- Email: [email protected]
Rates and risk factors of unplanned 30-day readmission following general and thoracic pediatric surgical procedures
Journal of Pediatric Surgery 52 (2017) 1239–1244
Contents lists available at ScienceDirect
Journal of Pediatric Surgery journal homepage: www.elsevier.com/locate/jpedsurg
Clinical Papers
Rates and risk factors of unplanned 30-day readmission following general and thoracic pediatric surgical procedures☆ Stephanie F. Polites a,⁎, Donald D. Potter a, Amy E. Glasgow b, Denise B. Klinkner a, Christopher R. Moir a, Michael B. Ishitani a, Elizabeth B. Habermann b a b
Division of Pediatric Surgery, Mayo Clinic, Rochester, MN Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN
a r t i c l e
i n f o
Article history: Received 14 September 2016 Received in revised form 10 November 2016 Accepted 28 November 2016 Key words: Readmission Pediatric surgery Quality
a b s t r a c t Background/Purpose: Postoperative unplanned readmissions are costly and decrease patient satisfaction; however, little is known about this complication in pediatric surgery. The purpose of this study was to determine rates and predictors of unplanned readmission in a multi-institutional cohort of pediatric surgical patients. Methods: Unplanned 30-day readmissions following general and thoracic surgical procedures in children b18 were identified from the 2012–2014 National Surgical Quality Improvement Program- Pediatric. Timedependent rates of readmission per 30 person–days were determined to account for varied postoperative length of stay (pLOS). Patients were randomly divided into 70% derivation and 30% validation cohorts which were used for creation and validation of a risk model for readmission. Results: Readmission occurred in 1948 (3.6%) of 54,870 children for a rate of 4.3% per 30 person–days. Adjusted predictors of readmission included hepatobiliary procedures, increased wound class, operative duration, complications, and pLOS. The predictive model discriminated well in the derivation and validation cohorts (AUROC 0.710 and 0.701) with good calibration between observed and expected readmission events in both cohorts (p N .05). Conclusions: Unplanned readmission occurs less frequently in pediatric surgery than what is described in adults, calling into question its use as a quality indicator in this population. Factors that predict readmission including type of procedure, complications, and pLOS can be used to identify at-risk children and develop prevention strategies. Level of evidence: III. © 2017 Elsevier Inc. All rights reserved.
Unplanned readmissions following surgical procedures are costly and disruptive to patients and their families. This complication is considered largely preventable and is being increasingly used as a quality indicator for hospitals and as a modifier for reimbursement. Initially, this was limited to certain adult procedures, as The Centers for Medicare and Medicaid Services (CMS) withhold up to 3% of reimbursement from hospitals with high readmission rates following coronary artery bypass grafting and joint replacement procedures [1]. However, it is now applicable to pediatric surgery as some states are imposing Medicaid and Children's Health Insurance (CHIP) reimbursement penalties for high readmission rates [2]. The CHIP Reauthorization Act of 2009 is in the Abbreviations: CMS, Centers for Medicare and Medicaid Services; CHIP, Children's Health Insurance; NACHRI, National Association of Children's Hospitals and Related Institutions; ACS, American College of Surgeons’; NSQIP-P, National Surgical Quality Improvement Program-Pediatric; GI, gastrointestinal; AUROC, area under the receiver operating characteristic. ☆ No external funding was secured for this study and the authors have no disclosures. ⁎ Corresponding author at: Department of Surgery, 200 1st Street SW, Rochester, MN 55901. Tel.: +1 507 255 5123. E-mail address: [email protected] (S.F. Polites). http://dx.doi.org/10.1016/j.jpedsurg.2016.11.043 0022-3468/© 2017 Elsevier Inc. All rights reserved.
process of creating hospital readmission quality measures [3]. Unfortunately the expansion of readmission penalties into the pediatric realm and use as a quality indicator has occurred in the absence of adequate evidence, as readmission following general and thoracic pediatric surgery is not well studied. Among more than 500,000 admissions at 72 National Association of Children's Hospitals and Related Institutions (NACHRI) hospitals, the all-cause unplanned readmission rate was 6.5% [4]. Another single institution study of all pediatric readmissions estimated nearly one third to be preventable [2]. In adults the readmission rate following general, vascular, and thoracic surgery in a national cohort was 7.8% and readmission was closely linked to the American Society of Anesthesiologists (ASA) class, length of stay (LOS), and postoperative complications [5]. It is unknown if these rates of readmission and risk factors can be extrapolated to pediatric general surgery. This information must be determined in order to inform reimbursement penalties and readmission prevention strategies. Thus, the purpose of this study was to determine the rate of unplanned and related postoperative readmissions in a national cohort of general and thoracic pediatric surgical patients and to determine predictors of readmission in this population.
1240
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1. Methods 1.1. Patient cohort and data source Children b18 years of age who underwent a general or thoracic surgical procedure were identified from the 2012–2014 American College of Surgeons' (ACS) National Surgical Quality Improvement ProgramPediatric (NSQIP-P) for inclusion in this study. Neonates who underwent operations b29 days after birth and those patients with missing data for postoperative length of stay were excluded. The NSQIP-P was designed to provide participating hospitals with riskadjusted performance measures [6]. A standardized 8-day sampling cycle of children who underwent inpatient or outpatient surgical procedures was used by dedicated clinical nurse abstracters to obtain data on more than 100 variables from patient medical records. Approximately 1400 cases were contributed annually from each institution. Surgical procedures were identified by current procedural terminology (CPT) code. In addition to demographic and procedural information, standardized definitions of comorbidities and postoperative complications were utilized. The follow-up period for NSQIP-P was 30 days from the index procedure. Any complications, readmissions, or death within this period were recorded. For patients who did not have follow up at the index hospital, telephone calls were made by NSQIP-P registrars to assess for these events. 1.2. Data collection and analysis CPT codes were used to identify patients who underwent a general or thoracic pediatric surgical procedure. Procedures were then categorized into one of eight groups—anorectal, general, gynecologic, head and neck, hepatobiliary, small and large intestine, thoracic, and upper gastrointestinal (GI). General included procedures such as appendectomy, cholecystectomy, and splenectomy (Appendix 1). The NSQIP-P collects information on up to five readmissions within the 30-day followup period and the NSQIP methodology for identifying readmissions has been validated as accurate [7]. Readmissions are classified as planned or unplanned and related or unrelated to the index procedures. Only unplanned, related readmissions were of interest and collected for this study. One reason for readmission is provided by ICD-9 diagnosis code. Demographic, comorbidity, operative, and postoperative information including complications were also collected for each patient. The unplanned readmission rate was determined for the entire cohort and for each procedure category. Since follow up in NSQIP is 30 days from the index procedure, patients had varied lengths at which they were at risk for readmission depending on their postoperative length of stay. For example, a patient discharged on postoperative day one had 29 days at risk for readmission while a patient discharged on postoperative day 16 had 14 days at risk. Patients who died postoperatively were also not at risk for readmission following death. Therefore, readmission rates per 30 person–days at risk with 95% confidence intervals for readmission were calculated based on methodology we have previously reported [8]. Only days following patients' discharge from their index procedure hospitalization which they were alive contributed to the denominator. The duration of readmission was not collected in NSQIP-P; therefore, each readmission was assumed to be 5 days based on previously reported methodology [8]. Patients still hospitalized at 30 days (n = 308), and those who died during their hospitalization or within 30 days from the index procedure (n = 4) were excluded. To identify and validate independent risk factors for readmission patients were randomly divided into a 70% derivation cohort and 30% validation cohort. Patients with at least one unplanned readmission related to the index surgical procedure were compared to non-readmitted patients in the derivation cohort using standard univariate analysis techniques. Only postoperative complications occurring prior to the readmission were included for readmitted patients. Continuous
variables were presented as mean (standard deviation) and categorical variables as n (percentage). Student's t tests were used to compare continuous variables and chi squared tests were used for categorical variables. Variables statistically significant on univariate analysis within the derivation cohort were used to create a stepwise Cox proportional hazards regression model for readmission using derivation cohort patients. Cox regression was used to account for varied postoperative length of day and subsequent days at risk for readmission within the 30-day follow-up period. Results were reported as hazard ratios (HR) with 95% confidence intervals and p values. Consistent with methods previously described in the literature by Donzé et al., β coefficients of significant variables were applied to the validation set [9]. Discriminatory power of both derivation and validation set models was assessed using area under the receiver operating characteristic (AUROC) analysis. Observed and expected readmission events were determined for both derivation and validation cohorts using Poisson regression and calibration testing was performed. Statistical significance was acknowledged when p b .05. Statistical analyses were performed using SAS (©SAS Institute, Cary, NC). 2. Results 2.1. Cohort characteristics Of 54,870 pediatric surgical patients included, the majority underwent general surgery procedures (61.1%) followed by upper GI (14.5%) and small and large intestine procedures (12.9%; Table 1). The mean age was 8.1 (5.8) years and 56.9% of patients were male. Outpatient procedures were performed in 14,440 patients (26.3%) and 28,166 (51.3%) underwent urgent or emergent procedures. Comorbidities were present in 28,154 (51.3%) patients. Median length of stay was 2.0 (1.0–5.0) days and 1819 (3.3%) of patients had postoperative complications. 2.2. Readmission rates and reasons Readmission occurred in 1948 (3.6%) patients with a resulting timedependent rate of 4.3% per 30 person–days for the overall cohort. The readmission rate was highest in patients who underwent hepatobiliary procedures (11.6% per 30 person–days) followed by small and large intestine (9.5% per 30 person–days) and anorectal procedures (5.3% per 30 person–days; Fig. 1). Patients who suffered complications also had higher readmission rates (15.8% per 30 person–days) than those without complications (4.0% per 30 person–days). Individual complications were infrequent and complications were subsequently analyzed as a dichotomous variable (Table 2). The ICD-9 diagnosis code reason for readmission was missing in 759 (38.5%) readmitted patients and the remaining readmitted patients had one of 298 unique diagnosis codes as their reason for readmission. The most frequent categories of reasons for readmission were bowel obstruction or constipation (11.8%), infection (7.8%), oral intolerance or dehydration (7.0%), and pain (6.7%). 2.3. Characteristics of readmitted patients There were no significant differences in age, sex, or race between readmitted and non-readmitted patients in the derivation cohort (n = 38,397; Table 1). Readmission was more frequent in patients with inpatient procedures, existing comorbidities, weight percentile b 5%, greater ASA class, increased operative duration, and greater wound contamination. While 2.5% of patients who underwent operations b60 min in duration were readmitted, readmission occurred in 6.9% of those who underwent operations N140 min (p b .001). Those with a postoperative complication and those with increased postoperative length of stay were also more frequently readmitted than those without, respectively.
