Clinical encounters require the creation of an enormous amount of documentation. This documentation is tedious, time-consuming, and, in practice, is usually created hours after the encounter has occurred. The requirement to create this documentation places a tremendous burden on the time of clinical staff, and due to pressing workloads, can lead to incorrect recall and clinical data errors. The project intends to prove that progress note portion of this documentation process can be automated by novel application of recent advances in machine learning. A software system will be built with Support Vector Machine theory at its core. The software system can learn from existing clinical progress notes, and then apply those learned method by auto-generating the subjective/analytical portions of the note. The project will also examine learnings from individual physician notes compared to collections of multiple physician notes in order to built a superior model. The project's impact on the quality, and cost, of care should be dramatic. Reductions in documentation error rates and increases in physician productivity equate to an incredible array of quantitative benefits. Qualitatively, the project should make providing healthcare a more enjoyable experience for all involved, given the reduction of administrative time on the part of the highly skilled clinical staff.
Clinical encounter documentation provides the detailed patient health data which enables all care providers to have an accurate picture of the patient's clinical activities. Currently, creating that documentation requires a manual, and time-consuming process, which reduces the amount of time clinicians spend with patients, and increases the possibility of data errors. We propose to build a machine learning-based software system that can automatically generate the required clinical documentation, thereby saving time and reducing errors, which will improve the overall quality of patient care.
