Search

ImportanceDiagnostic information from administrative claims and electronic health record (EHR) data may serve as an important resource for surveillance of vision and eye health, but the accuracy and validity of these sources are unknown.ObjectiveTo estimate the accuracy of diagnosis codes in administrative claims and EHRs compared to retrospective medical record review.Design, Setting, and ParticipantsThis cross-sectional study compared the presence and prevalence of eye disorders based on diagnostic codes in EHR and claims records vs clinical medical record review at University of Washington–affiliated ophthalmology or optometry clinics from May 2018 to April 2020. Patients 16 years and older with an eye examination in the previous 2 years were included, oversampled for diagnosed major eye diseases and visual acuity loss.ExposuresPatients were assigned to vision and eye health condition categories based on diagnosis codes present in their billing claims history and EHR using the diagnostic case definitions of the US Centers for Disease Control and Prevention Vision and Eye Health Surveillance System (VEHSS) as well as clinical assessment based on retrospective medical record review.Main Outcome and MeasuresAccuracy was measured as area under the receiver operating characteristic curve (AUC) of claims and EHR-based diagnostic coding vs retrospective review of clinical assessments and treatment plans.ResultsAmong 669 participants (mean [range] age, 66.1 [16-99] years; 357 [53.4%] female), identification of diseases in billing claims and EHR data using VEHSS case definitions was accurate for diabetic retinopathy (claims AUC, 0.94; 95% CI, 0.91-0.98; EHR AUC, 0.97; 95% CI, 0.95-0.99), glaucoma (claims AUC, 0.90; 95% CI, 0.88-0.93; EHR AUC, 0.93; 95% CI, 0.90-0.95), age-related macular degeneration (claims AUC, 0.87; 95% CI, 0.83-0.92; EHR AUC, 0.96; 95% CI, 0.94-0.98), and cataracts (claims AUC, 0.82; 95% CI, 0.79-0.86; EHR AUC, 0.91; 95% CI, 0.89-0.93). However, several condition categories showed low validity with AUCs below 0.7, including diagnosed disorders of refraction and accommodation (claims AUC, 0.54; 95% CI, 0.49-0.60; EHR AUC, 0.61; 95% CI, 0.56-0.67), diagnosed blindness and low vision (claims AUC, 0.56; 95% CI, 0.53-0.58; EHR AUC, 0.57; 95% CI, 0.54-0.59), and orbital and external diseases (claims AUC, 0.63; 95% CI, 0.57-0.69; EHR AUC, 0.65; 95% CI, 0.59-0.70).Conclusion and RelevanceIn this cross-sectional study of current and recent ophthalmology patients with high rates of eye disorders and vision loss, identification of major vision-threatening eye disorders based on diagnosis codes in claims and EHR records was accurate. However, vision loss, refractive error, and other broadly defined or lower-risk disorder categories were less accurately identified by diagnosis codes in claims and EHR data.

Measurements of visceral adipose tissue cross-sectional area and radiation attenuation from computed tomography (CT) scans provide useful information about risk and mortality. However, scan protocols vary, encompassing differing vertebra levels and utilizing differing phases of contrast enhancement. Furthermore, fat measurements have been extracted from CT using different Hounsfield Unit (HU) ranges. To our knowledge, there have been no large studies of healthy cohorts that reported reference values for visceral fat area and radiation attenuation at multiple vertebra levels, for different contrast phases, and using different fat HU ranges. Two-phase CT scans from 1,677 healthy, adult kidney donors (age 18–65) between 1999 and 2017, previously studied to determine healthy reference values for skeletal muscle measures, were utilized. Visceral adipose tissue cross-sectional area (VFA) and radiation attenuation (VFRA) measures were quantified using axial slices at T10 through L4 vertebra levels. T-tests were used to compare males and females, while paired t-tests were conducted to determine the effect (magnitude and direction) of (a) contrast enhancement and (b) different fat HU ranges on each fat measure at each vertebra level. We report the means, standard deviations, and effect sizes of contrast enhancement and fat HU range. Male and female VFA and VFRA were significantly different at all vertebra levels in both contrast and non-contrast scans. Peak VFA was observed at L4 in females and L2 in males, while peak VFRA was observed at L1 in both females and males. In general, non-contrast scans showed significantly greater VFA and VFRA compared to contrast scans. The average paired difference due to contrast ranged from 1.6 to − 8% (VFA) and 3.2 to − 3.0% (VFRA) of the non-contrast value. HU range showed much greater differences in VFA and VFRA than contrast. The average paired differences due to HU range ranged from − 5.3 to 22.2% (VFA) and − 5.9 to 13.6% (VFRA) in non-contrast scans, and − 4.4 to 20.2% (VFA) and − 4.1 to 12.6% (VFRA) in contrast scans. The − 190 to − 30 HU range showed the largest differences in both VFA (10.8% to 22.2%) and VFRA (7.6% to 13.6%) compared to the reference range (− 205 to − 51 HU). Incidentally, we found that differences in lung inflation result in very large differences in visceral fat measures, particularly in the thoracic region. We assessed the independent effects of contrast presence and fat HU ranges on visceral fat cross-sectional area and mean radiation attenuation, finding significant differences particularly between different fat HU ranges. These results demonstrate that CT measurements of visceral fat area and radiation attenuation are strongly dependent upon contrast presence, fat HU range, sex, breath cycle, and vertebra level of measurement. We quantified contrast and non-contrast reference values separately for males and females, using different fat HU ranges, for lumbar and thoracic CT visceral fat measures at multiple vertebra levels in a healthy adult US population.

