To the Editor:
We read the article by McIsaac et al.1 with great interest. The authors should be commended for attempting to estimate the effects of peripheral nerve blocks (PNBs) on healthcare resource use. These efforts could decrease the cost of health care without compromising patient health. However, we have a few points that we wish to pose to the authors, which may confound interpretation of the results.
First, PNBs are widely used to reduce pain after total knee arthroplasty (TKA). However, these techniques have shortcomings, such as inadequate pain control due to technical difficulty and inexperience. Multimodal analgesia has been introduced to overcome these shortcomings.2 The pain score is important to determine whether a nerve block is successful, but this retrospective design made it impossible to include pain scores.
Second, factors contributing to length of stay after TKA include preoperative, intraoperative, and postoperative variables. Elderly patients are more prone to postoperative complications. It is well documented that length of stay is associated with postoperative complications, such as cardiovascular complications, mechanical wounds, and infections.3 These variables may affect the results. However, these variables are not included in the analysis.
Third, the use of propensity score methods has increased significantly in recent years to evaluate treatment effects using observational data. These methods allow observational studies to be designed similar to randomized experiments. Four methods of using the propensity score have been described in the statistical literature, including matching, stratification, covariate adjustment, and weighting (inverse probability of treatment weighting; IPTW). It has been suggested that the last two methods directly estimate the effect of treatment, whereas the first two methods only group subjects rather than estimate the effect of treatment. Therefore, the latter two methods may be more sensitive to misspecification of the propensity score model than the first two methods.4 It would better to use the IPTW method to estimate treatment effects of PNB. Moreover, selecting similar propensity scores during matching allows the high and low propensity scores to be discarded. We are concerned that this portion will not represent all patients who have undergone TKA. The IPTW method would solve this problem.
The authors declare no competing interests.