Treatment Based Classification: Where Does it Fit In?

A recent webinar from the Healthcare Information Management Systems Society (HiMSS) called Improving Care with Predictive Analytics: The Next Generation of Clinical Decision Support offered this nice schematic of where decision rules fit into the "black box" that we call clinical decision making.

I've circled in red the "Rule Based" decision format used by TBC

CDS = clinical decision support, ANN = artifical neural network

Individual, idiosyncratic clinician decision making is not pictured in this schematic - just computer based.

I'm hopeful that, going forward, that these types of decision support will continue to expand to incorporate more of physical therapist practice.

I believe that handing off some of the automatic functions, such as documentation, will enable physical therapists to spend more time face-to-face with their patients.

My new book, pictured in the upper left hand corner of this page, describes the current state of Clinical Decision Support in  physical therapy and where we can go from here.

New Treatment Based Classification Chart with Hyperlinks

These are the Treatment Based Classification rules I've been using the past couple of years. Please use this chart, including the hyperlinks. Let me know if you see any ommissions or if you recommend a rule I have not included.

Rule	Author	Year	Level of Rule Development
Cancer Rule in Lower Back Pain Patients	Joines et al	2001	Cost Effectiveness study
Lumbar Manipulation Rule	Flynn et al	2002	Broad Validation
Lumbar Stabilization Rule	Hicks et al	2005	Derivation
Lumbar Traction Rule	Fritz et al	2007	Expert Consensus
Thoracic Manipulation Rule for Neck Pain	Cleland et al	2007	Broad Validation by Cleland in 2010 demonstrated NO rule is needed - ALL neck pain patients without "red flags" get manipulation
Ankle Mobilization for Inversion Injuries	Whitman et al	2007	Derivation
Hip Manipulation Rule for Knee Pain	Currier et al	2007	Derivation
Specific Directional Exercise for Lower Back Pain	Browder et al	2007	Expert Consensus
Lumbopelvic Manipulation for Anterior Knee Pain	Iverson et al	2008	Derivation
Cervical Traction Rule	Raney et al	2009	Derivation
Thoracic Manipulation Rule for Shoulder Pain	Mintken et al	2010	Derivation

This chart is regularly updated at www.BulletprootPT.com.

What is the role of the Physical Therapist's Diagnosis?

Another take on the problem of regional variation, excessive diagnostic testing and paying for quality...

A new study in the May 12th New England Journal of Medicine revealed a 'striking' new finding: that Medicare patients who move to high-intensity healthcare regions in the USA get diagnosed with more diseases, appear sicker and, ultimately, receive more invasive testing and diagnostic imaging services.

Excessive diagnostic testing is a problem because it increases costs and leads to unnecessary procedures.

Going forward, Medicare's Pay for Performance scheme will morph into a true outcomes-based health care system and will rely on diagnostic 'risk adjustments' to pay hospitals, doctors and physical therapists that treat sicker patients.

Some of these 'sicker' patients just appear sick because doctors are diagnosing too much - because they are paid to do so.

An example given in Dr. Maggie Mahar's The Health Care Blog of Regional Variations in Diagnostic Practices by Dr. Yunjie Song et al of men diagnosed with prostate cancer reveals that most of these men will never experience symptoms and will never need advanced cancer treatment (eg: surgery).

Subsequent outcomes measurement of treatments given to these men will reflect their cancer diagnosis and will increase the hospital's 'risk adjusted' payment - even though these men may never get 'sick' from cancer.

Physical Therapy Diagnosis

In physical therapy, many treatments are performed (eg: electric stimulation, ultrasound, diathermy or heat/ice) based on a pattern recognition approach.

Pattern recognition is efficient but subject to cognitive bias such as the following:

• Anchoring - you believe that ultrasound is an appropriate treatment for your current patient because you were taught this approach in PT school.
• Confirmation bias - you believe that ultrasound is the appropriate treatment for your current patient even though your patient mentions that they have used this modality previously to NO effect.
• Availability heuristic - you believe your current patient is a candidate for ultrasound because your last patient got better with ultrasound and she "just loves" ultrasound.

The alternative to pattern recognition is a 'rational analysis' approach - in Physical Therapy the best example of rational analysis is Treatment Based Classification (TBC) - where the physical therapist measures several, fundamental patient characteristics and 'decides' on a particular treatment approach.

An "ultrasound" decision rule could identify who WILL NOT benefit from ultrasound. Alas, no such rule exists.

No TBC rules exist for other vague, high-frequency conditions seen in physical therapy either, such as:

• future falls risk,
• general conditioning and
• pain control.

