The Dangers of Automated Clinical Decision-Making

We live in the age of convenience, and today’s consumers continually look to technology as a means of bypassing the manual—and often time-consuming—tasks that clutter our schedules. We text so we don’t have to call. We shop online so we don’t have to fight crowds—and food-court cravings—at the mall. We deposit checks using our smartphones so we don’t have to wait in line at the bank—if we even use checks at all (PayPal, anyone?). In short, we rely on technology to make our lives easier—and that dependence is now extending to the workplace, as well.

Business technology has become the oil that improves organizational efficiency—and financial viability—in the ever-evolving economic landscape. After all, if your business doesn’t figure out a way to automate the processes that are dragging down your productivity—and your profit margins—then one of your competitors will. But, the push for automation presents some unique challenges to professionals in certain industries, health care included. That’s because, while artificial intelligence is lightyears ahead of where it was 20, ten, or even five years ago, it’s still nowhere near ready to substitute the depth of thinking, reasoning, and decision-making inherent to clinical practice. And honestly, we wouldn’t want it to—because at that point, our value as clinical experts would be reduced to nothing more than a few lines of code. Furthermore, as Larry Benz and Tim Flynn have demonstrated in their research, the placebo effect of a therapist’s caring, empathetic interactions with his or her patients can greatly affect treatment outcomes—and you definitely can’t automate empathy.

That’s why I was so alarmed to see a recent ad for a rehab therapy software promising to automate the complexity determination process associated with the new CPT codes for PT and OT evaluations. You’ve almost certainly heard about these codes—which require providers to assign a level of complexity to each evaluative episode—by now (if not, then your practice has much bigger fish to fry than its choice of software solutions). Now, as I emphasized in my recent webinar with rehab therapy compliance expert Rick Gawenda, clinical decision-making is a critical component of accurate code selection. That’s because, even though the guidelines governing proper code selection are incredibly detailed, they cannot account for every single unique situation. You know as well as I do that no two patients are alike, and the assumption that each patient evaluation will fit neatly into the low, moderate, or high complexity bucket is nothing short of ridiculous (as proven by the onslaught of scenario-specific questions my team and I received during the webinar and via email in the days and weeks leading up to the January 1 code implementation date).

Fully automated, canned responses and notes cannot possibly account for all the nuances that influence the complexity of a patient evaluation. If it could, there wouldn’t be such a massive demand for education around the code criteria and how to apply it. And the fact that a software vendor is claiming the exact opposite troubles me for many reasons. Chief among them: Total clinical automation is an incredibly slippery slope. Yes, technology can (and should) help streamline the treatment and documentation process—by steering providers toward appropriate evidence-based tests and interventions, for example. But, technology will never replace the depth of clinical knowledge, experience, and expertise necessary to complete an evaluative episode—or to determine the appropriate plan of care for a particular patient.

Furthermore, if we are not in control of the data we are recording, then we put the quality of that data in jeopardy—and that does nothing to move us forward as individual practitioners or as an industry. Instead, we should be using technology that helps us take control of not only our clinical processes and data, but also our payments. After all, with third-party insurances continuing to up the ante on payment scrutiny, the need for greater efficiency in justifying our clinical decisions will only continue to grow. But, things get tricky—and potentially slippery—as we try to decide which areas of the care delivery process warrant automation.

The self-diagnosis and self-treatment problem is already dangerously pervasive in the neuromuscular space; plenty of people who should be seeking treatment from a rehab therapist mistakenly think they can achieve the same results by reading a couple of WebMD articles and finding a few exercise videos on YouTube. Plus, there are plenty of HEP programs and apps out there that claim to be the answer for the diagnoses that we see the most—including low back, knee, and/or shoulder pain. And if we, as therapists, buy in to the idea that a software system or app can make better decisions than we can, then we’re basically phasing ourselves out of the care delivery process—in which case we’d leave a lot of patients vulnerable to further functional decline.

Now, I’m not suggesting that therapists boycott clinical decision support (CDS) of any kind. In some cases, it can actually enhance therapy treatment and care plan development—by suggesting appropriate tests and measures, identifying red flags, or proposing possible treatment routes, for example. But, there’s a difference between using technology to help you do your job better and relying on it to do your job for you. And when the strength and accuracy of your data—and thus, your and your fellow therapists’ future as valued members of your patients’ care teams—are on the line, you can’t afford to fall into the habit of outsourcing your expertise as a skilled healthcare provider to a piece of software. And to bring it back to the issue of the new eval codes, once payers adopt differential payments based on the tiered levels of complexity—assuming they amass the data they need to assign accurate reimbursement rates, that is—any cookie-cutter, canned responses won’t get past auditors, anyway.

Sure, convenience is nice; most of us will probably never go through the trouble of looking up a number in the white pages again. But what’s easy isn’t always what’s best—and that’s certainly true when it comes to clinical judgement. So, whatever regulatory changes come down the pike—and trust me, there will be many more—I urge you to approach fully automated compliance solutions with caution. After all, many of those regulations will serve the overarching transition to a healthcare system rooted in value, and you certainly can’t automate that.


About the Author

Heidi Jannenga PT, DPT, ATC/L is the president and co-founder of WebPT, the leading practice management solution for physical, occupational, and speech therapists. Heidi leads WebPT’s product vision, company culture, and branding efforts, while advocating for the physical therapy profession on a national scale. She co-founded WebPT after recognizing the need for a more sophisticated industry-specific EMR platform and has since guided the company through exponential growth, while garnering national recognition. Heidi brings with her more than 15 years of experience as a physical therapist and multi-clinic site director as well as a passion for healthcare innovation, entrepreneurship, and leadership.

