Secondary Logo

Journal Logo

Healthcare Economics, Policy, and Organization: Original Clinical Research Report

Heterogeneity Among Hospitals in the Percentages of All Lumbosacral Epidural Steroid Injections Where the Patient Had Received 4 or More in the Previous Year

Pearson, Amy C. S. MD*; Dexter, Franklin MD, PhD, FASA; Epstein, Richard H. MD, FASA

Author Information
doi: 10.1213/ANE.0000000000004253



  • Question: What are the proportions of lumbosacral epidural injections administered to patients who have received 5 or more such injections during the preceding 364 days (ie, followed guidelines), and do these proportions vary among hospitals?
  • Findings: Among the 39 hospitals in Iowa that performed overall at least 1 steroid injection every 4 days, there were 6 hospitals at which the percentages of injections that were the fifth or greater significantly exceeded the overall prevalence of 1.91%; 14 of the 39 hospitals had prevalences significantly less.
  • Meaning: This heterogeneity warrants study of whether annual steroid injections per patient should be a clinical quality measure for the care received by patients with lower back pain or whether payment should be greater when injections are in accordance with guidelines.

Merit-based incentive payment system measures for pain medicine are for diagnostic evaluations (such as screening for depression),1,2 not for interventional treatments. We assess compliance with guidelines for the most common pain procedures, lumbosacral epidural steroid injections (“steroid injections”).3–5

The American Society of Interventional Pain Physicians’ Interventional Pain Management guidelines include that “in the treatment or therapeutic phase” (ie, when steroids are used), injections “should be repeated only as necessary according to medical necessity criteria, and it is suggested that these be limited to a maximum of 4 times per year.”6 This guideline matches evidence that first and second injections including steroids in a year have significant efficacy versus the injection of local anesthetic only.7 Among patients receiving repeat single-level transforaminal steroid injection, there was comparable benefit in pain reduction (based on numeric rating scale) from second injection compared with first, but too few data for confidence intervals sufficiently narrow to evaluate efficacy for a third injection within a year.8 Among a cohort with persistent pain after steroid injection for radicular pain due to lumbar disk herniation, a second injection reduced pain and functional impairment; no patient received a third injection.9 When caudal epidural steroid injections have been demonstrated to provide benefit relative to placebo for brief or long-term periods, the benefit was for the first and second injections.7,10,11 Yet, there are risks from the local administration of progressive, cumulative doses of steroid.12,13

The objective of our study was to quantify how many lumbosacral steroid injections patients received within 1-year periods. Multiple injections have economic and clinical consequences. We used the Iowa Hospital Association database of procedures performed at hospital-owned facilities statewide (ie, not a study of 1 network), regardless of payer (ie, not restricted to Medicare).14 The data used Healthcare Common Procedure Coding System (HCPCS) codes. This allowed us to analyze epidural steroid injections (ie, to exclude nonsteroid epidural injections); analyses of Medicare Part B data are without specification as to diagnostic or therapeutic.4,15 Using the dates of injections, the number of injections that the patient previously and subsequently received within 364 days (ie, 1 year) could be counted and compared among hospitals. We are unaware of another study reporting such managerial epidemiology or the existence of another publicly available database for such research.


The University of Iowa Institutional Review Board determined that this project (No. 201810757) does not meet the Federal regulatory definition of human subjects research. This retrospective, observational, cohort study was performed using deidentified patient data. The earliest date studied matched the date of the Iowa Hospital Association current method of patient deidentification.

The lumbosacral epidural steroid injections studied were those performed at all 121 nonfederal, hospital-owned surgical facilities in Iowa from July 1, 2012 through September 30, 2017.14 The 2 federal hospitals in Iowa, not contributing data to the Iowa Hospital Association, were the Veterans Affairs hospitals in Des Moines and in Iowa City. In the Iowa Hospital Association dataset, each procedure is attributed to the hospital owning the facility where the procedure was performed16; most facilities are in close geographic proximity to the hospital.17 The hospital data included not only the outpatient therapeutic procedures performed at hospitals, but also at facilities owned by those hospitals (eg, an ambulatory surgery center and medical office building of a pain medicine program).18,19 We henceforth refer to all the studied facilities as “hospitals.”

