Comparing the demographic data and outcomes of septic shock patients presenting

Ralph Bou Chebl, Nadim Kattouf, Mohamad Assaf, Saadeddine Haidar, Gilbert Abou Dagher, Sarah Abdul Nabi,Rana Bachir, Mazen El Sayed

Department of Emergency Medicine, American University of Beirut, Beirut 1107 2020, Lebanon

KEYWORDS: Sepsis; Septic shock; In-hospital mortality; Hospital teaching status; United States;Epidemiology; Outcome

INTRODUCTION

Sepsis is a public health burden and the most prevalent disease among intensive care unit (ICU) admissions.[1]Its incidence and prevalence have been gradually increasing over the years.[1]Sepsis also poses an important financial burden on the healthcare system; it accounts for 13% of hospitalization costs in the United States.[2]Moreover, sepsis patients who survive their ICU stay are at higher risk of immune system dysregulation, recurrent admissions and decreased quality of life.[1]The overall mortality rate of septic shock in the United States was 34.2% in 2018, with a median hospital length of stay reaching 16.5 d.[3,4]Similar numbers were also reported in Europe, where 10.4% of ICU admissions were attributed to septic shock, with a 28-day mortality reaching 36.7%.[5]Older age, medical comorbidities and an Acute Physiological Assessment and Chronic Health Evaluation (APACHE) score greater than 25 were associated with increased mortality in septic shock patients.[6-9]Several studies have shown that better outcomes in septic shock patients are due to early diagnosis and antibiotic use.[10-12]

Previous studies have examined the association between hospital teaching status and hospital mortality.Burke et al[13]found that teaching hospitals in the United States had significantly lower 30-day mortality rates than non-teaching hospitals. This was thought to be due to a higher level of medical expertise, availability of advanced technology and optimization of the staff’s duty hours.[13,14]These results were particularly more evident in pneumonia, acute coronary syndrome (ACS) and congestive heart failure patients.[15-17]

Given the lack of studies assessing the effect of hospital teaching status on septic shock-related mortality,the aim of this study was to examine the association between hospital teaching status and in-hospital mortality in septic shock patients in the United States.

METHODS

Study design

This was a retrospective observational study, using the 2016 Nationwide Emergency Department Sample(NEDS) Database, which was released in 2018. It is the largest validated database across the United States,containing over 30 million yearly emergency department(ED) visits.[18]It was developed for the Agency for Healthcare Research and Quality (AHRQ), Healthcare Cost and Utilization Project (HCUP).[18]The database includes ED visits from over 953 hospitals in 36 different states (686 hospitals were metropolitans) and the District of Colombia amounting to a total of 144,842,742 weighted ED visits.[19]To generate nationwide estimates,weights were created using the AHA survey of hospitals as the standard. These weights were used to obtain national and encounter level estimates for approximately 144 million ED visits in the NEDS 2016 database.[18]All contributing authors to this study finished the online HCUP Data Use Agreement training course and signed the agreement. An exemption for research using the deidentified database has been granted by the institutional review board of the American University of Beirut. Sepsis was defined according to the Sepsis-3 guidelines.[20]All septic shock patients who were 18 years of age or above,who were admitted to a metropolitan hospital, were included in the study. Patients below the age of 18 years or with an unrecorded age or who presented to nonmetropolitan or rural hospitals or with an ED disposition other than admitted, died, or transferred (home health care, against medical advice, not admitted/destination unknown, discharge alive/destination unknown but not admitted) were excluded. ICD-10 CM code R65.21 was used to identify septic shock patients. While selecting the study sample, we considered all diagnoses and not only the first diagnosis. Septic shock patients were divided into two groups: those who presented to a metropolitan teaching hospital and those who presented to metropolitan non-teaching hospital. A teaching hospital is defined as a hospital that has a residency program approved by the American Medical Association (AMA) and is part of the Council of Teaching Hospitals (COTH) or if the ratio of full-time residents to beds is at least 0.25 based on the AHA Annual Survey Database.[18]A hospital was defined as metropolitan or non-metropolitan based on the county in which it is located. Urban Influence Codes (UICs)were used to categorize hospitals as metropolitan or nonmetropolitan. Chronic medical conditions were identified using ICD-10 CM coding. The variables included in the analysis of our study were patient demographics, chronic conditions, hospital teaching status, length of stay, ED and hospital procedures, disposition and outcomes. The variable of interest was hospital teaching status. The primary outcome was hospital mortality.

