SPINE Volume 35, Number 21, pp E1115–E1119 ©2010, Lippincott Williams & Wilkins The Visual Analog Scale and a Five-Item Verbal Rating Scale Are Not Interchangeable for Back Pain Assessment in Lumbar Spine Disorders Anto`nia Matamalas, MD,* Manuel Ramírez, MD,* Sergi Mojal, Licentiate in Biostatistics,† Ana García De Frutos, MD,* Antonio Molina, MD, PhD,* Guillem Salo´, MD, PhD,* Andreu Llado´, MD, PhD,* and Enric Ca´ceres, MD, PhD* Study Design. Prospective study of patients with ing that VAS and VRS are not interchangeable and, there- chronic back pain from lumbar spine disorders. fore, a results obtained with the use of each scale cannot Objective. To evaluate the degree of interchangeabil- be compared. ity of a 100-mm visual analog scale (VAS) and a 5-point Key words: low back pain, visual analog scale, 5-point verbal rating scale (VRS) for the assessment of pain in- verbal rating scale, pain intensity, lumbar degenerative tensity. disease. Spine 2010;35:E1115–E1119 Summary of Background Data. The fact that VAS and Likert scales are highly intercorrelated does not mean that both types of scales can be used interchangeably. Reliable and accurate assessments of perceived low back Methods. A total of 151 patients (mean age, 52 Ϯ 14.6 years) undergoing elective spine surgery completed a pain are indispensable in clinical practice for diagnosis, 100-mm VAS and a discrete 5-category VRS corresponding selection of the most appropriate treatment modality, to the first item question of the core set (“How severe was and evaluation of treatment efficacy in patients with de- your back pain in the last week?”). Pain intensity on the VAS generative disc disease.1 Pain is a multifactorial unpleas- was rated using the same question than for the VRS. The ant sensory experience, which is influenced by different level of order-consistency (monotonic agreement), disor- dered pairs (D), percentage of agreement, and systematic aspects, such as the intensity of pain and the affective 2 disagreement (relative position), and relative concentration component. Pain intensity or pain severity is probably ([RC]) were estimated. VAS assessments were transformed the most frequently assessed dimension.3,4 The most into a discrete 5-category, with the cut-off VAS positions commonly used scales for measuring pain intensity are being defined by quintiles and equidistantly. the visual analog scale (VAS) and the verbal rating scales Results. For VAS defined equidistantly, monotonic agreement was 0.840, D was 0.080, and the percentage of (VRS). VAS consists of a 100-mm straight horizontal (or identical pairs was 53%. The corresponding figures for vertical) line labeled “no pain” at one end and “pain” as VAS defined by quintiles were 0.809, 0.096, and 27.8%. bad as could possible be on the other.5 In contrast to the Inconsistencies between the VAS and the VRS scales continuous VAS, Likert-type scales are comprised of a were also demonstrated by the marginal distributions, variable number of word categories (e.g., 2-point VRS with PR values of Ϫ0.005 (95% confidence interval [CI], Ϫ0.011 to Ϫ0.002) and RC values of 0.144 (95% CI, 0.137– and 11-point numerical rating scale), although the num- 0.152) for VAS defined equidistantly, and PR values of ber of levels needed to assess self-reported pain intensity 0.391 (95% CI, 0.384–0.397) and RC values of 0.265 (95% is an important issue that has yet to be resolved.6,7 CI, 0.255–0.275) for VAS defined by quintiles. Both VAS and Likert scales have been evaluated in Conclusion. The order-consistency level was low with terms of reliability, validity, and responsiveness. In gen- overlapping of pain records between the 2 scales, indicat- eral, both scaling methods seem to be reliable and valid for its use in routine clinical assessment.8–12 It has been shown that the correlation between both scales is highly From the *Service of Orthopaedic Surgery and Traumatology, Hospi- significant and most authors consider that both types of tal Universitario del Mar, Universitat Auto` noma de Barcelona, Barce- lona, Spain; and †Service of Methodological Consulting in Biomedical scales measure the same aspects of pain and can be used Research, Institut Municipal d’Investigacio´Me`dica (IMIM-Hospital interchangeably. Therefore, VAS and Likert scales are del Mar), Barcelona, Spain. indistinctively used for direct comparison of the results Acknowledgment date: February 24, 2009. First revision date: June 26, 2009. Second revision date: March 8, 2010. Acceptance date: March obtained in different studies. However, the fact that 2 12, 2010. scales are highly intercorrelated does not necessarily Presented as an oral communication at: The 9th Congress of the Euro- mean that they measure the same construct. Moreover, pean Federation of National Associations of Orthopaedics and Trau- matology (EFORT), Nice, France. 