Institutional Structures and Processes to Support Sepsis Care: A Multihospital Study

Abstract

OBJECTIVES:

To identify opportunities for improving hospital-based sepsis care and to inform an ongoing statewide quality improvement initiative in Michigan.

DESIGN:

Surveys on hospital sepsis processes, including a self-assessment of practices using a 3-point Likert scale, were administered to 51 hospitals participating in the Michigan Hospital Medicine Safety Consortium, a Collaborative Quality Initiative sponsored by Blue Cross Blue Shield of Michigan, at two time points (2020, 2022). Forty-eight hospitals also submitted sepsis protocols for structured review.

SETTING:

Multicenter quality improvement consortium.

SUBJECTS:

Fifty-one hospitals in Michigan.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

Of the included hospitals, 92.2% (n = 47/51) were nonprofit, 88.2% (n = 45/51) urban, 11.8% (n = 6/51) rural, and 80.4% (n = 41/51) teaching hospitals. One hundred percent (n = 51/51) responded to the survey, and 94.1% (n = 48/51) provided a sepsis policy/protocol. All surveyed hospitals used at least one quality improvement approach, including audit/feedback (98.0%, n = 50/51) and/or clinician education (68.6%, n = 35/51). Protocols included the Sepsis-1 (18.8%, n = 9/48) or Sepsis-2 (31.3%, n = 15/48) definitions; none (n = 0/48) used Sepsis-3. All hospitals (n = 51/51) used at least one process to facilitate rapid sepsis treatment, including order sets (96.1%, n = 49/51) and/or stocking of commonly used antibiotics in at least one clinical setting (92.2%, n = 47/51). Treatment protocols included guidance on antimicrobial therapy (68.8%, n = 33/48), fluid resuscitation (70.8%, n = 34/48), and vasopressor administration (62.5%, n = 30/48). On self-assessment, hospitals reported the lowest scores for peridischarge practices, including screening for cognitive impairment (2.0%, n = 1/51 responded “we are good at this”) and providing anticipatory guidance (3.9%, n = 2/51). There were no meaningful associations of the Centers for Medicare and Medicaid Services’ Severe Sepsis and Septic Shock: Management Bundle performance with differences in hospital characteristics or sepsis policy document characteristics.

CONCLUSIONS:

Most hospitals used audit/feedback, order sets, and clinician education to facilitate sepsis care. Hospitals did not consistently incorporate organ dysfunction criteria into sepsis definitions. Existing processes focused on early recognition and treatment rather than recovery-based practices.

KEY POINTS

Question: What structures and processes are used to support hospital-based sepsis care?

Findings: In this diverse multihospital quality improvement consortium, most hospitals had sepsis committees (98.0%, n = 50/51) and used audit and feedback (98.0%, n = 50/51), order sets (96.1%, n = 49/51), and clinician education (68.6%, n = 35/51) to facilitate sepsis care. Treatment protocols included guidance on antimicrobial therapy (68.8%, n = 33/48), fluid resuscitation (70.8%, n = 34/48), and vasopressor administration (62.5%, n = 30/48), and multiple recovery-based practices were rated lowest on hospital self-assessment.

Meaning: Hospitals use a variety of structures and processes to promote early sepsis treatment and identify later sepsis care as an area for improvement.

Sepsis is a leading cause of hospitalization in the United States and is estimated to contribute to approximately half of all hospital deaths (1, 2). Furthermore, patients who survive sepsis are at increased risk for new functional limitations, cognitive impairment, hospital readmission, and death for at least several months after the acute sepsis event resolves (3, 4). In recognition of its substantial morbidity and mortality, sepsis was identified as a global health priority by the World Health Organization in 2017 (5).

However, despite the global attention to sepsis, little is known about institutional approaches to implementing sepsis care. We evaluated the structures and processes in place to support sepsis care at hospitals across the state of Michigan. Specifically, we sought to understand the institutional processes for identifying, defining, and treating sepsis, as well as for educating clinicians and monitoring and improving sepsis care and outcomes. To do this, we surveyed 51 Michigan hospitals about sepsis policies and care processes at two different time points (2020, 2022). There were multiple goals to this work: 1) to understand hospital-based approaches to supporting sepsis care, 2) to establish a baseline assessment of sepsis practices in Michigan hospitals, 3) to identify opportunities to improve sepsis care, 4) to inform the ongoing quality improvement (QI) work of the Michigan Hospital Medicine Safety (HMS) Consortium, and 5) to provide other hospitals with a framework for evaluating current hospital-based sepsis practices and opportunities for improvement.

