STUDY DESIGN: A post hoc analysis. OBJECTIVE: Advances in machine learning (ML) have led to tools offering individualized outcome predictions for adult spinal deformity (ASD). Our objective is to examine the properties of these ASD models in a cohort of adult symptomatic lumbar scoliosis (ASLS) patients. SUMMARY OF BACKGROUND DATA: ML algorithms produce patient-specific probabilities of outcomes, including major complication (MC), reoperation (RO), and readmission (RA) in ASD. External validation of these models is needed. METHODS: Thirty-nine predictive factors (12 demographic, 9 radiographic, 4 health-related quality of life, 14 surgical) were retrieved and entered into web-based prediction models for MC, unplanned RO, and hospital RA. Calculated probabilities were compared with actual event rates. Discrimination and calibration were analyzed using receiver operative characteristic area under the curve (where 0.5=chance, 1=perfect) and calibration curves (Brier scores, where 0.25=chance, 0=perfect). Ninety-five percent confidence intervals are reported. RESULTS: A total of 169 of 187 (90%) surgical patients completed 2-year follow up. The observed rate of MCs was 41.4% with model predictions ranging from 13% to 68% (mean: 38.7%). RO was 20.7% with model predictions ranging from 9% to 54% (mean: 30.1%). Hospital RA was 17.2% with model predictions ranging from 13% to 50% (mean: 28.5%). Model classification for all three outcome measures was better than chance for all [area under the curve=MC 0.6 (0.5-0.7), RA 0.6 (0.5-0.7), RO 0.6 (0.5-0.7)]. Calibration was better than chance for all, though best for RA and RO (Brier Score=MC 0.22, RA 0.16, RO 0.17). CONCLUSIONS: ASD prediction models for MC, RA, and RO performed better than chance in a cohort of adult lumbar scoliosis patients, though the homogeneity of ASLS affected calibration and accuracy. Optimization of models require samples with the breadth of outcomes (0%-100%), supporting the need for continued data collection as personalized prediction models may improve decision-making for the patient and surgeon alike.

Background

The Kids Intracranial Injury Decision Support tool for Traumatic Brain Injury (KIIDS-TBI) tool is a validated risk prediction model for managing children with mild traumatic brain injuries (mTBI) and intracranial injuries. Electronic clinical decision support (CDS) may facilitate the clinical implementation of this evidence-based guidance.

Objective

Our objective was to evaluate the acceptability and usability of an electronic CDS tool for managing children with mTBI and intracranial injuries.

Methods

Emergency medicine and neurosurgery physicians (10 each) from 10 hospitals in the United States were recruited to participate in usability testing of a novel CDS prototype in a simulated electronic health record environment. Testing included a think-aloud protocol, an acceptability and usability survey, and a semi-structured interview. The prototype was updated twice during testing to reflect user feedback. Usability problems recorded in the videos were categorized using content analysis. Interview transcripts were analyzed using thematic analysis.

Results

Among the 20 participants, most worked at teaching hospitals (80%), freestanding children's hospitals (95%), and level-1 trauma centers (75%). During the two prototype updates, problems with clarity of terminology and navigating through the CDS interface were identified and corrected. Corresponding to these changes, the number of usability problems decreased from 35 in phase 1 to 8 in phase 3 and the number of mistakes made decreased from 18 (phase 1) to 2 (phase 3). Through the survey, participants found the tool easy to use (90%), useful for determining a patient's level of care (95%), and likely to improve resource use (90%) and patient safety (79%). Interview themes related to the CDS's ability to support evidence-based decision-making and improve clinical workflow proposed implementation strategies and potential pitfalls.

Conclusion

After iterative evaluation and refinement, the KIIDS-TBI CDS tool was found to be highly usable and useful for aiding the management of children with mTBI and intracranial injuries.