BACKGROUND Self-reported questions on blindness and vision problems are collected in many national surveys. Recently released surveillance estimates on the prevalence of vision loss used self-reported data to predict variation in the prevalence of objectively measured acuity loss among population groups for whom examination data are not available. However, the validity of self-reported measures to predict prevalence and disparities in visual acuity has not been established. OBJECTIVE This study aimed to estimate the diagnostic accuracy of self-reported vision loss measures compared to best-corrected visual acuity (BCVA), inform the design and selection of questions for future data collection, and identify the concordance between self-reported vision and measured acuity at the population level to support ongoing surveillance efforts. METHODS We calculated accuracy and correlation between self-reported visual function versus BCVA at the individual and population level among patients from the University of Washington ophthalmology or optometry clinics with a prior eye examination, randomly oversampled for visual acuity loss or diagnosed eye diseases. Self-reported visual function was collected via telephone survey. BCVA was determined based on retrospective chart review. Diagnostic accuracy of questions at the person level was measured based on the area under the receiver operator curve (AUC), whereas population-level accuracy was determined based on correlation. RESULTS The survey question, “Are you blind or do you have serious difficulty seeing, even when wearing glasses?” had the highest accuracy for identifying patients with blindness (BCVA ≤20/200; AUC=0.797). The highest accuracy for detecting any vision loss (BCVA CONCLUSIONS Although survey questions are not considered to be sufficiently accurate to be used as a diagnostic test at the individual level, we did find relatively high levels of accuracy for some questions. At the population level, we found that the relative prevalence of the 2 most accurate survey questions were highly correlated with the prevalence of measured visual acuity loss among nearly all demographic groups. The results of this study suggest that self-reported vision questions fielded in national surveys are likely to yield an accurate and stable signal of vision loss across different population groups, although the actual measure of prevalence from these questions is not directly analogous to that of BCVA.

Body composition measures derived from already available electronic medical records (computed tomography [CT] scans) can have significant value, but automation of measurements is needed for clinical implementation. We sought to use artificial intelligence to develop an automated method to measure body composition and test the algorithm on a clinical cohort to predict mortality. We constructed a deep learning algorithm using Google’s DeepLabv3+ on a cohort of de‐identified CT scans (n = 12,067). To test for the accuracy and clinical usefulness of the algorithm, we used a unique cohort of prospectively followed patients with cirrhosis (n = 238) who had CT scans performed. To assess model performance, we used the confusion matrix and calculated the mean accuracy of 0.977 ± 0.02 (0.975 ± 0.018 for the training and test sets, respectively). To assess for spatial overlap, we measured the mean intersection over union and mean boundary contour scores and found excellent overlap between the manual and automated methods with mean scores of 0.954 ± 0.030, 0.987 ± 0.009, and 0.948 ± 0.039 (0.983 ± 0.013 for the training and test set, respectively). Using these automated measurements, we found that body composition features were predictive of mortality in patients with cirrhosis. On multivariate analysis, the addition of body composition measures significantly improved prediction of mortality for patients with cirrhosis over Model for End‐Stage Liver Disease alone (P

Introduction: Cardiovascular disease remains the leading cause of death in the United States. Aortic calcification shows strong promise towards preventative screening efforts for cardiovascular risk in clinical and subclinical populations. While still relatively novel, granular distributions of aortic calcium in reference populations are absent from the literature. Methods: All participants of the Reference Analytic Morphomics Population, an approximation of vehicle occupants in the United States, with abdominal scans were considered for this analysis (n=4080). Analytic Morphomics was used to obtain vertebral-indexed anatomical measurements of the aorta and calcium characteristics (area, volume, density, and aortic wall obfuscation). Calcification was specifically identified as regions with a given morphology and pixel value three standard deviations above the defined central lumen zone. Results: Participant mean age was 33 (SD 22.6) and were mostly male (62.6%) due to the prevalence of males involved in trauma. Age and calcification distributions were right skewed among all vertebral levels and highest for the L3-L4 reference levels [~3% wall obfuscation (SD 10.5-11.3)]. Males demonstrated a progressive increase in total calcium volume from the 4 th decade of age, while females had comparatively smaller increases and a dramatic increase in total calcium after the 7 th decade. The largest increases in observed calcium volumes were observed in the 8 th decade for males (~2300 cm 3 ) and 7 th for females (~4500 cm 3 ). Conclusion: Aortic calcium prevalence was higher with reference to lower vertebral levels. Dramatic differential distributions of calcium were observed by age and sex.

