Treatment Based Classification and Skilled Physical Therapy

Treatment based classification slipped past me in 2003 when the American Physical Therapy Association’s Combined Section’s Meeting came to Tampa, Florida.

The lead author of the original manipulation derivation study in the December 2002 Spine journal presented a talk titled “Spinal Manipulation Predictor Variables” – or something like that.

I saw the words ‘Spinal Manipulation’ and that lure was sufficient to get me to drive an hour from my home south of Tampa Bay, across the world’s second-longest suspension bridge (at the time) - the Sunshine Skyway, and attend, for $110, Dr. Timothy Flynn’s talk on how he discovered the first five predictor variables pertinent to physical therapy.

Dr. Flynn talked for an hour and no doubt said many things that passed over my head. His description was clear enough for me to understand, however, a few simple and liberating facts:

1. Lumbar spinal manipulation is a basic motor skill that should be taught to every graduating physical therapist in the United States.
2. Physical therapists can determine who can, and who cannot benefit, from lumbar spinal manipulation with five easy tests done in, perhaps, five minutes.
3. Lumbar spinal manipulation is low risk, high reward – when performed on the right patient.
4. Lumbar spinal manipulation is underutilized by physical therapists in the United States, Ireland and, by implication, the entire world.

Dr Flynn was especially clear in describing the sophisticated statistics that were used to derive the five manipulation predictor variables.

The explanation went thusly…

"...many potential predictor variables were tossed into a ‘hopper’ and the five that ‘fell out’ formed the parsimonious data set."

I went home a looked up ‘parsimonious’ in the dictionary.

Despite Dr. Flynn’s best efforts, however, I spent the next 2-3 years busy with other tasks not directly related to manipulation or classification.

Cold, Hard Reality

It was only in February 2006, when I purchased a three-office, seven-therapist private practice that I seriously began looking for a completely evidence-based solution to the looming threat of Medicare audits (Florida was a RAC Demonstration state, along with New York and California) that I began to understand the true importance of decision rules.

Medical decision rules hold the promise of improving efficiency without sacrificing safety, improving resource allocation, cutting costs and improving functional outcomes113,115-120,132,137 (bibliography).

Much of the push for decision rules in medicine has come from institutional cost constraints at large public hospitals like Cook County Medical Center in Chicago where Dr. Brendan Reilly found that many patients sent to intensive care cardiac beds costing $3,000 per night might have been better managed with a pack of Rolaids.

The US Navy was the initial funding source for Dr. Lee Goldman’s Chest Pain prediction rule in 1982, when no American hospitals were willing to support a narrow validation study based on the belief that a computer algorithm couldn’t compete with human doctors’ decisions. Perhaps doctors’ traditional conservatism played a role, perhaps pride, perhaps fear that a computer could do better medicine was what kept Dr. Goldman from finishing his work for 14 years.

The Navy, however, didn’t have 14 years. The Navy ran nuclear submarines under the ocean that had to stay submerged for days or for weeks at a time for reasons of national security. The Navy would surface the sub to get a sick seaman to a land-based hospital. The Navy wanted to know, definitively, if their able seaman with chest pain needed:

• A cardiac intensive care bed.
• A cardiac telemetry bed with round-the-clock nursing.
• A pack of Rolaids and a bunk.

Without bowing to culture, pride or conservative sensibilities the US Navy sponsored Dr. Goldman’s initial studies that focused, not on the diagnosis, but on the outcome of the doctors’ decision – that is, which bed did the doctor decide on?

Providing the doctor with clear-cut treatment choices to follow and with a decision rule that leaves little room for interpretation are examples of ‘systems thinking’ – the type of thinking that says that errors, including errors that lead to less-than-optimal outcomes, happen when good, well-meaning people make mistakes – mistakes like forgetting to run a certain test or check a certain vital sign. Mistakes in physical therapy are rarely life-threatening – they often only mean that the patient doesn’t get better even when no new harm befalls the person.

Treatment based classification provides the physical therapist with a checklist that ‘backs up’ the examination process and prepares the therapist for the evaluation. As stated by Lee Goldman, MD:

“The modern approach to patient safety emphasizes “systems thinking” rather than individual cognitive mistakes or technical “slips”.

The goal is to create processes and solutions to prevent human errors, which are commonly made by competent individuals”.

Some therapists consider TBC to be “paint-by-the-numbers” or, what we called “cookbook physical therapy”, – something we swore we would never do in high-minded discussions in physical therapy school or after work in comfortable surroundings.

What gets missed in some of these discussions is the fast-paced reality of the physical therapy clinic where decisions get made on-the-fly, based on symptoms, or patient complaints, or lack of progress, or “that’s-what-I got-from-Joe-last-time” or “what-do-I-do-now?” While we all want to think that we can make decisions true to our biomechanically-correct hearts and to our patients’ needs the truth may be closer to the research of Atun Gawande, MD:

“Three decades of neuropsychology research have shown us numerous ways in which human judgment, like memory and hearing, is prone to systematic mistakes. The mind overestimates vivid dangers, falls into ruts, and manages multiple pieces of data poorly. It is unduly swayed by desire and emotion and even the time of day. It is affected by the order in which information is presented and how problems are framed.

