Great emphasis is currently placed on developing. Return to system within 30 days of discharge following pediatric non-shunt surgery.

Transcription

1 J Neurosurg Pediatrics 14: , 2014 AANS, 2014 Return to system within 30 days of discharge following pediatric non-shunt surgery Clinical article Samir Sarda, B.S., 1 Markus Bookland, M.D., 1 Jason Chu, M.D., 2 Mohammadali M. Shoja, M.D., 3 Matthew P. Miller, M.B.A., 4 Stephen B. Reisner, 1 Philip H. Yun, M.D., 1 and Joshua J. Chern, M.D., Ph.D. 1,2 1 Pediatric Neurosurgery Associates, Children s Healthcare of Atlanta; 2 Department of Neurosurgery, Emory University, Atlanta; 3 Division of Neurosurgery, Department of Surgery, University of Alabama at Birmingham, Alabama; and 4 Department of Information Systems and Technology, Children s Healthcare of Atlanta, Georgia Object. Hospital readmission after discharge is a commonly used quality measure. In a previous study, the authors had documented the rate of readmission and reoperation after pediatric CSF shunt surgery. This study documents the rate of readmission and reoperation after pediatric neurosurgical procedures excluding those related to CSF shunts. Methods. Between May 1, 2009, and April 30, 2013, 3098 non-shunt surgeries during 2924 index admissions were performed at a single institution. Demographic, socioeconomic, and clinical characteristics were prospectively collected in the administrative, business, and clinical databases. Clinical events within the 30 days following discharge were reviewed and analyzed. The following events of interest were analyzed for risk factor associations using multivariate logistic regression: return to the emergency department (ED), all-cause readmission, readmission to the neurosurgical service, and reoperation. Results. The number of all-cause readmissions within 30 days of discharge was 304 (10.4%, 304/2924). Admission sources consisted of the ED (n = 173), hospital transfers (n = 47), and others (n = 84). One hundred eighty of the 304 readmissions were associated with an operation, but only 153 were performed by the neurosurgical service (reoperation rate = 5.2%). These procedures included wound revisions (n = 30) and first-time shunt insertions (n = 35). The remaining 124 readmissions were nonsurgical, and only 54 were admitted to the neurosurgical service for issues related to the index non-shunt surgery. Thus, the rate of related readmission was 7.1% ([ ]/2924). A longer length of stay and admission to the neonatal intensive care unit during the index admission were associated with an increased likelihood of return to the ED and readmission. Certain procedures, such as baclofen pump insertion and intracranial pressure monitor placement, were also found to be associated with adverse clinical events in the 30-day period. Lastly, patients were more likely to a undergo reoperation if the index procedure had started after 3 p.m. Conclusions. The all-cause readmission rate within 30 days of discharge after a pediatric neurosurgical procedure was 10.4%, and the rate of related readmission was 7.1%. Whether these readmissions are preventable and to what extent they are preventable requires further study. ( Key Words patient readmissions reoperation quality assurance hydrocephalus Great emphasis is currently placed on developing quality measures to improve patient care, and studies examining readmission and reoperation rates following discharge are multiplying. 5,7,9,10 In the pediatric population, neurosurgery is considered to have one of the Abbreviations used in this paper: ACGME = Accreditation Council for Graduate Medical Education; CPT = current procedural terminology; ED = emergency department; ICP = intracranial pressure; ICU = intensive care unit; LOS = length of stay; NICU = neonatal ICU; PICU = pediatric ICU; VNS = vagus nerve stimulator. highest rates of readmission and reoperation. 3,8,11,15 This high rate is at least partially related to CSF shunt procedures, which tend to occur in medically complex patients and are almost always the most frequently performed procedures in a pediatric neurosurgery practice. Efforts to identify actionable areas of quality improvement regarding patients after shunt surgery are, therefore, a critical area of investigation, which the authors have previously addressed in a retrospective review of 1755 cases. 3 However, even in the most shunt-intensive pediatric neurosurgery practices, focusing entirely on postoperative shunt patients ignores 654 J Neurosurg: Pediatrics / Volume 14 / December 2014

