Reducing Maternal Mortality and Morbidity in Michigan: A Mixed Methods Analysis of the Impact of an Obstetric Hemorrhage Patient Safety Bundle

Abstract

The United States has the highest rate of maternal mortality despite outspending all other resource-rich countries in per-capita reproductive health care. Contributing to this problem are a combination of factors including disparities in patient socio-economic characteristics, changing population demographics, variation in medical practice, a fragmented healthcare system, and lack of quality reproductive health data. Nationally endorsed collections of evidence-based practices called patient safety bundles were developed to address this complex issue. Bundles present a structured way of improving the processes of care and include a set of evidence-based action steps that, when performed correctly and reliably, may contribute to improved patient outcomes. However, little is known about the real-world impact of bundles, including the optimal context, resources, and processes associated with implementation performance.

Since 2016, over 80 hospitals in Michigan participated in implementation of the Alliance for Innovation on Maternal Health’s obstetric hemorrhage patient safety bundle. The work presented in this dissertation assesses patient outcomes and factors that influenced the bundle implementation process across Michigan hospitals. An explanatory sequential mixed methods design informed by implementation science using the Tailored Implementation for Chronic Disease framework served as the basis for analysis of bundle implementation in Michigan hospitals between 2012 and 2019. Variation in, and factors associated with severe maternal morbidity were quantitatively evaluated across hospitals using the Michigan Inpatient Database (Aim 1). A qualitative study using in-depth semi-structured interviews with implementation stakeholders was conducted to assess experiences and context at select sites (Aim 2). Finally, factors contributing to implementation performance across hospitals with below- and above-average severe maternal morbidity rates were examined using a mixed methods approach (Aim 3).

The results from Aim 1 show that patients with hemorrhage – specifically targeted by this intervention – showed declining rates of severe maternal morbidity in both the pre- and post-implementation periods. Variation in severe maternal morbidity across hospitals was observed but not attributable to hospital factors, suggesting opportunities for tailoring of activities for further improvement in outcomes. In Aim 2, we found that implementation stakeholders were strongly supportive of quality improvement initiatives, yet cultural and motivational barriers prevented full integration of care practices, suggesting that strategies addressing attitudes and behaviors at the individual, team, and institutional level prior to implementation may improve effectiveness. An explanatory figure was developed to represent the relationships between themes and their influence on implementation: Altruism; Perceived impact despite limited data; Teamwork; Competing priorities; and Limited champions. In Aim 3 we found that despite variation in individual and system-level factors related to implementation, stakeholder experiences at hospitals with below- and above-average severe maternal morbidity rates were ultimately more similar than different. These findings suggest that systemic issues in medicine such as interprofessional culture or social or practice norms may be promising targets for improving performance.

Despite efforts to standardize obstetric care across Michigan hospitals with the obstetric hemorrhage bundle, variation in patient, provider, and hospital characteristics and context-specific cultural factors all influence implementation performance. Moreover, many of the factors identified as barriers to implementation are systemic to the medical field or hospital settings, presenting additional challenges. Taken together, variation in implementation performance and experiences across hospitals is related to the specific combination and respective strength of these factors. Bundle implementation is often understood as a straightforward process, but these results suggest that it may be quite the opposite.