| In this research, we study the manifestation of complexity via comorbidity in two different index diseases, diabetes and sepsis. Complexity, attributable to a number of factors, makes patients more difficult to diagnose and treat. From a modeling perspective, adequately representing more than one disease poses several challenges, including determining how to account for comorbidity, interactions between the diseases, as well as limitations in data sources. The goal of this research is to improve the understanding of complex patients via comorbidity in two index diseases from a systems perspective and provide recommendations for personalized care under three different settings. Using regression models, clustering methods and discrete-event simulation, we examine the challenge of multimorbidity by studying (i) the burden of comorbidity in diabetes and the role of sex and race in hospital outcomes using a national inpatient dataset, (ii) disease interaction and medication regimens in a combined simulation model of diabetes and its comorbidity of breast cancer using multiple data sources, and (iii) the implications of different definitions and guidelines on the diagnosis and treatment of septic patients with comorbidity using electronic health records.;In the hospitalized population, we show that the comorbidities of diabetes can impact inpatient outcomes including length of stay, total charges and discharge disposition. Although hospitalized women have better outcomes than men, the impact of diabetes is worse in women, particularly of minority races. The combined simulation model of diabetes and breast cancer allows us to focus on a specific comorbidity of diabetes. Here, we study outcomes under different assumptions about the relationship between diabetes, glycemic-control medication and breast cancer incidence. Our simulation experiments reveal that metformin improves cancer outcomes as well as lowers the risk of hypoglycemia in women. Further, we show that the risk for hypoglycemia across all races is a significant driver in delaying insulin initiation. Finally, in patients with sepsis, our results show that the time to first treatment is, on average, higher in patients with comorbidity, suggesting that sepsis progresses differently in individuals with a large number comorbidities. The model also shows significant differences in definition-driven outcomes, particularly when compared with administrative (ICD 9) codes available in patients' records.;This research provides models for decision making in the context of comorbidity in addition to the management and analysis of complex, longitudinal data. The findings from this research underscore consistently the importance of race and gender when caring for patients with multimorbidity. Despite the challenges and data limitations associated with modeling multiple diseases, our results highlight the need to move from a single-disease to a multi-disease focus. |