METHODS

Hospital Medicine Safety Consortium

The Michigan HMS Consortium is a Collaborative Quality Initiative sponsored by Blue Cross Blue Shield of Michigan (BCBSM) and Blue Care Network that aims to improve care and outcomes of hospitalized medical patients in the state of Michigan. The setting and design of HMS have been described previously (6, 7). Starting in 2020, participation in HMS was mandatory for all hospitals enrolled in BCBSM’s Value Partnership Program, which includes most large hospitals in Michigan. Of the 92 noncritical access, nongovernmental hospitals in Michigan, 51 (55.4%) participated in HMS during the study period.

Survey on Sepsis Practices

HMS surveys member hospitals twice yearly on care structures and processes to support ongoing QI work. Surveys are administered via Qualtrics (Qualtrics, Seattle, WA) to each hospital’s designated HMS representative who compiles answers from providers and staff at their institution. We surveyed 51 hospitals participating in HMS in December 2020 (time 1) and again in February 2022 (time 2). The survey included open- and closed-ended questions focused on sepsis QI approaches, sepsis recognition and treatment, antimicrobial stewardship, and self-assessment of sepsis practices. Specifically, hospitals were asked to evaluate 15 hospital-based practices using a 3-point Likert scale: 1) “we need to improve a lot,” 2) “we need to improve a little,” and 3) “we are good at this practice.” All sepsis-related survey questions are included in Appendix 1 (http://links.lww.com/CCX/B271) (2020) and Appendix 2 (http://links.lww.com/CCX/B271) (2022).

Review of Hospital Sepsis Documents

During the 2020 survey, hospitals were also asked to submit their institutional sepsis policies or protocols for review. Two authors (M.K.L., J.K.H.) independently reviewed the submitted sepsis documents. A guidebook was created to facilitate classification of institutional sepsis definitions and treatment recommendations. Hospital definitions of sepsis were classified as: “Sepsis-1” (infection + ≥ 2 basic systemic inflammatory response syndrome [SIRS] criteria) (8), “Sepsis-2” (infection + ≥ 2 diagnostic criteria) (9), “Sepsis-3” (infection + acute organ dysfunction) (10), “Hybrid” (infection + ≥ 2 basic SIRS criteria + acute organ dysfunction), “other” (sepsis definition provided but did not map to existing international sepsis definitions, e.g., SIRS-only or modified SIRS criteria), or “not defined” (eTables 1 and 2, http://links.lww.com/CCX/B271). Statements regarding antimicrobial therapy, fluid resuscitation, and vasopressor administration were abstracted and classified as “recommendations” (things one should do) vs. “suggestions” (things one could do). Differences in classification between the two reviewers were reconciled through discussion and a third reviewer (H.C.P.) as needed.

Statistical Analysis

Results of the sepsis survey and hospital sepsis document review are presented using standard descriptive statistics. We focus on the 2020 survey responses since these were collected at the same time as the hospital sepsis documents, but we report full results of the 2020 and 2022 survey responses in the online supplement. We assessed for differences in 2022 vs. 2020 survey self-assessment responses using McNemar’s test of paired data. We assessed for differences in hospital characteristics, sepsis document characteristics, and hospital self-assessments by hospital’s performance on the Centers for Medicare and Medicaid Services’ Severe Sepsis and Septic Shock: Management Bundle (SEP-1) measure using chi-square and two-sided tests. SEP-1 performance for 2021 was extracted from data.cms.gov, and hospital performance was classified as above median vs. at or below median for the HMS consortium. Data management and analysis were completed in Microsoft Access (Microsoft, Redmond, WA) and SAS, Version 9.4 (SAS Institute, Cary, NC). No Institutional Review Board review was necessary because this study did not fall under the board’s guidelines as human subjects research.

