If mechanical machines play an essential part in helping people to improve muscle strength, why can't digital machines enhance man's ability to increase the efficiency of the healthcare system (Thomas, 2019)? Machine learning (ML) uses business processes to study patterns of data resulting in predictions and recommendations (Bresnick, 2017).
Healthcare related administrative burdens include the business processes and associated costs that patients, providers, and insurers incur but are not directly related to the delivery of medical care (Gottlieb e Shephard, 2018). ML represents an approach to using data to make or recommend decisions with a focus on improving efficiency within the healthcare system and thus reducing administrative burdens.
Recent studies into the impact of administrative burdens on patients, providers, and insurers have found that costs associated with administrative burdens (2019) may exceed $496m (Gottlieb e Shephard, 2018). According to this same study for every $1B in revenue collected, the health care system requires 770 full-time workers to handle the administrative burdens. Considering there was more than $900b in revenue within the private commercial sector in 2014 (Becker, 2019), the number of resources applied to handle administrative burden is substantial. If the application of ML can help to decrease the resources spent on administrative burden by 10%, the savings could be extensive.
The purpose of healthcare is to improve or enhance a person’s quality of life. Administrative burdens directly impact a patient's quality of care. In one recent study, 1/3 of all patients studied had difficulty understanding their medical and drug coverage rules (Davio, 2019). The same survey revealed that nearly 50% of all patients surveyed received bills that were not adequately covered (Davio, 2019). With the advent of ML, patients can dramatically improve their quality of care through the removal of administrative burdens that contribute to delayed treatment decisions and inability to access medications in a timely/cost-effective manner.
Prescription renewal is a common administrative burden that impacts both patients and providers. Many providers have additional staff members whose sole responsibility it is to screen and respond to prescription renewals currently. Many smaller providers faced with the cost of a staff to handle prescription renewals only process renewals during in-person consultations. Using ML, it may be possible for a provider to analyze a requested renewal based on the patient's profile, type of medication, and other factors. The configured ML algorithm may approve certain types of prescription without human intervention.
If providers are ever-improving their ability to provide higher levels of care, techniques such as ML can help to highlight the need for automating guidelines and protocols (Trevena, 2019). Guidelines and protocols are rules that dictate the type of care and prescription medications that providers can order for a patient based on the patient’s insurance coverage.
Some guidelines and protocols are related to diagnoses, while many are related to insurer regulations. ML can enable providers to either remove or dramatically improve burdens that include claims submission, decision support, prescription renewals, and patient appointments.
Many providers use specially trained personnel whose primary function is to evaluate treatment protocols for applicability to the patient's insurance coverage. Because rules govern insurance coverage, it may be possible to configure the rules insurers use to approve a claim at the provider's point of care. Take the prescription of a new drug, for example. In some practices, an average of two days per week per physician is spent addressing burdens, including prior-authorization requests (Robeznieks, 2019). Using ML, it may be possible for the provider to determine the extent to which an insurer will cover a particular drug and handle any required prior authorization request during the initial patient consult.
In a recent study, Insurers spend upwards of 12% of total costs on billing and administrative (e.g., claims processing) services (Gee, 2019). Much of the 12% of total expenses are spent evaluating claims for potential readmission and other quality of care issues.
With ML, insurers may enable providers to more accurately identify patients with a high probability of re-admission and take steps to intercede with that patient before a catastrophic medical event. Reductions in the costs of these administrative burdens may lead to lower overall costs and trickle down into improved provider efficiency and increased patient access to care (Frakt, 2018).
ML is an emerging set of technologies and processes that provide users with advice, decision support, recommendations, and actions. uses formulas and algorithms to learn from experiences resulting in an activity or decision recommendation (Portugal et al.).
The first element of a potential ML solution is to define the problem and create a set of formulas and algorithms that result in an action or recommendation. Enabling patients to improve their ability to understand insurance coverage is a process based on methods and algorithms. In the current environment, if a physician selects the wrong diagnosis, the insurer may reject the claim leading to many days of activities to fix and address the issue (Djavaherian, 2019). Configuring an ML solution to adequately interpret a patient's coverage at the time of care may dramatically reduce the administrative burden spent revising treatments and claims once the insurer has denied a claim.