Physicians testing for pathology usually involves invasive, irradiating or expensive procedures which begs the question in medicine: Is more testing better? Does a medical diagnosis improve your chances? Does more healthcare lead to better health?

Diagnosis or Prognosis?

Fortunately, physical therapists already have tools to assess factors that contribute to outcome - and the authors of the diagnostic variation study seem to be aware of the importance of these tools...

“...measures of health risks reported by patients (e.g., smoking and exercise patterns) and functional status (physical, social, and role function) could be incorporated in risk-adjustment models.”

It seems likely that physical therapists, unlike physicians, could benefit their patients by additional testing, measurement and diagnosis of patient characteristics that contribute to outcome.

It also seems likely that the new role of the physical therapist under an outcomes-based healthcare system will be that of diagnostician: searching for risk factors that predict the outcome of an episode of physical therapy care, rather than primarily delivering interventions.

Is it possible that your search for diagnoses will be rewarded similar to, but more consistently, to the way physicians are rewarded for diagnosing today? And if diagnostic variations can be 'evened out' can we assume that this new system will be equitably rewarding for the patient?

PT Managers can now use TBC to set Long Term Goals

Physical therapy managers are always trying to find ways to:

• reduce paperwork (and spend more face-time with your patient)
• speed up productivity (and make more money)
• lower the costs of care (and make more money)
• see more patients (and keep your job)
• get the patients better (and get good outcomes)
• get the right stuff in the chart (for Medicare compliance)

Treatment based classification (TBC) uses a parsimonious (brief) set of tests to identify who will get better with which treatment.

For example, the Cervical Traction classification uses these tests (predictor variables) to identify patients who will respond well to cervical traction:

• Peripheralization with the C4-7 mobility test
• Age> 55
• Positive shoulder abduction test
• Positive cervical distraction test
• Positive Median nerve tension test A

The modifiable findings are in bold. Age is not modifiable. The others may change as a result of your care. The test results are expressed as percentages:

• The baseline chance of being in the cervical traction group is 44%.


• Three-of-five (3/5) positive tests implies a 79.2% chance of success with cervical traction.


• Four-of-five (4/5) positive tests implies a 94.8% chance of success with cervical traction.


• Five-of-five (5/5) positive tests implies a 100% chance of success with cervical traction.

What if we set up our goal in quantitative terms? Reduce the risk of needing cervical traction to the baseline level of 44%.

The long term goal would be written like this:

"Improve cervical traction risk classification from 94.8% to 44% in four weeks"

The goal is objective, quantifiable and easy to implement.

If you use TBC Templates, available here, you can check your patient's status with simple check boxes weekly.

Attach the template to your electronic documentation and eliminate lengthy, narrative descriptions of 'skilled therapy'.

Treatment based classification holds the promise of better physical therapy outcomes at lower cost.

Now, you can use templated (electronic or paper) TBC checklists to reduce your paperwork and speed up your productivity, too.

For Physical Therapists: How NOT to think like a turkey

Do you want to avoid thinking like a turkey?

To NOT think like a turkey, first consider HOW turkey's think...

It's the day after Thanksgiving and a newly hatched turkey has arrived.

He has the coop to himself with plenty of room and no competition for food, etc.

Then, some humans arrive at feeding time. They provide food and water.

After that, they turn the lights down low, turn up the bed warmers and play some soft music.

"My," observes the turkey, "these humans must really love me - they must really be my friends!"

The turkey observes this same process for 364 more days.

Each day is an independent observation that seems to confirm the initial observation - that humans are friendly benefactors of turkeys and would never do anything to harm a turkey.

Then comes Thanksgiving Day and the humans arrive at feeding time - only instead of food they carry hatchets and axes.

"I wonder what those are for?" thinks the turkey.

Needless to say, the turkey dies with a very surprised look on his face.

The Outcome is Dinner

Now, what did the turkey do wrong?

How could the turkey have come to a different conclusion about humans presented with the same set of facts?

What could the turkey have done differently to change the outcome?

The Fallacy of Inductive Logic

The turkey committed the fallacy of inductive logic, first described in 1748 by David Hume, a Scottish philosopher, economist and historian.

Hume's contribution to empirical thought is an important cornerstone of the scientific method and to evidence-based medicine.

"Inductive inference is reasoning from the observed behaviour of objects to their behaviour when unobserved...it is a question of how things behave when they go (in Hume's words)...

"beyond the present testimony of the senses, and the records of our memory."

...we tend to believe that things behave in a regular manner; i.e., that patterns in the behaviour of objects will persist into the future..."

