The Future Of Independent Practice And Ipas

Bradley J Erickson, MD PhD, is a practicing neuroradiologist at the Mayo Clinic, and is Medical Director for AI at Mayo Clinic. He also runs a research lab focused on applying AI to healthcare, and is Chief Medical Officer for Flow SIGMA, a company focused on applying IPA in healthcare.

Artificial Intelligence (AI) has captured the attention of many in the healthcare space. Its seeming ability to divine correct diagnoses from piles of pixels or tons of tables is almost magical. But a careful investigation reveals the many shortcomings when AI is applied to medicine. Too frequently, the models are very dependent on nuances in the data, sometimes are biased due to deficiencies in the training data, and lack the breadth of information required to make accurate predictions or diagnoses. As a result, there are now serious concerns about the validity of many AI tools in medicine.

A less magical form of AI, known as either Intelligent Process Automation (IPA) or Robotic Process Automation (RPA), is widely used in other industries, and has shown tremendous success in reducing costs while improving efficiency. The benefits derive from defining the business process that is used to make the product or service, and then using technology to either directly execute a step in that process or help to assure that a human properly executes the step. The distinction between RPA and IPA is whether a ‘robot’ (software or hardware) is used to execute the automated steps, and for our purposes, that distinction is less important. The key point is that there are programming languages (like BPMN— Business Process Model Notation) that can describe the process of providing healthcare. A workflow engine then executes this BPMN ‘program’, and thereby can help assure efficient and reliable execution of that process. Orchestration is another term used in this space, and it reflects the fact that there are often multiple parallel streams of activity that must be coordinated to provide high quality healthcare, which is another challenge that taxes human healthcare providers. For the rest of this article, IPA will be used to refer to this family of technologies.

Several major healthcare consultancy organizations have recognized the importance of IPA to healthcare. Most IPA applications in healthcare today are back office--that is, they assist in automating the billing process for hospitals and payors. Specifically, these tools can help assure that bills issued for a procedure match the steps, time, and devices that are reported in the procedure note. They can help collect documentation when a bill is issued, improving bill collection rates and times. On the payor side, IPA can assure that all proper documentation has been submitted and it supports the bill that has been submitted.

Despite some adoption of IPA into the back office of hospitals, there is essentially no use of IPA technology in patient care (the ‘front office’). That is not for the lack of a recognized need. A recent survey found that 98 percent of hospital leaders felt that care coordination was their top challenge. KPMG noted that “The challenge for healthcare providers is in how ‘enablers’ such as digital and analytical tools can be brought together in a connected way to better respond to the needs and preferences of consumers, while at the same time improving outcomes at a population level in a financially sustainable way.”

Frost and Sullivan argues that healthcare must include coordination (assuring the right resource is addressing the right problem at the right time), automation (having computers do as much of the work as possible), decision support (use AI and other technologies to help make good decisions), and coordination (making sure that required information is made available to the correct team members and patient).

A challenge is that Electronic Health Records (EHRs) are physician focused, not patient focused, resulting in disintegration of the patient experience. IPA can improve the patient experience by providing timely access to results and also give patients some decision making capability. Scheduling of tests is one such example, but healthcare testing is more complex than just finding an appointment time.There are often requirements that tests be done at a certain time (e.g. fasting blood glucose), in a certain sequence (e.g. doing an contrast-enhanced CT before an iodine uptake will invalidate the thyroid test), or with a certain delay to get results before a decision must be made (e.g. some lab tests must be sent to outside facilities, so having a follow up appointment before the results are back would be wasteful). An IPA solution can assist both in executing these rules for scheduling and educating the patient on these interactions and requirements. 

IPA can improve the physician experience by assuring important information is available at the time required. Today’s EHRs have extensive notification capability, but it is not intelligent. In fact, so many notifications are now given to physicians that they rapidly click through them without reading them, a phenomenon known as ‘Notification Overload”. The lack of an IPA capability in EHRs is a big problem. EHRs also tend to be monolithic, making it more challenging to integrate external technologies like RPA, though greater support of FHIR is helping to address this challenge. Experts suggest that IPA could assist in coordination of care and might save millions of lives each year.

As AI tools become more widely used in healthcare, there will be a need to monitor their performance, and likely, retrain them EPIC sepsis. This could require substantial human effort to identify the procedures or patients on which an AI was executed, collect the results, compare those results with some sort of truth, and perform an analysis. Collecting data for retraining the AI would be similar. IPA could be very helpful in performing all or nearly all of these steps, reducing the total cost of AI. In a similar fashion, IPA can also do similar automated quality assessments of nonAI tasks: they can identify when certain diagnoses are proposed and then find subsequent test results that can confirm the diagnosis; they can identify the need for follow up testing after a certain delay (e.g. follow up lung or breast abnormality in 6 months) and then determine if that testing was done, and if not, prompt the patient and/or physician to perform the follow up.

There is immense potential for IPA to impact the entire healthcare system, not only saving dollars but more importantly, saving lives. IPA has already demonstrated its value in other industries and in the back office of healthcare. It is essential that healthcare IT professionals learn about IPA and consider how it can be effectively applied to their hospital, for the benefit of patients, physicians, and society.

Robotic Process Automation (RPA) is defined as …“RPA allows computer software with AI and machine learning capabilities to handle high-volume repeatable tasks such as queries, calculations and transactions previously done by humans.”

Intelligent Automation (IA) and/ or IPA is defined as …”Intelligent Process Automation (IPA) is the collection of technologies that come together to manage, automate and integrate digital processes. The primary technologies that make up IPA include Digital Process Automation (DPA), Robotic Process Automation (RPA) and Artificial Intelligence (AI).” IPA and Hyperautomation are basically the same thing but HA is definitely enterprise IPA plus WFO.

Hyperautomation (HA) is defined as …“Hyperautomation is a businessdriven, disciplined approach that organizations use to rapidly identify, vet and automate as many business and IT processes as possible.”

Hyperautomation involves the orchestrated use of multiple technologies, tools or platforms, including:

• Artificial intelligence (AI)

• Machine learning                     

• Robotic process automation (RPA)

• Business process management (BPM) and intelligent business process management suites (iBPMS)

• Integration platform as a service (iPaaS)

• Event-driven software architecture

• Low-code/no-code tools

• Packaged software

• Other types of decision, process and task automation tools

Forbes 2021 … “Hyperautomation is the natural extension of RPA. It is the sophisticated use of technologies and science to deliver AI-based outcomes.

In fact, truly successful application of hyperautomation to technology happens when the real-life impact is almost imperceptible to the user. Hyperautomation differs from the automation (RPA) we’ve become used to because it combines a multitude of different automation technologies and processes and links them together. By integrating machine learning, AI, RPA, robotics and IoT, hyperautomation creates a seamless string of automated tasks, requiring very little human intervention or interpretation.”