Orthopedics

Feature Article 

Comparing Reported Complication Rates in Shoulder Arthroplasty Between 2 Large Databases

Matthew J. Salzler, MD; Ian D. Engler, MD; Andrew X. Li, MD; Anna H. Jorgensen, MD; Charles Cassidy, MD; David J. Tybor, PhD, MPH

Abstract

Large databases are commonly used to analyze surgical outcomes. Recent studies have suggested that there are differences in complication rates between databases across certain procedures, but the reasons for these differences are not fully understood. The goal of this study was to compare complications of shoulder arthroplasty across databases as well as to interpret the causes of any differences. The authors compared complication rates for shoulder arthroplasty as reported by the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS) and the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) from 2006 to 2010. The authors then restricted NIS data solely to hospitals that also contributed to NSQIP to provide a more direct comparison of the patient populations. The authors identified 48,287 discharges reported in NIS and 1679 discharges reported in NSQIP for patients who underwent shoulder arthroplasty. The complication rate for shoulder arthroplasty was significantly higher in the NIS population (12.6%; 95% confidence interval, 12.0%–13.2%) than in the NSQIP population (5.60%; 95% confidence interval, 4.59%–6.81%). When NIS data were restricted solely to hospitals that also participated in NSQIP, the rate of complications remained higher, at 13.4% (95% confidence interval, 11.2%–15.8%), and the rate increased relative to the nonrestricted data. The databases compared in this study had statistically significant differences in reported complication rates for shoulder arthroplasty. This difference persisted when NIS data were restricted to hospitals that also participated in NSQIP, suggesting that differences in database design contribute to important differences in data. Orthopedic surgeons and administrators must use caution when using complication rates derived from large database studies. [Orthopedics. 2020;43(2):113–118.]

Abstract

Large databases are commonly used to analyze surgical outcomes. Recent studies have suggested that there are differences in complication rates between databases across certain procedures, but the reasons for these differences are not fully understood. The goal of this study was to compare complications of shoulder arthroplasty across databases as well as to interpret the causes of any differences. The authors compared complication rates for shoulder arthroplasty as reported by the Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS) and the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) from 2006 to 2010. The authors then restricted NIS data solely to hospitals that also contributed to NSQIP to provide a more direct comparison of the patient populations. The authors identified 48,287 discharges reported in NIS and 1679 discharges reported in NSQIP for patients who underwent shoulder arthroplasty. The complication rate for shoulder arthroplasty was significantly higher in the NIS population (12.6%; 95% confidence interval, 12.0%–13.2%) than in the NSQIP population (5.60%; 95% confidence interval, 4.59%–6.81%). When NIS data were restricted solely to hospitals that also participated in NSQIP, the rate of complications remained higher, at 13.4% (95% confidence interval, 11.2%–15.8%), and the rate increased relative to the nonrestricted data. The databases compared in this study had statistically significant differences in reported complication rates for shoulder arthroplasty. This difference persisted when NIS data were restricted to hospitals that also participated in NSQIP, suggesting that differences in database design contribute to important differences in data. Orthopedic surgeons and administrators must use caution when using complication rates derived from large database studies. [Orthopedics. 2020;43(2):113–118.]

Large-scale databases maintained by federal agencies and private organizations are used to track patient demographics and admissions, cost of care, and medical and surgical outcomes and complication rates.1–3 Their size and design make them simple and powerful tools for quickly capturing large amounts of data.4 As medicine increasingly focuses on measuring outcomes, many of these databases can provide population-level data on complication rates for common procedures.5–9 This information is used to determine reimbursement rates in orthopedics, such as pay-for-performance measures and bundled payments.10 Further, in an era of increasing transparency of outcomes, these data are being presented to the public to allow them to compare the performance of physicians and hospitals.11

Large databases vary in size, purpose, detail, and method of data collection. In the United States, there are multiple large national databases established by organizations to collect patient demographics, associated diagnoses and procedures, and associated comorbidities and complications. The Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS) is one of the largest and tracks millions of patient encounters each year from a stratified systematic random sample of hospitals across the United States.3 Likewise, the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) collects data on surgical outcomes, among other data, from participating institutions.1

The current reliance on large databases and the substantial number of publications using them may have outpaced the evaluation of databases for consistency and accuracy. Recent studies have begun to compare national databases and have found discrepancies between them within the field of orthopedic surgery.12–17 However, it is unclear why reported complication rates differ.

