Utilization of Large-Scale Electronic Health Record Database to Understand Social Determinants of Health Disparities in the Diabetic Patient Population with Major Lower Limb Amputation.

Abstract

Objective: Poorly controlled diabetes is associated with higher risks of lower extremity amputations (LEAs). Social determinants of health, alternatively termed social drivers of health (SDoH), have been shown to play an important role in health outcomes. However, specific SDoH domains that impact poorly controlled diabetics (PC-diabetics) are undefined. This study aims to utilize Epic Cosmos, a cross-institutional electronic medical record (EHR) database, to identify these specific social risk factors in order to better inform and tailor support services.

Methods: Epic Cosmos is a HIPAA-compliant data platform with longitudinal EHR records of more than 300 million patients. PC-diabetics were queried based on having at least one A1c test ≥ 7.0%, and non-diabetic or controlled diabetic patients (N/C-diabetic) were identified as having an A1c test < 7.0% and none above 7.0% from 2014 to 2018. LEAs within 5 years of the A1c test were identified through selected CPT codes (27590, 27591, 27592, 27594, 27880, 27881, 27882, 27884). SDoH diagnoses specific to this patient cohort were extracted using ICD-10-CM Z-codes (Z55 - Z75). All statistical significance was calculated using Chi-square tests in Microsoft Excel and R Studio, with significant level set at p = 0.05.

Results: Our study included 4,512,612 PC-diabetic and 13,659,422 N/C-diabetic patients. The PC-diabetic cohort had 20,944 individuals with LEA while the N/C-diabetic group had 7,981 patients with LEA. In PC-diabetics with LEA, 28.41% had at least 1 SDoH diagnosis, in contrast to 12.20% of PC-diabetics without LEA or 11.80% of N/C-diabetics (p < 0.05). Patients with PC-diabetes and LEA are 3.02 times more likely to have problems related to housing/economic circumstances (Z59) and 2.54 times more likely to have problems related to lifestyle (Z72), compared to PC-diabetics without LEA. Our study also found high rates of smoking (Z72.0, 20.85% of total) in PC-diabetics with LEA. Moreover, African American PC-diabetics without LEA have higher baseline rates of SDoH diagnoses (14.90%) than their white/Caucasian counterpart (11.92%, p < 0.05).

Conclusions: Our analysis on PC-diabetics with LEA showed higher SDoH diagnosis rates in housing/economic and lifestyle challenges compared to PC-diabetics without LEA. Tobacco use was found to be a particularly notable risk factor among this patient cohort. This study demonstrates the feasibility of using large-scale EHR databases to generate hypotheses and guide future strategies to address SDoH risk factors.