Chronic opioid use poses an issue of pressing concern in the United States given the large burden it imposes in terms of morbidity and economic costs.1 Moreover, the past decade has seen sharp increases in opioid use2–5 with an accompanying increase in opioid-related adverse events and deaths.2,5 Of particular concern is opioid utilization in the perioperative period, in which developing literature suggests that patients undergoing surgery are at increased risk for prolonged opioid use.6–9 For example, 1 study estimated a 3% incidence of prolonged opioid use after major elective surgery,7 and the incidence of persistent pain or chronic opioid use after knee surgery has been estimated to range from 10% to 34%.10,11
In this light, a question with important clinical implications is whether the use of peripheral (eg, femoral nerve block) or neuraxial blockade (eg, epidural) can reduce the risk of chronic opioid use after surgery. Nerve blockade has been shown to reduce opioid requirements in the immediate postoperative period12–14 and has the potential to decrease the transition from acute to chronic pain by directly blocking transmission of pain impulses during the perioperative period, a theory known as preventive analgesia.15,16 Moreover, nerve blocks are a well-established modality for treating acute postoperative pain, which when severe is predictive of the development of chronic pain.17 However, there is limited evidence to date to suggest that nerve blockade is effective in reducing the risk of chronic opioid use after surgery with some studies demonstrating potential benefit in the cases of thoracotomy and abdominal surgery18 and other studies finding no effect.19
In this study, we tested the hypothesis that the use of nerve blockade was associated with a lower incidence of chronic opioid use among patients undergoing total knee arthroplasty (TKA). This can be considered an important public health concern given the frequency and volume of this procedure20 and rates of persistent postoperative pain and opioid use.21,22 In addition, we also examined whether nerve blockade was associated with other measures of long-term postoperative opioid use such as the number of prescriptions filled in the first postoperative year. Overall, a large base of literature suggests that the use of regional anesthesia for TKA can have moderate but beneficial effects on short-term outcomes related to the initial inpatient stay such as mortality, complication rates, length of stay, opioid use, and opioid-related adverse events.12,13,23 However, whether regional anesthesia can reduce the risk of chronic opioid use after surgery remains unknown. Using administrative claims data, we identified privately insured patients undergoing TKA between 2002 and 2012. We then examined whether the use of regional anesthesia was associated with a reduced risk of prolonged opioid use during the first postsurgical year.
We obtained a sample of administrative health claims provided by Marketscan (Truven Health Analytics, Ann Arbor, MI). Marketscan provides person-level data on utilization and spending for the care of people enrolled in private insurance plans through a participating employer, health plan, or government organization. In more recent years, the database contains information on the care of over 35 million beneficiaries and has contained fewer in earlier years. The data are frequently used in analyses of health care utilization and spending.24–27 Our data include all claims between 2001 and 2013.
The inpatient and outpatient claims data provide detailed information from specific encounters including diagnosis codes (International Classification of Diseases, 9th Revision), procedure codes (Current Procedural Terminology [CPT]), and date the service was provided. For pharmacy claims, the information provided includes fill date, quantity supplied, and number of days supplied. The data also provide the National Drug Code, which can be linked to Red Book data (Truven Health Analytics, Ann Arbor, MI) to obtain the generic name and dose of the prescribed drug. Because this study used deidentified data, approval from an institutional review board was not required.
Using the CPT code 27447, we constructed an initial sample consisting of 124,129 patients undergoing TKA between January 1, 2002, and December 31, 2012, who were continuously enrolled for at least 1 year before and 1 year after the date of their procedure. We then excluded patients who were younger than age 18 years or older than age 64 years at the time of their procedure (n = 79). In addition, because patients with cancer are at increased risk for opioid use,28 we used the methods described subsequently to exclude patients with a history of cancer (n = 3970), resulting in a final sample of 120,080 patients. From this sample, 17,184 patients (14.3%) received more than 1 TKA during the study period; for these patients, our analyses were restricted to their first TKA.
