Training method for giving treatment recommendations to a physician based on a propensity score and an outcome score

A computer-implemented method for providing treatment recommendation data, comprising:

• receiving actual patient data, which relates to a health condition of an actual patient;
• applying a trained function by a machine learning algorithm to the actual patient data, wherein the treatment suggestion data is generated by use of a treatment recommendation score,
• wherein at least one parameter of the trained function is based on a propensity score, which represents a relative frequency for a specific treatment to be chosen by a physician for specific health conditions of a former patient, and/or at least one parameter of the trained function is based on an outcome score, which is a measure of how health data and/or a health condition of the former patient changed while applying the specific treatment,
• providing the treatment recommendation data.

A providing system, comprising

• a first interface, configured for receiving actual patient data which relates to a health condition of an actual patient;
• a second interface, configured for providing the treatment recommendation data,
• a computation unit, configured for applying a trained function to the actual patient data, wherein the treatment recommendation data is generated,
• wherein at least one parameter of the trained function is based on a propensity score, which represents a relative frequency for a specific treatment to be chosen by a physician for specific health conditions of a former patient, and/or at least one parameter of the trained function is based on an outcome score, which is a measure of how health data and/or a health condition of the former patient changed while applying the specific treatment.

A training system, comprising

• a first training interface, configured for receiving a plurality of training data sets each relating to a former patient with each training data set comprising treatment data, which represents a former treatment chosen by a physician for the respective former patient, and patient data, which relates to a respective health condition of the respective former patient at different points in time,
• a training computation unit, configured for training a function based on the based on the training data sets,
• wherein the training computation unit is configured for determining a propensity score for each of the former treatments, wherein the propensity score represents a relative frequency for the respective treatment to be chosen for each of the health conditions,
• wherein the training computation unit is also configured for determining an outcome score for each of the former treatments, wherein the outcome score is a measure of how the health data and/or the health condition of the respective patient changed between the different points in time;
• and
• and wherein the training computation unit is also configured for concatenating the propensity score and the outcome score to a combined score,
• a second training interface, configured for providing the trained function.