Missing values, floor and ceiling effects
The response rate for COMI items was extremely good, with respondents in the general sample answering all of the individual COMI items. This finding is especially meaningful once we compare these results to those for the ODI. In the same sample of 353 individuals, a relatively high number of missing values (N=124; 35.1 %) was found for the “sex life” item of the ODI. This figure is slightly higher than that reported in previous studies; in the Polish COMI validation sample, for example, 23.0 % did not answer the “sex life” question of the ODI. Many other COMI validation studies have also shown a very low number of missing values [6, 17], implying that the COMI items in general as well as in the Slovene version do not ask about information that could be considered as too sensitive (and thus not answered) by the respondent.
While most of the COMI items did not exhibit any meaningful ceiling effects (they were all in the 0-20 % ideal range), the “work disability” item did show a slightly higher proportion of patients reporting the best status, although this was still far below the threshold that could be considered as adverse [14]. Interestingly, the same item has exceeded the ideal range of 0-20 % in many previous COMI validations [6, 13, 20, 21] with some of these studies also showing a greater ceiling effect for the “social disability” item, although this was not observed in our sample. In contrast, the floor effects in the present study were somewhat more prominent and affected five items, namely: “quality of life”, “work disability”, “function”, “social disability”, and, in particular, “symptom-specific well-being”, with values for the latter exceeding 70 % - the value considered adverse [14]. As with ceiling effects, this finding was not unexpected; a similar floor effect for the “symptom-specific well-being” item was also noted in many other COMI validation studies [6, 13, 20, 21] and could be attributed to the fact that the data were those of presurgical patients, who were generally in severe pain and not satisfied "to spend the rest of their life with their current symptoms". Overall, the findings for floor and ceiling effects are in line with those reported in other COMI validation studies, which further serves to support the validity of the Slovene adaptation of the instrument.
Construct validity
To assess the construct validity, the relationships between individual COMI items, the overall COMI score, and an already established questionnaire validated in the Slovene language, the Oswestry Disability Index, were analyzed. To confirm the hypothesis that the instruments measure a similar construct, the Spearman Rho should fall somewhere within the .40-.80 range [22]. In the present study, the overall COMI score correlated very well with the ODI (ρ = .76), demonstrating a relationship that was similar to, but slightly more pronounced than, that reported for the culturally relatively similar Polish version [6] as well as the Brazilian-Portuguese version [13]. The correlations with individual COMI items were also satisfactory; one item (“work disability”) showed a fair correlation with the ODI, three items (“symptom-specific well-being”, “leg pain”, and “social disability”) demonstrated good correlations with the ODI, and the remaining three items (“back pain”, “quality of life”, and “function”) showed correlations with the ODI that can be described as very good. These findings were largely in line with our expectations since previous studies [6] have also observed fair to good correlations between the individual COMI items and the overall ODI score. Overall, these results display a satisfactory construct validity of the Slovene version of the COMI low back questionnaire.
Responsiveness
The responsiveness analysis demonstrated that surgical patients from our sample showed a significant improvement in the median scores of individual COMI items as well as the overall COMI score, approximately three months after surgery. While our study is one of the few in the COMI literature that analyzed responsiveness, our result is very much in line with previous validation studies that did perform such analyses [7, 8, 17]. In the Hungarian validation study, for example, the effect size, pertaining to the change in mean scores 6 months after surgery, was large [17]. The analyses conducted on our sample with a different time frame (3 months) yielded a large effect size as well, showcasing the ability of the Core Outcome Measures Index to detect clinically important changes over time, even when the time period is relatively short.
Furthermore, the additional analyses also revealed that changes in COMI items and the overall COMI score differed between patients who rated the operation as being very helpful/helpful and those who perceived it as less helpful. Specifically, the COMI successfully captured a more pronounced improvement among “good outcome” patients, compared to “poor outcome” patients in all items except one (work disability remained relatively stable in both groups).
Test-retest reliability (stability)
The test-retest reliability of the overall COMI score (Slovene version) was found to be excellent despite the fact that the time lag in between the two COMI applications was significantly longer than in some previous COMI validations. Our findings thus represent an important contribution to the existing literature, supporting Mannion and colleagues [8] who found a significant reduction in COMI scores from pre-surgery to 3-months post-surgery, with the values then remaining stable up to 2 years after surgery. In our case, we have also observed a significant reduction in COMI scores post-surgery (the responsiveness part) and found that these values remained relatively unchanged approximately half a year after the therapeutic intervention (the reliability part; [23]).
Lastly, the minimum detectable change of the Slovene COMI total score (2.4) was slightly higher than that published for the Hungarian (1.6; [17]), Brazilian-Portuguese (1.7; [13]), and Polish version (1.8; [6]), but relatively similar to the French (2.0; [20]) and Norwegian version (2.2; [21]). As such, the MDC for the sum scale is only marginally poorer than in some former studies and within range of that reported for other low back pain outcome instruments [24, 25]. The Slovene version of COMI thus exhibits acceptable minimum detectable change, meaning that a change of more than 2.4 points at the COMI index needs to be observed to be labeled as a real change (and not the measurement error). However, as the MDC is largely dependent on the ICC and standard deviation (SD), future studies should investigate test-retest reliability and MDC on a sample of stable patients who fill out the questionnaire approximately 1-2 weeks apart instead of a longer time period as used in the present study.