Using pharmacy claims data, we isolated prescriptions for fentanyl (patch or oral form), hydrocodone, hydromorphone (oral form), methadone, morphine, oxymorphone, and oxycodone and excluded prescriptions containing hydrocodone in cough/cold formulations. Our primary outcome of interest was chronic opioid use within the first postsurgical year. Previous studies using administrative claims data have defined chronic opioid use as having filled 10 or more prescriptions or >120 days’ supply within a 1-year period.8 Because some opioid use is likely expected in the immediate postoperative period, we modified this definition to having filled 10 or more prescriptions or >120 days’ supply within the first year of surgery, excluding the first 90 postoperative days (ie, postoperative days 91–365).
Our independent variable of interest was whether the patient received a peripheral nerve block or a neuraxial block (eg, epidural or spinal) for their procedure. We identified the use of a nerve block by examining whether a claim with the following CPT codes was submitted along with the claim for the initial surgery: 64447/64448 (femoral nerve block or catheter), 64449 (lumbar plexus catheter), 64450 (other peripheral nerve block), 62318 (spinal block), and 62319 (epidural block).
In addition, we included a robust set of variables to adjust for potential confounders. These variables were chosen based on literature and/or biologic plausibility identifying them as risk factors for prolonged opioid use after surgery.7,29 First, we included age and sex, which were directly obtained from the claims data. Second, we included an extensive list of comorbidities including among others diabetes mellitus, congestive heart failure, peripheral vascular disease, and hypertension (see Table 1 for a full list of comorbidities). Patients were considered to have a history of the given comorbidity if they accumulated at least 2 claims with relevant International Classification of Diseases, 9thRevision codes in the 365 days before their surgery.30 Finally, we included whether the patient had any claims for the following prescription drugs or classes of prescription drugs in the year before surgery: benzodiazepines, antidepressants, gabapentin, pregabalin, antipsychotics, antihypertensives, antidiabetes medications, and antiasthma medications. A list of the specific medications included in each class can be found in Supplemental Digital Content, Appendix, http://links.lww.com/AA/B685.
We used a multivariable logistic regression to estimate the association between receiving a block and the risk for chronic opioid use after surgery. In this regression, the dependent variable is whether the patient met the criteria for postoperative chronic opioid use, and the independent variable of interest is whether the patient received a block. The coefficients from the regression can be used to obtain an adjusted relative risk (ARR), which is the relative reduction in the incidence of chronic opioid use for patients who received a block, after adjusting for age, gender, and year of surgery as well as medical comorbidities and preoperative prescription drug use as previously described. For example, if the ARR is 0.75, this would imply that the incidence of postoperative chronic opioid use was 25% less among patients who received a block.
We performed our analyses separately for 3 subgroups based on the intensity of opioid use in the year before surgery: opioid-naïve, intermittent opioid use, and chronic opioid use. Opioid-naïve patients were defined as patients who did not fill a prescription in the year before their surgery, whereas chronic use was defined as having filled more than 10 prescriptions or 120 days’ supply during this timeframe. Persons who filled at least 1 prescription for an opioid but who did not meet the criteria for chronic use were classified as intermittent users. We performed our analyses using Stata 14.0 (College Station, TX). The Hosmer-Lemeshow test (using deciles) was used to assess the goodness of fit for our logistic regression models.
The baseline analyses combined patients who received a peripheral neuraxial block into 1 group. To examine whether our results might differ by the type of block, we divided block patients into 2 groups: those who received a peripheral nerve block (eg, femoral nerve block, lumbar plexus block, other peripheral nerve block) and those who received a neuraxial block (eg, spinal or epidural) and performed a separate analysis in which we separately estimated the ARR for patients receiving a peripheral nerve block and the ARR for patients receiving a neuraxial block to patients who did not receive a block at all.
We explored the robustness of our results via several sensitivity analyses. First, we considered alternative specifications of our model in which the dependent variable was the number of opioid prescriptions, the number of days supplied, and the total oral morphine equivalents utilized in the postoperative period (postoperative days 91–365). To obtain the total oral morphine equivalents utilized during postoperative days 91 to 365, we converted each opioid prescription to milligrams of oral morphine by multiplying the quantity supplied by the dosage and a conversion factor.31
A unique issue arose for each of these outcomes because the majority of patients undergoing TKA (53.6%) did not fill any prescriptions for an opioid during postoperative days 91 to 365. When there are many observations for which the dependent variable is zero, a simple regression analysis will tend to be downward-biased; in other words, the estimated association will be lower in magnitude than the true association.32 To address this issue, we performed a 2-step analysis. In the first step, we estimated a multivariable logistic regression in which the outcome was whether the patient filled any opioid prescription at all, and the independent variables were the same variables used in our baseline analysis.