Data abstraction

The Healthcare Cost and Utilization Project (HCUPUS) is a family of healthcare databases developed through a Federal-State-Industry partnership. HCUP databases bring together the data collection efforts of hospitals, private data organizations and state data organizations. Quality review is checked by conducting descriptive analyses for all data elements and is implemented by an independent contractor for each year and data source. The validity of the numeric data is verified by checking their range against legal values/standard norms/maximum allowed value. The validity of the character data is assessed by verifying them against norms when feasible (e.g., diagnosis codes and procedure codes). Internal inconsistencies were assessed by comparing the values of related data elements. For example, a hysterectomy procedure should be performed for females and not for males. The admission date should fall before the discharge date.[18]The following link provides detailed information about the HCUP quality control procedures: https://www.hcup-us.ahrq.gov/db/quality.pdf.

Statisticial analysis

Descriptive analyses of the demographics,administrative and clinical characteristics were reported as the mean and 95% confidence interval (CI) for the continuous variables and as frequencies, percentages,and 95%CIfor the categorical variables. Some analyses were verified and confirmed by using a free online query system based on data from HCUP (HCUPnet).The Rao-Scott Chi-square test, a modified version of Pearson’s Chi-square test, and a general linear model for complex samples were used to compare the proportion of the categorical variables and the mean age between the two groups of patients (teaching vs. non-teaching).Binary logistic regression is not an accurate multivariate analysis for data abstracted from the NEDS database that was collected using complex survey sampling designs. Therefore, to avoid committing a type I error and producing biased estimates, a complex samples logistic regression model was performed by including all statistically and clinically significant variables. The following variables were imposed on the model: hospital teaching status; age; indicator of sex; infectious and parasitic disease; neoplasms; endocrine, nutritional, and metabolic diseases and immunity disorders; diseases of blood and blood-forming organs; diseases of the circulatory system; diseases of the respiratory system;diseases of the digestive system; diseases of the skin and subcutaneous tissue; region of hospital; average length of stay; hospital procedures (small bowel resection;colorectal resection; appendectomy; fluoroscopy of the biliary and pancreatic ducts; cholecystectomy and common duct exploration; laparoscopy [GI only];exploratory laparotomy; other non-operative upper GI therapeutic procedures; other operative upper GI therapeutic procedures; other non-operative lower GI therapeutic procedures; other operative lower GI therapeutic procedures; other GI diagnostic procedures;other non-operative GI therapeutic procedures; other operative GI therapeutic procedures and respiratory intubation and mechanical ventilation).

AP-value of ≤0.05 was used to indicate statistical significance. All statistical analyses were performed using SPSS version 24.0 (IBM Corp., USA).

RESULTS

A total of 388,552 septic shock patients were included in the study (Figure 1). The average age was 66.93 (95%CI: 66.84-67.03) years, and 51.7%were males. The majority of patients presented to a metropolitan teaching hospital (68.2%), and 31.8%presented to a metropolitan non-teaching hospital. The most common comorbid conditions were a history of infectious and parasitic disease (97.8%), endocrine,nutritional and immune disorders (92.9%), diseases of the circulatory system (87.5%), diseases of the genitourinary system (84.1%), diseases of the respiratory system (79.9%) and diseases of blood and blood-forming organs (63.3%). The average length of hospital stay was 11.76 d (95%CI: 11.67-11.84 d). A total of 31% of septic shock patients (n=120,602) died during their hospital stay (Table 1).