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1241
Table 1 Characteristics of All Patients and Readmitted vs Non-readmitted Patients in the Derivation Set. Patient characteristic
All patients N = 54,870
Age, mean (SD), years Female sex Race
White Unknown Black Other General Upper GI Small/Large Intestine Thoracic Gynecology Head and Neck Hepatobiliary Anorectal Inpatient Outpatient b5% 5–94% ≥95%
Procedure group1
Procedure type Weight percentage2
Premature birth3 Congenital malformation Any comorbidity Individual comorbidities
Diabetes mellitus Respiratory GI/Hepatobiliary Major cardiac Acute renal failure Neurologic Immunosuppression Nutrition Hematologic Active malignancy
Preoperative SIRS, sepsis, septic shock Preoperative transfusion or hematocrit b32 Case status ASA class
Operative time, minutes
Wound class
Preoperative length of stay, days Postoperative complication
4
Postoperative length of stay, days
Elective Emergent/urgent 1 2 3 4/5 b60 60–140 N140 Clean Clean/Contaminated Contaminated Dirty/Infected b2 ≥2 Yes No b2 2–4 N4
8.1 (5.8) 23,634 (43.1%) 40,089 (73.1%) 6647 (12.1%) 6444 (11.7%) 1690 (3.1%) 33,516 (61.1%) 7982 (14.5%) 7081 (12.9%) 3397 (6.2%) 1105 (2.0%) 696 (1.3%) 606 (1.1%) 487 (0.9%) 40,430 (73.7%) 14,440 (26.3%) 8865 (16.2%) 39,051 (71.2%) 6954 (12.7%) 4956 (9.0%) 9544 (17.4%) 28,154 (51.3%) 343 (0.6%) 7327 (13.4%) 7327 (13.4%) 2148 (3.9%) 153 (0.3%) 7707 (14.0%) 2265 (4.1%) 6429 (11.7%) 6032 (11.0%) 1458 (2.7%) 9621 (17.5%) 582 (1.1%) 26,704 (48.7%) 28,166 (51.3%) 16,590 (30.2%) 24,710 (45.0%) 12,515 (22.8%) 1055 (1.9%) 31,966 (58.3%) 17,083 (31.1%) 5821 (10.6%) 10,962 (20.0%) 25,196 (45.9%) 12,261 (22.3%) 6451 (11.8%) 48,678 (88.7%) 6192 (11.3%) 53,051 (96.7%) 1819 (3.3%) 28,486 (51.9%) 14,766 (26.9%) 11,618 (21.2%)
70% Derivation cohort n = 38,397 No readmission n = 37,061
Readmission n = 1336
8.1 (5.8) 15,951 (96.5%) 27,110 (96.5%) 4452 (96.8%) 4347 (96.8%) 1152 (95.8%) 22,812 (97.2%) 5413 (96.7%) 4600 (93.1%) 2312 (97.3%) 742 (98.1%) 455 (95.8%) 407 (91.3%) 320 (94.4%) 27,097 (96.0%) 9964 (98.1%) 5990 (96.1%) 26,300 (96.5%) 4771 (97.1%) 3337 (96.3%) 6387 (95.5%) 18,867 (95.9%) 241 (94.5%) 4948 (96.3%) 12,027 (95.5%) 1437 (95.2%) 93 (90.3%) 5082 (95.4%) 1497 (94.5%) 4241 (95.1%) 4024 (95.4%) 915 (94.6%) 6436 (95.2%) 378 (95.5%) 18,017 (96.3%) 19,044 (96.7%) 11,265 (97.1%) 16,716 (96.7%) 8392 (95.6%) 688 (94.4%) 21,759 (97.5%) 11,497 (95.8%) 3805 (93.1%) 7485 (97.4%) 17,021 (96.6%) 8421 (97.3%) 4134 (93.0%) 32,918 (96.6%) 4143 (95.8%) 35,868 (96.7%) 1193 (91.3%) 19,652 (98.4%) 9882 (95.8%) 7527 (93.0%)
8.2 (5.8) 574 (3.5%) 994 (3.5%) 148 (3.2%) 144 (3.2%) 50 (4.2%) 653 (2.8%) 184 (3.3%) 343 (6.9%) 64 (2.7%) 14 (1.9%) 20 (4.2%) 39 (8.7%) 19 (5.6%) 1143 (4.0%) 193 (1.9%) 246 (3.9%) 949 (3.5%) 141 (2.9%) 128 (3.7%) 304 (4.5%) 800 (4.1%) 14 (5.5%) 189 (3.7%) 562 (4.5%) 73 (4.8%) 10 (9.7%) 244 (4.6%) 87 (5.5%) 219 (4.9%) 195 (4.6%) 52 (5.4%) 324 (4.8%) 18 (4.5%) 683 (3.7%) 653 (3.3%) 331 (2.9%) 575 (3.3%) 389 (4.4%) 41 (5.6%) 547 (2.5%) 506 (4.2%) 283 (6.9%) 198 (2.6%) 592 (3.4%) 236 (2.7%) 310 (7.0%) 1155 (3.4%) 181 (4.2%) 1222 (3.3%) 114 (8.7%) 330 (1.7%) 437 (4.2%) 569 (7.0%)
p value .39 .98 .28
b.001
b.001 .009
.75 b.001 b.001 .09 .39 b.001 .003 .003 b.001 b.001 b.001 b.001 .005 b.001 .27 .08 b.001
b.001
b.001
.008 b.001 b.001
1 Percentages of patients readmitted by surgical procedure group are intended to describe the cohort and are not adjusted for length of stay in contrast to person–day rates of readmission by surgical procedure group in Fig. 1. 2 Weight percentages obtained by weight and sex per CDC guidelines. 3 Premature birth defined as ≤36 weeks gestation. 4 Includes any postoperative complication. Individual complications are outlined in Table 2.
