By Stuart Rabinowitz, MBA, MSHI and Emmanuel Iroanya, SAFe Agilist (SA)

In our previous HIMSS Clinical Informatics newsletter article, “Can Machine Learning (ML) Reduce Patient, Provider, and Insurer Administrative Burdens?" we defined administrative burden as essentially the administrative activities that add non-care-related costs into the healthcare system. We further defined ML as a set of technologies that use patterns and data to predict outcomes and make recommendations. In this article, we begin to explore the use of ML in supporting the expanding field of healthcare interoperability. Interoperability includes the “ability of different information systems, devices or applications to connect, in a coordinated manner, within and across organizational boundaries to access, exchange and cooperatively use data amongst stakeholders, to optimize the health of individuals and populations”(@HIMSS, 2016). This article reflects on specific areas (interoperability) where ML might be able to play a role in improving the administrative burden’s impact. As mentioned, Interoperability is ultimately the ability to transfer healthcare-related information across systems and organizational boundaries. In this article, we focus on reducing administrative burdens tied to interoperability.

Interoperability – Can It Reduce Administrative Burdens?

Administrative burdens represent the costs that patients, providers, and insurers incur as part of the approach to deliver healthcare. According to recent studies, it is estimated that 15-30 % (GOTTLIEB, 2018) of total healthcare spend goes towards administrative burdens. Specific types of interoperability-based administrative burdens that impact patients, providers, and insurers are described in Table 1.

 

 

Stakeholder Impact

Interoperability related administrative burden

Burden Description

Patients

Providers

Insurers

Billing and Claims

According to some studies, healthcare requires 770 workers to bill and collect $1b in revenue(GOTTLIEB, 2018)

Bills presented to patients are often incomprehensible, have not yet been settled with Insurers, and may not explain the meaning of all charges.

Some physicians may work with 10's if not 100's of insurers, including Medicare and Medicaid. This broad swath of Insurers leads to multiple types of billing mistakes (e.g., incorrect codes, and care justification)

Insurers are responsible for reviewing and approving all physician claims promptly so that patients understand their coverage. This process can sometimes take weeks to complete.

Sharing medical records

Medical records generally require physical printing and transportation unless the physicians happen to be in the same network.

Many patients keep physical or scanned copies of their medical records. Some providers still charge a fee to provide physical copies.

Providers may not accept patient delivered medical records instead requiring faxed copies from the other provider.

Automatic querying and limited consent requirements are organizational HIE policy decisions that

impact the volume of exchange, and ultimately the information available to providers to support optimal care(DOWNING; ADLER-MILSTEIN; PALMA; LANE et al., 2017)

Table 1 - Administrative Burdens

Value of Machine learning and Interoperability

Machine learning can be part of a technology suite that ensures all appropriate patient data is included in the doctor’s workflow(SHRYOCK, 2019). In a recent survey, a physician indicated that 70 to 80 percent of the information he needs is populated in the EHR, but the remaining requires checking portals or making phone calls to track it down”(SHRYOCK, 2019). I have experienced this directly when a provider doesn’t have access to a lab’s electronic system and requires faxes. I've shown my lab results to my provider on my phone because they couldn't get the fax to work. Each of these situations contains patterns of data that if appropriately identified, could be handled by machine-based rules (ML).

We understand the goal of healthcare interoperability is to connect healthcare systems electronically (to facilitate the electronic sharing of data). Unfortunately as CMS Administrator Seema Verma recently said, “Currently, the lack of patient access to price and quality information — and their [EHR] data—is a lost opportunity to keep our nation healthier and to drive down costs," According to Verma, “We're now left with a healthcare industry that still uses fax machines."(DREES, 2019). Verma continues to declare that more technological innovation is needed to “create a more efficient system”(DREES, 2019). One of the primary challenges facing successful interoperability is the “implementation of new quality metrics and requirements for public reporting”(SHANAFELT; NOSEWORTHY, 2017). These new metrics and reporting requirements put additional pressure on an interoperability environment that is already challenged with connecting and delivering information. So, where does one start when thinking about how to improve interoperability with ML? It is my opinion that two particular administrative burdens that could be enhanced through the use of interoperability and ML include improving billing and claims and medical record exchange.

Billing and Claims – Is it possible to send electronic bills that are understandable?