Background

Current management of children with minor head trauma (MHT) and intracranial injuries is not evidence-based and may place some children at risk of harm. Evidence-based electronic clinical decision support (CDS) for management of these children may improve patient safety and decrease resource use. To guide these efforts, we evaluated the sociotechnical environment impacting the implementation of electronic CDS, including workflow and communication, institutional culture, and hardware and software infrastructure, among other factors.

Methods

Between March and May, 2020 semi-structured qualitative focus group interviews were conducted to identify sociotechnical influences on CDS implementation. Physicians from neurosurgery, emergency medicine, critical care, and pediatric general surgery were included, along with information technology specialists. Participants were recruited from nine health centers in the United States. Focus group transcripts were coded and analyzed using thematic analysis. The final themes were then cross-referenced with previously defined sociotechnical dimensions.

Results

We included 28 physicians and four information technology specialists in seven focus groups (median five participants per group). Five physicians were trainees and 10 had administrative leadership positions. Through inductive thematic analysis, we identified five primary themes: (1) clinical impact; (2) stakeholders and users; (3) tool content; (4) clinical practice integration; and (5) post-implementation evaluation measures. Participants generally supported using CDS to determine an appropriate level-of-care for these children. However, some had mixed feelings regarding how the tool could best be used by different specialties (e.g. use by neurosurgeons versus non-neurosurgeons). Feedback from the interviews helped refine the tool content and also highlighted potential technical and workflow barriers to address prior to implementation.

Conclusions

We identified key factors impacting the implementation of electronic CDS for children with MHT and intracranial injuries. These results have informed our implementation strategy and may also serve as a template for future efforts to implement health information technology in a multidisciplinary, emergency setting.

Introduction

Nonalcoholic fatty liver disease (NAFLD) is an epidemic in the obese population. Bariatric surgery is known to reverse multiple metabolic complications of obesity such as diabetes, dyslipidemia, and NAFLD, but the timing of liver changes has not been well described.

Materials and methods

This was an IRB-approved, two-institutional prospective study. Bariatric patients received MRIs at baseline and after a pre-operative liquid diet. Liver biopsies were performed during surgery and if NAFLD positive, the patients received MRIs at 1, 3, and 6 months. Liver volumes and proton-density fat fraction (PDFF) were calculated from offline MRI images. Primary outcomes were changes in weight, body mass index (BMI), percent excess weight loss (EWL%), liver volume, and PDFF. Resolution of steatosis, as defined as PDFF < 6.4% based on previously published cutoffs, was assessed. Secondarily, outcomes were compared between patients who underwent laparoscopic sleeve gastrectomy (LSG) versus laparoscopic Roux-en-Y gastric bypass (LRYGB).

Results

From October 2010 to June 2015, 124 patients were recruited. 49 patients (39.5%) completed all five scans. EWL% at 6 months was 55.6 ± 19.0%. BMI decreased from 45.3 ± 5.9 to 34.4 ± 5.1 kg/m² and mean liver volume decreased from 2464.6 ± 619.4 to 1874.3 ± 387.8 cm³ with a volume change of 21.4 ± 11.4%. PDFF decreased from 16.6 ± 7.8 to 4.4 ± 3.4%. At 6 months, 83.7% patients had resolution of steatosis. Liver volume plateaued at 1 month, but PDFF and BMI continued to decrease. There were no statistically significant differences in liver volume or PDFF reduction from baseline to 6 months between the LSG versus LRYGB subgroups.

Conclusion

Patients with NAFLD undergoing bariatric surgery can expect significant decreases in liver volume and hepatic steatosis at 6 months, with 83.7% of patients achieving resolution of steatosis. Liver volume reduction plateaus 1-month post-bariatric surgery, but PDFF continues to decrease. LSG and LRYGB did not differ in efficacy for inducing regression of hepatosteatosis.

