J Neurosurg / Volume 120 / May 2014

J Neurosurg 120:1201–1211, 2014

1201

©AANS, 2014

Hospital readmission within 30 days of discharge is a major contributor to the high cost of health care in the US. Medicare payments for unplanned

30-day readmission episodes were responsible for $17.4 billion or roughly 17% of the total Medicare hospital pay-

ments for 2004.10,14 As a result, 30-day readmissions have become an important metric for measuring the quality of patient care. The Patient Protection and Affordable Care Act of 2010 authorized Medicare to use financial pen-alties to incentivize hospitals to reduce 30-day readmis-sions. Institutions have begun implementing programs to

Incidence and predictors of 30-day readmission for patients discharged home after craniotomy for malignant supratentorial tumors in California (1995–2010)

Clinical articleLogan P. Marcus, M.s.,1 Brandon a. Mccutcheon, M.P.P.,1 aBrahaM noorBakhsh, B.s.,1 raLitza P. Parina, M.P.h.,1 david d. gonda, M.d.,2 cLark chen, M.d., Ph.d.,2 david c. chang, Ph.d., M.P.h., M.B.a.,1 and BoB s. carter, M.d., Ph.d.2

1Department of Surgery and 2Division of Neurosurgery, University of California, San Diego, California

Object. Hospital readmission within 30 days of discharge is a major contributor to the high cost of health care in the US and is also a major indicator of patient care quality. The purpose of this study was to investigate the incidence, causes, and predictors of 30-day readmission following craniotomy for malignant supratentorial tumor resection.

Methods. The longitudinal California Office of Statewide Health Planning & Development inpatient-discharge administrative database is a data set that consists of 100% of all inpatient hospitalizations within the state of Califor-nia and allows each patient to be followed throughout multiple inpatient hospital stays, across multiple institutions, and over multiple years (from 1995 to 2010). This database was used to identify patients who underwent a craniotomy for resection of primary malignant brain tumors. Causes for unplanned 30-day readmission were identified by prin-ciple ICD-9 diagnosis code and multivariate analysis was used to determine the independent effect of various patient factors on 30-day readmissions.

Results. A total of 18,506 patients received a craniotomy for the treatment of primary malignant brain tumors within the state of California between 1995 and 2010. Four hundred ten patients (2.2%) died during the index surgical admission, 13,586 patients (73.4%) were discharged home, and 4510 patients (24.4%) were transferred to another facility. Among patients discharged home, 1790 patients (13.2%) were readmitted at least once within 30 days of discharge, with 27% of readmissions occurring at a different hospital than the initial surgical institution. The most common reasons for readmission were new onset seizure and convulsive disorder (20.9%), surgical infection of the CNS (14.5%), and new onset of a motor deficit (12.8%). Medi-Cal beneficiaries were at increased odds for readmis-sion relative to privately insured patients (OR 1.52, 95% CI 1.20–1.93). Patients with a history of prior myocardial infarction were at an increased risk of readmission (OR 1.64, 95% CI 1.06–2.54) as were patients who developed hydrocephalus (OR 1.58, 95% CI 1.20–2.07) or venous complications during index surgical admission (OR 3.88, 95% CI 1.84–8.18).

Conclusions. Using administrative data, this study demonstrates a baseline glioma surgery 30-day readmission rate of 13.2% in California for patients who are initially discharged home. This paper highlights the medical histories, perioperative complications, and patient demographic groups that are at an increased risk for readmission within 30 days of home discharge. An analysis of conditions present on readmission that were not present at the index surgical admission, such as infection and seizures, suggests that some readmissions may be preventable. Discharge planning strategies aimed at reducing readmission rates in neurosurgical practice should focus on patient groups at high risk for readmission and comprehensive discharge planning protocols should be implemented to specifically target the mitigation of potentially preventable conditions that are highly associated with readmission.(http://thejns.org/doi/abs/10.3171/2014.1.JNS131264)

key Words      •      brain tumor      •      readmission      •      neurosurgical outcomes

1201

Abbreviations used in this paper: DVT = deep venous throm-bosis; LOS = length of stay; OSHPD = Office of Statewide Health Planning & Development; PE = pulmonary embolism.

This article contains some figures that are displayed in color on line but in black-and-white in the print edition.

L. P. Marcus et al.

1202 J Neurosurg / Volume 120 / May 2014

reduce hospital readmissions in both medical and surgical specialties. Despite coordinated efforts from physicians, nurses, pharmacists, and lower-level care providers, many hospitals have struggled to understand and improve upon the factors that underlie high readmission rates.15

Few studies have examined the rehospitalization of patients after the neurosurgical care of brain malignancy. The purpose of this study was to determine the 30-day readmission rate for patients undergoing the resection of primary brain tumors and to identify which factors pre-dispose certain patient groups to rehospitalization.

MethodsData Source

The data source for this study was the California Of-fice of Statewide Health Planning & Development (OSH-PD) longitudinal inpatient-discharge administrative data-base for the years 1995 to 2010, obtained from the State of California OSHPD (http://www.oshpd.ca.gov/HID/ Products/PatDischargeData/PublicDataSet/). The Cali-fornia inpatient discharge database is an administrative, longitudinal database that represents a 100% sample of all inpatient discharges from California licensed hospi-tals. Each patient within the database is given a unique, masked patient identifier so that each patient may be fol-lowed throughout multiple inpatient hospital stays, across multiple institutions, and over multiple years within the state of California during the study period.