For this study, the current authors evaluated NIS and NSQIP because they are large nationwide databases with a subset of overlapping hospitals. These characteristics allow for a comparison of overall complication rates as well as a narrowed comparison between the overlapping hospitals. Notably, these databases vary in the period over which they assess complications. Whereas NIS includes only inpatient stay, NSQIP includes inpatient stay and the first 30 days postoperatively. They also vary in the method of data extraction; NIS uses claims data, whereas NSQIP uses reporting from specifically trained registered nurses. Therefore, multiple variables could explain the potential differences in the data, including the timing of data collection, the method of data collection, and sampling error.

The authors evaluated complication rates among patients undergoing shoulder arthroplasty across databases because this area has not yet been studied and this procedure is a commonly performed orthopedic surgery with a known and relatively high complication rate.18–20 The authors hypothesized that there would be a significant difference in complication rates for this procedure across the 2 databases, as has been seen with other orthopedic procedures. The authors further hypothesized that this difference would abate when the NIS data set was restricted solely to the subset of hospitals that were also included in NSQIP over the same period.

Materials and Methods

Data for the NIS are obtained by tracking millions of patient encounters each year from participating hospitals across the United States.3 The NIS data set is a sample of the State Inpatient Databases, a state-level collection of databases within NIS. Data are weighted to provide a nationally representative estimate of the US inpatient population. Each observation of the data set is collected through patient billing information that includes admission to discharge. Information collected includes patient demographics, associated diagnoses and procedures, comorbidities, and complications. Diagnoses, procedures, and complications are coded with the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM).

Data for the NSQIP are obtained by collecting surgical outcome data from participating institutions. Included in the database are patient demographics, diagnoses, procedures, complications, and comorbidities. Procedures are reported with the Current Procedural Terminology (CPT). Complications resulting from procedures that occur during up to 30 days of follow-up are interpreted from patient charts by registered nurses with specific training in NSQIP data collection.

The authors analyzed data from 2006 to 2010 that were obtained from 1051 hospitals included in NIS and 663 hospitals included in NSQIP, with 57 overlapping hospitals. Overlapping hospitals were defined as those that contributed to both NIS and NSQIP. To narrow data solely to overlapping hospitals, the authors used the published list of hospitals that contributed to NSQIP and restricted the NIS data to include only those 57 hospitals. Unlike NIS data, NSQIP data are not linked to the associated hospital, so the authors were unable to narrow the data set to solely the 57 overlapping hospitals. The final narrowed comparison was between the restricted NIS data set and the entire NSQIP data set.

The authors identified inpatient shoulder arthroplasty procedures in the NIS data with the following ICD-9-CM codes: total shoulder replacement (81.80), partial shoulder replacement (81.81), and other repair of shoulder (81.83).

For the NSQIP data, the authors used CPT codes to identify relevant procedures corresponding to the same NIS procedures. The following codes were used to identify shoulder arthroplasty procedures in the NSQIP data: shoulder replacement (23472) and shoulder hemiarthroplasty (23470). Revisions (23331, 23332) were excluded.

