The use of patient-reported outcomes instruments to gauge recovery after treatment and satisfaction with care has become more common. Reports in the medical literature often cite mean differences between pretreatment and posttreatment scores or between two different treatment groups. The reliance on mean differences does not provide the reader with the ability to incorporate the use of patient-reported outcomes into clinical practice.
The minimum clinically important difference has been defined as the smallest change that is meaningful to patients1,2. This difference has been estimated for a number of commonly used patient-reported outcomes measures, including the Short Form-36 (SF-36) and the Neck Disability Index (NDI). The SF-36 is a generic health status measure that was developed to provide an estimate of an individual's physical and mental health. It has been estimated that the minimum clinically important difference in SF-36 physical health component scores can range from 3 to 6 points3,4. The NDI is a disease-specific, patient-reported outcomes instrument that is used to assess an individual's level of disability. The minimum clinically important difference in NDI scores has been estimated to be 7.5 points5,6.
The Cervical Spine Outcomes Questionnaire (CSOQ) (see Appendix) is a multidimensional patient-reported outcomes instrument that was developed to specifically assess outcomes after anterior cervical spine decompression and arthrodesis in a Cervical Spine Research Society-funded outcomes study. The CSOQ provides estimates of neck and arm/shoulder pain, physical symptoms other than pain, functional disability, psychologic distress, and health-care utilization. Previous reports have shown the measure to have good psychometric properties7 and utility for tracking patient recovery8. We are not aware of any study in which the investigators estimated the minimum clinically important difference for the domain scores of the CSOQ.
Our goal was to determine the CSOQ scores that represent a minimum clinically important difference and a substantial clinical benefit for patients treated with anterior cervical spine decompression and arthrodesis for cervical degenerative disc disease. We also tested the influence of patient and clinical characteristics on change in CSOQ domain scores.
This study was approved by our institutional review board.
Patient Population
This study is a retrospective review of data that were prospectively collected from a cohort of patients undergoing anterior cervical spine decompression and arthrodesis at one of twenty-three national sites enrolled in the Cervical Spine Research Society Outcomes Study. In this study, operating surgeons were orthopaedists or neurosurgeons at centers that specialized in spine care. Inclusion criteria for the Cervical Spine Research Society Outcomes Study were the presence of cervical radiculopathy or myelopathy, use of an anterior surgical approach, adequate mental capacity to provide informed consent, patient age of more than eighteen years, complete enrollment information, and the ability to return for follow-up assessment. An additional criterion for inclusion in the current study was that the individual had to have provided complete information on the SF-36, the NDI, and the CSOQ at baseline and the six-month postoperative assessment. Of the 622 patients who had undergone surgery, 370 were excluded for the following reasons: posterior cervical spine surgery, 126; missing enrollment forms (including completion of the NDI, SF-36, and CSOQ), ninety-eight; missing six-month forms, eighty-five; absence of cervical radiculopathy or myelopathy, thirty-six; and requested withdrawal from the study, twenty-five. The remaining 252 individuals (40.5%) met all of the criteria and formed our study group.
This sample of participants tended to be middle-aged (mean age at surgery, 49.8 ± 10.4 years), white (87%), and well-educated (median education, fourteen years; interquartile range, eleven to fifteen years). Both sexes were equally represented (male, 55.6%); 142 (56.3%) of the participants reported smoking cigarettes at some point in the past, with fifty-one (20.2%) reporting current tobacco use. Eight participants (3.2%) were receiving unemployment benefits, thirty-four (13.5%) reported receiving Workers’ Compensation payments, and twenty-nine (11.5%) were receiving Social Security disability benefits. Of the 252 participants, 103 (40.9%) underwent single-level cervical spine surgery; ninety-seven (38.5%), two-level surgery; and the remaining fifty-two (20.6%), surgery involving three or more levels. Radiculopathy was the primary problem in 153 individuals (60.7%); myelopathy, in sixty-one (24.2%); and axial pain, in the remaining thirty-eight (15.1%).
Assessment Instruments
Baseline and postoperative evaluations included use of the NDI, the SF-36, and the CSOQ.
The NDI is a disease-specific instrument that is used to assess the impact of spinal disorders on ten aspects of daily living9. The NDI has excellent retest reliability (Pearson r > 0.80) and validity (moderately high correlations with the McGill Pain Questionnaire and a visual analog scale for pain). An expert panel convened by the journal Spine recommended use of this instrument because of its good psychometric properties5.
The SF-36 is a rater-administered survey that is used to assess the severity of an individual's functional disability related to his or her physical or emotional health10. This instrument is a reliable measure of disability and general health status and has been used extensively in similar populations. The SF-36 has excellent retest reliability (Pearson r = 0.93)11.
We used an anchor-based method that corresponded to a minimum clinically important difference and a substantial clinical benefit12. The health transition item of the SF-36, not used in calculating the SF-36 aggregate health scores, was used as the anchor. This item asks respondents: "Compared with one year ago today, how would you rate your health in general now?" Answer choices include "much worse," "somewhat worse," "about the same," "somewhat better," and "much better." Participants responding "somewhat better" and "much better" were considered to be experiencing a clinically important improvement in self-reported health outcome. Participants responding "much better" were considered to be experiencing a clinically important improvement in self-reported health outcome.
The CSOQ measures patient-reported outcomes across several broad factors, including pain severity (in the neck and arms/shoulders), functional disability, psychologic distress, physical symptoms other than pain, and health-care utilization (Table I)8. The CSOQ is a valid and reliable outcomes instrument (Pearson r > 0.80) for patients undergoing cervical spine surgery7.
