Mitigating structural racism to reduce inequities in sepsis outcomes: a mixed methods, longitudinal intervention study

Our team’s prior seminal work (the Leadership Saves Lives/LSL study), a novel, evidence-based organizational culture change intervention that focused on 30-day risk-standardized mortality rates (RSMR) for acute myocardial infarction (AMI), provides a robust foundation for the proposed study [29]. We intervened in 10 US hospital systems drawn from the Mayo Clinic Care Network (MCCN), working for 2 years with Guiding Coalitions (e.g., a wide range of clinicians, administrators and others both within and outside of the hospital), to reduce RSMR for AMI through improving key domains of organizational culture (i.e., learning and problem solving, senior management support, psychological safety). The 6 hospitals that demonstrated substantial shifts in culture showed a significantly greater decrease in RSMR than the 4 hospitals that did not experience such culture changes (RSMR reduced by 1.07 percentage points and 0.23 percentage points, respectively; P = 0.027). The evidence generated through LSL demonstrates that a coalition-based leadership intervention can change organizational culture with impact on a major clinical outcome [27]. Mechanisms for change included collaborating across historical, political, and organizational boundaries [30‐32]. The LSL intervention required close collaboration with organizations outside of the hospital; hospitals reported planning to ‘transfer LSL’ to other complex health outcomes such as heart failure and stroke, suggesting its broader applicability. In addition, our prior research has also shown that health system links with non-health community organizations are associated with improved readmission rates [33], and that efforts to generate social capital in surrounding geographies and invest in social supports such as housing and community engagement are tied to better outcomes for a number of complex health conditions [34‐36]. We anticipate similar types of collaborations with health care and non-health care partners will be developed by Guiding Coalitions in the proposed study. Finally, our validated measure of hospital organizational culture [28] will be adapted to address the domains of organizational culture that are required to mitigate structural racism, another important methodological advancement of the proposed study.

AA/B and LatinX populations have higher rates of complications leading to sepsis [37], deviations from standard sepsis care, higher in-hospital mortality, and higher all-cause and sepsis readmissions compared with NHW populations [7, 10]. Sepsis-related mortality for AA/B patients is nearly double that for NHWs across all age groups (rate ratio = 1.78) nationally [38]. AA/B patients are twice as likely to die from severe infections driven by antibiotic resistance as compared to NHWs; many of these are sepsis deaths [21]. Racial inequities are observed at all stages in the care pathway for patients at risk of sepsis. At the level of primary prevention, AA/B and LatinX people experience higher levels of chronic comorbidities, and lower rates of access to primary care [17, 19, 20], including vaccination (which reduces incidence of sepsis) [16]. At the level of care seeking, AA/B and LatinX patients have well-justified mistrust in engaging with the health care system [14], demonstrate lower awareness of sepsis [39], receive less timely and accurate triage [18, 40], and receive lower intensity use of diagnostics and treatment (e.g., less likely to receive ICU care, longer length of stay, less frequent referral for follow-up care). AA/B patients are more likely to experience poverty and lack of insurance [8], both of which are associated with poorer care and outcomes along the care continuum, although many of these inequities persist when controlling for insurance status [41‐43]. Finally, AA/B patients experience implicit bias at the provider and health care system levels; however, currently no studies examine the role of implicit bias in sepsis care [44‐46]. Sepsis Alliance, the nation’s first and leading sepsis organization, has emphasized the urgency of addressing racial inequities in sepsis care, prioritizing this issue in its mission and issuing a national call to action [21].

The conceptual model below (Fig. 1), which we developed based on synthesis of existing evidence on race-based inequities in healthcare and outcomes, shows how racism is manifest across the sepsis care continuum, and provides a starting point for our work to measure and intervene on the domains of organizational culture that are required to mitigate structural racism [8, 14, 47‐50]. At the foundational levels of the framework are ecological barriers to health equity, factors that are deeply embedded in our US society that perpetuate structural racism [51]. Through facilitated collaboration, Guiding Coalitions will increase their home systems’ attention to these factors. The next level defines domains of organizational culture associated with organizational capacity to mitigate structural racism. These domains, derived from our own work in organizational culture change and consistent with capacity for whole systems transformation, are the capacities that will be strengthened through the proposed intervention [27, 29, 30, 32]. The next level defines ways in which structural racism impacts the sepsis care pathway, moving from risk and primary prevention on the left to standards-based follow up care on the right. Participating sites will use their own data to name and prioritize the inequities most salient in their own systems and apply a proven strategic problem-solving process [52] to understand and address these symptoms of structural racism. Finally, at the highest level, we will measure the impact of these changes on racial inequities in care and outcomes for patients with sepsis.

