The Three Stages of Evacuation Behavior

The counter-intuitive finding that the relative risk of falling increased as the conflict progressed can be understood through a framework of three distinct behavioral stages, which has significant implications for how public health warnings should be issued over time:

• • Stage 1: Acute Shock and “Shelter-in-Place” (Days 1-3): Overwhelmed by unprecedented alert volumes, populations abandon regular movement. They remain inside or adjacent to shelters,Reference Dailey and Kaplan ^{7 ,} Reference Yantosik, Lerner and Maloney ⁸ minimizing physical transit and thereby drastically lowering the risk of falling per alert.

• • Stage 2: The “Complacency-Panic Mismatch” (Weeks 2-3): As alert frequency drops, civilians attempt to reclaim domestic routines (e.g., cooking, sleeping in regular bedrooms).Reference Agam, Godler and Calif ⁹ This creates cognitive complacency—environments are no longer primed for rapid escape (toys on floors, improper footwear). Yet, when a siren sounds, the biological “fight-or-flight” panic response is triggered at full force.Reference Milosevic, Milosevic and McCabe ^{10 ,} Reference Wise ¹¹ Executing a high-speed sprint through an unprepared environment causes the injury-to-alert ratio to surge.

• • Stage 3: Prolonged Routine and Fatigue: Cumulative stress, sleep deprivation, and eroded safety protocols compound physical clumsiness. As demonstrated by the extreme peaks in the injury-to-alert ratio during the final weeks of the operation, this prolonged phase maintains and even escalates high relative injury rates even as the external threat normalizes.Reference Artelli ^{12 –}Reference Zeidner and Hammer ¹⁴

The disproportionality of risk

The normalization of casualty data against the national population baseline ¹⁵ transforms the understanding of evacuation-related risk. The finding that adults over 80 are 4.5 times more likely to appear in ED records than their demographic weight suggests that the journey to shelter is not merely “difficult” for the elderly, but constitutes a high-risk event. This disproportionate toll validates the shift from generic alerts to specialized protection protocols for geriatric populations, as their baseline mobility constraints intersect fatally with the requirements of rapid evacuation.

Kin-Keeping: The Physical Toll of Preserving the Family

While physiological aging factors such as post-menopausal osteoporosis or sarcopenia certainly contribute to the higher incidence of fragility fractures and falls among older women,Reference Duque, Allen and Huang ^{16 -}Reference Testa, Callanan and Evans ¹⁸ they may not provide a complete explanation for the abrupt demographic shift observed from age 46 onwards. Instead, these biological vulnerabilities likely interact with the sociological burden of “kin-keeping,”Reference Ciciolla and Luthar ^{19 -}Reference Rosenthal ²¹ which emerges as a primary potential driver of indirect trauma in this cohort.

In familial structures, women frequently assume the role of primary caregivers. During the acute panic of an alarm, a kin-keeper’s immediate reaction is not solely self-preservation, but to gather and assist dependents.Reference Bateman and Edwards ²² For middle-aged women, this means carrying toddlers or assisting aging parents; for older women, it often involves physically supporting an ailing spouse.Reference McGhan, Loeb and Baney ^{23 ,} Reference Staight and Harvey ²⁴ This caregiving instinct severely compromises biomechanical stability. Running while bearing the weight of another human removes the ability to use arms to break a fall and divides cognitive attention.

This hypothesis is tragically supported by pediatric data. While older boys show higher injury rates due to impulsivity,Reference Chapple and Johnson ^{25 -}Reference Shulman, Harden and Chein ²⁷ the high concentration of injuries among infants and toddlers aged 0-4 (where the gender distribution is nearly equal) strongly suggests a mechanism of “passive trauma.” It is highly probable that these infants were not injured by their own impulsive running, but rather suffered trauma when their burdened, frantic caregivers fell while carrying them. This transformed their attempts at protection into a hazardous endeavor.

Mobility Constraints vs. Acute Flight Response

The stratification of injuries by warning time suggests two distinct mechanisms of morbidity. In 90-second zones, the prevalence of elderly casualties, who comprise 41.6% of all injuries in these areas, indicates that falls are primarily driven by baseline mobility constraints. While the longer warning window theoretically allows for a safer pace, the physical demands of reaching a protected space, often navigating obstacles or stairs in older urban environments, still pose a significant risk to the frailest individuals.

Conversely, the increased representation of younger, mobile adults (ages 26-65) in 0-15 and 30-second zones (accounting for 49.4% and 53.3% of casualties in these areas, respectively) illustrates the impact of an Acute Flight Response (or Panic-Induced Flight). When the warning window is extremely narrow, the physiological “fight-or-flight” response may trigger a frantic, high-speed evacuation. In these high-pressure scenarios, the urgency to reach safety severely compromises biomechanical stability, even in healthy individuals. This risk is likely exacerbated by the “kin-keeping” burden; the consistent injury rates of infants (0-4 years) across all timeframes suggest that regardless of the time available, a child’s safety remains dependent on the caregiver’s stability, which is most vulnerable during an acute, rapid rush.