Recent data suggests that 86% of consumers receive paper medical bills, 88% of providers accept paper checks from one or more payers and that 85% of providers would prefer the use of electronic payments (@MEDDATAINC, 2017). The same study found that 92% of consumers want to know their payment responsibility prior to a provider visit, 74% of consumers are confused by Explanation of Benefits (EOBs) and medical bills, and more than 73% of providers report that it takes one month or longer to collect from patients (@MEDDATAINC, 2017). Each of these data points demonstrates that even with advances in interoperability and EHR technology, there is still a need to reduce paper bills, make billing easier to understand, facilitate electronic payments, and improve patient or payer payment timing. Improving and understanding of EOBs and their underlying claims is one particular administrative burden that ML-powered interoperability could help. Improvements in understanding EOBs may lead to more informed patients and improved payment cycles.

Applying ML-powered interoperability to EOBs will enable providers and payers to provide more informative information resulting in billing statements delivered in readable and understandable format with machine-generated explanations. These technologies are configured to look for patterns within the underlying data and use ML to convert those patterns into readable documents. EOBs that often accompany billing or claims materials are often worded in complex language. With more transparent explanations, patients may be more likely to pay any resulting bills.

Value – Return on Investment

Why should the healthcare industry and government organizations invest precious resources into the use of ML as part of an interoperable healthcare environment? The first reason for continued investment is actual costs generated by billing and claims related activity. In one study, it was found that the administrative costs associated with billing and insurance-related activities for emergency department visits could be as high as 25%(@MEDDATAINC, 2017). According to recent data, there were more than 145m emergency room visits in 2016 (FastStats - Emergency Department Visits, 2019). When combined with an average cost of $1300 (COSTHELPER.COM, 2019) for an emergency room visit, reducing the administrative burden by %20 could result in a yearly cost savings of $80/visit ($325 avg burden). In the same study, 83% of Physician Practices (<five practitioners) said the slow payment of high-deductible plan patients are their top collection challenge. There is some school of thought that the paperwork and complexity of exchanging medical records (including claims) is slowed due to the length and complexity of paperwork.

Introducing ML into the creation of EOBs and other claims data may improve patient payment rates. In 2016 68% of patients failed to fully pay off medical bills up from 53% in 2015 (@MEDDATAINC, 2017). If for example, a physician practice has 200 patients with high-deductible plans and the average payment rate is 120 days, by improving the understanding of EOBs could reduce the payment cycle by just %5 would improve provider cash flow. This approach helps to directly address the more than 81% of surveyed physicians who complained of difficulties that practice staff have at communicating patient payment accountability (@MEDDATAINC, 2017).

Bibliography

@HIMSS. What is Interoperability? 2016-03-09 2016.

@MEDDATAINC. 2019 Medical Billing Statistics | MedData. 2017-10-26 2017.

COSTHELPER.COM. How Much Does An Emergency Room Visit Cost? - CostHelper.com. 2019.

DOWNING, N. L.; ADLER-MILSTEIN, J.; PALMA, J. P.; LANE, S. et al. Health information exchange policies of 11 diverse health systems and the associated impact on volume of exchange. Journal of the American Medical Informatics Association, 24, n. 1, p. 113-122, Jan 2017.

DREES, J. CMS Administrator Seema Verma: Despite billions spent, EHRs haven't delivered on interoperability. 2019.

FastStats - Emergency Department Visits. 2019-07-23T02:55:02Z/ 2019.

GOTTLIEB, J. How Large a Burden are Administrative Costs in U.S. Health Care? | Econofact. 2018-09-06 2018.

MARTIN, K. The next big EHR interoperability idea? 2018-08-29 13:38:37 2018.

PELSER, J. The Rise of the Robot Reporter. 20190205 2019.

SCRIPTSENDER. ScriptSender Reduces Fax Volume: Busy imaging center finds solution for "biggest problem" - ABI/INFORM Collection - ProQuest. 2019.

SHANAFELT, T. D.; NOSEWORTHY, J. H. Executive Leadership and Physician Well-being: Nine Organizational Strategies to Promote Engagement and Reduce Burnout. Mayo Clinic Proceedings, 92, n. 1, p. 129-146, Jan 2017.

SHRYOCK, T. Physicians' EHR work-arounds - ABI/INFORM Collection - ProQuest. 2019.