The Numerical Rating Scale (NRS) (which many articles write as NPRS, Numerical Pain Rating Scale) is the pain scale most widely used in the world: the person is asked for a number from 0 to 10, where 0 is "no pain" and 10 the "worst pain imaginable". It is quick, requires no materials and produces a figure that can be compared over time.
The problem is not collecting the number: it is reading it. The same 6/10 can mean very different things depending on how the question was asked, when it was asked and who is answering. And above all: when that 6 becomes 4, does that drop of 2 points really matter, or is it noise? This page answers that question, which is the most useful of all and almost never the best explained.
What the NRS is and what it really measures
The NRS is a one-dimensional tool: it measures one thing, the intensity of pain. It does not measure how much that pain limits movement, how much it disturbs sleep, how much it stops you working or training. These are different dimensions, which need different tools to measure them.
In clinical practice, convenient bands are used to talk: 0 no pain, 1-3 mild, 4-6 moderate, 7-10 severe. They are useful conventions in conversation, not diagnostic thresholds: nobody decides a treatment because a patient said 4 rather than 3.
The added value of the NRS lies not in the single number but in the series: fifteen values collected in the same way and at the same time of day tell a trajectory. A single value tells you about a moment, and it is a noisy moment.
How it is administered: the exact question changes the answer
This is the point almost everyone skips. "How much pain do you have?" is not one question: it is at least four different questions, and they produce different numbers.
- The pain right now, at this precise moment, perhaps sitting and at rest.
- The average pain over the last 24 hours or the last week.
- The worst pain experienced over the reference period.
- The minimum pain, that is, how low it goes on good moments.
How much does this difference matter? A 2026 study of 1,007 adults with chronic low back pain collected both a single assessment at enrolment and real-time readings through an app, three times a day for seven days. At enrolment the average intensity was 5.4 (median 5). The median of the real-time values was 1, and 54% of participants had a median of 0 or 1.
This is not a contradiction, and it is a clean finding: they are two different questions. Asking "how much does your back hurt" in a questionnaire evokes the memory of the worst moments; asking "how much does it hurt now" also captures all the hours in which the pain is silent. If the physiotherapist asks the first question at the initial visit and the second at follow-ups, they will see an improvement that never happened.
The operational rule is therefore simple and binding: you choose one formulation and never change it, for the whole course. Same time frame, ideally same time of day. In the same study, peak pain fell in the evening for 30% of participants: measuring in the morning or after dinner is not the same thing.
The patient gives the number, not the observer
It seems obvious, but it is one of the most documented errors. In a study of 113 non-paralysed critically ill patients in intensive care, agreement within one point between the value reported by the patient and the one attributed by the nurses was 73%, but high scores were systematically underestimated: a value of 4 or higher was reported by 33% of patients and recorded by observers in only 18% of cases.
The context is intensive care, not the clinic, and it should not be transferred mechanically. The principle, however, holds everywhere: when the person can speak, they are the one who gives the number, because observers tend to underestimate precisely strong pain.
How many points of change really matter
Here lies the most useful piece of information about the scale, and it is the one missing from almost every popular page: the minimal clinically important difference (MCID, or MIC in some texts), that is, the smallest change the patient perceives as a real improvement in their life, rather than as fluctuation.
The historical reference is a 2001 study that analysed 2,724 participants in ten placebo-controlled trials on chronic pain: diabetic neuropathy, postherpetic neuralgia, chronic low back pain, fibromyalgia and osteoarthritis. By comparing the change in NRS with the global judgement of change expressed by the patients themselves, the authors concluded that, on average, a reduction of about 2 points or about 30% represented a clinically important change.
On a separate track, an international panel of experts on low back pain reached in 2008 a consensus proposal in line with this: 2 points on the 0-10 NRS, 15 points on the 0-100 VAS, and 30% from the starting value as a useful threshold when the baseline is taken into account. The same authors state explicitly that the proposed values are not the final answer, but a common starting point, and that in the literature the estimation methods varied widely.
Why the starting point changes the calculation
This is the detail that is almost always summarised badly, and here the meaning is reversed.
In the 2001 study, the relationship between percentage change and the improvement reported by the patient remained constant regardless of starting pain. It was instead the raw points that depended on the baseline: those who started from higher values needed a larger drop in points to report the same degree of improvement.