The authors defined a complication as any patient encounter that resulted in at least one major outcome tracked by NSQIP. To accommodate differences in data collection methods, they created a list of ICD-9-CM codes in NIS to match the complications reported in NSQIP. First, the NSQIP complication groups were matched to Clinical Classification Software (CCS) categories, a system used by NIS to organize similar ICD-9-CM codes into groups. Within the CCS category, the ICD-9-CM codes that best matched the NSQIP complication group descriptions were selected. For example, the NSQIP complication of infection was matched to CCS 16.10.2.6 (postoperative infection), which was narrowed down to include the ICD-9-CM codes of 998.59 (other postoperative infection) and 998.51 (infected postoperative seroma). Irrelevant ICD-9-CM codes from that CCS, such as 519.01 (tracheostomy infection), were excluded. The same was done for NSQIP complications of wound dehiscence, pneumonia, deep venous thrombosis/pulmonary embolism, renal failure, urinary tract infection, coma, neurological injury (peripheral; shoulder), cardiac arrest requiring cardiopulmonary resuscitation, myocardial infarction, bleeding/transfusion, sepsis, septic shock, and death. The authors were unable to track 2 NSQIP complications, including unplanned intubation and failure to wean from a ventilator, in NIS. Therefore, these complications were excluded from analysis. In addition, the authors excluded graft, prosthesis, and implant failure as a complication because it could not be determined whether the diagnosis was the primary purpose of the procedure rather than an outcome.

The NSQIP data recorded patient age for all patients older than 89 years as “90+” to avoid potentially identifiable information. For calculation of mean age, the authors assigned the age of 90 years to those patients.

Statistical Analysis

Statistical analysis was performed with Stata software version 13 (Stata-Corp, College Station, Texas), to determine the complication rate and the 95% confidence interval. The authors compared complication rates for all shoulder arthroplasty procedures and performed separate comparisons for total shoulder arthroplasty and hemiarthroplasty procedures. A secondary analysis of all shoulder arthroplasty procedures was performed with the NIS data set restricted to hospitals that were also included in NSQIP. In NIS, annual population estimates of the number of procedures and complication rates were calculated with the SVYSET command in Stata to account for the complex survey design of the NIS data. To generate confidence intervals, strata with a single sampling unit were dropped from the analysis.

Results

Within the study period, the authors identified a total of 48,287 patients in NIS with a mean age of 68.8 years and 1679 patients in NSQIP with a mean age of 67.8 years who underwent shoulder arthroplasty. The complication rate for shoulder arthroplasty was significantly higher in NIS (12.6%; 95% confidence interval, 12.0%–13.2%) than in NSQIP (5.60%; 95% confidence interval, 4.59%–6.81%). When the NIS data set was restricted to institutions that also participated in NSQIP, the authors identified 3526 shoulder arthroplasty procedures within the study period, with a complication rate of 13.4% (95% confidence interval, 11.2%–15.8%) (Figure 1).

Graph of complication rates for shoulder arthroplasty from the Nationwide Inpatient Sample (NIS) and National Surgical Quality Improvement Program (NSQIP) databases with 95% confidence intervals (“shoulder arthroplasty”) and complication rates when the NIS data set was restricted to only the hospitals also participating in the NSQIP database (“matched shoulder arthroplasty”).

Figure 1:

Graph of complication rates for shoulder arthroplasty from the Nationwide Inpatient Sample (NIS) and National Surgical Quality Improvement Program (NSQIP) databases with 95% confidence intervals (“shoulder arthroplasty”) and complication rates when the NIS data set was restricted to only the hospitals also participating in the NSQIP database (“matched shoulder arthroplasty”).

In subgroup analysis, total shoulder procedures had a complication rate of 8.99% (95% confidence interval, 8.40%–9.63%) in NIS and 5.71% (95% confidence interval, 4.51%–7.21%) in NSQIP. The complication rate for shoulder hemiarthroplasty procedures was 17.9% (95% confidence interval, 17.1%–18.8%) in NIS and 5.35% (95% confidence interval, 3.72%–7.65%) in NSQIP.

