The Mechanical and Inflammatory Low Back Pain (MIL) Index
The Mechanical and Inflammatory Low Back Pain (MIL) Index
Content Validity. The 27 signs and symptoms items were reduced to an initial set of 11 through panel feedback and consensus agreement as detailed (Table 1). The reduction to the final set of seven items was achieved through factor analysis where four items were removed to leave the final MIL 7-item version. Two content validity index calculations were performed on both the items and the complete questionnaire to determine whether an item would be removed due to cross-loading (the presence of an item in both dimensions where loading is > 0.40).
Face Validity. All panel members agreed on the MIL being suitably indicative of a questionnaire to determine the presence of mechanical or inflammatory symptoms. All participants were able to complete the MIL without missing responses or additional assistance.
Psychometric Characteristics. Factor Analyses: Four items presented at >0.40 in both dimensions and these items were removed for cross-loading: "Pain when standing for a while"; "Pain on trunk extension"; "Palpatory pain of muscles"; and "Pain getting out of a chair". A flow chart of how the final MIL version was constructed and reduced from the initial 27-items to 7-items is presented (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart of how the final MIL was developed from the initial 27-item version to the final 7-item version.
The Kaiser-Meyer-Olkin measure produced a coefficient of 0.68, indicative of sampling adequacy, and the Bartlett's Test of Sphericity reached statistical significance (p < 0.001). Both supporting the factorability of the correlation matrix. There were 'two factors' prior to the 'inflection' point in the scree test with Eigenvalues >1.0, item-variance >5%, and a total cumulative variance of 51.7%. The rotated 'two-factor' solution showed strong loadings (Table 2).
The CFA of the two-factor model yielded a non-significant χ-test (χ2 = 14.80, df = 13, p = 0.37). The other fit indices were very satisfactory (NFI = 0.97, CFI = 0.98, and RMSEA = 0.029) (Figure 2) and the factor loadings of all variables were >0.40. The correlation coefficient between the two dimensions of 0.56 suggests a moderate relationship.
(Enlarge Image)
Figure 2.
The pathways, factor loading and goodness-of-fit indexes of the two-factor structure underlying the MIL.
Correlations between item-total factor Kendall's Tau are shown in Table 3. The items "morning pain on waking" and "pain on repetitive bending", both correlate highly with the ILBP component of the MIL questionnaire; while "pain on trunk flexion" and "pain on lateral bending" are factors more related to the MLBP component (Table 3).
Cronbach's α for the MLBP and ILBP factors was modest, being respectively at 0.68 and 0.72. The development of a combined index is justified given that the two factors are significantly and moderately associated. The MIL index is a pragmatic sum of the standardized scores with regression analysis of the two factors.
Baseline responses demonstrated normalized distribution for the 7-items. Normality was determined and means and variability of all measures are represented (Table 4). The consistency of the ILBP index, MLBP index and MIL score over time was high (ICC = 0.91; 95%CI =0.88–0.93, ICC = 0.93; 95%CI =0.90–0.96; ICC = 0.89; 95%CI =0.86–0.91, respectively).
The median score for the MLBP Index was 0.504. The 20, 40, 60 and 80 percentiles were −0.377, -0.097, 0.577 and 0.713 respectively. The median score for the ILBP Index was −0.344. The 20, 40, 60 and 80 percentiles were −1.028, -0.443, -0.055 and 1.159 respectively. MIL index are calculated as the sum of the standardised scores (MLBP and ILBP) and the values can be classified in five categories: very low, low, average and high, very high. The median score for MIL Index was 0.129. The 20, 40, 60 and 80 percentiles were −1.21, -0.296, 0.402 and 1.515 respectively.
The correlations between the factor ILBP, and RMQ and BAI measurements were practically identical but weak (r = 0.34, p < 0.001). The instruments that correlated weakly with the MLBP were the PCS, RMQ and BAI (r = 0.38, p < 0.001). Taking the factors ILBP and MLBP together, a significant but weak correlation is seen with the BAI and the RMQ, but virtually non-existent with the SF-12 PCS and SF-12 MCS (Table 5) apart from a very weak correlation with the PCS value and the combined MIL score.
The ROC analyses indicated that the AUCs (expressed in 95% confidence interval) for the specific low back pain questionnaires were from 0.74–0.92 for the RMQ and 0.51–0.65 for the BAI. In general, no significances were noted with the exception of the ILBP and the ILBP plus MLBP factors in the case of the RMQ value of state variable at 20%.
Readability was acceptable with a Flesch-Kincaid grade level at 6.8 and 68.5% reading ease.
Missing responses were acceptable with four responses found in three questions (1, 2, and 4) at a frequency of 5%. Completion time was 6.57 ± 3.03 minutes.