Inclusion and Exclusion Criteria and Definition of End Points

An index admission for surgical treatment of a pri-mary brain tumor was defined using a combination of ICD-9-CM diagnosis and procedure codes. Patients were included in this study if they were given both a diagnosis of supratentorial malignant brain tumor (191.0–191.5, 191.8, 198.9) while also receiving a procedure code for lobectomy (01.53), excision or destruction of tissue or lesion of brain (01.59), or open brain biopsy (01.14) during the same hos-pital stay.2,4 Patients who received a previous craniotomy for any diagnosis prior to their index craniotomy for supra-tentorial malignant brain tumor were excluded from this study. The primary end point examined in this study was an unplanned hospital readmission less than or equal to 30 days after inpatient discharge for surgical treatment of a primary brain tumor. Within the OSHPD longitudinal in-patient-discharge administrative database, unplanned hos-pital readmissions are defined as inpatient stays that are un-scheduled at the hospital 24 hours prior to patient admission (http://www.oshpd.ca.gov/HID/Products/PatDischarge Data/PublicDataSet/). Planned rehospitalizations were ex cluded from our analysis of readmission. Patients who were discharged to a location other than home after sur-gery or readmitted to the hospital from a location other than home were not included in our readmission analysis so as to eliminate the effect of inter- and/or intrahospital transfer on 30-day readmissions. However, all patients un-dergoing an index admission regardless of discharge loca-tion were included in our analysis of index surgical episode outcomes (Table 1).

Patient CharacteristicsPatient age, self-declared race/ethnicity, sex, expected

primary payer (Medicare, Medi-Cal, private insurance, or uninsured), admission source (home, residential care facil-ity, jail/prison, and others), type of admission (elective or nonelective), discharge disposition (home, died, residential care facility, jail/prison, and others), and calendar year were coded in the California inpatient discharge database. Due to a small sample size of certain ethnic groups within the database, patients of Native American race/ethnicity were combined with “other” race/ethnicity to avoid unstable co-efficients in the multivariate model. To assess the effect of general medical comorbidities, the Charlson comorbidity index was calculated for each patient using the method de-scribed by Romano and colleagues.17

The medical history for each patient was defined using ICD-9-CM diagnosis and procedure codes. A patient was determined to have a specific condition if he or she carried a specified ICD-9-CM code on the index surgical admis-sion or during any inpatient episode prior to the index sur-gical admission episode from 1995 to 2010 (Table 2).

Hospital CharacteristicsHospital identifier, county location of each hospital,

and total charges for each inpatient episode were coded within the California inpatient discharge database. The teaching status of each hospital was determined accord-ing to whether the hospital was affiliated with a neurosur-gical residency training program (http://www.societyns.org/match_information.html). Pediatric hospitals were defined as those hospitals solely dedicated to the treat-ment of pediatric patients.

Reasons for 30-Day ReadmissionTo identify the reasons for readmission, a cross-sec-

tional analysis of ICD-9-CM principal diagnosis codes present on readmission was conducted for all patients re-admitted within 30 days of discharge. Each ICD-9-CM principal diagnosis code with more than 1% prevalence in readmitted patients was placed into 1 of 15 disease catego-ries (Table 2). The ICD-9-CM codes comprising each of the 15 disease categories were used to tabulate the number of readmitted patients who belonged in each disease cat-egory. Because of the longitudinal nature of the data used in this study, we were able to determine which diagnoses were present on both discharge and readmission for each patient. We established the reason for each patient’s read-mission as a diagnosis that was not present on discharge but was present as a primary diagnosis on readmission.

Statistical AnalysisBivariate analysis of mean patient age, surgical length

of stay (LOS), Charlson score, and readmission status was performed using the Welch t-test. Bivariate analysis of pa-tient factors and readmission status was performed using the Pearson chi-square test. Multivariate analyses were performed using logistic regression models to determine the odds of hospital readmission within 30 days of surgi-cal discharge while adjusting for age, demographics, surgi-cal admission LOS, race/ethnicity, sex, expected primary

J Neurosurg / Volume 120 / May 2014

Incidence and predictors of 30-day readmission after craniotomy

1203

payer, hospital teaching status, California county, calendar year of surgery, medical history prior to surgical admis-sion, and complications arising during surgical admission.

Expected 30-day readmission rates were calculated for each patient using a multivariate logistic regression model for 30-day readmission. The expected likelihoods of 30-day readmission were then aggregated by hospital, and the ratio of observed to expected readmissions for each hospital was calculated. Statistical analysis was per-formed using commercially available software (STATA/AMP 10, Stata Corp. LP). All tests were 2-sided, and p values < 0.05 were considered statistically significant.

ResultsThere were 18,506 inpatient admissions for resection

of primary brain tumors in California between 1995 and 2010. Four hundred ten patients (2.2%) died during the index surgical admission, 13,586 patients (73.4%) were discharged home, and 4510 patients (24.4%) were trans-ferred to another facility (Table 1).

Among the home discharge patients, 1790 (13.2%) had at least 1 unplanned readmission within 30 days of discharge, and 483 (27%) of these readmitted patients were readmitted to a different hospital than the hospital from which they were discharged following the index sur-gical episode (Table 3). There were also 377 patients ex-cluded from our analysis who had a planned readmission within 30 days of discharge. A majority of these planned readmissions were for brain excision (28.6%), chemother-apy (22.0%), and radiotherapy (9.3%).

Index Admission EpisodesThe median age of patients undergoing tumor resec-

tion was 54 years. The majority of patients were privately insured non-Hispanic white patients who received neu-rosurgical care at nonacademic medical centers (Table 1). The median LOS for index surgical admission was 6 days, and the median total charges per index surgical ad-mission were $72,029 (Table 1). The most prevalent dis-eases in patients’ medical histories prior to surgery were hypertension (32.4%), tobacco use disorder (19.0%), and seizure and convulsive disorder (13.2%).