Hume's story uses chickens, not turkeys.

I have adapted Hume's story because turkeys resonate with Americans at this time of the year (Thanksgiving/Christmas).

Turkeys, chickens, David Hume and physical therapists are all searching for one universal constant - the TRUTH.

How can we find it?

For the physical therapist, TRUTH is the answer to these questions:

• What treatment will make my patient better?
• How long will the treatment take?
• How much better will my patient get?
• How much will the treatment cost?

In Hume's time the problem of induction was seen as a barrier to finding the truth. In other words, inductive reasoning (thinking like turkeys/chickens) did not seem to be the path to secular enlightenment (ie: the TRUTH).

Today, the foundation of the scientific method is hypothesis formation, usually by inductive reasoning.

The hypothetico-deductive model (thinking like physicians think) uses inductive reasoning to generate the initial hypothesis, and then to test the hypothesis:

1. Experience: What previous treatments have worked?


2. Why did the previous treatments work? What patient characteristics are unique to the responders (note: this example specifically describes Treatment Based Classification, or TBC).


3. Identify the patient characteristics (from the history or physical exam) that best predict which treatments worked (the outcome).


4. Test your theory.

What could the turkey have done different?

Find one human that likes to eat turkey. Then, the turkey could change his behavior to change the outcome (on Thanksgiving Day, run like hell).

In science, we question our assumptions - we test our theories and we learn from our failures. We're skeptics.

In physical therapy, we call this reflection.

How to Use Fear Avoidance Beliefs in your Physical Therapy Plan of Care

We recently held a Treatment Based Classification Seminar (TBC) at our physical therapy clinic in Palmetto, Florida.

Most of the physical therapists attending (14) had a fairly good awareness of TBC but one item stood out - Fear Avoidance Beliefs (FAB).

None of the attending therapists used even the FAB 'clinical shortcut' in their evaluation:

“I should not do physical activities which (might) make my pain worse.”

The clinical shortcut identifies elevated fear-avoidance beliefs early, so you could treat them. A number of seminar participants expressed interest in how physical therapists could screen for and treat patients' Fear Avoidance Beliefs in their plan of care.

• What are Fear Avoidance Beliefs and Behaviors?
• What techniques should physical therapists use when these findings are present?
• How can we help these people or should they be referred to medical providers?

What are Fear Avoidance Beliefs?

Fear Avoidance Beliefs are one type of psychosocial factors that include:

• depression,
• anxiety and
• job dissatisfaction,

...among other factors.

Persons experiencing an episode of acute pain are believed to manage the episode by 'confronting' the pain or by 'avoiding' the pain.

Persons who exhibit these 'avoidance' beliefs and behaviors are significantly more likely to experience disablement as a result of their acute pain.

How to Use Fear-Avoidance Beliefs in the Plan of Care?

1. Cognitive Behavioral Therapy (CBT)

Cognitive Behavioral Therapy (CBT) is a cool new tool that already fits the physical therapist skill set - although many of us may need additional training.

Some physical therapists may feel unprepared to render CBT but I would argue that the physical therapist is already well-suited to learn about CBT - it should complement our current 'toolbox'.

Cognitive behavioral therapy, within the context of our current practice, could be described as follows:

"Effective patient education by physical therapists appears to depend on the use of effective brief psycho-educational strategies that can address the cognitive and affective processes that motivate pain-related activity avoidance."

In other words, some of the same persuasive, coaxing, gentle, positive encouragement that most of us have used our entire careers to get patients more active.

Specific examples of some CBT techniques are as follows:

• keeping a diary of significant events and associated feelings, thoughts and behaviors;
• questioning and testing cognitions, assumptions, evaluations and beliefs that might be unhelpful and unrealistic;
• gradually facing activities which may have been avoided; and
• trying out new ways of behaving and reacting.
• Relaxation,
• mindfulness and
• distraction techniques are also commonly included.

George et al distinguished between the 'typical' educational approach of biomedical education and a fear-avoidance model of self-management.




2. Use FAB to screen for modalities

Childs et al found that use of electrotherapeutic or ultrasound modalities may encourage patients with elevated FAB to focus on their pain, avoid active 'confrontation' behaviors and lead to decreased outcomes.

Studies in America, Israel and the Netherlands tend to support the findings of poorer functional outcomes when modalities are used in the plan of care.

There is a significant chance that Medicare will decrease the relative value of modalities such as ultrasound or e-stim - even 'bundling' these modalities with other Common Procedural Terminology (CPT) codes, such as exercise, based on a lack of efficacy or effectiveness of modalities.