2.4. Independent predictors of readmission Twenty variables were initially entered into the model which discriminated well in the derivation cohort; AUROC 0.710, 95% CI 0.694–0.726 (Table 3). After adjustment, children who underwent head and neck, hepatobiliary, and small and large intestine operations remained at increased risk for readmission compared to general procedures while thoracic procedures carried the lowest risk (HR = 0.69, p = .009). Preoperative acute renal failure, neurologic comorbidity, preoperative systemic infection, wound classification, operative duration, and postoperative complications were also associated with readmission. The
strongest predictor of readmission was increased postoperative length of stay. Children hospitalized for N 4 days following their index procedure were more than 3 times as likely to be readmitted than children discharged within 2 days (HR = 3.12, p b .001). 2.5. Validation of prediction model The model discriminated well when applied to the validation cohort (AUROC 0.701, 95% CI 0.677–0.724), as the c statistic for the validation cohort remained within the confidence interval of the derivation cohort. Observed and expected readmission events were then determined and
1242
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
16%
12%
11.6% 9.5%
8% 5.3% 4%
5.0%
4.3%
3.7%
3.3% 1.8%
0%
Fig. 1. Readmission rates per 30 person–days by procedure category.
the O/E ratios were 1.03 and 1.07 in the derivation and validation cohorts, respectively, with good calibration (derivation p = 0.95; validation p = 0.36; Table 4). 3. Discussion In this study readmission rates were determined for general and thoracic pediatric surgical patients that account for varied postoperative length of stay and opportunity for readmission using a multiinstitutional data source that captures all readmissions within 30 days. Simple predictors for readmission were also identified that may be used to identify at-risk patients and direct efforts to reduce readmission accordingly. These results will help to inform policies and benchmarking related to readmission. Existing efforts are hampered by small sample sizes and inability to ascertain readmissions that did not occur at the index hospital [10]. The overall unplanned readmission rate per 30 person–days of 4.3% in this study is lower than the 5–8% described for adult general surgery patients using a similar data source as well as the 6.5% reported for all pediatric admissions at 72 children's hospitals [4,5,11,12]. While there are no large scale studies of readmission following general and thoracic
pediatric surgery specifically in the United States to serve as a reference, a readmission rate of 2% was identified in a United Kingdom study of more than 2000 pediatric surgery procedures [13]. Based on existing literature, unplanned readmission appears to be less frequent an occurrence following general pediatric surgical procedures than adult procedures, calling into question the use of readmission as a marker of quality in pediatric surgery. This determination would be aided by institution-specific pediatric general and thoracic surgery readmission rates which are not available in the NSQIP-P. Participating hospitals, however, do receive their readmission data and may compare this to the national rate described in this study to direct quality improvement. Consistent with existing literature, readmission rates varied by procedure type in this study and the highest rates were in children who underwent hepatobiliary and intestinal procedures. These procedures
Table 3 Multivariable Cox Proportional Hazard Regression Analysis Predicting Readmission in 70% Derivation Cohort. Covariate Procedure group (vs general)
Table 2 Complications in Readmitted vs Non-Readmitted Patients. Total (N = 54,870)
% Not readmitted (n = 52,922)
% Readmitted (n = 1948)
p value
Any complication/postoperative ocurrence1 No 53,051 (96.7%) 51,259 (96.6%) Yes 1819 (3.3%) 1663 (91.4%)
1792 (3.4%) 156 (8.6%)
b.001
Superficial Incisional SSI No 54,736 (99.8%) Yes 134 (0.2%)
52,797 (96.5%) 125 (93.3%)
1939 (3.5%) 9 (6.7%)
.05
Deep SSI2 No 54,425 (99.2%) Yes 445 (0.8%)
52,526 (96.5%) 396 (89.0%)
1899 (3.5%) 49 (11.0%)
b.001
Pulmonary complications No 54,663 (99.6%) Yes 207 (0.4%)
52,729 (96.5%) 193 (93.2%)
1934 (3.5%) 14 (6.8%)
.012
Sepsis/CL associated bloodstream infection No 54,691 (99.7%) 52,759 (96.5%) Yes 179 (0.3%) 163 (91.1%)
1932 (3.5%) 16 (8.9%)
b.001
SSI = surgical site infection. CL = central line. 1 Includes all postoperative occurrences and complications in NSQIP-P. Only the most frequent and clinically relevant are displayed subsequently in this table. 2 Deep SSI includes deep incisional and organ space infections.
Anorectal Gynecology Head and Neck Hepatobiliary Small and Large Intestine Thoracic Upper GI
Preoperative acute renal failure Neurologic comorbidity1 SIRS/Sepsis/Septic Shock within 48 h prior to index procedure Wound class Clean/Contaminated (vs clean) Contaminated Dirty/Infected Operative time 60–140 min (vs b60 min) N140 min Any complication Postoperative 2–4 days length of stay N4 days (vs b2 days)
β Hazard LCL Coefficient ratio
UCL
−0.46 −0.73 0.41 0.06 −0.001
1.59 0.77 2.40 1.69 1.59
0.98 0.45 1.48 1.17 1.33
2.58 .06 1.32 .34 3.89 b.001 2.44 .005 1.90 b.001
−0.84 −0.37 0.91
0.69 1.10 2.47
0.52 0.91 0.89 1.36 1.31 4.66
.009 .38 .005
0.18
1.30
1.05 1.62
.029
0.18
1.20
1.02 1.41
.033
0.08 0.26 0.65 0.19 0.41 0.29 0.78 1.14
1.08 1.29 1.92 1.21 1.51 1.34 2.17 3.12
0.91 1.05 1.53 1.06 1.26 1.09 1.84 2.60
1.29 1.60 2.40 1.39 1.81 1.65 2.57 3.74
p value
.36 .018 b.001 .005 b.001 .006 b.001 b.001
1 Indicates developmental delay, impaired cognitive status, seizure disorder, cerebral palsy, structural central nervous system disorder, or neuromuscular disorder as defined by NSQIP-P.
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1243
Table 4 Observed vs Expected Readmission Events.
Derivation cohort Validation cohort
Overall patients
Observed readmission events
Expected readmission events
O/E ratio
Calibration p value
38,390 16,480
1336 594
1296 558
1.03 1.07
.95 .36
were also the most likely to lead to unplanned readmission in adult general surgery patients [5]. Though general pediatric surgical procedures including appendectomy and cholecystectomy had relatively low rates of readmission in this study compared to more complex procedures, the absolute number of children readmitted after procedures in the general category was greater. Berry et al. found appendectomy to be one of the top 10 readmission diagnoses among all hospitalized children [4]. Prevention efforts should be targeted toward both low frequency procedures with high readmission rates and high frequency procedures with low readmission rates to optimize reductions in readmission. Many of the predictive factors identified in the model were not modifiable including procedure type, preoperative renal failure, and wound type. Operative time and postoperative duration of stay are modifiable in some circumstances but most likely reflect complexity of care. This has two implications. First, penalization of institutions for pediatric surgical readmissions through public reporting or reduced reimbursement should be adjusted for patient complexity. Second, it corroborates the conclusion of Srivastan and Keren that there is insufficient evidence to declare most pediatric readmissions preventable and hold hospitals responsible for them [14]. Quality of care and inpatient services are only one part of a framework that includes outpatient access to and quality of care and socioeconomic status [15]. A pediatric surgery risk calculator was recently created using NSQIP-P information including demographics, comorbidities, and procedure variables to estimate risk of postoperative morbidity and mortality [16]. Readmission was not an outcome for this risk calculator but application of the calculator to prediction of readmission risk would be an important area of study in the future. Nonetheless, it is in the best interest of pediatric surgical patients to minimize unplanned readmissions. Prevention efforts can be targeted using risk factors identified in the predictive model and factors associated with high readmission rates such as hepatobiliary procedures, postoperative complications, and longer postoperative length of stay. Minimally invasive techniques should be utilized when appropriate, as they have been shown to reduce length of stay in many pediatric surgical procedures; however, the exact impact on readmission requires study as these techniques may also have longer operative times [17]. Standardized, disease-specific care pathways may also be useful in decreasing readmission. Implementation of such guidelines for appendicitis at a children's hospital was associated with decreased 30-day readmission [18]. In the adult colon and rectal surgery and hepatobiliary literature, standardized enhanced recovery pathways emphasizing early mobilization, oral intake, and non-opioid pain control have been created and are associated with decreased length of stay [19,20]. Postoperative complications are known to be relatively infrequent when compared to adult surgical patients however remained predictive of readmission in this study, similar to studies of readmission risk factors in adults [11]. Deep surgical site infections and pulmonary complications were most frequent and this study provides impetus for working to prevent both these complications as well as readmission once complications have occurred. NSQIP-P provides a platform upon which these complications can be better understood. Institutions that participate in NSQIP-P receive their complication and readmission rates benchmarked against other hospitals and those hospitals with relatively high rates can examine instances of readmission closely to direct prevention efforts and then ideally share their experiences. The role of family educational material, post-discharge telephone calls, and scheduled follow up in reducing unplanned readmissions should also be investigated in at-risk children [21,22].
