Estimating EQ-5D values from the Neck Disability Index and numeric rating scales for neck and arm pain

Transcription

1 J Neurosurg Spine 21: , 2014 AANS, 2014 Estimating EQ-5D values from the Neck Disability Index and numeric rating scales for neck and arm pain Clinical article Leah Y. Carreon, M.D., M.Sc., Kelly R. Bratcher, R.N., C.C.R.P., Nandita Das, Ph.D., Jacob B. Nienhuis, M.Ed., and Steven D. Glassman, M.D. Norton Leatherman Spine Center, Louisville, Kentucky Object. The Neck Disability Index (NDI) and numeric rating scales (0 to 10) for neck pain and arm pain are widely used cervical spine disease specific measures. Recent studies have shown that there is a strong relationship between the SF-6D and the NDI such that using a simple linear regression allows for the estimation of an SF-6D value from the NDI alone. Due to ease of administration and scoring, the EQ-5D is increasingly being used as a measure of utility in the clinical setting. The purpose of this study is to determine if the EQ-5D values can be estimated from commonly available cervical spine disease specific health-related quality of life measures, much like the SF-6D. Methods. The EQ-5D, NDI, neck pain score, and arm pain score were prospectively collected in 3732 patients who presented to the authors clinic with degenerative cervical spine disorders. Correlation coefficients for paired observations from multiple time points between the NDI, neck pain and arm pain scores, and EQ-5D were determined. Regression models were built to estimate the EQ-5D values from the NDI, neck pain, and arm pain scores. Results. The mean age of the 3732 patients was 53.3 ± 12.2 years, and 43% were male. Correlations between the EQ-5D and the NDI, neck pain score, and arm pain score were statistically significant (p < ), with correlation coefficients of -0.77, -0.62, and -0.50, respectively. The regression equation ( NDI) + ( arm pain score) + ( neck pain score) to predict EQ-5D had an R-square of 0.62 and a root mean square error (RMSE) of The model using NDI alone had an R-square of 0.59 and a RMSE of The model using the individual NDI items had an R-square of 0.46 and an RMSE of The correlation coefficient between the observed and estimated EQ-5D scores was There was no statistically significant difference between the actual EQ-5D score (0.603 ± 0.235) and the estimated EQ-5D score (0.603 ± 0.185) using the NDI, neck pain score, and arm pain score regression model. However, rounding off the coefficients to fewer than 5 decimal places produced less accurate results. Conclusions. The regression model estimating the EQ-5D from the NDI, neck pain score, and arm pain score accounted for 60% of the variability of the EQ-5D with a relatively large RMSE. This regression model may not be sufficient to accurately or reliably estimate actual EQ-5D values. ( Key Words EQ-5D Neck Disability Index cost-utility analysis utilities outcome measurement cervical uality-adjusted life years (QALYs) are used as a measure of both health gains and losses due to different disease processes and the interventions used to treat these conditions. They account for both the quantity and the quality of life, and they are the arithmetic product of life expectancy and a measure of the quality of the remaining life years. 4,12,13,17,34 The measurement of life years is relatively straightforward. However, there are various ways of measuring the quality of life or health state during the years that an individual lives with the disease and its treatments. Health states are valued using a single score, anchored at 0 for death and 1 for Abbreviations used in this paper: EQ-5D-3L = EQ-5D 3 level; EQ-5D-5L = EQ-5D 5 level; NDI = Neck Disability Index; QALY = quality-adjusted life year; RMSE = root mean square error; VAS = visual analog scale. 394 perfect health. These health state values can be measured directly or indirectly. Direct methods, such as standard gamble, time tradeoff, and visual analog scale (VAS) assessments, are difficult to administer in the clinic setting. Thus, indirect methods, using preference-based measurement systems, such as the Quality of Well Being Scale, 20 the EQ-5D, 14 SF-6D, 6 and the Health Utilities Index, 15 are increasingly being used. 5 One year of perfect health generates 1 QALY, whereas 1 year in a health state valued at 0.5 generates half a QALY. Hence, an intervention that generates 4 additional years in a health state valued at 0.75 will generate 1 more QALY than an intervention that generates 4 additional years in a health state valued at 0.5. The Neck Disability Index (NDI) 31,32 and numeric rating scales for neck pain and arm pain 23 are widely used disease-specific measures in patients with degenerative cervical spine disorders. Recent studies have shown

