Early Detection of Meningitis Outbreaks: Application of Limited-baseline Data

There is no published study evaluating the performance of cumulative sum (CUSUM) algorithm on meningitis data with limited baseline period. This study aimed to evaluate the CUSUM performance in timely detection of 707 semi-synthetic outbreak days.

Simulated outbreaks were generated using syndromic data on fever and neurological symptoms from Mar 2010 to Mar 2013 in Hamadan Province, the west of Iran. The performance of CUSUM algorithms, numbered from 1 to 11, in timely detection of outbreaks was measured using sensitivity, specificity, false alarm rate, likelihood ratios and area under the receiver operating characteristics (ROC) curve.

The highest amount of sensitivity was related to algorithm11 (CUSUM_{(3-9 D11)}) and it was 52% (95% CI: 49%, 56%). Minimum amount of false alarm rate was related to CUSUM_{(1-7 D5)} algorithm equal to 8% (95% CI: 5, 10) and the best amount of positive likelihood ratio was related to CUSUM_{(1-7 D4)} equal to 4.97. CUSUM_{(1-7 D1)} has the best performance with AUC curve equal to 73% (95 CI%: 70%, 76%), as well.

The used approach in this study can be the basis for applying CUSUM algorithm in conditions that there is no access to recorded baseline data about under surveillance diseases or health events.