Modern fifth generation (5G) massive machine type communications (mMTC) including both Narrowband Internet of Things (NB-IoT) and LTE-M technologies utilize multi-channel ALOHA mechanism at the random access channel (RACH). This heavily affects services deployed for event-based anomaly detection that collectively detect a certain event in a limited spatial area. The rationale is that a single physical event leads to activation of numerous end systems that start compete for RACH resources resulting high collision probability. In this paper, we propose and evaluation a simple mechanism utilizing the downlink channel to inform end devices (ED) about the type of detected event. This allows us to effectively reduce the rate of redundant RACH attempts. Our numerical analysis shows that an XGBoost classifier accurately detects congestion level utilizing the received waveforms, achieving an accuracy of \(0.98-0.99\) . By leveraging this functionality, we show that starting from 50 EDs in the coverage of the cell operating synchronously, the proposed congestion reduction algorithm increases successful preamble detection probability from around 0.1 to 0.8. Note that in practical conditions, these values may change depending on the amount of background load.