Discussion

Database studies are common in the orthopedic literature and are frequently used for outcomes research on shoulder arthroplasty.21–24 As reported elsewhere in general and for other orthopedic procedures, large databases such as NIS and NSQIP have limitations.25–30 One limitation may be the accuracy and consistency of the data. These inconsistencies and inaccuracies have been shown by comparing outcomes between national databases for other procedures, but not, to the authors' knowledge, for shoulder arthroplasty.

The authors identified a significant difference in complication rates when comparing shoulder arthroplasty across these 2 databases. The complication rate in NIS for all shoulder arthroplasty procedures was more than twice as high as that in NSQIP. Because there is a small cohort of overlapping hospitals between NIS and NSQIP, the authors were able to perform a more refined comparison of the reported complication rates for the same procedures from 2006 to 2010 by restricting NIS data to the hospitals included in NSQIP. Rather than finding the same complication rate, or even a convergence of the complication rate, as the authors would expect when the patient populations became more similar, the NIS complication rate for shoulder arthroplasty was unexpectedly higher and more divergent from the NSQIP rate. The complication rates for shoulder arthroplasty fit within the range of previously reported complication rates, which vary widely—depending on the indication, procedure, and follow-up—from as low as 3.2% to as high as 25%.19,21,31–33

To the authors' knowledge, this is the first comparative study of surgical complication rates for shoulder arthroplasty between large databases and the first comparison performed for any orthopedic procedure that refined the data source to hospitals shared between databases.

There are many potential explanations for the difference in complication rates across the NIS and NSQIP databases (Table 1). The sampling frames of the 2 databases differ. By design, NIS data are nationally representative, whereas NSQIP data provide a convenience sample. This difference may result in samples with different preoperative comorbidities and hence a differential risk of complications. However, this difference is unlikely to explain the observed variation, as evidenced by the even larger gap in complication rates when the NIS data set was restricted to overlapping hospitals and patient encounters.

Differences Between NIS and NSQIP Databases

Table 1:

Differences Between NIS and NSQIP Databases

Another potential explanation could be the difference in the length of postsurgical follow-up between the databases. However, this difference is also unlikely to explain the difference because the NIS data, which showed a higher complication rate, had a shorter postoperative data collection period. If the length of follow-up had strongly influenced the complication rate, the opposite would be seen.

There could potentially be a difference as a result of misclassification of ICD-9-CM codes because they do not directly map to CPT codes. However, although the codes do not directly map, the authors used a broad number of CPT codes that should encompass all or nearly all of the procedures identified by ICD-9-CM codes.

The authors believe that the most likely explanation for their findings is the difference in the way in which complications are reported in each database, which may lead to inaccuracies. In NIS, complications are reported indirectly through billing and coding of the inpatient stay. Administrative claims data have been criticized for potential inaccuracy because of the disconnect between clinical care and billing.25–27 In NSQIP, registered nurses from each institution report complications based on a retrospective chart review. Although the discretion of well-trained nurses may more accurately define a complication than can billing software—and indeed the NSQIP method has been called a “sort of gold standard retrospective review”—the involvement of humans introduces additional bias.12 Nurses subjectively and voluntarily report complications that conceivably reflect the quality of care at their institutions, and they may be subconsciously biased to report lower complication rates. This explanation, which suggests that differences in these databases are inherent to the method of data collection, is concerning because it suggests that, despite the large size and careful sampling design of such databases, differences in complication rates always may be observed.

Earlier studies compared large databases and found discrepancies in complication rates after orthopedic surgery.12–14 Bohl et al15 compared patient demographics, comorbidities, and adverse events among patients with hip fractures within NIS and NSQIP and found large discrepancies in 4 of the 8 recorded comorbidities as well as 2 of the 12 recorded inpatient adverse events. In another study, Bohl et al12 examined the same data sets for lumbar fusion procedures and found large differences in comorbidities and inpatient adverse events, including greater than twofold discrepancies in rates of sepsis (0.38% in NIS, 0.81% in NSQIP) and acute kidney injury (1.79% in NIS, 0.21% in NSQIP). Bekkers et al16 found similar discrepancies within the total hip arthroplasty population between 2 large databases, and Bedard et al17 found variations in rates among 4 large databases within total knee arthroplasty.

