Prediction of Readmission and Complications After Pituitary Adenoma Resection via the National Surgical Quality Improvement Program (NSQIP) Database

Introduction Pituitary adenomas are common intracranial tumors (incidence 4:100,000 people) with good surgical outcomes; however, a subset of patients show higher rates of perioperative morbidity. Our goal was to identify risk factors for postoperative complications or readmission after pituitary adenoma resection. Methods We undertook a retrospective cohort study of patients who underwent surgery for pituitary adenoma in 2006-2018 by using the National Surgical Quality Improvement Program database. The main outcome measures were patient complications and the 30-day readmission rate. Results Among the 2,292 patients (mean age 53.3±15.9 years), there were 491 complications in 188 patients (8.2%). Complications and 30-day readmission have remained stable over time rather than declined. Unplanned readmission was seen in 141 patients (6.2%). Multivariable analysis demonstrated that hypertension (OR=1.6; 95% CI= 1.1, 2.1; p=0.005) and high white blood cell count (OR=1.08; 95% CI=1.03, 1.1; p=0.0001) were independent predictors of complications. Return to the operating room (OR=5.9, 95% CI=1.7, 20.2, p=0.0005); complications (OR=4.1, 95% CI=1.6, 10.6, p=0.004); and blood urea nitrogen (OR=1.08, 95% CI=1.02, 1.2, p=0.02) were independent predictors of 30-day readmission. Conclusion Using one of the largest datasets of pituitary adenoma patients, we identified perioperative factors most critical for patient outcome. One strength of this study is adjusting for cofactors that predict outcomes, which has not been done previously. Several patient biomarkers, namely white blood cell count and blood urea nitrogen, may serve as preoperative markers that might identify patients at higher risk. Control of blood pressure and renal disease may be perioperative management strategies that can impact the outcome.


Introduction
Tumors of the pituitary gland are the second most common intracranial tumor (incidence 4:100,000 people); their apparent incidence has increased over the past 30 years, likely because of advances in and accessibility of intracranial imaging as well as aging of the population [1][2][3]. Recent surgical trends have favored the endoscopic removal of pituitary tumors as an alternative to microscopic resection [4][5][6]. The prediction of peri-and postoperative outcomes has been emphasized in recent years with the increasing availability of electronic medical data, more formalized training for resection, objective measures of clinical outcomes, and improved understanding of the complication rates in pituitary adenoma resections [7][8][9]. Surgical databases offer the ability to review national data over long periods of time and identify subgroups of patients achieving better patient outcomes.
The purpose of this study was to identify the risk factors involved in the development of postoperative complications or readmission within 30 days after undergoing resection of a pituitary adenoma by using a national surgical database.

Data source
The American College of Surgeons National Surgical Quality Improvement Program (NSQIP) registry data from 2006 through 2018 were used for this study. The dataset provides perioperative data collected and recorded for the first 30 days after surgery at 400 hospitals that participate in the program throughout the U.S. [10]. The data are collected by trained research coordinators following an established protocol, and the database undergoes periodic data quality control. The common procedural terminology (CPT) codes 61546 (craniotomy for hypophysectomy or excision of pituitary tumor, intracranial approach), 61548 (hypophysectomy or excision of pituitary tumor, transnasal or transseptal approach, nonstereotactic), and 62165 (neuroendoscopy, intracranial; with excision of pituitary tumor, transnasal or trans-sphenoidal approach) were used for this study, and all patients with available variables were included. These review codes are the most consistent with identifying pituitary adenomas but are limited as they do not account for unlisted codes or other variations in practice that may be used to code for pituitary adenomas nationally. Institutional review board approval is not necessary because the data are de-identified upon entering into the registry.

Variables and outcomes
Missing variables were excluded during the tabulation of results. The primary outcomes were perioperative complications that occurred during the patient's admission and 30-day readmission rates, with each variable being distinctly coded in the database. A variable for "any complication" was generated using the prescribed variables in NSQIP except for the return to the operating room or 30-day readmission. Complications and return to the operating room were identified during the index surgery while 30-day readmission was after discharge and is a discrete coded variable in the NSQIP database. Patient demographic, biomarker, and clinical data were collected for classification.   Figure 1A) and a total of 141 (6.2%) had 30-day readmission (

Univariate analysis of 30-day readmission
A total of 141 patients (6.2%) underwent unplanned readmission within 30 days postoperatively, with 116 readmissions related to the original surgery (

