The use of machine learning for predictive modeling in healthcare provides great opportunities for knowledge discovery and decision support for clinicians. However, unequal performance of models across demographic groups can lead to disparities in health outcomes, potentially causing harm. Many approaches have been proposed to reduce the unfairness of machine learning models, and for fair model development, but these approaches may not consider or leverage the underlying differences between groups. The objective of this study was to utilize a collaborative learning approach to combine knowledge from models trained on the individual demographic groups. In the case of race groups, where the data was very imbalanced (10% Non Caucasian), collaborative learning effectively improved model fairness from the baseline with a drop in equalized odds (0.091 to 0.049) and a small drop in performance (AUC 0.846 to 0.823). Using sex groups for collaborative learning, where the distribution was not as imbalanced (35% female), did not lead to significant changes, suggesting that the collaborative approach has potential for developing a fair model in cases where the sensitive attribute distribution is heavily imbalanced.