RESULTS

All 51 hospitals participating in HMS in 2020 completed the December 2020 sepsis survey, and all completed the repeat survey in 2022. Of the 51 hospitals, 94.1% (n = 48/51) provided a hospital sepsis document for review. Hospital characteristics are presented in Table 1. To summarize, 92.2% (n = 47/51) were nonprofit, 88.2% (n = 45/51) were urban, 11.8% (n = 6/51) were rural, and 80.4% (n = 41/51) were teaching hospitals. Mean hospital size was 304 beds (range, 36–915).

TABLE 1.

Hospital Demographics, Sepsis Committee Composition, and Quality Improvement Practices

Hospital Characteristics	All Hospitals in the Cohort (n = 51)	SEP-1 Performance: Median or Lower (n = 29)	SEP-1 Performance: Greater than Median (n = 22)	p
SEP-1 performance, median (IQR)	51 (44–66)	44 (38–49)	62 (58–67)	Not applicable
Hospital bed size,a median (IQR)	317 (191–442)	360 (219–443)	292 (191–422)	0.449
Hospital bed size category,a n (%)				0.481
< 100	6 (11.8)	4 (13.8)	2 (9.1)
100–249	11 (21.6)	4 (13.8)	7 (31.8)
250–499	24 (47.1)	15 (51.7)	9 (40.9)
≥ 500	10 (19.6)	6 (20.7)	4 (18.2)
Urbanicity,b n (%)				0.383
Urban	45 (88.2)	27 (93.1)	18 (81.8)
Rural	6 (11.8)	2 (6.9)	4 (18.2)
Ownership status,c n (%)				0.625
Nonprofit	47 (92.2)	26 (89.7)	21 (95.5)
For-profit	4 (7.8)	3 (10.3)	1 (4.5)
Teaching hospital,c n (%)	41 (80.4)	28 (96.6)	13 (59.1)	0.001
Sepsis committee presence,d n (%)	50 (98.0)	29 (100.0)	21 (95.5)	0.431
Sepsis committee composition,d n (%)				0.976
Quality improvement staff	50 (98.0)	29 (100.0)	21 (95.5)
Physician	49 (96.1)	28 (96.6)	21 (95.5)
Pharmacist	46 (90.2)	27 (93.1)	19 (86.4)
Antibiotic stewardship representation	46 (90.2)	26 (89.7)	20 (90.9)
Nursing	46 (90.2)	27 (93.1)	19 (86.4)
Executive leadership	35 (68.6)	22 (62.9)	13 (59.1)
Rapid response team	32 (62.7)	22 (75.9)	10 (45.5)
Infectious diseases representation	31 (60.8)	18 (62.1)	13 (59.1)
Information and technology	30 (58.8)	16 (55.2)	14 (63.6)
Patient or family representative	1 (2.0)	1 (3.4)	0 (0.0)
Othere	26 (51.0)	17 (58.6)	9 (40.9)
Quality improvement practices,d n (%)				0.818
Audit and feedback	50 (98.0)	28 (96.6)	22 (100.0)
Formal education	35 (68.6)	22 (75.9)	13 (59.1)
Othere	27 (52.9)	16 (55.2)	11 (50.0)

IQR = interquartile range, SEP-1 = the Centers for Medicare and Medicaid Services’ Severe Sepsis and Septic Shock: Management Bundle.

^aData obtained from 2020 Michigan Certificate of Need Annual Survey, Basic Total Licensed Beds Utilization Statistics. Available at: https://www.michigan.gov/mdhhs/-/media/Project/Websites/mdhhs/Doing-Business-with-MDHHS/Health-Care-Providers/Certificate-of-Need/CON-Eval/Survey-Reports/2020/Beds/Report-010-Hospital-Beds-by-HSA.pdf?rev=16e5fcc69b6d4a03b17813f6028a34c3&hash=249DA44852A4FC41F010A3EA9D196034. Accessed June 21, 2022.