This study is limited by the retrospective nature in which these data were used. Furthermore, the NSQIP-P is a convenience sample of participating hospitals. Despite validation of the NSQIP readmission methodology in the literature there was potential for misclassification of planned vs unplanned and related vs unrelated readmissions. Many potential limitations in studying readmissions including person–time bias and loss to follow up were addressed in this study by the person–days analysis and use of a data source which ascertains readmission at 30 days in all patients. Validation of our model using patients from the same data source may have over-estimated accuracy and we hope to validate the identified risk factors externally in the future. Institutional information was not available therefore rates by hospital could not be assessed. While efforts were made to categorize procedures in similar groups there may be variation in disease severity within a single group that remains unaccounted. Specific reasons for readmission would also be helpful in directing prevention but were missing or ambiguous in many patients. Unplanned visits to the emergency department and outpatient clinic following discharge from a surgical hospitalization are not collected in NSQIP-P and should be studied in the future similar to inpatient readmissions. Lastly, pediatric patients are largely dependent on their family, social, and economic environment. These factors were not studied. 4. Conclusion In this large multicenter cohort of general and thoracic pediatric surgical patients the rate of readmission was 4.3% per 30 person–days and ranged from 1.8 to 11.6% per 30 person–days based on type of surgical procedure. Several risk factors were identified which can be used to adjust for patient complexity when benchmarking hospitals or determining reimbursement in addition to informing prevention efforts. This study supports existing belief that postoperative readmission requires further investigation as a quality indicator in the pediatric realm while providing direction for such efforts [14]. Appendix 1. CPT procedure groups Anorectal: 46,705, 46,715, 46,716, 46,730, 46,742, 45,130, 45,400. General: 38,100, 38,120, 38,780, 43,520, 43,800, 44,950, 44,960, 44,970, 44,979, 47,562, 47,563, 47,600, 49,000, 49,010, 49,320, 49,321, 49,322, 49,324, 49,421, 38,101, 38,570, 47,564, 47,605, 49,020. Gynecology: 57,335, 58,661, 58,720, 58,925, 58,940, 58,943, 56,805, 58,700. Head and neck: 38,510, 60,280. Hepatobiliary: 47,100, 47,120, 47,130, 47,701, 47,715, 47,780, 37,181, 47,122, 47,125, 47,785, 48,140. Small and large intestine: 43,880, 44,005, 44,020, 44,050, 44,055, 44,120, 44,125, 44,130, 44,140, 44,141, 44,143, 44,144, 44,150, 44,158, 44,160, 44,180, 44,186, 44,187, 44,188, 44,202, 44,204, 44,205, 44,210, 44,211, 44,227, 44,300, 44,310, 44,314, 44,320, 44,340, 44,345, 44,620, 44,625, 44,626, 44,640, 44,800, 45,112, 45,113, 45,120, 45,121, 45,397, 45,499, 46,740, 44,110, 44,312, 44,322, 44,602, 44,615, 49,605, 49,600. Thoracic: 19,260, 21,742, 21,743, 32,096, 32,097, 32,100, 32,480, 32,505, 32,601, 32,606, 32,607, 32,608, 32,650, 32,651, 32,652, 32,655, 32,662, 32,663, 32,664, 32,666, 39,220, 39,503, 39,545, 32,484, 32,653, 32,669, 32,673. Upper gastrointestinal: 39,541, 43,279, 43,280, 43,281, 43,325, 43,327, 43,332, 43,333, 43,500, 43,653, 43,659, 43,830, 43,840, 43,870, 43,820, 43,831, 43,832.
1244
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
References [1] Centers for medicare and medicaid services. Hospital quality initiative. Outcome measures. 2015. [Updated:September 25, 2015. Accessed August 16, 2016 at https://www.Cms.Gov/medicare/quality-initiativespatient-assessment-instruments/hospitalqualityinits/outcomemeasures.Html]. [2] Toomey SL, Peltz A, Loren S, et al. Potentially preventable 30-day hospital readmissions at a children's hospital. Pediatrics 2016;138. [3] Agency for healthcare research and quality. Chipra measures by chipra categoriesinitial core set and pqmp coe measure assignemnts. 2012. [Updated May 31, 2012. Accessed August 16, 2016 at http://www.Ahrq.Gov/sites/default/files/wysiwyg/ policymakers/chipra/pqmpmeasures.Pdf]. [4] Berry JG, Toomey SL, Zaslavsky AM, et al. Pediatric readmission prevalence and variability across hospitals. JAMA 2013;309:372–80. [5] Lucas DJ, Haider A, Haut E, et al. Assessing readmission after general, vascular, and thoracic surgery using acs-nsqip. Ann Surg 2013;258:430–9. [6] Bruny JL, Hall BL, Barnhart DC, et al. American college of surgeons national surgical quality improvement program pediatric: a beta phase report. J Pediatr Surg 2013; 48:74–80. [7] Sellers MM, Merkow RP, Halverson A, et al. Validation of new readmission data in the american college of surgeons national surgical quality improvement program. J Am Coll Surg 2013;216:420–7. [8] Su AW, Habermann EB, Thomsen KM, et al. Risk factors for 30-day unplanned readmission and major perioperative complications following spine fusion surgery in adults: a review of the national surgical quality improvement program (nsqip) database. Spine 2016. [9] Donze J, Aujesky D, Williams D, et al. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA Intern Med 2013;173:632–8.
[10] Khan A, Nakamura MM, Zaslavsky AM, et al. Same-hospital readmission rates as a measure of pediatric quality of care. JAMA Pediatr 2015;169:905–12. [11] Glance LG, Kellermann AL, Osler TM, et al. Hospital readmission after noncardiac surgery: the role of major complications. JAMA Surg 2014;149:439–45. [12] Merkow RP, Ju MH, Chung JW, et al. Underlying reasons associated with hospital readmission following surgery in the United States. JAMA 2015;313:483–95. [13] Sinha CK, Decker E, Rex D, et al. Thirty-days readmissions in pediatric surgery: the first U.K. experience. J Pediatr Surg 2016;51:1870–80. [14] Srivastava R, Keren R. Pediatric readmissions as a hospital quality measure. JAMA 2013;309:396–8. [15] Kangovi S, Grande D. Hospital readmissions–not just a measure of quality. JAMA 2011;306:1796–7. [16] Kraemer K, Cohen ME, Liu Y, et al. Development and evaluation of the american college of surgeons nsqip pediatric surgical risk calculator. J Am Coll Surg 2016;223: 685–93. [17] Billingham MJ, Basterfield SJ. Pediatric surgical technique: laparoscopic or open approach? A systematic review and meta-analysis. Eur J Pediatr Surg 2010;20:73–7. [18] Slusher J, Bates CA, Johnson C, et al. Standardization and improvement of care for pediatric patients with perforated appendicitis. J Pediatr Surg 2014;49:1020–4 [discussion 1024-1025]. [19] Lawrence JK, Keller DS, Samia H, et al. Discharge within 24 to 72 hours of colorectal surgery is associated with low readmission rates when using enhanced recovery pathways. J Am Coll Surg 2013;216:390–4. [20] Bond-Smith G, Belgaumkar AP, Davidson BR, et al. Enhanced recovery protocols for major upper gastrointestinal, liver and pancreatic surgery. Cochrane Database Syst Rev 2016;2, CD011382. [21] Harrison PL, Hara PA, Pope JE, et al. The impact of postdischarge telephonic followup on hospital readmissions. Popul Health Manag 2011;14:27–32. [22] Wu S, Tyler A, Logsdon T, et al. A quality improvement collaborative to improve the discharge process for hospitalized children. Pediatrics 2016;138.
