Skip to main content

Hitachi
Research & Development
  • Profile photo of researcher : Tatsuya Nakae

    Tatsuya Nakae

    Social Innovation Co-creation Center
    Societal Intelligence Co-creation Research Department
    Unit Manager


Screen display of Emergency Dispatch Demand Forecast System. Optimal teams placement is planned based on the prediction results.

Text by Reiko Imamura

Emergency calls to peak in 10 years

The number of emergency services dispatched from 119 emergency calls is increasing rapidly. The Fire and Disaster Management Agency counted 6.6 million cases in 2018, an 80 percent increase over the last 20 years. This may be attributable to the rising number of older adults living alone and summertime cases of heatstroke, but if left unaddressed, the trend is expected to peak in 10 years, making for significant delays in emergency response times. With time being of the essence when a patient is experiencing cardiac arrest, delayed response times may very well reduce survival rates.

In Japan, fire departments have taken various measures to reduce emergency response times. When statistical data and response team's experience suggest a higher rate of dispatches, the department may call in off-duty personnel or move the teams to standby according to the time of day. Alternatively, protocols currently under review may be introduced for the emergency dispatchers to judge whether a physician should accompany the response team at the time of the emergency call. Human experience and intuition, however, are limited. It can be better to address rising case numbers by expanding to a cross-jurisdictional target area and consider the agile placement of dispatch teams.

Knowing this need for quicker and more reliable decision-making, one fire department interviewed us about the possibility of forecasting emergency dispatch demand using AI.

Using in-house technology to show the basis of a forecast

Hitachi has already delivered to fire departments an emergency call command and control system, to which we also provide operation and maintenance. “Clients began analyzing and exploring the idea of using AI for forecasting demand four years ago. We decided to examine social issues and customer needs in house,” says Tatsuya Nakae of the Global Center for Social Innovation, Tokyo. After developing financial and healthcare systems, Nakae began working on fire departments' emergency response systems and coordinating response teams and hospitals in 2017. This time, Nakae is working as a researcher alongside the team tasked with developing the command and control system.

The newly developed, AI-based Emergency Dispatch Demand Forecast System divides the target area into meshes of one square kilometer, and for each mesh, it makes two predictions: the expected number of emergency calls and expected response times. The system uses 20–30 data values to make predictions, including population distribution, meteorological information, and hospital information. Large amounts of these data are fed to the AI so it may learn trends. The AI is then made to weigh temperature and precipitation, the day of the week or time of day, and other values to predict how many dispatches will be called for in that mesh. The system can also predict optimal dispatch team placement and how changes to placement will alter dispatch team response times. It is purported that a single mesh reveals results instantaneously.

One defining detail is that the system shows the basis for the results alongside its predictions. “The predictive accuracy of AI has become high,” says Nakae, “but people may feel uneasy if they don't know how it got to its conclusions, and they could stop using it. That's why our original technology displays what data the AI focused on and how much.” For example, suppose that the number of emergency dispatches around a certain train station is quite low and a dispatch team is placed nearby, yet the AI predicts a slow response time. Knowing the basis for this prediction—the hospital is far away, which extends the dispatch team's time in transit—gives decision-makers the confidence to give directives, like instructions for teams to relocate.

In some areas, the system is already in operation. “The AI recommends placement of dispatch teams with the shortest expected response times within 100-square kilometers,” says Nakae. “These recommendations inform dispatch team placements.”

There are few other examples of AI-based emergency dispatch demand forecasting being conducted in the field, and Hitachi is at the head of rapid implementation. In addition to further improving the system's accuracy, we aim to integrate not only fire departments but also hospitals and other institutions into the system in order to promote our overall optimization.

  • Page top