Assessment of Health Care Exposures and Outcomes in Adult Patients With Sepsis and Septic Shock

Key Points Question What types of health care exposures occur during the 30 days before hospitalization of a patient with sepsis or septic shock, and how common are these exposures? Findings In this cohort study of 1078 US adults with sepsis and septic shock across 10 states, most patients experienced sepsis onset outside of the hospital, had recent encounters with the health care system, and had a sepsis-associated pathogen documented; 42% of patients received antimicrobial drugs, chemotherapy, wound care, dialysis, or surgery in the 30 days before sepsis occurred. After controlling for other factors, an association was found between underlying comorbidities, such as cirrhosis, immunosuppression, and vascular disease, and 30-day mortality. Meaning The findings suggest that future efforts to improve outcomes among patients with sepsis and septic shock would benefit from examination of health maintenance practices and recent health care exposures as potential opportunities among high-risk patients.


Introduction
Sepsis is a serious public health issue, with an estimated 1.7 million adult cases annually in the United States, and it is potentially associated with 270 000 deaths. 1 Nearly 1 in 3 hospitalizations that end in death are associated with sepsis. 1 During the last 20 years, initiatives aimed at improving sepsis outcomes have focused on protocols that emphasize early recognition and standardized treatment in a hospital. 2 However, most sepsis cases begin outside of the hospital, encompassing diverse clinical presentations. 3,4 Because no confirmatory test exists, the diagnosis of sepsis is based on evidence of infection, organ dysfunction, and clinical judgment. Most large-scale descriptive studies of sepsis epidemiology using administrative claims data have not included detailed medical record review and are therefore subject to several limitations. 5 We performed detailed medical record reviews to describe the demographic and clinical characteristics, including health care exposures, pathogens, and outcomes, of persons diagnosed with sepsis and septic shock.

Methods
We used the Emerging Infections Program (EIP), a public health surveillance and research network supported by the Centers for Disease Control and Prevention, to collect data from the medical records of patients with sepsis and septic shock. 6 Each EIP site (California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee) identified 1 to 4 acute care hospitals in which to sample patients with sepsis for medical record review. Only general short-term acute care hospitals were eligible for inclusion. Eligible patients were discharged from participating hospitals between October 1, 2014, and September 30, 2015, with diagnosis codes for severe sepsis or septic shock (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 995.92 or 785.52). In each EIP site, staff members randomly selected approximately 100 adult patients (aged Ն18 years) and 100 pediatric patients (aged Յ18 years) across participating hospitals for medical record review; only adult patients are included in this article. For patients with more than 1 hospitalization during the period of interest, only the first hospitalization was eligible for inclusion. We excluded patients who had no clinical documentation that specifically indicated sepsis, septic shock, or similar terms in the medical records.
The Centers for Disease Control and Prevention determined the project to be a nonresearch activity. Each EIP site and hospital review board determined the project to be a nonresearch activity or approved the project as a research activity with a waiver of informed consent. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Data Collection
From September 1, 2017, to May 30, 2018, trained EIP staff members reviewed medical records using a standardized data collection form to abstract detailed information regarding patients' demographic and clinical characteristics.