2 EQ-5D from the NDI that there is a strong relationship between the SF-6D and the NDI, such that a simple linear regression (SF-6D = [ NDI]) allows for the estimation of an SF-6D value from the NDI alone. 8 Because of its ease of administration and scoring, the EQ-5D is increasingly being used as a measure of utility in the clinical setting. The purpose of this study is to determine if the EQ-5D can also be derived from commonly available cervical spine disease specific health-related quality of life measures, much like the SF-6D. Finding a simple means to convert the cervical spine disease specific measures to the EQ-5D will allow researchers to use previously collected NDI, neck pain score, and arm pain score data for cost-utility studies. Evaluating the relationship between these disease-specific measures and the EQ-5D may also help researchers and clinicians define the role of administering EQ-5D on a routine basis in the clinic. Methods All patients seen at the clinic presenting with neck and/or upper-extremity complaints were asked to complete the NDI, 31,32 numeric rating scale for neck pain (0 10) and arm pain (0 10), 23 and the EQ-5D 3 level (EQ-5D-3L). 14 These patients were seen at a single tertiary spine center comprising 6 spine fellowship trained surgeons with similar practice patterns; we included both surgical and nonsurgical patients, representing a typical adult spine practice population. After receiving institutional review board approval, we identified and included in the analysis all patients older than 18 years of age with degenerative cervical spine pathology who were seen between January 2011 and December Outcome Measures EQ-5D. The EQ-5D-3L 14 consists of 2 pages: a descriptive system and a visual analog scale (EQ VAS). The descriptive system has 5 dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) each with 3 levels (no problems, some problems, and extreme problems). The respondent indicates his/her health state by marking the box with the most appropriate statement in each of the 5 dimensions. The EQ VAS records the respondent s self-rated health on a vertical VAS where the end points are Best imaginable health state and Worst imaginable health state. The scoring algorithm for the US population was used to assign a value, or index score, that incorporates US population based preference weights. 1 The EQ-5D scores for the US population range from (worse than death) to 1.0 (full health), with a score of 0 indicating death. NDI. The NDI 31,32 is a 10-item self-administered questionnaire measuring disability in patients with neck pain. Each item is scored from 0 to 5 for a maximum score of 50; the higher the score, the greater the disability. In this study, a percentage score was used, so that the range of scores is The 10 items include pain intensity, personal care, lifting, reading, headaches, concentration, work, driving, sleeping, and recreation. Neck Pain and Arm Pain. The numeric rating scale consists of 2 items, 1 each for neck pain and arm pain, on the survey that asks, On a scale from 0 to 10, mark your level of neck (arm) pain discomfort, with 0 being none and 10 being unbearable. 23 Statistical Analysis Spearman and Pearson correlation coefficients for paired observations from multiple time points between EQ-5D and NDI were determined. Although these measures can be analyzed as continuous variables, some argue that these measures are long ordinal variables. Thus, both Spearman and Pearson correlations were calculated. Spearman correlation coefficients between EQ-5D and neck pain and arm pain scores were calculated as the neck pain and arm pain scores are considered discrete ordinal scales. Correlation coefficients between 0.60 and 0.80 indicate a marked degree of correlation, while coefficients between 0.80 and 1.00 indicate robust correlations. 