Development and External Validation of a Machine Learning Model for Prediction of Potential Transfer to the PICU.

OBJECTIVES: Unrecognized clinical deterioration during illness requiring hospitalization is associated with high risk of mortality and long-term morbidity among children. Our objective was to develop and externally validate machine learning algorithms using electronic health records for identifying ICU transfer within 12 hours indicative of a child's condition. DESIGN: Observational cohort study. SETTING: Two urban, tertiary-care, academic hospitals (sites 1 and 2). PATIENTS: Pediatric inpatients (age <18 yr). INTERVENTIONS: None. MEASUREMENT AND MAIN RESULTS: Our primary outcome was direct ward to ICU transfer. Using age, vital signs, and laboratory results, we derived logistic regression with regularization, restricted cubic spline regression, random forest, and gradient boosted machine learning models. Among 50,830 admissions at site 1 and 88,970 admissions at site 2, 1,993 (3.92%) and 2,317 (2.60%) experienced the primary outcome, respectively. Site 1 data were split longitudinally into derivation (2009-2017) and validation (2018-2019), whereas site 2 constituted the external test cohort. Across both sites, the gradient boosted machine was the most accurate model and outperformed a modified version of the Bedside Pediatric Early Warning Score that only used physiologic variables in terms of discrimination ( C -statistic site 1: 0.84 vs 0.71, p < 0.001; site 2: 0.80 vs 0.74, p < 0.001), sensitivity, specificity, and number needed to alert. CONCLUSIONS: We developed and externally validated a novel machine learning model that identifies ICU transfers in hospitalized children more accurately than current tools. Our model enables early detection of children at risk for deterioration, thereby creating opportunities for intervention and improvement in outcomes.

Eliminating Explicit and Implicit Biases in Health Care: Evidence and Research Needs.

Health care providers hold negative explicit and implicit biases against marginalized groups of people such as racial and ethnic minoritized populations. These biases permeate the health care system and affect patients via patient-clinician communication, clinical decision making, and institutionalized practices. Addressing bias remains a fundamental professional responsibility of those accountable for the health and wellness of our populations. Current interventions include instruction on the existence and harmful role of bias in perpetuating health disparities, as well as skills training for the management of bias. These interventions can raise awareness of provider bias and engage health care providers in establishing egalitarian goals for care delivery, but these changes are not sustained, and the interventions have not demonstrated change in behavior in the clinical or learning environment. Unfortunately, the efficacy of these interventions may be hampered by health care providers' work and learning environments, which are rife with discriminatory practices that sustain the very biases US health care professions are seeking to diminish. We offer a conceptual model demonstrating that provider-level implicit bias interventions should be accompanied by interventions that systemically change structures inside and outside the health care system if the country is to succeed in influencing biases and reducing health inequities.

Outcomes of Ileoanal Pouch Anastomosis in Pediatric Ulcerative Colitis Are Worse in the Modern Era: A Time Trend Analysis Outcomes Following Ileal Pouch-Anal Anastomosis in Pediatric Ulcerative Colitis.

BACKGROUND: Despite significant differences in surgical outcomes between pediatric and adult patients with ulcerative colitis (UC) undergoing colectomy, counseling on pediatric outcomes has largely been guided by data from adults. We compared differences in pouch survival between pediatric and adult patients who underwent total proctocolectomy with ileal pouch-anal anastomosis (IPAA). METHODS: This was a retrospective single-center study of patients with UC treated with IPAA who subsequently underwent pouchoscopy between 1980 and 2019. Data were collected via electronic medical records. We stratified the study population based on age at IPAA. Differences between groups were assessed using t tests and chi-square tests. Kaplan-Meier curves were used to compare survival probabilities. Differences between groups were assessed using a log-rank test. RESULTS: We identified 53 patients with UC who underwent IPAA before 19 years of age and 329 patients with UC who underwent IPAA at or after 19 years of age. Subjects who underwent IPAA as children were more likely to require anti-tumor nerosis factor (TNF) postcolectomy compared with adults (41.5% vs 25.8%; P < .05). Kaplan-Meier estimates revealed that pediatric patients who underwent IPAA in the last 10 years had a 5-year pouch survival probability that was 28% lower than that of those who underwent surgery in the 1990s or 2000s (72% vs 100%; P < .001). Further, children who underwent IPAA and received anti-TNF therapies precolectomy had the most rapid progression to pouch failure when compared with anti-TNF-naive children and with adults who were either exposed or naive precolectomy (P < .05). CONCLUSIONS: There are lower rates of pouch survival for children with UC who underwent IPAA following the uptake of anti-TNF therapy compared with both historical pediatric control subjects and contemporary adults.

