Contemporary Trends in Laparoscopy and Ovarian Sparing Surgery for Ovarian Torsion in the Pediatric Population

PURPOSE

Although total oophorectomy (TO) was historically performed in cases of nonviable-appearing ovaries, considerable evidence has demonstrated equivalent outcomes after ovarian sparing surgery (OSS) as well as long-term fertility preservation benefits. This study sought to compare outcomes of OSS and TO for patients with ovarian torsion.

METHODS

Females <21 years old admitted for ovarian torsion were identified from the Nationwide Readmissions Database (2016-2018) and stratified by OSS or TO. Propensity score-matched analysis (PSMA) utilizing >50 covariates (demographics, medical comorbidities, ovarian diagnoses, etc.) was constructed between those receiving TO and OSS.

RESULTS

There were 3,161 females (median 15 [12-18] years) with ovarian torsion, and concomitant pathologies included cysts (42%), benign masses (25%), and malignant masses (<1%). Open approaches were more common (52% vs. 48% laparoscopic), and ovarian resection (OSS or TO) was performed in 87% (39% OSS and 48% TO). OSS was more commonly performed with laparoscopic detorsions (60% vs. 40% TO), while TO was more frequent in open operations (59% vs. 41% TO; both p < 0.001). No differences in overall readmissions (7% OSS vs. 8% TO) or readmissions for recurrent torsion (<1% overall) and ovarian masses (<1%) were observed (both groups <1%; p = 0.612). After PSMA, laparoscopy was still utilized less frequently with TO (39% vs. 53%; p < 0.001) despite similar rates of malignant masses.

CONCLUSIONS

Overall, these data offer additional support for the current practice guidelines that give preference to OSS as the primary method of treatment for pediatric ovarian torsion in the majority of cases.

LEVEL OF EVIDENCE

III.

TYPE OF STUDY

Retrospective Comparative Study.

Primary Spontaneous Pneumothorax Outcomes in Children: A National Analysis

OBJECTIVE

Considerable variation in primary spontaneous pneumothorax (PSP) management exists in the pediatric population. This study aims to compare nationwide outcomes of children with PSP.

METHODS

The Nationwide Readmissions Database (2016 to 2018) was used to identify patients 1 to 18 years old with PSP. Trauma, secondary pneumothoraces, and elective admissions were excluded. Demographics and complications were compared among patients undergoing initial nonoperative management (NOM; observation or percutaneous drainage) or operative resection using standard statistical tests.

RESULTS

A total of 3,890 patients were identified with PSP (median age, 16 [interquartile range 14 to 17] years). Most (78%) underwent NOM, of which 17% failed requiring operative resection. Of the intent-to-treat cohort, 28% failed NOM during index admission or required repeat percutaneous drainage or operative resection on readmission. Patients treated by NOM had higher 30-day and overall readmission rates compared with operative resection (all P < 0.001). Ipsilateral recurrent pneumothorax was higher in those receiving NOM (13% vs 3%, P < 0.001). Patients from the lowest median household income quartile more frequently received NOM compared with the highest income quartile (82% vs 76%) with more readmissions.

CONCLUSIONS

Patients with PSP who underwent initial NOM experienced higher readmission rates than those receiving operative resection. Furthermore, socioeconomic status was associated with the utilization of nonoperative versus operative management.

Operative versus percutaneous drainage with fibrinolysis for complicated pediatric pleural effusions: A nationwide analysis

PURPOSE

Management of complicated pleural effusions and empyema using tube thoracostomy with intrapleural fibrinolysis versus surgical drainage has been debated for decades. However, there remains considerable variation in management with these approaches in the pediatric population. This study aims to compare the nationwide outcomes of pediatric patients with complicated pleural effusions.

METHODS

Patients <18 years old with a diagnosis of pleural effusion or empyema associated with pneumonia were identified from the Nationwide Readmissions Database (2016-2018). Demographics, hospital characteristics, and complications were compared among patients undergoing isolated percutaneous drainage (PD), percutaneous drainage with intrapleural fibrinolysis (PDF), or operative drainage (OD) using standard statistical tests.

RESULTS

5424 patients (age 4 [IQR 1-11] years) were identified with a pleural effusion or empyema who underwent percutaneous or surgical intervention. PD (22%) and OD (24%) were utilized more frequently than PDF (3%). Index complications, including bleeding and postprocedural air leak, were similar between groups. Those receiving PDF had lower index length of stay (LOS) and admission costs. Thirty-day and overall readmission rates were highest in patients receiving PD (15% and 24%) and OD (12% and 23%) versus PDF, all p < 0.001. Those receiving OD had fewer readmission complications including recurrent effusion or empyema, pneumonia, and bleeding. Overall readmission cost was highest in those receiving PD (p = 0.005).

CONCLUSION

In this nationwide cohort, PDF was associated with lower index admission cost, shorter LOS and lower rates of readmissions compared to OD. This knowledge should be used to improve selection of these treatments in this patient population.

TYPE OF STUDY

Retrospective Comparative LEVEL OF EVIDENCE: III.

Nationwide Analysis of Hospital admissions Prior to Hartmann's Procedure for Acute Diverticulitis

INTRODUCTION

Diverticular disease is one of the most common gastrointestinal diseases that require hospital admission. This study aims to identify trends in prior hospital admissions for patients that ultimately require a Hartmann's procedure for complicated diverticulitis.

METHODS

The Nationwide Readmissions Database for 2010-2014 was queried for all patients aged 18 years or older admitted with an ICD-9 code for colonic diverticulitis and end colostomy creation. Patients with prior hospital admissions were identified. The primary outcome was mortality after Hartmann's procedure. Secondary outcomes were prior hospital admission and previous percutaneous drain placement. Multivariable logistic regression was performed to control for confounding factors for each outcome and results were weighted for national estimates.