16 Linear regression modeling was performed to predict EQ-5D scores from the NDI, neck pain scores, and arm pain scores. The R-square and root mean square error (RMSE) was used to assess the ability of the model to estimate EQ-5D scores and to compare the performance of the various models. The R-square is reflective of the proportion of variation in the dependent variable (in this case, the EQ-5D score) that can be explained by the factors considered in the model (NDI and neck and arm pain scores). Thus, the better the factors are in predicting the dependent variable, the closer the R-square value is to 1.0. The RMSE provides a measure of the accuracy by which the model can predict the outcome of interest. Depending on model specifications, degrees of freedom, and sample size, a value of 0.05 or smaller is a reasonable error of approximation. Root mean square error values of 0.10 or greater will lead to unacceptable errors in approximation. 9 Results A total of 3732 patients were included in the analysis with a mean age of 53.3 ± 12.2 years; 43% of patients were male. Table 1 provides a summary of the outcome measures. All correlations between paired EQ-5D and NDI, neck pain score, and arm pain score were statistically significant (p < ). The strongest correlation was for EQ- 5D and NDI (Pearson = -0.77, Spearman = -0.79), followed by EQ-5D and neck pain score (Spearman = -0.64). The relationship between EQ-5D and arm pain score was the weakest (Spearman = -0.50), but it was still statistically significant (Table 2). Looking at the different dimen- TABLE 1: Summary statistics of all paired observations Clinical Outcome Score Parameter EQ-5D EQ VAS NDI Neck Pain Arm Pain mean SD minimum maximum

3 L. Y. Carreon et al. TABLE 2: Spearman correlation coefficients between EQ-5D dimensions and NDI items EQ-5D Dimensions NDI Items Mobility Self-Care Usual Activities Pain/Discomfort Anxiety/Depression EQ-5D VAS pain intensity personal care lifting reading headaches concentration work driving sleeping recreation neck pain arm pain sions of the EQ-5D and the individual items of the NDI, strong correlations were seen between the EQ-5D self-care dimension and NDI personal care (r = 0.623), EQ-5D usual activities dimension and NDI work (r = 0.626) and NDI recreation (0.584), and the EQ-5D pain/discomfort dimension and NDI pain intensity (r = 0.670) (Table 2). The regression equation using NDI, neck pain score, and arm pain score to predict EQ-5D accounted for 62% of the variability with an RMSE of (Table 3). The regression equation is as follows: EQ-5D = ( NDI) + ( arm pain score) + ( neck pain score). The model using NDI alone had an R-square of 0.60 and an RMSE of The model using the individual NDI items had an R-square of 0.46 and an RMSE of There was no statistically significant difference between the actual EQ-5D (0.603 ± 0.235) and the estimated EQ-5D score (0.603 ± 0.185) using the NDI, neck pain score, and arm pain score regression model. However, rounding off the coefficients to fewer than 5 decimal places produced less accurate results. Discussion The use of QALYs allows a unit to be assigned to the value of a health care intervention. The use of a single number reflects both the health state value losses and gains from the disease as well as the interventions used to treat the disease. This allows health care economists to determine the expected utility of an intervention and compare it to other interventions for the same or different conditions. In the clinical setting, health state utility values are usually measured indirectly, with the use of preference based self-administered questionnaires such as the EQ-5D or the SF-6D. Several studies have shown that, although the EQ-5D and SF-6D are both psychometrically reliable and valid measures of health state utility, they cannot be used interchangeably. 3,7,10,26,35,36 This has been found to be true even in patients with low-back pain 7,24,25,27 or in patients with neck pain. 35 Theoretical causes for these differences in- TABLE 3: Comparison of various models predicting the EQ-5D Stepwise Regression Results p Value R 2 RMSE NDI alone NDI, arm pain score NDI arm pain NDI, arm pain score, neck pain score NDI arm pain score neck pain score individual NDI items pain intensity personal care lifting reading headaches concentration work driving sleeping recreation individual NDI items, arm pain, neck pain pain intensity personal care lifting reading headaches concentration work driving sleeping recreation arm pain neck pain