3. Predictor variable in TBC
   
   
   • Flynn's lumbar manipulation
   • Child's thoracic manipulation and
   • Hicks' lumbar stabilization rules
   
   ...all use the Fear Avoidance Beliefs Questionnaire as a predictor variable - lower levels of fear avoidance behavior generally predict successful outcomes. Hicks' stabilization rule is the exception - higher levels of fear-avoidance beliefs predicts success in this group.
   


4. Risk factor under a Medicare alternative payment system.
   
   Between 12-15% of the variation in the outcome of industrial workers' injuries was due to psychosocial factors, like FAB, not the physical or personal factors that physical therapists typically measure - like straight leg raise and Manual Muscle Testing (ugh).
   
   Since the payment by Medicare under an alternative payment system is likely to be a 'case rate' - say, $800 for 10 visits - anybody over the 10th visit is a financial risk to the provider.
   
   
   Physical therapists will need better measurement tools to identify these people, these outliers, early and perhaps apply for extra dollars.
   
   Screening for 'outliers' under an alternative payment system to Medicare Fee for Service will require sensitive tests to predict who is likely to need 20 visits for LBP, not 10 visits.

Why TBC is not "Cookbook" medicine

I got it wrong!

This is first correction on Physical Therapy Diagnosis (that I'll admit to) - so don't go telling my wife or my mom!

I posted three days ago that a physical therapist could use a heuristic adjustment (a 'rule-of-thumb') to published estimates of pre-test probabilities for treatment based classification (TBC) groups, for example:

• Stabilization = 33%
• Lumbar Manipulation = 45%
• Thoracic Manipulation for Neck Pain = 54%


The example I used was of my own patient population here on the West Coast of Florida whose demographics I posted here.

My patients were measurably different from published norms for a stabilization group in the TBC derivation study.

I mistakenly recommended an downward adjustment to the pre-test probability based on age. To check my adjustment I e-mailed Dr. Steven McGee, author of Evidence Based Physical Diagnosis, who e-mailed back with the following:

"The only way to adjust published pretest probability is to measure your own clinical experience.

For example, in the last 100 patients you have seen with shoulder pain, how many have had rotator cuff disease?

This is the pretest probability figure to which you would apply Likelihood Ratios of diagnostic tests."

Steve McGee



Dr. McGee's book recommends a simple "bedside" approach to using likelihood ratios in the clinic that can assist physical therapists learning about TBC.

No calculators.

No nomograms.

Just you and the patient and a few simple tests and measures.

Dr. McGee's Mnemonic

Dr. McGee does a great job in his book but I'll try here to describe the clinical mnemonic:

Memorize 2,5 and 10: these are positive likelihood ratios. The approximate associated upward shifts in post-test probability are +15%, +30% and +45%.

Memorize 1/2, 1/5 and 1/10 (notice the pattern?): these are negative likelihood ratios. The approximate associated downward shifts in post-test probability are -15%, -30% and -45%.

Measuring Treatment Responders for 'Functional' Diagnoses

To measure the pre-test probability of an outcome is the same as a diagnosis. Outcomes what we are trying to improve in physical therapy TBC.

First, test every patient with lower back pain using your stabilization predictor tests:

1. Age less than 40
2. SLR > 90 degrees
3. Positive Prone Instability Test
4. Positive Gower's sign

Who is positive on the rule?

For example, The next 100 patients in my clinic who present with lower back pain who meet the stabilization inclusion criteria get tested as described above.

Those with at least 3 positive tests are my treatment responders - let's say 20 patients score at least 3 positive tests.

The pre-test probability is 20/100 = 20%.

At the Bedside

For a patient with 3 or more positive tests the positive likelihood ratio is 4.0. According to Dr. Magee's mnemonic the upward shift is between 15% and 30% (about 25%). Add that to my pre-test probability...

20% + 25% = 45%


Stabilization is no better than some randomly chosen clinical intervention for improving this patient's lower back pain. We should try to find an alternative intervention.

By using our own patients'characteristics along with published estimates of likelihood ratios we can "personalize" TBC. As Dr. McGee says in his book...

"...because the best estimate of pretest probability incorporates information from the clinician's own practice - how specific underlying diseases, risks and exposures make diseases more or less likely - the practice of evidence-based medicine is never "cookbook"."

Dr. McGee's paper on "Simplifying Likelihood Ratios" is a great way to get started using this valuable tool.

An excellent tutorial on physical therapy diagnosis and the use of pre-test probability is available from the University of Pittsburg and Dr. Julie Fritz.