of death and other outcomes among interhospital transfers (ie, patients transferred from other acute care hospitals) and to examine the impact of this potential bias on hospital outcomes profiles. The retrospective cohort study was conducted at a midwestern academic medical center with 40,820 adult medical and surgical patients from 1988 to 1991, of whom 38,946 were direct admissions and 1,874 were interhospital transfers. Hospital mortality, length of stay, and total charges in interhospital transfers and direct admissions were compared using multivariable regression methods that adjusted for admission severity of illness and other potential covariates (age, type of health insurance, diagnosis, emergent admission). Severity of illness was measured using the Medis-Groups methodology. Admission severity of illness was directly related (P<0.001) to rates of in-hospital death, length of stay, and charges, and was higher among interhospital transfers; 49% of transfers had moderate to high severity, compared with 35% of direct admissions (P<0.001) However, in a logistic regression model adjusting for severity and other covariates, the risk of in-hospital death was nearly two times (multivariable odds ratio, 1.99; 95% confidence interval [CI], 1.64-2.42) higher in transfers than in direct admissions. In linear regression models, length of stay and charges were 1.47 (95% CI, 1.42-1.53) and 1.40 (95% CI, 1.35-1.44) times higher, respectively, in transfers. Results were consistent in medical and surgical admissions, when examined separately, and among individual diagnostic categories. Based on their findings, the authors estimate that, independent of quality of care, severity adjusted mortality and length of stay would appear 17% and 8% higher, respectively, for hospitals in which 20% of patients were interhospital transfers than for hospitals in which 2% of patients were transfers. In an academic medical center, interhospital transfers had poorer severity adjusted outcomes than patients admitted directly. Failure to account for transfer status may produce biased performance profiles and, therefore, may create disincentives for hospitals to accept transfers from other acute care facilities.