Ileal pouch­anal anastomosis is the most common surgical approach for patients with ulcerative colitis undergoing total proctocolectomy. Outcomes are informed by heterogeneous adult data cohorts often predating anti-tumor necrosis factor uptake. We find that for children in the modern era pouch loss occurs at higher rates.

Using Wearable Biosensors to Predict Length of Stay for Patients with IBD After Bowel Surgery.

BACKGROUND AND AIMS: It remains unknown whether ambulation or sleep predicts postoperative length of stay for patients with IBD. We aim to identify the utility of wearable biosensors in predicting postoperative length of stay for patients with IBD. METHODS: Associations of postoperative length of stay with step count/sleep duration/sleep efficiency measured by wearable biosensors were examined. The best-fitting multivariable linear regression model predicting length of stay was constructed using stepwise model selection. RESULTS: Final sample included 37 patients. Shorter sleep duration on postoperative day 4 (r = 0.51, p = 0.043) or 5 (r = 0.81, p = 0.0045) or higher sleep efficiency on postoperative day 5 (r = - 0.77, p = 0.0098) was associated with a shorter length of stay. Additionally, a more positive change in sleep efficiency from postoperative day 4-5 was associated with a shorter length of stay (r = - 0.77, p = 0.024). The best-fitting multivariable linear regression model revealed Clavien-Dindo grade 1 (p = 0.045) and interaction between Clavien-Dindo grade 2/3a and mean daily steps (p = 0.00038) are significant predictors of length of stay. The following variables were not significantly associated with length of stay: mean daily steps/sleep duration/sleep efficiency, average rate of change in these three variables, and changes in step count between successive postoperative days 1-5, sleep duration between successive postoperative days 2-5, and sleep efficiency between successive postoperative days 2-4. CONCLUSION: We demonstrated the utility of activity and sleep data from wearable biosensors in predicting length of stay. Patients with more severe complications may benefit more (i.e., reduced postoperative length of stay) from increased ambulation. However, overall, sleep duration/efficiency did not predict length of stay.

Achieving Histologic Normalization in Ulcerative Colitis Is Associated With a Reduced Risk of Subsequent Dysplasia.

BACKGROUND: Complete histologic normalization is associated with improved clinical outcomes in ulcerative colitis (UC). However, it is currently unknown what effect achieving histologic normalization has on the development of dysplasia. METHODS: We performed a retrospective analysis of 495 patients with a confirmed diagnosis of UC from a tertiary center. Patients were categorized according to the best histologic assessment they had during their disease course: histologic normalization, histologic quiescence, or persistent histologic activity. We assessed dysplasia rates in these patient groups after achieving histologic normalization or histologic quiescence, or 8 years after UC diagnosis in those with persistent histologic activity. Kaplan-Meier graphs and Cox regression analyses were performed to estimate this effect. RESULTS: The incidence rate of dysplasia development after achieving histologic normalization was statistically significantly less when compared with the incidence rate after achieving histologic quiescence (P = 0.001) and in those with persistent histologic activity 8 years after UC diagnosis (P = 0.033). In multivariate analysis, at any point throughout UC duration, dysplasia development was statistically lower in those with histologic normalization (adjusted hazard ratio [aHR], 0.32; 95% confidence interval [CI], 0.13-0.81) but not in those with histologic quiescence (aHR, 0.52; 95% CI, 0.25-1.10), compared with those with persistent histologic inflammation. When assessing the time after achieving histologic normalization, histologic quiescence, or 8 years post UC diagnosis in those with persistent histologic activity, we found that patients with histologic normalization had a subsequent decreased risk of developing dysplasia (aHR, 0.09; 95% CI, 0.01-0.72), compared with patients without normalization. CONCLUSIONS: Histologic normalization is associated with a decreased risk in patients with UC of developing subsequent dysplasia, compared with patients without histologic normalization. These findings have implications for surveillance intervals.