An active member of the sports and private practice sections of the APTA, Heidi advocates for independent rehab therapy businesses, speaks as a subject-matter expert at industry conferences and events, and participates in local and national technology, entrepreneurship, and women-in-leadership seminars. In 2014, Heidi was appointed to the PT-PAC Board of Trustees, and in 2015, she was named PT of the Year by the Arizona Physical Therapy Association. She also serves as a mentor to physical therapy students and local entrepreneurs and leverages her platform to promote the importance of diversity, company culture, and overall business acumen for private practice rehab therapy professionals.

Heidi was a collegiate basketball player at the University of California, Davis, and remains a lifelong fan of the Aggies. She graduated with a bachelor’s degree in biological sciences and exercise physiology, went on to earn her master’s degree in physical therapy at the Institute of Physical Therapy in St. Augustine, Florida, and obtained her doctorate of physical therapy through Evidence in Motion. When she’s not enjoying time with her daughter Ava, Heidi is perfecting her Spanish, practicing yoga, or hiking one of her favorite Phoenix trails.


4 responses to “The Dangers of Automated Clinical Decision-Making

  1. Terry Abrams says:

    I agree with you one-hundred percent. Physical therapy is a human endeavor and leaving it in the hands of a complicated mathematical algorithm will take away that special touch of any therapist. It takes a human perspective to empathize with a patient and really invest time into in order for them to improve.

  2. Appreciate the read and comment Terry! The role of technology should play in PT is to support the human endeavor never replace.

  3. Thank you Heidi for bringing this matter to the attention of our profession. I would, however, change your title “The Dangers of Automated Clinical Decision-Making” . It should be written as “The beauty of Automated Clinical Decision-Making”. Read further for my point.

    Let me start by saying that I have been a Physical Therapist for 24 years now having worked, developed and directed programs in the hospital environment, rehab, and several outpatient facilities in private practice. I began my practice in May of 2002 and have grown triple fold since inception. I am a graduate of EPPM 2010. I can truly say I have experienced the “employee mentality” and the “employer mentality”. Like many of my colleagues, I have been through the “ringer’ that Medicare, Worker’s Compensation, and other commercial insurances have strung around our professions neck. Decreasing reimbursements, diminishing visit for treatment, PQRS, MG2 forms and other menacing rules and regulations that “steal us away for our hands on treatments have been hurled at our faces. In my experience, the key to a successful practice is the “AmaZing Experience” our patients get, BUT we need to get PAID for our services.

    The point you make referencing Larry and Tim, Heidi, actually supports using technology to our advantage keeping our hands on the patient and NOT the keyboard! This new evaluation system worries me, NOT because Systems 4 PT automates it using the data the patient and the PT enter into it, it is because NOT everyone is using this technology. We need uniformity. Despite the many courses out there that offer how to select the evaluation codes, some therapists will get it wrong. If I didn’t have this system, I too would make the same mistake! The insurance companies are providing the noose and we can choke ourselves with it. They are sitting back and laughing! They will use our ineptitude to decrease our reimbursement even more. They want to take our hands OFF the patients and inundate us with paperwork.

    The way you’ve phrased it, “computers take away clinical decision making and do all the work” goes too far. My EMR (Systems4PT) data mines my eval to find complexity-related topics, categorizes them and scores them per CMS rules. I add my clinical judgement and the eval code is defended with 100 – 200 words of air-tight evidence. This all takes 30 seconds and my clinical judgement drives the boat!

    This blog misses the point. “Use technology to take control of treatment processes and get paid”. Um… you left something out: Use technology to automate busy work and spend more time treating. After all, our job is not to pick out complexities that we’ve already documented in the eval.

    Rather than be defensive, “technology will reduce our job to a few lines of code”, change your perspective: Our value is patient progression. Patient progression is the result of hands-on treating. Technology automates busywork, so we can spend more time hands-on treating patients.

    Using technology that selects, categorizes, scores and defends complexity in 30 seconds, (all led by my clinical judgement), is a blessing. And it’s why I’m able to spend more time treating.

    More time treating, getting results and getting paid is where all of us want to be!!

    1. Thanks for your comment Jamie. Believe me, I am on your side when it comes to regulatory fatigue. It is frustrating that PTs have to jump through so many hoops just to get paid for the services we provide. But the truth is—and I know it’s tough for some of us to accept this—we really only have our ourselves to blame, because historically, we have failed to objectively demonstrate and prove the value we provide to our patients and the healthcare system at large. To our payers, we are nothing more than a cost, because the only real data they have on us is that associated with utilization rates and claims data. So, it’s no wonder they are constantly looking for ways to reduce that cost. And the only way for us to put a stop to that seemingly endless stream of reimbursement-slashing regulations—which, in addition to squeezing our profit margins, force us to complete more paperwork and take even more focus off of our patients—is to not only assert our value, but also prove it in a meaningful way. The new evaluation codes provide an opportunity—albeit a small one—for us to supply meaningful data about the complexity of the work we do. My contention with technology that completely automates that process is that it will lead therapy providers to de-emphasize the decision-making component of complexity selection—which in my opinion is the most important piece of the puzzle. After all, as I pointed out in my article, every patient case is different, and I honestly cannot imagine how a software program could account for all of the details and nuances that influence code selection. If we entrust our data efforts to such a program, we are putting our data integrity—and thus, the future of our profession (including how we are paid)—in someone (actually more like something) else’s hands. While that approach to coding may work now—after all, Medicare is paying the same rate for all three codes, so there’s no real incentive for intense scrutiny of our coding decisions—I believe that it will come back to bite us at some point down the road. And that could lead to even more regulatory strife in the future.

Leave a Reply

Your email address will not be published. Required fields are marked *