Purpose To longitudinally monitor liver fat before and after bariatric surgery by using quantitative chemical shift-encoded (CSE) MRI and to compare with changes in body mass index (BMI), weight, and waist circumference (WC). Materials and Methods For this prospective study, which was approved by the internal review board, a total of 126 participants with obesity who were undergoing evaluation for bariatric surgery with preoperative very low calorie diet (VLCD) were recruited from June 27, 2010, through May 5, 2015. Written informed consent was obtained from all participants. Participants underwent CSE MRI measuring liver proton density fat fraction (PDFF) before VLCD (2-3 weeks before surgery), after VLCD (1-3 days before surgery), and 1, 3, and 6-10 months following surgery. Linear regression was used to estimate rates of change of PDFF (ΔPDFF) and body anthropometrics. Initial PDFF (PDFF0), initial anthropometrics, and anthropometric rates of change were evaluated as predictors of ΔPDFF. Mixed-effects regression was used to estimate time to normalization of PDFF. Results Fifty participants (mean age, 51.0 years; age range, 27-70 years), including 43 women (mean age, 50.8 years; age range, 27-70 years) and seven men (mean age, 51.7 years; age range, 36-62 years), with mean PDFF0 ± standard deviation of 18.1% ± 8.6 and mean BMI0 of 44.9 kg/m2 ± 6.5 completed the study. By 6-10 months following surgery, mean PDFF decreased to 4.9% ± 3.4 and mean BMI decreased to 34.5 kg/m2 ± 5.4. Mean estimated time to PDFF normalization was 22.5 weeks ± 11.5. PDFF0 was the only strong predictor for both ΔPDFF and time to PDFF normalization. No body anthropometric correlated with either outcome. Conclusion Average liver proton density fat fraction (PDFF) decreased to normal (< 5%) by 6-10 months following surgery, with mean time to normalization of approximately 5 months. Initial PDFF was a strong predictor of both rate of change of PDFF and time to normalization. Body anthropometrics did not predict either outcome. Online supplemental material is available for this article. © RSNA, 2018.

Purpose

To compare longitudinal hepatic proton density fat fraction (PDFF) changes estimated by magnitude- vs. complex-based chemical-shift-encoded MRI during a weight loss surgery (WLS) program in severely obese adults with biopsy-proven nonalcoholic fatty liver disease (NAFLD).

Methods

This was a secondary analysis of a prospective dual-center longitudinal study of 54 adults (44 women; mean age 52 years; range 27-70 years) with obesity, biopsy-proven NAFLD, and baseline PDFF ≥ 5%, enrolled in a WLS program. PDFF was estimated by confounder-corrected chemical-shift-encoded MRI using magnitude (MRI-M)- and complex (MRI-C)-based techniques at baseline (visit 1), after a 2- to 4-week very low-calorie diet (visit 2), and at 1, 3, and 6 months (visits 3 to 5) after surgery. At each visit, PDFF values estimated by MRI-M and MRI-C were compared by a paired t test. Rates of PDFF change estimated by MRI-M and MRI-C for visits 1 to 3, and for visits 3 to 5 were assessed by Bland-Altman analysis and intraclass correlation coefficients (ICCs).

Results

MRI-M PDFF estimates were lower by 0.5-0.7% compared with those of MRI-C at all visits (p < 0.001). There was high agreement and no difference between PDFF change rates estimated by MRI-M vs. MRI-C for visits 1 to 3 (ICC 0.983, 95% CI 0.971, 0.99; bias = - 0.13%, p = 0.22), or visits 3 to 5 (ICC 0.956, 95% CI 0.919-0.977%; bias = 0.03%, p = 0.36).

Conclusion

Although MRI-M underestimates PDFF compared with MRI-C cross-sectionally, this bias is consistent and MRI-M and MRI-C agree in estimating the rate of hepatic PDFF change longitudinally.

Key points

• MRI-M demonstrates a significant but small and consistent bias (0.5-0.7%; p < 0.001) towards underestimation of PDFF compared with MRI-C at 3 T. • Rates of PDFF change estimated by MRI-M and MRI-C agree closely (ICC 0.96-0.98) in adults with severe obesity and biopsy- proven NAFLD enrolled in a weight loss surgery program. • Our findings support the use of either MRI technique (MRI-M or MRI-C) for clinical care or by individual sites or for multi-center trials that include PDFF change as an endpoint. However, since there is a bias in their measurements, the same technique should be used in any given patient for longitudinal follow-up.