Objective Measures of Physical Distancing in the Hospital During the COVID-19 Pandemic.

During the COVID-19 pandemic, hospitals published physical-distancing guidance and created dedicated respiratory isolation units (RIUs) for patients with COVID-19. The degree to which such distancing occurred between clinicians and patients is unknown. In this study, heat sensors from an existing hospital hand-hygiene monitoring system objectively tracked room entries as a proxy for physical distancing in both RIUs and general medicine units before and during the pandemic. The RIUs saw a 60.6% reduction in entries per room per day (from 85.7 to 33.8). General medicine units that cared for patients under investigation for COVID-19 and other patients experienced a 14.7% reduction in entries per room per day (from 76.9 to 65.1). While gradual extinction was observed in both units as COVID-19 cases declined, the RIUs had a higher degree of physical distancing. Although the optimal level of physical distancing is unknown, sustaining physical distancing in the hospital may require re-education and real-time monitoring.

The impact of the COVID-19 pandemic on nighttime room entries and sleep disruptions for pediatric patients.

OBJECTIVES/BACKGROUND: Sleep is critical to recovery, but inpatient sleep is often disrupted. During the COVID-19 pandemic, social distancing efforts to minimize spread may have improved hospitalized children's sleep by decreasing unnecessary overnight disruptions. This study aimed to describe the impact of these efforts on pediatric inpatient sleep using objective and subjective metrics. METHODS: Sleep disruptions for pediatric inpatients admitted prior to and during the COVID-19 pandemic were compared. Hand hygiene sensors tracking room entries were utilized to measure objective overnight disruptions for 69 nights pre-pandemic and 154 pandemic nights. Caregiver surveys of overnight disruptions, sleep quantity, and caregiver mood were adopted from validated tools: the Karolinska Sleep Log, Potential Hospital Sleep Disruptions and Noises Questionnaire, and Visual Analog Mood Scale. RESULTS: Nighttime room entries initially decreased 36% (95% CI: 30%, 42%, p < 0.001), then returned towards baseline, mirroring the COVID-19 hospital census. However, surveyed caregivers (n_pre = 293, n_post = 154) reported more disrupted sleep (p < 0.001) due to tests (21% vs. 38%), anxiety (23% vs. 41%), and pain (23% vs. 48%). Caregivers also reported children slept 61 fewer minutes (95% CI: -12 min, -110 min, p < 0.001). Caregivers self-reported feeling more sad, weary, and worse overall (p < 0.001 for all). CONCLUSIONS: Despite a decrease in objective room entries during the pandemic, caregivers reported their children were disrupted more and slept less. Caregivers also self-reported worse mood. This highlights the effects of the COVID-19 pandemic on subjective experiences of hospitalized children and their caregivers. Future work targeting stress and anxiety could improve pediatric inpatient sleep.

Identification of Risk Factors for Coexisting Sinusitis and Inflammatory Bowel Disease.