2 Return to system after non-shunt surgery a significant portion of pediatric neurosurgery cases. Potentially modifiable risk factors for 30-day readmission among the non-shunt operations would be expected to be divergent from those of patients undergoing shunt surgery, and they bear independent analysis. In this study, we review the outcomes of non-shunt neurosurgical procedures performed within a 4-year time period. The outcome measures used are emergency department (ED) visit, readmission, and reoperation within 30 days of discharge. Clinical events that ensued following a non-shunt surgery are documented. An analysis of the clinical, demographic, and socioeconomic factors associated with various clinical end points follows. Significant efforts were spent categorizing surgical procedures in clinically meaningful ways because we hypothesized that procedure types are strong prognosticators of clinical outcome. In the discussion, a comparison between findings in the shunt and non-shunt surgeries is provided. Methods Inclusion and Exclusion Criteria During 2924 index admissions between May 1, 2009, and April 30, 2013, 3098 non-shunt neurosurgical procedures were performed at a single institution (Children s Healthcare of Atlanta) by board-certified and board-eligible pediatric neurosurgeons and fellows. There were 155 index admissions that contained 2 or more neurosurgical procedures. The first neurosurgical procedure within the index admission was labeled as the index procedure. Index admissions included in this study were defined as those admissions involving a non-shunt neurosurgical procedure. Patients with existing shunts but who were without definitive signs of shunt failure at the time of the non-shunt index procedure were included. An admission was excluded from the study if: 1) the index procedure was associated with current procedural terminology (CPT) codes referring to a shunt (62190 to 62258); or 2) the index procedure was preceded by a shunt operation (identified with the CPT code) within 30 days prior to the index procedure; or 3) the index surgery was not shunt related but was later followed by a shunt insertion or revision during the same admission. Data Fields: Categorizing Neurosurgical Procedures All 3098 non-shunt neurosurgical procedures were categorized and subcategorized based on review of operative reports and electronic medical records, as shown in Table 1. The categories were adapted from the Accreditation Council for Graduate Medical Education (ACGME) Neurological Surgery defined case categories (CPT code mapping areas) with modifications. For instance, ventricular access device placement falls under minor procedure in the ACGME case categories, but is labeled as a standalone category in this study because of the frequency of this operation in the pediatric population. If a patient underwent 2 separate procedures in a single operation, the procedure of higher acuity, adjudicated by the senior author (J.J.C.), was designated to be the representative procedure and categorized as such. For example, J Neurosurg: Pediatrics / Volume 14 / December 2014 TABLE 1: Index non-shunt procedures list by surgical category (n = 3098) Non-Shunt Surgical Category Value (%)* total no. of procedures 3098 cranial procedures 1739 (56.1) craniotomy for tumor 347 (11.2) trauma craniotomy & craniectomy 270 (8.7) depressed skull fracture 91 subdural hematoma evacuation 62 epidural hematoma evacuation 59 craniotomy (craniosynostosis) 267 (8.6) suboccipital craniectomy for Chiari malformation 226 (7.3) craniotomy (seizure) 119 (3.8) placement of seizure grid 46 cranioplasty 105 (3.4) craniotomy (vascular) 29 (0.9) craniotomy (transsphenoidal) 8 (0.3) craniotomy (other) 25 (0.8) endoscopic procedures, including third ven- 87 (2.8) triculostomy stereotactic brain biopsy 34 (1.1) ventricular access device & external ventricular 222 (7.2) drain insertion spinal procedures 484 (15.6) laminectomy 404 (13.0) occult spinal dysraphism 222 cervical laminectomy for stenosis 39 spinal cord/spinal column neoplasm 32 spinal fusion 80 (2.6) others 875 (28.2) wound revision/washout 223 (7.2) minor procedures 207 (6.7) ICP monitor placement 90 application & removal of halo 38 lumbar puncture 21 placement of epidural lumbar drain 20 VNS insertion/revision 135/22, 157 (5.1) baclofen pump insertion/revision 114/22, 146 (4.7) brachial plexus/peripheral nerve 9 (0.3) spina bifida 58 (1.9) closure myelomeningocele 40 repair encephalocele 18 other procedures 75 (2.4) * Subcategories may not summate to total frequency, as they only describe the most significant contributors to each surgical category. if a patient underwent external ventricular drain placement and craniotomy for tumor resection during a single visit to the operating room, the craniotomy for tumor resection was categorized as the representative procedure. However, if the external ventricular drain and tumor resection were performed in 2 separate visits to the operating room, these 655