In practice: if someone starts at 8 and drops to 6, they have lost 2 points but only 25%. If they start at 4 and drop to 2, they have lost the same 2 points but 50%. The second case, with the same number of points, is almost always a more significant improvement for the person living it. That is why "minus 30%" is the more stable reading and "minus 2 points" is the shortcut, convenient but less reliable at the extremes of the scale.
In acute pain the numbers are different
The MCID is not a universal constant, and transferring it from chronic to acute pain is a mistake. In acute pain in the emergency department, measured with the 100 mm VAS, a study of 96 adults and 332 paired measurements estimated the minimal clinically significant difference at 13 mm, with a 95% confidence interval between 10 and 16 mm, confirming an earlier work. Another study, of 152 adults, instead found 9 mm (interval between 6 and 13 mm), with no differences by sex, by age above or below 50, or by traumatic cause of the pain.
Two warnings, both important. First: those values were measured on the VAS, not the NRS, and translating them into "about one point out of ten" is a convenient approximation, not a finding. Second: they refer to acute pain assessed over minutes or a few hours, a clinical situation that has nothing to do with a rehabilitation course of eight weeks.
What the MCID is not
It is not a threshold of success or failure. A patient who drops 1.5 points in six weeks has not "failed": they may have started sleeping, driving or carrying the shopping again. The MCID is a tool for interpreting numbers, not for discharging patients.
Nor is it a criterion that can be applied to a single measurement. It requires a pair of measures taken in the same way, at a sensible interval, after ruling out that the difference depends on the time of day or a bad night.
NRS, VAS and verbal scale: when they are equivalent and when they are not
The three most widespread scales are the NRS (a number from 0 to 10), the VAS (a mark on a 100 mm line between "no pain" and "worst possible pain") and the verbal scale or VRS (none, mild, moderate, severe).
A 2018 systematic review of 19 studies in adults concluded that all three are valid, reliable and suitable for clinical practice, but that the VAS is the hardest for the patient to complete, and that for general use the NRS has good sensitivity and produces data that are easy to analyse. The authors noted that the quality of the included studies was moderate to low.
On the ability to detect real differences in intensity, an experiment with 127 subjects exposed to a painful cold stimulus at four different temperatures found this order of sensitivity: NRS, then VAS, then VRS, then the faces scale. The differences between the scales were however small, and this is an experimental context, not a clinical one.
When they are not equivalent: in a study of 28 paediatric patients with sickle cell disease in the emergency department, VAS and NRS correlated strongly and moved together, but the VAS gave on average lower scores than the NRS, and agreement improved only at the highest pain levels. The authors' conclusion: they are similar, but not interchangeable.
In practical terms: switching from the VAS to the NRS halfway through a course can invent an improvement or erase one. You choose a scale at the start and keep it to the end.
The real limits of the NRS
It measures only one thing
The consequence of being one-dimensional is that intensity and interference are separate numbers and can diverge. In the study of 1,007 adults with chronic low back pain, the median of intensity recorded in real time was 1, while the median of interference with activities was 3: the same person, at the same moment, with two numbers that tell two different things.
The international recommendations for chronic pain studies reflect exactly this: they identify four distinct domains to measure separately, namely pain intensity on a 0-10 scale, physical function, emotional function and the patient's global impression of change. They also recommend using at least two different methods to judge whether a change is clinically important. If research is not satisfied with a single number, a rehabilitation course cannot be either.
It fluctuates with sleep, mood and context
The number is not a body temperature. In a study of 42 women with painful temporomandibular disorders, followed with electronic diaries, nights with better sleep quality were less often followed by pain flare-ups the next day, and days with higher levels of catastrophising were associated with a greater likelihood of a flare-up on the same day.
It is a small and specific sample, so no general rule can be drawn from it. But the signal is consistent with what is seen in the clinic: the same shoulder can score 5 after a bad night and 3 after a good one, with the same tissues. To explore how the cognitive component shapes perception, the pages on catastrophising about pain and on the psychological component of chronic pain are useful.
Older people with cognitive decline
Here the NRS may simply not work, and insisting produces invented numbers. A study of 129 patients with severe dementia (Mini-Mental State Examination score below 11) assessed the comprehension of three self-rating scales: 61% showed they understood at least one, and comprehension was significantly better for the verbal scale and the faces scale than for the visual analogue scale. When comprehension was present, the reliability of the measures was high.