^bClassified as rural vs. urban based on Rural-Urban Continuum Codes (RUCCs). RUCC 1 = counties in metro areas of 1 million population or more (n = 25/48 hospital documents, n = 25/51 hospital survey responses); RUCC 2 = counties in metro areas of 250,000 to 1 million population (n = 13/48, n = 14/51); RUCC 3 = counties in metro areas of fewer than 250,000 population (n = 4/48, n = 6/51); RUCC 4 = urban population of 20,000 or more, adjacent to a metro area (n = 0/48, n = 0/51); RUCC 5 = urban population of 20,000 or more, not adjacent to a metro area (n = 1/48, n = 1/51); RUCC 6 = urban population of 2,500 to 19,999, adjacent to a metro area (n = 2/48, n = 2/51); RUCC 7 = urban population of 2,500 to 19,999, not adjacent to a metro area (n = 3/48, n = 3/51); RUCC 8 = completely rural or < 2,500 urban population, adjacent to a metro area (n = 0/48, n = 0/51); and RUCC 9 = completely rural or < 2,500 urban population, not adjacent to a metro area (n = 0/48, n = 0/51). Data obtained from USDA 2013 Rural-Urban Continuum Codes. Available at: https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx. Accessed July 18, 2022.

^cData obtained from AHA’s Data Hub. Available at: https://guide.prod.iam.aha.org/guide/searchResults. Accessed April 28, 2021.

^dData obtained from Hospital Medicine Safety sepsis surveys administered to 51 participating hospitals in 2020.

^eSee eTable 3 (http://links.lww.com/CCX/B271) for description of items included in “other.”

Committee Structure and Quality Improvement Approaches

In 2020, 98.0% (n = 50/51) of hospitals reported having an institutional sepsis committee. Among these 50 hospitals, the composition of sepsis committees included representation from QI staff (100%, n = 50/50), physicians (98.0%, n = 49/50), nursing (94.0%, n = 47/50), pharmacists (92.0%, n = 46/50), antimicrobial stewardship team members (92.0%, n = 46/50), executive leadership (70.0%, n = 35/50), rapid response team members (64.0%, n = 32/50), infectious diseases (ID) specialists (e.g., ID physicians or pharmacists; 62.0%, n = 31/50), information and technology staff (60.0%, n = 30/50), and patient or family representatives (2.0%, n = 1/50) (Table 1). In 2022, 94.1% of hospitals (n = 48/51) had a sepsis committee. Compared with when hospitals were surveyed in 2020, committees less often included QI, antimicrobial stewardship, rapid response, or information technology representation in 2022 (eTable 3, http://links.lww.com/CCX/B271).

In 2020, all hospitals (100%, n = 51/51) used at least one QI approach to advance sepsis care in their institution. Among the included hospitals, 98.0% (n = 50/51) used audit and feedback for sepsis recognition (92.2%, n = 47/51), treatment (96.1%, n = 49/51), and/or outcomes (92.2%, n = 47/51); 68.6% (n = 35/51) provided formal education to medical personnel, including those practicing in the emergency department (62.7%, n = 32/51), hospital wards (62.7%, n = 32/51), and intensive care unit (ICU) (54.9%, n = 28/51) (Table 1); 60.8% (n = 31/51) had educational initiatives directed toward physicians, and 66.7% (n = 34/51) provided educational initiatives directed toward nurses (p = 0.38). In 2022, responses were similar apart from reduced use of audit and feedback targeted to either the hospital or unit level (eTable 3, http://links.lww.com/CCX/B271).

Approaches for Early Sepsis Identification and Treatment: Survey Results

In 2020, nearly all hospitals (98.0%, n = 50/51) had a standardized protocol or process for identifying sepsis, and many hospitals (80.4%, n = 41/51) had a formal process to screen for sepsis during transitions of care, such as upon presentation to the emergency department (72.5%, n = 37/51), admission to the hospital (52.9%, n = 27/51), and/or transfer to the ICU (37.3%, n = 19/51). Many hospitals (64.7%, n = 33/51) had a nurse-driven screening protocol (e.g., standing orders that nurses can activate to initiate sepsis evaluation), and 58.8% (n = 30/51) had a standard screening for sepsis during each nursing shift. Responses were similar for 2022 (eTable 3, http://links.lww.com/CCX/B271).

Hospitals reported using a variety of criteria to help identify sepsis, most commonly vital signs—including heart rate, respiratory rate, blood pressure, and temperature—(100%, n = 51/51), white blood cell count (98.0%, n = 50/51), lactate (92.2%, n = 47/51), known infection (92.2%, n = 47/51), and signs or symptoms of potential infection (88.2%, n = 45/51). Signs of organ dysfunction were also used to help identify sepsis in many hospitals, most commonly including abnormalities of creatinine (82.4%, n = 42/51), platelet count (80.4%, n = 41/51), bilirubin (76.5%, n = 39/51), and altered mental status (68.6%, n = 35/51).