Contents lists available at ScienceDirect
Journal of Pediatric Surgery journal homepage: www.elsevier.com/locate/jpedsurg
Clinical Papers
Rates and risk factors of unplanned 30-day readmission following general and thoracic pediatric surgical procedures☆ Stephanie F. Polites a,⁎, Donald D. Potter a, Amy E. Glasgow b, Denise B. Klinkner a, Christopher R. Moir a, Michael B. Ishitani a, Elizabeth B. Habermann b a b
Division of Pediatric Surgery, Mayo Clinic, Rochester, MN Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, MN
a r t i c l e
i n f o
Article history: Received 14 September 2016 Received in revised form 10 November 2016 Accepted 28 November 2016 Key words: Readmission Pediatric surgery Quality
a b s t r a c t Background/Purpose: Postoperative unplanned readmissions are costly and decrease patient satisfaction; however, little is known about this complication in pediatric surgery. The purpose of this study was to determine rates and predictors of unplanned readmission in a multi-institutional cohort of pediatric surgical patients. Methods: Unplanned 30-day readmissions following general and thoracic surgical procedures in children b18 were identified from the 2012–2014 National Surgical Quality Improvement Program- Pediatric. Timedependent rates of readmission per 30 person–days were determined to account for varied postoperative length of stay (pLOS). Patients were randomly divided into 70% derivation and 30% validation cohorts which were used for creation and validation of a risk model for readmission. Results: Readmission occurred in 1948 (3.6%) of 54,870 children for a rate of 4.3% per 30 person–days. Adjusted predictors of readmission included hepatobiliary procedures, increased wound class, operative duration, complications, and pLOS. The predictive model discriminated well in the derivation and validation cohorts (AUROC 0.710 and 0.701) with good calibration between observed and expected readmission events in both cohorts (p N .05). Conclusions: Unplanned readmission occurs less frequently in pediatric surgery than what is described in adults, calling into question its use as a quality indicator in this population. Factors that predict readmission including type of procedure, complications, and pLOS can be used to identify at-risk children and develop prevention strategies. Level of evidence: III. © 2017 Elsevier Inc. All rights reserved.
Unplanned readmissions following surgical procedures are costly and disruptive to patients and their families. This complication is considered largely preventable and is being increasingly used as a quality indicator for hospitals and as a modifier for reimbursement. Initially, this was limited to certain adult procedures, as The Centers for Medicare and Medicaid Services (CMS) withhold up to 3% of reimbursement from hospitals with high readmission rates following coronary artery bypass grafting and joint replacement procedures [1]. However, it is now applicable to pediatric surgery as some states are imposing Medicaid and Children's Health Insurance (CHIP) reimbursement penalties for high readmission rates [2]. The CHIP Reauthorization Act of 2009 is in the Abbreviations: CMS, Centers for Medicare and Medicaid Services; CHIP, Children's Health Insurance; NACHRI, National Association of Children's Hospitals and Related Institutions; ACS, American College of Surgeons’; NSQIP-P, National Surgical Quality Improvement Program-Pediatric; GI, gastrointestinal; AUROC, area under the receiver operating characteristic. ☆ No external funding was secured for this study and the authors have no disclosures. ⁎ Corresponding author at: Department of Surgery, 200 1st Street SW, Rochester, MN 55901. Tel.: +1 507 255 5123. E-mail address: [email protected] (S.F. Polites). http://dx.doi.org/10.1016/j.jpedsurg.2016.11.043 0022-3468/© 2017 Elsevier Inc. All rights reserved.
process of creating hospital readmission quality measures [3]. Unfortunately the expansion of readmission penalties into the pediatric realm and use as a quality indicator has occurred in the absence of adequate evidence, as readmission following general and thoracic pediatric surgery is not well studied. Among more than 500,000 admissions at 72 National Association of Children's Hospitals and Related Institutions (NACHRI) hospitals, the all-cause unplanned readmission rate was 6.5% [4]. Another single institution study of all pediatric readmissions estimated nearly one third to be preventable [2]. In adults the readmission rate following general, vascular, and thoracic surgery in a national cohort was 7.8% and readmission was closely linked to the American Society of Anesthesiologists (ASA) class, length of stay (LOS), and postoperative complications [5]. It is unknown if these rates of readmission and risk factors can be extrapolated to pediatric general surgery. This information must be determined in order to inform reimbursement penalties and readmission prevention strategies. Thus, the purpose of this study was to determine the rate of unplanned and related postoperative readmissions in a national cohort of general and thoracic pediatric surgical patients and to determine predictors of readmission in this population.
1240
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1. Methods 1.1. Patient cohort and data source Children b18 years of age who underwent a general or thoracic surgical procedure were identified from the 2012–2014 American College of Surgeons' (ACS) National Surgical Quality Improvement ProgramPediatric (NSQIP-P) for inclusion in this study. Neonates who underwent operations b29 days after birth and those patients with missing data for postoperative length of stay were excluded. The NSQIP-P was designed to provide participating hospitals with riskadjusted performance measures [6]. A standardized 8-day sampling cycle of children who underwent inpatient or outpatient surgical procedures was used by dedicated clinical nurse abstracters to obtain data on more than 100 variables from patient medical records. Approximately 1400 cases were contributed annually from each institution. Surgical procedures were identified by current procedural terminology (CPT) code. In addition to demographic and procedural information, standardized definitions of comorbidities and postoperative complications were utilized. The follow-up period for NSQIP-P was 30 days from the index procedure. Any complications, readmissions, or death within this period were recorded. For patients who did not have follow up at the index hospital, telephone calls were made by NSQIP-P registrars to assess for these events. 1.2. Data collection and analysis CPT codes were used to identify patients who underwent a general or thoracic pediatric surgical procedure. Procedures were then categorized into one of eight groups—anorectal, general, gynecologic, head and neck, hepatobiliary, small and large intestine, thoracic, and upper gastrointestinal (GI). General included procedures such as appendectomy, cholecystectomy, and splenectomy (Appendix 1). The NSQIP-P collects information on up to five readmissions within the 30-day followup period and the NSQIP methodology for identifying readmissions has been validated as accurate [7]. Readmissions are classified as planned or unplanned and related or unrelated to the index procedures. Only unplanned, related readmissions were of interest and collected for this study. One reason for readmission is provided by ICD-9 diagnosis code. Demographic, comorbidity, operative, and postoperative information including complications were also collected for each patient. The unplanned readmission rate was determined for the entire cohort and for each procedure category. Since follow up in NSQIP is 30 days from the index procedure, patients had varied lengths at which they were at risk for readmission depending on their postoperative length of stay. For example, a patient discharged on postoperative day one had 29 days at risk for readmission while a patient discharged on postoperative day 16 had 14 days at risk. Patients who died postoperatively were also not at risk for readmission following death. Therefore, readmission rates per 30 person–days at risk with 95% confidence intervals for readmission were calculated based on methodology we have previously reported [8]. Only days following patients' discharge from their index procedure hospitalization which they were alive contributed to the denominator. The duration of readmission was not collected in NSQIP-P; therefore, each readmission was assumed to be 5 days based on previously reported methodology [8]. Patients still hospitalized at 30 days (n = 308), and those who died during their hospitalization or within 30 days from the index procedure (n = 4) were excluded. To identify and validate independent risk factors for readmission patients were randomly divided into a 70% derivation cohort and 30% validation cohort. Patients with at least one unplanned readmission related to the index surgical procedure were compared to non-readmitted patients in the derivation cohort using standard univariate analysis techniques. Only postoperative complications occurring prior to the readmission were included for readmitted patients. Continuous
variables were presented as mean (standard deviation) and categorical variables as n (percentage). Student's t tests were used to compare continuous variables and chi squared tests were used for categorical variables. Variables statistically significant on univariate analysis within the derivation cohort were used to create a stepwise Cox proportional hazards regression model for readmission using derivation cohort patients. Cox regression was used to account for varied postoperative length of day and subsequent days at risk for readmission within the 30-day follow-up period. Results were reported as hazard ratios (HR) with 95% confidence intervals and p values. Consistent with methods previously described in the literature by Donzé et al., β coefficients of significant variables were applied to the validation set [9]. Discriminatory power of both derivation and validation set models was assessed using area under the receiver operating characteristic (AUROC) analysis. Observed and expected readmission events were determined for both derivation and validation cohorts using Poisson regression and calibration testing was performed. Statistical significance was acknowledged when p b .05. Statistical analyses were performed using SAS (©SAS Institute, Cary, NC). 2. Results 2.1. Cohort characteristics Of 54,870 pediatric surgical patients included, the majority underwent general surgery procedures (61.1%) followed by upper GI (14.5%) and small and large intestine procedures (12.9%; Table 1). The mean age was 8.1 (5.8) years and 56.9% of patients were male. Outpatient procedures were performed in 14,440 patients (26.3%) and 28,166 (51.3%) underwent urgent or emergent procedures. Comorbidities were present in 28,154 (51.3%) patients. Median length of stay was 2.0 (1.0–5.0) days and 1819 (3.3%) of patients had postoperative complications. 2.2. Readmission rates and reasons Readmission occurred in 1948 (3.6%) patients with a resulting timedependent rate of 4.3% per 30 person–days for the overall cohort. The readmission rate was highest in patients who underwent hepatobiliary procedures (11.6% per 30 person–days) followed by small and large intestine (9.5% per 30 person–days) and anorectal procedures (5.3% per 30 person–days; Fig. 1). Patients who suffered complications also had higher readmission rates (15.8% per 30 person–days) than those without complications (4.0% per 30 person–days). Individual complications were infrequent and complications were subsequently analyzed as a dichotomous variable (Table 2). The ICD-9 diagnosis code reason for readmission was missing in 759 (38.5%) readmitted patients and the remaining readmitted patients had one of 298 unique diagnosis codes as their reason for readmission. The most frequent categories of reasons for readmission were bowel obstruction or constipation (11.8%), infection (7.8%), oral intolerance or dehydration (7.0%), and pain (6.7%). 2.3. Characteristics of readmitted patients There were no significant differences in age, sex, or race between readmitted and non-readmitted patients in the derivation cohort (n = 38,397; Table 1). Readmission was more frequent in patients with inpatient procedures, existing comorbidities, weight percentile b 5%, greater ASA class, increased operative duration, and greater wound contamination. While 2.5% of patients who underwent operations b60 min in duration were readmitted, readmission occurred in 6.9% of those who underwent operations N140 min (p b .001). Those with a postoperative complication and those with increased postoperative length of stay were also more frequently readmitted than those without, respectively.
