Misdiagnosis is the leading cause for malpractice in healthcare and can result in improper treatments, disabilities, or death. Misdiagnoses have become increasingly more expensive as healthcare reimbursement shifts to a quality-based system. Endocarditis, an infection of the inner lining of the heart, is one of the more prevalent misdiagnosed diseases, and results in severe health complications as the infection spreads throughout the body. The objective of the study is to identify factors which might be associated with endocarditis diagnostic errors and deliver a proposal for specific areas of interest for future study.
Using inpatient and emergency department databases from the Healthcare Cost and Utilization Project (HCUP) data for California and New York between 2005 and 2012, we identify all patients who were diagnosed with endocarditis. We then link all visits of those patients that fall within a 60-day window—an approximation for the plausible progression of endocarditis development. We create a list of symptoms and diagnoses indicative of endocarditis and define patients with these primary diagnoses as having a missed opportunity to diagnose the disease. We then use statistical methods to identify common factors for these patients that might influence the probability of endocarditis misdiagnosis. Our goal is to predict patients with endocarditis diagnosis errors, or at the very least, identify factors which significantly impact misdiagnosis.
There are 10,498 patients in California and 7,773 patients in New York with plausible misdiagnoses, meaning they have a visit within 60 days of being diagnosed with endocarditis. Using statistical tests, we found a set of significant factors which are common among endocarditis patients suspected to be misdiagnosed. They align with the known symptoms of endocarditis, such as bacterial infections, prior heart defects, fever with unknown origin, etc. Twenty-five to seventy-four percent of patients with plausible misdiagnoses were diagnosed with at least one of the factors in our selected set of characteristics. The regression test confirmed that our suspected patients are more likely than other cases to reflect these factors.
We have implemented a data-driven approach to indicate demographic variables and medical conditions which affect the rate of misdiagnosis. Moreover, looking at patients’ records, it is likely that several of our statistically-identified factors may have serious diagnostic implications, which can be fully explored with future study. Various types of dispositions of the patient at discharge, and their race, sex, age, and length of stay at the hospital may be useful indicators to a doctor before he or she determines a patient’s final diagnosis.
Anh Pham, ’16
Polina Duneva, ’18
Economics & Business
Sponsor: Aaron Miller