Background Although imaging, endoscopy, and inflammatory biomarkers are associated with future Crohn disease (CD) outcomes, common laboratory studies may also provide prognostic opportunities. We evaluated machine learning models incorporating routinely collected laboratory studies to predict surgical outcomes in U.S. Veterans with CD. Methods Adults with CD from a Veterans Health Administration, Veterans Integrated Service Networks (VISN) 10 cohort examined between 2001 and 2015 were used for analysis. Patient demographics, medication use, and longitudinal laboratory values were used to model future surgical outcomes within 1 year. Specifically, data at the time of prediction combined with historical laboratory data characteristics, described as slope, distribution statistics, fluctuation, and linear trend of laboratory values, were considered and principal component analysis transformations were performed to reduce the dimensionality. Lasso regularized logistic regression was used to select features and construct prediction models, with performance assessed by area under the receiver operating characteristic using 10-fold cross-validation. Results We included 4950 observations from 2809 unique patients, among whom 256 had surgery, for modeling. Our optimized model achieved a mean area under the receiver operating characteristic of 0.78 (SD, 0.002). Anti-tumor necrosis factor use was associated with a lower probability of surgery within 1 year and was the most influential predictor in the model, and corticosteroid use was associated with a higher probability of surgery. Among the laboratory variables, high platelet counts, high mean cell hemoglobin concentrations, low albumin levels, and low blood urea nitrogen values were identified as having an elevated influence and association with future surgery. Conclusions Using machine learning methods that incorporate current and historical data can predict the future risk of CD surgery.

INTRODUCTION There is increasing recognition of the central role of muscle mass in predicting clinical outcomes in patients with liver disease. Muscle size can be extracted from computed tomography (CT) scans, but clinical implementation will require increased automation. We hypothesize that we can achieve this by using artificial intelligence. METHODS Using deep convolutional neural networks, we trained an algorithm on the Reference Analytic Morphomics Population (n = 5,268) and validated the automated methodology in an external cohort of adult kidney donors with a noncontrast CT scan (n = 1,655). To test the clinical usefulness, we examined its ability to predict clinical outcomes in a prospectively followed cohort of patients with clinically diagnosed cirrhosis (n = 254). RESULTS Between the manual and automated methodologies, we found excellent inter-rater agreement with an intraclass correlation coefficient of 0.957 (confidence interval 0.953-0.961, P < 0.0001) in the adult kidney donor cohort. The calculated dice similarity coefficient was 0.932 ± 0.042, suggesting excellent spatial overlap between manual and automated methodologies. To assess the clinical usefulness, we examined its ability to predict clinical outcomes in a cirrhosis cohort and found that automated psoas muscle index was independently associated with mortality after adjusting for age, gender, and child's classification (P < 0.001). DISCUSSION We demonstrated that deep learning techniques can allow for automation of muscle measurements on clinical CT scans in a diseased cohort. These automated psoas size measurements were predictive of mortality in patients with cirrhosis showing proof of principal that this methodology may allow for wider implementation in the clinical arena.

BackgroundIn response to the development of highly effective but expensive new medications, policymakers, payors, and health systems are considering novel and pragmatic ways to provide these medications to patients. One approach is to target these treatments to those most likely to benefit. However, to maximize the fairness of these policies, and the acceptance of their implementation, the values and beliefs of patients should be considered. The provision of treatments for chronic hepatitis C (CHC) in the resource-constrained context of the Veterans Health Administration (VHA) offered a real-world example of this situation, providing the opportunity to test the value of using Democratic Deliberation (DD) methods to solicit the informed opinions of laypeople on this complex issue.MethodsWe recruited Veterans (n = 30) from the VHA to attend a DD session. Following educational presentations from content experts, participants engaged in facilitated small group discussions to: 1) identify strategies to overcome CHC treatment barriers and 2) evaluate, vote on, and modify/improve two CHC treatment policies – “first come, first served” (FCFS) and “sickest first” (SF). We used transcripts and facilitators’ notes to identify key themes from the small group discussions. Additionally, participants completed pre- and post-DD surveys.ResultsMost participants endorsed the SF policy over the FCFS policy, emphasizing the ethical and medical appropriateness of treating the sickest first. Concerns about SF centered on the difficulty of implementation (e.g., how is “sickest” determined?) and unfairness to other Veterans. Proposed modifications focused on: 1) the need to consider additional health factors, 2) taking behavior and lifestyle into account, 3) offering education and support, 4) improving access, and 5) facilitating better decision-making.ConclusionsDD offered a robust and useful method for addressing the allocation of the scarce resource of CHC treatment. Participants were able to develop a modified version of the SF policy and offered diverse recommendations to promote fairness and improve quality of care for Veterans. DD is an effective approach for incorporating patient preferences and gaining valuable insights for critical healthcare policy decisions in resource-limited environments.