In the second step, we estimated a multivariable linear regression in which the dependent variable was our outcome of interest (number of prescriptions, number of days supplied, or number of oral morphine equivalents supplied during postoperative days 91–365), whereas the independent variables remained the same as in our baseline analysis. These analyses were restricted to patients who filled at least 1 prescription for an opioid during postoperative days 91 to 365 (ie, patients with nonzero opioid use). This 2-step approach has been used in other studies to obtain nonbiased estimates when a large proportion of observations assume a value of zero.33,34 It obtains unbiased estimates (ie, estimates that are likely to be close to the true value) because the first step estimates the effect of the given treatment on the probability that the outcome is nonzero, whereas the second step estimates the effect of a given treatment of the level of the outcome itself, conditional on the outcome being greater than zero. For this second set of analyses, we converted our independent variable (eg, number of prescriptions) to natural logs both because the distribution of the underlying data was fairly skewed and also to easily report the estimated association in percentage terms.
Finally, 1 potential source of bias arises because of our use of billing codes to identify the receipt of a block. Because these codes dictate payment for a block, their presence is likely to be specific—in other words, it is unlikely that a patient did not receive the block if the code is present. However, the presence of these codes may not be fully sensitive, particularly because anesthesia providers cannot bill for a block if the block was only intended to provide anesthesia for the procedure itself and was not intended to provide postoperative analgesia. In addition, it is possible that, rather than submitting a separate claim for the block, the anesthesia provider simply billed for the additional time used to place the block under the claim submitted for providing anesthesia for the knee arthroplasty itself.
Thus, it is possible that some patients who received blocks would not have an associated billing code and would be counted as part of the “no block” group, which could potentially bias our results toward finding no association.
We addressed this issue in 2 ways. First, although TKA could be performed under neuraxial blockade alone, it is unlikely that it was performed solely under peripheral nerve blockade; whereas the literature has described cases of TKA being performed under lumbar plexus and sciatic blocks,35,36 this approach is fairly uncommon. Therefore, 1 approach, which we performed as described previously, is to examine the robustness of our results to specifications in which we separately consider neuraxial and peripheral nerve blockade, because the latter should be unaffected by this bias.
As a second approach, we conducted a falsification analysis in which we first identified patients who did not receive a block but who were highly likely to have received a block based on demographics, medical comorbidities, and preoperative prescription drug use. If misclassification were a serious issue, it would most likely occur among this set of patients—patients who did not receive a block (as measured by the use of billing codes), but who, based on observable characteristics, were likely to have received a block. We then examined the robustness of our results to a falsification analysis where this set of patients (who in reality did not receive a block) was counted as having received a block. If the results from this analysis were similar to the results of our baseline analysis, this would provide reassurance that the misclassification of blocks was not a serious issue.
To implement this analysis, we first estimated a multivariable logistic regression in which the outcome was whether the patient received a block. The independent variables were the demographic factors, medical comorbidities, and preoperative prescription drug use described previously. The results from this regression were then used to estimate each patient’s predicted probability of receiving a block; in essence, this predicted probability is the percentage of patients with similar characteristics who ultimately received a block. We then repeated our baseline analyses, except that patients who did not receive a block but whose predicted probability of receiving a block was 75% or higher were also counted as block patients (in addition to patients who did receive a block).
Study Design and Sample Power
The construction of our study sample (such as the inclusion and exclusion criteria) as well as the elements of our baseline analysis (such as the choice of statistical model, choice of outcome, and choice of additional covariates) was determined before initiating our analysis. Our sensitivity analyses were conducted post hoc based on the finding of our baseline analysis.
We used a Bonferroni approach to correct for testing for our primary outcome across 3 subgroups, which would imply that a P value of .0167 (= .05/3) would be the cutoff for statistical significance. We did not adjust our P value cutoff for additional testing of our secondary outcomes. In presenting our results, we report unadjusted P values (ie, P values that are unadjusted for multiple hypothesis testing) but emphasize that the cutoff for statistical significance is the threshold outlined previously. In line with this approach, we report 98.3% confidence intervals (CIs) for all our analyses.