RESULTS

There were 90,162 patients admitted with complicated diverticulitis requiring end colostomy creation. Prior hospital admissions were found in 28.1% (n = 25,307) and 14.4% (n = 12,947) had a previous percutaneous drain placed during a prior admission. The overall mortality rate was 5.9% (n = 5314) after Hartman's procedure. The mortality rate for patients with prior hospital admissions was 8.7% (P < .001), and the mortality rate for patients with previous percutaneous drain placement was 4.3% (P < .001). After controlling for confounding factors including comorbidities, patients with prior admission had an increased risk of mortality (OR 1.48 [1.40-1.58], P < .001) and patients with previous percutaneous drain placement had a decreased risk of mortality (OR .66 [.60-.72], P < .001).

CONCLUSIONS

Hospitalizations for complications of diverticulitis are a costly burden to our healthcare system. By identifying those patients at high risk for readmission and emergency surgery, perioperative outcomes may be improved.

Complications while awaiting elective inguinal hernia repair in infants: Not as common as you thought

BACKGROUND

The dogma of early inguinal hernia repair in infants, especially those born prematurely, has dominated clinical practice owing to reports of a high frequency of incarceration and significant complications associated with untreated inguinal hernias. We aim to evaluate the frequency of complications after discharge with delayed surgery for inguinal hernia repair.

METHODS

The Nationwide Readmissions Database (2010-2014) was queried to identify infants diagnosed with inguinal hernia. We compared the frequency and characteristics of inguinal hernia repair performed during the index admission, discharge from the index admission without hernia repair, and unplanned readmissions.

RESULTS

We identified 33,530 infants (16,624 preterm and 16,906 full-term) diagnosed with an inguinal hernia during an index admission. For those infants diagnosed with an inguinal hernia at birth, inguinal hernia repair was performed during the birth admission for only a minority of both preterm (35%) and full-term infants (18%; P < .001). Of the infants discharged without hernia repair, 15% required nonelective readmission up to 1 year later, but only 2% of preterm and 1% of full-term infants actually underwent inguinal hernia repair during these unplanned readmissions. None of the readmitted infants underwent additional procedures suggestive of a strangulated hernia.

CONCLUSION

Complications among infants awaiting inguinal hernia repair may be substantially less common than previously reported, and the occurrence of significant associated morbidity is quite rare.

Hospital Readmissions for Hyperparathyroidism After Bariatric Surgery in the United States: A National Database Review

Introduction: The incidence and significance of hyperparathyroidism in patients after bariatric surgery have been established to some degree; however, the impact it has on the national healthcare system has not. We sought to assess the risk of readmission and related comorbidities in this patient population.

Methods: The Healthcare Cost and Utilization Project Nationwide Readmission Database was queried for all patients who underwent Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG). Multivariate logistic regression analysis was conducted to identify factors associated with readmission for hyperparathyroidism.

Results: A total of 915,792 patients between 2010 and 2015 were queried; 43.2% had undergone SG and 56.8% had RYGB. A total of 589 patients were readmitted for hyperparathyroidism; 80.8% were female and 68% had a Charlson comorbidity index ≥ 2. Factors associated with readmission were as follows: age 45-64 years (odds ratio [OR] 1.42, p=0.001), Medicare (OR 3.01, p<0.001) or Medicaid (OR 2.61, p<0.001) insurance status, lower median household income, renal failure (OR 17.14, p<0.001), hypertension (OR 2.89, p<0.001), and deficiency anemia (OR 2.62, p<0.01).

Conclusions: Parathyroid axis monitoring may provide benefits to predictably high-risk patients. Appropriate surveillance may decrease the impact of bariatric hyperparathyroidism readmission on the U.S. healthcare system.

Comparing Machine Learning Algorithms for Predicting Acute Kidney Injury

Prior studies have used vital signs and laboratory measurements with conventional modeling techniques to predict acute kidney injury (AKI). The purpose of this study was to use the trend in vital signs and laboratory measurements with machine learning algorithms for predicting AKI in ICU patients. The eICU Collaborative Research Database was queried for five consecutive days of laboratory measurements per patient. Patients with AKI were identified and trends in vital signs and laboratory values were determined by calculating the slope of the least-squares-fit linear equation using three days for each value. Different machine learning classifiers (gradient boosted trees [GBT], logistic regression, and deep learning) were trained to predict AKI using the laboratory values, vital signs, and slopes. There were 151,098 ICU stays identified and the rate of AKI was 5.6 per cent. The best performing algorithm was GBT with an AUC of 0.834 ± 0.006 and an F-measure of 42.96 per cent ± 1.26 per cent. Logistic regression performed with an AUC of 0.827 ± 0.004 and an F-measure of 28.29 per cent ± 1.01 per cent. Deep learning performed with an AUC of 0.817 ± 0.005 and an F-measure of 42.89 per cent ± 0.91 per cent. The most important variable for GBT was the slope of the minimum creatinine (30.32%). This study identifies the best performing machine learning algorithms for predicting AKI using trends in laboratory values in ICU patients. Early identification of these patients using readily available data indicates that incorporating machine learning predictive models into electronic medical record systems is an inevitable requisite for improving patient outcomes.

National Risk Factors for Child Maltreatment after Trauma: Failure to Prevent

The purpose of this study was to identify the risk factors for hospital readmission for child maltreatment after trauma, including admissions across different hospitals nationwide. The Nationwide Readmissions Database for 2010–2014 was queried for all patients younger than 18 years admitted for trauma. The primary outcome was readmission for child maltreatment. The secondary outcome was readmission for maltreatment presenting to a hospital different than the index admission hospital. A subgroup analysis was performed on patients without a diagnosis of maltreatment during the index admission. Multivariable logistic regression was performed for each outcome. There were 608,744 admissions identified and 44,569 (7.32%) involved maltreatment at the index admission. Readmission for maltreatment was found in 1,948 (0.32%) patients and 368 (18.89%) presented to a different hospital. The highest risk for readmission for maltreatment was found in patients with maltreatment identified at the index admission (odds ratios (OR) 9.48 [8.35–10.76]). The strongest risk factor for presentation to a different hospital was found with the lowest median household income quartile (OR 3.50 [2.63–4.67]). The subgroup analysis identified 647 (0.11%) children with readmission for maltreatment that was missed during the index admission. The strongest risk factor for this outcome was Injury Severity Score > 15 (OR 3.29 [2.68–4.03]). This study demonstrates that a significant portion of admissions for trauma in children and teenagers could be misrepresented as not involving maltreatment. These index admissions could be the only chance for intervention for child maltreatment. Identifying these at-risk individuals is critical to prevention efforts.