Sarcopenia has been shown to be an independent negative predictor in various diseases. The measurement of pre-defined criteria of skeletal muscle in patients with acute disease is usually unavailable. Therefore, we evaluate the psoas muscle area based on computed tomography (CT) imaging as an alternative for sarcopenia in an Asian trauma population. 939 trauma patients were enrolled and had CT imaging performed primarily for trauma indications. The cross-sectional area of psoas muscle at the base of the fourth lumbar vertebral was measured on these CTs. Psoas muscle index (PMI) was calculated and analyzed to determine sex-specific cut-off values to define the “extremely low psoas muscle index” (ELPMI) group. Psoas muscle index was significantly higher in males (1065.09 ± 230.51 mm2/m2 in males vs 719.57 ± 147.39 mm2/m2 in females, p

To analyze visual outcomes after treatment of choroidal hemangioma in the pre-photodynamic therapy (PDT) era versus PDT era.Retrospective, comparative case series.A total of 458 patients with circumscribed choroidal hemangioma.Comparison of hemangioma managed in the pre-PDT (1967-2001) era versus PDT (2002-2018) era.Visual acuity outcome.A total of 458 tumors were treated over this 51-year period. A comparison (pre-PDT [n = 220 cases] vs. PDT [n = 238 cases]) revealed PDT era patients were of older mean age (48.9 vs. 53.8 years, P = 0.002) and were more likely to have systemic hypertension (17.7% vs. 33.8%, P0.001), tumor location in the macula (57.4% vs. 67.5%, P = 0.01), subretinal fluid on OCT (33.3% vs. 70.7%, P = 0.01), and greater extent of overlying lipofuscin (P = 0.001). Findings of tumor basal diameter and thickness and fluorescein and indocyanine green angiography were no different in the 2 eras. Treatment (pre-PDT vs. PDT) included argon laser photocoagulation (42.1% vs. 0.4%), PDT (0% vs. 43.8%), transpupillary thermotherapy (0% vs. 0.4%), plaque radiotherapy (7.0% vs. 5.2%), external beam radiotherapy (1.4% vs. 1.3%), enucleation (0.9% vs. 0.4%), and observation (48.6% vs. 47.6%). After treatment, patients in the PDT era demonstrated better mean logarithm of the minimum angle of resolution visual acuity (1.28 vs. 0.51, P0.001) (Snellen equivalent 20/400 vs. 20/63, P0.001). Final visual acuity was ≥20/40 for those with entering vision of ≥20/40 (59.6% vs. 74.7%, P = 0.001) and for those with entering vision of 20/50-20/200 (25.4% vs. 47.3%, P0.001).Management of choroidal hemangioma in the PDT era has allowed for significantly better visual outcome compared with the pre-PDT era, with mean final visual acuity of 20/400 (pre-PDT era) versus 20/63 (PDT era).

BACKGROUND & AIMS: Computed tomography (CT) provides scans of the human body from which digitized features can be extracted. The aim of this study was to examine the role of these digital biomarkers for predicting subsequent occurrence of hepatocellular carcinoma (HCC) in cirrhotic patients. METHODS: A cohort of 269 patients with cirrhosis were recruited and prospectively followed for the occurrence of HCC in Taiwan. CT scans were retrospectively retrieved and computationally processed using analytic morphomics. A predictive score was constructed using Cox regression and the generalized iterative modeling method, maximizing the log-likelihood of the time to HCC development. An independent cohort of 274 patients from University of Michigan was utilized to examine the predictive validity of this score in a Western population. RESULTS: Of the 27 digitized features at the 12th thoracic vertebral level, 6 features were significantly associated with HCC occurrence. Two digitized features (fascia eccentricity and the bone mineral density) was able to stratify patients into high and low-risk groups with distinct cumulative incidence of HCC in both the training and validation cohorts (P = 0.015 and 0.044 respectively). When the two digitized features were tested in the Michigan cohort, only bone mineral density remained an effective predictor. CONCLUSION: Digitized features derived from the CT were effective in predicting subsequent occurrence of HCC in cirrhosis patients. The bone mineral density measured on CT was an effective predictor for patients in both Taiwan and USA.

BACKGROUND: Advances in imaging technology have the potential to transform the early diagnosis and treatment of hepatocellular carcinoma (HCC) through quantitative image analysis. Computational ‘radiomic’ techniques extract biomarker information from images which can be used to improve diagnosis and predict tumor biology. AIMS: We performed a systematic review of literature on radiomic features in HCC diagnosis and prognosis, with a focus on reporting metrics and methodologic standardization. METHODS: A systematic review was performed of all full-text articles published from inception through December 1, 2019. Standardized data extraction and quality assessment metrics were applied to all studies. RESULTS: A total of 54 unique studies were included for analysis. Radiomic features demonstrated good discriminatory performance to differentiate HCC from other solid lesions (c-statistics 0.66–0.95), predict microvascular invasion (c-statistic 0.76–0.92), predict early recurrence after hepatectomy (c-statistics 0.71–0.86), and predict prognosis after locoregional or systemic therapies (c-statistics 0.74–0.81). Common stratifying features for diagnostic and prognostic radiomic tools included analyses of imaging skewness, analysis of the peritumoral region, and feature extraction from the arterial imaging phase. The overall quality of the included studies was low, with common deficiencies in both internal and external validation, standardized imaging segmentation, and lack of comparison to a gold standard. CONCLUSIONS: Quantitative image analysis demonstrates promise as a non-invasive biomarker to improve HCC diagnosis and management. However, standardization of protocols and outcome measurement, sharing of algorithms and analytic methods, and external validation are necessary prior to widespread application of radiomics to HCC diagnosis and prognosis in clinical practice.

