Author: Rob Albritton, Senior Director, AI Center of Excellence
A recent Armed Forces Communications & Electronics Association (AFCEA) Health IT Summit provided trends and insights in emerging technology, automation, and leveraging data as a strategic asset to better serve citizens. Throughout the summit’s myriad sessions, one theme remained consistent: the need for better data management. Without a means to discover, interpret, and use data efficiently, information collected and stored for the purposes of improving patients’ health is often wasted. Here are two ways AI can help solve these challenges.
Problem: Agencies are awash in data but are not always able to use it in support of the mission.
Eighty percent of health care data is unstructured, in the form of doctor’s notes, medical test results, and images. Natural Language Processing (NLP), Machine Learning (ML) and Deep Learning (DL), all forms of AI, are inherent allies in the effort to turn this data into actionable information for agencies and the public.
NLP is a technology used to aid computers in understanding human spoken language and turning that language into executable commands. This makes data searchable, discoverable, retrievable, and shareable by allowing the user to employ natural language as the primary mechanism for entering search queries. NLP makes systems usable and data retrievable to authorized users who have a simple question about specific data, so users do not require special training to find the information they need. NLP can also reduce the time it takes to locate data, important in the health sector, where speed and accuracy can sometimes mean the difference between life and death.
However, traditional rule-based NLP approaches cannot read and interpret dynamic unstructured data without the aid of ML. Because reports, notes, and other unstructured data are created by humans with evolving language, the way the data is ingested and interpreted must always be adapting. Teaching the computer to recognize the nuances of language and meaning requires several steps. First, written words must be converted to text via Optical Character Recognition (OCR) models. Then text can be converted to features (vectorized) so machines can read and make sense of the data. Finally, ML and DL models can be trained over time to understand slight nuances in syntax, dialect, and other changing characteristics of human language. By working in tandem, NLP, ML, and DL are then able to not only make data discoverable but able to be analyzed and used for making more informed decisions.
Problem: There are efforts to bring major Health IT enhancements to federal agencies. Many of these are multiyear programs in midflight. These efforts could use a boost to meet deployment deadlines and achieve objectives.
Robotic Process Automation (RPA), a gateway technology to fully developed AI, is one solution to this problem. RPA is software created to do basic tasks across applications, just as human workers do. This emerging technology reduces the burden of repetitive, simple tasks on employees.
RPA is rapidly becoming a go-to resource to drive efficiencies and reduce processing time for repetitive, rote tasks. By adopting RPA, agencies can streamline operations, drive down costs, and move government employees into more challenging positions that require human capabilities like reasoning, analysis, creativity, and empathy. Using RPA, government agencies can push rote tasks — such as verifying that a form has been properly completed — to systems designed to process standardized information rapidly and with lower error rates. RPA can free staff to focus on mission-centric tasks, helping them to meet deadlines and achieve objectives.
The need for better data management is all around us, but AI is there, ready to harness, and we’ve seen its transformational capabilities. From interpreting handwritten medical records to making personalized health recommendations for patients to providing virtual care, AI can drastically improve health care. When your agency is ready to realize the power of AI by addressing data challenges and developing emerging technologies like ML, NLP, DL, and RPA, contact a member of Team Octo.