Mass Gathering Medicine: A Predictive Model for Patient Presentation and Transport Rates[1]

Summary

A 12-month survey of 201 Australian mass gatherings (combined audience >12 million) collected patient presentations and environmental factors under a standard reporting format. Across all event types the patient presentation rate was 0.99 per 1,000 attendees and the transport-to-hospital rate 0.027 per 1,000. Presentation rates fell slightly as crowd size grew, while relative humidity, crowd mobility, alcohol availability, and a bounded venue were associated with higher presentations. Three regression models were derived to predict presentation rates for future events.

So what

Establishes the baseline rates event planners still anchor to (~1 presentation/1,000; ~0.03 transports/1,000) and shows patient load is multifactorial — driven by weather, crowd behavior, alcohol, and venue boundary, not attendance alone. Predict with regression plus close attention to an event's own historical data.

Key findings

• Patient presentation rate 0.99/1,000; transport-to-hospital rate 0.027/1,000 (all event types).
• Presentation rate declined slightly as crowd size increased.
• Higher presentations with increased humidity, greater crowd mobility, alcohol availability, and bounded venues.
• Derived from 201 events / >12 million attendees; three predictive regression models produced.

Caveats & model limits

• Arbon (2001) — aggregate baseline (PPR/TTHR) Aggregate rates from Arbon 2001, NOT the regression model; use for order-of-magnitude context only.
• Arbon (2001) — aggregate baseline (PPR/TTHR) 201 events over 25,000 attendance, 12 months, Australia; 11,956 patients across all event types.
• Arbon (2001) — aggregate baseline (PPR/TTHR) At a recurring show the Arbon generic model correlated only r=0.12 with observed presentations (Zeitz 2005) and tended to overestimate - a history-based method did better there.
• Arbon (2001) — aggregate baseline (PPR/TTHR) Directional effects (crowd size, humidity, mobility, alcohol, boundedness, on-duty personnel) are documented but shown here as non-quantified flags, not applied to the number.
• Arbon (2001) — aggregate baseline (PPR/TTHR) Session 6: Option B baseline (BA). Aggregate rates verified from Arbon 2001 PDM. Regression coefficients NOT recovered - see arbon_2004 scaffold (still DRAFT). | Session 10: superseded by arbon_2001_regression; kept as humidity-free fallback.
• Arbon (2001) — full presentations regression Accounts for 64% of variance in presentations at 95% confidence (the SPORT term is weaker, p=0.0511).
• Arbon (2001) — full presentations regression Authors: unsuitable for events with historically very high or very low PPR; use WITH historical data where available.
• Arbon (2001) — full presentations regression PPR rises with temperature to the 25-29.9C band then DECLINES above 30C (behavioral adaptation); humidity is a more consistent predictor than temperature.
• Arbon (2001) — full presentations regression Requires relative humidity (%). Indoor/outdoor defaults to outdoor when unset (a large coefficient). Mixed crowd mobility maps to SEATS=0 (conservative).
• Arbon (2001) — full presentations regression At the index event both methods significantly under-predicted transports (actual 47 vs 27/22, p<0.05)
• Arbon (2001) — full presentations regression Session 10: Model A (Table 3) transcribed from the primary source via UK ILL (RapidILL 27181535); verified=1 on 19.25 validation-case reproduction. Supersedes arbon_2001_baseline. | model-card caveats appended per ZEITZ_MODEL_CARD_MEMO (BA delegation, trust-batch 2026-08-08)
• Arbon (2001) — transports from known presentations Requires a known/historical presentation count (TOTNUM). 34% of variance in transports at 95% confidence.
• Arbon (2001) — transports from known presentations Prospective use uses Model A -> Model B chaining with a stated limitation; Model C own coefficient is not published in the primary source.
• Arbon (2001) — transports from known presentations Session 10: Model B (Table 4) transcribed via UK ILL. Known-presentations transport pathway.
• ppr_bounded_focused Event-category rate reference; the regression models drive predictions.
• ppr_unbounded_extended Event-category rate reference.
• observed_staffing_rate OBSERVED practice, NOT a recommendation — never used to source a staffing line.
• staffing_skill_mix_observed Observed skill mix, not a recommendation.
• alcohol_concentration_finding Closest quantified alcohol finding in the literature.
• case_mix Case-mix reference (card caveat).
• temperature_caveat Card caveat: temperature is non-monotonic; do not extrapolate PPR linearly with heat.
• model_limits Card caveat on regression applicability.

Read the paper

• Publisher (DOI) ↗may be paywalled
• PubMed ↗abstract · free

Domains