Background Aortic wall calcification shows strong promise as a cardiovascular risk factor. While useful for visual enhancement of vascular tissue, enhancement creates heterogeneity between scans with and without contrast. We evaluated the relationship between aortic calcification in routine abdominal computed tomography scans (CT) with and without contrast. Methods Inclusion was limited to those with abdominal CT-scans with and without contrast enhancement within 120 days. Analytic Morphomics, a semi-automated computational image processing system, was used to provide standardized, granular, anatomically indexed measurements of aortic wall calcification from abdominal CT-scans. Aortic calcification area (ACA) and aortic wall calcification percent (ACP) and were the outcomes of interest. Multiple linear regression was used to evaluate the relationship of aortic measurements. Models were further controlled for age and sex. Stratification of measurements by vertebral level was also performed. Results A positive association was observed for non-contrast calcification in ACP β 0.74 (95% CI 0.72, 0.76) and ACA β 0.44 (95% 0.43, 0.45). Stratified results demonstrated the highest coefficient of determination at L2 for percent and L3 for area models [R2 0.91 (ACP) 0.74 (ACA)]. Adjusted lumber-level associations between non-contrast and contrast measurements ranged from (β 0.69–0.82) in ACP and (β 0.37–0.54) in ACA. Conclusion A straightforward correction score for comparison of abdominal aortic calcification measurements in contrast-enhanced and non-contrast scans is discussed. Correction of aortic calcification from CT scans can reduce scan heterogeneity and will be instrumental in creating larger cardiovascular cohorts as well as cardiovascular risk surveillance programs.

Frailty and sarcopenia are associated with mortality and poor outcomes among patients with cirrhosis. Frailty is multifactorial but due in part to sarcopenia and cognitive dysfunction. Data are limited regarding the correlation of bedside frailty and cognitive function measures with sarcopenia. To evaluate the correlations between frailty measures and muscle indices from computed tomography (CT). We prospectively enrolled 106 patients with clinically compensated cirrhosis (and no prior hepatic encephalopathy). All patients underwent CT scan and cognitive testing (via inhibitory control test, ICT), and were subject to hand grip, 30-s chair stands, mid-arm muscle area (MAMA), and a four-question algorithm based on the Sickness Impact Profile (SIP) predictive of minimal HE. We evaluated Spearman correlations between all measures as well as the sensitivity and specificity of each measure for falls. In total, 106 (35.3%) patients (55 men) had CT scans to measure skeletal muscle area and quality. Hand grip correlated strongly with skeletal muscle area (correlation coefficient 0.64, p

Purpose: To investigate features and outcomes of circumscribed choroidal hemangioma by patient age. Methods: Retrospective review of circumscribed choroidal hemangioma from 3/29/1967–6/4/2018 based on age at presentation (≤20 vs. >20–50 vs. >50 years). Results: There were 458 circumscribed choroidal hemangiomas diagnosed at mean age (13 vs. 41 vs. 64 years, p

Introduction: Aortic calcification can be utilized to assess cardiovascular risk. While contrast is useful for vascular enhancement in diagnostic imaging, enhancement creates heterogeneity between post and non-contrast scans and limits their direct comparability. Hypothesis: We hypothesized that post and non-contrast aortic calcification measures will correlate, and a correction score can be developed for statistical comparability. Methods: Retrospective CT-scans were obtained from the University of Michigan. Participants (N=330) received abdominal scans with and without contrast enhancement within 120 calendar days. Analytic Morphomics was used to obtain vertebral-indexed measurements of aortic calcium area, and aortic wall obfuscation percentage. Calcification was specifically identified as regions with a given morphology and pixel value five standard deviations above the defined central lumen zone. Pearson correlation and multiple linear regression were used to explain the relationship between aortic measurements with and without contrast. Regressions include calcification percent (Model 1), and area (Model 2). Independent variables were non-contrast measurements and dependent variables were contrast measurements, age, and sex. Results: Correlations of calcification percent ranged from 0.86 at T11 and 0.94 and L2. Correlations of calcification area ranged from 0.66 at T12 to 0.84 at L3. In Model 1, for every percent increase in post-contrast calcification, non-contrast calcification percent increased by 11% (β=1.11, p Conclusion: In conclusion, this research proposes a correction score for comparisons of abdominal aortic calcification measurements in post-contrast and non-contrast scans.

