Neck Disability Index (NDI) Calculator

Neck Disability Index (NDI) Calculator: Score, Interpret and Track Cervical Disability

The Neck Disability Index is one of the most widely used self-report measures of neck-related disability, and it is also one of the more frequently mis-scored patient-reported outcomes in clinical use. It takes a patient three to five minutes to complete and a clinician about thirty seconds to score by hand — but the number it produces is only as useful as the threshold you compare it against, and many published thresholds don't generalize well across every caseload.

Use the calculator on this page to score the ten sections, convert to a percentage, place the result in Vernon's severity bands, and compare it against a previous visit. The rest of this page explains what that number actually supports — and, just as importantly, what it does not. For a broader look at how the NDI fits alongside other outcome measures, see Spry's PROMs hub.

Scoring note: at least 8 of the 10 sections need to be completed for the calculator to return a score. Below that threshold, the pro-rated percentage is considered too unreliable to interpret.

What the Neck Disability Index measures

The NDI is a ten-item patient-reported outcome measure of how much neck pain interferes with everyday activity. Howard Vernon and Silvano Mior published it in 1991 in the Journal of Manipulative and Physiological Therapeutics, adapting the structure of the Oswestry Disability Index after obtaining permission from its primary author, J. Fairbank. That lineage is why the two instruments look so alike: ten sections, six statements each, a raw score doubled to give a percentage.

What it measures is self-reported disability, not pain intensity, not tissue pathology, and not impairment. Two patients with identical MRI findings and identical numeric pain ratings routinely produce NDI scores twenty points apart, because the instrument is asking how much the neck is costing them, not how much it hurts. That is the point of it — and the reason it should sit alongside a pain scale rather than replace one.

The NDI is validated in mechanical neck pain, whiplash-associated disorder, cervical radiculopathy, cervicogenic headache and post-surgical cervical populations, and it has been translated into more than twenty languages.

The 10 sections of the NDI

Each section presents six statements. The patient ticks the single statement that best describes them today; the first statement in each section scores 0 and the last scores 5.

How to score the NDI

The arithmetic itself is simple. The interpretation of an incomplete form is where clinicians diverge.

Step 1 — sum the raw score. Add the ten section scores. The range is 0 to 50, where 0 is no disability and 50 is complete disability.

Step 2 — convert to a percentage. Multiply the raw score by two. A raw 26 becomes 52%. Equivalently, (raw ÷ 50) × 100. Both conventions appear in the literature; the percentage is the more commonly documented format in clinical practice, and it is what the calculator above reports as the headline number.

Step 3 — pro-rate any skipped sections. Reduce the denominator by five for every section left blank, then convert:

NDI % = (obtained score ÷ (5 × sections answered)) × 100

Example: 16 points across 9 answered sections = (16 ÷ 45) × 100 = 35.5%.

Here is the part almost no other NDI page states plainly: the original NDI publication contains no missing-data rule at all. Vernon's own 2008 review describes two competing conventions in circulation — reducing the denominator, or substituting the mean of the completed items into the missing one — and warns that if three or more items are missing, "the overall score may be suspect."

The practical consequence is that you must pick a method, apply it consistently, and record it. Switching methods between visits will manufacture change that did not happen. The calculator above uses the reduced-denominator method, states so on every pro-rated result, and refuses to produce a score below eight completed sections.

NDI score interpretation: the five severity bands

Vernon's original categories remain the most widely cited reference points, though — as noted below — they are a convention rather than a validated diagnostic standard.

Verify the responses before acting on them — consider serious underlying pathology or a comprehension problem with the form.

Two details are worth double-checking, since they vary across sources. First, the top band begins at raw 35 (rather than "over 35"), so a raw score of exactly 35 falls into the complete-disability band. Second, raw 25–34 converts to 50–68%; some sources round this slightly differently, so it's worth confirming which convention a given tool is using.

Treat these bands as Vernon's original convention rather than validated diagnostic cut-offs. Vernon himself notes they "have been revised by several authors," and no consensus revision has replaced them. They are useful for communicating with a patient or a payer; they are not a triage algorithm.