Health systems and their surrounding communities will be recruited from the membership of the MCCN, a group of independent health systems that purchase access to Mayo Clinic's tools and expertise, and that are vetted and selected for their commitment to high-quality patient experience and outcomes. These members represent all four census regions and all nine census divisions, serving sufficiently diverse patient populations to demonstrate the changes in racial inequities anticipated through the proposed work. We will use random sampling with a purposeful component [56] to select sites that include at least 5% proportion of AA/B clients and are diverse in geography and teaching status. Of MCCN’s 37 domestic (US) members, approximately 20 have a minimum patient population greater than 5% AA/B [average 12%]. We will work with each site to establish a Guiding Coalition [57] that represents the components of the system (within and beyond the health sector) required to address the impact of structural racism on sepsis care and outcomes (Table 1). Each site will be asked to name four core coalition members who will serve as senior champions for the project representing health systems administration, clinical leadership, and community engagement. These core members will participate in the cross-hospital forums and will lead the creation of the full coalition in their site, adapting guidance for coalition membership and structure to their unique context.

Based on our prior success in similar organizational change efforts [29], the 2.5 year intervention will include three components: 1) a series of five semiannual virtual, cross-site forums attended by four key members of each Guiding Coalition; 2) a series of four one-day workshops onsite with the full coalition at each hospital; and 3) a web-based platform to allow sites to share experiences and to serve as a repository for program resources. The content will build on and extend the previously published LSL curriculum [29]. Semiannual forums will bring sites together as a learning community for advice and problem solving on culture change to address structural racism in sepsis care. The sequence of forums will move sites from understanding and ‘buying in’ to the evidence base to sharing implementation challenges and successes. In the annual and closing forums, sites will vote for the “STAR” (Striving to Achieve the Remarkable) site that best exemplifies commitment to the objectives of the learning community, a highly motivational activity in LSL. On site workshops will build leadership capacity within the full Guiding Coalition to mitigate structural racism as it manifests in sepsis care. Each of the four workshops will include one full day of content, scheduled based on site preferences. The core curriculum for the workshops will incude both how to work (improving the domains of organizational culture that are required to mitigate structural racism and the underlying ecological barriers to health equity) and what to work on (strategic problem solving [52] to address root causes of inequities in sepsis care and outcomes in their site, see Additional File 1; Supplemental Fig. 2 for examples). In the first workshop, we will orient Guiding Coalition members to the intervention and the evidence base, promote reflection on their own measures of inequities in sepsis care and baseline organizational culture (survey measure described below), and name a problem statement and objective on which the group will focus their root cause analysis process. We will also invest in effective working relationships and processes among coalition members, including representation and role clarity [58], decision-making [59], and accountability [60]. Between workshops, the coalitions will be tasked with making and measuring progress toward addressing identified priority root causes of racial inequities in sepsis care in their home systems. At subsequent workshops, members will report on their progress, with the goal of solving implementation challenges and further developing their individual and group capacity for adaptive leadership [25, 26]. Additional workshop content will include diagnosing and shaping organizational culture, engaging conflict productively [61], using levels of analysis to diagnose organizational challenges [62, 63], and working with power and hierarchy [64, 65]. Content will be tailored to the local context by adjusting the timing of these modules to meet teams’ most pressing needs, adapting the specific examples and experiential learning exercises used in each module, articulating linkages between these content areas and ongoing work in each site, and encouraging each site to focus on root causes of inequities in sepsis care and treatment that are most salient in their environment. The facilitation team will include expert facilitators from across the United States, with backgrounds in organizational development, leadership education, anti-racist organizational change, and clinical quality improvement; most facilitators will be drawn from a faculty pool with deep experience developed through prior successful projects. Facilitators will work in teams of two or three, including a lead facilitator and additional experts tailored to the needs of each site. To promote standardization across sites while allowing for adaptation to local context, facilitators will use a fidelity checklist for each workshop. At the end of each workshop, the lead facilitator will prepare a structured written summary of decisions and action items for report back to the Guiding Coalition and to the full facilitation team. Between workshops, a web-based information sharing platform will serve two primary functions: (1) to serve as an accessible, up-to-date repository of project materials and references and (2) to support direct communication across hospital teams for sharing of successes, barriers, and project updates.

All Guiding Coalition members at each system (n = 8 x ~ 15 members = 120) will be invited to participate in a survey using a novel instrument drawn from previously validated scales (Additional File 3) [28], which includes 30 items within 4 domains: learning and problem solving, senior leadership support, structures and processes that support change, and psychological safety. Example item: “We have goals and metrics in place to guide our efforts to reduce racial inequities in care of patients with sepsis.” Responses will be entered using the sliding bar function in Qualtrics, with anchors “never” to “always.” To ensure that our instrument captures new concepts related to anti-racist practice, we will conduct cognitive interviews [70, 71] (n = 8–10) with respondents similar to Guiding Coalition members, and revise our draft instrument as needed [28, 72‐75]. Participants will complete the secure, web-based survey at the launch, midpoint and end of the intervention. To assess changes in 4 domains of organizational culture and overall, we will use hierarchical generalized linear modeling to estimate the association between mean scores and time, accounting for clustering of individual respondents within systems. We will use statistical analysis to determine significant changes at the level p < 0.05 for overall culture and each of the 4 domains for each system and the full sample. Analyses will be conducted using SAS V.9.4 and independently corroborated by two analysts. In addition to assessing the impact of the intervention in this study, the survey will be the first (to our knowledge) to measure aspects of organizational culture required for anti-racist systems change.

