Resources / Blog

How ML Boosts Efficiency in Healthcare Information Management

Partnerships in machine learning have more than just intrinsic value. In fact, developing partnerships with industry leaders has many advantages for both parties. Intelligent document processing can transform an organization’s ability to extract information from varied document types and be a great resource for improving workflow.

A common issue for people working in healthcare is that health information data and patient records can be difficult to analyze. Traditional patient health files contain information from several sources such as electronic records, faxed forms, scanned files, written physician and nurse notes, and other miscellaneous provider additions.

Instead of being able to do a keyword search similarly to other electronic documents, reviewers must skim through documents line by line to pull the information they need. Also, pertinent is usually not all in a centralized location, making reviewing patient information a timely and tedious manual process.

 

Difficulties with this process

In the case of one of our customers, a national pharmacy benefits management company, patient records are an integral part of processing prescription requests and refills. Once these requests are received from a provider such as a physician’s office, hospital, or pharmacy, employees are then responsible to wade through up to hundreds of pages of medical records to ensure any prescriptions filled are appropriate and do not conflict with any information located within the patients’ files. They must also ensure any prescribed medications fit into the parameters of each patient’s specific health insurance drug policy.
 
The process entails a series of reviews that start with basic demographic information such as patient name, height, and weight. It then progresses to clinical reviews performed by licensed pharmacists. There can also be instances where the requests received or a patient’s records may be missing information that is required to move the request forward, so there is an element of back and forth between the parties involved, which increases the manual processing time. This also contributes to a higher margin of error.

More than intrinsic value

In this and other similar customer cases, Infinia ML used our machine learning and healthcare industry expertise to configure a platform that performs an automated review of patient health records. The model is able to extract data and relevant information related to the prescription request processing process. After the application of the model, our customers are now able to decrease the time and cost associated with patient health document review compared to the previous document review process. They are now able to save thousands of hours of manual document processing time and reallocate valuable resources elsewhere within the organization.

A vital aspect of the model’s application involved integrating it into the customer’s existing software environment, or bringing it “on-prem.” Due to the sensitive nature of healthcare information as well as efforts to remain in compliance with health information regulations, it was imperative that we were able to seamlessly harmonize our model into the customers’ systems. As a result, employees are able to streamline their workflow and not have to experience productivity disruptions. 

Another common theme in a lot of Infinia ML’s projects is the human-in-the-loop workflow which was another key element of this project. Per Ben Schellner, Sr. Data Scientist, “humans are good at detailed work, whereas computers are really good at doing ‘dumb’ things really fast with lots of data. The power of the human-in-the-loop workflow is that we can scan large sources of data and produce a range of lists of things that we think are useful, and then trust the human who reviews at the end to use his or her experience and domain expertise to make the final decision.” 

We are able to sift through all of the “noise” of the data and present the most relevant information to the experts, in this case, the humans processing requests, and present it to them so that they can make a final determination based on their expertise. In other words, our customers receive the best of both worlds by benefiting from the fast processing of information made available by automation and the expertise of human brains, working together to achieve a common goal.

The value of partnerships

The value of partnerships in machine learning and automation is priceless. Aside from project facilitation being more streamlined, communication between partners and customers is further enhanced by the involvement of a partnership. Furthermore, from a customer perspective, there is an additional level of trust and responsibility attached to projects. As most endeavors are a very involved process by both parties, it is always fruitful for them to be attached to a working relationship that encourages positive communication and collaboration.