Incidence of recurrent intussusception in young children: A nationwide readmissions analysis

BACKGROUND/PURPOSE

Recurrent intussusception following successful nonoperative reduction has previously been reported with a frequency of 8%-12% based on data from individual institutions. Meanwhile, the timing of discharge after successful reduction continues to be debated. Here, we evaluate readmissions for recurrent intussusception in young children using a large-scale national database.

METHODS

The National Readmissions Database (2010-2014) was queried to identify young children (age < 5 years) diagnosed with intussusception. We compared procedures performed during the index admission and frequency of readmissions for recurrent intussusception. Results were weighted for national estimates.

RESULTS

We identified 8289 children diagnosed with intussusception during an index admission. These patients received definitive treatment with nonoperative reduction alone (43%), surgical reduction (42%), or bowel resection (15%). Readmission for recurrent intussusception was required for 3.7% of patients managed with nonoperative reduction alone, 2.3% of patients that underwent surgical reduction, and 0% of those that underwent bowel resection. Median time to readmission was 4 days after nonoperative reduction, and only 1.5% of these patients experienced recurrence within 48 h of discharge.

CONCLUSIONS

Recurrent intussusception may be substantially less common than previously reported. Our findings support the practice of discharge shortly after successful nonoperative reduction.

TYPE OF STUDY

Retrospective, prognosis study.

LEVEL OF EVIDENCE

III.

Nationwide analysis of mortality and hospital readmissions in esophageal atresia

PURPOSE

The purpose of this study is to identify determinants of mortality and hospital readmission in infants born with esophageal atresia ± tracheoesophageal fistula.

METHODS

The Nationwide Readmissions Database (2010-2014) was queried for newborns with a diagnosis of esophageal atresia. Outcomes included mortality and readmissions at 30-day and 1-year.

RESULTS

3157 patients were identified, of which 54% were male. 81% had an additional congenital anomaly, and 35% had VACTERL association. Overall mortality at index hospitalization was 11% (n = 360) and was significantly higher with additional congenital anomalies (13%), VACTERL (19%), and Spitz classification II/III (18%) vs. isolated esophageal atresia/tracheoesophageal fistula (4%), all p < 0.001. After esophageal atresia repair (n = 2179), 10% (n = 212) were readmitted within 30 days and 26% (n = 563) within 1 year, with 17% admitted to different hospitals. Common diagnoses during readmission were GERD (54%), infections (42%), failure to thrive (17%), tracheomalacia (14%), and esophageal stricture (10%). Unplanned readmissions accounted for 85% of readmissions. A large number underwent operative procedures, most commonly esophageal dilation (17%) and fundoplication/gastrostomy (12%).

CONCLUSION

Our study has uncovered a high likelihood of complications and unplanned readmission within the first year of life for newborns with esophageal atresia. Coordinated multidisciplinary care may help to decrease unnecessary readmissions and improve outcomes in this vulnerable population.

TYPE OF STUDY

Retrospective comparative analysis.

LEVEL OF EVIDENCE

Level III.

Nationwide Outcomes and Risk Factors for Reinjury After Penetrating Trauma

BACKGROUND

Previous studies have shown that a notable portion of patients who are readmitted for reinjury after penetrating trauma present to a different hospital. The purpose of this study was to identify the risk factors for reinjury after penetrating trauma including reinjury admissions to different hospitals.

METHODS

The 2010-2014 Nationwide Readmissions Database was queried for patients surviving penetrating trauma. E-codes identified patients subsequently admitted with a new diagnosis of blunt or penetrating trauma. Univariable analysis was performed using 44 injury, patient, and hospital characteristics. Multivariable logistic regression using significant variables identified risk factors for the outcomes of reinjury, different hospital readmission, and in-hospital mortality after reinjury.

RESULTS

There were 443,113 patients identified. The reinjury rate was 3.5%. Patients presented to a different hospital in 30.0% of reinjuries. Self-inflicted injuries had a higher risk of reinjury (odds ratio [OR]: 2.66, P < 0.05). Readmission to a different hospital increased risk of mortality (OR: 1.62, P < 0.05). Firearm injury on index admission increased risk of mortality after reinjury (OR: 1.94, P < 0.05).

CONCLUSIONS

This study represents the first national finding that one in three patients present to a different hospital for reinjury after penetrating trauma and have a higher risk of mortality due to this fragmentation of care. These findings have implications for quality and cost improvements by identifying areas to improve continuity of care and the implementation of penetrating injury prevention programs.