Simultaneously, the relative under-representation of the elderly in these acute warning zones (0-15 seconds) suggests a potential self-regulatory behavioral pattern. Recognizing that their mobility constraints make reaching a designated shelter within seconds impossible, many older adults may opt for a “shelter-in-place” strategy, remaining in their immediate location rather than attempting a high-risk sprint. This decision, while increasing potential exposure to direct threats, effectively mitigates the immediate biomechanical risk of fall-related trauma.

Global Public Health Implications

The findings from this study extend far beyond the geopolitical borders of Israel and armed conflict. As global early warning systems become more sophisticated,Reference Calvo-Solano and Quesada-Román ^{28 ,} Reference Debnath and Dhanekula ²⁹ the “warning window” is universally shrinking. This compressed timeframe is highly relevant globally:

• • Natural Disasters: Systems like the US ShakeAlert for earthquakes,Reference Lux, Smith and Böse ³⁰ tornado sirens in the American Midwest,Reference Jackson, Cutter, Gall and Rubin ³¹ and tsunami warning systems in the Pacific RimReference Mori, Satake and Cox ³² all rely on the public taking immediate physical action within seconds to minutes.

• • The Universality of the “Mismatch”: The “Complacency-Panic Mismatch” is a universal human response to prolonged, intermittent threats.Reference Aguirre ³³ Public health agencies managing prolonged disaster warnings (e.g., hurricane seasons, wildfire watch periods) must recognize that the physical risk of evacuation does not remain static but can exponentially multiply as populations attempt to normalize their daily lives amidst the ongoing threat.

• • Redefining Vulnerability: Global disaster preparedness protocols often categorize the elderly and children as vulnerable because they cannot run fast enough.Reference Onyejesi, Alsabri and Del Castillo Miranda ³⁴ This study provides compelling evidence suggesting that the caregivers themselves—particularly middle-aged and older females—may become highly vulnerable specifically because they are assisting others. International evacuation guidelines must shift to address the biomechanics of caregiving under duress.

• • Urban and Architectural Constraints: As demonstrated by the high injury rates in affluent, densely populated municipalities, structural vulnerability is not solely a function of socioeconomic deprivation. Global urbanization and high-rise living mean that longer warning times (e.g., 90 seconds) can become hazardous “traps” if the evacuation route involves navigating multiple flights of stairs. Global protocols must tailor physical evacuation guidelines to the specific built environment of the population.

Preventive Strategies and Interventions

The leadership of the Ministry of Health has placed a central emphasis on the planning of emergency policy that underscores the primacy of addressing the needs of vulnerable populations, particularly older adults. Within this policy framework, a series of tactical measures were undertaken, including the preparation of health maintenance organizations and the provision of guidance regarding structured, controlled discharge processes. These processes were accompanied by targeted provisions ensuring access to medications, mobility and accessibility aids, guidance for families, and transfers to protected facilities for continued inpatient care. Hospitals were instructed to establish a “discharge staff”—comprising social workers, allied health care professionals, nurses, and doctors—to engage in the process of clinical and functional assessment.

In addition, “pop-up” rehabilitation units were established to enable the continuity of rehabilitation processes in protected structures, aiming to prevent physical and functional deterioration while safeguarding elderly patients and maintaining continuity of care.

A further critical component of this systemic effort was addressing the out-of-hospital risk environment. In response to the escalating burden of these indirect casualties, the Israel Ministry of Health launched a targeted, multi-lingual public awareness campaign on March 8, 2026. Disseminated across mass media platforms, the campaign emphasized simple, practical, and actionable measures for the home environment. Providing tailored guidance on fall prevention en route to protected spaces, the initiative sought to mitigate the acute stress and rapid rushing that primarily precipitate these falls.

Key public health guidelines included preemptively clearing physical obstacles from evacuation routes, ensuring adequate lighting, utilizing closed, non-slip footwear, and keeping mobility and sensory aids readily accessible. Furthermore, the messaging explicitly advised against running, advocating instead for a swift but measured pace, while urging the public to actively monitor and assist vulnerable neighbors during alerts.

More generally, effective public health strategies must move beyond the basic instruction to “seek shelter immediately” and actively incorporate behavioral guardrails and structural modifications:

• • Dynamic Messaging: Messaging must combat the “complacency-panic mismatch” by evolving throughout the conflict. Given that the highest injury-to-alert ratios were recorded in the final weeks of the operation (peaking at 88.8 on April 5), sustained and repetitive campaigns are essential to remind populations that domestic safety protocols must not be neglected as routines normalize.

• • Addressing the Caregiver Burden: Guidelines must explicitly address the biomechanics of kin-keeping. Rather than frantic running while carrying a child, campaigns should advocate for a “swift but measured pace” and pre-planned household roles to distribute the burden. Crucially, in zones with immediate warning times (0-15 seconds), the emphasis should shift from “reaching the primary shelter” to “seeking the nearest protected space” to minimize the high-velocity sprint. For older adults and individuals with significant mobility impairments in these areas, public health guidelines should explicitly prioritize immediate protective actions within the current room (e.g., lying on the floor below window level) over the attempt to reach a distant shelter. In such extreme time-compressed scenarios, the clinical risk of a catastrophic fall may outweigh the marginal protective benefit of a rapid evacuation attempt.