This study had several limitations. Because NSQIP does not identify data by hospital, the data set could not be refined to certain hospitals that also participated in NIS. Ideally, the authors would have refined both data sets to solely the overlapping hospitals, but the nature of NSQIP prevented that. Nevertheless, the finding of a diverging complication rate when NIS data were narrowed to make the hospitals and patients in the databases more similar shows that the discrepancy between the databases likely is not related to hospital and patient sampling. Additionally, the authors reported an overall complication rate and did not report data on specific complications. Including this information would have strengthened this study, but the absence of this analysis does not invalidate the results. Of note, this study adhered to the methods put forth by Stulberg and Haut28 for the proper use of NIS data for research.

The US health care system is trending toward increased use of outcome metrics and complications for reimbursements and comparison of hospitals and surgeons. Although large-scale prospective studies offer the truest estimate of complication rates and other outcomes, these studies are often expensive and take years to perform. For this reason, database studies are increasingly used to provide this information. Undoubtedly, these studies offer many advantages, and they have their place in outcomes research. This study shows the flaws in database data on the complications of total shoulder arthroplasty. Further studies on the accuracy of data and the consistency between databases for other orthopedic procedures would be of value to ensure the use of the highest-quality data in calculations of performance and reimbursements for hospitals and surgeons.

Conclusion

The 2 large databases compared in this study had statistically significant differences in reported complication rates for shoulder arthroplasty. This difference remained and increased when 1 data set was restricted to overlapping hospitals. This discrepancy is mostly likely caused by the use of varying reporting methods. Based on this information, careful consideration should be given to reported complication rates for shoulder arthroplasty and potentially other procedures, based on a single large database.