Multi-Center Outcomes of Chlorhexidine Oral Decontamination in Intensive Care Units

Background: The efficacy of oral chlorhexidine (oCHG) for decontamination in intensive care unit (ICU) patients is controversial. The purpose of this study was to evaluate the effect of oCHG decontamination on the incidence of pneumonia, sepsis, and death in ICU patients. Methods: The Philips eICU database version 2.0 was queried for patients admitted to the ICU for ≥48 hours in 2014-2015. The primary outcome of interest was death in the ICU. Secondary outcomes were a diagnosis of pneumonia or sepsis. Patients with pneumonia or sepsis diagnosed within the first 48 hours of ICU admission were excluded from the outcome analyses. Univariable analysis was performed comparing age, gender, race, severity of illness scores, hospital characteristics, and oCHG order. Multivariable logistic regression was performed using univariable results with p < 0.05. Results: Of the 64,904 patients from 186 hospitals, 22.1% (n = 14,333) had oCHG ordered. The overall mortality rate was 6.9% (n = 4,449) and the mortality rate in patients receiving oCHG was 10.6% (n = 1,518; p < 0.001). After controlling for confounding factors, oCHG remained an independent risk factor for death (odds ratio [OR] 1.25; 95% confidence interval [CI] 1.16-1.34). After excluding patients with an early diagnosis of pneumonia, the overall pneumonia incidence was 2.6% (n = 1,431) and the incidence in patients having oCHG was 4.2% (n = 517; p < 0.001). However, multivariable logistic regression revealed no significant difference in the risk of pneumonia with oCHG (OR 0.97; 95% CI 0.85-1.09). After excluding patients with an early diagnosis of sepsis, the overall rate of sepsis was 1.8% (n = 949) and for patients with oCHG, the rate was 3.3% (n = 388; p < 0.001). After controlling for other confounders, oCHG remained an independent risk factor for sepsis (OR 1.37; 95% CI 1.19-1.59). Conclusions: A chlorhexidine mouthwash order is associated with increased odds of death and sepsis without decreased odds of pneumonia in a heterogeneous cohort of ICU patients. Additional studies are needed to understand better the effect of oCHG on outcomes.

Identifying Populations at Risk for Child Abuse: A Nationwide Analysis

PURPOSE

Child abuse is a national, often hidden, epidemic. The study objective was to determine at-risk populations that have been previously hospitalized prior to their admission for child abuse.

METHODS

The Nationwide Readmissions Database (NRD) was queried for all children hospitalized for abuse. Outcomes were previous admissions and diagnoses. χ² analysis was used; significance equals p < 0.05.

RESULTS

31,153 children were hospitalized for abuse (half owing to physical abuse) during the study period. 11% (n = 3487) of these children had previous admissions (one in three to a different hospital), while 3% (n = 1069) had multiple hospitalizations. 60% of prior admissions had chronic conditions, and 12% had traumatic injuries. Children with chronic conditions were more likely to have sexual abuse (89% vs. 57%, p < 0. 001) and emotional abuse (75% vs. 60%, p < 0. 01). 25% of chronic diagnoses were psychiatric, who were also more likely to have sexual and emotional abuse (47% vs. 5.5% and 10% vs. 1%, all p < 0. 001).

CONCLUSION

This study uncovers a hidden population of children with past admissions for chronic conditions, especially psychiatric diagnoses that are significantly associated with certain types of abuse. Improved measures to accurately identify at-risk children must be developed to prevent future childhood abuse and trauma.

LEVEL OF EVIDENCE

Level III.

TYPE OF STUDY

Retrospective comparative study.

Validating the ATLS Shock Classification for Predicting Death, Transfusion, or Urgent Intervention

BACKGROUND

The Advanced Trauma Life Support (ATLS) shock classification has been accepted as the conceptual framework for clinicians caring for trauma patients. We sought to validate its ability to predict mortality, blood transfusion, and urgent intervention.

MATERIALS AND METHODS

We performed a retrospective review of trauma patients using the 2014 National Trauma Data Bank. Using initial vital signs data, patients were categorized into shock class based on the ATLS program. Rates for urgent blood transfusion, urgent operative intervention, and mortality were compared between classes.

RESULTS

630,635 subjects were included for analysis. Classes 1, 2, 3, and 4 included 312,404, 17,133, 31, and 43 patients, respectively. 300,754 patients did not meet criteria for any ATLS shock class. Of the patients in class 1 shock, 2653 died (0.9%), 3123 (1.0%) were transfused blood products, and 7115 (2.3%) underwent an urgent procedure. In class 2, 219 (1.3%) died, 387 (2.3%) were transfused, and 1575 (9.2%) underwent intervention. In class 3, 7 (22.6%) died, 10 (32.3%) were transfused, and 13 (41.9%) underwent intervention. In class 4, 15 (34.9%) died, 19 (44.2%) were transfused, and 23 (53.5%) underwent intervention. For uncategorized patients, 21,356 (7.1%) died, 15,168 (5.0%) were transfused, and 23,844 (7.9%) underwent intervention.

CONCLUSIONS

Almost half of trauma patients do not meet criteria for any ATLS shock class. Uncategorized patients had a higher mortality (7.1%) than patients in classes 1 and 2 (0.9% and 1.3%, respectively). Classes 3 and 4 only accounted for 0.005% and 0.007%, respectively, of patients. The ATLS classification system does not help identify many patients in severe shock.

Comparing Machine Learning Algorithms for Predicting Acute Kidney Injury

Prior studies have used vital signs and laboratory measurements with conventional modeling techniques to predict acute kidney injury (AKI). The purpose of this study was to use the trend in vital signs and laboratory measurements with machine learning algorithms for predicting AKI in ICU patients. The eICU Collaborative Research Database was queried for five consecutive days of laboratory measurements per patient. Patients with AKI were identified and trends in vital signs and laboratory values were determined by calculating the slope of the least-squares-fit linear equation using three days for each value. Different machine learning classifiers (gradient boosted trees [GBT], logistic regression, and deep learning) were trained to predict AKI using the laboratory values, vital signs, and slopes. There were 151,098 ICU stays identified and the rate of AKI was 5.6 per cent. The best performing algorithm was GBT with an AUC of 0.834 ± 0.006 and an F-measure of 42.96 per cent ± 1.26 per cent. Logistic regression performed with an AUC of 0.827 ± 0.004 and an F-measure of 28.29 per cent ± 1.01 per cent. Deep learning performed with an AUC of 0.817 ± 0.005 and an F-measure of 42.89 per cent ± 0.91 per cent. The most important variable for GBT was the slope of the minimum creatinine (30.32%). This study identifies the best performing machine learning algorithms for predicting AKI using trends in laboratory values in ICU patients. Early identification of these patients using readily available data indicates that incorporating machine learning predictive models into electronic medical record systems is an inevitable requisite for improving patient outcomes.