Key Points Question Can deep learning recurrent neural network (RNN) models using raw longitudinal data extracted directly from electronic health records outperform conventional regression models in predicting the risk of developing hepatocellular carcinoma (HCC)? Findings This prognostic study included 48 151 patients with hepatitis C virus (HCV)–related cirrhosis in the national Veterans Health Administration who had at least 3 years of follow-up after the diagnosis of cirrhosis. Deep learning RNN models outperformed conventional linear regression models and could be used to identify patients with HCV-related cirrhosis at high risk of developing HCC. Meaning The findings of this study suggest that RNN models could have multiple applications in clinical practice and could be applied to HCC outreach and surveillance strategies., This prognostic study examines whether deep learning recurrent neural network (RNN) models that use raw longitudinal data extracted directly from electronic health records outperform conventional regression models in predicting the risk of developing hepatocellular carcinoma (HCC)., Importance Deep learning, a family of machine learning models that use artificial neural networks, has achieved great success at predicting outcomes in nonmedical domains. Objective To examine whether deep learning recurrent neural network (RNN) models that use raw longitudinal data extracted directly from electronic health records outperform conventional regression models in predicting the risk of developing hepatocellular carcinoma (HCC). Design, Setting, and Participants This prognostic study included 48 151 patients with hepatitis C virus (HCV)–related cirrhosis in the national Veterans Health Administration who had at least 3 years of follow-up after the diagnosis of cirrhosis. Patients were identified by having at least 1 positive HCV RNA test between January 1, 2000, to January 1, 2016, and were followed up from the diagnosis of cirrhosis to January 1, 2019, for the development of incident HCC. A total of 3 models predicting HCC during a 3-year period were developed and compared, as follows: (1) logistic regression (LR) with cross-sectional inputs (cross-sectional LR); (2) LR with longitudinal inputs (longitudinal LR); and (3) RNN with longitudinal inputs. Data analysis was conducted from April 2018 to August 2020. Exposures Development of HCC. Main Outcomes and Measures Area under the receiver operating characteristic curve, area under the precision-recall curve, and Brier score. Results During a mean (SD) follow-up of 11.6 (5.0) years, 10 741 of 48 151 patients (22.3%) developed HCC (annual incidence, 3.1%), and a total of 52 983 samples (51 948 [98.0%] from men) were collected. Patients who developed HCC within 3 years were older than patients who did not (mean [SD] age, 58.2 [6.6] years vs 56.9 [6.9] years). RNN models had superior mean (SD) area under the receiver operating characteristic curve (0.759 [0.009]) and mean (SD) Brier score (0.136 [0.003]) than cross-sectional LR (0.689 [0.009] and 0.149 [0.003], respectively) and longitudinal LR (0.682 [0.007] and 0.150 [0.003], respectively) models. Using the RNN model, the samples with the mean (SD) highest 51% (1.5%) of HCC risk, in which 80% of all HCCs occurred, or the mean (SD) highest 66% (1.2%) of HCC risk, in which 90% of all HCCs occurred, could potentially be targeted. Among samples from patients who achieved sustained virologic response, the performance of the RNN models was even better (mean [SD] area under receiver operating characteristic curve, 0.806 [0.025]; mean [SD] Brier score, 0.117 [0.007]). Conclusions and Relevance In this study, deep learning RNN models outperformed conventional LR models, suggesting that RNN models could be used to identify patients with HCV-related cirrhosis with a high risk of developing HCC for risk-based HCC outreach and surveillance strategies.

Background: Frailty has been shown to be an independent negative predictor of surgical outcomes in geriatric patients. Traditional measurements of frailty are impractical in emergency settings, and computed tomography (CT)-measured skeletal muscle mass has been proposed as an alternative. However, the cutoff values of these CT metrics for frailty are still unknown, and their impact on abdominal emergencies in the elderly population is unclear.Study Design: A total of 462 young trauma patients aged 18-40 years were analyzed to establish sex-specific reference cutoff values for the CT-measured muscle index (MI) and muscle gauge (MG) values. The impacts of low MI and MG values were investigated in 1192 elderly patients (aged ≥ 65 years) undergoing abdominal surgery.Results: The sex-specific cutoff values for MI and MG were determined by adopting European Working Group on Sarcopenia in Older People (EWGSOP) guidelines. The correlation between MG and ageing was significantly stronger than that between MI and ageing. With regard to the MG, the L4 psoas muscle gauge (L4 PMG) was further investigated in an elderly cohort owing to its high predictive value and ease of use in the clinical setting. A low L4 PMG value was an independent risk factor for overall complications and mortality in elderly patients with abdominal emergencies.Conclusion: The current study was the largest study investigating the correlations between MG values and ageing in the Asian population. Frailty, as indicated by a low L4 PMG value, may help surgeons during preoperative decision making regarding geriatric patients with abdominal emergencies.

Background High-dose methotrexate (HD-MTX) is used to treat primary central nervous system lymphoma (PCNSL), but potential differences in MTX clearance (CL) due to obesity have not been studied. We characterized the relationship between HD-MTX CL and computed tomography (CT)-generated body composition (morphomic), body size descriptors, and laboratory measurements in a cohort of obese and non-obese patients with PCNSL. Methods Medical records from adult patients with PCNSL treated with HD-MTX over a 10-year period were queried. Individuals with CT data within 30 days of the first cycle of treatment were included. Population pharmacokinetic analysis was performed using a 2-compartment base structural model. We specifically compared body surface area (BSA) to standard body size, morphomic, and renal function estimation methods as covariates of HD-MTX CL. Results The final data set consisted of non-obese (n=45) and obese (n=28) patients with 291 observations (3-7 samples per patient) with a mean (standard deviation) weight of 69.8 (11.6) kg and 104 (14.9) kg, respectively (p=0.0001). Vertebral body height was more informative than BSA of MTX CL. Similarly, a CL model incorporating age, albumin, and serum creatinine was more informative than kidney function equations and body size. The final model of MTX CL was based on age, albumin, serum creatinine, and vertebral body height. Conclusions Common clinical variables coupled with vertebral body height are more predictive of first cycle MTX CL than BSA, alternate body size descriptors, and commonly used kidney function equations.

