Mortality Rate After Nonelective Hospital Admission
Total Page：16
File Type：pdf, Size：1020Kb
PAPER Mortality Rate After Nonelective Hospital Admission Rocco Ricciardi, MD, MPH; Patricia L. Roberts, MD; Thomas E. Read, MD; Nancy N. Baxter, MD, PhD; Peter W. Marcello, MD; David J. Schoetz, MD Objective: We hypothesized that the mortality rate af- patients with nonelective hospital admissions during the ter nonelective hospital admission is higher during week- 5-year study period: 6 842 030 during weekends and ends than weekdays. 23 149 591 during weekdays. Inpatient mortality was re- ported in 185 856 patients (2.7%) admitted for nonelec- Design: Retrospective cohort analysis. tive indications during weekends and 540 639 (2.3%) dur- ing weekdays (P Ͻ .001). The regression revealed Setting: Patients admitted to hospitals in the Nation- significantly higher mortality during weekends for 15 of wide Inpatient Sample, a 20% sample of US community 26 (57.7%) major diagnostic categories. The weekend hospitals. effect remained, and mortality was noted to be 10.5% higher during weekends (odds ratio, 1.10; 95% confi- Patients: We identified all patients with a nonelective dence interval, 1.10-1.11) compared with weekdays af- hospital admission from January 1, 2003, through De- ter adjusting for all other variables with the imputed data cember 31, 2007, in the Nationwide Inpatient Sample. set. Next, we abstracted vital status at discharge and calcu- lated the Charlson comorbidity index score for all pa- Conclusions: These data demonstrate significantly worse tients. We then compared odds of inpatient mortality af- outcomes after nonelective admission during the week- ter nonelective hospital admission during the weekend end compared with weekdays. Although the underlying compared with weekdays, after adjusting for diagnosis, mechanism of this finding is unknown, it is likely that age, sex, race, income level, payer, comorbidity, and hos- factors such as differences in hospital staffing and ser- pital characteristics. vices offered during the weekend compared with week- days are causal and mutable. Main Outcome Measure: Mortality rate. Results: Discharge data were available for 29 991 621 Arch Surg. 2011;146(5):545-551 HE HEALTH CARE SYSTEM IN during weekends for several urgent medi- the United States is rap- cal and surgical diagnoses. However, many idly evolving to provide the of these studies6-14 have focused on a single highest-quality care at the diagnosis or set of diagnoses. We hypoth- most reasonable cost. Al- esized that differences in mortality rates Tthough Americans anticipate receiving based on day of admission are present across quality health care based on the most the spectrum of clinical diagnoses. Thus, sound scientific knowledge available,1 we undertook a study to evaluate mortal- many patients do not receive optimal ity rate as a function of admission day across care.2,3 Differences in outcome and qual- a wide range of medical and surgical diag- ity of care have been described for mea- noses for patients admitted to hospitals sures of health and life expectancy in re- within the United States. Author Affiliations: Lahey lation to race, ethnicity, sex, educational Clinic, Department of level, income, geographic location, dis- METHODS Colorectal Surgery, Tufts ability status, and sexual orientation.4 University Medical School, These observations have led us to more Burlington, Massachusetts critical analyses of outcomes to ensure eq- DATA SOURCE (Drs Ricciardi, Roberts, Read, uitable and reliable high-quality care. Marcello, and Schoetz); and We obtained all-payer discharge data from Janu- Department of Surgery and Because marked variability exists in ary 1, 2003, through December 31, 2007, via Li Ka Shing Knowledge health care outcomes among patients in the Nationwide Inpatient Sample (NIS) of the Institute, St. Michael’s Hospital, hospitals, the inpatient setting has been an Healthcare Cost and Utilization Project of the University of Toronto, Toronto, area of relatively intense study.5 Many stud- Agency for Healthcare Research and Quality. Ontario, Canada (Dr Baxter). ies have identified increased mortality rates The NIS—the largest source of all-payer hos- ARCH SURG/ VOL 146 (NO. 5), MAY 2011 WWW.ARCHSURG.COM 545 ©2011 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 09/26/2021 pital discharge information in the United States—contains data Bureau and categorized as Northeast, Midwest, South, and from approximately 7 million to 8 million hospital stays per West.15 Hospital rurality was categorized as rural or urban based year in 1000 hospitals in 35 states.15 It approximates a 20% strati- on Core Based Statistical Area codes. Before 2004, all metro- fied sample of US community hospitals, including large uni- politan statistical