Orchestrating performance of healthcare networks subjected to the compound events of natural disasters and pandemic

The study addresses how compounding hazards—a natural disaster and an epidemic—simultaneously impact healthcare systems. While pandemics elevate demand and require measures that reduce transmission, natural disasters can damage infrastructure, disrupt services, displace populations into shelters, and force behaviors that increase contagion risk. The authors identify a major research gap: existing analyses typically consider natural disasters and epidemics separately and often assume static hospital capacity, offering no framework to quantify their combined effects or the dynamic spatiotemporal variability of capacity and demand. Motivated by the 2018 Camp Fire in Butte County, California occurring in the broader context of COVID-19, the study examines how event timing and mitigation strategies influence hospital performance, patient access, and outcomes when wildfire and epidemic stressors interact.

Prior work on healthcare systems under natural disasters has examined household risk, surge and patient demand models, fragility and resilience, and recovery, often focusing on single hospitals and hazards such as earthquakes and climate-related events. Studies on pandemics investigated system challenges, anticipated surges, staffing ratios, and demand management, frequently assuming constant hospital capacity. Collectively, this body of literature treats disasters and epidemics independently and lacks an analytical framework to capture their joint, dynamic impacts on capacity and demand, especially under conflicting mitigation needs (e.g., evacuation vs. distancing). The increasing frequency and intensity of wildfires under climate change further underscore the need to evaluate compounding hazard effects on healthcare networks.

Case study and scope: The authors analyze Butte County, California, using the 2018 Camp Fire scenario in Paradise and a counterfactual overlay of a COVID-19 epidemic in the same timeframe. The healthcare network includes acute care hospitals (e.g., Enloe Medical Center, Oroville Hospital, Orchard Hospital, and the Feather River Hospital prior to its evacuation/closure), as well as non-acute facilities (nursing homes, clinics, hospice, etc.). Data sources include wildfire perimeters and damage, hospital capacity and utilization (beds, ER, ICU, ventilators), staff residential locations, transportation networks, and COVID-19 case trajectories from multiple countries. Modeling framework: The study modifies a previously developed healthcare network model to dynamically capture hospital functionality (capacity and quality) under compound stressors. Hospitals are decomposed into service units: ER, inpatient, ICU, and ICU with mechanical ventilators. A patient-driven health-seeking and hospital-interaction model routes patients from census tracts to facilities based on proximity, capacity, and service availability, accounting for support lifelines (transportation, supplies, coordination), alternative care for non-critical cases, and mitigation actions that alter staff/space/supplies. Disease transmission: A modified SEIR model extends classical compartments to ten states: susceptible (S), exposed (E), infective (I), quarantined (Q), hospitalized (H), ICU/admitted (C), mechanical ventilator (V), recovered (R), deceased (D), and an active-positive tally. It models spread within and between census tracts and among evacuees and host zones, allowing time-varying reproduction numbers and mitigation effects. Wildfire evacuations are represented as temporarily reducing protection (increasing susceptibility), with shelter conditions elevating transmission. Hospitalization and service needs (ER, inpatient, ICU, ventilator) are derived from quarantine and severity probabilities using rates from CDC, ECDC, and Zhou et al., with uncertainty quantified via Monte Carlo simulation (e.g., 100,000 trials). Scenarios: The analysis varies (1) the relative timing between wildfire occurrence (tw) and epidemic progression/peak (te), examining cases where wildfire occurs before, during, or after the epidemic peak; and (2) disease spread profiles calibrated to multiple hotspots (Hubei/China, Iran, Italy, Spain, Germany, U.S.). The model jointly simulates wildfire-induced capacity disruptions (e.g., evacuation/closure of Feather River Hospital, staff displacement, smoke-related respiratory demand, transportation degradation) and epidemic-induced surges. Metrics include average waiting time, treatment time, days hospitals are overwhelmed, ratios of untreated patients by service line, and overall system functionality. Mitigation strategies: The study tests feasible measures used in prior events: augmenting ER throughput (backup beds, alternative staff, higher patients-per-bed-per-hour), expanding non-ICU physical beds for non-critical patients (non-acute units, unoccupied beds), early discharge policies, staff reallocation from Feather River Hospital to mitigate shortages, increasing shelters and lowering occupancy to enable distancing with enhanced testing and quarantine, and improving PPE to reduce healthcare worker infection. An optimization model then locates and sizes a temporary backup (field) hospital to minimize untreated patients and reduce waiting times, determining bed and ventilator needs under each spread scenario.