Missing data
Of the patients found to have had a return to the operating room (n=91), a total of 68 did not have the proper CPT codes or International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes to allow us to accurately identify the procedure performed and the specific reason. Additionally, there were 523 patients of the 2,292 that did not have a definite "readmission" or "non-readmission" code. As such, we excluded those patients from our data analysis during the analysis of risk factors impacting the readmission rate. For the multivariate analysis, similar results were obtained when performing sensitivity analysis with the removal of highly significant variables from models as well as the removal of variables with >10% missing data. Reasons for return to the operating room were often missing (n=68), but common explanations included hematoma evacuation (n=6), cerebrospinal fluid (CSF) leak (n=4), resection of additional tumor (n=8) and other unrelated procedures (n=5). Similarly, reasons for 30-day readmission were commonly absent (n=50) or unknown (n=54).

Study findings
The results of this study suggest that multiple factors have a significant influence on complication rates or 30-day readmission after the resection of pituitary adenomas. Among all of the evaluated factors, preoperative hypertension and elevated WBC count were significant predictors of perioperative complications. These are notable as they may be potentially modifiable risk factors. Furthermore, preoperative BUN levels, perioperative complications, and return to the operating room were significant predictors of 30-day readmission. Interestingly, operative time and ASA classification were important factors for complications and 30-day readmission but were not important in the multivariate analysis.
Surprisingly, complications and 30-day readmission rates have remained stable over time rather than declined despite improved surgical techniques and standardized treatment strategies.

Clinical factors
These results suggest that some modifiable factors are associated with patient outcomes (e.g., hypertension, WBC, BUN), but it is unclear how better management can improve patient outcomes. Preoperative serological biomarkers, namely WBC count and BUN levels, were the most promising for predicting outcome and could be objective metrics of underlying patient health. WBC count, and specifically neutrophil-tolymphocyte ratio, has been shown to be an important factor in predicting outcomes for a variety of diseases, including head trauma [11], subarachnoid hemorrhage [12,13], and tumors [14,15]. Clinically, high levels of BUN preoperatively could be an indication of kidney dysfunction and be a major contributing factor to the development of postoperative urinary tract infections, which was one of the most common complications in our cohort. High levels of BUN have also been shown to be a risk factor for the development of pneumonia after surgery [16]. Similarly, return to the operating room played an important role in 30-day readmission (OR=4.1), but it is unclear how preventable this is. While some return to the operating room may be unavoidable, the clear association of repeat early surgery and poor outcomes should be notable because it highlights the importance of the surgeon and treatment team in patient care. The lack of improvement in complications and 30-day readmission also suggests additional room for patient outcome improvement.

Patient complication after pituitary adenoma surgery
Multiple single-center database studies and meta-analysis reviews have evaluated outcomes after pituitary adenoma surgery. Agam et al. [6] evaluated 1,153 consecutive patients undergoing trans-sphenoidal microscopic and endoscopic approaches for pituitary adenoma resection. The authors noted a median hospital stay of three days, perioperative death rate of 0.1%, an overall complication rate of 17.0%, and surgical complication rate of 6.8%. The most common surgical and medical complications were postoperative CSF leak (2.6%) and bacteremia/sepsis (0.5%), respectively. Microscopic and endoscopic approaches showed no differences in surgical complications (6.4% vs. 8.8%, p=0.2) or endocrinological complications (11.4% vs. 11.8%, p=0.9). Risk factors for complications were prior transsphenoidal surgery, preoperative vision loss, and the presence of invasion on MRI. These results reflect the low overall complication rate of patients undergoing pituitary tumor surgery and delineate between different surgical approaches. The 30-day readmission rate was not addressed, but high-risk features on imaging during pituitary adenoma resection were notable. Limitations of these data include the retrospective nature of the data analysis and limited external validity due to the study predominantly including data from a single surgeon.
Another multicenter study evaluated 982 patients undergoing endoscopic pituitary surgery at six international centers between 2002 and 2014 [17]. This study demonstrated a median hospital stay of five days and an overall adverse event rate of 23.8%. Risk factors predicting complications included intraventricular tumor extension and previous radiation. Reoperation occurred in 6.5% of patients, with risk factors including intraventricular extension and younger age. CSF leak risk was associated with female sex, high body mass index (BMI), lower age, and intraventricular extension. These results are interesting in supporting imaging findings (i.e., intraventricular extension) that correlate with risk. The strengths of this study include its more homogeneous sample of endoscopic approaches and multiple centers; however surgical treatments were done at high-volume centers with experts in the field. Thus, the results of this study may not apply to surgeons at all centers.