National Risk Factors for Child Maltreatment after Trauma: Failure to Prevent

The purpose of this study was to identify the risk factors for hospital readmission for child maltreatment after trauma, including admissions across different hospitals nationwide. The Nationwide Readmissions Database for 2010-2014 was queried for all patients younger than 18 years admitted for trauma. The primary outcome was readmission for child maltreatment. The secondary outcome was readmission for maltreatment presenting to a hospital different than the index admission hospital. A subgroup analysis was performed on patients without a diagnosis of maltreatment during the index admission. Multivariable logistic regression was performed for each outcome. There were 608,744 admissions identified and 44,569 (7.32%) involved maltreatment at the index admission. Readmission for maltreatment was found in 1,948 (0.32%) patients and 368 (18.89%) presented to a different hospital. The highest risk for readmission for maltreatment was found in patients with maltreatment identified at the index admission (odds ratios (OR) 9.48 [8.35-10.76]). The strongest risk factor for presentation to a different hospital was found with the lowest median household income quartile (OR 3.50 [2.63-4.67]). The subgroup analysis identified 647 (0.11%) children with readmission for maltreatment that was missed during the index admission. The strongest risk factor for this outcome was Injury Severity Score > 15 (OR 3.29 [2.68-4.03]). This study demonstrates that a significant portion of admissions for trauma in children and teenagers could be misrepresented as not involving maltreatment. These index admissions could be the only chance for intervention for child maltreatment. Identifying these at-risk individuals is critical to prevention efforts.

Readmissions to an alternate hospital in patients undergoing vascular intervention for claudication and critical limb ischemia associated with significantly higher mortality

BACKGROUND

Hospital readmissions with 30 days after vascular surgical interventions have been associated with increased morbidity, mortality, and cost. Readmission rates, now a Centers for Medicare and Medicaid Services quality measure, have been studied in databases that have excluded certain payer types and states and have not accounted for readmission to a hospital different from that of the index admission. More accurate and nationally representative data are needed, because this fragmentation of care could lead to flawed conclusions. The purpose of the present study was to examine the incidence and risk factors for readmission to a nonindex hospital for patients admitted for claudication or critical limb ischemia (CLI). We also examined how this disruption of patient care affects mortality.

METHODS

The 2013 to 2014 Nationwide Readmissions Database was queried for all patients admitted for claudication or CLI who had undergone angioplasty, lower extremity bypass, or aortobifemoral bypass. The outcomes of interest were 30- and 365-day readmission rates to any hospital, 30- and 365-day readmission rates to a nonindex hospital, and mortality rates. Multivariable logistic regression was used to identify risk factors for readmission to a nonindex hospital. The most common readmission diagnoses and diagnosis-related groups were identified.

RESULTS

A total of 92,769 patients had been admitted with peripheral vascular disease (33,055 with claudication and 59,714 with CLI). The 30- and 365-day readmission rate was 8.97% and 21.49% and 19.26% and 40.36%, for claudication and CLI, respectively. Of the 30- and 365-day readmissions, 20.47% and 24.92% had occurred at a nonindex hospital, respectively. Significantly higher mortality rates were found for patients with 30- or 365-day readmissions to different hospitals (odds ratio, 1.4 and 1.8, respectively). Multivariable analysis revealed that procedural indication and angioplasty are not significant risk factors for readmission to a different hospital. However, female sex, length of stay >7 days, and Charlson Comorbidity Index >3 remained significant risk factors for nonindex readmissions. The most common disease groups for nonindex readmission were "septicemia and disseminated infections" (6.5%), "heart failure" (6.4%), "other vascular procedures" (6.1%), and "amputation of lower limb except toes" (4.0%).

CONCLUSIONS

Previously unreported, ≥1 in 4 readmissions after lower extremity vascular procedures for peripheral vascular disease will occur at a nonindex hospital. This fragmentation of care is associated with increased mortality and has serious implications for guiding outcome and quality measures. With a sizeable portion of patients missed by current metrics, concern exists that providers are using flawed data. Further study into social- and patient-specific risk factors might provide methods to prevent these readmissions and improve outcomes in this difficult patient population.

Variations in Nationwide Readmission Patterns after Umbilical Hernia Repair

Up to one in three readmissions occur at a different hospital and are thus missed by current quality metrics. There are no national studies examining 30-day readmission, including to different hospitals, after umbilical hernia repair (UHR). We tested the hypothesis that a large proportion were readmitted to a different hospital, that risk factors for readmission to a different hospital are unique, and that readmission costs differed between the index and different hospitals. The 2013 to 2014 Nationwide Readmissions Database was queried for patients admitted for UHR, and cost was calculated. Multivariate logistic regression identified risk factors for 30-day readmission at index and different hospitals. There were 102,650 admissions for UHR and 8.9 per cent readmissions, of which 15.8 per cent readmissions were to a different hospital. The most common reason for readmission was infection (25.8%). Risk factors for 30-day readmission to any hospital include bowel resection, index admission at a for-profit hospital, Medicare, Medicaid, and Charlson Comorbidity Index ≥ 2. Risk factors for 30-day readmission to a different hospital include elective operation, drug abuse, discharge to a skilled nursing facility, and leaving against medical advice. The median cost of initial admission was higher in those who were readmitted ($16,560 [$10,805–$29,014] vs $11,752 [$8151–$17,724], P < 0.01). The median cost of readmission was also higher among those readmitted to a different hospital ($9826 [$5497–$19,139] vs $9227 [$5211–$16,817], P = 0.02). After UHR, one in six readmissions occur at a different hospital, have unique risk factors, and are costlier. Current hospital benchmarks fail to capture this sub-population and, therefore, likely underestimate UHR readmissions.