Thirty-Day Outcomes and 30-Day ReadmissionsThree hundred sixteen patients (2.3%) died within 30

days of home discharge (Table 3). A total of 1790 patients were rehospitalized within 30 days of discharge, and the median time to first readmission was 11 days. Although a majority of readmitted patients had only 1 readmission episode, 11.8% of readmitted patients had multiple read-

TABLE 1: Patient characteristics and surgical episode outcome among 18,506 patients undergoing craniotomy for primary malignant brain tumor in California (1995–2010)

Variable Value

mean age in yrs (median) 51.6 (54.0)race/ethnicity (%) n = 18,324 non-Hispanic white 13,546 (74.0) African American 626 (3.4) Hispanic 2772 (15.1) Asian 931 (5.1) Native American/other 449 (2.5)sex (%) n = 18,506 male 10,684 (57.7) female 7822 (42.3)medical history prior to surgical episode (%) n = 18,506 hypertension 5988 (32.4) tobacco use disorder 3513 (19.0) seizure & convulsive disorder 2449 (13.2) cerebrovascular disease 2160 (11.7) chronic pulmonary disease 1771 (9.6) diabetes mellitus 1672 (9.0) lipid disorder 1118 (8.0) obesity & overweight 1061 (5.7) speech & language disorder 893 (4.8)  motor deficit 857 (4.6) moderate or severe liver disease 779 (4.2) myocardial infarction 598 (3.2) congestive heart failure 474 (2.6) cerebral edema & compression of the brain 432 (2.3) hydrocephalus 375 (2.0) gait & coordination dysfunction 372 (2.0) peripheral vascular disease 220 (1.2) vision & optic disorder 186 (1.0) malaise & fatigue 171 (0.9) peptic ulcer disease 110 (0.6) renal disease 61 (0.3) mild liver disease 55 (0.3)mean Charlson comorbidity index (median) 2.7 (2.0)expected primary payer (%) n = 15,996 Medicare 3903 (24.4) Medi-Cal 1562 (9.8) private insurance 10,132 (63.3) uninsured 369 (2.3)hospital teaching status (%) n = 18,506 neurosurgical residency program 5881 (31.8) no neurosurgical residency program 12,625 (68.2)hospital pediatric status (%) n = 18,506 nonpediatric hospital 17,918 (96.8) pediatric hospital 588 (3.2)mean LOS on surgical admission in days (median) 8.2 (6.0)

(continued)

TABLE 1: Patient characteristics and surgical episode outcome among 18,506 patients undergoing craniotomy for primary malignant brain tumor in California (1995–2010) (continued)

Variable Value

mean charges for surgical admission in $ (median) 95,906 (72,029)surgical admission mortality (%) 410 (2.2)patients discharged home (%) 13,586 (73.4)

L. P. Marcus et al.

1204 J Neurosurg / Volume 120 / May 2014

missions and there were a total of 2022 30-day readmis-sion episodes. The median LOS during a readmission was 4 days, and the median hospital charges per 30-day read-mission episode were $20,296. The diagnoses that were most often established as reasons of readmission were seizure and convulsive disorder (20.9%), surgical infec-tion of the CNS (14.5%), and motor deficit (12.8%).

Differences Between Readmitted and Nonreadmitted Patients

Patients readmitted within 30 days had a 2-day lon-

ger median LOS on their index surgical episode than nonreadmitted patients. There were significant racial/eth-nic differences noted between the 2 groups. Readmitted patients were more frequently African American or His-panic than nonreadmitted patients. Patients with at least one 30-day hospital readmission were more often Medi-care and Medi-Cal beneficiaries and less often privately insured. Other observed differences between readmitted and nonreadmitted patients are noted in Table 4.

The results of the multivariate analyses are presented in Table 5. In the multivariate regression model, longer

TABLE 2: Diagnosis and procedure codes (ICD-9-CM) used to define patient comorbidities and complications

Category Code*

hypertension 401.0, 401.1, 401.9tobacco use disorder 305.1, V15.82seizure & convulsive disorder 345.10, 345.11, 345.2, 345.3, 345.40, 345.41, 345.50, 345.51, 345.70, 345.71, 345.8, 345.9, 780.39cerebrovascular disease 430, 431, 432.0, 432.1, 432.9, 433.00, 433.01, 433.10, 433.11, 433.20, 433.21, 433.30, 433.31, 433.80,

433.81, 433.90, 433.91, 434.00, 434.01, 434.10, 434.11, 434.90, 434.91, 435.0, 435.1, 435.2, 435.3, 435.8, 435.9, 436, 437.0, 437.1, 437.2, 437.3, 437.7, 437.8, 437.9, 438.0, 438.89, 438.9

chronic pulmonary disease †lipid disorder 272.0, 272.1, 272.2diabetes mellitus †obesity & overweight 278.00, 278.01, 278.02, V85.36, V85.37, V85.38, V85.39, V85.4speech & language disorder 784.3, 784.41, 784.42, 784.5, 784.51, 784.52, 784.59, 784.60, 784.61, 784.69, 937.5,‡ 937.2,‡ 438.10,

438.11, 438.12, 438.19moderate or severe liver disease †motor deficit 342.0, 342.1, 342.8, 342.9, 344.0, 344.1, 344.3, 344.4, 344.6, 344.81, 344.89, 344.9, 781.4, 781.94, 799.3,

