Internet of Things (IoT) systems continuously generate digital representations of people, objects or physical phenomena when they can be made available over the Internet. IoT provides powerful opportunities for seamless data collection and transfer of information, which is useful for healthcare providers to temporally diagnose patients remotely by analyzing their data from IoT information systems when they are far away. In this study, a mayfly optimized deep Learning framework, IOT Healthcare Monitoring (OLOHM) has been proposed to mitigate the risk of having event memory potential health issues and help Emergency services operate more effectively. The proposed method provides the real-time monitoring of health parameters, including the level of oxygen, pulse rate, and blood pressure, through the wearable sensors. These continuous vital readings are sent to cloud databases via gateway devices via Bluetooth or Wi-Fi. These data streams are pre-processed, and features extracted, and the Mayfly Optimization algorithm was utilized to optimize the use of features to ensure that the most appropriate features are considered. The appropriate features are put into a Bidirectional Gated Recurrent Unit (BiGRU) model to predict health status. The outputs are then available in real-time for healthcare providers, families, hospitals, and emergency services to react in a medically timely manner. This system achieves accuracies of 99.25% and over simple Hierarchical Hidden (SHH), and Deep Neural Network (DNN), and Bagged Decision Stump Decision Tree classifiers, again confirming the use of 95.23%, 98% and 97.35% respectively