A systematic literature review of predicting diabetic retinopathy, nephropathy and neuropathy in patients with type 1 diabetes using machine learning

Background: Diabetic retinopathy, nephropathy and neuropathy in patients with type 1 diabetes (T1D) are microvascular complications that can adversely impact disease prognosis and incur greater healthcare costs. Early identification of patients at risk of these microvascular complications using predictive models through machine learning (ML) can be helpful in T1D management. The objective of current review was to systematically identify and summarize published predictive models that used ML to assess the risk of diabetic nephropathy, retinopathy and neuropathy in T1D patients. Methods: A targeted review of English literature was undertaken in PubMed (http://www.ncbi.nlm.nih. gov/pubmed) and Google Scholar (http://scholar.google.com/) from January 1, 2016 to May 31, 2019. Eligible articles were also identified from cross-references. Following concepts were used in combination to conduct the search queries: diabetes, retinopathy, nephropathy, neuropathy, microvascular complication, risk/predictive model, and ML/artificial intelligence/data mining. Results: A total of 3,769 hits were found from all sources combined, duplicates were removed, titles and abstracts were screened, 61 studies underwent full-text review and a total of six studies met the eligibility criteria. Among them, four studies had developed risk models using data obtained from T1D patients alone, whereas two used data from both T1D and type 2 diabetes (T2D) patients. There was only one study that evaluated all three types of microvascular complications while the other five focused on one individual complication, i.e., either diabetic retinopathy, nephropathy or neuropathy. Only two studies evaluated time to developing a complication. The other four studies assessed complications as either binary (yes/no) or categorical (multiple levels). Prediction models were built using cross-sectional data from survey questionnaire (n=1, Iran) and longitudinal data (n=5) which were further classified as sources of electronic medical records (EMR) (n=3, US: 1, Europe: 2), clinical trial (n=1, US) and prospective study (n=1, Europe). Common predictors across studies as well as across types of microvascular complications included age, gender, diabetes duration, BMI, blood pressure, lipid level, and mean or a single HbA1C value. Commonly used ML algorithms included classification and regression tree (CART) and random forest (RF) (CART/RF, n=3), support vector machines (SVMs, n=2), logistic regression (LR, n=2) and neural networks (NNs, n=1). Model performance was evaluated using area under curve (AUC, n=4) and accuracy (n=2). Only half (n=3) of the included studies tested their developed models in an external dataset of patients with T1D. Conclusions: Overall, very few studies reported predictive models for diabetic retinopathy, nephropathy and neuropathy using ML specifically for T1D patients. Future research that utilizes contemporary clinical data from T1D patients to predict the three types of microvascular complications is needed.