Risk of Typical Diabetes-Associated Complications in Different Clusters of Diabetic Patients: Analysis of Nine Risk Factors

Abstract

Objectives: Diabetic patients are often diagnosed with several comorbidities. The aim of the present study was to investigate the relationship between different combinations of risk factors and complications in diabetic patients. Research design and methods: We used a longitudinal, population-wide dataset of patients with hospital diagnoses and identified all patients (n = 195,575) receiving a diagnosis of diabetes in the observation period from 2003–2014. We defined nine ICD-10-codes as risk factors and 16 ICD-10 codes as complications. Using a computational algorithm, cohort patients were assigned to clusters based on the risk factors they were diagnosed with. The clusters were defined so that the patients assigned to them developed similar complications. Complication risk was quantified in terms of relative risk (RR) compared with healthy control patients. Results: We identified five clusters associated with an increased risk of complications. A combined diagnosis of arterial hypertension (aHTN) and dyslipidemia was shared by all clusters and expressed a baseline of increased risk. Additional diagnosis of (1) smoking, (2) depression, (3) liver disease, or (4) obesity made up the other four clusters and further increased the risk of complications. Cluster 9 (aHTN, dyslipidemia and depression) represented diabetic patients at high risk of angina pectoris “AP” (RR: 7.35, CI: 6.74–8.01), kidney disease (RR: 3.18, CI: 3.04–3.32), polyneuropathy (RR: 4.80, CI: 4.23–5.45), and stroke (RR: 4.32, CI: 3.95–4.71), whereas cluster 10 (aHTN, dyslipidemia and smoking) identified patients with the highest risk of AP (RR: 10.10, CI: 9.28–10.98), atherosclerosis (RR: 4.07, CI: 3.84–4.31), and loss of extremities (RR: 4.21, CI: 1.5–11.84) compared to the controls. Conclusions: A comorbidity of aHTN and dyslipidemia was shown to be associated with diabetic complications across all risk-clusters. This effect was amplified by a combination with either depression, smoking, obesity, or non-specific liver disease.

Real-World Data Analysis During the First-Wave Coronavirus Pandemic from a Diabetic Unit in Central North Italy: Old Oral Hypoglycaemic Agents and Insulin Show to be Associated with Higher Hospital Admissions and Total Mortality and Glifozines Seem to Protect from Total Mortality in Covid-19+ T2DM Patients

“…These data also highlight the importance of correctly approaching the diabetic patient, regarding phenotype [23], cardiovascular risk [28] and diabetic complications [24][25][26] and comorbidities [27]. As far as we can appropriately characterize the T2DM patient's profile, thus tailoring his treatment needs, we could be able to better prevent complications in chronic and in the acute setting of care to improve outcomes even during a severe acute illness such as Covid-19.…”

Correlation between myocardial perfusion imaging findings and future cardiac events in patients with type 2 diabetes mellitus

“…This together with the prevalence of diabetic cardiomyopathy contributes to a higher cardiovascular mortality [18,19]. Furthermore, diabetic patients have higher prevalence of cardiovascular co-morbidities as compared to patients without diabetes [20,21]. Two-thirds of diabetic patients will die of heart or vascular disease, and patients with CAD and diabetes mellitus have worse outcomes and a much higher cardiac event rate than their nondiabetic counterparts [22,23].…”

Using Domain Knowledge and Data-Driven Insights for Patient Similarity Analytics

“…However, it remains a challenge to analyze and derive insights from the huge volume of EHR data, which are multivariate, heterogeneous, and sparse. These analyses involve finding similar patients for patient stratification [ 11 , 12 , 13 ], diagnosis prediction [ 14 , 15 ], medical prognosis [ 16 , 17 ], or treatment recommendations [ 18 , 19 , 20 ]. With patient similarity analytics, personalized models can be built based on the retrieved cohort of similar patients, thus furthering the development of personalized medicine.…”

“…Despite the fact that many studies had proposed their own similarity metrics belonging to these two categories, some limitations exist in the proposed approaches. First, many of the proposed approaches were only applicable to datasets with a low-level of granularity, where the datasets only consisted of limited types of variables, such as only using a series of International Classification of Diseases (ICD) codes as data input [ 17 , 21 ]. Moreover, most of the proposed approaches were solely based on data-driven insight.…”