MCID and MDC: why a single threshold is the wrong question

This is where most NDI content stops being useful. The question a clinician actually has at reassessment is: did this patient really change? Answering it requires two different numbers, not one.

• Minimum detectable change (MDC) is the smallest change that exceeds measurement error. Below the MDC, you cannot distinguish improvement from noise.
• Minimal clinically important difference (MCID) is the smallest change the patient regards as worthwhile. It answers a different question, and it is typically derived from anchors such as a Global Rating of Change scale.

Published NDI values for both vary by a factor of five, and the variation is not random — it tracks the population.

Values are on the 0–50 raw scale unless marked otherwise.

Young and colleagues (2018) explain the spread directly: thresholds for patients without upper-extremity symptoms (MCID 5.5) are, in their words, "substantially lower than previously reported values for mixed samples with radicular symptoms" (8.5), and they conclude that "psychometric properties of self-report instruments should be population specific."

Read that as a workflow instruction. If your caseload is mechanical neck pain without arm symptoms and you apply the radiculopathy MCID of 8.5 points, you will document real, patient-valued improvement as "no significant change" — and you will do it consistently, in the direction that hurts authorisation.

There is one more wrinkle worth knowing. Vernon's own 2008 review recommends "a minimum clinically important change value of 3 to 5 points," well below the 7.5–8.5 that later studies support, because the MDC estimates he was working from (Vos: under 2 points; Trouli: 1.78) were far smaller than Pool's outlying 10.4. Later studies with larger samples have tended to support values nearer the higher figure. If you need a single number for a protocol, 7.5 points / 15% is a reasonable pooled estimate — but it is an estimate, not a universal cutoff, so state the population it came from and prefer a population-matched value when one is available.

Is the NDI reliable? What the psychometrics actually show

A 2024 meta-analysis by Saltychev and colleagues pooled 79 studies and gives the most defensible summary available.

PropertyPooled valueRange across studiesTest–retest reliability (ICC)0.91 (95% CI 0.90–0.93)0.50 to 0.96 — low values cluster in radiculopathy and mechanical samples with longer retest intervalsInternal consistency (Cronbach's α)> 0.810.72–0.93; the original 1991 paper reported 0.80Responsiveness (AUC)0.74 (95% CI 0.68–0.80)0.57 (Cleland 2006, radiculopathy) to 0.90 (Stratford 1999)Floor / ceiling effectsNone detected in most studiesA genuine NDI strength — it discriminates at both extremes

Convergent validity is strong against the measures you would expect: r = 0.86 with the Neck Pain and Disability Scale, 0.88 with the Northwick Park Questionnaire in whiplash, 0.77–0.80 with the Neck Bournemouth Questionnaire, 0.73–0.81 with the Patient-Specific Functional Scale, and 0.75 with the DASH. Correlations with pain intensity are weaker and more variable (0.38–0.89), which is what you want from a disability measure — if it tracked pain perfectly it would be redundant.

One honest caveat: Cleland and colleagues (2006) rated the NDI's construct validity as poor in cervical radiculopathy specifically, with responsiveness at AUC 0.57 — barely better than chance. In that population, the NDI is the weakest link in a battery, not the anchor.

Where the NDI falls short

It may not be measuring one thing

Dimensionality is genuinely contested. The 2024 meta-analysis found 13 studies supporting unidimensionality against 15 finding two or three dimensions, concluding only that "in most situations" the NDI can be treated as unidimensional. Rasch analyses have been less forgiving: Walton and MacDermid (2013) report poor model fit (χ² = 89.1, p < 0.001), local dependency between the driving and lifting items, and disordered response thresholds. Practically, this means the total score should be read with some caution despite its apparent precision. Section-level change is often the more informative signal.

Some items behave differently for different patients

Differential item functioning has been demonstrated by sex — women tend to score worse than men at equivalent levels of disability — and the driving item is problematic for anyone without regular access to a car, which matters for urban, low-income and international populations alike.