3 S. Sarda et al. were then counted as 2 separate procedures and given equal weight in the statistical analysis. Other Data Fields Demographic, socioeconomic, and clinical characteristics were prospectively collected electronically with the following software: EPIC ASAP (emergency department), OpTime (surgical department), EpicCare Inpatient (inpatient service), EpicCare Ambulatory (outpatient clinic visits), and ADT (admission-discharge-transfer application). Data from the individual databases were then exported into a single operational database on an Oracle platform (Oracle 11g), available for queries by physicians and administrators. This database was not restricted to neurosurgical patients and included all visits to the system. The Information Technology Department of the hospital system was responsible for the maintenance and verification of the information accuracy. With few exceptions, when individual charts were reviewed for this study, the accuracy of the data was confirmed. Demographic data fields that were collected included age, primary language, race/ethnicity, zip code, and primary payer status (public assistance and self-pay vs private insurance). Data fields that were related to the surgeries included surgeon identity, hospital length of stay (LOS), procedure length, starting time of the procedure, admission source (elective admission vs admission from ED), CPT code and surgeon-dictated procedure comments (a short description entered by operating room circulating nurses), and operative reports. If surgery began after 3 p.m. but before 7 a.m., it was labeled as an after-hours surgery. Admissions to pediatric and neonatal intensive care units (PICUs and NICUs, respectively) were also captured. Statistical Methods Multivariate logistic regression analysis was used to assess whether risk factors independently correlated with dependent variables. The multivariate model was constructed based on the following independent variables: age (in months), self-pay or public assistance (yes/no), Caucasian (yes/no), English speaking (yes/no), address zip code belonging in the Atlanta metropolitan area (yes/ no), admission from ED (yes/no), ICU stay after surgery (yes/no), LOS (in days), procedure start time after 3 p.m. (yes/no), and surgery types (see below). Procedure length was not used as an independent variable to construct the model because it is strongly correlated with the type of surgery. Dichotomized dependent variables included ED returns (n = 353), all-cause readmissions (n = 304), neurosurgery readmissions (n = 207), and neurosurgery reoperations (n = 153). There were at least 10 outcomes for each independent variable eligible to be included in the logistic regression model. The association between surgeon identities and dependent variables was separately analyzed using univariate analysis and no association was found. Furthermore, univariate associations of surgery types (Table 1) with various dependent variables were analyzed, and those surgery types that were found to be significant (p < 0.05) were entered into the multivariate analysis. Results Categorizing Neurosurgical Procedures There were 3098 non-shunt surgeries in 2924 index admissions over the 4-year period. Descriptive statistics of the surgeries and patient cohort are provided in Tables 1 and 2, respectively. There were 155 index admissions during which more than 1 surgery was performed. The most common clinical scenarios were as follows: subdural grid placement followed by removal of grid and corticectomy (n = 46), placement of an external ventricular drain followed by tumor resection (n = 43), and trauma craniectomy followed by cranioplasty (n = 22). The 4 most commonly performed cranial procedures were craniotomy for tumor resection, craniotomy for traumatic injury, craniosynostosis repair, and Chiari malformation decompression. Together, they represented 1110 of 3098 cases. Craniotomy for epilepsy and cranioplasty procedures, which included patients with congenital skull defects and traumatic brain injuries, were the next two most commonly performed procedures (Table 1). Among the spinal procedures, treatment for occult spinal dysraphism, including filum lipoma, dermal sinus tract, dermoid tumor, diastematomyelia, and lipomyelomeningocele, was the most common. In the category of others, it was notable that the number of cases of wound revision and washout was significant (n = 223, 7.2% of all cases). Furthermore, the number of cases involving intracranial pressure (ICP) monitor placement (n = 90), vagus nerve stimulator placement (VNS, n = 157), and baclofen pump placement (n = 146) totaled 393 (12.7% of all cases). This last grouping deserved special attention because these patients were likely to be medically complex and the case volume was not insignificant. Sources of Readmissions Within 30 Days of Discharge A summary of readmission sources and events occurring after the readmission is shown in Fig. 1. The allcause readmission rate within 30 days of non-shunt surgery was 