Early warning systems for malaria outbreaks in Thailand: an anomaly detection approach

Malaria Journal

Table 3 Results from method comparison

Method	Ubon (2014)³	Ubon (2015)³	Yala (2016)	SSK (2017)³	KCN (2017)³	KCN (2022)	Tak (2022)	Verified Anomalies detected (%)	Total Reported⁴
Statistical profiling	✘⁵	✘	✔⁶	✘	✘	✘	✘	1 (14%)	882
Predictive confidence interval	✔	✘	✔	✔	✘	✘	✘	3 (43%)	2356
Unsupervised clustering	✘	✘	✘	✘	✘	✘	✘	0 (0%)	75
Density-based profiling	✘	✘	✘	✘	✘	✘	✘	0 (0%)	5
Density-based profiling w/T&P¹	✘	✘	✘	✘	✘	✘	✘	0 (0%)	452
Historical average	✘	✔	✔	✔	✔	✔	✔	6 (86%)	10875
Weekly Case previous year	✔	✘	✔	✔	✔	✔	✔	6 (86%)	30449
Monthly case 4 years	✘	✘	✘	✘	✘	✘	✘	0 (0%)	5577
Weekly 3 year median²	✔	✔	✔	✔	✘	✔	✔	6 (86%)	32630

¹Temperature and precipitation included in analysis
²Baseline method used in Thailand (BIOPHICS)
³Ubon: Ubon Ratchathani, SSK: SI Sa Ket, KCN: Kanchanaburi
⁴Total anomalies repored for each method when applied to all malaria cases between 2012 and 2022
⁵Symbol showing that this method was not able to trigger anomaly alerts at least 14 days before this verified outbreak observation
⁶Symbol showing that this method was able to trigger anomaly alerts at least 14 days before this verified outbreak observation

ISSN: 1475-2875