Variations in Nationwide Readmission Patterns after Umbilical Hernia Repair

Up to one in three readmissions occur at a different hospital and are thus missed by current quality metrics. There are no national studies examining 30-day readmission, including to different hospitals, after umbilical hernia repair (UHR). We tested the hypothesis that a large proportion were readmitted to a different hospital, that risk factors for readmission to a different hospital are unique, and that readmission costs differed between the index and different hospitals. The 2013 to 2014 Nationwide Readmissions Database was queried for patients admitted for UHR, and cost was calculated. Multivariate logistic regression identified risk factors for 30-day readmission at index and different hospitals. There were 102,650 admissions for UHR and 8.9 per cent readmissions, of which 15.8 per cent readmissions were to a different hospital. The most common reason for readmission was infection (25.8%). Risk factors for 30-day readmission to any hospital include bowel resection, index admission at a for-profit hospital, Medicare, Medicaid, and Charlson Comorbidity Index ≥ 2. Risk factors for 30-day readmission to a different hospital include elective operation, drug abuse, discharge to a skilled nursing facility, and leaving against medical advice. The median cost of initial admission was higher in those who were readmitted ($16,560 [$10,805-$29,014] vs $11,752 [$8151-$17,724], P < 0.01). The median cost of readmission was also higher among those readmitted to a different hospital ($9826 [$5497-$19,139] vs $9227 [$5211-$16,817], P = 0.02). After UHR, one in six readmissions occur at a different hospital, have unique risk factors, and are costlier. Current hospital benchmarks fail to capture this subpopulation and, therefore, likely underestimate UHR readmissions.

Nationwide risk factors for hospital readmission for subsequent injury after motor vehicle crashes

OBJECTIVE

Some drivers involved in motor vehicle crashes across the United States may be identified as at risk of subsequent injury by a similar mechanism. The purpose of this study was to perform a national review of the risk factors for hospitalization for a new injury due to a subsequent motor vehicle crash. It was hypothesized that presenting to a different hospital after subsequent injury would result in worse patient outcomes when compared to presentation at the same hospital.

METHODS

The Nationwide Readmissions Database for 2010-2014 was queried for all inpatient hospitalizations with injury related to motor vehicle traffic. The primary patient outcome of interest was subsequent motor vehicle crash-related injury within 1 year. The secondary patient outcomes were different hospital subsequent injury presentation, higher Injury Severity Score (ISS), longer length of stay (LOS), and in-hospital death after subsequent injury. The analysis of secondary patient outcomes was performed only on patients who were reinjured. Univariable analysis was performed for each outcome using all variables during the index admission. Multivariable logistic regression was performed using all significant (P < .05) variables on univariate analysis. Results were weighted for national estimates.

RESULTS

During the study period, 1,008,991 patients were admitted for motor vehicle-related injury; 12,474 patients (1.2%) suffered a subsequent injury within 1 year. From the reinjured patients, 32.9% presented to a different hospital, 48.9% had a higher ISS, and 22.1% had a longer LOS. The in-hospital mortality rate after subsequent injury was 1.1%. Presentation to a different hospital for subsequent injury was associated with a longer LOS (odds ratio [OR] = 1.32; 95% confidence interval [CI], 1.20-1.45; P < .01) and a higher ISS (OR = 1.38; 95% CI, 1.27-1.49; P < .01). Motorcyclists were more likely to suffer subsequent injury (OR = 1.39; 95% CI, 1.32-1.46; P < .01) and motorcycle passengers were more likely to present to a different hospital with a subsequent injury (OR = 2.49; 95% CI, 1.73-3.59; P < .01). Alcohol abuse was associated with subsequent injury (OR = 1.12; 95% CI, 1.07-1.18; P < .01).

CONCLUSIONS

Nearly a third of patients suffering subsequent motor vehicle crash-related injury after an initial motor vehicle crash in the United States present to a different hospital. These patients are more likely to suffer more severe injuries and longer hospitalizations due to their subsequent injury. Future efforts to prevent these injuries must consider the impact of this fragmentation of care and the implications for quality and cost improvements.

Nationwide outcomes and costs of laparoscopic and robotic vs. open hepatectomy

The safety of hepatectomy continues to improve and it holds a key role in the management of benign and malignant hepatic lesions. Laparoscopic and robotic approaches to hepatectomy are increasingly utilized. The purpose of this study was to compare outcomes and costs of laparoscopic and robotic vs. open approaches to hepatectomy and to determine the national nonelective postoperative readmission rate, including readmission to other hospitals. The Nationwide Readmission Database from 2013 to 2014 was queried for all patients undergoing hepatectomy. Patients undergoing laparoscopic and robotic hepatectomies were compared to patients undergoing open hepatectomy. Multivariate logistic regression was implemented to determine the odds ratios (OR) for non-elective readmission within 45 days. There were 10,870 patients who underwent hepatectomy from 2013 to 2014 and 724 (6.7%) were approached with laparoscopic or robotic technique. The robotic cohort had lower mean cost of the index admission ($24,983 ± $18,329 vs. open $32,391 ± $31,983, p < 0.001, 95% CI - 18,292 to 534), shorter LOS (4.5 ± 3.8 vs. lap 6.8 ± 6.0 vs. open 7.6 ± 7.7 days, p < 0.01), and were less likely to be readmitted within 45 days (7.9% vs. 13.0% lap vs. 13.8% open, p = 0.05). The robotic cohort was slightly younger (mean age 57.5 ± 13.5 vs. lap 60.1 ± 13.8 vs. open 58.9 ± 13.7, p < 0.05), and no significant differences were seen by Charlson Comorbidity Index. Anastomosis of hepatic duct to GI tract carried higher odds of mortality (OR 2.87, p < 0.01) and higher odds of readmission (OR 1.40, p < 0.01). LOS above 7 days increased odds of readmission (OR 2.24, p < 0.01). Nearly one-fifth of patients readmitted after hepatectomy present to a different hospital. Robotic hepatectomy was associated with favorable cost and readmission outcomes compared to laparoscopic and open hepatectomy patients, despite similar patient comorbid burdens and patient's age. Length of stay over 7 days and anastomosis of hepatic duct to GI tract are strong risk factors for readmission and mortality.

