TL;DR
This paper introduces a novel framework to quantify the resilience and adaptability of US healthcare systems to successive disruptions, using millions of patient records from the COVID-19 pandemic. The study reveals that while US healthcare systems show substantial adaptability, their overall resilience is moderate, with significant disparities observed across racial groups and essential services.
What Problem Does This Paper Address?
The COVID-19 pandemic exposed the vulnerability of healthcare systems to large-scale, prolonged disruptions. While "resilience" (the ability to absorb, recover, and adapt) is a widely discussed concept, there's a significant lack of empirical evidence and quantitative tools to measure a system's adaptability – its capacity to learn from and adjust to successive shocks. Existing research often focuses on single disruptions or the static aspects of recovery, leaving a gap in understanding how systems perform and evolve through multiple, overlapping crises. This paper aims to fill that gap by developing and applying a framework to quantify both resilience and adaptability in the context of the COVID-19 pandemic's dual disruption waves.
Methods
The authors developed a novel quantification framework to assess healthcare system resilience and adaptability. This framework builds upon existing models for single disruptions and extends them to handle multiple successive disruptions. Key components include:
- Disruption Characterization: Each disruption is characterized by its amplitude (severity), duration, disruption rate, and recovery rate.
- Adaptability Index ($\rho$): This index is calculated based on the change in disruption rates between consecutive disruptions. A positive $\rho$ indicates improved adaptability (lower disruption rate in the second wave), while a negative $\rho$ suggests a lack of adaptability.
- Resilience Index ($r$): This index measures the total loss of patient visits relative to expected performance, normalized to a range of 0 to 1, where 1 signifies perfect resilience.
- Predictive Modeling: A model was developed to estimate expected patient visits in the absence of disruptions, using factors like physician EMR adoption, state population, and physician-to-population ratios.
- Statistical Analysis: Pearson correlation coefficients were used to examine the relationships between resilience/adaptability and factors like pandemic severity, physician shortages, and socioeconomic variables.
Datasets
The primary dataset used in this study is the Healthjump EMR dataset, which aggregates electronic medical records and practice management system data from over 70,000 hospitals and clinics across the US, covering over 33 million patients and 1 million healthcare providers, with records dating back to 1995.
Additional data sources include:
- Johns Hopkins Coronavirus Resource Center: For cumulative COVID-19 infection cases by state.
- 2019 State Physician Workforce Data: For physician abundance data by state.
- CDC/ATSDR Social Vulnerability Index (SVI): For socioeconomic factors like poverty, unemployment, uninsured rates, and demographic proportions.
- National Center for Health Statistics and US Census Bureau: For validation data on patient visits, emergency department visits, and hospital discharges.
Key Insights and Ideas
- Two Waves, Not One: The COVID-19 pandemic, contrary to some prior assumptions, presented healthcare systems with two distinct, successive waves of disruption, offering a natural experiment for studying adaptive responses.
- Adaptability Outshines Resilience: US healthcare systems demonstrated substantial adaptability ($\rho = 0.58$), meaning they learned from the first disruption to better manage the second. However, overall resilience ($r = 0.70$) was only moderate, indicating significant losses in patient visits still occurred.
- Chronic Care is More Robust: Services like chronic disease care showed higher resilience and adaptability compared to maternal care, likely due to their long-term nature and flexible delivery models.
- Racial Disparities are Stark: Black and Hispanic populations experienced significantly more severe disruptions and lower resilience compared to White and Asian groups, highlighting deep-seated socioeconomic inequalities.
- Physician Abundance is Key: The number of physicians in a region emerged as a critical factor in determining healthcare system resilience and adaptability, with states having fewer physicians being more vulnerable.
Key Results
- Overall Adaptability: US healthcare systems achieved an average adaptability index of $\rho = 0.58 \pm 0.05$.
- Overall Resilience: The average resilience index was $r = 0.70 \pm 0.03$.
- Service-Specific Resilience: Chronic disease care had a resilience of $r = 0.76 \pm 0.03$, while maternal care had $r = 0.74 \pm 0.03$.
- Racial Group Resilience: Asian populations had the highest resilience ($r_{Asian} = 0.80 \pm 0.02$), followed by White ($r_{White} = 0.74 \pm 0.03$), Black ($r_{Black} = 0.73 \pm 0.03$), and Hispanic ($r_{Hispanic} = 0.72 \pm 0.03$) groups.
- Socioeconomic Factors: Resilience was negatively correlated with poverty and unemployment rates, and positively correlated with physician abundance.
Limitations and Future Work
The authors acknowledge several limitations:
- Data Bias: The EMR dataset may overrepresent states with high EMR adoption, potentially biasing results for regions with lower adoption. Missing attributes in the dataset could also introduce bias.
- Retrospective Analysis: The current measures are retrospective. Future work should aim for real-time estimates for proactive decision-making.
- Scope: The study focused on specific essential services and did not include direct COVID-19 patient care. Future research could broaden the scope to include other crises (e.g., climate change-related diseases) and explore higher-resolution guidance for individual system components.
Future research directions include:
- Developing a more comprehensive quantification of resilience, potentially including "excessive recovery" as a dimension.
- Investigating the impact of intertwined disruptions from different crises.
- Conducting comparative analyses with past outbreaks like SARS.
- Delving into sophisticated computational methods to incorporate intricate connections between healthcare facilities, resource management, workforce, and supply chains.
Summary updated