In this study, the decisions of chiropractic clinicians to recommend secondary and tertiary preventive care, MC, for recurrent and persistent LBP were tested using theoretically defined indications. We tested models in which three, two, one and none of the indications were used. We propose that, in clinical reality, this information is weighed together consciously or subconsciously to form a clinical decision.
The results largely confirm the findings of the previous studies in the area [10, 13, 14]. That is, the theoretical construct previously identified was found to reflect reality. The clinical encounter is always tailored to the individual patient, but clinicians are clearly using some overarching principles when recommending MC. The clinicians in this study weigh these factors together when deciding on MC. The presence of many previous episodes was found to be the main indicator for such care in the clinical encounter. This might suggest that clinicians are viewing MC mainly as secondary prevention aimed at preventing future episodes, and is in line with the MC intent described in previous studies [10, 13, 14].
The accuracy of the predictive models was examined using ROC curves. The area under the ROC curve for the factor “many previous episodes”, suggests that the predictive accuracy of this model is better than “long duration” and far better than “definite improvement”. It is interesting that the sensitivity of the best model is less accurate than the specificity. Thus, the absence of many previous episodes more accurately predicts a decision not to recommend MC (specificity 0.72), than the presence of many previous episodes predicts a MC recommendation (sensitivity 0.53). From previous studies, it is known that clinicians consider a number of factors before recommending MC, factors such as psychosocial situation, work demands, patient motivation and so on . We did not record these variables in this study, nor can we know if clinicians use some other, maybe tacit, knowledge in their decision process.
By adding the clinicians as a factor in the multi-level regression model, the model fit was improved. We conclude that the heterogeneity among chiropractors in regards to recommending MC is substantial. This is also in line with the findings of a previous study .
Further, the initial hypothesis was not confirmed in full, as not all patients with three predictors were regarded as MC candidates and a majority of patients with no predictors also were given this recommendation (keeping in mind that the latter is a small group). Again, it is possible that some other unknown or unrecorded variables were considered in these cases, making the decisions go either way depending on the type and presence or absence of that information. This could possibly explain the fact that even the patients with no predictors to a large extent (73%) were regarded as MC candidates. A previous study explained the clinician’s intent of continuing treatment despite the lack of progress in terms of taking on the role as a health coach . Further, a recent consensus process among the chiropractic profession described “wellness care” with a primary preventive intent: to promote general health including counseling on behavior related to diet, exercise and tobacco . We did not investigate these aspects of the MC decision, and this subgroup (with no predictors) was very small, rendering conclusions subject to caution.
It is important to note that we do not know if the patients involved in this study that were deemed “MC candidates” were actually given the MC recommendation, if they accepted it and what the outcome of that preventive strategy was.
The major limitation of this study is the scarcity of variables, which is a result of time restrictions in the clinic. The objective was to test a theoretical construct, which was possible using the available data which are part of the normal clinical encounter. However, it would have been possible to add an explorative element to the study with data concerning psychosocial factors, motivation, work demands etc. which would possibly add an explanatory value to the results. Still, as the main hypothesis was largely confirmed, the theoretical construct was found to be reflective of reality to some extent. The results are also restricted by the detail of the available data. Had more categories been added to previous duration and previous episodes, detailed associations regarding subgroups may have been explored. Both variables are self-reported and may be subject to memory bias. For previous duration, the cut point of 30 days the previous year has been used in several studies [26–28] and found to be useful in separating patients with good and poor long term prognosis. For episodes, no evidence-based definition exists , and the decision to ask the patients to remember whether they had many (≥4) or fewer (<4) was based upon discussions with clinicians.