Background & Aims Body composition, particularly sarcopenia, is associated with mortality in patients with decompensated cirrhosis undergoing transplant evaluation. Similar data are limited for non-transplant eligible or compensated patients. Methods A total of 274 patients with cirrhosis were followed prospectively for ≤5 years after a CT scan. We utilized Analytic Morphomics® to measure body composition (fat, muscle, and bone) which was rendered into relative values (percentiles) in relation to a reference population. The model for end-stage liver disease (MELD) score was used as a reference model for survival prediction. We validated our models in a separate cohort. Results Our cohort had a mean Child-Pugh score of 7.0 and a mean MELD of 11.3. The median follow-up time was 5.05 years. The proportion of patients alive at 1, 3 and 5 years was 86.5%, 68.0%, and 54.3%; 13 (4.6%) underwent liver transplantation. Child-Pugh B/C (vs. A) cirrhosis was associated with decreased muscle, subcutaneous, and visceral fat area but increased subcutaneous/visceral fat density. Decreased normal density muscle mass was associated with mortality (hazard ratio [HR] 0.984, p, Graphical abstract, Highlights • Features of body composition can predict clinical outcomes in patients with cirrhosis awaiting liver transplantation. • Data are lacking regarding long-term outcomes among patients with compensated disease. • We show that features of muscle and fat are associated with decompensation and risk of death across the spectrum of cirrhosis. • CT scans obtained for unrelated clinical purposes can be analyzed as a digital risk biomarker for patients with compensated cirrhosis.

Measurements of skeletal muscle cross-sectional area, index, and radiation attenuation utilizing clinical computed tomography (CT) scans are used in assessments of sarcopenia, the loss of skeletal muscle mass and function associated with aging. To classify individuals as sarcopenic, sex-specific cutoffs for ‘low’ values are used. Conventionally, cutoffs for skeletal muscle measurements at the level of the third lumbar (L3) vertebra are used, however L3 is not included in several clinical CT protocols. Non-contrast-enhanced CT scans from healthy kidney donor candidates (age 18–40) at Michigan Medicine were utilized. Skeletal muscle area (SMA), index (SMI), and mean attenuation (SMRA) were measured at each vertebral level between the tenth thoracic (T10) and the fifth lumbar (L5) vertebra. Sex-specific means, standard deviations (s.d.), and sarcopenia cutoffs (mean-2 s.d.) at each vertebral level were computed. Associations between vertebral levels were assessed using Pearson correlations and Tukey’s difference test. Classification agreement between different vertebral level cutoffs was assessed using overall accuracy, specificity, and sensitivity. SMA, SMI, and SMRA L3 cutoffs for sarcopenia were 92.2 cm2, 34.4 cm2/m2, and 34.3 HU in females, and 144.3 cm2, 45.4 cm2/m2, and 38.5 HU in males, consistent with previously reported cutoffs. Correlations between all level pairs were statistically significant and high, ranging from 0.65 to 0.95 (SMA), 0.64 to 0.95 (SMI), and 0.63 to 0.95 (SMRA). SMA peaks at L3, supporting its use as the primary site for CT sarcopenia measurements. However, when L3 is not available alternative levels (in order of preference) are L2, L4, L5, L1, T12, T11, and T10. Healthy reference values reported here enable sarcopenia assessment and sex-specific standardization of SMA, SMI, and SMRA in clinical populations, including those whose CT protocols do not include L3.

Over 40% of patients with cirrhosis will develop hepatic encephalopathy (HE). HE is associated with decreased survival, falls, motor vehicle accidents, and frequent hospitalization. Accordingly, we aimed to develop a tool to risk-stratify patients for HE development. We studied a population-based cohort of all patients with cirrhosis without baseline HE (N=1,979) from the Veterans Administration from Michigan, Indiana, and Ohio (1/1/2005-12/31/10) using demographic, clinical, laboratory, and pharmacy data. The primary outcome was the development of HE. Risk-scores were constructed with both baseline and longitudinal data (annually updated parameters) and validated using bootstrapping. The cohort had mean age of 58.0±8.3 years, 36% had hepatitis C, 17% had ascites. Opiates, benzodiazepines, statins, and nonselective beta-blockers were taken at baseline by 24%, 13%, 17%, and 12%. Overall, 863(43.7%) developed HE within 5 years. In multivariable models, risk factors (HR, 95%CI) for HE included higher bilirubin (1.07, 1.05-1.09) and nonselective beta-blocker use (1.34, 1.09-1.64), while higher albumin (0.54, 0.48-0.59) and statin use (0.80, 0.65-0.98) were protective. Other clinical factors, including opiate and benzodiazepine use were not predictive. The AUROC for HE using the 4 significant variables in baseline and longitudinal models were 0.68 (0.66-0.70) and 0.73 (0.71-0.75), respectively. Model effects were validated and converted into a risk score. A score ≤0 in our longitudinal model assigns a 6% 1-year probability of HE while a score >20 assigns a 38% 1-year risk. Conclusion: Patients with cirrhosis can be stratified by a simple risk-score for HE that accounts for changing clinical data. Our data also highlight a role for statins in reducing cirrhosis complications including HE. This article is protected by copyright. All rights reserved.

