EvacKalma: Evacuation center crowding balancer and camp disease outbreak detector

Problem Statement

When a typhoon, flood, or volcanic eruption forces a municipality to evacuate, families naturally walk to the single most familiar school or covered court. As a result, one elementary school ends up packed with 700 evacuees sharing two overflowing toilets—violating the Ligtas Pinoy Centers Act (RA 12076) and Sphere humanitarian standards—while another designated evacuation facility 600 meters away sits at 20 percent capacity. Within 48 to 72 hours of severe crowding and poor sanitation, secondary disasters strike inside the camp: outbreaks of acute watery diarrhea, measles, conjunctivitis, and acute respiratory infections among infants and seniors. Barangay Health Workers record camp clinic visits on paper notebooks that only reach the Municipal Health Officer days later, after an outbreak has already spread across classrooms.

Proposed Solution

EvacKalma is a two-phase shelter balancing and syndromic outbreak warning system for local government units and barangay health teams:

1. Pre-Evacuation Smart Routing: As preemptive evacuation begins, EvacKalma tracks real-time headcount, toilet-to-person ratios, potable water supply, lactation spaces, and backup generator status across all barangay and municipal shelters. Using constrained optimization, it assigns and texts each purok or special-needs household (pregnant women, insulin/dialysis patients, seniors, families with pets) to the nearest safe shelter that still has sanitation and medical capacity.
2. In-Camp SPEED Outbreak Surveillance: Once shelters are occupied, Barangay Health Workers submit a 30-second daily tally via SMS or offline web form using the 21 syndromic categories of the DOH-WHO Surveillance in Post Extreme Emergencies and Disasters (SPEED) system (such as loose watery stools, fever with rash, cough/difficulty breathing, high blood pressure). EvacKalma runs a Poisson and Cumulative Sum (CUSUM) anomaly detector across rooms and centers, alerting the Municipal Health Officer the same hour a symptom cluster crosses the epidemic threshold so isolation rooms, ORS packets, and clean water tankers are dispatched immediately.

Target Market

• Primary users: Barangay Health Workers, Camp Managers (CSWDO/MSWDO social workers), Municipal Health Officers, and evacuating families.
• Public-sector buyers: City and Municipal DRRM Offices, Municipal Social Welfare and Development Offices (MSWDO), and Provincial Health Offices implementing RA 12076 (Ligtas Pinoy Centers Act).
• Humanitarian partners: Philippine Red Cross, DSWD Disaster Response Management Bureau (DRMB), and UNICEF/WHO WASH clusters.

How AI Is Used

• Multi-objective shelter capacity optimization: Balances walking distance, flood-free road access, Sphere toilet ratios (maximum 20 persons per toilet), and medical vulnerability matching using integer linear programming.
• Syndromic outbreak anomaly detection: Applies spatio-temporal Poisson/EARS (Early Aberration Reporting System) and ML classification to small-sample daily symptom counts to distinguish normal baseline illness from an emerging camp outbreak of cholera/rotavirus or measles.
• Taglish SMS camp intake parser: Extracts structured family headcounts, infant formula needs, maintenance medicine shortages, and syndromic tallies from free-text SMS sent by busy camp volunteers.
• Automated DSWD DROMIC and DOH SPEED report generator: Compiles age- and sex-disaggregated camp situation reports and WASH supply requisitions in seconds.

Community & Ecosystem Integration

• Built around Philippine DOH SPEED and DSWD DROMIC standards: Uses the exact syndromic definitions and camp registry fields Philippine social workers and nurses already train on.
• Protects vulnerable evacuees: Ensures pregnant mothers, infants, PWDs, and seniors are routed to rooms with accessible toilets and power rather than sleeping on wet gym floors.
• Works over basic SMS: Camp volunteers can report headcounts and symptom tallies without mobile data.

Demo Scenario (3 minutes for judges)

1. Show a municipality with 5 evacuation centers during a typhoon. Center A (San Roque Elementary) hits 92 percent occupancy and 45 persons per toilet, turning red on the map.
2. EvacKalma automatically updates the outbound SMS routing for Purok 5 and Purok 6, directing incoming families and two dialysis patients to Center B (Municipal Gym, 35 percent full, generator active, 400 meters away).
3. Switch to Day 2 of camp operations: a Barangay Health Worker texts "SPEED CenterA: 6 watery diarrhea, 4 lagnat sa Room 3, ubos na drinking water".
4. The SPEED anomaly detector triggers a Red Outbreak Alert for Acute Watery Diarrhea in Center A (crossing the 2-sigma threshold), auto-drafts an urgent water-chlorination and ORS dispatch order for the Municipal Health Officer, and flags Room 3 for isolation.

30-Hour Build Plan

Suggested stack: Python FastAPI, Google OR-Tools, statsmodels / scikit-learn (Poisson/CUSUM aberration detection), Gemini API for SMS parsing, PostgreSQL/PostGIS, Next.js + Leaflet, SMS gateway. Public data / APIs: DOH-WHO SPEED (Surveillance in Post Extreme Emergencies and Disasters) Operations Manual, DSWD DROMIC public camp situation reports, OpenStreetMap school and covered-court locations, Sphere Handbook WASH standards.

Research & Literature References

1. Volker Winkler, Gerard Krause, et al., “Post-disaster health impact of natural hazards in the Philippines in 2013,” Global Health Action, Vol. 9, Issue 1, 31320, 2016. DOI: 10.3402/gha.v9.31320
   • How EvacKalma uses this study: Uses the empirical disease-onset timelines and syndrome distributions recorded across Philippine evacuation centers (acute respiratory infections, acute watery diarrhea, fever, and hypertension spikes) to calibrate the camp outbreak warning thresholds.
2. S. Tante et al., “Which surveillance systems were operational after Typhoon Haiyan?,” Western Pacific Surveillance and Response Journal, Vol. 6, Suppl. 1, pp. 66-70, 2015. DOI: 10.5365/wpsar.2015.6.2.HYN_012
   • How EvacKalma uses this study: Addresses the paper’s key operational finding that SMS-based SPEED syndromic reporting is the fastest surveillance channel after a super-typhoon when internet and paper courier networks fail.

Risks & Privacy Notes

• Aggregate syndromic counts only: Daily SPEED reports transmit only aggregate symptom counts per room or shelter (for example, “4 cases of fever”), never broadcasting individual patient names over unencrypted SMS.
• Human clinical verification: An automated outbreak flag prompts an immediate inspection by the Municipal Health Officer or public health nurse rather than diagnosing individual patients.
• Family unity rule: The shelter router never splits members of the same household across different evacuation centers.

AI Use Disclosure (draft)

EvacKalma uses linear programming (OR-Tools) for shelter load balancing, statistical aberration detection (CUSUM/Poisson) for syndromic outbreak alerts, and an LLM to parse unstructured Taglish SMS reports from camp volunteers. AI coding tools helped scaffold the dashboard UI and synthetic camp simulation data. All health surveillance thresholds follow public DOH SPEED guidelines.

Rubric Scoring