What is a person’s ‘Rehabilitation Potential’? Often this question is asked by one clinician of another about a patient. Still, there is a second interpretation, “What do we mean by Rehabilitation Potential when we ask that question?” I will discuss these questions, but before doing so, I will consider two further crucial questions, “Can we, based on any evidence, predict or know whether a patient will benefit from the involvement of a specific rehabilitation team in contrast to not being involved with the team?” and “Can we predict who will benefit from a specific rehabilitation intervention?”
Table of Contents
Context - evidence
Healthcare relies heavily on randomised trials when considering what interventions or strategies are effective. In such trials, the design compares the outcomes between two (or more) groups of patients, where the groups received different interventions or differing amounts of a single intervention.
The analysis relies on summary data from each group. Consequently, if there is a difference unlikely to be due to chance, one may draw the valid conclusion that the probability of a better outcome will be increased for any similar patients receiving the intervention as again receiving the comparator intervention. However, one cannot know whether the improvement was distributed across all patients with a statistically normal distribution or whether some patients improved significantly and others did not.
Occasionally one may identify clear-cut factors that predict who will or will not benefit. Most examples probably arise from other basic research on specific biological processes involved in the treatment. For example, some chemotherapy agents for cancer only help if the patient has a particular gene. In the same way, if the bacterium causing an infection is known to be resistant to an antibiotic, then using that antibiotic is unlikely to help.
In a few circumstances, evidence supports particular treatment combinations; for example, this applies to chemotherapy for acute leukaemia and antimicrobial therapy for tuberculosis.
In other words, most evidence supporting the effectiveness of healthcare treatments only allows a probabilistic approach to the population of suitable patients. Identifying individual patients from within the population known to benefit is only possible when basic research on the treatment identifies a biologically likely marker. Further research confirms that putative markers predict their validity.
Context - rehabilitation
There is good evidence that rehabilitation benefits almost any population of patients, and many individual treatments are supported by good evidence. Thus, rehabilitation is in a similar position to many other healthcare treatments. There is also some early research into mechanisms underlying a few specific interventions, but not with the specificity needed to identify likely factors predicting benefit for a subset of patients.
In addition to the lack of sufficiently detailed research able to identify specific markers that might predict who will or, equally importantly, who will not benefit from rehabilitation, many other characteristics make it improbable that we can develop evidence-based criteria for patients who will benefit.
First, rehabilitation is based on a multi-professional team giving many interventions, many of which will interact or be mutually interdependent. Many beneficial interactions may be unknown or not recognised; for example, providing information on prognosis and empathetic support while delivering a specific therapy may be more effective than the therapy.
Second, the outcome spans several domains such as activities, pain, distress, and social interactions. The most important effects are specific to the patient, difficult to measure, and not easily categorised. Many factors will affect most outcomes, not just a single treatment.
In other words, rehabilitation is a complex intervention applied to a complex situation, with many factors working through different processes underlying change and benefit. It is improbable that research will ever identify specific factors that will determine with certainty who will benefit from a rehabilitation intervention or overall rehabilitation.
Context - predictive value
The predictive value of any prognostic factor is a third vital consideration. The calculations are illustrated here. Several crucial matters must be remembered.
The characteristics of a test – its sensitivity and specificity – depend on the underlying frequency of the actual cases. A test will be quite sensitive if the population has a reasonable proportion of cases but very insensitive if there are few cases. In other words, you can only use a validated test on a population identical to the population used to develop and validate the test. For example, a test designed for aphasia in people with left hemisphere stroke will be much less sensitive in people with traumatic brain injury.
Second, there is a trade-off between sensitivity and specificity. A test that detects almost all cases will also select many non-cases and vice versa.
Considering the assessment of the potential to benefit from rehabilitation, if the criterion or criteria are sent to ensure that everyone with potential is included, many people who will not benefit will also be included. Conversely, if the criteria are set so that only people who will benefit are included, many people who would benefit will be excluded.
There is no evidence that any criteria in any population can select a subpopulation of patients who will benefit from rehabilitation, with the remaining population failing to benefit from rehabilitation. This statement also applies to almost all specific rehabilitation treatments. Nonetheless, there are frequent references to predictive tests and selecting people with potential rehabilitation. What is the explanation?
One explanation is that some people confuse the prediction of outcome with the prediction of benefit (from rehabilitation). The effect is to select people who will have a better outcome. In the uncontrolled context of clinical practice, it is incorrectly interpreted as showing the patients have benefitted from the rehabilitation service.
Most studies of recovery take all patients receiving a range of different rehabilitation interventions and differing amounts (including, sometimes, none), and they investigate the relationship between initial factors and the change in a measure of function or other outcomes such as returning home. They do not allow any conclusion about rehabilitation.
A few prognosis studies involve people in trials, but these usually combine both groups when looking for factors predictive of change.
