The Spine Blog

Saturday, January 25, 2014

Risk Factors for Readmission Following Anterior Cervical Fusion

Readmission following surgery is a major driver of healthcare costs and is a major target for improvement from the perspective of payers and hospitals. Thirty day readmission rate is already used as a marker for quality of care and will likely affect reimbursement in the future. As such, hospitals and surgeons are strongly motivated to better understand why patients are readmitted and then take steps to reduce readmission rates. Dr. Kim and his colleagues from Northwestern performed the current study to evaluate risk factors for complications and readmissions following anterior cervical fusion (ACF) surgery (including both ACDF and corpectomy). They used the National Surgical Quality Improvement Program (NSQIP) database and included 2,320 ACF patients undergoing surgery in 2011, of whom 59 were readmitted—a 30 day readmission rate of 2.5%. Age over 65 years, diabetes, hypertension, prior stroke, prior cardiac revascularization procedure, higher ASA score, and having been admitted to the hospital for at least 24 hours prior to surgery all increased the risk of readmission in a univariate analysis. Given that 32% of readmitted patients had complications recorded in the database, the authors also wanted to evaluate risk factors for complications, and they found that age over 65 years, being admitted to the hospital for over 24 hours prior to surgery, and longer OR time were all independent risk factors for complications in a multivariate analysis. They also performed an analysis to determine independent risk factors for readmission using a multivariate model with complication as a covariate and found that the only significant risk factor for readmission was having a complication (hypertension was borderline significant). Based on these results, they concluded that avoiding complications would be the best way to reduce readmission rates.


This paper demonstrates both the power and limitations of large, administrative database studies. Using the NSQIP database, the authors were quickly able to gather data on over 2,320 ACF patients, a task that would likely involve years of enrollment in a multicenter study. While this seems like a relatively large sample size, with a readmission rate of only 2.5%, the authors were left with only 59 patients who were readmitted. With this relatively small cohort of readmissions, their ability to perform multivariate analysis to evaluate risk factors became quite limited as the size of the subgroups with certain characteristics became quite small (i.e. there was only one patient with COPD in the readmitted group). An additional limitation of administrative database studies looking at complications is that only complications recorded in the database can be studied. The authors pointed out that complications such as dysphagia, hardware migration or failure, hematoma, and CSF leak are not captured in this database and were likely responsible for some of the readmissions. Given that almost all patients who are readmitted within 30 days are readmitted for some type of “complication”, the fact that only 32% of the readmitted cohort had a complication listed indicates how many of the relevant complications were not captured. The distinction they make between a complication and a readmission leads to some analytical confusion as they performed a multivariate analysis to determine if there were risk factors for readmission independent of having a complication. Not surprisingly, in a model including “having a complication” as a covariate, there were no significant independent risk factors for readmission (hypertension had a p-value of 0.051). It would have been interesting if they had performed a similar analysis without “having a complication” as a covariate to determine the independent risk factors for readmission, which was really the question at hand. Not surprisingly, age, non-elective surgery (as indicated by being admitted for over 24 hours prior to surgery), and longer operative time were risk factors for complications. It seems likely that tumor, trauma, infection, and myelopathy patients have higher rates of complications and readmissions than those undergoing ACDF for degenerative conditions with radiculopathy, though the underlying diagnosis was not included as a variable in the models. This paper provides a good benchmark 30 day readmission rate following ACF (2.5%). Future studies will need to further explore risk factors for readmission, both so that patients with these characteristics can be under closer surveillance and so that these factors can be used in risk adjustment models when evaluating institutions’ readmission rates.


Please read Dr. Kim’s article on this topic in the January 15 issue. Does this paper change your views on readmission following ACF? Let us know by leaving a comment on The Spine Blog.

Adam Pearson, MD, MS

Associate Web Editor