References

  1. Maggard-Gibbons M. The use of report cards and outcome measurements to improve the safety of surgical care: the American College of Surgeons National Surgical Quality Improvement Program. BMJ Qual Saf. 2014;23(7):589–599. https://doi.org/10.1136/bmjqs-2013-002223 PMID: doi:10.1136/bmjqs-2013-002223 [CrossRef]24748371
  2. Stey AM, Russell MM, Sugar CA, et al. Extending the value of the National Surgical Quality Improvement Program claims dataset to study long-term outcomes: rate of repeat ventral hernia repair. Surgery. 2015;157(6):1157–1165. https://doi.org/10.1016/j.surg.2014.12.027 PMID: doi:10.1016/j.surg.2014.12.027 [CrossRef]25731782
  3. Healthcare Cost and Utilization Project. NIS overview. Retrieved from www.hcup-us.ahrq.gov/nisoverview.jsp
  4. Bohl DD, Grauer JN, Leopold SS. Editor's spotlight/Take 5: Nationwide Inpatient Sample and National Surgical Quality Improvement Program give different results in hip fracture studies. Clin Orthop Relat Res. 2014;472(6):1667–1671. https://doi.org/10.1007/s11999-014-3559-0 PMID: doi:10.1007/s11999-014-3559-0 [CrossRef]24711130
  5. Ahldén M, Samuelsson K, Sernert N, Forssblad M, Karlsson J, Kartus J. The Swedish National Anterior Cruciate Ligament Register: a report on baseline variables and outcomes of surgery for almost 18,000 patients. Am J Sports Med. 2012;40(10):2230–2235. https://doi.org/10.1177/0363546512457348 PMID: doi:10.1177/0363546512457348 [CrossRef]22962296
  6. D'Apuzzo MR, Browne JA. Obstructive sleep apnea as a risk factor for postoperative complications after revision joint arthroplasty. J Arthroplasty. 2012;27(8 suppl):95–98. https://doi.org/10.1016/j.arth.2012.03.025 PMID: doi:10.1016/j.arth.2012.03.025 [CrossRef]22917083
  7. Molina CS, Thakore RV, Blumer A, Obremskey WT, Sethi MK. Use of the National Surgical Quality Improvement Program in orthopaedic surgery. Clin Orthop Relat Res. 2015;473(5):1574–1581. https://doi.org/10.1007/s11999-014-3597-7 PMID: doi:10.1007/s11999-014-3597-7 [CrossRef]
  8. Rahr-Wagner L, Thillemann TM, Pedersen AB, Lind MC. Increased risk of revision after anteromedial compared with transtibial drilling of the femoral tunnel during primary anterior cruciate ligament reconstruction: results from the Danish Knee Ligament Reconstruction Register. Arthroscopy. 2013;29(1):98–105. https://doi.org/10.1016/j.arthro.2012.09.009 PMID: doi:10.1016/j.arthro.2012.09.009 [CrossRef]23276417
  9. Sedrakyan A, Romero L, Graves S, et al. Survivorship of hip and knee implants in pediatric and young adult populations: analysis of registry and published data. J Bone Joint Surg Am. 2014;96(suppl 1):73–78. https://doi.org/10.2106/JBJS.N.00541 PMID: doi:10.2106/JBJS.N.00541 [CrossRef]25520422
  10. Wei DH, Hawker GA, Jevsevar DS, Bozic KJ. Improving value in musculoskeletal care delivery: AOA critical issues. J Bone Joint Surg Am. 2015;97(9):769–774. https://doi.org/10.2106/JBJS.N.00841 PMID: doi:10.2106/JBJS.N.00841 [CrossRef]25948524
  11. Calderwood MS, Kleinman K, Murphy MV, Platt R, Huang SS. Improving public reporting and data validation for complex surgical site infections after coronary artery bypass graft surgery and hip arthroplasty. Open Forum Infect Dis. 2014;1(3):ofu106. https://doi.org/10.1093/ofid/ofu106 PMID: doi:10.1093/ofid/ofu106 [CrossRef]
  12. Bohl DD, Russo GS, Basques BA, et al. Variations in data collection methods between national databases affect study results: a comparison of the Nationwide inpatient Sample and National Surgical Quality Improvement Program databases for lumbar spine fusion procedures. J Bone Joint Surg Am. 2014;96(23):e193. https://doi.org/10.2106/JBJS.M.01490 PMID: doi:10.2106/JBJS.M.01490 [CrossRef]25471919
