Applications of healthcare analytics in reducing hospitalization days

In this thesis, we employ healthcare analytics to inform system-level changes at Massachusetts General Hospital that could lead to a significant reduction in avoidable hospitalization days and improvement in patients outcomes. The first area of focus is around avoidable bed-days in the ICU. Many surgical patients experience non-clinical delays when they transfer from the ICU to a subsequent general care unit where they are expected to continue their recovery. As a result, they spend a longer time in the ICU than necessary. In spite of several studies that suggest out-of-ICU transfer delays are quite common, there is little work that quantifies the impact on patient recovery. Using multiple statistical approaches including regression and matching, we obtain a robust result that suggests that non-clinical transfer delays from the ICU delay the patient's recovery as well as extend the hospital LOS. Specifically, the analysis shows that each day that the patient is delayed in the ICU for non-clinical reasons increases hospital LOS by 0.71 days (p-value < 0.01) and the patient's progress of care by 0.32 days (p-value < 0.01), on average. The second area of focus is concerned with bed-days from heart failure (HF) admissions. Much of the current work in reducing HF hospitalizations promotes interventions after the patient is hospitalized, aiming to prevent subsequent hospitalizations within 30 days. In contrast, we focus on reducing overall hospitalizations from the general HF population. We first analyze the outpatient access for these patients before they are admitted to the hospital (mostly) through the Emergency Department. One of the main findings is that in more than half of these admissions, the patient did not have a completed appointment with any outpatient clinic (Primary Care, Cardiology, or Home Health) during the two weeks prior to hospitalization. This reveals the need for improved outpatient-based preventative measures to manage HF patients. To partially address this challenge, we develop a predictive model using a logistic regression to predict the risk of a HF-related admission within the next six months. The model performs quite well with an out-of-sample AUC of 0.78.

Department dc:contributor.department

Massachusetts Institute of Technology. Operations Research Center.

Grantor dc:publisher

Massachusetts Institute of Technology

Year dc:date.issued

2018

Author dc:creator

• Furtado, Jazmin D. (Jazmin Dahl)

Advisor dc:contributor.advisor

• Retsef Levi.

• Operations Research Center.

Statement dc:rights

• MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission.

Licence dc:rights.uri

• http://dspace.mit.edu/handle/1721.1/7582

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/1721.1/119355

OAI identifier oai:identifier

oai:dspace.mit.edu:1721.1/119355