Figure 1. Inclusion and exclusion flowchart. ED: emergency department.

A total of 43.7% and 3.1% of septic shock patients underwent minor and major diagnostic procedures,respectively. A total of 83.1% and 19.1% underwent minor and major therapeutic procedures, respectively. A total of 49.4% of the septic shock patients were intubated and were mechanically ventilated during their hospital stay (Table 2).

Table 1. The characteristics and disposition of the septic shock patients presenting to the emergency department

Table 2. Procedures undergone by the septic shock patients

Patients presenting to a metropolitan non-teaching hospital were older (68.01 years vs. 66.43 years,P＜0.001). There was no significant difference in gender based on hospital teaching status (51.4% males in nonteaching hospitals vs. 51.8% in teaching hospitals,P=0.190). Patients presenting to a metropolitan teaching hospital were found to have higher rates of malignancy(18.7% vs. 16.6%,P＜0.001), endocrine, nutritional and immune disorders (93.0% vs. 92.7%,P=0.029), diseases of the respiratory system (80.1% vs. 79.3%,P=0.003),diseases of blood and blood forming organs (63.8%vs. 62.1%,P＜0.001), diseases of the digestive system(58.4% vs. 56.4%,P＜0.001) and diseases of the skin and subcutaneous tissue (24.9% vs. 22.0%,P＜0.001).Respiratory intubation and mechanical ventilation rates among septic shock patients were higher in metropolitan teaching hospitals (50.5% vs. 46.9%,P＜0.001).Exploratory laparotomy (0.5% vs. 0.3%,P＜0.001) and all non-surgical gastrointestinal procedures (7.5% vs.5.9%) were significantly higher in the metropolitan teaching hospitals. However, small bowel resection(0.4% vs. 0.3%,P＜0.001), appendectomy (0.6% vs.0.4%,P＜0.001) and cholecystectomy (1.3% vs. 0.9%,P＜0.001) were significantly higher in the metropolitan non-teaching hospitals. The average length of hospital stay was significantly longer in metropolitan teaching hospitals(12.47 d vs. 10.20 d,P＜0.001). Hospital mortality was significantly higher in metropolitan teaching hospitals(32.4% vs. 28.1%,P＜0.001) (Table 3).

After adjusting for confounders, septic shock patients presenting to metropolitan teaching hospitals had greater odds of in-hospital mortality than patients presenting to metropolitan non-teaching hospitals (adjustedOR=1.295,95%CI: 1.256-1.335,P＜0.001) (Table 4).

Table 3. The characteristics, disposition and hospital procedures of the septic shock patients presenting to the emergency department based on hospital teaching status

Table 4. Crude and adjusted odds ratio of mortality in septic shock patients

DISCUSSION

This study examined the relationship between hospital teaching status and in-hospital mortality in septic shock patients using the 2016 NEDS database. Inhospital mortality was found to be significantly higher in teaching hospitals than in non-teaching hospitals.After adjusting for confounders, septic shock patients presenting to metropolitan teaching hospitals had higher odds of mortality compared to metropolitan non-teaching hospitals. Furthermore, septic shock patients presenting to teaching hospitals were found to have a higher percentage of medical comorbidities, were more likely to be intubated or placed on mechanical ventilation and had a longer hospital length of stay.

Septic shock treatment evolved over time. It was first centered around early goal-directed therapy (EGDT),which was resource dependent and invasive involving the use of central venous catheters, arterial lines and blood transfusions.[21,22]Three randomized clinical trials have showed that EGDT was not better than standard care(consisting of early identification and early antibiotic administration).[10-12]Their findings were corroborated later by a meta-analysis.[22]Studies performed between 2016 and 2020 also recommended the administration of vasopressors via a peripheral line, further reducing the need for central venous catheters.[23-26]Given this switch to less invasive and less resource dependent care, one would expect no difference in septic shockrelated mortality between teaching and non-teaching metropolitan hospitals.

