Kosloff et al. Chiropractic & Manual Therapies (2015) 23:19
Page 8 of 10
measures, albeit more direct surrogate measures, than simply using the exposure to chiropractic visits. Our study was also limited to replication of the case – control design described by Cassidy, et al. [32]. For pragmatic reasons, we did not attempt to conduct a case-crossover design. While the addition of a case- crossover design would have provided better control of confounding variables, Cassidy, et al. [32] showed the results were similar for both the case control and case crossover studies. The findings of this case – control study and previous retrospective research underscore the need to rethink how to better conduct future investigations. Researchers should seek to avoid the use of surrogate measures or use the least indirect measures available. Instead, the focus should be on capturing data about the types of ser- vices and not the type of health care provider. In alignment with this approach, it is also important for investigators to access contextual data (e.g., from electronic health records), which can be enabled by qualitative data analysis computer programs [58]. The acquisition of the elements of clinical encounters – in- cluding history, diagnosis, intervention, and adverse events – can provide the infrastructure for more action- able research. Because of the rarity of VBA stroke, large data sets (e.g., registries) containing these elements will be necessary to achieve adequate statistical power for making confident conclusions. Until research efforts produce more definitive results, health care policy and clinical practice judgments are best informed by the evidence about the effectiveness of manipulation, plausible treatment options (including non-thrust manual techniques) and individual patient values [20]. Conclusions Our findings should be viewed in the context of the body of knowledge concerning the risk of VBA stroke. In contrast to several other case – control studies, we found no significant association between exposure to chiropractic care and the risk of VBA stroke. Our sec- ondary analysis clearly showed that manipulation may or may not have been reported at every chiropractic visit. Therefore, the use of chiropractic visits as a proxy for manipulation may not be reliable. Our results add weight to the view that chiropractic care is an unlikely cause of VBA strokes. However, the current study does not exclude cervical manipulation as a possible cause or contributory factor in the occurrence of VBA stroke. Competing interests All authors are employees of UnitedHealth Group – a U.S based commercial health care company. The authors declare that they have no other competing interests.
[43,46]. Cassidy, et al. [32] conducted a sensitivity analysis to determine the effect of diagnostic misclassification bias. Their conclusions did not change when the effects of mis- classification were assumed to be similarly distributed be- tween chiropractic and PCP cases. A particular limitation in using administrative claims data is the paucity of contextual information surround- ing the clinical encounters between chiropractors/PCPs and their patients. Historical elements describing the oc- currence/absence of recent trauma or activities reported in case studies [47-51] as potential risk factors for VBA stroke were not available in claims data. Confidence was low concerning the ability of claims data to provide ac- curate and complete reporting of other health disorders, which have been described in case – control designs as being associated with the occurrence of VBA stroke e.g., migraine [52] or recent infection [53]. Symptoms and physical examination findings that would have permitted further stratification of cases were not reported in the claims data. The reporting of clinical procedures using current pro- cedural terminology (CPT) codes presented additional shortcomings concerning the accuracy and interpretation of administrative data. One inherent constraint was the lack of anatomic specificity associated with the use of standardized procedural codes in claims data. Chiropractic manipulative treatment codes (CPT 98940 – 98942) have been formatted to describe the number of spinal regions receiving manipulation. They do not identify the particular spinal regions manipulated. Also, treatment information describing the type(s) of manipulation was not available. When SMT was re- ported, claims data could not discriminate among the range of techniques including thrust or rotational ma- nipulation, various non-thrust interventions e.g., mech- anical instruments, soft tissue mobilizations, muscle energy techniques, manual cervical traction, etc. Many of these techniques do not incorporate the same bio- mechanical stressors associated with the type of mani- pulation (high velocity low amplitude) that has been investigated as a putative risk factor for VBA stroke [54-56]. It seems plausible that the utility of future VBA stroke research would benefit from explicit descriptions of the particular type of manipulation performed. Moreover, patient responses to care – including any adverse events suggestive of vertebral artery dissection or stroke-like symptoms – were not obtainable in the data set used for the current study. In the absence of performing comprehensive clinical chart audits, it is not possible to know from claims data what actually transpired in the clinical encounter. Fur- ther, chart notes may themselves be incomplete or other- wise fail to precisely describe the nature of interventions [57]. Therefore, manipulation codes represent surrogate
Made with FlippingBook HTML5