areas were considered urban, and nonmet- versity hospitals and smaller regional facilities. The database ropolitan statistical areas were classified as rural.15 The teach- provides information regarding patient demographics, socio- ing status of the hospital was obtained from the American economic factors, admission profiles, hospital profiles, state Hospital Association Annual Survey of Hospitals. A hospital is codes, discharge diagnoses, procedure codes, total charges, and considered to be a teaching hospital if it has an American Medi- vital status at hospital discharge. Along with other hospital dis- cal Association–approved residency program, is a member of charge databases, the NIS has been used to review trends in sur- the Council of Teaching Hospitals, or has a ratio of full-time gical care and outcomes,16 volume outcome relationships,17 and equivalent interns and residents to beds of 0.25 or higher.15 disparities in care.18 A data use agreement is held by the Agency for Healthcare Research and Quality, and our study was con- VITAL STATUS sidered exempt by the Lahey Clinic Institutional Review Board. The data set permits identification of vital status at the time of PATIENT SELECTION discharge. The variable is coded as died during hospitalization AND PREDICTOR VARIABLE or did not die during hospitalization. Deaths that occurred af- ter discharge are not identifiable from our data set.15 All patients discharged during the time frame sampled were in- cluded. We used the elective variable to exclude all patients with STATISTICAL ANALYSIS an admission for elective reasons and included only those pa- tients with nonelective admission.15 Thus, patients with emer- Statistical analyses were performed using SAS statistical soft- gency and urgent indications for admission were included. ware, version 9.2 (SAS Institute Inc, Cary, North Carolina). We The data set permits identification of admission day as a used t tests to analyze continuous variables and 2 tests for cat- weekend or weekday. We recorded this variable as admitted egorical variables. Results were considered statistically signifi- during a weekend (ie, Saturday or Sunday) or a weekday (ie, cance at P Ͻ .05, and all statistical tests were 2-tailed. We in- Monday through Friday).15 cluded all covariates in our regression model. The analyses were conducted with and without missing variables. To confirm re- sults, we performed imputation of missing data using the mul- COVARIATES tiple imputation procedure from SAS Institute Inc. Imputa- tion substitutes missing values with plausible values that Our analysis adjusted for the following covariates: age, sex, race, characterize the uncertainty regarding the missing data. This income level, payer, major diagnostic categories (subgroup- process results in valid statistical inferences that properly re- ings of diagnosis-related groups), and the Charlson comorbid- flect the uncertainty due to missing values, for example, con- ity index score. Age was included as a continuous variable. Sex fidence intervals with the correct probability coverage.25 The was entered as a dichotomous variable. Race was divided into imputed data set was then analyzed by using standard logistic white, black, Hispanic, Asian or Pacific Islander, Native Ameri- regression for the complete data. can, or other. Income level was categorized into quartiles per Last, to assess whether the effect of admission day differed estimated median household income of residents in the pa- 15 as a function of diagnosis, we tested for interactions between tient’s zip code. The median income quartiles are classified admission day and major diagnostic category. Because of the as follows: $0 to $38 999, $39 000 to $47 999, $48 000 to 15 significant interaction between these variables, we reanalyzed $62 999, and $63 000 or more. the effect of admission day on mortality rate for each indi- Payer was recorded as follows: Medicare, Medicaid, private vidual major diagnostic category with the same covariates in including health maintenance organization, self-pay, no charge, the larger analysis described herein. or other.15 Major diagnostic categories were used to adjust for diagnoses and reflect larger groupings of diagnostic-related groups made available in the provided data set and download- RESULTS able for review from the US Department of Health and Human 19 Services, Centers for Medicare and Medicaid Services. Major COHORT diagnostic categories have been used to evaluate hospitaliza- 20 21 22 tion risk, mortality risk, and other outcomes. We also evalu- From January 1, 2003, through December 31, 2007, a total ated comorbidity with the Deyo modification of the Charlson comorbidity index.23 Briefly, we ascertained the presence of 17 of 40 095 587 discharges were recorded at NIS hospi-