29 May–1 June, 2008. considerable interindividual variability and overlap of The manuscript submitted does not contain information about medical VAS scores and VRS ratings have been reported, and device(s)/drug(s). there is no enough data to conclude that one scale type is No funds were received in support of this work. No benefits in any 1,12,13 form have been or will be received from a commercial party related better than the other. directly or indirectly to the subject of this manuscript. In spine patients, the scarce evidence on the results of Address correspondence and reprint requests to Manuel Ramírez, MD, different treatment methods of degenerative lumbar Service of Orthopaedic Surgery and Traumatology, Hospital Universi- tario del Mar, Universitat Auto` noma de Barcelona, Passeig Marítim spine diseases requires the use of standardized outcome 25–29, E-08003 Barcelona, Spain; E-mail: [email protected] measures to allow international comparisons.14 In this E1115 E1116 Spine • Volume 35 • Number 21 • 2010 respect, the disparity in results published on the inter- Table 1. Frequency Distributions of Paired Data of changeability of VAS and VRS scales and the need for Condensed 5-Category VAS Versus the Discrete VRS in standardized measurement tools make necessary to de- Case of Maximum Order-Consistency termine to what extent both scales can be compared for VAS the assessment of results in lumbar degenerative disease. The objective of this study was to evaluate the degree of VRS 12345 interchangeability of VAS and a 5-point Likert scale for None 11000 2 back pain assessment in patients with lumbar spine dis- Mild 0522210 orders undergoing elective spine surgery. Moderate 1298121 Severe 0 1 8 33 20 62 Materials and Methods Very severe 0 1 2 17 36 56 210216058151 Between June 2007 and July 2008, all consecutive patients with VAS indicates visual analog scale; VRS, verbal rating scale. lumbar degenerative disease scheduled to undergo elective spine surgery at the Spinal Unit of the Department of Ortho- pedic Surgery and Traumatology of Hospital Universitario del Mar in Barcelona, Spain, were eligible to participate in the ping when responses obtained on the 2 scales were found in study. Illiterate patients and those who were unable to self- different levels of pain. In addition, pairs should be ordered, complete the study questionnaires were excluded from the that is, low scores on the VAS should be consistent with low study as were those who refused to participate. The assess- scores on the VRS. The number of disordered pairs was calcu- ments were conducted in accordance with the declaration of lated and defined the measure of disorder, D. Systematic dis- Helsinki and written informed consent was obtained from all agreement was measured by means of the relative position (RP) participants. The study was approved by the Ethics Committee and the relative concentration (RC) with possible values rang- Ϫ of Hospital Universitariodel Mar. ing from 1 to 1. The RP estimates the probability of the pain All patients were admitted to the hospital 24 hours before assessments on 1 scale being shifted toward higher or lower surgery, and were administered a series of patient-based out- categories relative to the other. The RC estimates the differ- come assessments instruments, including the Spanish version of ences between the probability of the pain assessments on 1 scale the Oswestry Low Back Pain Disability Questionnaire,15,16 the being concentrated to the other and vice versa. The distribution ϫ pain item of the core set developed by Deyo et al,17,18 the SF-36 of the pairs of data was evaluated by means of 5 5 contin- 13 Health Survey,19 a VAS for pain intensity, and a questionnaire gency tables (Table 1). The level of order-consistency was on work status. The discrete 5-category VRS corresponded to estimated by the coefficient of monotonic agreement (MA), the first item question of the core set (How severe was your which is defined by the difference between the probabilities of back pain in the last week?), with the eligible alternatives of ordered and disordered pairs of assessments. The range of pos- ϫ “none,” “mild,” “moderate,” “severe,” and “very severe.” The sible values of MA is 1 to 1. The degree of concordance VAS scale was a 100-mm long horizontal line labeled no pain at between VAS and VRS instruments was also assessed with one end and “worst pain possible” at the other end. In order to kappa statistics. The R statistical package was used for data avoid bias in the wording for the patients’ experienced pain analysis. intensity level, they were asked to rate their pain intensity on Results the VAS using the same question than for the VRS. There were 293 patients undergoing spine surgery dur- Statistical Analysis ing the study period.