Algorithms enable providers to analyze patient data to make improved decisions and recommendations. In my experience, providers spend a significant amount of time handling prescription renewals and prior authorizations. An ML-based solution could take a patients' renewal request and evaluate the patients' records to determine if the requested renewal is allowable (based on the patients' electronic record profile) and then either approve the request or pass the request to the proper reviewer. In this scenario, it might relieve providers of the resources required to approve standard prescription renewals. In the case of prior authorizations, we previously discussed the use of ML to move insurer prescription decision making to the point of care, enabling providers to handle prior authorization requests at the time of care.
Many insurers are focused on working with patients and providers to reduce readmissions. One particular area of readmissions where ML may hold promise is in identifying early-stage heart failure (HF). Kaiser Permanente has recently stated that HF is the leading cause of hospital readmission among patients 65+ and that providing patients with HF or suspected HF diagnosis may significantly lower the odds that a patient with HF may be re-admitted (Permanente, 2019). Recently, ML was applied to a population of patients with potential HF indicators. The population, evaluated through the model, was determined to be a candidate for enabling care teams to “target interventions at their most high-risk HF patients and improve overall clinical outcomes by treating the whole picture of the patient, not just a single diagnosis” (Golas et al., 2018). Identifying patients with potential HF risk factors as early as possible is critical to ensuring the patient's HF is addressed appropriately.
The views and opinions expressed in this content or by commenters are those of the author and do not necessarily reflect the official policy or position of HIMSS or its affiliates.
Experience the education, innovation and collaboration of the HIMSS Global Health Conference & Exhibition with on-demand sessions available as your schedule allows. We’re also supplementing our HIMSS20 programming with critical COVID-19 content.
BECKER. 17 statistics on the current state of US healthcare spending, finances: Here are 17 interesting facts and statistics on finance in the U.S. healthcare industry, all in one place, 2019. Disponível https://www.beckershospitalreview.com/finance/17-fascinating-statistics-on-the-current-state-of-us-healthcare-spending-finances.html.
BRESNICK, J. Machine Learning in Healthcare: Defining the Most Common Terms. 2017-07-06 2017. Disponível https://healthitanalytics.com/news/machine-learning-in-healthcare-defining-the-most-common-terms.
DAVIO, K. For Seriously Ill Patients, the American Healthcare System Adds to the Burden of Disease. 2019. Disponível https://www.ajmc.com/focus-of-the-week/for-seriously-ill-patients-the-american-healthcare-system-adds-to-the-burden-of-disease.
DJAVAHERIAN, C. Industry Voices—Patients aren’t cattle, so deliver a personalized process and improve patient outcomes - ABI/INFORM Collection - ProQuest. 2019. Disponível https://search-proquest-com.proxy.library.nyu.edu/abicomplete/docview/2269209485/6C5588AB00B04AEAPQ/3?accountid=12768.
FRAKT, A. The Astonishingly High Administrative Costs of U.S. Health Care. 20180716 2018. Disponível https://www.nytimes.com/2018/07/16/upshot/costs-health-care-us.html.
GEE, E. Excess Administrative Costs Burden the U.S. Health Care System - Center for American Progress. 2019. Disponível https://www.americanprogress.org/issues/healthcare/reports/2019/04/08/468302/excess-administrative-costs-burden-u-s-health-care-system/.
GOLAS, S. B. et al. A machine learning model to predict the risk of 30-day readmissions in patients with heart failure: a retrospective analysis of electronic medical records data. BMC Medical Informatics And Decision Making [BMC Med Inform Decis Mak] 2018v. 18, 2018. ISSN 1.
GOTTLIEB, J.; SHEPHARD, M. How Large a Burden are Administrative Costs in U.S. Health Care? | Econofact. 2018-09-06 2018. Disponível https://econofact.org/how-large-a-burden-are-administrative-costs-in-health-care.
PERMANENTE, K. Early Follow-up with Patients after Hospitalization for Heart Failure Reduces Readmission Rates - ABI/INFORM Collection - ProQuest. 2019. Disponível https://search-proquest-com.proxy.library.nyu.edu/abicomplete/docview/1773525852/5FCC1F4B046F4795PQ/3?accountid=12768.
ROBEZNIEKS, A. Prior authorization is a major practice burden. How do you compare? , 2019. Disponível https://www.ama-assn.org/practice-management/sustainability/prior-authorization-major-practice-burden-how-do-you-compare.
THOMAS, M. Ultra-modern medicine: Examples of machine learning in healthcare. 2019. Disponível https://builtin.com/artificial-intelligence/machine-learning-healthcare.
TREVENA, L. RACGP - Minimally disruptive medicine. 2019. Disponível https://www1.racgp.org.au/ajgp/2018/april/minimally-disruptive-medicine.