Diabetes Disease Severity Was Not Associated with Risk of Deep Infection or Revision After Shoulder Arthroplasty

Abstract

Background: Prior studies have identified diabetes and disease severity (defined using hemoglobin A1c [HbA1c]) as potential risk factors for complications after shoulder arthroplasty. Evaluations of diabetes status and risk of adverse outcomes beyond the 30-day window either are limited or have not accounted for disease severity. Further, measures of diabetes severity other than HbA1c have yet to be investigated in a shoulder arthroplasty population.

Questions/purposes: (1) Are diabetes status and glycemic control associated with adverse events, including deep infection, all-cause revision, and 90-day readmission after shoulder arthroplasty? (2) Is postoperative HbA1c associated with revision risk? (3) Is there a threshold of preoperative HbA1c that best identifies patients with diabetes who are at higher risk of 3-year deep infection, 1-year all-cause revision, or 90-day readmission? (4) Can the Adapted Diabetes Complications Severity index (aDCSI) be used as an alternative measure of diabetes severity in evaluating the risk of deep infection, all-cause revision, and 90-day readmission and identification of patients with diabetes at higher risk for these events? (5) Is there a difference between elective and traumatic shoulder arthroplasty patients?

Methods: We conducted a retrospective registry-based cohort study using Kaiser Permanente's Shoulder Arthroplasty Registry (2005-2015). Primary shoulder arthroplasties were classified as patients with and without diabetes. Patients with diabetes were further evaluated using two disease severity measures (1) HbA1c, with good glycemic control classified as preoperative HbA1c < 7.0% and poor control defined as HbA1c ≥ 7.0%; and (2) aDCSI, classified as mild (score of 0-2) or severe (score ≥ 3) diabetes. Cox regression was used to evaluate the risk of deep infection and revision according to diabetes status and disease severity; conditional logistic regression was used for 90-day readmission. Time-dependent 1-year postoperative HbA1c was used to evaluate revision risk in Cox regression. All models were adjusted for covariates and stratified by elective versus trauma shoulder arthroplasty. Receiver operating characteristic curves were generated for HbA1c and aDCSI to determine whether a threshold exists to identify patients at higher risk of deep infection, all-cause revision, or 90-day readmission. The study sample consisted of 8819 patients; 7353 underwent elective shoulder arthroplasty and 1466 underwent shoulder arthroplasty due to trauma. For elective shoulder arthroplasty, 1430 patients (19%) had diabetes, and among the patients who underwent arthroplasty due to trauma, 444 (30%) had diabetes.

Results: Patients with diabetes who underwent elective shoulder arthroplasty and had poor glycemic control had a higher likelihood of 90-day readmission compared with patients without diabetes (OR, 1.5; 95% CI, 1.0-2.1; p = 0.032). No association was found for patients with diabetes who underwent shoulder arthroplasty due to trauma. No association was found between postoperative HbA1c and revision risk. Receiver operating characteristic curve analysis suggested preoperative HbA1c performed poorly at differentiating adverse events. When using aDCSI, patients with severe diabetes who underwent both elective and traumatic shoulder arthroplasty had a higher likelihood of 90-day readmission compared with patients without diabetes (OR, 1.6; 95% CI, 1.2-2.2; p = 0.001 and OR, 1.8; 95% CI, 1.2-2.7; p = 0.005, respectively). Similar to HbA1c, the aDCSI was a poor classifier in differentiating adverse events.

Conclusions: Of the longer-term outcomes evaluated, more-severe diabetes was only found to be associated with an increase in 90-day readmissions after shoulder arthroplasty; a stronger association was found when using the aDCSI in identifying diabetes severity. Arbitrary cutoffs in HbA1c may not be the best method for determining risk of postoperative outcomes. Future work investigating perioperative diabetes management should work to identify and validate measures, such as the aDCSI, that better identify patients at higher risk for postoperative outcomes and, more importantly, whether outcomes can be improved by modifying these measures with targeted interventions.

Level of evidence: Level III, therapeutic study.

Fig. 1

Receiver operating characteristic curves with cutoff values of preoperative HbA1c and aDCSI score for patients with diabetes. The reference line (dashed black line) is a guideline of a test unable to diagnose patients with the outcome of interest (deep infection, all-cause revision, and 90-day readmission) better than chance, or where the AUC = 0.50. The AUC for elective procedures was as follows: using HbA1c – infection = 0.53, revision = 0.57, readmission = 0.52; using aDCSI score – infection = 0.63, revision = 0.49, readmission = 0.63. The AUC for trauma procedures was as follows: using HbA1c – infection = 0.46, revision = 0.56, readmission = 0.52; using aDCSI score – infection = 0.67, revision = 0.41, readmission = 0.71; aDCSI = adapted diabetes complications severity index; AUC = area under the curve; HbA1c = hemoglobin A1c.