Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


37969.pdf (230.03 KB)

ETD Abstract Container

Abstract Header

Regression/Decision Trees to Predict the Severity of Intervention Needed for COVID-19 Positive Patients Using Baseline Emergency Department Vitals at Presentation

Hoehn, Jonathan
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613745329872462

Abstract Details

2020, MS, University of Cincinnati, Medicine: Biostatistics (Environmental Health).
Introduction: The Coronavirus Disease 2019 (COVID-19) pandemic continues to hugely impact daily life nearly 1 year after its discovery. While vaccine development is underway and testing continues to increase, new cases of the severe acute respiratory disease coronavirus 2 (SARS-CoV-2) continue to increase in the United States and elsewhere. Though mitigating the spread of the disease is vital to decreasing new cases and mortality, being able to predict the severity of the disease to help accelerate treatment for patients with confirmed positives is also essential. Methods: In April and May of 2020, patients presenting at UCMC’s Emergency Department (ED) with suspected COVID-19 had blood drawn with an IRB approved waiver of informed consent. Data was abstracted from the patients’ medical records and their blood was sent to the CCHMC lab for further testing. Of the samples collected, 52 were confirmed positive by reverse-transcription polymerase chain reaction (RT-PCR) test using a standard-of-care nasopharyngeal swab. In addition to the data abstracted, each positive patient was given a severity score several times during their time in the ED. A regression/decision tree was built using the 49 of the 52 positive cases that had complete data abstraction with maximum severity score as the primary outcome. Cross validation was then performed using an 80%/20% training/testing split to determine the tree’s accuracy. Results: The full tree accurately predicted the maximum severity score for 39/49 patients (79.59%). However, after cross validation (1000 iterations), the mean accuracy of the decision tree was found to be 44.18% with a standard deviation of 14.75%. Conclusion: While a regression/decision tree such as the one originally built could be easily implemented to predict the progression of the disease, further work is needed. Machine learning techniques could be used to refine the model but the largest weakness is the very small dataset.
Marepalli Rao, Ph.D. (Committee Chair)
Opeolu Adeoye, M.D. (Committee Member)
7 p.
Biostatistics

Recommended Citations

Citations

  • Hoehn, J. (2020). Regression/Decision Trees to Predict the Severity of Intervention Needed for COVID-19 Positive Patients Using Baseline Emergency Department Vitals at Presentation [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613745329872462

    APA Style (7th edition)

  • Hoehn, Jonathan. Regression/Decision Trees to Predict the Severity of Intervention Needed for COVID-19 Positive Patients Using Baseline Emergency Department Vitals at Presentation. 2020. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613745329872462.

    MLA Style (8th edition)

  • Hoehn, Jonathan. "Regression/Decision Trees to Predict the Severity of Intervention Needed for COVID-19 Positive Patients Using Baseline Emergency Department Vitals at Presentation." Master's thesis, University of Cincinnati, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613745329872462

    Chicago Manual of Style (17th edition)

ucin1613745329872462
94
© 2020, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.