Semester

Fall

Date of Graduation

2020

Document Type

Dissertation

Degree Type

PhD

College

School of Pharmacy

Department

Pharmaceutical Systems and Policy

Committee Chair

Usha Sambamoorthi

Committee Co-Chair

Nilanjana Dwibedi

Committee Member

Traci LeMasters

Committee Member

Amit Ladani

Committee Member

Wenhui Wei

Abstract

Osteoarthritis (OA) is a degenerative arthritis affecting over 30 million Americans most of whom are over 65 years or older. Its clinical management is complicated by several disease- and treatment-specific factors. These include the co-occurrence of cardiovascular and gastrointestinal disorders (CV-GID), the inappropriate use of non-steroidal anti-inflammatory drugs (NSAID) to manage pain, and the risk of certain age-related chronic conditions like Alzheimer’s disease and related dementia (ADRD). Moreover, older adults with OA are at a higher risk of CV-GID, inappropriate NSAID use, and ADRD. Additionally, these factors can also affect one another in both a positive and a negative way. For example, the long-term use of NSAID has been shown to increase the risk for cardiovascular and gastrointestinal disorders. On the other hand, their use has been shown to decrease the risk of ADRD in some studies. NSAID use is disproportionately higher among older adults, so the benefits or risks associated with such use should be taken into account while making treatment decisions. However, there is a gap in our understanding of the clinical and demographic factors that increase the risk of co-occurring CV-GID, inappropriate NSAID use, and ADRD in older adults with OA. This dissertation pursued three related aims to fill this knowledge gap: 1) identify the leading predictors of CV-GID; 2) identify the leading predictors of inappropriate NSAID use; and 3) examine whether duration of NSAID use is a leading predictor of ADRD and how other factors affect this relationship using a combination of machine learning techniques. All three aims used a retrospective, longitudinal, cohort study design using de-identified commercial health insurance insurance claims data from Optum De-identified Clinformatics Data Mart for years 2015 through 2017. OA was identified from these data using a combination of International Classification of Disease – 9th Revision and 10th Revision (ICD-9 and ICD-10) codes. Using a random forest classifier, we identified age, cardiac arrhythmia, and the duration of opioid use to be the top three leading predictors of CV-GID in our study cohort. In the second aim, we found that around 13% of older adults with OA were prescribe NSAIDs not in accordance with their CV and GI risk profile (i.e. inappropriate NSAID use). Using an eXtreme Gradient Boosting classifier and Shapley Additive eXplanations, we found durations of non-selective and selective NSAID use to be the top two predictors of inappropriate NSAID use. Older adults with low CV and high GI or high CV and low GI risk were also identified to be more likely to be treated with inappropriate NSAIDs. Lastly, results from our third aim showed that duration of NSAID use was among the top five leading predictors of ADRD in our study cohort. With the help of interpretable machine learning techniques, we found that the effect of NSAID duration on ADRD varied by factors like age, gender, and OA-related pain. In summary, the results of these aims suggest a need to target certain patient-level factors for better management of older adults with OA to reduce their risk of CVGID, NSAID-related adverse events, and ADRD. The results from this dissertation also highlight the need for further research to identify the causal links between the leading factors and the outcomes. The clinical management of OA is complex, and the knowledge obtained from this dissertation could help ease its burden for clinicians and patients.

Embargo Reason

Publication Pending

Download

Share

COinS