The literature showed that acute coronary syndrome,pneumonia and heart failure patients presenting to teaching hospitals have lower mortality.[15-17]This finding was attributed to the greater expertise of treating teams in teaching hospitals as well as the greater availability of advanced technology and higher medical resident to patient ratio.[13,17]There is a scarcity of evidence examining the association between septic shock mortality and hospital teaching status. One study showed a higher mortality rate for septic shock patients in teaching hospitals. This was explained by the higher severity of sepsis cases in teaching hospitals.[27]

Several reasons can explain why septic shock patients had greater odds of mortality in teaching hospitals. Khwannimit et al[7]showed that patients who presented to teaching hospitals were more likely to have chronic medical conditions and increased odds of hospital mortality. Furthermore, another study showed that there was a correlation between the number of medical comorbidities and septic shock related inhospital mortality.[28]

Another potential explanation for the higher mortality in teaching hospitals is the higher intubation rate and mechanical ventilation in septic shock patients.Mechanical ventilation can predispose patients to lung injury as well as ventilator associated pneumonia (VAP).Delbove et al[29]showed that among sepsis patients, those who were intubated had a higher mortality rate than those who were not intubated. Furthermore, patients who developed VAP had a higher mortality rate and longer ICU stay.[30,31]

We found that septic shock patients presenting to metropolitan teaching hospitals had more comorbidities at baseline and higher intubation rates during their hospital stay. This might indicate that they were sicker or were more complicated to manage medically. In addition,they had a longer length of hospital stay, which can increase the patients’ risk of hospital acquired infection and further increase their morbidity. The above factors could have contributed to the higher odds of in-hospital mortality in metropolitan teaching hospitals.

Limitations

This study is limited by its retrospective design. The generalizability to other populations could be limited by the fact that the database only includes hospitals inthe United States. In addition, there can be variability in coding from one ED to another which might affect the selection of septic shock patients by not including all eligible patients. However, the ICD-10 CM for septic shock was used to obtain the data from the NEDS database to catch as many patients as possible. In the NEDS database, each admission from the ED to the hospital leads to the generation of one discharge record.A transfer from one ED to another would result in two discharge records: one from the initial hospital and one from the receiving hospital. If both hospitals participate in the database, both records will be included. Transfer from one ED to another ED occurred in only 1.5% of the NEDS records.[18]The low percentage of patients who fit the previous category would less likely skew the results. Furthermore, several variables, such as APACHE scores, SOFA scores, vital signs, lab parameters and therapeutic measures (intravenous fluid requirements and vasopressor requirements) were not included in the database and play a role in mortality comparison. Finally,it is important to note that information regarding hospital performance metrics, resident training level at teaching hospitals or the level of expertise of physicians at nonteaching hospitals was not available, which might have influenced the outcomes of the study.

CONCLUSION

Septic shock patients presenting to a metropolitan teaching hospital had significantly higher odds of mortality than those presenting to metropolitan nonteaching hospitals. Patients presenting to teaching hospitals had higher rates of pre-existing medical conditions. They also had higher rates of intubation and mechanical ventilation and longer lengths of hospital stay. These findings are interesting given the changing nature of sepsis management. Further studies are needed to study the association between hospital teaching status and mortality in septic shock patients.

Funding:None.

Ethical approval:An exemption for research using the deidentified database has been granted by the institutional review board of the American University of Beirut.

Conflicts of interest:The authors have no commercial associations or sources of support that might pose a conflict of interest.

Contributors:Concept (RBC, NK, MA, SH, GAD, MS), design(RBC, NK, MA, SH, GAD, MS), definition of intellectual content(RBC, NK, MA, SH, GAD, MS), literature search (RBC, NK,MA, SH, GAD, SAN, MS), data acquisition (RB), data analysis(RB), statistical analysis (RB), manuscript preparation (RBC, NK,MA, SH, GAD, SAN, MS), manuscript editing and manuscript review (RBC, NK, MA, SH, GAD, SAN, MS).