Influence of Weekend Hospital Admission on Short-Term Mortality After Intracerebral Hemorrhage

Approximately 37 000 to 52 400 people experience an intracerebral hemorrhage (ICH) each year in the United States.^1,2 Primary ICH accounts for between 8% and 15% of all strokes,³ and with the increasing age of our population as well as other changes in the demographic composition of the United States, the number of cases of ICH is expected to increase significantly over the next 50 years.² Mortality rates for these patients are quite high, with some estimating a 30-day mortality rate close to 50%.⁴

Prior studies suggest that mortality rates for certain diagnoses are affected by whether the patient is admitted to the hospital on the weekend. This has been dubbed the “weekend effect,”⁵ a concept that has understandably called into question the idea that patients receive equal care regardless of when they present to the hospital. This literature has demonstrated the presence of this “weekend effect” for a variety of diagnoses, including stroke, myocardial infarction, pulmonary embolism, and ruptured abdominal aortic aneurysm.^6–9 For this study, we hypothesized that patients with ICH had a higher mortality risk if they were admitted to the hospital on the weekends than if they were admitted during the week.

Methods

Data for this study were drawn from inpatient hospital discharge records available from the Agency for Healthcare Research and Quality Health Care Utilization Project Nationwide Inpatient Sample (NIS) data set for 2004.¹⁰ This is the largest all-payer inpatient care database in the United States. The NIS data set for 2004 includes records from over 8 million hospitalizations from 1004 US hospitals across 37 states. This represents roughly 20% of all community hospitals in the United States, including academic hospitals.

The study population included all hospital admissions with ICH listed as the principal cause of the admission. Appropriate patients were identified and extracted from the NIS data using the International Classification of Diseases, 9th Revision, Clinical Modification diagnosis code for intracerebral hemorrhage, 431.¹¹ Multivariable logistic regression analysis was used to calculate the adjusted odds of death occurring within 7 days of admission among patients with ICH who were admitted on either Saturday or Sunday compared with the adjusted odds of death for this event among patients admitted during other days of the week. The odds of death associated with weekend admission were adjusted for differences in patient age by decade, sex, the hospital’s quartile of ICH admission volume, the density of patient’s locality of residence (metropolitan area with more than 1 million residents, metropolitan area with 1 million or fewer residents, micropolitan area, or nonurban local), median income of the patient’s locality of residence (<$25 000, $25 000 to $30 000, $30 001 to $35 000, or >$35 000), and by primary payer (Medicare, Medicaid, private insurance, self-pay, no charge, other payer). Baseline mortality risk was also adjusted for differences in comorbid disease among patients measured by assessing the independent effect of 30 categories of comorbid disease defined using the method of Elixhauser et al.¹²

Multivariable logistic regression analysis was also used to calculate the adjusted risk of in-hospital death within 14 and 30 days of admission using the same multivariable model formulation. Cox proportional hazards regression was used to calculate the adjusted overall hazard of in-hospital death associated with admission on either Saturday or Sunday adjusted for the same set of covariates included in the logistic models. The c statistic was used to assess each multivariable model’s capacity to discriminate between survivors and decedents. The proportion of the total log likelihood of the outcome explained by each multivariable model was measured using the maximum rescaled R² statistic.

Results

The 2004 NIS data set included 13 821 hospital discharge abstract records for patients with ICH. In this study population, 26.8% of the discharge records were for patients hospitalized on either a Saturday or Sunday. Complete data for each model covariate were available for 13 441 records (Table 1).

Weekend hospital admission was a statistically significant independent predictor of death within 7 days (OR, 1.14; 95% CI, 1.05 to 1.25), within 14 days (OR, 1.15; 95% CI, 1.05 to 1.25), and within 30 days (OR, 1.15; 95% CI, 1.05 to 1.25; Table 2). The multivariable logistic regression model demonstrated adequate discrimination among patients with in-hospital death within 7 days (c=0.65) and explained 7% of the total log likelihood of in-hospital death within 7 days predicting death within 7 days with similar discrimination and explanatory performance obtained using the same model to predict death within 14 days and death within 30 days (supplemental Table I, available at http://stroke.ahajournals.org).

The adjusted hazard of in-hospital death (hazard ratio, 1.12; CI, 1.05 to 1.20) calculated for this study population indicates that the overall risk of in-hospital death is 12% higher for patients admitted on either Saturday or Sunday. Survival probabilities up to 28 days after admission for patients admitted on either a Saturday or Sunday are plotted in the Figure along with that of patients admitted on other days of the week.

