Authors: Daniel I. McIsaac, M.D., M.P.H., F.R.C.P.C. et al
Anesthesiology published on December 15, 2016.
Background: Although peripheral nerve blocks decrease pain after total knee arthroplasty, the population-level impact of nerve blocks on arthroplasty resource utilization is unknown.
Methods: We conducted a population-based cohort study using linked administrative data from Ontario, Canada. We identified all adults having their first primary knee arthroplasty between 2002 and 2013. Using propensity scores to adjust for measureable confounders, we matched nerve block patients to a patient who did not receive a block. Within the matched cohort, we estimated the independent association of blocks with outcomes (length of hospital stay [primary]; and readmissions, emergency department visits, and falls [secondary]).
Results: One hundred seventy-eight thousand two hundred fourteen patients were identified; 61,588 (34.6%) had a block. The mean hospital stay was 4.6 days with a block compared to 4.8 without. After matching, there was a statistically significant decrease in the length of stay in the block group (relative risk, 0.98; 95% CI, 0.97 to 0.99; P < 0.001). Blocks were associated with a significant decrease in readmissions (relative risk, 0.87; 95% CI, 0.79 to 0.88; P < 0.001) but not emergency department visits (relative risk, 1.02; 95% CI, 0.98 to 1.05) or falls (relative risk, 1.37; 95% CI, 0.90 to 2.08). The association of blocks with length of stay after 2008 was inconsistent; overall, they were associated with longer stays; however, single-shot blocks were associated with shorter stays, while continuous techniques prolonged the length of stay.
Conclusions: Nerve blocks in total knee arthroplasty patients were associated with statistically significant reductions in length of stay and readmissions, but not emergency department visits or falls. The significance of these findings at the patient level and in contemporary practice requires further exploration in prospective randomized studies at low risk of indication bias.