Return to the operating room
We found that 91 patients (4.0%) had a return to the operating room after the initial surgery. CPT codes indicating the reason for reoperation were listed for 23 of the 91 patients (25.2%). Of those 23 patients, the most common reasons for reoperation were further tumor resection, CSF leak/dural repair, and hematoma. Further tumor resection was the primary reason for reoperation, which suggests a limitation of interpreting the NSQIP dataset. Although some of these patients may have been indicated for a return to the operating room to excise the residual tumor, others may have returned to the operating room for other complications or indications, with the original CPT or ICD-9 code acting as a basic proxy.
CSF leaks are among the most common complications of pituitary adenoma surgery [26]. Our data show that seven patients (30.4%) returned to the operating room because of a CSF leak or need for a dural repair as indicated by CPT codes (15570, 20937, 30520, 31287, 63707). CSF leaks can be variable, based on the surgical approach, patient factors (e.g., BMI) [22], and initial repair techniques, making them extremely difficult to predict. Strickland et al. [29] performed a meta-analysis of the literature from 1995 through 2016 evaluating CSF rhinorrhea in 1002 patients. These results showed that half of all patients (n=26) that developed a postoperative CSF leak did not have a noted intraoperative leak or empirical sellar repair.

Limitations
The main limitation of this study is the incomplete data from the NSQIP dataset that were used as the only source of patient information and outcomes for this study. Missing data points differed depending on the variable included in the database. In addition, long-term follow-up and more granular outcome data (e.g., tumor resection rate, quality-of-life measures) were not available. No cases were coded as CPT 62165 (neuroendoscopy, intracranial; with excision of pituitary tumor, transnasal or transsphenoidal approach), but numerous cases had CPT 15740 (flap; island pedicle) or 15750 (flap; neurovascular pedicle), which would only be associated with endoscopic approaches. While these approaches are fairly specific to pituitary adenoma approaches, they may not capture all the relevant patients. These data also suggest that endoscopic approaches were being performed in the NSQIP dataset but could not be distinguished from open approaches. Despite being a prospectively collected surgical database that has been broadly used in the surgical literature, further studies are needed to better account for patient heterogeneity, and potentially including imaging could aid in this problem.
NSQIP was designed to track patients for 30 days after the initial operation, which could lead to a lower number of complications than actually occurred because of a lack of long-term follow-up, with late complications being a known issue with pituitary surgery [30]. In addition, the importance of WBC, hypertension, and BUN is limited by the relatively low odds ratios for predicting outcomes. Despite available literature supporting the importance of these factors overall, more work is needed to evaluate a causal relationship with clinical factors and worsened risk with pituitary adenoma surgery. Lastly, the obvious risk factors of patient complication and return to the OR may also be challenging to modify.
Notwithstanding these limitations, this study helps to better understand the potential risk factors that can lead to a higher incidence of complications and readmission in patients undergoing surgical treatment of pituitary tumors regardless of the operative approach. Because of the nature of the NSQIP program and the dataset, there is less selection bias that can help generate more generalizable and valid results. With hundreds of hospitals around the country tracking and submitting results to this program for over a decade, this database and patient population give us the best overall picture of the comorbidities and preoperative risk factors most likely to be associated with negative outcomes in our own patients.

Conclusions
Our results suggest that increased preoperative WBC count, return to the operating room, and hypertension were associated with increased complication rate during pituitary adenoma resections. Furthermore, complications, elevated preoperative BUN, and return to the operating room were independent predictors of 30-day readmission. WBC and BUN are potential biomarkers of risk and are readily available clinically. All of these factors can be potentially modifiable. Further work can potentially refine these and other risk factors to generate appropriate predictive models and cutoff values. Over time, complication rates and 30-day readmission after the resection of pituitary tumors have remained stable. These findings are limited by the coding of the NSQIP and missing data. The strength of this data is in the use of a prospective, curated, multicenter dataset. Understanding the effect of comorbidities on potential complications and readmission rates is vital to patient care and can present new areas of prospective investigation. The challenges for optimizing these factors in patients remain an area of ongoing research.

Additional Information Disclosures
Human subjects: Consent was obtained or waived by all participants in this study. University of Utah IRB issued approval Not necessary. Institutional review board approval is not necessary because the data are deidentified upon entering into the registry. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: Michael Karsy declare(s) personal fees from Cyrus Surgical. James Evans declare(s) personal fees from Mizuho. Michael Karsy declare(s) personal fees from Thieme Medical Publishing. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.