And if we doctors believed that, with all our training and experience, we escape such fallibilities, the notion was dashed when researchers put us under the microscope.”

Why do Doctors Distrust Clinical Prediction Rules?

Medicine is a profession of people caring for people. For all the science, statistics, dollars and debates that go on about who should get it and who should pay for it the essential interaction in medicine is the face-to-face encounter with the patient in front of the doctor saying, “I need help”.

Help is what we provide be it in counseling, coaching, supportive care, surgery, physical therapy, pharmacologic agents, mental health services or end-of-life hospice care. Help is an emotional word.

The problem, as I see it, is that treatment-based classification is seen as a completely emotionless, distant process:

1. assess pre-test probability (example: stabilization sub-group 33%)
2. measure predictor variables
3. apply treatment
4. Bye bye

Chris Anderson, in The Long Tail, describes probabilistic systems, like Wikipedia, Google and classification predictor rules as…

“…operating on the alien logic of probabilistic statistics - a matter of likelihood rather than certainty.

But our brains aren’t wired to think in terms of statistics and probability. We want to know whether an encyclopedia entry is right or wrong. We want to know that there’s a wise hand (ideally human) guiding Google’s results.

We want to trust what we read.”

Anderson describes a phenomenon that is evident among clinical physical therapists and physicians who resist “cookbook’ medicine wherein algorithmic simplicity seems to disregard the face-to-face complexity that characterizes (and sometimes confuses) clinical care.

The Crisis of Complexity

The Second Ecumenical Council of the Vatican, or Vatican II, opened under Pope John XXIII on October 11, 1962 and closed under Pope Paul VI on December 8, 1965.

Vatican II was destined to become known as the “opening of windows” when the Church would become open to all. Masses would now be said in native tongues instead of Latin, laws proscribing rewards and punishment rescinded in favor of ‘flexibility’ and ‘participation’. Catholicism became ’relative’ to one’s personal desires.

Attendance at Catholic Masses dropped 50% between 1964 and 1981. The Church was losing its influence.

Al Ries and Jack Trout were public relations consultants working in Atlanta, Georgia at the time when they were contacted by a group of lay persons who were concerned about the Catholic Church’s position in the minds of the people.

The laity asked Ries and Trout to help them answer this simple question…

“What is the role of the Catholic Church in the modern world?”

The Church had thought that their role was to be the ‘Teacher of the Law' while Ries and Trout found that the people (and, indeed, scripture) found that the role of the church was to be the ‘Teacher of the Word’.

‘Word’ vs. ‘Law’

Seems like such a simple, obvious distinction, doesn’t it?

Ries and Trout believe that simple positioning statements best clarify organizations and ideas in the minds of people. However, church leaders seemed to prefer their role as teachers of the Law:

“Experience has shown that a positioning exercise is a search for the obvious…

...The human mind tends to admire the complicated and dismiss the obvious as being too simplistic”

Did the Catholic Church change to become teachers of the Word? Not according to Trout and Ries. What happened to the Church since 1981? The results seem evident.

Classification predictor rules, despite their statistical challenge, promise to simplify clinical physical therapy decision-making in a way that challenges the established clinical order.

The Skilled Process of Classification Prediction Rules

‘Skilled’ physical therapy segued from inpatient facilities to outpatient settings in the early part of the 21st century as one of the answers to the problem of yearly 30% increases in ‘per beneficiary’ costs. Each outpatient physical therapy Medicare patient was increasingly consuming more and more services.

In order to control costs, Medicare would more stringently monitor the 'process' of outpatient pysical therapy care.

Skilled physical therapy initially meant ‘safety’ in the provision of the following:

1. movement
2. ambulation
3. range-of-motion
4. bed mobility
5. pressure relief

...in settings where patients with multiple, chronic and progressive conditions might not be expected to show ‘progress’. When the ‘skilled’ concept began to be applied to outpatient physical therapy it came with adjunctive requirements of supervision like the following:

• “8-minute rule”
• timed vs. un-timed codes
• ‘59’ modifiers

Medicare auditors in outpatient physical therapy now look for safety and progression in the therapists’ note as evidence of ‘skilled’ care. One of Medicare’s definitions of ‘skilled care’ (there are more than one):

“Services must… require, for example, the expertise, knowledge, clinical judgment, decision making and abilities of a therapist that assistants, qualified personnel, caretakers or the patient cannot provide independently.

A clinician may not merely supervise, but must apply the skills of a therapist by actively participating in the treatment of the patient during each Progress Report Period.

In addition, a therapist’s skills may be documented, for example, by

• the clinician’s descriptions of their skilled treatment
• the changes made to the treatment due to a clinician’s assessment of the patient’s needs on a particular treatment day
• changes due to progress the clinician judged sufficient to modify the treatment toward the next more complex or difficult task.

