Management Of Hospitalized Acute Kidney Injury Based On Data Mining And Intelligent Alerts

OBJECTIVEThe present study aimed to investigate the burden of hospitalized acute kidney injury through data mining,especially the hyperkalemia associated with acute kidney injury.We further explored the clinical value of alerts,and clinical practices of acute kidney injury.METHODS?Utilizing the electronic medical information of the patients in the cardiology divisions,we conducted a retrospective case-control study to observe the incidence of hyperkalaemia associated with acute kidney injury,and analyze the risk factors.?A prospective,randomized,controlled study was designed to investigate the value of electronic alerts,applied in intensive care units and cardiovascular divisions.?The online questionnaire-based cross-sectional survey was used to investigate the Chinese doctors' attitudes towards acute kidney injury and its management in China.RESULTS?37,837 patients were included in this study.1571(4.3%)patients with AKI were detected.The detection rates of acute kidney injury stage 1-3 were 3.6%,0.3%and 0.4%,respectively.Hyperkalaemia occurred in 517 patients(1.4%).The incidence of hyperkalaemia in patients with acute kidney injury was higher than that in patients without acute kidney injury(10.1%vs.1.0%,P<0.001).The incidences of hyperkalaemia in patient with acute kidney injury stage 1-3 were 2.6%,13.9%and 20.6%,respectively.Multiple logistic regression analysis demonstrated that the acute kidney injury stages were risk factors for hyperkalaemia.Acute kidney injury and hyperkalaemia were related to delayed hospital stay,increased hospitalization expenses,renal replacement therapy and in-hospital mortality.Based on acute kidney injury,the combination of hyperkalemia could significantly increase clinical burden and adverse outcomes.?We ran electronic alerts in intensive care units and cardiovascular divisions.Artificial diagnosis of acute kidney injury was considered as "gold standard".The system automatically diagnosed the acute kidney injury by serum creatinine.Patients were randomly assigned to alert group(467 patients)and non-alert group(408 patients).Only the alerrt group could receive the alert pop-up windows.The sensitivity and specificity were 99.8%and 97.7%respectively.Youden index and accuracy were 97.5%and 98.1%,respectively.The diagnosis rates of acute kidney injury(7.9%and 2.7%,P= 0.001)and expanded acute kidney injury(acute kidney injury and multiple organ failure,16.3 and 6.1%,P<0.001)in alert group were higher than that in non-alert group.The prevalence of nephrology consultation in the alert group was higher than that in the non-alert group(9:0%and 3.7%,P=0.001).There was no significant difference in the prevalence dialysis,rehabilitation of renal function or death in the two groups.?From February 20 to August 15,2017,We received 1,289 respondents of the questionnaires.Among the 1,289 respondents from secondary and tertiary hospitals in 30 provinces,718(55.7%)were nephrologists,94.3%had the ability to evaluate glomerular filtration rates,and 98.8%could evaluate urinary protein excretion.However,nearly half of all respondents reported that easy methods for spot urine creatinine-adjusted urinary protein assessments were unavailable.Only 54.2%of respondents reported that they had received nutritional education for renal diseases.66.4%and 76.3%of respondents reported nephrology referrals and nephrology consultations.CONCLUSIONSBy analyzing the information from acute kidney injury in high-risk wards,the present study found that acute kidney injury and hyperkalemia were associated with adverse clinical outcomes and increased medical burden.Although the electronic alert improved the diagnosis rate of acute kidney injury,it had no significant effect on the clinical prognosis.Combined with the results from the questionnaire,it was indicated that it may associated with the insufficient knowledge and training of acute kidney injury from clinical physicians.Therefore,it is necessary to strengthen clinical training and develop clinical protocols to further obtain the potential benefits of the electronic alert in acute kidney injury.