All hospitals (n = 51/51) reported using at least one formal process to facilitate rapid sepsis treatment, including use of a formal order set (96.1%, n = 49/51) and stocking of commonly used antibiotics in at least one clinical setting (92.2%, n = 47/51), such as the emergency department (92.2%, n = 47/51), ICU, (70.6%, n = 36/51), and/or hospital wards (60.8%, n = 31/51). About one-third of hospitals (31.4%, n = 16/51) used nurse-driven protocols to expedite the initiation of sepsis treatment regimens.

In 2020, 96.0% (n = 48/51) of hospitals reported promoting antimicrobial stewardship during sepsis care. Common strategies included review of antibiotics for appropriateness (70.6%, n = 36/51), provision of an antibiogram (76.5%, n = 39/51), provision of education regarding the importance of antimicrobial stewardship to emergency department staff (70.6%, n = 36/51) and/or inpatient staff (62.7%, n = 32/51), and assisting clinicians in antimicrobial selection (78.4%, n = 40/51) (eTable 3, http://links.lww.com/CCX/B271). Hospitals primarily reported that principles of antibiotic stewardship were incorporated into their sepsis care well (49.0%, n = 25/51) or extremely well (23.5%, n = 12/51).

Institutional Sepsis Definitions and Identification: Document Review

Of the 48 hospital sepsis documents received, 66.7% (n = 32/48) included a specific definition of sepsis used at their institution, and 79.2% (n = 38/48) included information on their institution’s sepsis screening process. On double-review, institutional sepsis definitions were classified as Sepsis-1 in 18.8% (n = 9/48), Sepsis-2 in 31.3% (n = 15/48), Hybrid in 6.3% (n = 3/48), and other in 10.4% (n = 5/48). No hospitals used the Sepsis-3 definition (eTable 4, http://links.lww.com/CCX/B271). Institutional guidance on sepsis screening included information on which patients to screen (89.6%, n = 43/48), suggestions for sepsis reassessment (22.9%, n = 11/48), and guidance on specific criteria for sepsis identification, with the most common criteria being infection (60.4%, n = 29/48) and SIRS criteria (45.8%, n = 22/48). Organ dysfunction was included in the hospital sepsis definition in only 6.3% (n = 3/48) of submitted protocols.

Institutional Treatment Recommendations: Document Review

All 48 hospital sepsis documents included treatment recommendations, but the scope of guidance differed. Of the submitted documents, 68.8% (n = 33/48) provided guidance on antimicrobial therapy, 70.8% (n = 34/48) on fluid resuscitation, and 62.5% (n = 30/48) on vasopressor administration; 62.5% (n = 30/48) addressed all three domains (eFig. 1, http://links.lww.com/CCX/B271). Among the 33 hospital protocols that provided guidance on antimicrobial therapy, 93.9% (n = 31/33) recommended (n = 29) or suggested (n = 2) blood cultures be drawn before antibiotic administration. Recommendations on antimicrobial timing were provided by 87.9% (n = 29/33), including 54.5% (n = 18/33) that recommended administration within 1 hour, and 33.3% (n = 11/33) that recommended administration within 3 hours. Guidance on antimicrobial selection was provided by 69.7% (n = 23/33), including 69.7% (n = 23/33) that recommended (n = 13) or suggested (n = 10) using broad-spectrum coverage and 39.4% (n = 13/33) that provided specific recommendations (n = 6) or suggestions (n = 7) for antibiotic selection by site of infection. Only one document (3.0%) included suggestions for antibiotic stewardship, including guidance on duration of therapy and de-escalation (eFig. 1, http://links.lww.com/CCX/B271).