In the “Results” section, we detail the sample sizes for our 3 subgroups (opioid-naïve patients, intermittent opioid users, and chronic users) as well as the rates of block use and the unadjusted incidence of chronic opioid use among patients in the subgroup who did not receive a block. Using these values as well as the adjusted significance threshold of 0.0167, our sample would be adequately powered to detect a decreased incidence of 0.38 percentage points (20% relative decrease) among opioid-naïve patients, 0.74 percentage points (12% relative decrease) among intermittent opioid users, and 2.62 percentage points (4% relative decrease) among chronic users. Thus, our analyses were adequately powered to find differences in incidence of <3 percentage points—and for 2 of our subgroups, <1 percentage point—differences thought to be too small to be of clinical significance. Specifically, after 1 previous study, we considered an ARR of 0.8 (ie, 20% relative reduction) to be clinically significant.19
Our sample consisted of 120,080 patients with 53,177 (44.2%) receiving a nerve block for their procedure. Among patients who received a block, 47,123 (88.6%) received a femoral nerve block, 293 (0.55%) received a lumbar plexus block, and 1918 (3.61%) received another type of peripheral nerve block. One hundred thirty-eight patients (0.26%) received a spinal block, and 5119 patients (9.63%) received an epidural. The percentages sum to more than 100 because some patients received more than 1 type of block. A total of 59,298 patients were classified as opioid-naïve (ie, did not fill any opioids before their surgery), of whom 26,183 (44%) received a block. Similarly, 5582 of 13,254 (42%) chronic opioid users (persons who filled at least 10 prescriptions or 120 days’ supply before their surgery) received a block, whereas 21,142 of 47,528 (44%) intermittent users (persons who filled at least 1 prescription for an opioid before surgery but who did not meet the criteria for chronic use) received a block.
There were no differences in terms of age, degree of opioid use in the year before surgery, or use of benzodiazepines, antidepressants, gabapentin, antipsychotics, antihypertensives, or antidiabetes medications in the year before surgery between patients who did and did not receive a nerve block. Patients receiving a block were more likely to have used pregabalin and antiasthma medications in the year before surgery (Table 1). Among the comorbidities we examined, the prevalence of congestive heart failure, chronic obstructive pulmonary disease, diabetes mellitus, and psychosis was lower among patients who received a block, whereas the prevalence of chronic kidney disease and depression (7.16% vs 6.68%) was higher. Patients who received a block were also more likely to be male. However, although statistically significant, these differences were modest in clinical terms (Hedge’s g <0.05 for all characteristics).
The unadjusted incidence of chronic opioid use in the first postoperative year was 1.78% (98.3% CI, 1.62–1.96) among opioid-naïve patients who received a block (Figure 1) compared to 1.81% (98.3% CI, 1.67–1.96) among opioid-naïve patients who did not (P = .744). For patients who used opioids chronically in the year before surgery, the unadjusted incidence of chronic opioid use in the first postoperative year was 67.6% (98.3% CI, 66.4–68.8) for patients who received a nerve block and 67.8% (98.3% CI, 66.8–68.9) among those who did not (P = .761). Among patients who used opioids intermittently in the year before surgery, the unadjusted incidence of chronic opioid use after surgery was 6.08% (98.3% CI, 5.77–6.41) for patients who received a block and 6.15% (98.3% CI, 5.86–6.44) for those who did not (P = .787).
After adjusting for patient demographics, comorbidities, and preoperative medication use, the ARR for chronic opioid use after surgery was 0.984 (98.3% CI, 0.847–1.14, P = .794) among opioid-naïve patients, 1.02 (98.3% CI, 0.933–1.11, P = .617) for patients who used opioids intermittently before surgery, and 0.986 (98.3% CI, 0.957–1.02, P = 0.257) for patients with chronic opioid use before surgery (Figure 2). Put together, these results suggest no association between block use and the risk of postsurgical chronic opioid use. These findings were robust to alternative measures of postoperative opioid use (Table 2) such as whether a patient filled any opioid prescriptions during postoperative days 91 to 365 or the total number of prescriptions filled.