The injections were identified using the 5 HCPCS codes for lumbosacral epidural injection, interlaminar, or transforaminal (Supplemental Digital Content, Table A, The injections were limited to procedures performed during the same ambulatory visit as injection of methylprednisolone, triamcinolone, dexamethasone, or betamethasone (Supplemental Digital Content, Table A,

Analyses by patient were based on the date of each steroid injection. For calculations by patient, the first date that each patient underwent a steroid injection was determined. The number of steroid injections during the subsequent 364 days was calculated. These results are presented as raw counts to motivate future investigations; they are unrelated to our primary objective. Figure 1 legend includes counts excluding the final 1 year.

Figure 1.
Figure 1.:
Patients who underwent different numbers of lumbosacral epidural steroid injections during the 1 year following the first such injection. There were 48,270 such patients (ie, the sums of the numbers of patients shown in the figure sum to 48,270). From among the 121 nonfederal hospitals in the State of Iowa, there were 112 hospitals with at least 1 such injection. This highlights that most patients received 1 injection (54.1%). However, there were 4 or more additional injections during subsequent 364 days for 1.27% of patients (ie, 5 or more as listed in the Table). This is shown by using a logarithmic vertical axis. Because the date of each steroid injection was used, the number of steroid injections during the 1 year means the total count within 365 days, not during the calendar year. When calculations were repeated excluding patients having the first injection during the final 1 year, the corresponding percentages were 54.3% instead of 54.1% (a difference of prevalences of 0.26%) and 1.28% instead of 1.27% (a difference of 0.01%), respectively.

Analyses by injection also were based on the date of each steroid injection. For calculations by steroid injection from the date that each injection was performed, the number of steroid injections during the preceding 364 days was calculated. Our primary objective was to quantify injections because payments are made per administration. Comparisons among hospitals were limited to hospitals (1) with the same years of data such that influences of starting and ending dates were the same, and (2) with substantial sample sizes (see Results). Comparisons were made among percentages, principally the percentage of steroid injections in which the patient had received at least 4 steroid injections during the preceding 364 days thus making the current injection the fifth or greater. We expected to have sufficient sample sizes to compare hospitals based on this end point because steroid injections were the most common interventional pain procedures among the studied hospitals during the studied dates (see Table 1 and Figure 1 of Pearson and Dexter5). We used Pearson χ2 test with the exact P value calculated using Monte-Carlo simulation to 8-digit accuracy (StatXact 11; Cytel, Inc, Cambridge, MA).a Confidence intervals for percentages were calculated using the Clopper–Pearson method.21,22 We deliberately selected this method because it is conservative (ie, confidence intervals wider than needed to assure that reported significant differences among hospitals were reliably so). Bonferroni adjustments of the 95% confidence intervals were used for the same reasons (eg, as compared with the Holm–Šidák method).23 The Strengthening The Reporting of OBservational Studies in Epidemiology (STROBE) checklist is in Supplemental Digital Content, Table B,

Patient demographics also were studied. For each demographic group (eg, male), we calculated the count of steroid injections for the denominator and the count of those steroid injections that were the patient’s fifth or greater within 364 days as the numerator. Patient ages were binned in 10-year categories, except among patients ≥85 years of age, in which all such patients were in the ≥75-year upper age category for privacy protection.


There were 48,270 unique patients who each underwent at least 1 lumbosacral epidural steroid injection (“steroid injection”). The patients received care at 112 hospitals’ facilities. Most patients received no additional steroid injections within 364 subsequent calendar days after the first steroid injection (54.1%). There were 4 or more additional steroid injections within 364 days of the first steroid injection for 1.27% of patients (ie, the injection was the fifth or greater). Raw counts are presented in Figure 1.

