Why should I release my block time when Copient Health prompts me to?
What Do We Predict?
We predict utilization for block days in the near future. But more importantly, we use that prediction to decide whether or not to alert you to release time from a block day. We look ahead at upcoming block days and use a machine learning algorithm to predict how much of an upcoming block day you will fill up. If the algorithm predicts the block will not be fully utilized, you receive a message prompting you to release some OR time from that day.
How Does It Work?
We start with historical surgical case data (with no patient information) and historical block schedules. Our machine learning models are "trained" from that historical data. A trained model uses inputs of real-time booking and other data to predict utilization for an upcoming block day. If the algorithm predicts the block will not be fully utilized, you'll receive an alert prompting you to release time.
What's a Block Alert?
A Block Alert is a message you receive letting you know that an upcoming block will not be fully utilized. The Block Alert will contain a recommended amount of time for you to release. That is the time our algorithm is confident will not be filled. You can choose to release the recommended amount of time, or whatever amount of time you prefer, by clicking the prompt.
How Accurate Are The Block Alerts?
The short answer is that the Block Alert accuracy is chosen by your facility, usually a figure of 90% or higher. If you want to know more, keep reading.
By separating the historical data set into "training" and "test" data, we're able to use the training data to train the machine learning model, and the "test" data to measure its prediction accuracy. We look at the ratio of (correct block alerts/all block alerts) from the test data as a block alert accuracy measure. We can then use that measure to determine how many days in advance it's safe to alert you. This ensures that the accuracy of the Block Alert is above the facility-set limit. A practical example: If the Block Alert accuracy limit was set by your facility at 90%, that would mean that 9 out of 10 block days for which you received Block Alerts notifications would have been correct (i.e., that we predicted you wouldn’t fill your entire block, and you didn’t).