A therapist’s skill may also be required for safety reasons…”

Is Classification a Skill ?

Let’s look again at my brief tongue-in-cheek vignette on how to apply CPR’s:

1. assess pre-test probability (example: stabilization sub-group 33%)
2. measure predictor variables
3. apply treatment

Measuring predictor variables is not new and one could argue that measurement is a skilled component of the physical therapy service. Most of the predictor variables are self-report or historical findings ‘mashed up’ to form the CPR.

Applying treatment may or may not be skilled, depending on the treatment and the degree of repetition the Medicare auditor finds in your chart. Remember, Medicare expects the physical therapist to document ‘progression’ in the note as a means of establishing ‘skill’.

We are not yet paid for outcomes so we will still be judged on the process by which we provide care.

I was once called by a physical therapist undergoing a Medicare audit asking for help – she couldn’t understand why she was being audited when she spent nearly one hour (and charged 4 units) with each patient, giving massage and ‘myofascial release’, because they were ‘so much worse off’ than the ‘average’ Medicare patient. Her services were repetitive in nature, not restorative. Her argument was that she needed to provide so much massage in order to get her patient ‘ready’ to participate in restorative exercise.

Her patient population was unique but not in the way that she described it.

Because each physical therapists patient population is unique I believe there is an opportunity to provide a skilled service unique to each physical therapist: estimation of the size of the pre-test probability.

For example: the derivation study for the lumbar stabilization CPR had a pre-test probability of 33%.

Evidence-based medicine precepts would have me compare my patient population to the study population described by Hicks before applying the stabilization CPR.

Hicks’ population had the following:
1. Mean age of 42 years
2. 57% female
3. 41 days duration of symptoms
4. 46% had distal symptoms
5. 33% had a sudden onset
6. Average pain (VAS) 4.5 points.
7. Baseline Oswestry scale 30 points
8. Baseline Fear Avoidance Beliefs – Work 14 points
9. Baseline Fear Avoidance Beliefs – Physical Activity 15 points

My population has the following:
1. Mean age of 57 years
2. 61% female
3. 18 days duration of symptoms
4. 55% had distal symptoms (higher rate of compressive dysfunction)
5. 23% had a sudden onset
6. Average pain (VAS) 5.5 points.
7. Baseline Oswestry scale 42 points (greater disability)
8. Baseline Fear Avoidance Beliefs – Work 7 points (retired persons don’t work)
9. Baseline Fear Avoidance Beliefs – Physical Activity 18 points

Unique factors about my population may also include the following:
• Retirement community
• Southern climate (lots of flip-flops and sunburns)
• 60% Medicare (lower out-of-pocket but fixed income)
• Second-highest rate of spinal fusion surgery spending in the United States (Bradenton-Sarasota MSA)

Let’s say I get a patient with 3/4 predictor variables for success with lumbar spinal stabilization. Hicks’ algorithm gives a 4.0 positive likelihood ratio (+LR) which implies about a 30% upward shift in probability favoring stabilization – assuming my population is exactly like his.

But, I think my population is different: older, stiffer, has a higher incidence of spinal compressive disorders, more disability and more co-morbidities. I think I should adjust my pre-test probability of stabilization responders downward – but by how much?

There are two ways to find the pre-test probability for your population. The first way is to adjust published estimates based on heuristics, or ‘rules-of-thumb’. My population is about 1.5 times older than Hicks’ population and correspondingly stiffer so I could adjust Hicks’ pre-test probability down by the inverse of 1.5 (67%).

The other way is to generate a 2x2 table, test everybody, apply the CPR and assess the responders and non-responders. You will be able to calculate your own values for sensitivity, specificity, pre-test probability, post-test probability, positive and negative likelihood ratios.

Since I have a full caseload and little time for outside research I will use the heuristic adjustment to Hicks’ rule:

33% x 0.67 = 22% adjusted pre-test probability

¾ predictor variables in Hicks' classification has a +LR = 4.0 which shifts probability upward approximately 30% .

30% + 22% = 52% adjusted post-test probability

…which is little better than chance. I could do about as good flipping coins and allocating patients to lumbar stabilization or some other treatment.

So, Is This Skilled Physical Therapy?

Whether or not I choose to use stabilization I think that using Hick’s rule implies that, yes, my care is skilled.

In outpatient care our only competitive advantage is that we are closer to and more in contact with our patients than in other settings, which tend to be larger, institutionalized settings. We need to know our patients better than anyone else.

Steven McGee, MD argues convincingly that the use of likelihood ratios requires greater patient contact and understanding, not less:

“…because the best estimate of pretest probability incorporates information from the clinicians’ own practice – how specific underlying diseases, risks and exposures make disease more or less likely – the practice of evidence-based medicine is never “cookbook”.”

We measure what can be measured in order to quantify the impact on the pre-test probability. Physical therapists need a ‘culture of measurement’ in order to routinely collect and understand these data.