Among the 34 hospital sepsis documents providing guidance on fluid resuscitation, 85.3% (n = 29/34) specified indication(s) to administer fluids. The most common indications were hypotension (79.4%, n = 27/34), including mean arterial pressure less than 65 mm Hg (76.5%, n = 26/34), systolic blood pressure less than 90 mm Hg (73.5%, n = 25/34), and systolic blood pressure greater than or equal to 40 mm Hg below baseline (35.3%, n = 12/34). Additionally, 70.6% (n = 24/34) used lactate greater than 4 mmol/L, 5.9% (n = 2/34) lactate greater than 2 mmol/L, and 5.9% (n = 2/34) decrease in urine output (< 0.5 mL/kg/hr for 2 hr) as criteria to initiate fluid resuscitation (eFig. 1, http://links.lww.com/CCX/B271). Contraindications to fluid resuscitation were identified in 23.5% (n = 8/34) of documents that addressed fluids, most commonly heart failure (11.8%, n = 4/34), reduced left ventricular ejection fraction (11.8%, n = 4/34), end-stage renal disease (8.8%, n = 3/34), presence of a left ventricular assist device (5.9%, n = 2/34), and patient/family refusal (5.9%, n = 2/34). Fluid volume recommendations were included in 91.2% (n = 31/34) of documents addressing fluid resuscitation, including 88.2% (n = 30/34) that recommended (n = 29) or suggested (n = 1) 30 mL/kg, 11.8% (n = 4/34) that recommended (n = 1) or suggested (n = 3) administering fluid in 1 L increments, and 8.8% (n = 3/34) that recommended an initial 500 mL bolus (eFig. 1, http://links.lww.com/CCX/B271). Recommendations on fluid type were included in 85.3% (n = 29/34), all of which recommended crystalloid fluids. Several documents (20.6%, n = 7/34) provided no guidance beyond recommending crystalloids; 32.4% (n = 11/34) suggested either 0.9% normal saline or balanced solutions; 29.4% (n = 10/34) recommended 0.9% normal saline; and 2.9% (n = 1/34) recommended balanced solutions.

Recommendations on vasopressor use were included in 30 of 48 documents (62.5%), including 15 (50.0%) that specified a vasopressor of choice (all norepinephrine). Indications for vasopressor initiation were identified in all of these 30 documents, most commonly hypotension (93.3%, n = 28/30), lactate greater than 4 mmol/L (13.3%, n = 4/30), lactate greater than 2 mmol/L (3.3%, n = 1/30), and persistent hypotension after fluid resuscitation (83.3%, n = 25/30) (eFig. 1, http://links.lww.com/CCX/B271).

Self-Assessment of Sepsis Practices: Survey Results

Hospital self-assessment was highest (i.e., had the highest proportion of “we are good at this” responses) for measurement of initial lactate (66.7% in 2020, 52.9% in 2022), medication reconciliation at discharge (49.0% in 2020, 45.1% in 2022), and daily reassessment of antibiotics (49.0% in 2020 and 2022). Self-assessment was lower for several other practices, including measurement of repeat lactate (33.3% in 2020, 21.6% in 2022), timely recognition of sepsis (31.4% in 2020, 27.5% in 2022), and scheduling timely outpatient follow-up (23.5% in 2020, 27.5% in 2022). In both 2020 and 2022, hospitals had the lowest self-assessment for the following practices: provision of an initial 30 mL/kg fluid bolus (5.9% in both years), screening for mental health impairment at discharge (3.9% in 2020, 5.9% in 2022), providing anticipatory guidance regarding potential new symptoms/morbidity after sepsis (3.9% in 2020, 5.9% in 2022), and screening for cognitive impairment at discharge (2.0% in 2020, 5.9% in 2022) (Fig. 1; and eFig. 2, http://links.lww.com/CCX/B271). Self-assessment responses were similar between 2020 and 2022 for all practices assessed (eTable 3, http://links.lww.com/CCX/B271).

Figure 1.

Hospital self-assessment of sepsis care practices. As part of the sepsis surveys administered in 2020 to 51 hospitals participating in the Michigan Hospital Medicine Safety Consortium, hospitals were asked to self-assess performance in various sepsis practices using a 3-point Likert scale: 1) “we need to improve a lot” (light green in figure), 2) “we need to improve a little” (medium green in figure), and 3) “we are good at this practice” (dark green in figure). See eFigure 2 (http://links.lww.com/CCX/B271) for a comparison with 2022 survey responses. *Providing anticipatory guidance to patients/family members regarding potential new symptoms/morbidity after sepsis.