• • Household Environmental Modifications: Checklist-based programs should target home “hazard-free zones,” securing rugs and ensuring accessible mobility aids. In high-SES and densely populated urban areas, specific focus must be placed on stair safety and the risks associated with internal stairs in multi-level homes, accounting for the 51.4% of casualties residing in these affluent clusters.

• • Addressing Socioeconomic and Architectural Disparities: Our findings revealed a counterintuitive trend: the majority of casualties (51.4%, n = 1,485) resided in high-SES municipalities (clusters 7-10). This distribution is likely influenced by an “evacuation bias.” During the operation, populations from high-risk border zones—which predominantly consist of lower-SES municipalities (clusters 1-4)—underwent mass evacuation. Consequently, the relative proportion of middle- and high-SES populations remaining in the active warning zones artificially increased, naturally skewing the national casualty demographics.

  However, even accounting for this demographic shift, structural vulnerabilities and variations in health-seeking behaviors play a crucial role. High-SES populations often reside in multi-level private homes or large apartments with internal stairs, which pose a unique falling hazard during a rapid rush to a shelter. Furthermore, higher-SES individuals may be more inclined to visit an ED for minor injuries. Consequently, preventive interventions must be bifurcated: high-SES areas require targeted campaigns on stair safety and internal “luxury” hazards, while lower-SES neighborhoods require continued focus on public shelter accessibility and infrastructure safety where reporting might be lower but environmental risks remain high.

• • Tailored Architectural Interventions: Our findings suggest that structural vulnerability is influenced by the built environment. In municipalities with longer warning times (e.g., Bat Yam, Or Yehuda), where elderly populations face the challenge of navigating multi-story buildings, policy should prioritize the accessibility of floor-level shelters and the improvement of communal stairwell lighting.

Limitations

This study has several limitations. First, it relies on retrospective ED administrative data, which inherently captures only injuries severe enough to warrant hospital admission or emergency care. Consequently, our reported incidence rates likely underestimate the true community-level burden of evacuation-related falls, as minor injuries (e.g., mild sprains, contusions) managed at community clinics or self-treated at home are excluded. This reliance on ED data introduces potential selection biases that may impact the reported demographic distributions. For instance, older adults are clinically more likely to be referred to the ED following a fall due to the higher risk of fractures, whereas younger adults might self-manage similar falls, potentially exaggerating the observed age disparities. Furthermore, differences in health care-seeking behaviors among various socioeconomic groups may skew the data, as higher-SES populations might be more inclined or have better logistical access to utilize ED services for non-life-threatening injuries. Second, the database lacks granular narrative details regarding the exact micro-mechanism of the fall (e.g., tripping over a rug vs. slipping on stairs) and lacks direct confirmation of the caregiver status at the exact moment of injury, requiring the use of demographic proxies to infer “kin-keeping” behaviors.

Third, regarding exposure metrics, the use of aggregated daily national alert counts as a denominator for the injury-to-alert ratio (Figures 1 and 2) does not account for the geographic distribution of sirens. Regional surges in alert volume in specific high-risk zones may not be fully captured in the national ratio, potentially masking localized risk dynamics. Furthermore, while this study attempted to stratify casualties by municipal warning time categories (15-90 seconds), this analysis was limited by the use of municipal residence as a proxy for the warning window at the time of injury. Crucially, this official polygon-based categorization relies solely on the formal Home Front Command siren activation time. It does not account for the widespread phenomenon of “advance warnings” (e.g., third-party mobile applications, broadcast media alerts, or the auditory crossover of sirens from adjacent municipalities). These unofficial early alerts often provide civilians with additional, unrecorded seconds prior to the localized siren, meaning the actual time a casualty had to react and seek shelter may have differed from their official municipal designation. However, it should be noted that data completeness in the expanded cohort was exceptionally high, with specific warning-time data available for 99.3% of the sample (n = 2,919).

Additionally, the displacement and evacuation of civilian populations from high-risk border zones likely altered the demographic profile of casualties in the 0-15 and 30-second zones. Because these evacuated border communities are disproportionately represented in lower socioeconomic clusters (1-4), this mass displacement created an inherent “evacuation bias.” This bias not only leads to an underestimation of the risk to elderly populations in these areas, as many vulnerable residents were evacuated early in the conflict, but it also limits our ability to isolate the true effect of SES on fall risk, as the remaining population exposed to sirens nationwide was structurally skewed toward higher-SES demographics.

Finally, the spatial analysis normalized absolute cases by municipal population size, which can introduce statistical noise in smaller communities. Nevertheless, the consistency of findings across broad socioeconomic clusters (n = 2,887) suggests that the overarching trends remain robust across the national population.

Future prospective research should incorporate qualitative interviews and granular, GPS-linked temporal data to further validate the biomechanical and behavioral hypotheses presented here.