4 EQ-5D from the NDI clude dissimilarities in the descriptive components leading to the measurement of distinct constructs, variation in the period of recall, and the use of different valuation methods (time tradeoff for the EQ-5D and chained standard gamble for the SF-6D). These differences in the health state utility values derived from the EQ-5D and SF-6D are substantial enough that several cost-effectiveness studies present their results using both the EQ-5D and the SF-6D as part of the sensitivity analysis. 29,30 If the relationship between the EQ-5D and NDI were as robust as the relationship between the SF-6D and NDI, researchers would be able to perform this sensitivity analysis using NDI data alone. Thus, data sets that only collected cervical spine disease specific outcome measures may have limited use in cost-utility analysis if the EQ-5D was not concurrently collected. Data from the current study suggest that the relationship between the EQ-5D and the NDI is not strong enough to allow for a valid estimation of EQ-5D scores from the NDI using a regression equation. This is true even when neck pain and arm pain scores are included in the equation or when individual items of the NDI are used in the equation instead of the NDI score alone. None of the equations performed well in predicting EQ-5D utility scores as reflected in the relatively large RMSE, a measure of the difference between the values predicted by the equation and the actual values (Fig. 1). There may be several reasons why the NDI cannot provide a valid estimate of the EQ-5D. The instruments may have different descriptive components and therefore measure very different constructs. This is evidenced by lack of strong correlations between the majority of the NDI items and the EQ-5D dimensions. Only those components that measure similar constructs had strong correlations: EQ-5D self-care and NDI personal care, EQ-5D usual activities and NDI work and NDI recreation, and the EQ-5D pain/discomfort and NDI pain intensity. No strong correlations were seen between the EQ-5D dimensions of mobility or anxiety/depression and any of the NDI items. Several studies have also shown a difference in the distribution of scores between the EQ-5D and SF- 6D. Whereas the SF-6D scores and NDI are dispersed and normally distributed, the EQ-5D tends to cluster or skew 35,36 or have a bimodal distribution. 7,27 Looking at the scatterplot between the NDI and the EQ-5D scores, a floor effect is evident as well as gaps in the EQ-5D values (Fig. 2). The relationship between the EQ-5D and the NDI may not be defined by a simple linear regression. However, using nonlinear regressions did not produce robust estimates either. The current EQ-5D in wide use (the EQ-5D-3L, which has 3 choices for each dimension) may not be sensitive enough to measure small changes in health, especially in patients with milder conditions. Several studies on the EQ-5D in the general population 2,11,19,21,22,28,33 as well as in patients with low-back pain 7,24,25,27 showed a ceiling effect. In a study of the EQ-5D in 346 patients with neck pain, Whitehurst and Bryan showed a floor effect, which is similar to the current study. 