In-depth, in-person interviews will be conducted by trained members of the research team with different disciplinary backgrounds, using a standard interview guide (Additional File 4), beginning with a grand tour question [76]: “Tell me about your experiences in working to address racial inequities in sepsis care,” with structured probes to encourage participants to describe specific examples or vignettes. The sample will include Guiding Coalition members and additional key informants [56] as needed to achieve thematic saturation [77] at each site. As is typical with in-depth interviewing [78], we anticipate interviews will be approximately 1 h in length. Interviews will be audiotaped and transcribed by independent, professional transcriptionists to enhance data reliability. At each site, we will also observe 2–8 h of daily routines in relevant community and hospital settings tailored to the sepsis project at each site.

Observations will be completed by an experienced researcher using a standard observation template (Additional File 5) [79, 80] including context, stakeholders, and events, synthesized as field notes at the completion of the site visit. Intervention teams will participate in a debriefing session immediately upon return from site visits; reflections will be included in the qualitative data set for analysis. Analysis: Transcribed interview and ethnographic data will be analyzed by a 5-person multidisciplinary team using the constant comparative method of qualitative data analysis [81]. Coding of the data will include a series of iterative steps, beginning with an initial ‘start list’ of codes [82] to be refined during analysis of transcripts from subsequent interviews. We will conduct iterative coding and analysis with each wave of data collection, using negotiated consensus to resolve differences in interpretation, until a final code structure is established [82]. We will implement longitudinal analysis using matrices (rows = single case, group or theme, columns = time) to uncover patterns and interrelationships in change over time within and across health systems [83]. Data will be entered into ATLAS.ti to facilitate analysis. We will systematically search for disconfirming evidence [84], interview multiple participants in each health system for triangulation [85], and maintain a detailed audit trail to document analytic decisions [76]. To our knowledge, this will be one of the largest longitudinal qualitative data sets in the country [83], and could support sub-investigations on the nature of structural racism and health [31, 86].

Integration will be accomplished through merging in a mixed methods matrix [69, 87] at the final stage of analysis (Additional File: 1 Supplemental Fig. 4). For each site, we will integrate quantitative trend data (overall and by each domain of the survey) and qualitative data. We will classify each site as having positive or no change based on meeting either quantitative (statistically significant change in overall readiness score from baseline to follow up) or qualitative criteria (marked shift in culture characterized by substantial, consistent, specific illustrations of notable changes in three to four of the domains of culture and little to no disconfirming evidence). We will use comparative analysis to describe differences between the systems that achieve substantial change and those that do not. Finally, we will examine differences in clinical outcomes as described in Aim 3 below. Note: analysts will be blinded to all clinical outcomes data during this analysis.

We will conduct a comparative interrupted times series (ITS) analysis to examine the relation between the intervention and racial/ethnic differences in sepsis clinical outcomes between AA/B and NHW patients and between LatinX and NHWs. The analysis will be performed separately for our three key outcomes: 1) time to antibiotics, 2) in-hospital mortality, and 30-day readmission and stratified by response to the intervention (change/no change) (Additional File 1; Supplemental Fig. 5). We will consider a pre-exposure to the intervention time beginning 2.5 years (10 quarters) prior to starting the intervention and a postexposure period comprising 2.5 years (10 quarters) following the start of the intervention, with estimates made at every calendar quarter. ITS will be utilized to determine the difference in quarterly rates over time between changers (systems that responded to the intervention) and non-changers, health systems that did not respond to the intervention. Subsequently, ITS allows us to detect if an intervention has impact above and beyond secular trends when randomization is not a feasible option, and is ideal for evaluation of data collected across multiple equally spread time points by testing the changes in the slope of an outcome over time [96]. Additionally, ITS requires at least eight observations pre and post intervention [97]. We will utilize generalized mixed effect models in conjunction with segmented regression to evaluate the effectiveness of our intervention [98]. The model will control for patient-level characteristics, including age, sex, insurance status, Charlson Comorbidity index, and admission through the Emergency Department or transfer from another acute care facility, and hospital-level characteristics including size, academic vs community hospital, number of intensive care beds, annual sepsis case volume. We will also account for seasonality based on calendar quarter by including a “season” term alone and interacted with the treatment indicator and will assess for any residual autocorrelation using autoregressive integrated moving average (ARIMA). To allow a marginal interpretation of results, we will use a linear probability model for all analyses. Because outcomes of patients within a hospital are expected to be correlated, we will account for these non-standard variance–covariance structures by clustering at the hospital level. All coefficients will be modeled as fixed effects.