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1241
Table 1 Characteristics of All Patients and Readmitted vs Non-readmitted Patients in the Derivation Set. Patient characteristic
All patients N = 54,870
Age, mean (SD), years Female sex Race
White Unknown Black Other General Upper GI Small/Large Intestine Thoracic Gynecology Head and Neck Hepatobiliary Anorectal Inpatient Outpatient b5% 5–94% ≥95%
Procedure group1
Procedure type Weight percentage2
Premature birth3 Congenital malformation Any comorbidity Individual comorbidities
Diabetes mellitus Respiratory GI/Hepatobiliary Major cardiac Acute renal failure Neurologic Immunosuppression Nutrition Hematologic Active malignancy
Preoperative SIRS, sepsis, septic shock Preoperative transfusion or hematocrit b32 Case status ASA class
Operative time, minutes
Wound class
Preoperative length of stay, days Postoperative complication
4
Postoperative length of stay, days
Elective Emergent/urgent 1 2 3 4/5 b60 60–140 N140 Clean Clean/Contaminated Contaminated Dirty/Infected b2 ≥2 Yes No b2 2–4 N4
8.1 (5.8) 23,634 (43.1%) 40,089 (73.1%) 6647 (12.1%) 6444 (11.7%) 1690 (3.1%) 33,516 (61.1%) 7982 (14.5%) 7081 (12.9%) 3397 (6.2%) 1105 (2.0%) 696 (1.3%) 606 (1.1%) 487 (0.9%) 40,430 (73.7%) 14,440 (26.3%) 8865 (16.2%) 39,051 (71.2%) 6954 (12.7%) 4956 (9.0%) 9544 (17.4%) 28,154 (51.3%) 343 (0.6%) 7327 (13.4%) 7327 (13.4%) 2148 (3.9%) 153 (0.3%) 7707 (14.0%) 2265 (4.1%) 6429 (11.7%) 6032 (11.0%) 1458 (2.7%) 9621 (17.5%) 582 (1.1%) 26,704 (48.7%) 28,166 (51.3%) 16,590 (30.2%) 24,710 (45.0%) 12,515 (22.8%) 1055 (1.9%) 31,966 (58.3%) 17,083 (31.1%) 5821 (10.6%) 10,962 (20.0%) 25,196 (45.9%) 12,261 (22.3%) 6451 (11.8%) 48,678 (88.7%) 6192 (11.3%) 53,051 (96.7%) 1819 (3.3%) 28,486 (51.9%) 14,766 (26.9%) 11,618 (21.2%)
70% Derivation cohort n = 38,397 No readmission n = 37,061
Readmission n = 1336
8.1 (5.8) 15,951 (96.5%) 27,110 (96.5%) 4452 (96.8%) 4347 (96.8%) 1152 (95.8%) 22,812 (97.2%) 5413 (96.7%) 4600 (93.1%) 2312 (97.3%) 742 (98.1%) 455 (95.8%) 407 (91.3%) 320 (94.4%) 27,097 (96.0%) 9964 (98.1%) 5990 (96.1%) 26,300 (96.5%) 4771 (97.1%) 3337 (96.3%) 6387 (95.5%) 18,867 (95.9%) 241 (94.5%) 4948 (96.3%) 12,027 (95.5%) 1437 (95.2%) 93 (90.3%) 5082 (95.4%) 1497 (94.5%) 4241 (95.1%) 4024 (95.4%) 915 (94.6%) 6436 (95.2%) 378 (95.5%) 18,017 (96.3%) 19,044 (96.7%) 11,265 (97.1%) 16,716 (96.7%) 8392 (95.6%) 688 (94.4%) 21,759 (97.5%) 11,497 (95.8%) 3805 (93.1%) 7485 (97.4%) 17,021 (96.6%) 8421 (97.3%) 4134 (93.0%) 32,918 (96.6%) 4143 (95.8%) 35,868 (96.7%) 1193 (91.3%) 19,652 (98.4%) 9882 (95.8%) 7527 (93.0%)
8.2 (5.8) 574 (3.5%) 994 (3.5%) 148 (3.2%) 144 (3.2%) 50 (4.2%) 653 (2.8%) 184 (3.3%) 343 (6.9%) 64 (2.7%) 14 (1.9%) 20 (4.2%) 39 (8.7%) 19 (5.6%) 1143 (4.0%) 193 (1.9%) 246 (3.9%) 949 (3.5%) 141 (2.9%) 128 (3.7%) 304 (4.5%) 800 (4.1%) 14 (5.5%) 189 (3.7%) 562 (4.5%) 73 (4.8%) 10 (9.7%) 244 (4.6%) 87 (5.5%) 219 (4.9%) 195 (4.6%) 52 (5.4%) 324 (4.8%) 18 (4.5%) 683 (3.7%) 653 (3.3%) 331 (2.9%) 575 (3.3%) 389 (4.4%) 41 (5.6%) 547 (2.5%) 506 (4.2%) 283 (6.9%) 198 (2.6%) 592 (3.4%) 236 (2.7%) 310 (7.0%) 1155 (3.4%) 181 (4.2%) 1222 (3.3%) 114 (8.7%) 330 (1.7%) 437 (4.2%) 569 (7.0%)
p value .39 .98 .28
b.001
b.001 .009
.75 b.001 b.001 .09 .39 b.001 .003 .003 b.001 b.001 b.001 b.001 .005 b.001 .27 .08 b.001
b.001
b.001
.008 b.001 b.001
1 Percentages of patients readmitted by surgical procedure group are intended to describe the cohort and are not adjusted for length of stay in contrast to person–day rates of readmission by surgical procedure group in Fig. 1. 2 Weight percentages obtained by weight and sex per CDC guidelines. 3 Premature birth defined as ≤36 weeks gestation. 4 Includes any postoperative complication. Individual complications are outlined in Table 2.