When broken down by type of block, neither peripheral nor neuraxial blocks were associated with statistically significant decreases in the risk for chronic opioid use (Table 3). For peripheral nerve blocks, the estimated ARR (relative to receiving no block) ranged from 0.969 (98.3% CI, 0.852–1.10, P = .630) for opioid-naïve patients to 1.04 (98.3% CI, 0.963–1.12) for chronic opioid users. In the case of neuraxial blocks, the estimated ARR (relative to receiving no block) ranged from 0.853 (98.3% CI, 0.704–1.03, P = .102) for chronic opioid users to 1.11 (98.3% CI, 0.845–1.46, P = .452) for opioid-naïve patients.
To account for potential misclassification, we performed a falsification analysis in which we defined block patients to be (1) those with appropriate claim for a block; and (1) those without a claim but whom we predicted would have been extremely likely (>75% probability) to have received a block (Table 4). Overall, the point estimates and CI from this analysis largely match the point estimates and CI from our baseline analysis.
Opioid use is increasing in the United States, leading to concerns about the concomitant increase in opioid overutilization and opioid-related adverse events. In addition, a growing body of literature suggests that surgery increases the risk of chronic opioid use. Put together, these 2 facts suggest the importance of examining whether the choice of anesthesia technique can reduce the risk of chronic opioid use. In this article, we used health care utilization data to examine whether the use of peripheral (eg, femoral nerve block) blockade or neuraxial (eg, epidural blockade) was associated with a decrease in postoperative chronic opioid use among opioid-naïve patients undergoing TKA. Overall, our results suggest no significant association between the use of nerve blocks and the risk of chronic opioid use in the first year after surgery. Moreover, in general, both the point estimates and lower 95% CI of our estimates argue against any clinically meaningful decrease. We also found that nerve blockade was not associated with any reductions in alternative measures of long-term opioid use after surgery such as the number of prescriptions filled or total oral morphine equivalents utilized in the first postsurgical year.
Although previous work23 has examined whether nerve blockade has been effective in lowering opioid use in the short term—typically during the initial inpatient stay—for joint arthroplasty, our analysis represents a pioneering exploration into the potential association between nerve blockade and opioid use in the longer term. In theory, the use of nerve blockade may reduce long-term opioid use by preventing the transmission of pain impulses during the perioperative period,15 effectively mitigating the severity of acute pain and thus reducing the risk of transition from acute to chronic pain. In addition, by lowering—or even eliminating—opioid requirements during the immediate postoperative period, the use of nerve blockade may reduce chronic opioid use by reducing opioid exposure for susceptible individuals.37,38 Empirical evidence for this benefit has been lacking, however, with a recent study suggesting that neuraxial blockade has no effect on persistent opioid use for patients undergoing abdominal surgery.19 Consistent with this work, we also find no effect in the case of knee arthroplasty.
Our results should be viewed in light of a number of limitations. Like with any observational study, we cannot exclude the possibility that residual confounding could have biased our results. For example, some potentially relevant covariates (eg, body mass index, race) are not available in the data; however, these factors are not expected to be strong confounders (ie, be a strong determinant of exposure and outcome). It could be that physicians used blocks more often for patients they judged most likely to be at risk for postsurgical chronic opioid use based on information only available to them. However, we do note that using the information available, we did not find significant differences between patients who did and did not receive a block for many of the possible confounders we examined such as age and opioid use before surgery. Moreover, even when statistically significant, the magnitude of the differences was small, and in general, our results suggest that patients receiving a block were at lower risk of postsurgical chronic opioid use given that, where differences existed, this group generally experienced a lower prevalence of most comorbidities.
In addition to residual confounding, our results could have been biased by an inability to fully capture the use of nerve blockade. Our approach relied on billing codes to identify the presence of a nerve block. Because these codes dictate payment for a block, their presence is likely to be specific. However, the presence of these codes may not be fully sensitive, particularly because anesthesia providers cannot bill for a block if the block was only intended to provide anesthesia for the procedure itself and was not intended to provide postoperative analgesia. Although we suspect that the incidence of blocks intended solely for intraoperative analgesia is low, patients who received these blocks would not have an associated code and would therefore be counted as part of the “no block” group, which could bias our results toward finding no association. To address this issue, we performed 2 sensitivity analyses. Our results were robust to these sensitivity analyses, suggesting that that inability to fully capture the use of nerve blockade using billing data is unlikely to be significantly affecting our results. Moreover, we also note that the percentage of patients receiving a block in our study (44.3%) was roughly similar to or higher than the percentages found in other studies using alternative methods of measuring block use (such as clinical and electronic medical record data).39,40 Although not fully definitive, this finding also suggests that billing data were largely able to capture the use of blocks. However, our data are limited in that we are not able to characterize the duration of nerve blockade for patients who received continuous peripheral or neuraxial blockade (ie, the use of nerve catheters).