There were 103,750 steroid injections performed over the studied 1916 days. The Table shows the demographics of the patients who underwent these steroid injections. Among the steroid injections, 1.50% was the fifth steroid injection for the patient within the preceding 364 days, and 1.91% was the fifth or greater steroid injection (Table). The percentages differed among hospitals (P < .00001). The percentages did not differ significantly between transforaminal and interlaminar routes (P = .76; Supplemental Digital Content, Table A, Only 0.09% of the steroid injections were among patients receiving steroid injections at >1 hospital’s facilities during the year (Table).

Demographics of the Lumbosacral Epidural Steroid Injections Studied

Figure 2 shows that among the 39 hospitals in Iowa that performed overall at least 1 steroid injection every 4 days, there were 6 hospitals at which the percentages of injections that were the fifth or greater significantly exceeded the overall prevalence of 1.91% (range: 3.0%–6.4%).8,24,b There were 14 of the 39 hospitals with prevalences significantly less than the overall prevalence (Figure 2). Hospital was a significantly stronger predictor of these percentages than patient characteristics (eg, primary payer) (Table; Figure 2).25,26

Figure 2.
Figure 2.:
Prevalences of lumbosacral epidural steroid injections that were the fifth or greater such injections for the patient within 365 days. For example, suppose that at a hospital a patient received 1 injection today and 5 injections during the preceding 364 calendar days. Then, that patient contributed 6 to the hospital’s denominator and 2 to the hospital’s numerator. The figure includes the 39 hospitals in Iowa performing overall at least 1 such injection every 4 days. The overall prevalence was 1.91%, shown by the horizontal line. At the hospital with the 1 accredited pain medicine fellowship in the state, the prevalence was 0.4% (N = 10/2589), not significantly different from the 0.1% (N = 3/2087, Fisher exact test P = .16) reported by Mayo Clinic8 and 0.4% (N = 2/540, P = .99) reported by Wooridul Spine Hospital, Seoul, Korea.24 The confidence intervals are 2-sided, but only the limit relevant to the comparison is displayed. There were 14 hospitals (in blue) with the prevalences of fifth or greater injections significantly smaller than the overall prevalence, while Bonferroni adjusting for the 39 comparisons. There were 6 (in red) with significantly greater than the overall prevalences. The Cramér V (0.123, SE = 0.004) comparing the 39 hospitals was larger than for each of the 5 patient variables in the Table; all P < .00001 after Bonferroni adjustment for the 5 comparisons. SE indicates standard error.

We performed 3 additional analyses to evaluate the validity of our evaluation of the percentages of steroid injections that were the fifth steroid injection or greater during the year for the patient. First, we tested convergent validity. There was significant positive correlation among the 39 hospitals between the percentages of steroid injections that were (1) the fifth or greater during the year and (2) the sixth or greater during the year. The Spearman rank correlation was 0.915 (standard error [SE] = 0.022), P < .00001. Second, we performed another test of convergent validity. There was significant negative correlation among the 39 hospitals between the percentages of steroid injections that were (1) the fifth or greater during the year and (2) the first within the year (Supplemental Digital Content, Figure A, The Spearman rank correlation was −0.792 (SE = 0.067), P < .00001. Third, we evaluated discriminant validity. There was lack of significant correlation among the 39 hospitals between the (1) percentages of steroid injections that were the fifth or greater during the year and (2) count of steroid injections performed. The Spearman correlation was 0.267 (SE = 0.159), P = .10, and Pearson correlation was 0.155 (SE = 0.134), P = .35.


In the current paper, we showed reliably (P < .00001) heterogeneity among hospitals in the prevalences of repeated lumbosacral epidural steroid injections that were the patient’s fifth or greater during the year. The reliability of the results of multiple hospitals differing from the overall prevalence was assured further by our use of conservative statistical methods.21–23 The magnitude of effect of hospital was substantively larger than the effects of patient demographics (Table; Figure 2). The usefulness of the results was based on steroid injections being the most common of the interventional pain procedures,3–5 especially at the many small hospitals that we studied (Figures 1–3 of Pearson and Dexter5). The validity of the end point was supported by positive correlation with the counts of sixth or greater injections (convergent validity), negative correlation with the counts of first within the year (convergent validity), and by lack of correlation with total counts of steroid injections (discriminative validity).