BACKGROUND: This study aimed to analyze the association of coexisting sinusitis and IBD, establish significant factors involved in their development, and enable further biological correlation between these two diseases. METHODS: The IBD and Sinusitis Study at UChicago Medicine (TISSUe) is a retrospective, single-center study. We reviewed patients to confirm IBD and chronic sinusitis diagnoses. Case-control propensity score matching was performed using matched controls with IBD only or sinusitis only. Statistical methods included Chi-squared test and Wilcoxon rank sum test. Logistic regression analysis was performed, and factors were considered significant if p<0.05. RESULTS: Stratifying 214 patients with coexisting IBD and sinusitis, 176 patients had IBD first and 38 patients had sinusitis first. Multivariable analysis of factors associated with subsequent disease with matched controls determined that duration of disease, UC, steroid exposure ever, and younger age of IBD diagnosis were associated with subsequent sinusitis in patients with IBD; steroid exposure ever and duration of sinusitis were significantly associated with subsequent IBD in patients with sinusitis. CONCLUSIONS: This study suggests that IBD maintenance therapies are not associated with increased risk of sinusitis, as proposed by adverse events in clinical trial data; rather, UC diagnosis and duration of disease may be more influential in sinusitis development. While further studies are necessary, this study also demonstrates that sinusitis precedes IBD in some patients, probing its biological association with IBD and possible classification as an extraintestinal manifestation.

Black Race and Public Insurance Are Predictive of Inappropriate Evaluation of Iron Deficiency Anemia and Diarrhea.

BACKGROUND AND AIMS: Socioeconomic status, race, and insurance can impact healthcare delivery and utilization in several chronic disease states. The primary aim of our study was to determine whether race and insurance status are predictors of having an appropriate workup for celiac disease and inflammatory bowel disease (IBD) when presenting with iron deficiency anemia (IDA) and chronic diarrhea. METHODS: Medical records of patients seen at the University of Chicago Medical Center between January 1, 2006, and September 20, 2017, were reviewed. Patients with two separate encounters within 6 months associated with the diagnosis codes for both IDA and chronic diarrhea were identified. Patients without a diagnosis code for IBD and celiac disease were further grouped as those that had an "appropriate" workup and those that did not. Factors associated with the appropriate evaluation were analyzed by univariate and multivariate logistic regression. RESULTS: In total, 899,701 records were searched. A total of 83 patients fit inclusion into the study (8 IBD, 3 CD, 72 neither IBD or CD). Black race was associated with a 91% decreased odds of having the appropriate workup on univariate (OR 0.090, 95%CI 0.017-0.475, p = 0.005) and age-adjusted multivariate analysis (OR 0.095, 95% CI 0.017-0.527, p = 0.007). Public insurance status was significantly associated with a 90% decreased odds of appropriate workup on univariate (OR 0.102, 95% CI 0.024-0.438, p = 0.002) and age-adjusted multivariate analysis (OR 0.104, 95% CI 0.021-0.513, p = 0.005). CONCLUSIONS: Black race and public insurance were significantly associated with not having an appropriate workup for IBD and celiac disease when presenting with iron deficiency and chronic diarrhea.

Improving Hospitalized Children's Sleep by Reducing Excessive Overnight Blood Pressure Monitoring.

BACKGROUND: Although required for healing, sleep is often disrupted during hospitalization. Blood pressure (BP) monitoring can be especially disruptive for pediatric inpatients and has few clinical indications. Our aim in this pilot study was to reduce unnecessary overnight BP monitoring and improve sleep for pediatric inpatients. METHODS: The intervention in June 2018 involved clinician education sessions and updated electronic health record (EHR) orders that enabled the forgoing of overnight BP checks. The postintervention period from July 2018 to May 2019 examined patient-caregiver surveys as outcome measures. These surveys measured inpatient sleep and overnight disruptions and were adopted from validated surveys: the Patient Sleep Questionnaire, expanded Brief Infant Sleep Questionnaire, and Potential Hospital Sleep Disruptions and Noises Questionnaire. Uptake of new sleep-friendly EHR orders was a process measure. Reported patient care escalations served as a balancing measure. RESULTS: Interrupted time series analysis of EHR orders (npre = 493; npost = 1472) showed an increase in intercept for the proportion of patients forgoing overnight BP postintervention (+50.7%; 95% confidence interval 41.2% to 60.3%; P < .001) and a subsequent decrease in slope each week (-0.16%; 95% confidence interval -0.32% to -0.01%; P = .037). Statistical process control of surveys (npre = 263; npost = 131) showed a significant increase in sleep duration for patients older than 2, and nighttime disruptions by clinicians decreased by 19% (P < .001). Annual estimated cost savings were $15 842.01. No major adverse events in patients forgoing BP were reported. CONCLUSIONS: A pilot study combining EHR changes and clinician education safely decreased overnight BP checks, increased pediatric inpatient sleep duration, and reduced nighttime disruptions by clinicians.

