I think this is one of the cases where the proverb fits: "shit in - shit out".

If the patients are not well informed and may differ in their understanding and attitude towards the scale, if the data obtained from is not comparable.

As long as you have no detailed information about all the paitents' understanding and motivation (e.g. what other patients served as "reference", and how precisely did they serve as reference, how did the patient "adjust" his choice etc) there won't be any sensible way to correct or normalize your data.

There is at least one very important thing one can learn from this experiment:

The quality of the data is fundamental for all further analysis. This quality depends much on the study design and the protocol (methods). Analysing/interpreting data without being sure that the data is really ok is a waste of time.

The most hopeful feature is that you have before and after measurements, so you can include the before measurement as a covariate in your comparison. The problem with the poor instruction will be to introduce measurement error, thus making significant differences between the groups harder to detect.

I'm with Jochen on this. No amount of rescaling of data will remove measurement error, which seems to be the critical problem here.

You don´t need to check for normality (if you have at least n=30; check central limit theorem), Visual Analogue Scale (VAS) have been tested and accepted by anesthesiology and surgical community like the most practical and aproximately normal scale to measure pain, check old anesthesiology and neurology journals articles (this problem was solved in clinical research a lot of years ago),other scales are more accurate (e.g. Neurologists scales), BUT less practical in clinical research.

aahhh...and to compare use Wilcoxon rank test or Sign test for categorical data (both are non-parametric tests, you don´t need to have a "normal distribution of data") in IBM-SPSS  is set to 99%, good luck!!!

VAS measures the worst pain the patient has ever experienced. For women it is usually delivering a baby. For men it is often no more than a visit to the dentist. Vastly different amounts of pain yet both rate a 10 for that individual. What you are really measuring here is not the absolute amount of pain but the amount of change in the present pain. Have faith in your raw data.

Very interesting issue! I appreciate the innovative approach proposed by Dr Yarnold.

I agree with Dr Bravo, the easiest way for comparison purpose is to adopt a non parametric test like Wilcoxon rank test.

I just want to put your attention to another interesting approach proposed by Silverman et al.

http://www.ncbi.nlm.nih.gov/pubmed/8317727

http://www.ncbi.nlm.nih.gov/pubmed/23743256

chaque patient est unique et son histoire précisément modifiera son intensité de douleur.

Bien éduquer le patient à l'outil et revenir sur les principes fondamentaux.

Le patient communicant s'auto évalue (gold)et donc doit prendre l'outil en main et déplacer le curseur lui-même (reproductibilité).QUAND IL Y A UN DOUTE revenir aux explications.Voir aussi si sousjacent il n'y a pas des craintes,anxiété,peur,souffrance qui augmente la douleur.

Demander avis à son collègue

Depuis 40 ans sur le terrain j'évalue les patients avec bon nombre d'outils différents adapté à la population et je suis bien certaine de ma justesse.

Maintenant pour les data cela reste un autre problème de la qualité des évaluations des équipes et l'écriture dans les dossiers infirmiers!

Former non stop le personnel à l'usage de l'EVA (VAS).

Quand rescue réevaluer dans l'heure le patient.

  As pain is unpleasant sensory and emotional experience so Visual analogue scale  assessment is clearly highly subjective.

I agree with Dr Bravo, In clinical research VAS is more sensitive to small changes than are simple descriptive ordinal scales in which symptoms are rated so you don´t need to check for normality as it is individuals pain experience.

Pain is what the patient says it is. Why is the pain score being measured?  Usually to compare scores over time in that patient, so the patient is the standard for themselves. Changes in score are what matter, not absolute values. Each animal species and culture has its' own characteristics; vocalisation / drama remain relatively constant with the individual.  So, no normalisation is needed.

Pain responses using the visual analog scale are often non-normally distributed. As described previously, if the sample size is more than 30 in each arm, you do not need to worry about non-normally distributed data since the mean is thought to reflect the mean in the target population. Use of independent t-tests, rather than non-parametric tests is acceptable even though pain is in fact ordinal.  Non-parametric tests are less likely to show a statistical difference than parametric tests and are less preferred for that reason.  Yes,  the VAS definitely has a ceiling effect in patients with very high levels of pain who move on to more extreme levels of pain (e.g. laboring women).  Other issues exist with use of the tool (i.e. end aversion, difficulty with understanding it in the elderly or those with language issues, scoring it/measurement error since it is non-graduated). When used as a single measure there are also issues with reliability -- that is why most validated questionnaires are made up of multiple questions representing different domains of the pain experience meant to be measured.  The issue with assessing changes in pain when retrospective at least is whether patients can accurately recall their previous pain levels when doing so.  This is a problem when the testing interval is long. Yes. adjusting for baseline differences in groups makes sense if looking at changes in pain over time in a pre-test post-test scenario.  .

Thank you All for your views on this topic. In this issue, we met with the opinions of experts from various professions which can be beneficial for everyone to construct his own complex view. The question is still open, followers give feedback in case of new answer.

I do not subscribe to Andrews answer, although it seems to be sensible.

Even the differences in pain ratings may have a different meaning, and it may be valuable (if available!) to calibrate at least the differences.

Example:

You may imagine the case where persons are get a control and an actual treatment (e.g. holding the hand 5s into water of 4°C and 0°C). Imagine that men are much more "pain-responsive" than women, rating the pain for 4°C an 8, where as women rate it 5. For the icecold water, women can rate 8 or even 9, what is 3-4 units more, but man can only maximally give a 2 unit higher rating. You would conclude that icecold water (compared to 4°C) induces more (additional) pain in women than in men.

Is this conclusion reasonable? Does it makle sense? Is it helpful?

It may be, but it also may not be. I think, to come to some biologically relevant conclusion one needs some information how comparable the "8" of men is to the "5" of women, by some relevant external measure. And this seems to be a calibration (or, more precisely, one particular conclusion from a calibration). Maybe I was not very concrete or I have chosen a bad example but I hope you get the point.

Three methods are used by neurologists and anesthesiologists to assess level of pain, two are not very practical, I found VAS was the most extended test used in clinical research, for children using categorical rule marked with emoticons and for adults graphical rule marked from 0 to 100, when I checked journal articles the difference between authors was inclusion criteria, how example don t include rural workers with people working in urban offices (e.g. banks), check journal articles using VAS and pain as keywords in www.pubmed.org, because the pain perception and limits may be different. Remember, pain isn't normal, that's because disease are not normal because our specie introduced disease prevent measure or treatment, altering the natural outcome.