To assess and compare the diagnostic performance of magnitude-reconstruction chemical-shift-encoded magnetic resonance imaging (MRI-M) and complex-reconstruction chemical-shift-encoded magnetic resonance imaging (MRI-C) for diagnosis of hepatic steatosis in subjects with severe obesity without known non-alcoholic fatty liver disease (NAFLD), using contemporaneous histology as reference. This is an IRB-approved, HIPAA-compliant, two-center, cross-sectional study of a larger prospective trail that recruited patients without known NAFLD consecutively between October 2010 and March 2015 to undergo research MRI exams 1-2 days prior to clinical-care weight-loss surgery. Proton denisty fat fraction (PDFF) was estimated using MRI-M and MRI-C. Liver biopsies were obtained intraoperatively. Using histologically-determined presence of steatosis as the reference standard, receiver operating characteristics (ROC) analyses were used to identify MRI-M- and MRI-C-derived PDFF thresholds for diagnosing steatosis. Bootstrapped-based tests were used to compare their diagnostic performace. A total of 81 patients (67 female, 14 maile, average age 48.2) were recruited for this study. MRI-M and MRI-C had areas under the ROC curve of 0.951 and 0.947, respectively, for diagnosing hepatic steatosis. for MRI-M, the Youden-index-based PDFF threshold of 6.5% provided 0.87 sensitivity (95% confidence internal: 0/75, 0.95), 0.96 specificity (0.81, 0.99), and 0.90 total accuracy (0.82, 0.96). For MRI-C, a PDFF threshold of 6.8%, provided a 0.90 sensitivity (0.77, 0.96), 0.96 specificity (0.81, 0.99), and 0.91 total accuracy (0.83, 0.97). Differences in performance parameters between MRI-M and MRI-C were not statistically significant. Conclusion: MRI-M- and MRI-C-derived PDFF is accurate for non-invasive diagnosis of hepatic steatosis in subjects with severe obesity.

Background

When evaluating children with mild traumatic brain injuries (mTBIs) and intracranial injuries (ICIs), neurosurgeons intuitively consider injury size. However, the extent to which such measures (eg, hematoma size) improve risk prediction compared with the kids intracranial injury decision support tool for traumatic brain injury (KIIDS-TBI) model, which only includes the presence/absence of imaging findings, remains unknown.

Objective

To determine the extent to which measures of injury size improve risk prediction for children with mild traumatic brain injuries and ICIs.

Methods

We included children ≤18 years who presented to 1 of the 5 centers within 24 hours of TBI, had Glasgow Coma Scale scores of 13 to 15, and had ICI on neuroimaging. The data set was split into training (n = 1126) and testing (n = 374) cohorts. We used generalized linear modeling (GLM) and recursive partitioning (RP) to predict the composite of neurosurgery, intubation >24 hours, or death because of TBI. Each model's sensitivity/specificity was compared with the validated KIIDS-TBI model across 3 decision-making risk cutoffs (<1%, <3%, and <5% predicted risk).

Results

The GLM and RP models included similar imaging variables (eg, epidural hematoma size) while the GLM model incorporated additional clinical predictors (eg, Glasgow Coma Scale score). The GLM (76%-90%) and RP (79%-87%) models showed similar specificity across all risk cutoffs, but the GLM model had higher sensitivity (89%-96% for GLM; 89% for RP). By comparison, the KIIDS-TBI model had slightly higher sensitivity (93%-100%) but lower specificity (27%-82%).

Conclusion

Although measures of ICI size have clear intuitive value, the tradeoff between higher specificity and lower sensitivity does not support the addition of such information to the KIIDS-TBI model.