Nationally Representative Readmission Factors Associated with Endovascular versus Open Repair of Abdominal Aortic Aneurysm

BACKGROUND

Hospital readmissions are tied to financial penalties and thus significantly influence health-care policy. Many current studies on readmissions lack national representation by not tracking readmissions across hospitals. The recently released Nationwide Readmission Database is one of the most comprehensive national sources of readmission data available, making it an invaluable resource to understand this critically important health policy issue.

METHODS

The Nationwide Readmission Database for 2013 and 2014 was queried for adult patients with abdominal aortic aneurysm (441.4) undergoing endovascular (39.71) or open (38.44) repair. Outcomes examined were overall/initial admission mortality and overall/30-day readmissions. Multivariate logistic regression for these outcomes was also performed on multiple readmission factors.

RESULTS

Fifty-three thousand four hundred seventeen patients underwent abdominal aortic aneurysm repair (47,431 endovascular aortic repair [EVAR] versus 5,986 open surgical repair [OSR]). Significant differences were found for EVAR versus OSR on overall readmissions, initial admission cost, readmission costs, length of stay, days to readmission, and overall/initial admission mortality. Multivariate logistic regression analysis found that length of stay > 30, Charlson Comorbidity Index > 1, discharge disposition, and female sex were all significant predictors of 30-day readmission. Repair type was significantly associated with 30-day readmissions; however, it was not a significant factor for overall readmissions.

CONCLUSION

There are significant differences in costs, prognosis, and readmission rates for EVAR versus OSR. Given that these differences are being used to create "acceptable" readmission rates, disbursement quotas among hospitals, and subsequent penalties for providers outside the expected rates, it is only prudent to obtain the most accurate information to guide those policies.

LEVEL OF EVIDENCE

Care management/epidemiological, level IV.

Teenage Trauma Patients Are at Increased Risk for Readmission for Mental Diseases and Disorders

BACKGROUND

Most studies of readmission after trauma are limited to single institutions or single states. The purpose of this study was to determine the risk factors for readmission after trauma for mental illness including readmissions to different hospitals across the United States.

MATERIALS AND METHODS

The Nationwide Readmission Database for 2013 and 2014 was queried for all patients aged 13 to 64 y with a nonelective admission for trauma and a nonelective readmission within 30 d. Multivariable logistic regression was performed for readmission for mental diseases and disorders.

RESULTS

During the study period, 53,402 patients were readmitted within 30 d after trauma. The most common major diagnostic category on readmission was mental diseases and disorders (12.1%). The age group with the highest percentage of readmissions for mental diseases and disorders was 13 to 17 y (38%). On multivariable regression, the teenage group was also the most likely to be readmitted for mental diseases and disorders compared to 18-44 y (odds ratio [OR] 0.45, P < 0.01) and 45-64 y (OR 0.24, P < 0.01). Other high-risk comorbidities included HIV infection (OR 2.4, P < 0.01), psychosis (OR 2.2, P < 0.01), drug (OR 2.0, P < 0.01), and alcohol (OR 1.4, P < 0.01) abuse.

CONCLUSIONS

Teenage trauma patients are at increased risk for hospital readmission for mental illness. Efforts to reduce these admissions should be targeted toward individuals with high-risk comorbidities such as HIV infection, psychosis, and substance abuse.

Predicting Mortality in the Surgical Intensive Care Unit Using Artificial Intelligence and Natural Language Processing of Physician Documentation

The purpose of this study was to use natural language processing of physician documentation to predict mortality in patients admitted to the surgical intensive care unit (SICU). The Multiparameter Intelligent Monitoring in Intensive Care III database was used to obtain SICU stays with six different severity of illness scores. Natural language processing was performed on the physician notes. Classifiers for predicting mortality were created. One classifier used only the physician notes, one used only the severity of illness scores, and one used the physician notes with severity of injury scores. There were 3838 SICU stays identified during the study period and 5.4 per cent ended with mortality. The classifier trained with physician notes with severity of injury scores performed with the highest area under the curve (0.88 ± 0.05) and accuracy (94.6 ± 1.1%). The most important variable was the Oxford Acute Severity of Illness Score (16.0%). The most important terms were "dilated" (4.3%) and "hemorrhage" (3.7%). This study demonstrates the novel use of artificial intelligence to process physician documentation to predict mortality in the SICU. The classifiers were able to detect the subtle nuances in physician vernacular that predict mortality. These nuances provided improved performance in predicting mortality over physiologic parameters alone.

Predicting Mortality in the Surgical Intensive Care Unit Using Artificial Intelligence and Natural Language Processing of Physician Documentation

The purpose of this study was to use natural language processing of physician documentation to predict mortality in patients admitted to the surgical intensive care unit (SICU). The Multiparameter Intelligent Monitoring in Intensive Care III database was used to obtain SICU stays with six different severity of illness scores. Natural language processing was performed on the physician notes. Classifiers for predicting mortality were created. One classifier used only the physician notes, one used only the severity of illness scores, and one used the physician notes with severity of injury scores. There were 3838 SICU stays identified during the study period and 5.4 per cent ended with mortality. The classifier trained with physician notes with severity of injury scores performed with the highest area under the curve (0.88 ± 0.05) and accuracy (94.6 ± 1.1%). The most important variable was the Oxford Acute Severity of Illness Score (16.0%). The most important terms were “dilated” (4.3%) and “hemorrhage” (3.7%). This study demonstrates the novel use of artificial intelligence to process physician documentation to predict mortality in the SICU. The classifiers were able to detect the subtle nuances in physician vernacular that predict mortality. These nuances provided improved performance in predicting mortality over physiologic parameters alone.

