Abstract:
This study employs machine learning using electronic general practice records to enhance patient safety in New Zealand. Objectives include identifying healthcare harm, harm severity and preventability. This research uses existing
data and knowledge from an extensive retrospective review of general practice records. Three years of records were obtained for 9076 patients who attended 44
general practices. All records were de-identified and reviewed by eight general practitioners for evidence of patient harm. 1505 patients experienced 2972
harms over the three-year period. Although most harms were minor, around one in four harms were moderate or severe, with 45 patients hospitalised and 11
patients dying. This review identified preventable patient safety issues but required extensive time, expert knowledge and resources for collecting, screening and analysing data. Given the availability of this data, a great opportunity exists for developing a machine learning-based system to rapidly
screen records according to consistent criteria for patient harm, severity and preventability. There are many examples of health care applications where the use of machine learning techniques has been proven to improve patient care and decision making. Artificial Intelligence working in conjunction with experts moves us further towards reducing healthcare harm in New Zealand general practice.
