Date of Award

January 2019

Document Type

Open Access Thesis

Degree Name

Medical Doctor (MD)



First Advisor

Kasia Lipska


Patients with diabetes mellitus are at risk for two acute metabolic complications: severe hyperglycemia and hypoglycemia. These acute complications are costly and associated with significant morbidity and mortality, but are preventable with delivery of high-quality care. The purpose of this work is to focus on a subset of these complications which are iatrogenic, i.e., caused by medical treatment. Hospital-acquired diabetic ketoacidosis (DKA) is an iatrogenic complication as it occurs when a patient with known diabetes experiences DKA while hospitalized for other reasons. Hypoglycemia is an adverse effect of treatment and thus, by definition, all hypoglycemia resulting from the use of glucose-lowering medications in the outpatient setting is iatrogenic. Reducing the occurrence of these iatrogenic complications of diabetes can improve patient health outcomes and reduce costs. However, prevention requires targeted interventions based on a detailed understanding of precipitating factors. In order to address these iatrogenic complications, we performed two analyses to examine factors driving their occurrence.

The first analysis is a retrospective chart review of hospitalized adults with diabetes who developed DKA during a hospital admission at a single local hospital. Twenty-seven patients were included in this analysis over 5 years. The patients were predominantly White (70.4%) and middle-aged (average age 53.4 years). Most had a documented diagnosis of type 1 diabetes (59.3%) and all but 1 patient were on insulin at home. At the time of DKA, 51.9% were on medicine or neurology services, 33.3% on surgery or ob/gyn, and 14.8% on podiatry. Using common cause analysis, the most prevalent reason for DKA was a problem with insulin dosing, including missed doses of insulin (n=7, 25.9%) and insulin dose reductions of 50% or greater (n=8, 29.6%). The remaining cases were caused by steroids (n=4, 13.8%), infection (n=4, 13.8%), and acute stress associated with surgery or shock (n=4, 13.8%).

The second analysis is a retrospective analysis of factors that mediate severe hypoglycemia requiring an ED visit or hospitalization in an insured population in California. A total of 305,310 adults with diabetes were included in this analysis. Among the full cohort, the rate of severe hypoglycemia requiring an ED visit or hospitalization was 7.4 per 1,000 person-years, but this varied significantly by race. Among Black vs White patients, the rates were 13.64 vs 9.27 per 1,000 person-years, respectively. Given the significance of these racial disparities, factors mediating these disparities were further explored. Differences in insulin use by race were not significant, and racial disparities persisted among patients on insulin. Rates of hypoglycemia among Black vs White patients on insulin were 34.72 [95% CI 30.09, 38.87] vs 27.14 [25.38, 28.98] per 1000 person-years, respectively. Factors mediating the racial differences in ED visits and hospitalizations for severe hypoglycemia were investigated using literature review and clinical expert input and a directed acyclic graph (DAG) was created to depict the causal relationships of the proposed mediator variables. Analytic work for this project is ongoing. To analyze our DAG, we plan to assess the causal impact of each proposed mediator variable by using inverse probability weighting to estimate counterfactual disparity measures.

Together, these projects demonstrate the importance of thorough analysis of factors that mediate and precipitate iatrogenic complications. In the case of hospital-acquired DKA, interventions targeting inappropriate insulin dosing among hospitalized

patients with diabetes could potentially prevent over 50% of cases. For severe outpatient hypoglycemia, quantifying the causal impact of each proposed mediator variable in the DAG will reveal high-yield opportunities to address disparities in hypoglycemia. Ongoing work on both projects continues to improve understanding of these problems and will ultimately facilitate implementation of targeted prevention strategies.


This is an Open Access Thesis.

Open Access

This Article is Open Access