<SURF Student Zhenzhen Liu Decodes Death Data>
July 29, 2020 • jainendra joshi • business



Zhenzhen Liu, a senior majoring in statistics  and machine learning , has always been interested in data-

driven analysis of social issues. For the past year, Liu has applied her passion toward Carnegie Mellon

University's  (CHRS) under the direction of ,robin mejia statistics manager, and ,jay aronson the center’s


Liu had followed with interest as another Center researcher, Ben Klingensmith,analyzed data to

understand causes of death in jails   throughout the Commonwealth of Pennsylvania. "Deaths are an

important indicator of public health," Liu said. "Elevated death rates means something is going wrong."

Liu applied for and earned a  (SURF) grant to pursue a project looking at causes of deaths in

Pennsylvania's prisons.

"Deaths are an important indicator of public health. Elevated death rates means something is going wrong." — Zhenzhen Liu

County jails differ from state prisons in that jails typically hold people for shorter lengths of time while they

await trial (if they have been denied bail or cannot afford to pay it) or serve short sentences that do not

merit transfer to a prison. The populations and the causes of death are different in the two types of

facilities, and Liu was interested in helping to quantify just how so.

In recent years, the CHRS has built a robust partnership with the ,pennsylvania prison society  the oldest

human rights organization in the United States, founded two months prior to the signing of the U.S.

Constitution. Historically, the Prison Society has worked to provide direct services to incarcerated people.

Today, with the help of CMU, they want to understand data related to people in custody to inform their

work advocating for policy reform. The Society provided data from the Pennsylvania Department of

Corrections, and CHRS obtained additional publicly available data.


Armed with scores of raw data on deaths within Pennsylvania prison facilities between 2000 and 2019,

Liu set to work sifting through inconsistent abbreviations and lack of categorization. "I thought I could

work with the dataset to develop meaningful ways to compare death rates," Liu said.

In looking through the data, she realized that some deaths seemed to be recorded with an observational

description of the cause rather than a recognized disease code. This may be because most jurisdictions

in Pennsylvania rely on a coroner system rather than a medical examiner to determine cause of death

and fill in the death certificate. A medical examiner, Liu explains, is a physician with a background in

pathology whereas a coroner is an elected person with different qualifications in different geographical

regions. So, too, do the abbreviations or jargon used in recording differ by region. "We want a medical

examiner to evaluate the cause of death data," Liu said.

Liu's next step will be to compare the prison death data to that of non-incarcerated people in

Pennsylvania who fit the same demographic profile. "This will help us develop a statistical understanding

of whether incarcerated people are dying from similar things as the general population," Mejia said.

Aronson points out that another goal of the CHRS is to promote careful recording of cause of death,

because limited or incomplete reporting could obfuscate vital details. "Sure, an incarcerated person could

have died as a result of a heart attack," Aronson said, "but the fact that the individual was denied

potentially life-saving care is an important piece of the story."

Liu said, "Ultimately, we want to be able to do a facility level analysis to identify if one prison is better or

worse than the others in certain metrics." Such comparisons could identify best practices.

Naming those metrics is another challenge the CHRS faces. Mejia said, "there is very little information

available about measuring prison quality." She had hundreds of pages of documents from the American

Bar Association and the United Nations that discussed indicators and standards. Liu volunteered to comb

those documents to identify metrics by which researchers could measure prisons, in terms of the health

of their populations.

Mejia said, "We want to understand the differences between facilities. We’re creating indicators in

Pennsylvania both to understand what’s happening in this state and as models to apply elsewhere."

At the conclusion of the summer SURF project, LIU will continue analyzing data for the Center as part of

her senior thesis.

"It's been amazing to have Zhenzhen with us for the past year," said Aronson. "I continue to be impressed

by her work and commitment to helping us to support the Pennsylvania Prison Society and their work to

hold the state accountable inside facilities where far too many people are being held."