10.4% (304/2924). Within 30 days of discharge, there were 353 ED visits that resulted in 173 readmissions (49%). Forty-seven readmissions were transfers from outside hospitals. There were 1536 postoperative neurosurgery clinic visits within 30 days after discharge, leaving a 30-day return to clinic rate of 52.5%. Eleven direct admissions from the neurosurgical outpatient clinic were identified. All readmissions were designated to a hospital inpatient service. A related readmission was defined in this study as any readmission to the neurosurgical service. Surgery may or may not have been performed within the readmission. Children who were admitted to other services but later underwent neurosurgical procedures were classified under neurosurgical reoperation. As shown in Fig. 1 (gray boxes), the rate of related readmission within 30 days of discharge after non-shunt surgery was 7.1% (207/2924). Neurosurgical Reoperations One hundred eighty of the 304 all-cause readmissions 656 J Neurosurg: Pediatrics / Volume 14 / December 2014

4 Return to system after non-shunt surgery TABLE 2: Patient characteristics and other variables associated with index admissions for non-shunt surgery (n = 2924) Variable Value Range continuous variables mean age ± SD (yrs) 7.17 ± mean LOS ± SD (days) 7.38 ± mean procedure length ± SD (min) 116 ± mean household income per zip code ($) 49,657 13, ,674 categorical variables (%) admission from ED before index non-shunt surgery 635 (21.7) PICU admission 1355 (46.3) NICU admission 195 (6.7) English as primary language 2671 (91.3) Caucasian 1836 (62.8) Atlanta metropolitan zip codes 999 (34.2) surgery started after 3 p.m. & before 7 a.m. 386 (13.2) Medicaid/Medicare as the primary payee 1398 (47.8) LOS 3 days 1794 (61.4) were associated with 1 or more operations during the hospitalization, and 153 of those operations were performed by the neurosurgical service (Fig. 1). As such, the rate of neurosurgical operations within 30 days of discharge was 5.2% (153/2924). Of these 153 operations, there were 98 non-shunt procedures, which included wound revision (n = 30), craniotomies (n = 15), laminectomies (n = 12), insertions of external ventricular drains (n = 8), placement of lumbar drains (n = 6), and removal of intrathecal baclofen pumps (n = 5), among others. Notably, even though this study had excluded index shunt surgeries but did include patients with existing shunts, there were 20 Fig. 1. Thirty-day clinical events after 2924 index admissions for non-shunt surgeries over a 4-year period. The second row of boxes reflects only those events that occurred within 30 days. Gray boxes in the last row represent related readmissions. NSGY = neurosurgery. Return to OR = return to operating room. J Neurosurg: Pediatrics / Volume 14 / December

5 S. Sarda et al. shunt revisions within 30 days after the index non-shunt surgeries. These 153 neurosurgical reoperations were further divided into 2 groups: planned and unplanned operations. A clinical scenario from the planned group would be, for example, after a cranial encephalocele closure, a child is followed as an outpatient and brought in for an elective shunt insertion after identification of progressive hydrocephalus. There were 73 planned readmissions for neurosurgical procedures, including shunt insertions/revisions (n = 23), wound revisions (n = 11), and craniotomies (n = 8). In the unplanned reoperation group (n = 80), the most common operation was urgent or emergency shunt insertion in patients whose index surgeries were either ventricular access device insertions or myelomeningocele closures (n = 18). Additionally, 6 shunt revisions were performed in patients with previously existing shunts. Other reoperations included 18 wound revisions and 6 baclofen pump revisions. In the risk factor analysis provided below, special attention was paid to this patient group who underwent unplanned reoperations. Risk Factors Associated With Patient Subgroups Multivariate analysis was used to identify clinical and demographic factors that were predictive of clinical events after non-shunt surgeries. The following 5 clinical events were designated as the dependent variables: return to ED, all-cause readmissions, all neurosurgical readmissions (related readmissions), neurosurgical reoperations, and unplanned neurosurgical reoperations. The result of the multivariate analysis is summarized in Table 3. With return to the ED as a dependent variable (n = 353), multivariate logistic regression identified LOS to be a statistically significant risk factor. Stereotactic brain biopsies and insertion of baclofen pumps and vagus nerve stimulators were 3 types of procedures also identified as significant risk factors for returning to the ED. Patient payee status was not a significant risk factor. For both all-cause and neurosurgical readmissions, NICU