• Download figure
• Download PowerPoint

Discussion

The data presented here indicate that patients admitted on the weekend with ICH have increased risk of in-hospital death compared with patients admitted on the weekdays with the same diagnosis. After adjusting for patient demographics and comorbidities, weekend admission was found to increase the overall risk of in-hospital death by 12%. Once a medical hypothesis, the concept of the “weekend effect” has become increasingly examined in the medical literature.^{5,7,8,13–19}

Arguably the most substantial study investigating at the “weekend effect” is the work of Bell and Redelmeier published in 2001.⁷ They looked at approximately 3.8 million acute care hospital admissions in Ontario, Canada, over a 10-year period. Their retrospective analysis examined the 100 leading causes of death over that time and found that 23 of these diagnoses had higher mortality rates when admitted on the weekends. These included diagnoses of pulmonary embolism, ruptured abdominal aortic aneurysm, and acute epiglottitis. Interestingly, their analysis used ICH as one of 3 control groups, because they found no difference in in-house mortality rates for the 10 987 patients with ICH, which is in contrast to our findings. What is perhaps most striking between the 2 studies is the observed difference in in-house mortality rates. Bell and Redelmeier reported a mortality rate of 44% regardless of whether a patient was admitted on the weekend or during the week. On the other hand, our analysis demonstrated a mortality rate of 30% for all patients with ICH at 30 days after admission with the previously mentioned increased OR for those patients admitted on the weekend. Certainly there are differences in the healthcare systems between the United States and Canada, and the patients in our study were treated nearly 16 years after the first patients were treated in the study by Bell and Redelmeier; however, it was not possible to evaluate any potential differences in the treatment of the patients in the 2 studies.

Recently, investigations have been conducted looking at the “weekend effect” with regard to several neurological conditions. In 2007, Saposnik et al⁶ examined the effect of weekend admission on patients with ischemic stroke. They performed a retrospective analysis similar to ours using data from the Hospital Morbidity Database, a Canadian database managed by the Canadian Institute for Health Information. They looked at 26 676 hospital admissions for ischemic stroke over a 1-year period and showed that weekend admission was associated with higher mortality rates at 7 days and at discharge when compared with weekday admission. Two years earlier, Hasegawa et al⁹ analyzed 1134 admissions to Japanese stroke units with similar conclusions as those of Saposnik et al.⁶ They found that weekend admission was associated with unfavorable outcome and increased mortality risk.⁹ Although their study size was significantly smaller than Saposnik et al, they analyzed not only mortality, but outcome determined by the modified Rankin scale, an additional analysis that unfortunately could not be performed using databases such as the one in our study or that of Saposnik et al. Other groups have since reported the presence of a “weekend effect” for patients with stroke.²⁰

There are several potential reasons to explain the observed weekend effect in the present study. Weekend shifts often have fewer physicians and nurses with a relative lack of experience and decreased familiarity with patients that certainly may contribute to our findings.^7,21–23 A portion of the “weekend effect” may also be due to the availability of interventional procedures, as they are often less likely to be performed on the weekends.²⁴ Although this may account for worse outcomes for some diagnoses, it is less likely to affect the patients in this study, because the vast majority of patients with ICH receive treatment that is largely supportive, and surgical management has not been proven to be of any benefit.²⁵ It is also conceivable that the patients who present on the weekends have a higher severity of ICH than those on the weekdays, a finding that has been previously described with acute coronary syndrome.²⁶ We have no reason, however, to suspect that there was a higher percentage of severe ICH cases admitted on the weekend. Another possible explanation lies in the lack of uniformity that exists in the management of patients with ICH. Patients are often admitted directly to the floor, where it may be harder to adequately manage blood pressure or occasionally intracranial pressure, and patients with ICH are admitted to the care of physicians with different training backgrounds, including neurologists, neurosurgeons, internists, and hospitalists.

Although we feel that this study has substantial implications, there are several inherent limitations of studies using databases such as the NIS that should be recognized. It is impossible to randomize patients into a prospective study looking at the day of admission, and as such, virtually all of the studies looking at the “weekend effect” have used retrospective analyses of patient databases.^{5–8,13,14,16–18,24,27,28} There are also drawbacks specific to the NIS that may have influenced this particular study. The NIS Health Care Utilization Project data identifies whether patients were admitted on a weekend day (Saturday or Sunday), but it does not identify the specific day of the week that patients were admitted. Thus, it was not possible to distinguish Friday night or holiday admissions. This characteristic of the available data is a limitation in examining the true differences in mortality outcomes between weekend and weekday admissions in this study.

Differences in actual hospital staffing levels and staff expertise at the time a patient is hospitalized are not measured in this study. Weekend admission is only a surrogate measure of differences in hospital staffing, and this indirect relationship is a significant limitation. Factors other than hospital staff levels and experience that occur more or less often for patients with weekend admission may also contribute to the observed difference in patient mortality risk. Another limitation is the potential for unadjusted confounders. We were not able to measure and adjust for potentially important confounders such as differences in cause of death, admission grade and stroke severity, level of consciousness, imaging findings, size of hemorrhage, the need for cerebrospinal fluid diversion, whether the ICH was evacuated, and graded outcome scales. Although limited in detail, analysis of the NIS data provides a population-based perspective on the association among in-hospital death, weekend admission, and other potential explanatory factors that is unmatched by any other source.^29–37

Conclusion

Our study found that patients with ICH who were admitted on weekends had a higher adjusted risk of inpatient death than patients admitted on weekdays. It should be the right of every patient to receive and the goal of every physician to provide the same standard of care regardless of the day or time of admission. Nevertheless, a discrepancy in hospital resources and personnel available on the weekends is often apparent. Certainly there needs to be a substantial amount of institutional interest in auditing current systems and in implementing new systems to eliminate this difference, and there needs to be departmental interest in ensuring that every patient under their care receives a uniform quality of care.

Footnotes