Results
Phase 1: The MIL Development
Content Validity. The 27 signs and symptoms items were reduced to an initial set of 11 through panel feedback and consensus agreement as detailed (Table 1). The reduction to the final set of seven items was achieved through factor analysis where four items were removed to leave the final MIL 7-item version. Two content validity index calculations were performed on both the items and the complete questionnaire to determine whether an item would be removed due to cross-loading (the presence of an item in both dimensions where loading is > 0.40).
Face Validity. All panel members agreed on the MIL being suitably indicative of a questionnaire to determine the presence of mechanical or inflammatory symptoms. All participants were able to complete the MIL without missing responses or additional assistance.
Phase 2: MIL Validation
Psychometric Characteristics. Factor Analyses: Four items presented at >0.40 in both dimensions and these items were removed for cross-loading: "Pain when standing for a while"; "Pain on trunk extension"; "Palpatory pain of muscles"; and "Pain getting out of a chair". A flow chart of how the final MIL version was constructed and reduced from the initial 27-items to 7-items is presented (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart of how the final MIL was developed from the initial 27-item version to the final 7-item version.
The Kaiser-Meyer-Olkin measure produced a coefficient of 0.68, indicative of sampling adequacy, and the Bartlett's Test of Sphericity reached statistical significance (p < 0.001). Both supporting the factorability of the correlation matrix. There were 'two factors' prior to the 'inflection' point in the scree test with Eigenvalues >1.0, item-variance >5%, and a total cumulative variance of 51.7%. The rotated 'two-factor' solution showed strong loadings (Table 2).
The CFA of the two-factor model yielded a non-significant χ-test (χ2 = 14.80, df = 13, p = 0.37). The other fit indices were very satisfactory (NFI = 0.97, CFI = 0.98, and RMSEA = 0.029) (Figure 2) and the factor loadings of all variables were >0.40. The correlation coefficient between the two dimensions of 0.56 suggests a moderate relationship.
(Enlarge Image)
Figure 2.
The pathways, factor loading and goodness-of-fit indexes of the two-factor structure underlying the MIL.
Correlations between item-total factor Kendall's Tau are shown in Table 3. The items "morning pain on waking" and "pain on repetitive bending", both correlate highly with the ILBP component of the MIL questionnaire; while "pain on trunk flexion" and "pain on lateral bending" are factors more related to the MLBP component (Table 3).
Cronbach's α for the MLBP and ILBP factors was modest, being respectively at 0.68 and 0.72. The development of a combined index is justified given that the two factors are significantly and moderately associated. The MIL index is a pragmatic sum of the standardized scores with regression analysis of the two factors.
Baseline Responses and Test–Retest Reliability
Baseline responses demonstrated normalized distribution for the 7-items. Normality was determined and means and variability of all measures are represented (Table 4). The consistency of the ILBP index, MLBP index and MIL score over time was high (ICC = 0.91; 95%CI =0.88–0.93, ICC = 0.93; 95%CI =0.90–0.96; ICC = 0.89; 95%CI =0.86–0.91, respectively).
Normal Reference Values as Standardized Scores of the Mechanical and Inflammatory Low Back Pain (MIL) Index
The median score for the MLBP Index was 0.504. The 20, 40, 60 and 80 percentiles were −0.377, -0.097, 0.577 and 0.713 respectively. The median score for the ILBP Index was −0.344. The 20, 40, 60 and 80 percentiles were −1.028, -0.443, -0.055 and 1.159 respectively. MIL index are calculated as the sum of the standardised scores (MLBP and ILBP) and the values can be classified in five categories: very low, low, average and high, very high. The median score for MIL Index was 0.129. The 20, 40, 60 and 80 percentiles were −1.21, -0.296, 0.402 and 1.515 respectively.
Convergent Validity
The correlations between the factor ILBP, and RMQ and BAI measurements were practically identical but weak (r = 0.34, p < 0.001). The instruments that correlated weakly with the MLBP were the PCS, RMQ and BAI (r = 0.38, p < 0.001). Taking the factors ILBP and MLBP together, a significant but weak correlation is seen with the BAI and the RMQ, but virtually non-existent with the SF-12 PCS and SF-12 MCS (Table 5) apart from a very weak correlation with the PCS value and the combined MIL score.
Discriminant Validity
The ROC analyses indicated that the AUCs (expressed in 95% confidence interval) for the specific low back pain questionnaires were from 0.74–0.92 for the RMQ and 0.51–0.65 for the BAI. In general, no significances were noted with the exception of the ILBP and the ILBP plus MLBP factors in the case of the RMQ value of state variable at 20%.
Practical Characteristics
Readability was acceptable with a Flesch-Kincaid grade level at 6.8 and 68.5% reading ease.
Missing responses were acceptable with four responses found in three questions (1, 2, and 4) at a frequency of 5%. Completion time was 6.57 ± 3.03 minutes.
Source...