Characterizing pediatric inpatient sleep duration and disruptions.

OBJECTIVE: To contextualize inpatient sleep duration and disruptions in a general pediatric hospital ward by comparing in-hospital and at-home sleep durations to recommended guidelines and to objectively measure nighttime room entries. METHODS: Caregivers of patients four weeks - 18 years of age reported patient sleep duration and disruptions in anonymous surveys. Average at-home and in-hospital sleep durations were compared to National Sleep Foundation recommendations. Objective nighttime traffic was evaluated as the average number of room entries between 11:00pm and 7:00am using GOJO brand hand-hygiene room entry data. RESULTS: Among 246 patients, patients slept less in the hospital than at home with newborn and infant cohorts experiencing 7- and 4-h sleep deficits respectively (Newborn: 787 ± 318 min at home vs. 354 ± 211 min in hospital, p < 0.001; Infants: 703 ± 203 min at home vs. 412 ± 152 min in hospital, p < 0.01). Newborn children also experienced >2 h sleep deficits at home when compared to NSF recommendations (Newborns: 787 ± 318 min at home vs. 930 min recommended, p < 0.05). Objective nighttime traffic measures revealed that hospitalized children experienced 7.3 room entries/night (7.3 ± 0.25 entries). Nighttime traffic was significantly correlated with caregiver-reported nighttime awakenings (Spearman Rank Correlation Coefficient: 0.83, p < 0.001). CONCLUSION: Hospitalization is a missed opportunity to improve sleep both in the hospital and at home.

Caregiver and Staff Perceptions of Disruptions to Pediatric Inpatient Sleep.

STUDY OBJECTIVES: Sleep is critical to a child's health and well-being, but children are likely to sleep less and be awakened more often during the night in the hospital than at home. To date no studies have compared caregiver, nurse, and physician perspectives of nighttime sleep disruptions in the pediatric general medicine setting. Our aim was to assess caregiver, nurse, and physician perspectives on the most frequent in-hospital disruptors of sleep for pediatric patients. Additionally, we evaluated the degree of agreement of those opinions between the caregivers and medical team. METHODS: Caregivers, nurses, and physicians were surveyed using the Potential Hospital Sleep Disruption and Noises Questionnaire (PHSDNQ) regarding their opinions on factors that disrupt sleep. Caregiver responses were collected via a convenience sample of patients hospitalized from February to August 2017 and hospital staff was surveyed once regarding overall perception. The perceived percentage of patients disrupted by each factor was calculated and compared among groups using chi-square tests. Using caregiver rank order based on mean response as the reference gold standard, the absolute differences of nurse and physician rank orders were summed and analyzed using a two-sample test of proportion. In addition, staff was asked knowledge and empowerment questions about how to maximize patient sleep in the hospital and responses were compared using chi-square tests. RESULTS: A total of 162 caregivers, 77 nurses (84% response rate), and 81 physicians (90% response rate) completed surveys. Checking vital signs (50%), nurse/physician interruption (49%), and continuous pulse oximetry (38%) were the three most prevalent disruptors of pediatric inpatient sleep as reported by caregivers. Significant differences were observed between caregiver, nurse, and physician responses for pain, anxiety, alarms, noise, and tests (P ≤ .001 for all). Both nurse and physician rank orders were discordant when compared to caregivers; there was no significant difference between the two staff groups. When compared to physicians, nurses reported doing more to help children sleep in the hospital (33% versus 94%, P < .001). CONCLUSIONS: Although caregivers report medical interventions such as checking vital signs, nurse/physician interruption, and continuous pulse oximetry as the most frequent disruptors of inpatient pediatric sleep, pediatric staff has poor insight into these disruptions.