Statistical Analysis
To calculate the minimum clinically important difference, we used receiver operating characteristics curve analysis to plot the sensitivity and specificity for each patient-reported outcome. We chose the optimal cutoff for the receiver operating characteristics curve to maximize the sensitivity and specificity12. This value represented the minimum clinically important difference. We also used receiver operating characteristics curve analysis and chose the optimal cutoff to calculate the substantial clinical benefit. We used the minimum clinically important difference and the substantial clinical benefit to classify each individual and compared those classifications with the anchor stratification.
In addition, the standard error measure was calculated to represent the precision of the measurement for each patient-reported outcome instrument in this population. The standard error measure is calculated as the product of the sample standard deviation and the square root of one minus the test-retest reliability of the instrument. Published test-retest reliability estimates were used for each scale: 0.8 for the NDI9, 0.93 for the SF-364, and 0.90 for the CSOQ7.
A secondary analysis was performed to determine the influence of patient and clinical factors on change in CSOQ domain scores. The patient factors were age and sex, and the clinical factor was the duration of current symptoms. We used a t test to compare the mean changes in the CSOQ domain scores among patients stratified by sex. Spearman correlation coefficients were estimated to examine the relationship between the change in the CSOQ domain scores and age and duration of current symptoms.
Significance was set at p < 0.05.
Source of Funding
Data for this report were collected in the Cervical Spine Research Society Outcomes Study, which is funded through the Cervical Spine Research Society with contributions from Aesculap, DePuy Spine, Medtronic Sofamor Danek, Stryker Spine (Osteonics), and Synthes Spine.
Outcome studies frequently rely on reports of the mean change on self-reported instruments measuring physical and mental health to denote the clinical benefit of a certain procedure. Without clear metrics against which to judge these changes to clinical practice, these outcomes studies have limitations. Minimum clinically important difference and substantial clinical benefit thresholds provide the metric by which researchers and clinicians can gauge mean changes in scores on self-reported instruments of health. The current study demonstrated that, among individuals who have undergone anterior cervical spine decompression and arthrodesis, the changes in the CSOQ domain scores representing a minimum clinically important difference and a substantial clinical benefit were 0.13 to 0.24 point and 0.20 to 0.30 point, respectively. In each case, these metrics exceed the standard error of measurement for these domain scores, indicating that the minimum clinically important difference and substantial clinical benefit values are greater than what could be accounted for by measurement error. Thus, these values represent a true change in the domain scores associated with clinical benefits. Women experienced a greater decrease in neck pain after surgery than did men, but we found no sex-related difference in the other CSOQ domain scores. Baseline preoperative age was not associated with the change in domain scores. An increase in the duration of current symptoms was associated with a smaller improvement in domain scores from baseline for neck pain and arm/shoulder pain, but not for the other domains.
There are three major methods for estimating the minimum clinically important difference and substantial clinical benefit: distribution-based, opinion-based, and anchor-based. With distribution-based methods, these clinical metrics are estimated on the basis of the distribution of scores in an untreated/unaffected population. Although these metrics are easy to calculate, limitations of this method include variability across studies and the absence of a clinical component13. Opinion-based methods are based on a consensus between patients and experts regarding minimum clinically important differences. Although expert opinion regarding changes in outcome measures may be useful, this method is rarely used. With anchor-based methods, an external patient-reported measure of change in status (e.g., current health state) is used to determine these metrics. We used the anchor-based method to determine the minimum clinically important difference and substantial clinical benefit with use of receiver operating characteristics curves.
Other investigators have estimated the minimum clinically important difference for the NDI3,5. The values that we calculated are higher than those reported by Cleland et al.5. It is important to note that those authors studied patients undergoing nonoperative care for spinal abnormalities5. Differences among our study, the study by Cleland et al.5, and the study by Carreon et al.3 may be related to greater expectations for benefit on the part of patients undergoing surgical treatment, a longer follow-up time, or greater disability at baseline. Carreon et al. reported the minimum clinically important difference on the NDI in a population of patients who had undergone cervical spine surgery and had been followed for a minimum of one year after that surgery. Our findings indicate greater values for the minimum clinically important difference than those that they reported.
There are several limitations to our study. First, minimum clinically important difference and substantial clinical benefit values are specific to the populations in which they are developed. Thus, our findings are applicable to those undergoing anterior cervical spine decompression and arthrodesis for cervical degenerative disc disease but may not be applicable to patients undergoing nonoperative care or other types of surgical intervention. Second, we used baseline preoperative and six-month postoperative assessments to generate the minimum clinically important difference and substantial clinical benefit values. The thresholds reported in this study may not be appropriate for evaluation of shorter-term or longer-term change in domain scores on the CSOQ. Third, we did not examine threshold values to identify individuals who experienced a worsening of health status because of the limited number of patients who reported that finding when answering our anchor question.
Patients with a 0.20-point change in neck pain or a 0.24-change in arm/shoulder pain experienced a clinically important improvement in the self-reported outcome after surgery for cervical degenerative disc disease. Similarly, patients with a 0.30-point change in neck pain or arm/shoulder pain experienced a substantial clinical benefit after surgery.
Our study has provided metrics with which changes in the domain scores of the CSOQ can be compared to denote a minimum clinically important difference and a substantial clinical benefit. This instrument provides clinicians and researchers with an additional valid and reliable multidimensional outcomes measure with which to evaluate the progress of their patients after surgery. This information, coupled with previous reports of the psychometric stability of the CSOQ, should increase the clinical utility of this patient-reported outcomes instrument in cervical spine surgery.