2.4. Independent predictors of readmission Twenty variables were initially entered into the model which discriminated well in the derivation cohort; AUROC 0.710, 95% CI 0.694–0.726 (Table 3). After adjustment, children who underwent head and neck, hepatobiliary, and small and large intestine operations remained at increased risk for readmission compared to general procedures while thoracic procedures carried the lowest risk (HR = 0.69, p = .009). Preoperative acute renal failure, neurologic comorbidity, preoperative systemic infection, wound classification, operative duration, and postoperative complications were also associated with readmission. The
strongest predictor of readmission was increased postoperative length of stay. Children hospitalized for N 4 days following their index procedure were more than 3 times as likely to be readmitted than children discharged within 2 days (HR = 3.12, p b .001). 2.5. Validation of prediction model The model discriminated well when applied to the validation cohort (AUROC 0.701, 95% CI 0.677–0.724), as the c statistic for the validation cohort remained within the confidence interval of the derivation cohort. Observed and expected readmission events were then determined and
1242
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
16%
12%
11.6% 9.5%
8% 5.3% 4%
5.0%
4.3%
3.7%
3.3% 1.8%
0%
Fig. 1. Readmission rates per 30 person–days by procedure category.
the O/E ratios were 1.03 and 1.07 in the derivation and validation cohorts, respectively, with good calibration (derivation p = 0.95; validation p = 0.36; Table 4). 3. Discussion In this study readmission rates were determined for general and thoracic pediatric surgical patients that account for varied postoperative length of stay and opportunity for readmission using a multiinstitutional data source that captures all readmissions within 30 days. Simple predictors for readmission were also identified that may be used to identify at-risk patients and direct efforts to reduce readmission accordingly. These results will help to inform policies and benchmarking related to readmission. Existing efforts are hampered by small sample sizes and inability to ascertain readmissions that did not occur at the index hospital [10]. The overall unplanned readmission rate per 30 person–days of 4.3% in this study is lower than the 5–8% described for adult general surgery patients using a similar data source as well as the 6.5% reported for all pediatric admissions at 72 children's hospitals [4,5,11,12]. While there are no large scale studies of readmission following general and thoracic
pediatric surgery specifically in the United States to serve as a reference, a readmission rate of 2% was identified in a United Kingdom study of more than 2000 pediatric surgery procedures [13]. Based on existing literature, unplanned readmission appears to be less frequent an occurrence following general pediatric surgical procedures than adult procedures, calling into question the use of readmission as a marker of quality in pediatric surgery. This determination would be aided by institution-specific pediatric general and thoracic surgery readmission rates which are not available in the NSQIP-P. Participating hospitals, however, do receive their readmission data and may compare this to the national rate described in this study to direct quality improvement. Consistent with existing literature, readmission rates varied by procedure type in this study and the highest rates were in children who underwent hepatobiliary and intestinal procedures. These procedures
Table 3 Multivariable Cox Proportional Hazard Regression Analysis Predicting Readmission in 70% Derivation Cohort. Covariate Procedure group (vs general)
Table 2 Complications in Readmitted vs Non-Readmitted Patients. Total (N = 54,870)
% Not readmitted (n = 52,922)
% Readmitted (n = 1948)
p value
Any complication/postoperative ocurrence1 No 53,051 (96.7%) 51,259 (96.6%) Yes 1819 (3.3%) 1663 (91.4%)
1792 (3.4%) 156 (8.6%)
b.001
Superficial Incisional SSI No 54,736 (99.8%) Yes 134 (0.2%)
52,797 (96.5%) 125 (93.3%)
1939 (3.5%) 9 (6.7%)
.05
Deep SSI2 No 54,425 (99.2%) Yes 445 (0.8%)
52,526 (96.5%) 396 (89.0%)
1899 (3.5%) 49 (11.0%)
b.001
Pulmonary complications No 54,663 (99.6%) Yes 207 (0.4%)
52,729 (96.5%) 193 (93.2%)
1934 (3.5%) 14 (6.8%)
.012
Sepsis/CL associated bloodstream infection No 54,691 (99.7%) 52,759 (96.5%) Yes 179 (0.3%) 163 (91.1%)
1932 (3.5%) 16 (8.9%)
b.001
SSI = surgical site infection. CL = central line. 1 Includes all postoperative occurrences and complications in NSQIP-P. Only the most frequent and clinically relevant are displayed subsequently in this table. 2 Deep SSI includes deep incisional and organ space infections.
Anorectal Gynecology Head and Neck Hepatobiliary Small and Large Intestine Thoracic Upper GI
Preoperative acute renal failure Neurologic comorbidity1 SIRS/Sepsis/Septic Shock within 48 h prior to index procedure Wound class Clean/Contaminated (vs clean) Contaminated Dirty/Infected Operative time 60–140 min (vs b60 min) N140 min Any complication Postoperative 2–4 days length of stay N4 days (vs b2 days)
β Hazard LCL Coefficient ratio
UCL
−0.46 −0.73 0.41 0.06 −0.001
1.59 0.77 2.40 1.69 1.59
0.98 0.45 1.48 1.17 1.33
2.58 .06 1.32 .34 3.89 b.001 2.44 .005 1.90 b.001
−0.84 −0.37 0.91
0.69 1.10 2.47
0.52 0.91 0.89 1.36 1.31 4.66
.009 .38 .005
0.18
1.30
1.05 1.62
.029
0.18
1.20
1.02 1.41
.033
0.08 0.26 0.65 0.19 0.41 0.29 0.78 1.14
1.08 1.29 1.92 1.21 1.51 1.34 2.17 3.12
0.91 1.05 1.53 1.06 1.26 1.09 1.84 2.60
1.29 1.60 2.40 1.39 1.81 1.65 2.57 3.74
p value
.36 .018 b.001 .005 b.001 .006 b.001 b.001
1 Indicates developmental delay, impaired cognitive status, seizure disorder, cerebral palsy, structural central nervous system disorder, or neuromuscular disorder as defined by NSQIP-P.
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
1243
Table 4 Observed vs Expected Readmission Events.
Derivation cohort Validation cohort
Overall patients
Observed readmission events
Expected readmission events
O/E ratio
Calibration p value
38,390 16,480
1336 594
1296 558
1.03 1.07
.95 .36
were also the most likely to lead to unplanned readmission in adult general surgery patients [5]. Though general pediatric surgical procedures including appendectomy and cholecystectomy had relatively low rates of readmission in this study compared to more complex procedures, the absolute number of children readmitted after procedures in the general category was greater. Berry et al. found appendectomy to be one of the top 10 readmission diagnoses among all hospitalized children [4]. Prevention efforts should be targeted toward both low frequency procedures with high readmission rates and high frequency procedures with low readmission rates to optimize reductions in readmission. Many of the predictive factors identified in the model were not modifiable including procedure type, preoperative renal failure, and wound type. Operative time and postoperative duration of stay are modifiable in some circumstances but most likely reflect complexity of care. This has two implications. First, penalization of institutions for pediatric surgical readmissions through public reporting or reduced reimbursement should be adjusted for patient complexity. Second, it corroborates the conclusion of Srivastan and Keren that there is insufficient evidence to declare most pediatric readmissions preventable and hold hospitals responsible for them [14]. Quality of care and inpatient services are only one part of a framework that includes outpatient access to and quality of care and socioeconomic status [15]. A pediatric surgery risk calculator was recently created using NSQIP-P information including demographics, comorbidities, and procedure variables to estimate risk of postoperative morbidity and mortality [16]. Readmission was not an outcome for this risk calculator but application of the calculator to prediction of readmission risk would be an important area of study in the future. Nonetheless, it is in the best interest of pediatric surgical patients to minimize unplanned readmissions. Prevention efforts can be targeted using risk factors identified in the predictive model and factors associated with high readmission rates such as hepatobiliary procedures, postoperative complications, and longer postoperative length of stay. Minimally invasive techniques should be utilized when appropriate, as they have been shown to reduce length of stay in many pediatric surgical procedures; however, the exact impact on readmission requires study as these techniques may also have longer operative times [17]. Standardized, disease-specific care pathways may also be useful in decreasing readmission. Implementation of such guidelines for appendicitis at a children's hospital was associated with decreased 30-day readmission [18]. In the adult colon and rectal surgery and hepatobiliary literature, standardized enhanced recovery pathways emphasizing early mobilization, oral intake, and non-opioid pain control have been created and are associated with decreased length of stay [19,20]. Postoperative complications are known to be relatively infrequent when compared to adult surgical patients however remained predictive of readmission in this study, similar to studies of readmission risk factors in adults [11]. Deep surgical site infections and pulmonary complications were most frequent and this study provides impetus for working to prevent both these complications as well as readmission once complications have occurred. NSQIP-P provides a platform upon which these complications can be better understood. Institutions that participate in NSQIP-P receive their complication and readmission rates benchmarked against other hospitals and those hospitals with relatively high rates can examine instances of readmission closely to direct prevention efforts and then ideally share their experiences. The role of family educational material, post-discharge telephone calls, and scheduled follow up in reducing unplanned readmissions should also be investigated in at-risk children [21,22].