  13. Davenport DL, Holsapple CW, Conigliaro J. Assessing surgical quality using administrative and clinical data sets: a direct comparison of the University HealthSystem Consortium Clinical Database and the National Surgical Quality Improvement Program data set. Am J Med Qual. 2009;24(5):395–402. https://doi.org/10.1177/1062860609339936 PMID: doi:10.1177/1062860609339936 [CrossRef]19584374
  14. Simorov A, Bills N, Shostrom V, Boilesen E, Oleynikov D. Can surgical performance benchmarking be generalized across multiple outcomes databases: a comparison of University HealthSystem Consortium and National Surgical Quality Improvement Program. Am J Surg. 2014;208(6):942–948. https://doi.org/10.1016/j.amjsurg.2014.08.016 PMID: doi:10.1016/j.amjsurg.2014.08.016 [CrossRef]25440482
  15. Bohl DD, Basques BA, Golinvaux NS, Baumgaertner MR, Grauer JN. Nationwide Inpatient Sample and National Surgical Quality Improvement Program give different results in hip fracture studies. Clin Orthop Relat Res. 2014;472(6):1672–1680. https://doi.org/10.1007/s11999-014-3559-0 PMID: doi:10.1007/s11999-014-3559-0 [CrossRef]24615426
  16. Bekkers S, Bot AGJ, Makarawung D, Neuhaus V, Ring D. The National Hospital Discharge Survey and Nationwide Inpatient Sample: the databases used affect results in THA research. Clin Orthop Relat Res. 2014;472(11):3441–3449. https://doi.org/10.1007/s11999-014-3836-y PMID: doi:10.1007/s11999-014-3836-y [CrossRef]25091226
  17. Bedard NA, Pugely AJ, McHugh M, et al. Analysis of outcomes after TKA: do all databases produce similar findings?Clin Orthop Relat Res. 2018;476(1):52–63. https://doi.org/10.1007/s11999.0000000000000011 PMID: doi:10.1007/s11999.0000000000000011 [CrossRef]29529616
  18. Favard L, Katz D, Colmar M, Benkalfate T, Thomazeau H, Emily S. Total shoulder arthroplasty. Arthroplasty for glenohumeral arthropathies: results and complications after a minimum follow-up of 8 years according to the type of arthroplasty and etiology. Orthop Traumatol Surg Res. 2012;98(4 suppl):S41–S47. https://doi.org/10.1016/j.otsr.2012.04.003 PMID: doi:10.1016/j.otsr.2012.04.003 [CrossRef]22583895
  19. Gadea F, Alami G, Pape G, Boileau P, Favard L. Shoulder hemiarthroplasty: outcomes and long-term survival analysis according to etiology. Orthop Traumatol Surg Res. 2012;98(6):659–665. https://doi.org/10.1016/j.otsr.2012.03.020 PMID: doi:10.1016/j.otsr.2012.03.020 [CrossRef]22944393
  20. Kiet TK, Feeley BT, Naimark M, et al. Outcomes after shoulder replacement: comparison between reverse and anatomic total shoulder arthroplasty. J Shoulder Elbow Surg. 2015;24(2):179–185. https://doi.org/10.1016/j.jse.2014.06.039 PMID: doi:10.1016/j.jse.2014.06.039 [CrossRef]
  21. Ponce BA, Oladeji LO, Rogers ME, Menendez ME. Comparative analysis of anatomic and reverse total shoulder arthroplasty: in-hospital outcomes and costs. J Shoulder Elbow Surg. 2015;24(3):460–467. https://doi.org/10.1016/j.jse.2014.08.016 PMID: doi:10.1016/j.jse.2014.08.016 [CrossRef]
  22. Jiang JJ, Toor AS, Shi LL, Koh JL. Analysis of perioperative complications in patients after total shoulder arthroplasty and reverse total shoulder arthroplasty. J Shoulder Elbow Surg. 2014;23(12):1852–1859. https://doi.org/10.1016/j.jse.2014.04.008 PMID: doi:10.1016/j.jse.2014.04.008 [CrossRef]25156959
  23. Issa K, Pierce CM, Pierce TP, et al. Total shoulder arthroplasty demographics, incidence, and complications: a Nationwide In-patient Sample database study. Surg Technol Int. 2016;29:240–246. PMID:27608744
  24. Jain N, Pietrobon R, Hocker S, Guller U, Shankar A, Higgins LD. The relationship between surgeon and hospital volume and outcomes for shoulder arthroplasty. J Bone Joint Surg Am. 2004;86(3):496–505. https://doi.org/10.2106/00004623-200403000-00006 PMID: doi:10.2106/00004623-200403000-00006 [CrossRef]14996874