stay during index admission and undergoing a minor procedure during index admission were significant risk factors. In addition, a longer LOS correlated with an increased likelihood of being readmitted, though not necessarily to the neurosurgical service (Table 3). In the group of patients who underwent neurosurgical procedures (planned and unplanned) after readmission, NICU patients and patients who underwent minor procedures were at risk. Interestingly, if the index procedure was started after 3 p.m., this significantly predisposed patients to subsequent operations. In the analysis of the group of patients who experienced unplanned visits to the operating room, procedure start time after 3 p.m. was identified as the only statistically significant risk factor by multivariate regression analysis. Discussion Rationale for the Study A descriptive investigation of clinical events that occur within 30 days of discharge following non-shunt surgery is provided in this study. It first establishes the percentage of readmissions associated with patient care provided by the neurosurgical service. Without an accurate depiction, the neurosurgeon may be held accountable for readmissions that are beyond his or her influence. 13,15 Second, the study highlights whether potentially modifiable risk factors are present. If they were identified, protocols targeting these areas may be implemented as quality improvement initiatives. In designing this study, the first and foremost consideration was to separate shunt and non-shunt surgeries. Previously we had documented the strong relationship between shunt surgery itself and early patient readmissions. 3,14 This correlation, while vital to a complete apprehension of factors driving early readmission after all TABLE 3: Statistically significant risk factors from multivariate analysis associated with various clinical end points after index non-shunt surgery* Dependent Variables No. of Patients Identified Risk Factors OR (95% CI) p Value return to the ED 353 LOS (days) baclofen pump insertion VNS insertion stereotactic brain biopsy all-cause readmissions 304 LOS (days) NICU admission minor procedures ETV neurosurgical readmissions 207 NICU admission minor procedures neurosurgical reoperations 153 procedure start after 3 p.m. NICU admission minor procedures 1.01 ( ) 2.20 ( ) 2.13 ( ) 3.06 ( ) 1.01 ( ) 3.54 ( ) 2.86 ( ) 2.28 ( ) 3.61 ( ) 2.32 ( ) 1.77 ( ) 5.39 ( ) 2.87 ( ) unplanned neurosurgical reoperations 80 procedure start after 3 p.m ( ) * CI = confidence interval; ETV = endoscopic third ventriculostomy; OR = odds ratio. 658 J Neurosurg: Pediatrics / Volume 14 / December 2014

6 Return to system after non-shunt surgery pediatric neurosurgical procedures, is a powerful confounder in statistical models. Carefully teasing the impact of shunt surgery apart from non-shunt neurosurgical procedures allows for a clearer analysis of clinical factors that are not shunt related. A second consideration was to include patients with shunts whose index procedures were not shunt related. These were patients with shunts who were without definitive signs of shunt failure at the time of a non-shunt index procedure. Ideally, one may wish that these patients should be excluded. However, to do so, one would have to rely on problem and diagnosis lists in the administrative database to identify these patients. Knowing the inaccuracy of the recorded ICD-9 codes in the administrative database based on our previous review of medical records, we did not feel confident that these patients could be accurately identified. Therefore it is out of practical consideration that these patients were included. At-Risk Patient Groups After baclofen pump insertion and revision, VNS insertion and revision, and stereotactic brain biopsy, there is an elevated risk of return to the ED within 30 days of the procedure. Importantly, the same procedure categories were not found to be at increased risk for readmissions or reoperations (Table 3). As these patients do not typically warrant readmission to the hospital, their ED complaints can be assumed to be very subtle or too mild to rise to the level of inpatient care. This suggests an ED utilization pattern that may be independent of medical necessity and may be amendable to modifications via patient education or other social support targeted programs. One may speculate that the high anxiety and medical complexity of these patients may have attributed to this observation. Neonatal ICU stay during the index admission was identified as a significant risk factor in all-cause readmissions, neurosurgery readmissions, and all neurosurgery reoperations (planned and unplanned; Table 3). The unique challenges and needs of these premature infants have been well documented. NICU patients suffer from higher rates of infection and ventilator dependence than their non-ni- CU counterparts, factors that complicate postsurgical care and predispose patients to early returns to the hospital. 