A variation on this theme uses the categorisation of change seen during rehabilitation to identify patients who change considerably and patients who change very little. Usually, the term used is ‘responder’, meaning a person who changes significantly during rehabilitation and is assumed to be someone who has benefited.
This conclusion is invalid. The study is simply identifying factors associated with a better outcome. The same elements will be identified in a study of people who do not receive rehabilitation.
To find factors predicting the ability to benefit from rehabilitation, you must take the following steps.
First, you need to decide what two interventions you will compare. What will be the alternative if one is a complete inpatient rehabilitation service from a multidisciplinary team? This might be “whatever else is available”, which has the advantage of being realistic (i.e. what happens now) but would be very difficult to conduct as few patients would agree to participate even if an ethical committee passed it.
This initial step will also require a definition of the relevant patient population, such as all people referred to a service or all patients with disability seven days after stroke. One must remember that the findings will only apply to the population defined.
Next, select a few items that you think predict that a patient will do better receiving the rehabilitation on offer than receiving the alternative service(s). This is difficult when there is no evidence, but a selection of some criteria used at present would be reasonable.
This step will also require you to specify the outcome of interest. This might be pretty restricted, such as independence in personal activities of daily living or level of mobility, or general, such as quality of life or living outside residential care.
Third, use the criteria to categorise all recruited patients as having or not having potential. Then, and only then, randomise the patients in each group to the rehabilitation group or alternative group.
Next, collect the outcome data at an appropriate time, usually months later.
Finally, test your hypothesis: in the group receiving rehabilitation, the outcome will be better in the group with potential; in the group receiving the alternative, the outcome in the group with potential will be less good than the outcome in the group with potential who did receive rehabilitation; the effect in the group without potential will be equivalent in both groups.
To achieve adequate power, you are likely to need over 1000 patients.
I advise against trying this because
- it won’t be easy to identify possible predictive factors and would take several iterations to reach a better set of criteria
- it will be effectively impossible to conduct the research for practical reasons
- the results will only apply to the selected population, the selected alternative, and the selected outcome.
I have discussed this problem at least twice, once when using criteria to select patients for rehabilitation and once when considering the prediction of benefit. In each case, I proposed the same solution as I am setting out here.
An axiom is “a statement or proposition which is regarded as being established, accepted, or self-evidently true” [OED], and I will give mine here.
Axiom 1. The benefit includes all possible outcomes.
From a patient’s perspective, the benefit covers any aspect of their life that the intervention improves. The benefit is not restricted to improvement in capability in or performance of activities.
Axiom 2. Rehabilitation may help anyone.
The evidence I have reviewed shows that any patient at any age with any malady (disease, disability, illness, sickness) and at any stage of that malady can benefit from rehabilitation given by a multi-professional team.
Axiom 3. No criteria can predict response.
The evidence given above suggests that no evidence-based criteria can identify with any certainty patients who will or will not benefit from rehabilitation.
If these axioms are accepted, it follows that a rehabilitation team should see all patients and that the team will then decide on what is best for the patient. To expand on this further, I wish to emphasise several points.
The question the person (or team) who sees the potential patient needs to ask is, “What is the best next step for this person, among the available alternatives?” It is essential that the person does not answer the question, “Is this person suitable for our service?” The latter question prioritises the service; the former prioritises the person. Services need to be person-centred.
In practice, this means that the person making the decision must know enough about the services available to the patients to decide which is most likely to benefit the person most. They must also be prepared to take the person on even if they are not a ‘perfect fit’ with their service.
The team involved should also be sufficiently confident to contemplate treatment trials. One cannot ever know that something will work, and they should try interventions that have a reasonable chance of benefiting the patient. This helps management because it forces the team and the patients to evaluate the effect of an intervention and, when doing this, to discuss what might happen if there is no benefit.
On a larger scale, the admission to the service (as an inpatient or outpatient) can be seen as a ‘trial of treatment’. However, the move to another service or place must be undertaken positively and planned, not simply as a matter of removing the patient.
Why the solution does not work (yet).
This solution does not work because there is no consistent and coherent organisation of services used by and available to people needing and receiving rehabilitation. I have emphasised this in several places for Long Covid, at least twice, and for Traumatic Brain Injury, and I have suggested a way forward that has had no effect (yet).
I draw the following conclusions. The concept of rehabilitation potential is fatally flawed because the evidence suggests that any person with a disability may benefit from rehabilitation. There is no evidence to allow rational, fair selection into rehabilitation services. All criteria reflect personal (or team) prejudice and are used by commissioners to justify rationing.
Second, the crucial underlying problem is the abject failure of the NHS at all levels to develop a coherent, consistent, coordinated and collaborative approach to rehabilitation. This arises from a failure to give any significant attention to rehabilitation, with an increasing tendency to fragment yet further existing services, for example, by developing separate services for Long Covid. The recent development of Integrated Care Systems seems a serious misnomer because they are not integrating rehabilitation services within health services or across health, social services, housing and education services.