Wildfire-only impacts (Camp Fire): Model estimates indicate healthcare system functionality dropped by about 18%, average waiting time increased by 35%, and patient treatment time decreased by 9%. Closure of Feather River Hospital (17% of county staffed beds) permanently shifted demand to other facilities in the study timeframe. Inpatient demand increased approximately 14% at Enloe, 19% at Orchard Hospital, and 15% at Oroville. Epidemic-only impacts (COVID-19 counterfactual in Butte County): Patient waiting time would increase more than fivefold, treatment time would be minimized due to high demand, many ER-seeking patients would be sent home untreated, and hospitals would be overwhelmed for up to 135 days depending on the disease transmission scenario. Compound hazard impacts (wildfire plus epidemic): Evacuation and sheltering amplify transmission, especially when coinciding with the epidemic’s acceleration or peak. Quarantined (confirmed) cases increase substantially relative to the epidemic-only case—reported increases include about 35% and up to 59% depending on wildfire timing (before, during, or after the epidemic peak). Timing sensitivity is pronounced: - Wildfire 15 days before epidemic peak (Day 50): faster spread among evacuees/shelters; average waiting time roughly doubles; hospitals overwhelmed for about 245 days. - Wildfire at the epidemic peak (Day 80): waiting time increases by about 8.3-fold; hospitals overwhelmed for about 223 days. - Wildfire during decline: waiting time about five times higher than normal; hospitals overwhelmed for about 189 days. Service-specific overloads and untreated patients: When wildfire occurs during the epidemic acceleration phase, the ratio of patients leaving ER without being seen is maximized; even with enhanced ER capacity (e.g., increasing ER throughput to 3 patients per bed per hour and capacities at focal hospitals to ~600, 30, and 200 patients/day), worst-case ER demand can still be about double capacity, leading to thousands of untreated ER patients. For inpatient services, when wildfire occurs before the epidemic peak, single-day demand can be ~82% higher than capacity. Effectiveness of mitigation strategies: Increasing the number of shelters, reducing shelter occupancy to maintain distancing, and enhancing testing/quarantine for shelter residents can reduce service demand by more than 64%, shorten the number of overwhelmed hospital days by ~30%, reduce average waiting time by ~28%, and increase overall healthcare functionality by ~50%. Providing additional PPE to align healthcare worker infection rates with the general population slightly improves staffing (shortage −0.12%) and waiting time (−0.1%) but can marginally increase untreated patients (up to +1.2%) due to changes in bed availability dynamics. If staff transfers from Feather River Hospital are not implemented, average staff shortage is ~6%, overwhelmed days increase by ~27%, waiting time by ~26%, and total functionality decreases by ~26%; if ICU staffing gaps are not covered, untreated ICU patients increase by ~3.25-fold. Not leveraging non-acute care beds raises untreated patient counts and increases overwhelmed days by ~10%. Resource optimization: An optimally placed temporary field hospital in Oroville can substantially alleviate overloads; required staffed beds range from about 148 (Germany-like spread) to 392 (U.S.-like spread) across ER, inpatient, and ICU, with up to 19 mechanical ventilators.

The findings show that the interplay between wildfire-induced disruptions and epidemic transmission can severely degrade healthcare access, with outcomes highly sensitive to the relative timing of events. Evacuations and shelter crowding can amplify disease spread, while loss of hospital capacity and staff displacement intensify surges across ER, inpatient, ICU, and ventilator services. Modeling both hazards jointly, rather than adding their effects or considering epidemic pressures alone, is essential to capture spatiotemporal dynamics of demand, capacity, and patient routing. The results demonstrate that well-targeted mitigation—expanding shelters with distancing and testing, mobilizing staff, using non-acute beds for surge capacity, and strategically deploying a temporary field hospital—can markedly reduce waiting times and untreated patients and improve network functionality. These insights inform emergency planning by quantifying how policy levers and resource allocation decisions affect patient outcomes during compounding events.

The study introduces a framework to quantify and manage the compound impacts of a natural disaster and an epidemic on a regional healthcare network. Using Butte County during the Camp Fire with a concurrent COVID-19 epidemic as a testbed, the analysis shows that combined hazards can produce unprecedented strain, strongly modulated by event timing. Incorporating practical mitigation strategies and optimizing resource allocation—particularly shelter management, staff redistribution, use of non-acute beds, and temporary field hospitals—substantially enhances patient access and system performance. Future research should adapt and validate this framework across different communities and hazard types, incorporate evolving epidemic features (e.g., vaccination, variants, mobility restrictions), integrate richer real-time data streams to reduce epidemiological uncertainty, and extend optimization to multi-objective, multi-region deployments under resource constraints.

Generalizability may vary due to differences in community characteristics, hazard types, and sociocultural adherence to protective measures. The analysis assumes evacuees remain within Butte County without strict lockdown entry restrictions and that no vaccine is available during the study timeframe; relaxing these assumptions could alter outcomes. Disease transmission parameters were estimated from disparate international settings and are subject to uncertainty; available data were limited to the study period. Some hospital capacity and staffing adaptations (e.g., seamless staff transfers, effective use of non-acute beds, and immediate implementation of ER throughput enhancements) may be optimistic in real-world conditions.