Glycemic and lipid variability for predicting complications and mortality in diabetes mellitus using machine learning
Recent evidence directs that HbA1c and lipid variability is useful for risk stratification in diabetes mellitus. With the increasing prevalence of diabetes, there is a need for new parameters to monitor the disease. Random survival forest (RSF) can be used for survival analysis. It is a class of machine learning algorithms. Hence, Sharen Lee and colleagues have conducted research published in BMC Endocrine Disorder Journal under the title “Glycemic and lipid variability for predicting complications and mortality in diabetes mellitus using machine learning”. The summary of this study is given below:
Objective:
To examine the predictive value of glycemic and lipid variability towards a varied range of adverse outcomes in diabetes.
To elaborate on accurate risk prediction using RSF.
Method:
It is a retrospective cohort study consisting of type 1 and types 2 diabetic patients. Participants have prescribed insulin at outpatient clinics of Hong Kong public hospitals, between 1st January to 31st December 2009. Standard deviation (SD) and coefficient of variation were used in order to measure the variability of HbA1c, total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglyceride. The main outcome is all-cause mortality. Secondary outcomes were diabetes-related complications.
Findings:
Increased HbA1c and lipid parameters are linked with an elevated risk in both diabetic complications and all-cause mortality. Additionally, HbA1c variability was found to be positively associated with increased neutrophil-lymphocyte ratio (NLR) and a number of hypoglycemia episodes. There is also the presence of interactions between the value and variability of different lipid parameters.
Limitation:
The authors acknowledge a few limitations to study as follow:
Firstly, there was missing data, under-coding, and coding errors. As this is an observational study, it can only establish correlation, not causation. Additionally, the duration of diabetes was not recorded, subjects have gone through large changes in the therapeutic options, management guidelines, and treatment targets throughout follow-up, there was a lack of data on the patient’s body mass index and lifestyle factors, such as alcoholism, smoking, and diet.
Image Credit : People photo created by drobotdean – www.freepik.com