Finally, we note 2 additional limitations. First, our measure of opioid utilization is based on prescriptions that were filled by the patient. Although this is a common method of measuring utilization in the pharmacoeconomics and health policy literature,41 it is worth noting that we do not know whether the patient actually consumed the drug after filling the prescription. Second, our study was limited to privately insured patients younger than age 65 years who received TKA and may not extend to patients 65 years of age and older, who typically receive health insurance from Medicare, the public health insurance program for elderly persons in the United States. Because elderly patients represent the majority of patients who undergo knee arthroplasty and may be particularly susceptible to opioid-induced adverse events, understanding whether nerve blockade may reduce the risks of chronic opioid use in this population is an important area for further study.
Our findings are interesting because, consistent with previous work,19 they suggest that the use of nerve blockade as specifically described here may not confer any long-term benefits with regard to opioid use. This by no means takes away from a large body of literature finding nerve blockade is associated with reductions in opioid use in the short term.12,13,23 However, the block types studied did not entail a large population of patients having received a prolonged peripheral blocks (eg, >24 hours of analgesic duration) in both the L2–L4 and L4–S3 distributions. Otherwise, this is analogous to functional outcome data, which support the benefits of nerve blockade after joint arthroplasty in the short term but fail to definitely prove benefits in longer term function outcomes.42 Why nerve blockade appears to confer short-term benefits without affecting longer term outcomes remains an area for further research. One possibility is rooted in the fact that postoperative pain after knee arthroplasty is in essence mediated by the femoral and sciatic nerves, and that in practice, most practitioners tend to block the femoral nerve only to allow for postoperative rehabilitation efforts. However, a consequence of this approach is that pain from the sciatic nerve is left unblocked. Although this incomplete blockade may explain why peripheral nerve blockade would not reduce the transition to chronic opioid use after knee arthroplasty, it is important to note that we also found no association between neuraxial blockade and the risk of chronic opioid use after surgery. Another possibility—which would apply to peripheral and neuraxial blockade—is that the duration of nerve blockade offered at present is not of sufficient duration. For example, a recent study suggests that patients with persistent postsurgical neuropathic pain after joint arthroplasty typically present with heightened pain scores on postoperative day 5, long after most catheters used for nerve block are removed or single-injection nerve blocks resolve.43
In sum, our results suggest that nerve blockade as described here was not associated with a reduction in chronic opioid use after TKA. Moreover, we found no statistically or clinically significant association for any of our alternative measures of opioid utilization, suggesting that the use of nerve blockade may not reduce opioid use more generally in the long term. However, these results were based on observational data, were focused on a single surgery, and could have been influenced by other factors. Further research efforts investigating the effects of regional anesthesia techniques on long-term opioid use should include prospective studies, other surgical procedures and patient populations, and nerve block interventions of a specified duration that match the trajectory of postoperative pain expected.
Name: Eric C. Sun, MD, PhD.
Contribution: This author helped to design the study, conduct the study, analyze the data, and write the manuscript.
Name: Brian T. Bateman, MD, MSc.
Contribution: This author helped to design the study, analyze the data, and write the manuscript.
Name: Stavros G. Memtsoudis, MD, PhD.
Contribution: This author helped to design the study, analyze the data, and write the manuscript.
Name: Mark D. Neuman, MD.
Contribution: This author helped to design the study, analyze the data, and write the manuscript.
Name: Edward R. Mariano, MD, MAS.
Contribution: This author helped to design the study, analyze the data, and write the manuscript.
Name: Laurence C. Baker, PhD.
Contribution: This author helped to design the study and write the manuscript.
This manuscript was handled by: Richard Brull, MD, FRCPC.
1. Leider HL, Dhaliwal J, Davis EJ, Kulakodlu M, Buikema AR. Healthcare costs and nonadherence among chronic opioid users. Am J Manag Care. 2011;17:32–40.