Purpose The prognosis of patients with hepatocellular carcinoma (HCC) undergoing transarterial chemoembolization (TACE) is often uncertain. We aimed to utilize analytic morphomics, a high-throughput imaging analysis, to assess if body composition is predictive of post-TACE survival. Materials and Methods We included patients from a single center (Ann Arbor VA)who had TACE as the primary treatment forHCC and had a pre-treatment computed tomography scans. Univariate analysis and multivariate conditional inference tree analysis were utilized to identify the morphomic characteristics predictive of 1-year survival. Results were validated in an external cohort(University of MichiganHealth System) ofHCC patientswho underwent TACE as their primary treatment. Results In the 75 patients in the derivation cohort, median survival was 439 (interquartile range, 377 to 685) days from receipt of TACE, with 1-year survival of 61%. Visceral fat density (VFD) was the only morphomic factor predictive of overall and 1-year survival (p < 0.001). Patients with VFD above the 56th percentile had a 1-year survival of 39% versus 78% for those below the 56th percentile. VFD also correlated with 1-year survival in the external validation cohort (44% vs. 72%, p < 0.001). In a secondary analysis, patients with higher VFD were significantly more likely to experience hepatic decompensation after TACE (p < 0.001). Conclusion VFD served as an objective predictor of mortality in patients undergoing TACE, possibly through its ability to predict hepatic decompensation. VFD may serve as a radiographic biomarker in predicting TACE outcomes.

Introduction: The severity of injury from motor vehicle crashes (MVCs) depends on complex biomechanical factors, and the bodily features of the injured person account for some of these factors. By assuming that vulnerable road users (VRUs) have limited protection resulting from vehicles and safety equipment, the current study analyzed the characteristics of fat distribution measured by computed tomography (CT) imaging and investigated the existence of a “cushion effect” in VRUs. Materials and Methods: This retrospective study enrolled 592 VRUs involved in MVCs who underwent CT scans. Visceral fat area and subcutaneous fat cross-sectional area were measured and adjusted according to total body area (TBA) and are presented as the visceral fat ratio and the subQ fat ratio (subcutaneous fat ratio). Risk factors for serious abdominal injury (maximum abbreviated injury scale (MAISabd ≥ 3)) resulting from MVCs were determined by univariate and multivariate analysis. Results: MAISabd ≥ 3 was observed in 104 (17.6%) of the patients. The subQ fat ratio at the L4 vertebral level was significantly lower in the MAISabd ≥ 3 group than in the MAISabd &lt, 3 group (24.9 ± 12.0 vs. 28.1 ± 11.9%, p = 0.015). A decreased L4 subQ fat ratio was associated with a higher risk for MAISabd ≥ 3 in multivariate analysis (odds ratio 0.063, 95% CI 0.008–0.509, p = 0.009). Conclusion: The current study supported the “cushion effect” theory, and protection was apparently provided by subcutaneous fat tissue. This concept may further improve vehicle and safety designation in the future.

Grace L Su,1 Douglas M Baughman,2 Qinqin Zhang,3 Kasra Rezaei,2 Aaron Y Lee,2 Cecilia S Lee2 1Lewis Katz School of Medicine, Temple University, Philadelphia, PA, 2Department of Ophthalmology, 3Department of Bioengineering, University of Washington, Seattle, WA, USA Purpose: The aim of this study was to compare physician preferences regarding the commercially available spectral-domain (SD) optical coherence tomography angiography (OCTA) and swept-source (SS) OCTA prototype device.Design: Comparative analysis of diagnostic instruments was performed.Patients and methods: Subjects at the University of Washington Eye Institute and Harborview Medical Center were prospectively recruited and imaged with the Zeiss SD OCTA (HD-5000, Angioplex) and Zeiss SS OCTA (Plex Elite, Everest) devices on the same day. The study included 10 eyes from 10 subjects diagnosed with a retinal/choroidal disease. Deidentified images were compiled into a survey and sent to retina specialists in various countries. The survey presented masked SD and SS images of each eye for each retinal sublayer side by side. Respondents were asked about their image preference and impact on clinical management. A priori and post hoc preferences for SD vs SS were collected.Results: Fifty-four retina specialists responded to the survey. Median years in practice was3.00 (interquartile range [IQR] 1.50–17.00). At baseline, 23 (48%) physicians owned an OCTA machine. The majority of physician responses showed a preference for the SS over SD OCTA, independent of the retinal pathology shown (n=454 overall responses, 74%). Nevertheless, themajority indicated that both SD and SS would be equally valuable in informing clinical decisions (n=374 overall responses, 61%).Conclusion: These findings indicate that the majority of retina specialists surveyed prefer SS over SD OCTA based on image quality, regardless of the retinal pathology shown. Regarding the clinical utility of each modality, the majority of physicians perceive SD and SS as equallyeffective. Keywords: swept-source optical coherence tomography angiography, spectral-domain optical coherence tomography angiography, physician preference