Electronic alert outpatient protocol improves the quality of care for the risk of postcontrast acute kidney injury following computed tomography.

BACKGROUND: Prevention and diagnosis of postcontrast acute kidney injury (AKI) after contrast-enhanced computed tomography is burdensome in outpatient department. We investigated whether an electronic alert system could improve prevention and diagnosis of postcontrast AKI. METHODS: In March 2018, we launched an electronic alert system that automatically identifies patients with a baseline estimated glomerular filtration rate of <45 mL/min/1.73 m2, provides a prescription of fluid regimen, and recommends a follow-up for serum creatinine measurement. Participants prescribed contrast-enhanced computed tomography at outpatient department before and after the launch of the system were categorized as historical and alert group, respectively. Propensity for the surveillance of postcontrast AKI was compared using logistic regression. Risks of AKI, admission, mortality, and renal replacement therapy were analyzed. RESULTS: The historical and alert groups included 289 and 309 participants, respectively. The alert group was more likely to be men and take diuretics. The most frequent volume of prophylactic fluid in historical and alert group was 1,000 and 750 mL, respectively. Follow-up for AKI was more common in the alert group (adjusted odds ratio, 6.00; p < 0.001). Among them, incidence of postcontrast AKI was not statistically different. The two groups did not differ in risks of admission, mortality, or renal replacement therapy. CONCLUSION: The electronic alert system could assist in the detection of high-risk patients, prevention with reduced fluid volume, and proper diagnosis of postcontrast AKI, while limiting the prescribing clinicians' burden. Whether the system can improve long-term outcomes remains unclear.

Spatial autocorrelation may bias the risk estimation: An application of eigenvector spatial filtering on the risk of air pollutant on asthma.

Air pollutants are major risk factors for respiratory diseases, particularly asthma, socially and spatially correlated. Many existing environment-asthma-related studies, however, have evaluated the impact of crude trends at the largest district level, which accounts only for temporal effects and may produce biased results with spatial autocorrelation. This study aimed to investigate how the spatial autocorrelation affects the air pollution effect estimations (sulfur dioxide [SO2], nitrogen dioxide [NO2], carbon monoxide [CO], and particulate matter [PM10]) on daily asthma emergency department (ED) visits in two metropolitan areas in Korea (Seoul Metropolitan Area [SMA] and Busan Metropolitan City, Ulsan Metropolitan City, Gyeongsangnamdo [BUG]). We applied eigenvector spatial filter (ESF) to the spatio-temporal model to remove spatial autocorrelation and distributed lag nonlinear model (DLNM) to explore nonlinear patterns between air pollutant concentration and lagged days on the three models including aggregated model (a temporal model), spatial model without ESF, and spatial model with ESF (both are spatio-temporal models). The effect of SO2 was not statistically significant for asthma ED visits in the aggregated model for SMA (cumulative relative risks [CRR] = 0.99, confidence intervals [CI]: 0.93-1.05), while the effect was statistically significant in the spatial model with ESF (CRR = 1.10, CI: 1.08-1.12). NO2 and CO were positively correlated to asthma ED visits in the spatial model without ESF (CRR = 0.84, CI: 0.81-0.86; 0.91, 0.89-0.94, respectively), but the spatial model with ESF showed significant risks (CRR = 1.21, CI: 1.18-1.24; 1.13, 1.11-1.16). Moreover, the spatial model with ESF successfully removed spatial autocorrelation (P-values for Moran's I 0.83-0.98) and demonstrated the highest model fit (McFadden's pseudo R2 0.42-0.43 for SMA and 0.26-0.27 for BUG) among the three models. Our findings demonstrate how ESF can be introduced into spatial correlation to remove bias and construct more reliable models.

Real-Time Clinical Decision Support Based on Recurrent Neural Networks for In-Hospital Acute Kidney Injury: External Validation and Model Interpretation.

BACKGROUND: Acute kidney injury (AKI) is commonly encountered in clinical practice and is associated with poor patient outcomes and increased health care costs. Despite it posing significant challenges for clinicians, effective measures for AKI prediction and prevention are lacking. Previously published AKI prediction models mostly have a simple design without external validation. Furthermore, little is known about the process of linking model output and clinical decisions due to the black-box nature of neural network models. OBJECTIVE: We aimed to present an externally validated recurrent neural network (RNN)-based continuous prediction model for in-hospital AKI and show applicable model interpretations in relation to clinical decision support. METHODS: Study populations were all patients aged 18 years or older who were hospitalized for more than 48 hours between 2013 and 2017 in 2 tertiary hospitals in Korea (Seoul National University Bundang Hospital and Seoul National University Hospital). All demographic data, laboratory values, vital signs, and clinical conditions of patients were obtained from electronic health records of each hospital. We developed 2-stage hierarchical prediction models (model 1 and model 2) using RNN algorithms. The outcome variable for model 1 was the occurrence of AKI within 7 days from the present. Model 2 predicted the future trajectory of creatinine values up to 72 hours. The performance of each developed model was evaluated using the internal and external validation data sets. For the explainability of our models, different model-agnostic interpretation methods were used, including Shapley Additive Explanations, partial dependence plots, individual conditional expectation, and accumulated local effects plots. RESULTS: We included 69,081 patients in the training, 7675 in the internal validation, and 72,352 in the external validation cohorts for model development after excluding cases with missing data and those with an estimated glomerular filtration rate less than 15 mL/min/1.73 m2 or end-stage kidney disease. Model 1 predicted any AKI development with an area under the receiver operating characteristic curve (AUC) of 0.88 (internal validation) and 0.84 (external validation), and stage 2 or higher AKI development with an AUC of 0.93 (internal validation) and 0.90 (external validation). Model 2 predicted the future creatinine values within 3 days with mean-squared errors of 0.04-0.09 for patients with higher risks of AKI and 0.03-0.08 for those with lower risks. Based on the developed models, we showed AKI probability according to feature values in total patients and each individual with partial dependence, accumulated local effects, and individual conditional expectation plots. We also estimated the effects of feature modifications such as nephrotoxic drug discontinuation on future creatinine levels. CONCLUSIONS: We developed and externally validated a continuous AKI prediction model using RNN algorithms. Our model could provide real-time assessment of future AKI occurrences and individualized risk factors for AKI in general inpatient cohorts; thus, we suggest approaches to support clinical decisions based on prediction models for in-hospital AKI.