There is no psychosocial dimension at all

The NDI contains nothing on frustration, low mood, anxiety or fear of movement, despite these being among the strongest prognostic factors in persistent neck pain. It also has no item for computer work, sport, dizziness or medication effects. Across eleven commonly reported problems in mechanical neck pain, the NDI covers six; across nine common whiplash problems it covers three. Pair it with a Fear-Avoidance Beliefs Questionnaire when the presentation is persistent, and with a Dizziness Handicap Inventory when cervicogenic dizziness is in the picture.

A shorter NDI version exists

The NDI-5 (Walton & MacDermid, 2013) keeps Personal Care, Concentration, Work, Driving (rescored) and Recreation, dropping Pain Intensity, Headaches and Sleeping as symptom- rather than function-based, and Reading and Lifting for statistical reasons. In its validation studies it showed strong Rasch fit (χ² = 23.6, p = 0.26; person separation index 0.79), ICC 0.89–0.92 and effect sizes 0.56–0.87, though it has been tested in far fewer populations than the full NDI. It is not the field standard, but it may be worth considering for a new outcomes battery, with that smaller evidence base in mind.

NDI, NPRS, PSFS or the Neck Bournemouth Questionnaire?

These are not competitors so much as different jobs.

MeasureWhat it does bestKey numbersUse it whenNDIStandardised, comparable disability across patientsICC 0.88–0.91; MCID ~7.5/50Default disability measure for any cervical presentationNPRSPain intensity, in secondsICC 0.67; MDC 2.6; MCID 1.5Alongside the NDI, not instead of it — pain intensity and disability capture different thingsPSFSPatient-nominated, individualised activity limitationr = 0.73–0.81 with NDIThe patient's goal is specific and the NDI has hit its ceilingNeck BournemouthAdds the psychosocial dimensions the NDI omitsHigher internal consistency than NDI; SRM 1.17 vs 1.21Persistent presentations where distress is central — but note the evidence base is much thinner

Gay and colleagues (2007) compared the NDI and Neck Bournemouth Questionnaire head-to-head and found they "performed comparably" with similar sensitivity to change, though the inter-questionnaire correlation was only moderate (Spearman 0.46–0.57) — they are not interchangeable. A broader systematic review (Schellingerhout, 2012) favours the NDI overall, rating it positively for internal consistency, content validity, structural validity, hypothesis testing and responsiveness, while finding "minimal evidence" for the Neck Bournemouth Questionnaire.

Administering the NDI in a clinical workflow

1. Capture a baseline at evaluation, before the first intervention. An NDI collected after treatment has begun is not a baseline, and every subsequent change calculation inherits that error.
2. Let the patient complete it themselves. Reading the statements aloud changes response patterns. Hand over the form or a tablet and give them three to five minutes.
3. Re-administer on a fixed cadence — commonly every four to six visits, at any significant change in status, and at discharge. Fixed intervals make the trend line interpretable; opportunistic re-testing does not.
4. Document the raw score, the percentage, the number of sections completed and the pro-rating method if any section was skipped. The percentage alone is not auditable.
5. Compare against the population-matched MDC first, then the MCID. A change that does not clear the MDC should be interpreted cautiously, however encouraging it looks, since it may not exceed normal measurement variation.
6. Store it where it is queryable. An NDI score buried in a note body cannot be trended, reported to a payer, or used for clinic-level benchmarking. In an EMR that treats outcome measures as structured data, the trend, the change score and the MCID comparison come for free.

That last point is where an outcome measure stops being paperwork and starts being leverage — for utilisation review, for demonstrating value to referral sources, and for the clinician's own reasoning at visit six. Spry captures the NDI score, the trend across visits and the MCID comparison automatically as structured data at the point of care, instead of leaving it to be recalculated by hand from a note. See how Spry handles outcome measures as trendable data.

Licensing and permissions

The NDI is copyrighted (Vernon & Mior, 1991) and distributed under licence by Mapi Research Trust. It is free to access for students, clinicians in practice and non-funded academic users; fees may apply for funded academic users, healthcare organisations, commercial users and IT companies. Electronic or web-based implementations should be confirmed directly with Mapi Research Trust, since terms for digital use can differ from paper use. Requests are handled through eProvide, and translations may carry their own separate rights.

Frequently asked questions

What is a good NDI score?

