A healthcare analytics company has developed a machine learning model to predict patient outcomes based on historical medical data. During testing, the model demonstrates high accuracy and performs well on the training dataset, but once deployed in a real-world production environment, its accuracy drops significantly when processing new, unseen patient records. The company needs to improve the model's ability to generalize and perform well on new data, ensuring reliable predictions in the production setting. What would be the most effective approach to fix this problem?