Association With SEP-1 Performance Scores

SEP-1 performance scores were median 53 (interquartile range, 44–66; range, 29–84). Twenty-two hospitals were above median, whereas 29 were at or below median. Hospital characteristics stratified by SEP-1 performance are presented in Table 1. Of the twenty-two hospitals with above median SEP-1 performance, 59.1% (n = 13/22) were teaching vs. 96.6% (n = 28/29) of hospitals below median (p ≤ 0.001). Otherwise, hospital characteristics did not differ by SEP-1 performance. Likewise, there were no differences in institutional sepsis document characteristics (eTable 4, http://links.lww.com/CCX/B271). For hospital self-assessment of sepsis practices, 17.2% (n = 5/29) of hospitals with at or below median SEP-1 performance responded “we are good at this practice” for timely recognition of sepsis, whereas 50.0% (n = 11/22) of hospitals with greater than median SEP-1 performance responded “we are good at this practice,” (p = 0.013). For all other self-assessed sepsis practices, there were no statistically significant differences by SEP-1 performance (Table 2).

TABLE 2.

2020 Hospital Self-Assessment, Stratified by SEP-1 Performance

Hospital-Based Sepsis Practice	n (%) Reporting “My Hospital Is Good At This Practice”		p
	SEP-1 Performance: Median or Lower (n = 29)	SEP-1 Performance: Greater than Median (n = 22)
Early sepsis care
Measurement of initial lactate	19 (65.5)	15 (68.2)	0.842
Blood cultures before antibiotics	10 (34.5)	9 (40.9)	0.638
Delivery of antibiotics within 3 hr	6 (20.7)	10 (45.5)	0.059
Timely recognition of sepsis	5 (17.2)	11 (50.0)	0.013
Provision of fluid bolus	0 (0.0)	3 (13.6)	0.074
Ongoing sepsis care
Daily reassessment of antibiotics	15 (51.7)	10 (45.5)	0.657
Daily reassessment of diagnosis	9 (31.0)	9 (40.9)	0.465
Peridischarge care
Medication reconciliations	15 (51.7)	10 (45.5)	0.657
Scheduling of timely outpatient follow-up	7 (24.1)	5 (22.7)	0.906
Discussing goals of care	1 (3.5)	3 (13.6)	0.303
Screening for functional impairment	3 (10.3)	1 (4.6)	0.625
Providing anticipatory guidance	1 (3.5)	1 (4.6)	1.0
Screening for mental health impairment	1 (3.5)	1 (4.6)	1.0
Screening for cognitive impairment	1 (3.5)	0 (0.0)	1.0

SEP-1 = the Centers for Medicare and Medicaid Services’ Severe Sepsis and Septic Shock: Management Bundle.

DISCUSSION

In this study of 51 diverse hospitals in Michigan, the majority of hospitals had sepsis committees and used audit and feedback, order sets, and clinician education to facilitate sepsis care. However, the scope of these structures and processes varied and was rarely comprehensive. Most hospital policies and protocols focused on recognition and early management of sepsis, and only two-thirds included guidance on all three key domains of antimicrobial therapy, fluid resuscitation, and vasopressor administration. Additionally, while most hospitals indicated they have processes in place to promote antimicrobial stewardship and a high degree of confidence in their implementation of these practices, our review of policies revealed only one document that directly addressed aspects of antimicrobial stewardship within the protocol. This discordance suggests that hospital sepsis protocols may need to be updated to reflect current hospital practices.

A second finding of the study was that hospitals had better self-assessments of early vs. later sepsis care. Self-assessment was highest for measurement of initial lactate and daily reassessment of antibiotics/sepsis diagnosis. Meanwhile, multiple peridischarge and recovery-based practices, such as scheduling of timely outpatient follow-up, screening for new limitations, and provision of anticipatory guidance, were rated among the lowest of the practices evaluated. These findings are consistent with other studies showing low delivery of recovery-oriented practices and suggest that later sepsis care may be a key opportunity for improvement of hospital sepsis management (11, 12). Sepsis QI initiatives have traditionally focused on early recognition and treatment. Over the past decade, however, it has been recognized that sepsis contributes not only to acute mortality, but also to long-term morbidity, including functional decline and cognitive impairment (3, 13). Although recent guidelines have been updated to address peridischarge sepsis care, our findings suggest there is an opportunity for better implementation of these practices (14, 15).