438.20, 438.30, 438.40, 438.50, 438.82, 438.83myocardial infarction †cerebral edema & compression of brain 348.4, 348.5hydrocephalus 331.3, 331.4, 02.2,‡ 02.34‡congestive heart failure †gait & coordination dysfunction 781.2, 438.81, 438.84, 781.3, 93.22‡peripheral vascular disease †vision & optic disorder 368.46, 368.47, 377.49, 377.75, 378.81, 379.50, 379.56, 438.7malaise & fatigue 780.7, 780.71, 780.79peptic ulcer disease †renal disease †mild liver disease †nonspecific CNS surgical complication 996.2, 996.75, 997.00, 997.01, 997.09general infection 039.1, 039.8, 041.00, 041.01, 041.02, 041.03, 041.04, 041.09, 041.10, 041.11, 041.12, 041.19, 041.2, 041.3,

041.4, 041.5, 041.6, 041.7, 041.81, 041.82, 041.83, 041.84, 041.85, 041.86, 041.89, 041.9, 999.31, 999.39sepsis & septicemia 003.1, 038.0, 038.10, 038.11, 038.12, 038.19, 038.2, 038.3, 038.40, 038.41, 038.42, 038.43, 038.44, 038.49,

038.8, 038.9, 790.7, 995.91, 995.92urinary tract infection 590.10, 590.11, 590.80, 595.0, 595.3, 595.4, 599.0, 997.5DVT, PE, & venous complications 415.10, 415.11, 415.19, 416.2, 451.0, 451.11, 451.19, 451.2, 451.82, 451.83, 451.84, 451.89, 451.9, 453.1,

453.2, 453.40, 453.41, 453.42, 453.51, 453.52, 453.6, 453.80, 453.81, 453.82, 453.9surgical infection of the CNS 324.0, 326, 998.51, 998.59, 01.31,‡ 01.25,‡ 01.24,‡ 01.59,‡ 03.4,‡ 036.0, 047.9, 049.9, 320.0, 320.1, 320.2,

320.3, 320.7, 320.81, 320.82, 320.89, 320.9, 321.0, 321.1, 321.2, 322.0, 322.1, 322.2, 322.9, 323.01, 323.02, 323.9, 324.1, 324.9

* All codes are ICD-9-CM codes unless otherwise indicated.†  ICD-9 code categorical classification from Romano and colleagues.17

‡ ICD-9 procedure code.

J Neurosurg / Volume 120 / May 2014

Incidence and predictors of 30-day readmission after craniotomy

1205

LOSs were associated with greater odds of subsequent readmission. Medi-Cal beneficiaries were at an increased likelihood of readmission relative to the privately insured (OR 1.52, 95% CI 1.20–1.93) as were patients with his-tories of prior myocardial infarctions (OR 1.64, 95% CI 1.06–2.54).

The development of hydrocephalus during the index surgical admission was associated with a higher likeli-hood of 30-day readmission (OR 1.58, 95% CI 1.20–2.07) as was the development of a deep venous throm-bosis (DVT), pulmonary embolism (PE), or other venous complication (OR 3.88, 95% CI 1.84–8.18). Urinary tract infections during the index surgical episode were associ-ated with lower odds of 30-day readmission (OR 0.52, 95% CI 0.29–0.94).

Readmission Rates by HospitalForty hospitals (20.7%) had higher than expected

30-day readmission rates after craniotomy for resection of primary malignant brain tumors during the study pe-

riod (Fig. 1). Among patients discharged home during the study period, 501 craniotomies were performed at a pediatric hospital within California and 133 (26.6%) of these patients had at least one 30-day readmission, while 13,085 craniotomies were performed at a nonpediatric hospital and 1657 (12.7%) of these patients had at least 1 readmission. Additionally, among home-discharged pa-tients, 8885 craniotomies were performed at nonteaching hospitals with a readmission rate of 13.3%, while 4701 craniotomies were performed at teaching hospitals with a readmission rate of 12.9%.

DiscussionNeurosurgical Cost Burden of 30-Day Readmission

In this study it was found that 13.2% of all patients discharged home after craniotomy for tumor resection were rehospitalized within 30 days of discharge. Each re-admission represented an additional $20,296 in median

TABLE 3: Thirty-day outcomes among 13,586 patients discharged home following craniotomy for primary malignant brain tumor

Variable Value

30-day mortality (%) inpatient 81 (0.6) outpatient 235 (1.7) total 316 (2.3)30-day readmission (%) 1790 (13.2) patients w/ 1 readmission episode 1577 patients w/ 2 readmission episodes 196 patients w/ 3 readmission episodes 15 patients w/ 4 readmission episodes 2 total 30-day readmission episodes 2022 patients w/ readmission occurring at a different hospital from surgical site (%) 483 (27.0)  mean days to first 30-day readmission (median) 12.3 (11.0) mean LOS during 30-day readmission episode in days (median) 5.9 (4.0) mean charges ($) per 30-day readmission episode (median) 44,249 (20,296)reason for readmission by diagnosis (%) seizure & convulsive disorder 20.9 surgical infection of the CNS 14.5  motor deficit 12.8 DVT, PE, & venous complications 11.3 general infection 11.1 cerebral edema & compression of the brain 10.1 hydrocephalus 9.6 speech & language disorder 8.5 urinary tract infection 8.5 cerebrovascular disorder 6.3  nonspecific CNS surgical complication  6.3 sepsis & septicemia 4.9 malaise & fatigue 2.6 gait & coordination dysfunction 2.0 vision & optic disorder 1.3

L. P. Marcus et al.

1206 J Neurosurg / Volume 120 / May 2014

TABLE 4: Comparative analysis of nonreadmitted patients and patients with at least one 30-day readmission after home discharge

