Abstract
BACKGROUND:
The Oswestry Disability Index (ODI) is one of the most common condition specific outcome measures used in the management of spinal disorders. But there is insufficient study on healthy
populations and long term test-retest reliability. This is important because healthy populations are often used for control groups in low back pain interventions, and knowing the reliability of the controls affects the interpretation of the findings of these studies.
OBJECTIVE:
The purpose of this study is to determine the long term test-retest reliability of ODI in office workers.
METHODS:
Participants who have no chronic low back pain history were included in study. Subjects were assessed by the Turkish-ODI 2.0 (e-forms) on 1st, 2nd, 4th, 8th, 15th, 30th days to determine the stability of ODI scores over time. The study began with 58 (12 female, 46 male) participants. 36 (3 female, 33 male) participated for the full 30 days.
STATISTICS:
Kolmogorov-Smirnov and Friedman tests were used. Test-retest reliability was evaluated by using nonparametric statistics. All tests were done by using SPSS-11.
RESULTS:
There was no statistically significant difference among the median scores of each day. (χ= 6.482, p > 0.05).
CONCLUSION:
The difference between median score of the days with 1st day was neither statistically nor clinically significant. ODI has long term test re-test reliability in healthy subjects over a 1 month time interval.
Introduction
The Oswestry Disability Index (ODI) is one of the most common condition specific outcome measure used in the management of spinal disorders. ODI measures the intensity of pain and effect of low back pain (disability) on nine different daily living activities (lifting, ability to care for oneself, ability to walk, ability to sit, sexual function, ability to stand, social life, sleep quality, and ability to travel). Each question (subscale) consists of 6 statements which describes least amount (0) of to severe disability (5). Patient performs a self-assessment by selecting the a statement for each question which most closely resembles his/her situation. Disability is evaluated by calculating total score. The total score can vary from 0–50 or 0% to 100% . Total score from 0% – 20% reflects minimal disability, 21% – 40% reflects moderate disability, 41% – 60% reflects severe disability. 61% -80% and 81% – % 100 are the high score bands.
Due to its widespread use there are a large number of studies in the literature related to ODI. ODI studies can be divided in to: Development studies (the studies related to version revision and the basic psychometric properties of ODI) (S2) [2, 13], cross cultural adaptation and translation studies (S1) [4, 15], use of ODI as an outcome measure on different populations (S3) [6, 11], technology adaptation studies (S4) [1].
Since 1980, studies in all groups have been developed in a natural scientific progression. But in 2000, Fairbank who is the developer of ODI and one of the important scientist on ODI studies clearly stated the need for studies on healthy subjects [12].
In ODI studies, healthy subjects are included only in control groups [6, 7]. So, Fairbank may have had doubts on the control group response to ODI. In physiotherapy practice, therapies related to low back pain takes 10 to 21 days. If a home exercise program is prescribed, the patient will be followed up to 6–8 weeks. This may be a very long time period because the test re-test studies of ODI have not been evaluated for more than two weeks. In this time period the stability of the control group in time may be lost and test re-test reliability may be corrupted.
Also in previous studies fundamental statistical properties of healthy controls like distribution were not clearly stated [6, 7]. There may be a floor and ceiling effect; Healthy controls may cluster at a point like zero or clinically non-significant score value (less than 9) and may not have a normal distribution.
Also there are additional factors which may affect long term test re-test reliability like menstrual disorders. Menstrual disorders on ODI in long time period is not well known but Smith et al. found evidence for influence of menstrual disorders on low back pain in Japanese Nurses [16].
The purpose of this study is to investigate the long term stability of healthy male office workers’ response to ODI.
Material and method
Population
Participants who have no chronic low back pain history were included in this study. Participants were excluded who had: low back pain history, low back pain during the study, menstrual disorder related low back pain, or previous operations or physical therapy related to low back pain.