2. Vital Signs: Overdoses of Prescription Opioid Pain Relievers—United States, 1999–2008, Morbidity and Mortality Weekly Report. 2011.Atlanta, GACenters for Disease Control and Prevention
3. Daubresse M, Chang HY, Yu Y, et al. Ambulatory diagnosis and treatment of nonmalignant pain in the United States, 2000–2010. Med Care. 2013;51:870–878.
4. Wright EA, Katz JN, Abrams S, Solomon DH, Losina E. Trends in prescription of opioids from 2003–2009 in persons with knee osteoarthritis. Arthritis Care Res (Hoboken). 2014;66:1489–1495.
5. Bohnert AS, Valenstein M, Bair MJ, et al. Association between opioid prescribing patterns and opioid overdose-related deaths. JAMA. 2011;305:1315–1321.
6. Alam A, Gomes T, Zheng H, Mamdani MM, Juurlink DN, Bell CM. Long-term analgesic use after low-risk surgery: a retrospective cohort study. Arch Intern Med. 2012;172:425–430.
7. Clarke H, Soneji N, Ko DT, Yun L, Wijeysundera DN. Rates and risk factors for prolonged opioid use after major surgery: population based cohort study. BMJ. 2014;348:g1251.
8. Raebel MA, Newcomer SR, Reifler LM, et al. Chronic use of opioid medications before and after bariatric surgery. JAMA. 2013;310:1369–1376.
9. Sun EC, Darnall BD, Baker LC, Mackey S. Incidence of and risk factors for chronic opioid use among opioid-naive patients in the postoperative period. JAMA Intern Med. 2016;176:1286–1293.
10. Beswick AD, Wylde V, Gooberman-Hill R, Blom A, Dieppe P. What proportion of patients report long-term pain after total hip or knee replacement for osteoarthritis? A systematic review of prospective studies in unselected patients. BMJ Open. 2012;2:e000435.
11. Rozet I, Nishio I, Robbertze R, Rotter D, Chansky H, Hernandez AV. Prolonged opioid use after knee arthroscopy in military veterans. Anesth Analg. 2014;119:454–459.
12. Chan EY, Fransen M, Parker DA, Assam PN, Chua N. Femoral nerve blocks for acute postoperative pain after knee replacement surgery. Cochrane Database Syst Rev. 2014;5:CD009941.
13. Richman JM, Liu SS, Courpas G, et al. Does continuous peripheral nerve block provide superior pain control to opioids? A meta-analysis. Anesth Analg. 2006;102:248–257.
14. Memtsoudis SG, Poeran J, Cozowicz C, Zubizarreta N, Ozbek U, Mazumdar M. The impact of peripheral nerve blocks on perioperative outcome in hip and knee arthroplasty—a population-based study. Pain. 2016;157:2341–2349.
15. Dahl JB, Moiniche S. Pre-emptive analgesia. Br Med Bull. 2004;71:13–27.
16. Liu SS, Buvanendran A, Rathmell JP, et al. A cross-sectional survey on prevalence and risk factors for persistent postsurgical pain 1 year after total hip and knee replacement. Reg Anesth Pain Med. 2012;37:415–422.
17. Kehlet H, Jensen TS, Woolf CJ. Persistent postsurgical pain: risk factors and prevention. Lancet. 2006;367:1618–1625.
18. Andreae MH, Andreae DA. Regional anaesthesia to prevent chronic pain after surgery: a Cochrane systematic review and meta-analysis. Br J Anaesth. 2013;111:711–720.
19. Ladha KS, Patorno E, Liu J, Bateman BT. Impact of perioperative epidural placement on postdischarge opioid use in patients undergoing abdominal surgery. Anesthesiology. 2016;124:396–403.
20. Cram P, Lu X, Kates SL, Singh JA, Li Y, Wolf BR. Total knee arthroplasty volume, utilization, and outcomes among Medicare beneficiaries, 1991–2010. JAMA. 2012; 308: 1227–1236.
21. Brander VA, Stulberg SD, Adams AD, et al. Predicting total knee replacement pain: a prospective, observational study. Clin Orthop Relat Res. 2003;416:27–36.
22. YaDeau JT, Lin Y, Mayman DJ, et al. Pregabalin and pain after total knee arthroplasty: a double-blind, randomized, placebo-controlled, multidose trial. Br J Anaesth. 2015;115:285–293.