30-Day ReadmissionsVariable None (%) At Least 1 (%) p Value*

mean age in yrs (median) 48.4 (50.0) 47.2 (51.0) 0.031mean LOS on surgical admission in days (median) 6.2 (4.0) 9.2 (6.0) <0.001mean Charlson comorbidity index (median) 2.4 (2.0) 2.6 (2.0) <0.001race/ethnicity <0.001 non-Hispanic white 75.0 69.3 African American 2.9 3.8 Hispanic 14.7 19.5 Asian 5.0 4.8 Native American/other 2.6 2.6sex 0.032 male 58.5 61.2 female 41.5 38.8medical history hypertension 26.5 30.1 0.001 tobacco use disorder 18.2 17.3 0.345 cerebrovascular disease 8.5 9.3 0.257 chronic pulmonary disease 8.1 9.5 0.037 lipid disorder 7.7 9.8 0.003 diabetes mellitus 7.0 9.1 0.001 obesity & overweight 4.9 6.2 0.022 moderate or severe liver disease 3.2 4.3 0.016 myocardial infarction 2.3 3.7 <0.001 congestive heart failure 1.4 2.8 <0.001 peripheral vascular disease 0.8 1.1 0.287 peptic ulcer disease 0.4 0.6 0.236 renal disease 0.2 0.3 0.517 mild liver disease 0.2 0.4 0.038complications during surgical admission seizure & convulsive disorder 21.6 19.4 0.046 speech & language disorder 8.9 9.7 0.275  motor deficit 8.0 9.6 0.029 cerebral edema & compression of the brain 3.6 4.0 0.362 hydrocephalus 5.2 11.6 <0.001  nonspecific CNS surgical complication 1.0 2.5 <0.001 gait & coordination dysfunction 2.0 2.9 0.019 vision & optic disorder 2.6 3.1 0.246 malaise & fatigue 0.5 0.5 0.760 general infection 1.5 3.7 <0.001 sepsis & septicemia 0.3 1.3 <0.001 urinary tract infection 2.5 3.9 <0.001 DVT, PE, & venous complications 0.4 1.2 <0.001 cerebrovascular disorder 3.2 3.3 0.827 surgical infection of the CNS 0.8 2.4 <0.001expected primary payer <0.001 Medicare 17.9 20.7 Medi-Cal 9.2 16.3

(continued)

J Neurosurg / Volume 120 / May 2014

Incidence and predictors of 30-day readmission after craniotomy

1207

hospital charges on top of the $72,029 in charges for the index neurosurgical admission. While it is expected that a small fraction of all operative patients will be rehos-pitalized due to the new onset of unpreventable medical conditions following any surgical procedure, reducing the number of preventable unplanned 30-day readmissions could create significant cost savings for health care pay-ers. It was determined that a 40% reduction in the number of 30-day readmission episodes for brain tumor patients undergoing craniotomy within the state of California would eliminate 606 hospitalizations and create more than $12 million in cost savings.

Lowering readmission rates could also create cost savings by preventing redundant medical charges. In this study, it was observed that 27.0% of readmitted pa-tients were rehospitalized at a different location than the site that provided initial surgical care. Because provid-ers within different hospital systems typically operate in isolation from one another, it is reasonable to expect that the care of these readmitted patients included significant unnecessary health care costs in the form of redundant imaging, workup, and care.

In spite of the significant cost and morbidity burden that 30-day readmission places on patients, payers, and providers, it is important to note that most surgical cen-ters in California have observed readmission rates similar to those expected by multivariate modeling. In the present study, it was shown that 20.7% of hospitals had an observed to expected ratio of 30-day readmission rates that was sig-nificantly greater than 1. This implies that almost 80% of the hospitals in California performing craniotomy for tu-mor resection operated at an expected level of care qual-ity during the study period, and only one-fifth of hospitals would be financially penalized in the future according to Centers for Medicare & Medicaid Studies guidelines for 30-day readmission.3 While the scope of this study focused on patient-level factors, it is believed that future investiga-tion into specific hospital-level factors associated with higher than expected readmission rates is warranted.

Predictors of ReadmissionMany studies in the surgical literature regarding 30-

day readmissions establish postoperative complications as a common risk factor for readmission across multiple specialties.5,6,8,11,16 The present study similarly showed that certain complications arising during the index surgical admission increase the odds of readmission for patients after the resection of brain malignancy. Specifically, pa-tients who developed hydrocephalus during the surgical admission had 58% greater odds of readmission than those who did not develop hydrocephalus. Furthermore, patients who experienced a DVT, PE, or venous com-plication during surgical admission were nearly 4 times more likely to be readmitted within 30 days of discharge. It is thus proposed that more astute monitoring and man-agement of both hydrocephalus and venous complications in the perioperative period may reduce the frequency of 30-day patient rehospitalizations.

Previous studies have found that longer LOSs during surgical hospitalization are associated with an increased risk of readmission, with the assertion being that LOS serves as a surrogate for postoperative complications.11,18 The results of the present study showed a similar trend of increasing odds of readmission for LOSs greater than 1 week. We propose that longer LOSs are inherently as-sociated with a greater level of postoperative complica-tions and greater risk of acquiring nosocomial infections, which may increase the risk of readmission.

Certain components of a patient’s medical histo-ry were also shown to be associated with readmission. While other studies have shown an association between comorbidities and readmission, these studies have fo-cused solely on comorbidities present at admission as a predictor of readmission. The present study, in contrast, used a longitudinal database and was able to characterize the presence of pathologies throughout the portion of a patient’s medical history occurring during the study pe-riod rather than being limited to those pathologies pres-ent on initial surgical admission. Patients who had any history of myocardial infarction were observed to have a 64% greater likelihood of 30-day readmission than those patients without a history of myocardial infarction. This is the first study (to our knowledge) that examined pa-tients’ medical histories rather than comorbidities present

TABLE 4: Comparative analysis of nonreadmitted patients and patients with at least one 30-day readmission after home discharge (continued)

30-Day ReadmissionsVariable None (%) At Least 1 (%) p Value*

expected primary payer (continued) <0.001 private insurance 70.2 61.0 uninsured 2.7 2.0hospital teaching status 0.443 neurosurgical residency program 34.7 33.8 no neurosurgical residency program 65.3 66.2hospital pediatric status nonpediatric hospital 87.6 73.5 <0.001 pediatric hospital 12.4 26.5

*  Values in bold are statistically significant.