A third finding of this study was that hospitals predominantly used SIRS criteria and presence of infection to identify sepsis, with fewer hospitals using acute organ dysfunction as a sign of sepsis. Most hospitals used either the Sepsis-1 or Sepsis-2 definition, and none used the Sepsis-3 definition. This shows a disconnect between current local and international sepsis definitions and highlights potential confusion following recent sepsis definition updates (16). Additionally, it may reflect differences in definitional criteria across SEP-1, diagnostic coding, and current international consensus definitions. It is important to recognize that sepsis may present as acute organ dysfunction without clear signs and symptoms of infection on initial presentation. Prior studies have shown that obvious signs of infection, such as fever, are associated with faster time-to-antimicrobials as compared with generic signs of illness severity, such as hypotension (17). However, delays in time-to-antimicrobials are more strongly associated with risk of mortality among patients with hypotension or shock, underscoring the importance of early recognition of sepsis in patients presenting with hypotension (18–20).

A fourth finding of the study was that while nearly all hospitals had an institutional sepsis committee, the composition of committees varied across hospitals. Most committees included representation from QI staff, nursing, antimicrobial stewardship, physicians, executive leadership, and rapid response teams, but almost none included patient or family representation. Participation among QI staff, rapid response, and information technology staff decreased from 2020 to 2022, which we hypothesize may reflect staffing shortages following the COVID-19 pandemic.

Our findings build on existing literature on sepsis QI (largely single-center studies), by reporting on structures and processes used across a diverse cohort of hospitals (21–23). Sepsis is a new focus for HMS, and our surveys and data collection were conducted before initiation of any QI work. As such, the findings presented reflect the baseline status of sepsis care in Michigan. Future work is needed to understand the association between hospital structures/policies and sepsis management and outcomes. We found no association between characteristics of institutional sepsis policies and SEP-1 performance category, but we hypothesize there may be associations with more granular measures of sepsis management.

Our study had several limitations. First, the hospitals in our study were limited to those participating in HMS, so may not extrapolate to other areas of the country. However, given the large, diverse sample of hospitals included in HMS, and several consistent findings across these hospitals, we suspect our findings are not unique to Michigan, but rather reflective of the broader U.S. landscape. Second, data were collected via survey, and it is possible that response options may not have fully reflected the range or nuance of hospital-based structures for sepsis. Third, hospital self-assessment may not reflect actual performance, although responses were notably consistent between 2020 and 2022 despite respondent turn-over. Fourth, we examined for differences in hospital structures/policies and SEP-1 performance but found no meaningful differences. We also evaluated for differences in hospital self-assessment of sepsis practices by SEP-1 performance and found only one practice (timely recognition of sepsis) that varied significantly by SEP-1 performance, with a higher percentage of hospitals with greater than median SEP-1 scores responding “we are good at this practice,” vs. hospitals with median or lower SEP-1 scores. For all other self-assessed practices, there were no meaningful differences by SEP-1 performance. Future work is needed to understand the association between hospital-based structures/policies and more granular measures of sepsis management and outcomes. Strengths of the study include the setting of a mature continuous QI consortium, which allowed for a 100% survey response rate. Furthermore, HMS representatives are accustomed to reaching out to relevant colleagues to best answer survey questions outside their scope of practice.

CONCLUSIONS

In this diverse multihospital cohort, we found that most hospitals relied on SIRS criteria and/or signs of infection to identify sepsis, with few hospitals incorporating organ dysfunction in their definition of sepsis. Expanding hospital sepsis definitions to include organ dysfunction criteria may better facilitate sepsis identification and rapid treatment initiation. Additionally, most hospitals used a multifaceted approach to implementing best practices for sepsis. The majority of hospitals had sepsis committees and used audit and feedback, order sets, and clinician education to facilitate sepsis care. The scope and content of hospital-based sepsis policies varied, with less than two-thirds addressing three core domains of antimicrobial therapy, fluid resuscitation, and vasopressors. On self-assessment, peridischarge and recovery-based practices were noted as areas for improvement. Standardizing sepsis structures and broadening processes to address peridischarge care may aid in implementation of these practices and help mitigate sepsis-related morbidity and mortality, both in the acute setting and postdischarge.

Supplementary Material

cc9-5-e1004-s001.pdf ^{(754.5KB, pdf)}