Lower is better. A score of 0–8% is Vernon's "no disability" band; 10–28% is mild. But "good" depends entirely on where the patient started — a drop from 68% to 40% is an excellent outcome even though 40% still sits in the moderate band. Judge the change, not the absolute value.

What does an NDI score of 20 mean?

A raw score of 20 falls in the 15–24 range, which converts to 40% and sits in Vernon's moderate disability band. At this level, several activity domains are typically affected — for example concentration, sleep or lifting — rather than one isolated area. As with any single score, it matters more in context: whether it's rising, falling or holding steady across visits, than as a one-time snapshot.

How do you calculate an NDI score?

Add the ten section scores (0–5 each) for a raw total out of 50, then multiply by two for a percentage. If a section was skipped, divide the obtained points by (5 × sections answered) and multiply by 100.

What is the MCID for the Neck Disability Index?

The pooled value across 79 studies is roughly 7.5 points out of 50, or 15% — treat this as a pooled estimate rather than a universal cutoff. Population-specific values differ substantially: 8.5 points in cervical radiculopathy, 5.5 points in mechanical neck pain without upper-limb symptoms, and 3.5 points in chronic non-specific neck pain. Use the value matched to your population when one is available.

What is the difference between MCID and MDC on the NDI?

MDC is the smallest change that exceeds measurement error — commonly around 10% (5 of 50 points) for the NDI. MCID is the smallest change the patient considers worthwhile. A change should clear the MDC before you interpret it at all, and clear the MCID before you call it clinically important.

How long does the NDI take to complete?

Three to five minutes for most patients, and about thirty seconds to score by hand. The calculator on this page scores it instantly.

What if a patient skips the driving section?

Pro-rate: divide the obtained points by (5 × sections answered) and multiply by 100. This is the most commonly skipped section, and skipping it is legitimate for a patient who does not drive. Record that you pro-rated, and use the same method at every reassessment.

Can the NDI diagnose a neck condition?

No. It quantifies self-reported disability. It cannot identify a cause, and it does not screen for red flags such as fracture, myelopathy or vascular pathology — those require separate clinical assessment.

Is the NDI free to use?

It is free for students, clinicians in practice and non-funded academic users. Funded academic users, healthcare organisations, commercial users and IT companies may be charged a licence fee. Requests are handled by Mapi Research Trust.

References

1. Vernon H, Mior S. The Neck Disability Index: a study of reliability and validity. J Manipulative Physiol Ther. 1991;14(7):409–415.
2. Vernon H. The Neck Disability Index: state-of-the-art, 1991–2008. J Manipulative Physiol Ther. 2008;31(7):491–502.
3. Saltychev M, et al. Psychometric properties of the Neck Disability Index amongst patients with chronic neck pain: systematic review and meta-analysis. Disabil Rehabil. 2024;46(23):5415–5431.
4. Young IA, et al. Reliability, construct validity and responsiveness of the Neck Disability Index and Numeric Pain Rating Scale in patients with mechanical neck pain. Physiother Theory Pract. 2018.
5. Young BA, et al. Responsiveness of the Neck Disability Index in patients with mechanical neck disorders. Spine J. 2009;9(10):802–808.
6. Cleland JA, et al. Psychometric properties of the Neck Disability Index and Numeric Pain Rating Scale in patients with mechanical neck pain. Arch Phys Med Rehabil. 2008;89(1):69–74.
7. Walton DM, MacDermid JC. A brief 5-item version of the Neck Disability Index shows good psychometric properties. Health Qual Life Outcomes. 2013;11:108.
8. Schellingerhout JM, et al. Measurement properties of disease-specific questionnaires in patients with neck pain: a systematic review. Qual Life Res. 2012;21(4):659–670.
9. van der Velde G, et al. Rasch analysis provides new insights into the measurement properties of the Neck Disability Index. Arthritis Rheum. 2009;61(4):544–551.
10. Gay RE, Madson TJ, Cieslak KR. Comparison of the Neck Disability Index and the Neck Bournemouth Questionnaire. J Manipulative Physiol Ther. 2007;30(4):259–262.