  25. Bozic KJ, Chiu VW, Takemoto SK, et al. The validity of using administrative claims data in total joint arthroplasty outcomes research. J Arthroplasty. 2010;25(6 suppl):58–61. https://doi.org/10.1016/j.arth.2010.04.006 PMID: doi:10.1016/j.arth.2010.04.006 [CrossRef]20570479
  26. Best WR, Khuri SF, Phelan M, et al. Identifying patient preoperative risk factors and postoperative adverse events in administrative databases: results from the Department of Veterans Affairs National Surgical Quality Improvement Program. J Am Coll Surg. 2002;194(3):257–266. https://doi.org/10.1016/S1072-7515(01)01183-8 PMID: doi:10.1016/S1072-7515(01)01183-8 [CrossRef]11893128
  27. Sarrazin MSV, Rosenthal GE. Finding pure and simple truths with administrative data. JAMA. 2012;307(13):1433–1435. https://doi.org/10.1001/jama.2012.404 PMID: doi:10.1001/jama.2012.404 [CrossRef]22474208
  28. Stulberg JJ, Haut ER. Practical guide to surgical data sets: Healthcare Cost and Utilization Project National Inpatient Sample (NIS). JAMA Surg. 2018;153(6):586–587. https://doi.org/10.1001/jamasurg.2018.0542 PMID: doi:10.1001/jamasurg.2018.0542 [CrossRef]29617533
  29. Khera R, Angraal S, Couch T, et al. Adherence to methodological standards in research using the National Inpatient Sample. JAMA. 2017;318(20):2011–2018. https://doi.org/10.1001/jama.2017.17653 PMID: doi:10.1001/jama.2017.17653 [CrossRef]29183077
  30. Pugely AJ, Martin CT, Harwood J, Ong KL, Bozic KJ, Callaghan JJ. Database and registry research in orthopaedic surgery: Part I. Claims-based data. J Bone Joint Surg Am. 2015;97(15):1278–1287. https://doi.org/10.2106/JBJS.N.01260 PMID: doi:10.2106/JBJS.N.01260 [CrossRef]26246263
  31. Cuff DJ, Pupello DR. Comparison of hemiarthroplasty and reverse shoulder arthroplasty for the treatment of proximal humeral fractures in elderly patients. J Bone Joint Surg Am. 2013;95(22):2050–2055. https://doi.org/10.2106/JBJS.L.01637 PMID: doi:10.2106/JBJS.L.01637 [CrossRef]24257664
  32. Shields E, Iannuzzi JC, Thorsness R, Noyes K, Voloshin I. Postoperative morbidity by procedure and patient factors influencing major complications within 30 days following shoulder surgery. Orthop J Sports Med. 2014;2(10):2325967114553164–2325967114553164. https://doi.org/10.1177/2325967114553164 PMID: doi:10.1177/2325967114553164 [CrossRef]
  33. Simovitch RW, Gerard BK, Brees JA, Fullick R, Kearse JC. Outcomes of reverse total shoulder arthroplasty in a senior athletic population. J Shoulder Elbow Surg. 2015;24(9):1481–1485. https://doi.org/10.1016/j.jse.2015.03.011 PMID: doi:10.1016/j.jse.2015.03.011 [CrossRef]25958214

Differences Between NIS and NSQIP Databases

VariableNISNSQIPCould Variable Explain Variation?
Data collection periodInpatient stayInpatient stay + 30 days postoperativeNo—increased complications in NIS
Data collection methodAdministrative/claimsNurse chart reviewYes
Hospitals sampledNIS-specific stratified, systematic, random sampleNSQIP-specific convenience sample of invited hospitalsUnlikely—increased complications in narrowed analysis
DataRepresentative/weightedVoluntary/unweightedPossible
Complication codingICD-9-CMDatabase-specific binary variablesUnlikely—matched in the current study
Authors

The authors are from the Department of Orthopaedics (MJS, IDE, CC), Tufts Medical Center, and the Department of Public Health and Community Medicine (DJT), Tufts University School of Medicine (AXL, AHJ), Boston, Massachusetts.

The authors have no relevant financial relationships to disclose.

Correspondence should be addressed to: Matthew J. Salzler, MD, Department of Orthopaedics, Tufts Medical Center, 800 Washington St, Boston, MA 02111 ( MSalzler@tuftsmedicalcenter.org).

Received: November 14, 2018
Accepted: January 14, 2019
Posted Online: January 13, 2020

10.3928/01477447-20200107-05

Sign up to receive

Journal E-contents