This retrospective analysis included patient data from the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database from 2005 to 2018. This project is exempt from the Institutional Review Board as this database is de-identified, without direct patient involvement.

Patients aged ≥18 years, who underwent elective single-level PCDF, were identified and included based on the Current Procedural Terminology (CPT) code 22600, in conjunction with a cervical decompression code (63045, 63001, 63015, or 22210). We evaluated multilevel fusion using CPT 22614. Patients were excluded if they had CPT codes for anterior, thoracic, and/or lumbar procedures; deformity, revision, nonelective, or spinal tumor surgery; if they were admitted from non-home settings; discharged anywhere other than home or rehabilitation; or had emergency surgery, preoperative systemic inflammatory response syndrome or sepsis, disseminated cancer, or decompression without posterior cervical fusion.

The NSQIP database provides data on patients treated as inpatient or outpatient as defined by the NSQIP Participant Use Data File [6]. Patients were categorized into either inpatients or outpatients based on the provided variable.

Primary outcomes were 30-day readmission, reoperation, and morbidity. Readmission includes any inpatient stay at the same or another hospital related to the surgical procedure [6]. The NSQIP database did not collect readmission data until 2011. Reoperation and morbidity outcomes were collected from the start of the dataset in 2005. Reoperation includes all major surgical procedures requiring unplanned operating room returns [6]. Morbidity includes infectious, pulmonary, cardiac, renal, neurological, hematologic, and thromboembolic complications reported in the dataset [6,7]. Primary outcomes, as well as specific complications, were compared between inpatient and outpatient PCDF.

Predictors of primary outcomes were analyzed amongst the entire cohort. Variables evaluated as potential predictors included patient demographic, comorbidity, preoperative lab values, and procedural factors (Table 1). Variables with <80% of available data were excluded from the multivariate analysis to avoid distorted results [7].

Analyses were completed using the IBM SPSS ver. 28.0 (IBM Corp., Armonk, NY, USA). Demographic, comorbidity, laboratory, and procedural factors were individually analyzed for baseline differences between inpatients and outpatients using the Student t-test for continuous and chi-square or Fisher’s exact test for categorical variables. Additionally, the above factors were individually analyzed for association with the primary outcomes using univariate logistic regression. Potential predictor variables from the univariate analyses that were significant (p<0.05) [8], as well as surgical setting, were then evaluated for significance (p<0.05) as independent predictors and control variables in a series of multivariate logistic regression analyses of the primary outcomes.

Unadjusted analysis revealed that outpatients had significantly lower rates of readmission (4.7% versus 8.8%, p=0.020), reoperation (1.7% versus 3.8%, p=0.038), and morbidity (4.5% versus 11.2%, p<0.001) compared to inpatients (Table 1). Inpatient PCDF was associated with urinary tract infection (UTI) (1.6% versus 0%, p=0.018) and transfusion (4.4% versus 0.3%, p<0.001) (Table 2).

After adjusting for significant baseline differences and primary outcome predictors, the multivariate analysis revealed that readmission (p=0.236; odds ratio [OR], 1.436), reoperation (p=0.524; OR, 1.317), and morbidity (p=0.716; OR, 1.108) no longer statistically differed between surgical setting (Tables 3–5). Additionally, UTI and stroke no longer statistically differed between surgical settings (Table 2).

Readmissions occurred in 620 patients (8.6%). The multivariate analysis revealed age (p=0.008; OR, 1.012), unexpected weight loss (p=0.045; OR, 2.444), and increased creatinine (p<0.001; OR, 2.233) as independent readmission predictors (Table 3).

Reoperations occurred in 335 patients (3.8%). ASA-class of ≥3 (p=0.028; OR, 1.406) and prior open wound (p=0.040; OR, 2.557) predicted reoperation on multivariate analysis (Table 4). Reoperations most occurred due to site-related complications (65.2%), followed by hardware-related complications (16.3%). No statistically significant differences were found in the reasons for reoperation between inpatient and outpatient settings (Fig. 1).

Morbidity occurred in 972 patients (10.9%). Discharge to rehabilitation (p<0.001; OR, 1.412), ASA-class of ≥3 (p=0.008; OR, 1.296), decreased hematocrit (p<0.001; OR, 1.700), and operative time (p<0.001; OR, 1.005) predicted morbidity on multivariate analysis (Table 5).

An increasing trend toward performing spine surgery was found in the outpatient setting, but literature evaluating outpatient PCDF is limited. In the present study, 30-day outcomes were statistically similar between inpatients and outpatients, indicating that PCDF can safely be performed in the outpatient setting.

The outcomes observed in the present study are consistent with current literature on outpatient spinal procedures [9,10]. Segal et al. [8] demonstrated that rates of readmission, reoperation, and complication were statistically similar between inpatient and outpatient cervical disc replacement. Bovonratwet et al. [9] showed that rates of bleeding events requiring transfusion for posterior lumbar fusion were significantly lower in the outpatient setting, without significant differences between inpatient and outpatient readmission or reoperation rates. Khanna et al. [11] demonstrated a significantly lower overall complication rate for anterior cervical discectomy and fusion in the outpatient setting compared with the inpatient setting, without significant differences in readmission and mortality rates. Low complication rates for outpatient discectomy have been well-described [10,12]. Inpatient and outpatient outcomes did not statistically differ despite significant variability in resources available among outpatient facilities, such as equipment availability, laboratory draw access, and specialty physician support. This may indicate that features intrinsic to outpatient surgery foster a safe and efficient operative environment [13]. This is evidenced by reduced operative time, lower cost, and statistically similar complication and readmission rates.