As summarized above, the American Society of Interventional Pain Physicians’ guidelines suggest that steroid injections “be limited to a maximum of 4 times per year.”6 Clinicians have incremental knowledge about individual patients’ response to guide the decision of another steroid injection. Specifically, patients responding to steroid injection when assessed 2 weeks later were more likely to be responders at 2 months than those who were nonresponders at 2 weeks (odds ratio, 6.49; 95% confidence interval, 5.38–7.84).27 However, although cohort and placebo-controlled trials have shown significant efficacy from first and second steroid injections, efficacy data are lacking about additional steroid injections.7–11 Furthermore, there is overall brief duration of significantly improved functional status following interlaminar injection versus placebo.28,29 Lumbar epidural steroid injections (inclusive of interlaminar and transforaminal) provide small gains in quality-adjusted life-years.30 Long-term efficacy is limited.31

We were careful not to risk adjust our comparisons among hospitals for percentages of steroid injections that were the fifth or greater within the year, because of the results of the Table (see footnotes a and c). The strongest and the second strongest predictors among patient characteristics were payer and age, respectively. US Medicare patients had the greatest observed prevalences of steroid injections being the fifth or greater versus patients of other payers (Table); prior authorization (approval) was not needed for payment. This effect of payer matches that of Virk et al32 who compared the percentage of patients from multiple states who were ≥65 years of age with lumbar degenerative condition and underwent at least 1 injection, not necessarily steroid. Patients with Medicare fee-for-service insurance had greater prevalence of injection than those with commercial insurance (eg, Medicaid Part C).32 Creating Figure 2 while adjusting for heterogeneity in patient characteristics would be permissive of a hospital that preferentially performed more steroid injections in excess of 4 among patients receiving Medicare (ie, ≥65 years of age). National guidelines do not differentiate Medicare patients from other patients for the recommendation of a maximum of 4 or greater injections.6 Neither clinical trials nor observational studies suggest greater efficacy among patients ≥65 years of age.7,8,10,31 Rather, they suggest the opposite. The complications of epidural steroid injections are greater with older age (ie, among patients receiving Medicare).12,13 Regardless of our rationale, 1 limitation of our findings is that they may apply only to the United States, because of the substantial apparent influence of Medicare payment policies on the observed clinicians’ behavior.

Our article has the additional limitation that, although the results can be applied promptly by hospitals in the State of Iowa, by clinicians revising their practices, or by payers, further consideration of the value and practicality of using the frequency of steroid injections as a clinical quality measure would require prospective observational data collected from interventional pain medicine practices including their patients’ outcomes.33 We expect that data would need to be collected prospectively because the methodology that we used would likely not be available from other secondary data sources. Specifically, the Iowa Hospital Association data included recording of HCPCS codes including medications (unlike Medicare Part B4), tracking of all patients among hospitals (unlike Premier and other commercial databases), and dates of procedures (unlike state ambulatory procedure data available through the Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project). We are unaware of the existence of a publicly available database that includes the necessary data not only from hospital-owned facilities, but also independently owned offices.34,c Future studies also should evaluate the extent to which the best clinical quality measure is the percentage of injections that were the fifth or greater within the year versus the efficacy of the first or second injection received by the patient, as these end points are correlated (Supplemental Digital Content, Figure A, Most patients (54.1%) received 1 steroid injection, and likewise most injections (52.9%) were first injections. Yet, for the treatment of acute radiculopathy, a second steroid injection frequently reduces pain.8 Having had just 1 injection could reflect near-complete resolution of pain or could represent barriers to receiving appropriate care such as insurance prior authorization.32 Finally, if procedures performed in non–hospital-owned offices were added, whether that would change the distribution of procedures per clinician sufficiently to warrant analysis by clinician rather than by group is unknown.

In conclusion, although most patients (≥53.6%) received only 1 lumbosacral steroid injection within 365 days, 1.27% of patients received 5 or more. While several hospitals in Iowa had significantly less than 1.91% of steroid injections being the fifth or greater, several hospitals had significantly greater than 1.91% of injections being the fifth or greater. The heterogeneity warrants systematic study of whether a threshold of the number of steroid injections within 365 days could be a clinical quality measure or whether payment should be incentivized for complying with guidelines.