Using artificial intelligence to predict prolonged mechanical ventilation and tracheostomy placement

BACKGROUND

Early identification of critically ill patients who will require prolonged mechanical ventilation (PMV) has proven to be difficult. The purpose of this study was to use machine learning to identify patients at risk for PMV and tracheostomy placement.

MATERIALS AND METHODS

The Multiparameter Intelligent Monitoring in Intensive Care III database was queried for all intensive care unit (ICU) stays with mechanical ventilation. PMV was defined as ventilation >7 d. Classifiers with a gradient-boosted decision trees algorithm were created for the outcomes of PMV and tracheostomy placement. The variables used were six different severity-of-illness scores calculated on the first day of ICU admission including their components and 30 comorbidities. Mean receiver operating characteristic curves were calculated for the outcomes, and variable importance was quantified.

RESULTS

There were 20,262 ICU stays identified. PMV was required in 13.6%, and tracheostomy was performed in 6.6% of patients. The classifier for predicting PMV was able to achieve a mean area under the curve (AUC) of 0.820 ± 0.016, and tracheostomy was predicted with an AUC of 0.830 ± 0.011. There were 60.7% patients admitted to a surgical ICU, and the classifiers for these patients predicted PMV with an AUC of 0.852 ± 0.017 and tracheostomy with an AUC of 0.869 ± 0.015. The variable with the highest importance for predicting PMV was the logistic organ dysfunction score pulmonary component (13%), and the most important comorbidity in predicting tracheostomy was cardiac arrhythmia (12%).

CONCLUSIONS

This study demonstrates the use of artificial intelligence through machine-learning classifiers for the early identification of patients at risk for PMV and tracheostomy. Application of these identification techniques could lead to improved outcomes by allowing for early intervention.

Same-Hospital Re-Admission Rate Is Not Reliable for Measuring Post-Operative Infection-Related Re-Admission

BACKGROUND

Post-operative infections cause morbidity, consume resources, and are an important quality measure in assessing and comparing hospitals. Commonly used metrics do not account for re-admission to a different hospital. The Nationwide Readmissions Database (NRD) tracks re-admissions across United States (US) hospitals. Infection-related re-admission across US hospitals has not been studied previously.

PATIENTS AND METHODS

The 2013 NRD was queried for admissions with a primary International Classification of Diseases and Related Health Problems, 9th revision, Clinical Modification code for the most frequently performed operations. Non-elective all-cause, infection-related, and different hospital 30-day re-admission rates were calculated, using All Patient Refined Diagnosis Related Groups codes. Multi-variable logistic regression identified risk factors for re-admission.

RESULTS

Of 826,836 surviving to discharge, 39,281 (4.8%) had an unplanned re-admission within 30 days, occurring at a different hospital 20.5% of the time. The most common reason for re-admission was infection (25.1%). Orthopedic and spinal procedures were at highest risk for all-cause and infection-related different hospital re-admission. Infection-related different hospital re-admission risk factors included: Length of stay >30 days (odds ratio [OR] 2.28 [1.62-3.21], p < 0.01), age ≥65 years (OR 1.56 [1.38-1.76], p < 0.01), and Charlson Comorbidity Index >1 (OR 1.14 [1.01-1.28], p < 0.01) and differed from predictors of same-hospital infectious re-admission. Non-elective surgical procedure (OR 0.79 [0.72-0.87], p < 0.01) and initial hospitalization at a large hospital (OR 0.66 [0.59-0.74], p < 0.01) were protective.

CONCLUSION

A substantial proportion of post-operative re-admissions are missed by same-hospital re-admission data. All-cause and infection-related post-operative re-admissions to a different hospital are affected by unique patient and institution-specific factors. Re-admission reduction programs, quality metrics, and policy based on same hospital re-admission data should be updated to incorporate different hospital re-admission.

Risk factors for nonelective 30-day readmission in pediatric assault victims

PURPOSE

Hospital readmission in trauma patients is associated with significant morbidity and increased healthcare costs. There is limited published data on early hospital readmission in pediatric trauma patients. As presently in healthcare outcomes and readmissions rates are increasingly used as hospital quality indicators, it is paramount to recognize risk factors for readmission. We sought to identify national readmission rates in pediatric assault victims and identify the most common readmission diagnoses among these patients.

METHODS

The Nationwide Readmission Database (NRD) for 2013 was queried for all patients under 18years of age with a non-elective admission with an E-code that is designed as assault using National Trauma Data Bank Standards. Multivariate logistic regression was implemented using 18 variables to determine the odds ratios (OR) for non-elective readmission within 30-days.

RESULTS

There were 4050 pediatric victims of assault and 92 (2.27%) died during the initial admission. Of the surviving patients 128 (3.23%) were readmitted within 30days. Of these readmitted patients 24 (18.75%) were readmitted to a different hospital and 31 (24.22%) were readmitted for repeated assault. The variables associated with the highest risk for non-elective readmission within 30-days were: length of stay (LOS) >7days (OR 3.028, p<0.01, 95% CI 1.67-5.50), psychoses (OR 3.719, p<0.01, 95% CI 1.70-8.17), and weight loss (OR 4.408, p<0.01, 95% CI 1.92-10.10). The most common readmission diagnosis groups were bipolar disorders (8.2%), post-operative, posttraumatic, or other device infections (6.2%), or major depressive disorders and other/unspecified psychoses (5.2%).

CONCLUSIONS

Readmission after pediatric assault represents a significant resource burden and almost a quarter of those patients are readmitted after a repeated assault. Understanding risk factors and reasons for readmission in pediatric trauma assault victims can improve discharge planning, family education, and outpatient support, thereby decreasing overall costs and resource burden. Psychoses, weight loss, and prolonged hospitalization are independent prognostic indicators of readmission in pediatric assault patients.

LEVEL OF EVIDENCE

Level IV - Prognostic and Epidemiological - Retrospective Study.