5 EBP Numbers You Should Discuss With Your Patients

A central tenant of evidence-based practice is to incorporate the patient’s values and beliefs into treatment plan creation. For most clinical diagnostic tests used by PTs and ATs, it isn’t crucial to explain the sensitivity and specificity of the tests. However, there are aspects of the proposed intervention that the patient needs to understand to be a true partner in the plan of care. Here are 5 EBP numbers you might want to discuss with your patients.

1. Absolute Risk Reduction (ARR)

Absolute Risk Reduction is a basic statistic that can be calculated from a well-conducted randomized clinical trial. Assuming the authors provide the proportion of patients in each group (the control and experimental) who either achieved a desirable outcome or failed to achieve it, it’s calculated as follows:

ARR = CER – EER

Where:

• The Control Event Rate (CER) is the number of patients exposed to the control intervention who failed to achieve the desirable outcome divided by the total number of patients exposed to the control intervention.
• The Experimental Event Rate (EER) is the number of patients exposed to the experimental intervention who failed to achieve the desirable outcome divided by the total number of patients exposed to the experimental intervention.

If the experimental intervention was successful, then the CER will be greater than the EER.

2. Number Needed to Treat (NNT)

The ARR itself is not particularly illuminating, but its reciprocal – NNT – can be useful in talking with patients.

NNT = 1/ARR.

Fundamentally, the NNT answers the question:

“How many patients would need to be exposed to the experimental treatment instead of the control treatment to have one additional positive outcome?”

Suppose the CER is 0.44 and the EER is 0.18. The ARR is then 0.26, and 1/0.26 = 3.8. It is recommended to avoid fractions in the NNT, so the 3.8 should be rounded up to 4 (always round up NNT to the next whole number to avoid excess optimism). So, in this case, for every four patients treated with the experimental intervention, there will be one additional good outcome.

3. Absolute Risk Increase (ARI)

A patient may not only fail to benefit from a particular intervention, they may also experience harm from it. Harm is unlikely in most physical rehabilitation settings, but patients should be fully informed.

When considering adverse events, a statistic conceptually similar to ARR can be calculated from the proportion of patients in each group who experience the adverse event. Clear as mud? Consider this example:

Suppose out of 50 patients in the control group, four experience an adverse event, giving a CER of 4/50 = 0.08. Of 70 patients in the experimental group, nine experience the same adverse event, yielding an EER of 9/70 = 0.13. Subtracting the EER from the CER gives a value of -0.05.

When the EER is greater than the CER, the resulting negative value is referred to as the Absolute Risk Increase.

ARI = CER – EER (calculated for adverse events)

4. Number Needed to Harm (NNH)

The reciprocal of the absolute value of the ARI (i.e., disregarding the negative sign) is then the Number Needed to Harm.

NNH = 1/ | ARI |

In the above case, the NNH is 20 (from 1/.05). For every 20 patients treated with the experimental intervention, there will be one additional adverse event.

When the NNH results in a fractional value, it should be rounded to the nearest smaller whole number, again to avoid an overly optimistic view.

5. Likelihood of Being Helped or Harmed (LHH)

The number needed to treat and the number needed to harm can be combined into a single value, the Likelihood of Being Helped or Harmed. The LHH merges the NNT and the NNH into a single value to illustrate both the potential harm and benefit from the intervention. The LHH is the ratio of the reciprocals of the NNT and NNH, so the formula is:

LHH = (1/NNT)/(1/NNH)

Using the examples above, the LHH is .25/.05 = 5. Therefore, the patient is 5-times more likely to have a good outcome from the intervention than to be harmed by it.

Put into Practice

One more example:

Suppose a novel (experimental) rehabilitation program for patients who have undergone a rotator cuff repair results in 12 of 32 patients in the control group failing to achieve full function, and 5 of 41 patients in the experimental group failing to achieve full function. This gives an NNT of 4.

Of the 32 patients in the control group, 4 experienced a recurrent tear of the rotator cuff during the rehabilitation, and of the 41 patients in the experimental group, 8 experienced a recurrent tear. This gives an NNH of 14.

Divide NNH by NNT, and we get an LHH of 3.5. So, the patient is 3.5-fold more likely to gain full function than to experience a recurrent tear if the experimental intervention is chosen.

An important aspect of failing to achieve a good outcome or experiencing harm from an intervention is the magnitude of the result. An NNH of 20 if the harm is paralysis due to a spinal cord injury is very different from an NNH of 20 if the harm is moderate delayed-onset muscle soreness. There are methods to adjust the LHH given the severity of various outcomes, but the NNT, NNH and LHH are good starting points to help patients become true participants in their care.