The data used in this study can be obtained from the Iowa Hospital Association.14 Craig Jarvie, MMR, Senior Planning Analyst at University of Iowa Health Care, managed the Iowa Hospital Association data at the University of Iowa and prepared the data fields for us.


Name: Amy C. S. Pearson, MD.

Contribution: This author helped design the study and write the manuscript.

Name: Franklin Dexter, MD, PhD, FASA.


This author helped design the study, analyze the data, and write the manuscript.

Name: Richard H. Epstein, MD, FASA.


This author helped write the manuscript.

This manuscript was handled by: Nancy Borkowski, DBA, CPA, FACHE, FHFMA.


aAnalyses were not performed with clustering by patient, because there were 934 unique patients among the 1979 injections that were the fifth or greater within the preceding 364 days. These were 53% of patients contributing 1, 76% contributing 1 or 2, 84% contributing 1–3, and 91% contributing 1–4. Corresponding percentages for the 945 unique combinations of hospital and patient were 54%, 77%, 84%, and 91%, respectively.

bAmong the 39 hospitals in Figure 2, the relevant upper or lower (2-sided) Bonferroni-adjusted confidence limits were <2.5%, the maximum being 2.38%. Instead using the criterion of hospitals with overall at least 1 steroid injection every 3 days, there were 8 fewer hospitals, 4 fewer outliers (all negative), and no change in the maximum confidence limit. In contrast, using the criterion of hospitals with overall at least 1 steroid injection every 5 days, there were 9 additional hospitals, but just 1 additional outlier (a negative one), and 3 hospitals with confidence limit >2.5%, the maximum being 3.14%.

cIn Iowa, procedure data are not obtained from the independent ambulatory surgery centers.34