35 Thus, the current EQ-5D-3L may have limited use as an aid in medical decision making and measurement of treatment effects in clinical practice. In response, the EuroQol group is currently developing the EQ-5D 5 level (EQ-5D-5L), expanding the number of levels of severity in each dimension to 5 instead of Unfortunately, value sets based on preferences directly elicited from representative general population samples are not yet available. Future studies using the EQ-5D-5L may show a more robust relationship with the NDI allowing for estimation of the EQ-5D from the NDI. Conclusions The regression model estimating the EQ-5D from the NDI, neck pain score, and arm pain score accounted for 60% of the variability of the EQ-5D with a relatively large RMSE. This regression model may not be sufficient to accurately or reliably estimate actual EQ-5D values. Fortunately, the EQ-5D is easy to administer and score and may not substantially increase respondent burden. However, Fig. 1. Scatterplot showing actual EQ-5D values versus EQ-5D values estimated from the NDI. Fig. 2. Scatterplot showing NDIs plotted against the actual EQ-5D scores demonstrating a floor effect as well as gaps in EQ-5D values. 397

5 L. Y. Carreon et al. the use of existing data sets that only have the NDI cannot be used to determine QALY values with the EQ-5D. Disclosure Dr. Glassman is an employee of Norton Healthcare, has received support of non study-related clinical or research effort from Norton Healthcare, is a patent holder with Medtronic, served as a consultant to Medtronic until July 2011, served on a board at Medtronic until July 2011, and is President of the Scoliosis Research Society (unpaid). Dr. Carreon is an employee of Norton Healthcare; has received research grants from Norton Healthcare and AO Spine paid directly to the Scoliosis Research Society, and from the OREF; previously received funds for travel expenses and accommodations from the OREF, Department of Defense, Association for Collaborative Spine Research, and NIH; has received honoraria for participation in Review Panels for Medtronic and the Children s Tumor Fund; is a member of the Editorial Advisory Board for Spine and Spine Journal; and is a Board Member on the University of Louisville Institutional Review Board. Medtronic provided funds directly to the database company; no funds were paid directly to an individual or an individual s institution (January 2002 to September 2009). NuVasive provides funds directly to the database company; no funds are paid directly to an individual or an individual s institution. Author contributions to the study and manuscript preparation include the following. Conception and design: Carreon. Acquisition of data: all authors. Analysis and interpretation of data: Carreon. Draft ing the article: Carreon. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Carreon. Statistical analysis: Carreon. Administrative/technical/ma terial support: all authors. References 1. Agency for Healthcare Research and Quality: Calculating the U.S. population-based EQ-5D Index Score. Rockville, MD: Agency for Healthcare Research and Quality, 2005 ( [Accessed May 7, 2014] 2. Badia X, Schiaffino A, Alonso J, Herdman M: Using the EuroQoI 5-D in the Catalan general population: feasibility and construct validity. Qual Life Res 7: , Barton GR, Sach TH, Avery AJ, Jenkinson C, Doherty M, Whynes DK, et al: A comparison of the performance of the EQ-5D and SF-6D for individuals aged 45 years. Health Econ 17: , Bloom BS: Use of formal benefit/cost evaluations in health system decision making. Am J Manag Care 10: , Brazier J, Deverill M, Green C: A review of the use of health status measures in economic evaluation. J Health Serv Res Policy 4: , Brazier J, Roberts J, Deverill M: The estimation of a preference-based measure of health from the SF-36. J Health Econ 21: , Brazier J, Roberts J, Tsuchiya A, Busschbach J: A comparison of the EQ-5D and SF-6D across seven patient groups. Health Econ 13: , Carreon LY, Anderson PA, McDonough CM, Djurasovic M, Glassman SD: Predicting SF-6D utility scores from the neck disability index and numeric rating scales for neck and arm pain. Spine (Phila Pa 1976) 36: , Chen F, Curran PJ, Bollen KA, Kirby J, Paxton P: An empirical evaluation of the use of fixed cutoff points in RMSEA test statistic in structural equation models. Sociol Methods Res 36: , Conner-Spady B, Suarez-Almazor ME: Variation in the estimation of quality-adjusted life-years by different preferencebased instruments. Med Care 41: , Devlin N, Hansen P, Herbison P: Variations in self-reported health status: results from a New Zealand survey. N Z Med J 113: , Dickson M, Hurst J, Jacobzone S: Survey of pharmacoeco nom ic assessment activity in eleven