Opioid Utilization Following Lumbar Arthrodesis

Trends and Factors Associated With Long-Term Use

Piyush Kalakoti, MD; Nathan R. Hendrickson, MD; Nicholas A. Bedard, MD; Andrew J. Pugely, MD


Spine. 2018;43(17):1208-1216. 

In This Article

Materials and Methods

Data Source

A national commercial claims database, accessed through the PearlDiver Research Program (PearlDiver Inc, Fort Wayne, IN),[22] was the data source for this investigation. The Humana Inc. (https://www.humana.com/) claims dataset through the PearlDiver Research Program contains over 20 million de-identified, HIPAA compliant medical billing records from across the US. Unlike commonly utilized administrative databases and registries, the Humana data permit longitudinal tracking of patients and events based on unique patient identifiers and comprises of data including commercially insured patients as well as Medicare beneficiaries. Clinical data are sequenced using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and Current Procedural Terminology (CPT) codes.

Study Design, Cohort Definition, and Exposure Groups

In a retrospective, population-based observational cohort study, patients registered in the data source that underwent lumbar arthrodesis (ICD-9 procedure code: 81.06–81.08) between 2007 through 2015 were identified. Cohort definition relied upon grouping patients based upon surgical approach for lumbar fusion. This included those undergoing anterior lumbar interbody fusions (ALIFs, 81.06), posterior/transforaminal lumbar interbody (P/TLIF, 81.08), posterolateral fusion (PLF, 81.07), or a combination of ALIFs and PLF (presence of ICD-9 procedure code 81.06 and 81.08) (Online supplement: Table S1, http://links.lww.com/BRS/B362).

The inclusion cohort was limited to include only those active in the dataset at least 3 months prior and 1 year following their lumbar surgery dates. A postoperative period of up to 1 year was deemed appropriate to assess narcotic utilization rates as opposed to 2-year period. The latter would have considerably reduced cohort size and also could have introduced bias relating to narcotic consumption other than those related to spinal conditions or newer events. Patients were then classified as opioid users (OUs) or opioid naive (ON). Patients with an active opioid prescription within 3 months before lumbar arthrodesis were characterized as OU, while those without a preoperative narcotic prescription comprised the latter cohort. Opioid prescription filling was based upon tracking records with an active prescription filling of common oral and transdermal opioids except for tramadol. (Online supplement: Table S2, http://links.lww.com/BRS/B362). Data on explanatory variables (exposures) were extracted and included age, gender, and other a priori selected comorbidities known to influence outcomes such as diabetes mellitus, morbid obesity, presence of psychiatric comorbidity, fibromyalgia, osteoporosis, low back pain, bowel and bladder dysfunction, motor deficits, drug and alcohol abuse (Online supplement: Table S1, http://links.lww.com/BRS/B362).


The primary outcome of interest was trends in monthly opioid prescription filling rates across the two defined study groups (OU vs. ON) up to 1 year following lumbar arthrodesis. To compare postoperative opioid consumption rates, we averaged preoperative monthly proportion of OUs 3 months before surgery as a baseline. Subgroup analysis based upon approach to lumbar arthrodesis was performed to assess pre and postoperative monthly opioid prescription filling rates. Secondary outcome was to assess risk factors associated with opioid use at 1 year (prolonged use) in patients undergoing lumbar arthrodesis, including for common surgical approaches such as ALIF, P/TLIF, and PLF. On the basis of the identified risk factors, we created a clinical calculator (app) predicting 1-year narcotic usage in these patients [http://neuro-risk.com/opiod-use/]. (Online Supplement: Table S3, http://links.lww.com/BRS/B362)

Statistical Analysis

Descriptive and inferential statistical techniques were used. Pearson Chi-square test was utilized to assess differences in proportions across two or more groups for postoperative differences in narcotic prescription filling rates. Unadjusted impact of previously defined exposure groups such as age at surgery, gender, and comorbidities on prolonged postoperative opioid use (at 1-year) following fusion was reported as unadjusted odds ratios (ORs) along with 95% confidence interval (95% CI). Multivariable log-binomial (generalized linear model) models were constructed to assess the risks associated with opioid usage at 1-year postoperative period (Figure S1). Regression diagnostics was performed for all models including assessment of deviance and calibration testing using Hosmer-Lemeshow (HL) test (P > 0.10). All statistical tests were two-sided, and the significance level was derived at an alpha set at 5%. All analysis was performed using the SAS software (version 9.4; SAS Institute, Cary, NC).