1,4 In other words, reducing readmission and reoperation rates among neurosurgical non-shunt NICU patients may be impossible without general advances in neonatal care or, at the very least, will require multidisciplinary care to ensure that all postoperative and nonoperative medical issues are thoroughly resolved prior to discharge. With the same analytical results but for a different reason, patients who underwent minor procedures, such as ICP monitor placement, lumbar puncture, and lumbar drain placement, were also found to be at risk for readmissions and reoperations. Despite the label of minor procedure, these patients were selected for invasive diagnostic testing to resolve diagnostic dilemmas. This raised the pretest probability of the target adverse events (readmission and reoperation) and, in this manner, influenced the observed posttest probability. We found that minor procedures are 2.8 times more likely to be followed by J Neurosurg: Pediatrics / Volume 14 / December 2014 another neurosurgical procedure (both planned and unplanned) within 30 days compared with other procedure types. However, minor procedures are not more likely to be followed by an unplanned neurosurgical procedure (Table 3). This suggests that the operations after minor index procedures were likely to be staged or planned. Comparing Results of Shunt and Non-Shunt Surgeries In Table 4, selected clinical and socioeconomic factors are listed to highlight the differences between patients who underwent shunt and non-shunt surgeries. Compared with non-shunt patients, patients with shunts had a shorter LOS and were more likely to be admitted from the ED, less likely to be admitted to ICU afterward, and more likely to have Medicaid listed as payee. This last characteristic was unexpected and had not been previously reported, to the best of our knowledge. As predicted, the rates of reoperation and readmission were higher following shunt surgeries than non-shunt surgeries. Perhaps the most interesting result from the statistical analysis was the finding that with unplanned neurosurgical reoperations as the dependent variable, having a procedure start after 3 p.m. was a significant risk factor (Table 3). This finding matches the finding from our previously conducted shunt readmission study, 3 thereby confirming that after-hours surgery is a promising factor for future study (Table 4). It is crucial to point out that procedure start time is only a surrogate marker (similar to hospital LOS). It is a variable that can be confounded by status of patient condition at presentation, care delivery in the ED, reduced staff in the operating room, and surgeon fatigue, just to name a few. The measurements of these factors were imperfect and nearly impossible at times, and their modifiability varies. There might even be more unknown unmeasured factors. While a multivariate analysis is able to eliminate confounding effects of multiple risk factors, it can only adjust for measured variables. Presenting to the hospital in a poor clinical state or having emergency cases that often bring in patients with severe conditions or trauma after hours is likely not a modifiable risk factor to be considered in preventability studies. Studies are ongoing at our institution to further delineate the causes of this association. Limitations of the Study The categorization of procedures continues to represent a challenge in this study and others. 2,4,6,12 To a certain degree, it is subjective in nature, and in this study it was determined by a single surgeon. Even though the ACGME coding practice was followed, modifications were made to reflect the needs of pediatric patients. Variation in the accuracy and reliability of the assignment can potentially alter the analysis results. The decision to include patients with shunts whose index procedures were not shunt-related could be controversial. As previously discussed, these patients were included because of practical reasons. Lastly, even though this study includes a large number of surgeries, they originate from a single institution, therefore the results are limited to a small collection of practicing physicians in one locality. 659