Definitions
The date of sepsis diagnosis was defined as the first date of clinical documentation by a health care practitioner of either sepsis or an associated term in the patient's medical record. Community-onset cases were defined as cases in which the date of sepsis diagnosis occurred during the first 3 days of hospitalization (in which the date of admission was day 1). Sepsis cases for which the date of sepsis diagnosis occurred after day 3 of hospitalization were classified as hospital-onset cases.
We defined prehospital medical treatment as the receipt of an antimicrobial drug, chemotherapy, wound care, dialysis, or surgery in the 30 days before admission, and we defined a prehospital medical device as the presence of a urinary catheter, central line, mechanical ventilation, gastrostomy or jejunostomy tube, or tracheostomy in the 30 days before admission. We collected information regarding whether the patient had an outpatient medical encounter in the 7 days before hospital admission, including the date of the visit and the type of setting (eg, urgent care or medical subspecialty clinic). We defined health care exposure as the receipt of any prehospital medical treatment in the 30 days before admission, the presence of a prehospital medical device in the 30 days before admission, the occurrence of any outpatient medical encounter within 7 days of the sepsis diagnosis, or the occurrence of a stay of at least 2 days at a health care facility in the 30 days before admission. We determined the presence of systemic inflammatory response syndrome based on clinical information from the date of the sepsis diagnosis. 7 Infections that were documented in the medical record as being present within 7 calendar days before or after the sepsis diagnosis and noted to be associated with sepsis in the discharge summary were included in this analysis. We defined organ dysfunction as any of the following documented in the medical record within 7 calendar days before or after sepsis diagnosis: receipt of invasive mechanical ventilation or noninvasive positive pressure ventilation, receipt of vasopressor medications (ie, norepinephrine, epinephrine, dopamine, phenylephrine, and vasopressin), systolic blood pressure less than 90 mm Hg, or altered mental status. We defined septic shock as 2 or more criteria of systemic inflammatory response syndrome plus the receipt of a vasopressor medication within 1 calendar day of sepsis diagnosis.

Pathogen Data and Death Certificates
Culture and culture-independent diagnostic test information was collected for each patient. We developed an algorithm based on specimen, diagnostic test, and organism type to identify organisms that were likely to be sepsis-associated pathogens (eTable 1 in the Supplement

Statistical Analysis
We performed descriptive analyses of demographic and clinical variables stratified by age. We used logistic regression analysis of complete cases with stepwise selection (entry and stay P values were set to .20 and .05, respectively) to assess risk factors associated with mortality within 30 days after sepsis diagnosis. Model selection included age group, septic shock status, presence of a sepsisassociated pathogen, and organ dysfunction within 7 days of sepsis diagnosis as a priori patient-level risk adjustment factors. Multicollinearity was ruled out using variance inflation factors. Analyses were

Results
We collected information from 28 individual hospitals across 10 EIP sites, with a median of 37 patients (interquartile range [IQR], 25-50 patients) included from each hospital. We excluded 22 patients with medical records that did not include clinical documentation of sepsis or an associated term.

Health Care Exposures
In total, 654 patients (60.7%) had at least 1 health care exposure before their hospital admission for sepsis ( Table 2)

