PREDICTING CHRONIC COMORBID CONDITIONS OF TYPE 2 DIABETES IN NEWLY-DIAGNOSED DIABETIC PATIENTS
Author(s)
Razavian N1, Smith-McLallen A2, Nigam S2, Blecker S1, Schmidt AM1, Sontag D1
1New York University, New York City, NY, USA, 2Independence Blue Cross, Philadelphia, PA, USA
OBJECTIVES: Type 2 Diabetes(T2D) and its comorbid conditions are imposing large burdens on health-care systems worldwide. Once a patient is diagnosed with T2D, their risk for immediate chronic comorbid conditions is less quantified. We aim to build risk prediction models for onset of cardiovascular, cerebrovascular, renal and eye severe conditions for newly-diagnosed patients with T2D. METHODS: Our cohort includes 4.1million insurance beneficiaries of age≥18yrs between 2005 and 2013. Of these subjects, 34,411 patients were newly-diagnosed with T2D, had continuous enrollment for 6months pre- and 5years post-onset, and did not have the comorbid condition diagnosis before T2D. We developed a prediction model for each condition based on L1-regularization method, which selects few(less than 350 per condition) relevant risk factors from approximately 27K general variables derived from claims data. We defined comorbid conditions based on the validated diabetes complications severity index. RESULTS: For patients newly-diagnosed with T2D, we can predict new onset of cardiovascular conditions with Area Under Curve(AUC)=0.69±1e-4; cerebrovascular conditions with AUC=0.73±1e-4; renal complications with AUC=0.71±1e-4; eye conditions with AUC=0.65±1e-4. Our method discovers risk factors, some of which have significant (p≤0.001) differences in odds ratios(OR) between genders/age subgroups. Top differences include: High albumin in blood for middle-aged patients(40≤age<65yrs) (male-OR 1.3[0.9-1.9] vs. female-OR 2.1[1.5-3.4]) in cardiovascular events; High erythrocyte distribution width in old patients(age>65) (male-OR 1.6[1.3-2.1] vs. female-OR 1[0.8-1.2]) in cerebrovascular events; High creatinine in middle aged patients (male-OR 4.6[3.4-6.0] vs. female-OR 5.7[4.1-7.9]), high urea nitrogen for middle-age patients (male-OR 2.2[1.7-2.7] vs. female-OR 2.7[2.11-3.5]), and decreasing thyrotropin for old patients (male-OR 1.1[0.9-1.5] vs. female-OR 0.6[0.4-0.7]) in renal events. CONCLUSIONS: Our proposed model shows promise in risk prediction and risk factor discovery for 4 comorbid complications of T2D in patients newly-diagnosed with T2D. Further research is needed to understand how these predictions translate to prevention and delay of the complications onset.
Conference/Value in Health Info
2015-05, ISPOR 2015, Philadelphia, PA, USA
Value in Health, Vol. 18, No. 3 (May 2015)
Code
PDB5
Topic
Epidemiology & Public Health
Topic Subcategory
Disease Classification & Coding
Disease
Diabetes/Endocrine/Metabolic Disorders