1. Savoy ML, O’Gurek DT. Screening Your Adult Patients for Depression. Fam Pract Manag. 2016;23:16–20.
2. CMS2v8, Preventive care and screening: screening for depression and follow-up plan. Available at: Accessed October 27, 2018.
3. Manchikanti L, Pampati V, Hirsch JA. Retrospective cohort study of usage patterns of epidural injections for spinal pain in the US fee-for-service Medicare population from 2000 to 2014. BMJ Open. 2016;6:e013042.
4. Manchikanti L, Soin A, Mann DP, Bakshi S, Pampati V, Hirsch JA. Comparative analysis of utilization of epidural procedures in managing chronic pain in the Medicare population: pre and post affordable care act. Spine (Phila Pa 1976). 2019;44:220–232.
5. Pearson ACS, Dexter F. Observational study of the distribution and diversity of interventional pain procedures among hospitals in the State of Iowa. Pain Physician. 2019;22:e157–e170.
6. Manchikanti L, Abdi S, Atluri S. An update of comprehensive evidence-based guidelines for interventional techniques in chronic spinal pain. Part II: guidance and recommendations. Pain Physician. 2013;16:S49–S283.
7. Manchikanti L, Singh V, Cash KA, Pampati V, Damron KS, Boswell MV. A randomized, controlled, double-blind trial of fluoroscopic caudal epidural injections in the treatment of lumbar disc herniation and radiculitis. Spine (Phila Pa 1976). 2011;36:1897–1905.
8. Murthy NS, Geske JR, Shelerud RA. The effectiveness of repeat lumbar transforaminal epidural steroid injections. Pain Med. 2014;15:1686–1694.
9. Joswig H, Neff A, Ruppert C, Hildebrandt G, Stienen MN. Repeat epidural steroid injections for radicular pain due to lumbar or cervical disc herniation: what happens after ‘salvage treatment’? Bone Joint J. 2018;100-B:1364–1371.
10. Manchikanti L, Cash KA, McManus CD, Pampati V, Smith HS. One-year results of a randomized, double-blind, active controlled trial of fluoroscopic caudal epidural injections with or without steroids in managing chronic discogenic low back pain without disc herniation or radiculitis. Pain Physician. 2011;14:25–36.
11. Iversen T, Solberg TK, Romner B. Effect of caudal epidural steroid or saline injection in chronic lumbar radiculopathy: multicentre, blinded, randomised controlled trial. BMJ. 2011;343:d5278.
12. Bicket MC, Chakravarthy K, Chang D, Cohen SP. Epidural steroid injections: an updated review on recent trends in safety and complications. Pain Manag. 2015;5:129–146.
13. Kerezoudis P, Rinaldo L, Alvi MA. The effect of epidural steroid injections on bone mineral density and vertebral fracture risk: a systematic review and critical appraisal of current literature. Pain Med. 2018;19:569–579.
14. Iowa Hospital Association Outpatient Data Request. Available at: Accessed August 8, 2018.
15. Data Dictionary for Use with the IHA Outpatient Database, Revenue Records. Available at: Accessed October 27, 2018.
16. Data Dictionary for Use with the IHA Outpatient Database, Appendix I, Hospitals. Available at: Accessed August 9, 2018.
17. Elhag D, Dexter F, Elhakim M, Epstein RH. Many US hospital-affiliated freestanding ambulatory surgery centers are located on hospital campuses, relevant to interpretation of studies involving ambulatory surgery. J Clin Anesth. 2018;49:88–91.
18. Carver College of Medicine. Department of Anesthesia, Chronic Pain Medicine. Available at: Accessed August 9, 2018.
19. Accreditation Council for Graduate Medical Education (ACGME) – Public Advanced Program Search. Available at: Accessed August 8, 2018.
20. HCPCS – General Information. Available at: Accessed August 9, 2018.
21. Clopper CJ, Pearson ES. The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika. 1934; 26:404–413.
22. Hahn GJ, Meeker WQ. Statistical Intervals. A Guide for Practitioners. 1991:New York, NY: Wiley; 82100–84,105.
23. Holm–Šidák Method. Available at: Accessed November 2, 2018.
24. Lee JH, Lee SH. Can repeat injection provide clinical benefit in patients with lumbosacral diseases when first epidural injection results only in partial response? Pain Physician. 2016;19:E283–E290.
25. Cramér’s V. Available at: Accessed November 3, 2018.
26. US Geological Survey statistical interpretation. Available at: Accessed May 15, 2019.
27. El-Yahchouchi C, Wald J, Brault J. Lumbar transforaminal epidural steroid injections: does immediate post-procedure pain response predict longer term effectiveness? Pain Med. 2014;15:921–928.
28. Arden NK, Price C, Reading I, et al.; WEST Study Group. A multicentre randomized controlled trial of epidural corticosteroid injections for sciatica: the WEST study. Rheumatology (Oxford). 2005;44:1399–1406.
29. Wilson-MacDonald J, Burt G, Griffin D, Glynn C. Epidural steroid injection for nerve root compression. A randomised, controlled trial. J Bone Joint Surg Br. 2005;87:352–355.
30. Carreon LY, Bratcher KR, Ammous F, Glassman SD. Cost-effectiveness of lumbar epidural steroid injections. Spine (Phila Pa 1976). 2018;43:35–40.
31. Zhai J, Zhang L, Li M. Epidural injection with or without steroid in managing chronic low back and lower extremity pain: a meta-analysis of ten randomized controlled trials. Int J Clin Exp Med. 2015;8:8304–8316.
32. Virk SS, Phillips FM, Khan SN. Factors affecting utilization of steroid injections in the treatment of lumbosacral degenerative conditions in the United States. Int J Spine Surg. 2018;12:139–148.
33. Doan L, Patel H, Aronova Y, Gharibo C. Variations in interlaminar epidural steroid injection practice patterns by interventional pain management physicians in the United States. Pain Physician. 2018;21:E493–E499.
34. Sioux City Journal, Iowa Not Regulating All Outpatient Surgery Centers. Available at: Accessed February 11, 2019.

Supplemental Digital Content

Copyright © 2019 International Anesthesia Research Society