7 S. Sarda et al. TABLE 4: Comparison of results from index shunt and non-shunt surgeries over a 4-year period Variable Shunt Surgeries* Non-Shunt Surgeries no. of surgeries clinical & other factors associated w/ index admission (%) admission from ED before index surgery 961 (54.8) 635 (21.7) ICU admission after surgery 354 (20.2) 1355 (46.3) Medicaid/Medicare as primary payee 1010 (57.5) 1398 (47.8) surgery started after 3 p.m. & before 7 a.m. 512 (29.2) 386 (13.2) LOS 3 days 702 (40) 1794 (61.4) mean procedure length ± SD (min) 42.0 ± ± 107 rate of readmission & reoperations (%) all-cause readmissions 290 (16.5) 304 (10.4) neurosurgical readmissions 216 (12.3) 207 (7.1) all operations following readmission 184 (10.5) 180 (6.2) neurosurgical operations following readmission 165 (9.4) 153 (5.2) identified risk factors return to ED Atlanta zip codes OR 1.39, 95% CI (p = 0.008) Medicaid payee OR 1.31, 95% CI (p = 0.037) LOS (days) OR 1.01, 95% CI (p = 0.032) baclofen pump insertion OR 2.20, 95% CI (p = 0.017) VNS Insertion OR 2.13, 95% CI (p = 0.028) stereotactic brain biopsy OR 3.06, 95% CI (p = 0.019) neurosurgical reoperations (planned & unplanned) procedure start after 3 p.m. OR 1.77, 95% CI (p = 0.016) NICU admission OR 5.39, 95% CI (p = 0.002) minor procedures OR 2.87, 95% CI (p = 0.005) neurosurgical unplanned reoperation procedure start after 3 p.m. OR 1.45, 95% CI (p = 0.044) OR 2.10, 95% CI (p = 0.019) * From Chern et al., All of the shunt reoperations were unplanned. Conclusions The all-cause readmission rate within 30 days of discharge after non-shunt surgery was 10.4%, and neurosurgery readmission rate was 7.1%. The rate of neurosurgery reoperation within 30-day was 5.2%. Undergoing surgery starting after 3 p.m. was a significant and potentially modifiable risk factor. Selecting risk factors that are modifiable and devising intervention protocols will be the logical next step. Disclosure The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Author contributions to the study and manuscript preparation include the following. Conception and design: Chern. Acquisition of data: Sarda, Reisner, Yun. Analysis and interpretation of data: Chern, Sarda, Shoja. Drafting the article: Bookland, Chu. Statistical analysis: Shoja. Administrative/technical/material support: Miller. References 1. Ambalavanan N, Carlo WA, McDonald SA, Yao Q, Das A, Higgins RD: Identification of extremely premature infants at high risk of rehospitalization. Pediatrics 128:e1216 e1225, Baaj AA, Uribe JS, Nichols TA, Theodore N, Crawford NR, Sonntag VKH, et al: Health care burden of cervical spine fractures in the United States: analysis of a nationwide database over a 10-year period. Clinical article. J Neurosurg Spine 13:61 66, Chern JJ, Bookland M, Tejedor-Sojo J, Riley J, Shoja MM, Tubbs RS, et al: Return to system within 30 days of discharge following pediatric shunt surgery. Clinical article. J Neurosurg Pediatr 13: , Drake JM, Riva-Cambrin J, Jea A, Auguste K, Tamber M, Lamberti-Pasculli M: Prospective surveillance of complications in a pediatric neurosurgery unit. Clinical article. J Neurosurg Pediatr 5: , Feudtner C, Pati S, Goodman DM, Kahn MG, Sharma V, Hutto JH, et al: State-level child health system performance and the likelihood of readmission to children s hospitals. J Pediatr 157: e1, Fokkema M, Hurks R, Curran T, Bensley RP, Hamdan AD, Wyers MC, et al: The impact of the present on admission indicator on the accuracy of administrative data for carotid endarterectomy and stenting. J Vasc Surg 59:32 38.e1, Groman RF, Rubin KY: Neurosurgical practice and health 660 J Neurosurg: Pediatrics / Volume 14 / December 2014

8 Return to system after non-shunt surgery care reform: moving toward quality-based health care delivery. Neurosurg Focus 34(1):E1, Hain PD, Gay JC, Berutti TW, Whitney GM, Wang W, Saville BR: Preventability of early readmissions at a children s hospital. Pediatrics 131:e171 e181, Iannuzzi JC, Chandra A, Kelly KN, Rickles AS, Monson JR, Fleming FJ: Risk score for unplanned vascular readmissions. J Vasc Surg 59: e1, Joynt KE, Jha AK: Thirty-day readmissions truth and consequences. N Engl J Med 366: , Raval MV, Dillon PW, Bruny JL, Ko CY, Hall BL, Moss RL, et al: Pediatric American College of Surgeons National Surgical Quality Improvement Program: feasibility of a novel, prospective assessment of surgical outcomes. J Pediatr Surg 46: , Shah MN, Stoev IT, Sanford DE, Gao F, Santiago P, Jaques DP, et al: Are readmission rates on a neurosurgical service indicators of quality of care? Clinical article. J Neurosurg 119: , Tsai TC, Joynt KE, Orav EJ, Gawande AA, Jha AK: Variation in surgical-readmission rates and quality of hospital care. N Engl J Med 369: , Wong JM, Ziewacz JE, Ho AL, Panchmatia JR, Bader AM, Garton HJ, et al: Patterns in neurosurgical adverse events: cerebrospinal fluid shunt surgery. Neurosurg Focus 33(5):E13, Wrubel DM, Riemenschneider KJ, Braender C, Miller BA, Hirsh DA, Reisner A, et al: Return to system within 30 days of pediatric neurosurgery. Clinical article. J Neurosurg Pediatr 13: , 2014 Manuscript submitted March 3, Accepted August 29, Please include this information when citing this paper: published online October 17, 2014; DOI: / PEDS Address correspondence to: Joshua J. Chern, M.D., Ph.D., 5455 Meridian Mark Rd. NE, Ste. 540, Atlanta, GA joshua. chern@choa.org. J Neurosurg: Pediatrics / Volume 14 / December