Search

BackgroundApplications of machine learning for the early detection of diseases for which a clear-cut diagnostic gold standard exists have been evaluated. However, little is known about the usefulness of machine learning approaches in the decision-making process for decisions such as insulin initiation by diabetes specialists for which no absolute standards exist in clinical settings. ObjectiveThe objectives of this study were to examine the ability of machine learning models to predict insulin initiation by specialists and whether the machine learning approach could support decision making by general physicians for insulin initiation in patients with type 2 diabetes. MethodsData from patients prescribed hypoglycemic agents from December 2009 to March 2015 were extracted from diabetes specialists’ registries, resulting in a sample size of 4860 patients who had received initial monotherapy with either insulin (n=293) or noninsulin (n=4567). Neural network output was insulin initiation ranging from 0 to 1 with a cutoff of >0.5 for the dichotomous classification. Accuracy, recall, and area under the receiver operating characteristic curve (AUC) were calculated to compare the ability of machine learning models to make decisions regarding insulin initiation to the decision-making ability of logistic regression and general physicians. By comparing the decision-making ability of machine learning and logistic regression to that of general physicians, 7 cases were chosen based on patient information as the gold standard based on the agreement of 8 of the 9 specialists. ResultsThe AUCs, accuracy, and recall of logistic regression were higher than those of machine learning (AUCs of 0.89-0.90 for logistic regression versus 0.67-0.74 for machine learning). When the examination was limited to cases receiving insulin, discrimination by machine learning was similar to that of logistic regression analysis (recall of 0.05-0.68 for logistic regression versus 0.11-0.52 for machine learning). Accuracies of logistic regression, a machine learning model (downsampling ratio of 1:8), and general physicians were 0.80, 0.70, and 0.66, respectively, for 43 randomly selected cases. For the 7 gold standard cases, the accuracies of logistic regression and the machine learning model were 1.00 and 0.86, respectively, with a downsampling ratio of 1:8, which were higher than the accuracy of general physicians (ie, 0.43). ConclusionsAlthough we found no superior performance of machine learning over logistic regression, machine learning had higher accuracy in prediction of insulin initiation than general physicians, defined by diabetes specialists’ choice of the gold standard. Further study is needed before the use of machine learning–based decision support systems for insulin initiation can be incorporated into clinical practice.

Aim Materials and Methods Results Conclusions To investigate the association between low‐density lipoprotein‐cholesterol (LDL‐C) levels and coronary artery disease (CAD) incidence based on combining high‐density lipoprotein‐cholesterol (HDL‐C) levels and glucose status.In this retrospective cohort study, we used data from a nationwide claims database (1,524,289 individuals without a history of CAD or familial hypercholesterolaemia; 2008–2019). Cox proportional hazards modelling identified the risk of incident CAD by a novel combination of four HDL‐C levels, seven LDL‐C levels and glucose status.During the follow‐up period (mean: 5.5 years), 8301 (0.99/1000 person‐years) events occurred. The risk of CAD increased from lower LDL‐C levels accompanied by lower HDL‐C levels regardless of the glucose status. Using the most favourable levels of HDL‐C and LDL‐C (i.e. 60–99 mg/dL and <80 mg/dL, respectively) as references, the hazard ratios (95% confidence interval) for the group with HDL‐C levels <40 mg/dL and LDL‐C levels <80 mg/dL were 2.74 (1.47–5.11), 2.52 (1.30–4.91) and 2.85 (1.68–4.84) for normoglycaemia, borderline glycaemia and diabetes, respectively. Comparison of the most favourable levels of HDL‐C and LDL‐C with their least favourable levels (i.e. <40 mg/dL and 180–199 mg/dL, respectively) revealed that the risk of new‐onset CAD exhibited a 19‐, nine‐ and seven‐fold increase in individuals with normoglycaemia, borderline glycaemia and diabetes, respectively.To prevent CAD, LDL‐C levels should be strictly controlled in patients with low HDL‐C levels regardless of glucose tolerance. Individualized treatment, which involves setting target LDL‐C levels based on glucose tolerance and HDL‐C values, is required. [ABSTRACT FROM AUTHOR]

To determine associations between severity of hypertension and risk of starting dialysis in the presence or absence of diabetes mellitus (DM). A nationwide database with claims data on 258 874 people with and without DM aged 19-72 years in Japan was used to elucidate the impact of severity of hypertension on starting dialysis. Initiation of dialysis was determined from claims using International Classification of Diseases-10 codes and medical procedures. Using multivariate Cox modeling, we investigated the severity of hypertension to predict the initiation of dialysis with and without DM. Hypertension was significantly associated with the initiation of dialysis regardless of DM. The incidence of starting dialysis in those with systolic blood pressure (SBP) ≤119 mm Hg and DM (DM+) was almost the same as in those with SBP ≥150 mm Hg and absence of DM (DM-). In comparison with SBP ≤119 mm Hg, SBP ≥150 mm Hg significantly increased the risk of the initiation of dialysis about 2.5 times regardless of DM+ or DM-. Compared with DM- and SBP ≤119 mm Hg, the HR for DM+ and SBP ≥150 mm Hg was 6.88 (95% CI 3.66 to 12.9). Although the risks of hypertension differed only slightly regardless of the presence or absence of DM, risks for starting dialysis with DM+ and SBP ≤119 mm Hg were equivalent to DM- and SBP ≥150 mm Hg, indicating more strict blood pressure interventions in DM+ are needed to avoid dialysis. Future studies are required to clarify the cut-off SBP level to avoid initiation of dialysis considering the risks of strict control of blood pressure. [ABSTRACT FROM AUTHOR]

Aims: We attempted to clarify whether the multiple criteria for metabolic syndrome (MetS) can sufficiently predict cardiovascular disease, whether waist circumference (WC) should be required, and whether sex-specific thresholds for each component are necessary. Only a few large-scale studies among East Asians have addressed the ability of MetS to predict cardiovascular disease., Methods: We analyzed the data of 330,051 men and 235,028 women aged 18-74 years with no history of coronary artery disease (CAD) or cerebrovascular disease (CVD) from a nationwide Japanese claims database accumulated during 2008-2016. The association of each MetS component with CAD or CVD (CAD/CVD), MetS associated with CAD/CVD according to various criteria, and utility of modified criteria with more specific optimal values for each component were examined using multivariate Cox regression and receiver operating characteristic (ROC) analysis., Results: During the study, 3,934 men (1.19%) and 893 women (0.38%) developed CAD/CVD. For each current MetS criteria, there was a 1.3- to 2.9-fold increased risk of CAD/CVD. Optimal thresholds for predicting CAD/CVD were WCs of 83 and 77 cm, triglycerides levels of 130 and 90 mg/dl, high-density lipoprotein cholesterol levels of 50 and 65 mg/dl, blood pressures of 130/80 and 120/80 mmHg, and fasting plasma glucose levels of 100 and 90 mg/dl for men and women, respectively. The existing MetS criteria and modified criteria were not significantly different in predicting CAD/CVD, but using the modified criteria markedly increased the prevalence of MetS and percentage of people with MetS developing CAD/CVD., Conclusions: Although various criteria for MetS similarly predicted CAD/CVD, the new criteria greatly reduced the number of high-risk individuals, especially women, overlooked by the current criteria.