countries. OECD Health Working Papers, No. 4 ( ) [Accessed May 7, 2014] 13. Drummond MF, Sculpher MJ, Torrance GW, O Brien BJ, Stoddart GL: Methods for the Economic Evaluation of Health Care Programmes, ed 3. Oxford, UK: Oxford University Press, EuroQol Group: EuroQol a new facility for the measurement of health-related quality of life. Health Policy 16: , Feeny D, Furlong W, Torrance GW, Goldsmith CH, Zhu Z, DePauw S, et al: Multiattribute and single-attribute utility functions for the Health Utilities Index Mark 3 system. Med Care 40: , Franzblau AN: A Primer of Statistics for Non-Statisticians. New York: Harcourt, Gold MR, Siegel JE, Russel LB, Weinstein MC: Cost-Effectiveness in Health and Medicine. New York: Oxford University Press, Herdman M, Gudex C, Lloyd A, Janssen M, Kind P, Parkin D, et al: Development and preliminary testing of the new fivelevel version of EQ-5D (EQ-5D-5L). Qual Life Res 20: , Johnson JA, Pickard AS: Comparison of the EQ-5D and SF- 12 health surveys in a general population survey in Alberta, Canada. Med Care 38: , Kaplan RM: Health-related quality of life in cardiovascular disease. J Consult Clin Psychol 56: , Kind P, Dolan P, Gudex C, Williams A: Variations in population health status: results from a United Kingdom national questionnaire survey. BMJ 316: , Luo N, Johnson JA, Shaw JW, Feeny D, Coons SJ: Self-reported health status of the general adult U.S. population as assessed by the EQ-5D and Health Utilities Index. Med Care 43: , McCaffery M, Beebe A: Pain: Clinical Manual for Nursing Practice. Baltimore: Mosby Co, McDonough CM, Grove MR, Tosteson TD, Lurie JD, Hilibrand AS, Tosteson AN: Comparison of EQ-5D, HUI, and SF- 36-derived societal health state values among Spine Patient Outcomes Research Trial (SPORT) participants. Qual Life Res 14: , McDonough CM, Tosteson TD, Tosteson AN, Jette AM, Grove MR, Weinstein JN: A longitudinal comparison of 5 preference-weighted health state classification systems in persons with intervertebral disk herniation. Med Decis Making 31: , Sach TH, Barton GR, Jenkinson C, Doherty M, Avery AJ, Muir KR: Comparing cost-utility estimates: does the choice of EQ-5D or SF-6D matter? Med Care 47: , Søgaard R, Christensen FB, Videbaek TS, Bünger C, Christiansen T: Interchangeability of the EQ-5D and the SF-6D in long-lasting low back pain. Value Health 12: , Sullivan PW, Lawrence WF, Ghushchyan V: A national catalog of preference-based scores for chronic conditions in the United States. Med Care 43: , Tosteson AN, Lurie JD, Tosteson TD, Skinner JS, Herkowitz H, Albert T, et al: Surgical treatment of spinal stenosis with and without degenerative spondylolisthesis: cost-effectiveness after 2 years. Ann Intern Med 149: , Tosteson AN, Skinner JS, Tosteson TD, Lurie JD, Andersson GB, Berven S, et al: The cost effectiveness of surgical versus nonoperative treatment for lumbar disc herniation over two years: evidence from the Spine Patient Outcomes Research Trial (SPORT). Spine (Phila Pa 1976) 33: ,

6 EQ-5D from the NDI 31. Vernon H: The Neck Disability Index: state-of-the-art, J Manipulative Physiol Ther 31: , Vernon H, Mior S: The Neck Disability Index: a study of reliability and validity. J Manipulative Physiol Ther 14: , Wang H, Kindig DA, Mullahy J: Variation in Chinese population health related quality of life: results from a EuroQol study in Beijing, China. Qual Life Res 14: , Weinstein MC, Siegel JE, Gold MR, Kamlet MS, Russell LB: Recommendations of the Panel on Cost-Effectiveness in Health and Medicine. JAMA 276: , Whitehurst DG, Bryan S: Another study showing that two preference-based measures of health-related quality of life (EQ-5D and SF-6D) are not interchangeable. But why should we expect them to be? Value Health 14: , Whitehurst DG, Bryan S, Lewis M: Systematic review and empirical comparison of contemporaneous EQ-5D and SF-6D group mean scores. Med Decis Making 31:E34 E44, 2011 Manuscript submitted June 13, Accepted May 5, Please include this information when citing this paper: published online June 6, 2014; DOI: / SPINE Address correspondence to: Leah Y. Carreon, M.D., M.Sc., Norton Leatherman Spine Center, 210 E. Gray St., Ste. 900, Louisville, KY leah.carreon@nortonhealthcare.org. 399