L. P. Marcus et al.

1208 J Neurosurg / Volume 120 / May 2014

TABLE 5: Odds of 30-day readmission among patients discharged home following craniotomy for primary brain tumor in California (1995–2010)*

Variable Adjusted OR (95% CI) p Value†

LOS on surgical admission (days) ≤1 1.12 (0.71–1.78) 0.620 2 0.82 (0.60–1.11) 0.203 3 1.10 (0.85–1.41) 0.480 4 1.00 (reference) 5 1.16 (0.86–1.55) 0.329 6 1.25 (0.92–1.69) 0.149 7–13 1.64 (1.29–2.07) <0.001 14–20 2.17 (1.53–3.09) <0.001race/ethnicity non-Hispanic white 1.00 (reference) African American 1.26 (0.85–1.87) 0.252 Hispanic 0.86 (0.69–1.06) 0.150 Asian 0.85 (0.61–1.18) 0.328 Native American/other 1.00 (0.67–1.53) 0.987sex male 1 (reference) female 0.93 (0.80–1.07) 0.282medical history hypertension 0.97 (0.81–1.17) 0.777 tobacco use disorder 0.89 (0.73–1.09) 0.258 cerebrovascular disease 1.73 (0.43–7.01) 0.438 chronic pulmonary disease 0.99 (0.76–1.29) 0.949 lipid disorder 1.17 (0.90–1.52) 0.233 diabetes mellitus 1.21 (0.92–1.59) 0.175 obesity & overweight 1.00 (0.71–1.41) 0.989 moderate or severe liver disease 1.25 (0.84–1.86) 0.275 myocardial infarction 1.64 (1.06–2.54) 0.026 congestive heart failure 1.06 (0.57–1.97) 0.863 peptic ulcer disease 1.03 (0.29–3.74) 0.956 mild liver disease 2.89 (0.71–11.75) 0.137complications during surgical admission seizure & convulsive disorder 0.90 (0.76–1.08) 0.255 speech & language disorder 1.06 (0.84–1.34) 0.608  motor deficit 0.93 (0.73–1.21) 0.603 cerebral edema & compression of the brain 1.01 (0.72–1.43) 0.940 hydrocephalus 1.58 (1.20–2.07) 0.001  nonspecific CNS surgical complication 1.03 (0.52–2.03) 0.931 gait & coordination dysfunction 0.98 (0.62–1.54) 0.913 vision & optic disorder 1.22 (0.84–1.79) 0.297 malaise & fatigue 0.84 (0.29–2.43) 0.752 general infection 1.82 (0.91–3.64) 0.089 sepsis & septicemia 0.80 (0.24–2.70) 0.720 urinary tract infection 0.52 (0.29–0.94) 0.032 DVT, PE, & venous complications 3.88 (1.84–8.18) <0.001 cerebrovascular disorder 0.41 (0.10–1.71) 0.221 surgical infection of the CNS 0.63 (0.26–1.55) 0.316

(continued)

J Neurosurg / Volume 120 / May 2014

Incidence and predictors of 30-day readmission after craniotomy

1209

on admission, and we propose that additional investiga-tion be undertaken to better understand the relationship between medical history and readmission.

Lastly, it was found that Medi-Cal beneficiaries had increased odds for readmission relative to privately in-sured patients (OR = 1.52). Initially, it was believed that perhaps Medi-Cal patients were at increased odds of re-admission because of premature discharge after surgery. However, on further analysis it was shown that the me-dian LOS during surgical admission was 4 days for pri-vately insured patients and 8 days for Medi-Cal beneficia-ries. As a result, it is proposed that certain factors within the population of Medi-Cal patients predispose them to

longer LOSs and greater odds of rehospitalization. We believe that a randomized clinical trial is necessary to better understand the factors that increase the likelihood of readmission among Medi-Cal patients. The results of such future studies could prove to be essential to our un-derstanding of 30-day readmissions in neurosurgery. Ad-ditionally, it is possible that there may be differences in both the surgical technologies used and the index surgical procedures performed on Medi-Cal beneficiaries versus privately insured patients that increase the likelihood of readmission in the Medi-Cal group. Early application of technologies such as intraoperative MRI, image-guided surgery platforms, and cranial tractography may create

TABLE 5: Odds of 30-day readmission among patients discharged home following craniotomy for primary brain tumor in California (1995–2010)* (continued)

Variable Adjusted OR (95% CI) p Value†

expected primary payer Medicare 1.00 (0.72–1.39) 0.992 Medi-Cal 1.52 (1.20–1.93) 0.001 private insurance 1.00 (reference) uninsured 0.69 (0.44–1.11) 0.125hospital teaching status no neurosurgical residency program 1.00 (reference) neurosurgical residency program 1.07 (0.89–1.29) 0.448hospital pediatric status nonpediatric hospital 1.00 (reference) pediatric hospital 1.64 (1.00–2.70) 0.051

* Other covariates not included in Table 5 are patient age, California county, and calendar year.†  Values in bold are statistically significant.