This study is limited by the retrospective nature in which these data were used. Furthermore, the NSQIP-P is a convenience sample of participating hospitals. Despite validation of the NSQIP readmission methodology in the literature there was potential for misclassification of planned vs unplanned and related vs unrelated readmissions. Many potential limitations in studying readmissions including person–time bias and loss to follow up were addressed in this study by the person–days analysis and use of a data source which ascertains readmission at 30 days in all patients. Validation of our model using patients from the same data source may have over-estimated accuracy and we hope to validate the identified risk factors externally in the future. Institutional information was not available therefore rates by hospital could not be assessed. While efforts were made to categorize procedures in similar groups there may be variation in disease severity within a single group that remains unaccounted. Specific reasons for readmission would also be helpful in directing prevention but were missing or ambiguous in many patients. Unplanned visits to the emergency department and outpatient clinic following discharge from a surgical hospitalization are not collected in NSQIP-P and should be studied in the future similar to inpatient readmissions. Lastly, pediatric patients are largely dependent on their family, social, and economic environment. These factors were not studied. 4. Conclusion In this large multicenter cohort of general and thoracic pediatric surgical patients the rate of readmission was 4.3% per 30 person–days and ranged from 1.8 to 11.6% per 30 person–days based on type of surgical procedure. Several risk factors were identified which can be used to adjust for patient complexity when benchmarking hospitals or determining reimbursement in addition to informing prevention efforts. This study supports existing belief that postoperative readmission requires further investigation as a quality indicator in the pediatric realm while providing direction for such efforts [14]. Appendix 1. CPT procedure groups Anorectal: 46,705, 46,715, 46,716, 46,730, 46,742, 45,130, 45,400. General: 38,100, 38,120, 38,780, 43,520, 43,800, 44,950, 44,960, 44,970, 44,979, 47,562, 47,563, 47,600, 49,000, 49,010, 49,320, 49,321, 49,322, 49,324, 49,421, 38,101, 38,570, 47,564, 47,605, 49,020. Gynecology: 57,335, 58,661, 58,720, 58,925, 58,940, 58,943, 56,805, 58,700. Head and neck: 38,510, 60,280. Hepatobiliary: 47,100, 47,120, 47,130, 47,701, 47,715, 47,780, 37,181, 47,122, 47,125, 47,785, 48,140. Small and large intestine: 43,880, 44,005, 44,020, 44,050, 44,055, 44,120, 44,125, 44,130, 44,140, 44,141, 44,143, 44,144, 44,150, 44,158, 44,160, 44,180, 44,186, 44,187, 44,188, 44,202, 44,204, 44,205, 44,210, 44,211, 44,227, 44,300, 44,310, 44,314, 44,320, 44,340, 44,345, 44,620, 44,625, 44,626, 44,640, 44,800, 45,112, 45,113, 45,120, 45,121, 45,397, 45,499, 46,740, 44,110, 44,312, 44,322, 44,602, 44,615, 49,605, 49,600. Thoracic: 19,260, 21,742, 21,743, 32,096, 32,097, 32,100, 32,480, 32,505, 32,601, 32,606, 32,607, 32,608, 32,650, 32,651, 32,652, 32,655, 32,662, 32,663, 32,664, 32,666, 39,220, 39,503, 39,545, 32,484, 32,653, 32,669, 32,673. Upper gastrointestinal: 39,541, 43,279, 43,280, 43,281, 43,325, 43,327, 43,332, 43,333, 43,500, 43,653, 43,659, 43,830, 43,840, 43,870, 43,820, 43,831, 43,832.
1244
S.F. Polites et al. / Journal of Pediatric Surgery 52 (2017) 1239–1244
References [1] Centers for medicare and medicaid services. Hospital quality initiative. Outcome measures. 2015. [Updated:September 25, 2015. Accessed August 16, 2016 at https://www.Cms.Gov/medicare/quality-initiativespatient-assessment-instruments/hospitalqualityinits/outcomemeasures.Html]. [2] Toomey SL, Peltz A, Loren S, et al. Potentially preventable 30-day hospital readmissions at a children's hospital. Pediatrics 2016;138. [3] Agency for healthcare research and quality. Chipra measures by chipra categoriesinitial core set and pqmp coe measure assignemnts. 2012. [Updated May 31, 2012. Accessed August 16, 2016 at http://www.Ahrq.Gov/sites/default/files/wysiwyg/ policymakers/chipra/pqmpmeasures.Pdf]. [4] Berry JG, Toomey SL, Zaslavsky AM, et al. Pediatric readmission prevalence and variability across hospitals. JAMA 2013;309:372–80. [5] Lucas DJ, Haider A, Haut E, et al. Assessing readmission after general, vascular, and thoracic surgery using acs-nsqip. Ann Surg 2013;258:430–9. [6] Bruny JL, Hall BL, Barnhart DC, et al. American college of surgeons national surgical quality improvement program pediatric: a beta phase report. J Pediatr Surg 2013; 48:74–80. [7] Sellers MM, Merkow RP, Halverson A, et al. Validation of new readmission data in the american college of surgeons national surgical quality improvement program. J Am Coll Surg 2013;216:420–7. [8] Su AW, Habermann EB, Thomsen KM, et al. Risk factors for 30-day unplanned readmission and major perioperative complications following spine fusion surgery in adults: a review of the national surgical quality improvement program (nsqip) database. Spine 2016. [9] Donze J, Aujesky D, Williams D, et al. Potentially avoidable 30-day hospital readmissions in medical patients: derivation and validation of a prediction model. JAMA Intern Med 2013;173:632–8.
[10] Khan A, Nakamura MM, Zaslavsky AM, et al. Same-hospital readmission rates as a measure of pediatric quality of care. JAMA Pediatr 2015;169:905–12. [11] Glance LG, Kellermann AL, Osler TM, et al. Hospital readmission after noncardiac surgery: the role of major complications. JAMA Surg 2014;149:439–45. [12] Merkow RP, Ju MH, Chung JW, et al. Underlying reasons associated with hospital readmission following surgery in the United States. JAMA 2015;313:483–95. [13] Sinha CK, Decker E, Rex D, et al. Thirty-days readmissions in pediatric surgery: the first U.K. experience. J Pediatr Surg 2016;51:1870–80. [14] Srivastava R, Keren R. Pediatric readmissions as a hospital quality measure. JAMA 2013;309:396–8. [15] Kangovi S, Grande D. Hospital readmissions–not just a measure of quality. JAMA 2011;306:1796–7. [16] Kraemer K, Cohen ME, Liu Y, et al. Development and evaluation of the american college of surgeons nsqip pediatric surgical risk calculator. J Am Coll Surg 2016;223: 685–93. [17] Billingham MJ, Basterfield SJ. Pediatric surgical technique: laparoscopic or open approach? A systematic review and meta-analysis. Eur J Pediatr Surg 2010;20:73–7. [18] Slusher J, Bates CA, Johnson C, et al. Standardization and improvement of care for pediatric patients with perforated appendicitis. J Pediatr Surg 2014;49:1020–4 [discussion 1024-1025]. [19] Lawrence JK, Keller DS, Samia H, et al. Discharge within 24 to 72 hours of colorectal surgery is associated with low readmission rates when using enhanced recovery pathways. J Am Coll Surg 2013;216:390–4. [20] Bond-Smith G, Belgaumkar AP, Davidson BR, et al. Enhanced recovery protocols for major upper gastrointestinal, liver and pancreatic surgery. Cochrane Database Syst Rev 2016;2, CD011382. [21] Harrison PL, Hara PA, Pope JE, et al. The impact of postdischarge telephonic followup on hospital readmissions. Popul Health Manag 2011;14:27–32. [22] Wu S, Tyler A, Logsdon T, et al. A quality improvement collaborative to improve the discharge process for hospitalized children. Pediatrics 2016;138.