Discussion
Sepsis is an important public health challenge, and characterization of the disease course and health care exposures of patients with sepsis in the days or weeks before hospitalization may help to identify opportunities for improving outcomes. In this cohort of patients, which included in-depth medical record reviews of more than 1000 patients from diverse geographic areas, we observed that most adult patients with sepsis had outpatient or other health care facility exposures or medical  Abbreviation: G-tube, gastrostomy tube; J-tube, jejunostomy tube; PEG-tube, percutaneous endoscopic gastrostomy tube.
a Includes any medical treatment in the 30 days before hospital admission, the presence of a medical device in the 30 days before hospital admission, the occurrence of any outpatient medical encounter within 7 days of the sepsis diagnosis, and the occurrence of a stay of 2 or more days at a health care facility in the previous 30 days. materials to help patients, caregivers, and different health care practitioners recognize the signs and symptoms of sepsis.
We identified factors associated with mortality in adult patients who were hospitalized with sepsis; in particular, we found that increasing age, the presence of organ dysfunction, and selected underlying conditions were associated with death. Of note, we observed that the lack of influenza or pneumococcal vaccination was also associated with mortality, although to a lesser extent than underlying conditions and severity of illness. Although these vaccinations offer protection against specific sepsis-associated pathogens, we hypothesize that vaccination status also serves as a surrogate for broader health care access and health maintenance practices that were unmeasured in our analysis. 11,12 Several analyses have indicated that underlying conditions are associated with sepsis outcomes and may be as important to sepsis outcomes as hospital care. Hatfield et al 4 reported that markers of health status are associated with sepsis mortality even after accounting for in-hospital care, suggesting that efforts that encourage patients to seek care before the onset of organ failure could be associated with reductions in sepsis mortality. Rhee et al 13 reviewed the medical records of 198 deceased patients with sepsis to assess the extent to which these deaths were preventable. They found that 23% of the cohort experienced some level of suboptimal care but considered only 12% of deaths to be potentially preventable, suggesting that patient factors, such as comorbidities and baseline health status, play an important role in sepsis outcomes. Our analysis also highlights the association of underlying comorbidities and severity of illness with sepsis mortality, suggesting that sepsis education and preemptive care may be particularly important for specific patient populations or health care practitioners. The use of risk stratification models for outpatients who present with acute infections could improve the identification of patients with the greatest risk of disease progression who may benefit from closer monitoring. 14 Infection source control and the identification of sepsis-associated pathogens are important components of sepsis treatment that can guide the selection and treatment duration of antimicrobial drugs. 2 Even in our population of patients, who were identified through the use of explicit administrative codes for severe sepsis and septic shock, nearly 40% of patients did not have a sepsisassociated pathogen identified in our analysis; 30% of patients had no documented sepsisassociated infection on their discharge summary, despite having a billing code for severe sepsis or septic shock. It is possible that previous antibiotic medication exposure (identified in 30% of the patients in our analysis) was a factor in the low percentage of patients for whom a sepsis-associated pathogen was identified. In addition, we did not find an association between the identification of a sepsis-associated pathogen and mortality within 30 days after sepsis diagnosis, which could be owing to the prevalence of broad-spectrum antimicrobial treatment among patients with suspected sepsis, regardless of whether a sepsis-associated pathogen had been identified. Better diagnostic tests for sepsis and infectious diseases are needed to quickly and accurately identify patients with sepsis and infections associated with specific pathogens and to improve antimicrobial drug use and minimize antimicrobial-associated risks. 15,16 Limitations This study has several limitations. Patients with sepsis and septic shock were identified through the use of administrative codes along with confirmation that at least 1 health care practitioner had documented sepsis in the patient's medical record rather than through the application of an objective definition based on physiological or laboratory criteria. Explicit codes for severe sepsis and septic shock are specific but insensitive, 1 and we may have captured more severe cases but missed cases of sepsis that were less severe or less easily diagnosed. We only considered an infection to be associated with sepsis or septic shock if it was documented in the discharge summary as associated with sepsis; we may have misclassified infection types among patients or included patients who may not have had sepsis or septic shock.
Because this analysis relied on data abstraction from hospital medical records, the data obtained might be incomplete. Information documented during outpatient clinic visits or nursing home stays, on receipt of home health care services or vaccines, or on indication for prehospital antibiotic medications may have been unavailable. We also limited the period before sepsis hospitalization during which these factors were assessed. In particular, the proportion of patients with health care exposures, including outpatient visits, prehospital medical treatment, and prehospital medical devices, could have been underestimated.
We used an algorithm (eTable 1 in the Supplement) to identify sepsis-associated pathogens based on available clinical information; organisms could have been misidentified as sepsis-associated pathogens. Furthermore, we included a limited number of clinical factors to define organ dysfunction (Table 1); we excluded hepatic dysfunction, thrombocytopenia, and kidney dysfunction because it was difficult to determine the true baseline values for each patient and to identify whether an abnormal laboratory value was associated with sepsis or septic shock. We only included patients with sepsis or septic shock in our analysis, and we cannot draw conclusions about the likelihood that risk factors are associated with the onset of sepsis or septic shock.

Conclusions
Our data indicate that, across all age groups, most adult patients with sepsis have chronic underlying illnesses, and a substantial percentage have prehospital opportunities for care that could be used to disrupt the progression from infection to sepsis, hospital admission and, in almost one-third of cases, death. Public health and medical professionals can work to ensure that sepsis educational initiatives reach a wide array of outpatient health care settings and practitioners as well as patients and hospital-based practitioners to raise awareness of sepsis as an important public health problem.