The NRS was not among the scales tested in that study, so it cannot be said that it performs worse than the others. The useful finding is a different one, and it holds in general: a significant share of people with advanced cognitive decline cannot use a self-rating scale, so it should be verified before trusting the number, and verbal or picture-based scales tend to be understood better than analogue ones. When comprehension is lacking, observational tools are needed.
Children
A 2019 systematic review of self-report pain measures in children and adolescents aged 3 to 18 selected 8 tools out of 60 examined. For acute pain, the 11-point NRS, the Faces Pain Scale-Revised and the Colour Analogue Scale received a strong recommendation. For postoperative and chronic pain the authors could only make weak recommendations, and no measure was recommended under the age of 6, where the literature is still insufficient.
So with children the NRS can be used, but not at any age and not for every type of pain. On how a child's pain is really read, and on when recurring pain deserves a closer look, we have a dedicated page: growing pains: what they really are and Ri-Hub Junior, our physiotherapy service for children and adolescents.
How it is used in physiotherapy to decide whether something works
The NRS enters clinical reasoning at three moments, and in none of them does it decide on its own.
- At the start of the course, to fix the baseline with a written, repeatable question. It is the value on which the 30% will be calculated.
- Within a single session, before and after the work, to understand how the tissue responds to load in the acute phase. It is safety and dosage information, not a measure of treatment effectiveness.
- Over time, as a series, to understand whether the trajectory is going down. It is the only one of the three that says anything about effectiveness.
Confusing the second with the third is the most common error. A pain that drops by 2 points at the end of each session and returns to the starting value the next day for eight weeks is not a course that is working: it is a temporary relief, repeated. Conversely, a pain that does not move within a single session but goes from 7 to 4 across a six-week series is working very well.
This is why the number alone is not enough. Alongside the NRS, the physiotherapist should collect at least one indicator of function (how far you can walk without stopping, how many degrees of shoulder elevation, whether you can sleep on the painful side, whether you have gone back to training) and keep an eye on the behavioural component, for example the avoidance of movement described on the page on kinesiophobia. A patient who drops from 6 to 5 but has gone back to running is improving more than one who drops from 6 to 4 and keeps avoiding the stairs.
At Ri-Hub, the NRS values reported by the patient are recorded at every session and shown as a graph in the Hub section of the app, together with the functional goals: it serves exactly to read the trajectory rather than the single number, for both the physiotherapist and the person.
What to bring to your physiotherapist
Three habits are enough to make your numbers useful rather than decorative.
- Always answer the same question. If the first time you gave the average pain for the week, keep doing that, and say so explicitly.
- Note the context too, when it is unusual: a terrible night, a day of moving house, a stressful period. It helps you avoid mistaking noise for worsening.
- Bring a function measure, not only the intensity. What you can do today that you could not two weeks ago, or the reverse.
Need a consultation?
Let's see whether a course makes sense for you: a few questions about your case, you register, and you see who will follow you: a physiotherapist registered with the Albo, whom you can write to in chat whenever you like. If you would rather talk it through, the first consultation is free, lasts twenty minutes and commits you to nothing: they will arrive having already read what you wrote, and together you decide whether and how to go on.
Sources
- Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale, Pain, 2001
- Interpreting change scores for pain and functional status in low back pain: towards international consensus regarding minimal important change, Spine, 2008
- Prospective validation of clinically important changes in pain severity measured on a visual analog scale, Annals of Emergency Medicine, 2001
- Does the clinically significant difference in visual analog scale pain scores vary with gender, age, or cause of pain?, Academic Emergency Medicine, 1998
- Interpreting the clinical importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations, The Journal of Pain, 2008
- A systematic review of the pain scales in adults: which to use?, American Journal of Emergency Medicine, 2018
- Validity of four pain intensity rating scales, Pain, 2011
- A comparison of pain assessment measures in pediatric sickle cell disease: visual analog scale versus numeric rating scale, Journal of Pediatric Hematology/Oncology, 2015
- Comparison of different pain scoring systems in critically ill patients in a general ICU, Critical Care, 2008
- Pain in severe dementia: self-assessment or observational scales?, Journal of the American Geriatrics Society, 2006
- Recommendations for selection of self-report pain intensity measures in children and adolescents: a systematic review, Pain, 2019
- The contribution of sleep quality and psychological factors to the experience of within-day pain fluctuations among individuals with temporomandibular disorders, The Journal of Pain, 2024
- In-field ecological momentary assessment from wearable motion sensors and self-report in a chronic low back pain cohort, JOR Spine, 2026
Last source check: August 2026.