Methods

In this nationwide cross-sectional study primary hip and knee arthroplasty performed between January 1, 2009 and December 31, 2019 were identified from the Medicare physician service records. These records were compiled by the Centers for Medicare and Medicaid Services (CMS), and after deidentification, were made available for research, known as the Limited Data Set (LDS). Specifically, physician records associated with a 5% sample of Medicare beneficiaries, equivalent to the records from approximately 2.5 million enrolees, formed the basis of this study. The population of interest included elderly Medicare patients (ages 65 and above) who received a primary hip or primary knee arthroplasty during the study period. Since the CMS data is deidentified, it was exempt from review by the Institutional Review Board.

Procedures were identified using Current Procedural Terminology (CPT) codes: 27,125, 27,130, and 27,132 for hip arthroplasty, and 27,446 and 27,447 for knee arthroplasty. Diagnoses were coded using ICD-9-CM (before October 1, 2015) and ICD-10-CM (thereafter). Steps were taken to retain only one record per procedure and to derive laterality from ICD-10 codes, CPT modifiers, or related procedures. Laterality was identified in approximately 95% of arthroplasties.

Three measures were used for evaluation: (a) the frequency of joint arthroplasty performed Friday, the 13th compared to other Fridays, (b) septic and aseptic revision risk for arthroplasty performed on Friday, the 13th and other Fridays (c) survival and mortality of patients with arthroplasty performed on Friday the 13th and other Fridays. The CPT codes used to identify hip revisions were: 27,134, 27,137, 27,138, as well as 27,090 and 27,091. For knee revisions, the CPT codes used were: 27,486 and 27,487, as well as 27,488. Septic revision is defined by having a diagnosis of 996.66 (ICD-9) or T84.5x (ICD-10) indicating infection and inflammation reaction of internal joint prosthesis in the 30 days on or before the date of the revision arthroplasty. The 30-day window was chosen to capture infection diagnoses that were temporally linked to the revision procedure, consistent with established perioperative infection definitions in claims-based research. Revisions without finding any diagnosis indicating infection and inflammation are classified as aseptic. Date of death was identified from the Medicare enrolment data, also provided by the CMS. Individuals were tracked from the arthroplasty date until occurrence of the outcome (revision), until death, until end of enrolment, or until end of study on December 31, 2019, which ever came first.

Survival analysis techniques were used to analyse these outcomes. The Kaplan-Meier (KM) method with the Fine and Gray sub-distribution adaptation was used to calculate the cumulative incidence rate of the septic and aseptic revisions. Septic and aseptic revisions were considered mutually exclusive, and therefore “competing risk” events. For patients undergoing bilateral or staged arthroplasty, each procedure was treated as an independent event, with laterality-specific follow-up for revision outcomes. Multiple procedures per patient were permitted in the analysis. Cases with missing covariate information were excluded from multivariable models. The Fine and Gray technique was used to account for death as a competing event [9]. For mortality or survivorship, the conventional KM method was appropriate. The semiparametric Cox regression was used, also accounting for competing risk, to investigate these outcomes and to compare the risk among patients receiving their arthroplasty on Friday 13th or control dates, after adjusting for a number of potential confounding factors. The demographic factors included: age, sex, race, resident region, and Medicare buy-in (as a surrogate for patient’s economic status). Clinical factors included were osteoporosis, obesity, diabetes mellitus, rheumatoid disease, chronic kidney disease, tobacco dependence, regular use of anti-coagulant, regular NSAID use, prior osteoporotic fracture, hypertensive disease, ischemic heart disease, cerebrovascular disease, COPD, and congestive heart failure. These conditions could appear as either primary or secondary diagnosis, but at least 2 mentions of such condition in the prior year was required to reduce false or “suspect-only” diagnosis. Socioeconomic measures for the patients’ county of residence (median household income, educational attainment, poverty rate, unemployment, urban-rural classification) were obtained from publicly available U.S. Census and federal data sources. All data processing and statistical analyses were performed using the SAS statistical software (Version 9.4, Cary, NC) and significance was determined at α = 0.05.