Fig. 1. Observed/expected 30-day readmission rates by hospital. The triangles represent each California (CA) hospital’s observed readmission rate divided by the expected readmission rate for that hospital as calculated from our multivariate model of readmissions. A hospital that has an observed rate equal to the predicted/expected rate has a triangle at 1 (red line). The error bars represent the 95% CI around each triangle. Forty (20.7%) of 193 hospitals had observed/expected 30-day readmission rates that were significantly greater than 1 (p ≤ 0.05).

L. P. Marcus et al.

1210 J Neurosurg / Volume 120 / May 2014

differences in neurooncological care based on patient in-surance status. The current population-based study, how-ever, is unable to resolve the differences that may exist in the types of procedures performed between insurance groups. We hope that future investigations will better de-tail the disparities in surgical care provided to Medi-Cal beneficiaries compared with their privately insured coun-terparts.

Clinical ApplicationsA better understanding of the factors that influence

the likelihood of readmission for neurosurgical patients will allow providers to institute programs for reducing 30-day readmissions. Previous studies have begun to de-fine the types of systematized action that hospitals may take to reduce readmission rates, and we believe that the results of the present study highlight potential interven-tions that require further investigation.

Because the two most common diagnoses present on readmission were seizure and CNS infection, we suggest that better and more aggressive seizure prophylaxis and antibiotic regimens could potentially reduce readmission rates after the resection of brain malignancy. Furthermore, there is evidence to suggest that better patient education and medication counseling as part of a discharge bundle approach decreases unplanned acute care utilization at 30 days.9,12 A discharge bundle for neurosurgical patients after tumor resection could include education about the symptoms of surgical site infection and seizure disor-der that warrant a return to the clinic. This intervention would allow early identification and treatment of develop-ing postoperative morbidity in an outpatient setting rather than through inpatient readmission. A neurosurgical dis-charge bundle could also include patient counseling from a clinical pharmacist about correct dosing schedules and durations of pharmacotherapies. While the time and re-sources required for the implementation of this discharge bundle approach may not be feasible for all patients, the patient groups at risk for readmission highlighted in this study may benefit from this intervention.

Postdischarge interventions directed at these high-risk patient groups could also prove to be of benefit in reducing 30-day readmissions.7 The Agency for Health-care Research and Quality has established timely follow-up appointment scheduling as an important component of the discharge workflow that may reduce readmission.1 The time course for scheduling follow-up appointments in current neurosurgical practice varies between institu-tions; however, the results of this study establish 11 days as the median time to 30-day readmission. Implicit in this finding is that half of all readmissions will occur after 11 days postdischarge, and we believe that this is an im-portant time point for surgeons to identify patients with developing pathologies that have the potential to cause rehospitalization.

Strengths and LimitationsThis is the first investigation to analyze 30-day re-

admissions in neurosurgical practice, and it is important to address both the strengths and the limitations in the

present study. First, analyses that use large administra-tive databases rely on the use of ICD-9 diagnosis codes to identify patients and their outcomes. These ICD-9 diag-nosis codes are used for hospital billing, and lack clinical data such as histological grades, radiological diagnoses, and physiological parameters. As a consequence, the cur-rent study was not able to investigate the epidemiology of readmission for specific types of intracranial malignancy. Nevertheless, we believe that this investigation represents an important step toward a better macro understanding of the readmission process following craniotomy for intra-cranial malignancy. We expect that future studies will be more granular (that is, specifically tailored to each unique tumor classification) by drawing upon the conclusions of this report to better elucidate the differences in readmis-sion between distinct classifications of brain tumors.

Second, ICD-9 diagnosis coding is often performed by lower level clinical professionals rather than by expe-rienced clinicians and surgeons. As such, inaccuracies in diagnostic and procedural coding may be introduced to the data set. While the use of ICD-9 coding may be asso-ciated with potential inaccuracies, it is unlikely that these inaccuracies would be biased or systematic in a manner that would affect our conclusions. Instead, the use of administrative data in this study increased the statisti-cal power of our analysis by allowing for a more robust sample size and a reduction in Type II errors.

Third, unlike a prospective study, the nature and processes related to readmission are inferred. We took several steps to clarify the types of patients being stud-ied in this report. First, we focused on patients who were discharged home. While there are likely preventable re-admissions in patients who are discharged to rehabilita-tion or nursing facilities, we wished to focus on patients who had achieved the best initial outcomes following craniotomy and were discharged home. We also sought to avoid potentially confounding effects of patients who were transferred from one facility to another facility for a higher level of care in this report. We plan to report on patterns of transfer of care in a future report. Second, we chose to analyze readmission episodes that were un-planned. In some patients, a practice pattern of biopsy followed by readmission for definitive craniotomy may be performed. As a result, we excluded all planned read-missions from our analysis to avoid characterizing this practice pattern as a preventable 30-day readmission. Our exclusion of planned readmissions is also in accordance with the Centers for Medicare and Medicaid Services recommendations for defining preventable readmission.3 Lastly, we chose to exclude patients who received a previ-ous craniotomy from our analysis to ensure that we cap-tured surgical episodes for primary brain tumors.

We believe that an additional advantage of the lon-gitudinal California inpatient discharge database used in this study was the ability to identify readmissions to different hospitals. Most investigations looking at 30-day readmission after surgery rely on same-hospital readmis-sion data found in single institutional electronic medical records or databases such as the National Surgical Quali-ty Improvement Program. However, Nasir and colleagues have shown that same-hospital readmission rates may

J Neurosurg / Volume 120 / May 2014

Incidence and predictors of 30-day readmission after craniotomy

1211

not be effective surrogates for true readmission rates.13 The use of a longitudinal data set that follows each pa-tient across multiple hospitals and across multiple years allowed us to avoid the underestimation of true readmis-sion rates noted in other studies.

