MIT Sloan Health Systems Initiative
HSI Funds Five New Research Projects for 2023
With this round of funding, we will have funded 19 projects since the Spring of 2019. Eight are complete, six are ongoing, and five are new. Two of the new projects are with faculty whom we have not funded before, Alexey Makarin and Rahul Mazumder.
New Research Projects
Prof. Georgia Perakis and her team (Prof. Dessislava Pachamanova. Omar Skali Lami and students from the MIT Operations Research Center) have been collaborating with UMass Memorial Hospital in Worcester, MA on a number of initiatives to reduce backlog in services in a variety of departments. For this project, the team is evaluating the backlog for the MRI resources and whether it makes sense to expand this service over the weekend even though it is more expensive to operate then than during the week.
There are three components to this research project: 1) demand forecasting, 2) a framework and assessment based on those predictions, and 3) a method to make this approach workable and implementable. Perakis notes that many sophisticated healthcare operations approaches remain on paper because they are too difficult to implement.
Component one: Develop a forecasting model for the demand for in-patient, out-patient and emergency patients for MRI services on the weekends. Rate which parts of the demand are most beneficial for the hospital metrics such as overall patient LOS.
Component two: Based on the predictions forecasted in component one, develop a prescriptive framework for capacity management that also assesses the need to expand capacity at any given time.
Component three: Here is where the framework and models go from paper to being useful. Perakis and team intend to develop an interpretable framework for implementing their findings. “A critical part of this project”, Perakis emphasizes, “would be to translate our findings into recommendations that can be implemented within the hospital operational priorities.”
The Quest Blueprint for Wellness health risk assessment program has two components: health screenings and programs to address health issues found via those screenings. For this project, Prof. Joe Doyle seeks to understand which employees are taking advantage of these programs, and mechanisms that would increase participation. He will also explore the cost effectiveness of these offerings.
The first part of the project involves a retrospective analysis of the data Quest collects in its health risk assessments as well as information on program participation. This analysis will provide information on the relationships among program involvement and employees’ health, satisfaction and productivity.
Doyle will use behavioral economics techniques to uncover insights that may increase employees’ take-up of the health programs. Specifically, these techniques can bring to light barriers to participation more effectively than employee surveys. They can ultimately provide guidance for ways to reduce these barriers that will make a difference in enrollment and employee health.
Some of the behavioral economics techniques are:
- Default: automatically enrolling some employees during orientation or reorientation
- Reminders: helping busy employees take part by reminding them of the benefits, popularity or low cost of the programs
- Financial incentives: encouraging participation with features such as the timing or amount of incentives
- Commitment contracts: agreeing to receive something in the future to overcome current procrastination
Doyle plans to recommend options for increasing take-up of the wellness programs that will be of benefit to both the employees and to the company. Insights and techniques from this study can then be adapted to other organizations.
Harm reduction is an approach to reducing drug-related deaths by focusing on educating drug users how to use in a way that minimizes health-related risks. Prof. David Rand and his co-researcher Dr. Elizabeth Paci from the Boston University School of Medicine seek to investigate whether TikTok, since it is extremely effective in targeting users, may be a unique opportunity to reach at-risk youth with a harm reduction message.
The first step will be to identify the hashtags that are associated with harm reduction. This exercise combined with interviews ought to result in a list of relevant hashtags and the number of views associated with each one. Then, using this list, Rand and Paci will collect several hundred TikTok videos across a wide variety of tags and engagement level that are all related to drug use practices. Analyses of these videos will allow them to assess how much of the drug-use related content is from official channels versus peer-to-peer, what practices are being encouraged and levels of engagement for the different types of videos.
Ultimately, the goal of this project is to use statistical and machine learning approaches to produce models that provide insight into what factors are relevant and which combinations best encourage engagement levels. Of particular interest will be using these models to determine which kinds of harm reduction content perform best.
Prof. Rahul Mazumder intends to mine prior work in which researchers studied the effect of COVID-19 on mental health of former NIMH patients and volunteer participants. He plans to use machine learning tools to analyze those data collected over time in order to identify the best interventions to benefit employees’ mental health. Too often this type of data is difficult to interpret. The main goal of the project is to identify subgroups of employees who are at a higher risk for the development of adverse mental health conditions.
Some of the questions Mazumder seeks to answer are:
- To what extent do baseline personal characteristics such as demographics (e.g., age, gender, occupation) or clinical history (e.g., past history of depression) determine propensity to low mood or high anxiety?
- To what extent do changes in working environment or personal circumstances increase or decrease that propensity?
- What personal characteristics are risk or resilience factors in the face of changes in working environment or personal circumstances?
- How do these within- and between-subject factors interact to predispose individuals to pandemic-related anxiety and resilience?
Mazumder proposes to develop interpretable machine learning tools that will both analyze this complex data and shed light on the most effective interventions. Armed with these tools, he hopes to be able to translate the employee feedback and determine which changes are most likely to improve productivity, employee retention and worker well-being.
Smartphones and geolocation services ushered in enormous changes in the online dating scene. Yet, the causal relationships among online dating applications and young adults’ dating behavior and health outcomes are not well understood. Prof. Alexey Makarin notes that almost all of the existing research on this topic relies on correlation analyses.
His contribution to the research on online dating technologies investigates causal relationships between using dating apps and health by focusing on Tinder. Tinder is the market leader among the US college student population. According to a PEW Research Center Study, in 2022, 80% of people ages 18 and 29 in the US used Tinder at least once. Markarin’s preliminary analysis suggests that Tinder led to a “sharp and lasting increase in reported dating and sexual activity” as well as an increase in the “reported instances of sexual assault and STDs…[but] no change to the number of reported relationship problems.”
Makarin intends to build on this analysis by conducting an original survey of students who were in college around the time of Tinder’s rollout. Also, he will pilot an experiment that will monitor the amount of time students spend on Tinder, incentivize them to reduce time spent on the dating app, and study the impact of that incentive offer on students’ mental health.