The incidence of low back pain was evaluated via an analogue visual scale prior to the study to make sure that participants met the inclusion criteria. The subjects with non zero score were excluded from the study. Participants who could not speak Turkish fluently could not complete questionnaires by themselves, or who do not want to join the study were also excluded.
The paper based Turkish 2.0 version of ODI was used. Questionnaires were given to participants at day 1, 2, 4, 8, 15, 30.
Most of the subjects were administrative staff members who worked 8 hours per day while sitting and performing most of their tasks by using computers. The subjects whose questionnaires were not returned in the same day were excluded from the study.
The study was approved by the ethics committee of Hacettepe University. All subjects were informed about the study and written consent was given.
Reliability
In this study more than one time interval (on 1st, 2nd, 4th, 8th, 15th, 30th days) was used to determine the stability of ODI scores.
Statistics
There is not enough knowledge about fundamental statistical properties of ODI response of healthy population. By this reason a statistical analysis algorithm was designed to cover two possible cases. In first step a Kolmogorov-Smirnov test was used to check whether the data was normally distributed or not.
If distribution of ODI responses are normal and there is no floor and ceiling effect a parametric approach will be enough. In this situation a test for investigation of statistical significance of means in dependent samples (ANOVA) can be sufficient to investigate the statistical significance of differences between ODI day scores on the 1st day through the 30th day. In the same way test-retest reliability can be evaluated by correlation coefficients (Pearson correlation coefficient or intraclass correlation coefficient).
If distribution of ODI responses are not normal and there is a floor and ceiling effect a non-parametric approach have to be used. Median and number of participants in each quartiles will have significance. In this situation a Friedman test will be used to investigate the statistical significance of the differences between the ODI scores on the 1st day through the 30th day.
All tests will be done by using SPSS 11.
Results
At the end of this study 36 (3 female, 33 male) out of 58 (12 female, 46 male) subjects were left. Fallout was due to 16 subjects whose questionnaires were not retrieved in the same day and 6 subjects whose questionnaires were not properly answered. ODI responses are not normally distributed. There is a floor and ceiling effect which was evaluated by a non-parametric approach. The mean score of 216 questionnaires (6×36 = 216) was 1.9 ± 3.6 (median 0, IQR 0–2) (See Table 1). The total possible score is 50, the mean scores are very low (in the 0–20% range), as would be expected from healthy participants. The scores of all subjects were between 0 and 20. ODI daily scores were not normally distributed
Discussion
The scores in our study were not normally distributed. This is a previously expected result on healthy subjects. It was expected that ODI scores of healthy subjects have values near to zero and this will cause a distortion on distribution curve. For this reason, statistical analysis should be done by non parametric approaches. In previous studies researches did not check the normal distribution of healthy subject’s ODI scores. They directly used interclass correlation coefficient (ICC) to test the test retest reliability. But normal distribution is one of the important conditions to use ICC to test the test retest reliability.
The mean score in our study (1.9 ± 3.6) was lower than the previous studies (10, min 2.2 and max 12). 56% – % 72 of the subjects had zero score and 92% – 100% of the subjects’ score were less then or equal to 10 (Table 1). This showed ODI score of healthy population was accumulated in the 0 –≤10 interval. Score changes from baseline was between –0.1 ± 3 and 0.9 ± 3 (Table 2 and Fig. 1). The reported clinical significant scores of ODI were between 4 and 15 [6]. The changes were so small and clinically not significant.
Limitations
The most important limitation of this study is fewer female participants were included. The reason is menstrual pain reported by female participants in pre interviews which may distort the data distribution. Future studies may focus on stability of ODI in female population or possible errors in ODI scores associated with menstrual pain.
Conclusions
The most important finding of this study was that the difference between median score of day 1 was neither statistically nor clinically significant from days 2, 4, 8, 15, or 30. Also some results are obtained which may contribute to clarification of fundamental statistical properties of healty population. ODI scores of healthy participants was not normally distributed and there is a floor and ceiling effect. The conclusion is that ODI has long term test re-test reliability in healthy subjects over a 1 month time interval.