23. Macfarlane AJ, Prasad GA, Chan VW, Brull R. Does regional anesthesia improve outcome after total knee arthroplasty? Clin Orthop Relat Res. 2009;467:2379–2402.
24. Baker L, Bundorf MK, Royalty A. Private insurers’ payments for routine physician office visits vary substantially across the United States. Health Aff (Millwood). 2013;32:1583–1590.
25. Chernew ME, Sabik LM, Chandra A, Gibson TB, Newhouse JP. Geographic correlation between large-firm commercial spending and Medicare spending. Am J Manag Care. 2010;16:131–138.
26. Dor A, Koroukian S, Xu F, Stulberg J, Delaney C, Cooper G. Pricing of surgeries for colon cancer: patient severity and market factors. Cancer. 2012;118:5741–5748.
27. Dunn A, Liebman E, Pack S, Shapiro AH. Medical care price indexes for patients with employer-provided insurance: nationally representative estimates from MarketScan Data. Health Serv Res. 2013;48:1173–1190.
28. Jarlbaek L, Hansen DG, Bruera E, Andersen M. Frequency of opioid use in a population of cancer patients during the trajectory of the disease. Clin Oncol (R Coll Radiol). 2010;22:199–207.
29. Macrae WA. Chronic post-surgical pain: 10 years on. Br J Anaesth. 2008;101:77–86.
30. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–1139.
31. Principles of Analgesic Use in the Treatment of Acute Pain and Cancer Pain. 2008.6th ed. Glenview, ILAmerican Pain Society
32. Amemiya T. Regression analysis when the dependent variable is truncated normal. Econometrica. 1973;41:997–1016.
33. Goldman DP, Jena AB, Lakdawalla DN, Malin JL, Malkin JD, Sun E. The value of specialty oncology drugs. Health Serv Res. 2010;45:115–132.
34. O’Donnell O, Doorslaer EV, Wagstaff A, Lindelow M. Nonlinear Models for Health and Expenditure Data, Analyzing Health Equity Data Using Household Survey Data: A Guide to Techniques and Their Implementation. 2008Washington, DCThe World Bank131–145.
35. Williams BA, Ibinson JW, Gould AJ, Mangione MP. The incidence of peripheral nerve injury after multimodal perineural anesthesia/analgesia does not appear to differ from that following single-drug nerve blocks (2011–2014). Pain Med. 2016 [Epub ahead of print].
36. Williams BA, Ibinson JW, Mangione MP, et al. Research priorities regarding multimodal peripheral nerve blocks for postoperative analgesia and anesthesia based on hospital quality data extracted from over 1,300 cases (2011–2014). Pain Med. 2015;16:7–12.
37. Sinha R. Stress and addiction: a dynamic interplay of genes, environment, and drug intake. Biol Psychiatry. 2009;66:100–101.
38. Sinha R. Modeling stress and drug craving in the laboratory: implications for addiction treatment development. Addict Biol. 2009;14:84–98.
39. Cozowicz C, Poeran J, Zubizarreta N, Mazumdar M, Memtsoudis SG. Trends in the use of regional anesthesia: neuraxial and peripheral nerve blocks. Reg Anesth Pain Med. 2016;41:43–49.
40. Fleischut PM, Eskreis-Winkler JM, Gaber-Baylis LK, et al. Variability in anesthetic care for total knee arthroplasty: an analysis from the anesthesia quality institute. Am J Med Qual. 2015;30:172–179.
41. Goldman DP, Joyce GF, Zheng Y. Prescription drug cost sharing: associations with medication and medical utilization and spending and health. JAMA. 2007;298:61–69.
42. Ilfeld BM, Shuster JJ, Theriaque DW, et al. Long-term pain, stiffness, and functional disability after total knee arthroplasty with and without an extended ambulatory continuous femoral nerve block: a prospective, 1-year follow-up of a multicenter, randomized, triple-masked, placebo-controlled trial. Reg Anesth Pain Med. 2011;36:116–120.
43. Lavand’homme PM, Grosu I, France MN, Thienpont E. Pain trajectories identify patients at risk of persistent pain after knee arthroplasty: an observational study. Clin Orthop Relat Res. 2014;472:1409–1415.
Supplemental Digital Content
Copyright © 2017 International Anesthesia Research Society