ConclusionsIn this report, we determined the current 30-day re-

admission rate in patients undergoing brain tumor surgery who were discharged home in California. Certain medi-cal conditions that are highly prevalent on readmission such as seizure may highlight the need for targeted in-terventions to improve patient education and therapy sur-rounding seizure prevention. In addition, this work lays the foundational design for targeted prospective studies of readmission and targeted interventions to reduce 30-day readmission. We also demonstrated the potential cost savings associated with reduction in readmission rates following brain tumor surgery.

Disclosure

The authors report no conflict of interest concerning the mate-rials or methods used in this study or the findings specified in this paper.

Author contributions to the study and manuscript prepara-tion include the following. Conception and design: Marcus, Chen, Chang. Acquisition of data: Chang. Analysis and interpretation of data: all authors. Drafting the article: Marcus. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Carter. Statistical analysis: Carter, Marcus, McCutcheon, Noorbakhsh, Parina, Gonda, Chang. Administrative/technical/mate-rial support: Noorbakhsh, Chang. Study supervision: Carter, Gonda, Chen, Chang.

References

1. Agency for Healthcare Research and Quality: Re-engineered discharge project dramatically reduces return trips to the hospital. Rockville, MD: Agency for Healthcare Research and Quality, 2011 (http://www.ahrq.gov/research/mar11/0311RA1.htm) [Accessed January 21, 2014]

2. Barker FG II, Curry WT Jr, Carter BS: Surgery for primary su-pratentorial brain tumors in the United States, 1988 to 2000: the effect of provider caseload and centralization of care. Neuro Oncol 7:49–63, 2005

3. Centers for Medicare and Medicaid Services: Readmissions Reduction  Program. Baltimore: Centers for Medicare and Medicaid Services, 2013 (http://cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html) [Accessed January 22, 2014]

4. Curry WT Jr, Carter BS, Barker FG II: Racial, ethnic, and so-cioeconomic disparities in patient outcomes after craniotomy for tumor in adult patients in the United States, 1988-2004. Neurosurgery 66:427–438, 2010

5. Hannan EL, Racz MJ, Walford G, Ryan TJ, Isom OW, Bennett E, et al: Predictors of readmission for complications of coro-nary artery bypass graft surgery. JAMA 290:773–780, 2003

6. Hannan EL, Zhong Y, Lahey SJ, Culliford AT, Gold JP, Smith CR, et al: 30-day readmissions after coronary artery bypass graft surgery in New York state. JACC Cardiovasc Interv 4: 569–576, 2011

7. Hansen LO, Young RS, Hinami K, Leung A, Williams MV: Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med 155:520–528, 2011

8. Hendren S, Morris AM, Zhang W, Dimick J: Early discharge and hospital readmission after colectomy for cancer. Dis Co-lon Rectum 54:1362–1367, 2011

9. Jack BW, Chetty VK, Anthony D, Greenwald JL, Sanchez GM, Johnson AE, et al: A reengineered hospital discharge program to decrease rehospitalization: a randomized trial. Ann Intern Med 150:178–187, 2009

10. Jencks SF, Williams MV, Coleman EA: Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med 360:1418–1428, 2009

11. Kassin MT, Owen RM, Perez SD, Leeds I, Cox JC, Schnier K, et al: Risk factors for 30-day hospital readmission among general surgery patients. J Am Coll Surg 215:322–330, 2012

12. Koehler BE, Richter KM, Youngblood L, Cohen BA, Prengler ID, Cheng D, et al: Reduction of 30-day postdischarge hospi-tal readmission or emergency department (ED) visit rates in high-risk elderly medical patients through delivery of a tar-geted care bundle. J Hosp Med 4:211–218, 2009

13. Nasir K, Lin Z, Bueno H, Normand SL, Drye EE, Keenan PS, et al: Is same-hospital readmission rate a good surrogate for all-hospital readmission rate? Med Care 48:477–481, 2010

14. Patient Protection and Affordable Care Act of 2010, Pub Law 111–148, 124 Stat. 127, Sec. 6301 (March 23, 2010) (http://www.gpo.gov/fdsys/pkg/PLAW-111publ148/pdf/PLAW-111publ148.pdf) [Accessed January 21, 2014]

15. Rau J: Medicare to penalize 2,217 hospitals for excess readmis-sions. Kaiser Health News. August 13, 2012 (http://www.kai serhealthnews.org/stories/2012/august/13/medicare-hospitals-readmissions-penalties.aspx) [Accessed January 14, 2014]

16. Reddy DM, Townsend CM Jr, Kuo YF, Freeman JL, Goodwin JS, Riall TS: Readmission after pancreatectomy for pancreatic cancer in Medicare patients. J Gastrointest  Surg 13:1963–1975, 2009

17. Romano PS, Roos LL, Jollis JG: Adapting a clinical comorbid-ity index for use with ICD-9-CM administrative data: differ-ing perspectives. J Clin Epidemiol 46:1075–1090, 1993

18. Schneider EB, Hyder O, Brooke BS, Efron J, Cameron JL, Edil BH, et al: Patient readmission and mortality after colorectal surgery for colon cancer: impact of length of stay relative to other clinical factors. J Am Coll Surg 214:390–399, 2012

Manuscript submitted July 2, 2013.Accepted January 16, 2014.Please include this information when citing this paper: pub-

lished online March 7, 2014; DOI: 10.3171/2014.1.JNS131264.Address correspondence to: Bob S. Carter, M.D., Ph.D., UC San

Diego Division of Neurosurgery, 200 W. Arbor Dr., San Diego, CA 92103-8893. email: bobcarter@gmail.com.