Moreover, the outcomes presented in this study are consistent with current literature that evaluates PCDF. Long-term PCDF studies have demonstrated reoperation rates ranging from 4% to 27% [14,15]. Our early reoperation rate of 3.8% for the entire cohort is in-line with these findings. Complication and readmission profiles of PCDF have been incompletely evaluated thus far. Current literature on PCDF provides varying data on complication rates; however, most studies demonstrate postoperative complication rates ranging from 15% to 25% [5,16,17]. Commonly reported complications or adverse events included axial pain, temporary neurologic deficit, and wound infection [5]. The PCDF literature provides scant data on readmission rates. Snyder et al. reported 30-day readmission rates following PCDF of 10% for non-home discharges and 6% for home discharges [18]. Cole et al. [19] reported a 30-day readmission rate of 9.9% for the posterior approach to multilevel degenerative cervical disease. Despite inadequate extensive data, these findings are consistent with our overall 30-day readmission rate of 8.6%.

Few studies have reported on predictors of poor early outcomes in PCDF. Snyder et al. [18] found that dependent functional status, diabetes, ASA-class of >2, older age, and female gender were significant predictors of non-home discharge, severe post-discharge adverse events, and unplanned readmission following PCDF. Tetreault et al. [20] found that a longer operation duration was predictive of perioperative complications. Similarly, our study found that older age independently predicted readmission, and ASA-class of ≥3 and longer operative time predicted morbidity on multivariate analysis. However, we did not obtain similar findings for dependent functional status, diabetes, and female gender. Several other studies have found operative time to predict complications and morbidity in spinal surgery [21,22]. The suggested benefit of reducing operative time further supports the shifting trend towards outpatient spinal surgery.

Baseline health status is often considered in deciding on inpatient or outpatient surgery for patients. Analysis revealed that this was evident in our present study, as patients in the inpatient cohort were more likely older and have medical comorbidities. Notably, outpatients had significantly greater rates of obesity and smoking, further suggesting that they were overall healthier by being selected for outpatient surgery, despite having these negative health attributes. These discrepancies were addressed by comparing readmission, reoperation, and morbidity before and after baseline difference adjustments. Before adjustment, the outpatient cohort had significantly lower readmission, reoperation, and morbidity rates. After adjustment, no statistically significant difference was found between the surgical settings in any of the measured variables.

Operative time was longer in the inpatient group. Similar findings were observed in the comparison of inpatient to outpatient cervical disc replacement [8]. On average, inpatient spinal surgery requires approximately 30 minutes more operative time than outpatient [23]. Our study showed that outpatient PCDF was 53 minutes faster on average. Some of this difference in operation time may be attributed to the performance of more complex cases in inpatient settings; however, the statistically similar reoperation, readmission, and morbidity rates, even after adjusting for baseline differences between cohorts, support the idea that outpatient surgery provides similar efficacy as evidenced by the increased number of performed levels.

The NSQIP database provides access to a large number of nationally represented patients from multiple institutions and highly relevant variables, allowing for generalizability and the development of meaningful predictive models [24]. Trained reviewers collect data for the database while adhering to stringent variable definitions, rendering it more reliable than other large-scale administrative databases [25].

Limitations exist with the NSQIP database, which uses hospital billing data to capture inpatient and outpatient status, and thus may not accurately reflect the length of stay [9]. Consequently, some researchers have treated patients with a zero-day length of stay as an outpatient, and greater than zero as an inpatient [9,10]. However, this method can introduce a potential bias by creating modified inpatient and outpatient groups, with fewer and more number of healthier patients, respectively. Additionally, determining that an unexpected negative intraoperative event would have resulted in an intended outpatient case to require inpatient admission is impossible; however, the utilization of billing data to determine inpatient-versus-outpatient status would suggest that the vast majority of any outpatient-converted-to-inpatient patient stay would have been appropriately coded as an inpatient. Therefore, we used the inpatient versus outpatient variable provided by the NSQIP dataset as-is, as has been previously done [8,10]. The detailed indications for outpatient surgery are not provided; however, various patient-related factors that are surrogates for outpatient selection criteria were controlled for in the analysis. Additionally, the database does not account for the evolution of spinal implants, technology, and techniques that have occurred over the period (i.e., facet fusion devices, navigation, robotics, etc.). Furthermore, a disproportionately greater number of cases are performed in inpatient settings compared to those in outpatient settings, which may potentially provide a temporal bias based on recent improvements in surgical technique, technology, and ambulatory surgery centers. Finally, posterior cervical decompression has many different types, such as total laminectomy, hemilaminectomy, laminotomy, or foraminotomy, which are not differentiated in the analysis. However, every included patient in the study underwent fusion, which would more strongly influence the overall morbidity and invasiveness than the utilized decompression method. Therefore, the decompression method is likely less influential given that every patient underwent fusion.