Research

Mohammad Akbarpour, Cody Cook, Aude Marzuoli, Simon Mongey, Abhishek Nagaraj, Matteo Saccarolak, Pietro Tebaldi, Shoshana Vasserman, and Hanbin Yang
Researchers from Harvard, Stanford, the University of Chicago, and Berkeley created a methodology for deciding which locations in a city to re-open and which to keep closed. The authors utilized a wide variety of data sources to build their models, including electronic medical records from the COVID-19 Research Database. After its completion, the model was then tested across three different cities: Chicago, Sacramento, and New York.

Faraz S. Ahmad, Iben M. Ricket, Bradley G. Hammill, Lisa Eskenazi, Holly R. Robertson, Lesley H. Curtis, Cecilia D. Dobi, Saket Girotra, Kevin Haynes, Jorge R. Kizer, Sunil Kripalani, Mathew T. Roe, Christianne L. Roumie, Russ Waitman, W. Schuyler Jones, and Mark G. Weiner
This study outlines the recruitment process for the ADAPTABLE study (Aspirin Dosing: a Patient-Centric Trial Assessing Benefits and Long-Term Effectiveness), a pragmatic, randomized, open-label clinical trial that tested the optimal dose of aspirin for secondary prevention of atherosclerotic cardiovascular disease events. Investigators identified 650,000 potential eligible patients and recruited them at community sites, linking together data from the 40 sites in the Patient-Centered Outcomes Research Network to understand eligibility. The study was ultimately able to successfully enroll 15,076 patients in a significantly lower-cost way than most trials.

John Hargraves and Daniel Kurowski
Using mortality data, the Health Care Cost Institute found that daily deaths in the United States were over 10 percent higher in 2020 than they had been in previous years. The data showed dramatic variations by geography. Since the end of March, New York daily death figures were double those from previous years. While New York deaths appeared to be declining by April, other states were on the rise.

Jordan B. Strom, Kamil F. Faridi, Neel M. Butala, Yuansong Zhao, Hector Tamez, Linda R. Valsdottir, J. Matthew Brennan, Changyu Shen, Jeffrey J. Popma, Dhruv S. Kazi, and Robert W. Yeh
The study attempted to see whether passively collected data could substitute for adjudicated outcomes to reproduce the magnitude and direction of treatment effect observed in cardiovascular clinical trials. By linking together a variety of data sets, including clinical trials and medicare inpatient claims, the researchers concluded that the clinical trial and the claims data produced magnitudinally and directionally consistent results for the primary endpoints, though less so for the secondary endpoints.

Abel N. Kho, Jingzhi Yu, Molly Scannell Bryan, Charon Gladfelter, Howard S. Gordon, Shaun Grannis, Margaret Madden, Eneida Mendonca, Vesna Mitrovic, Raj Shah, Umberto Tachinardi, and Bradley Taylor
In conjunction with the All of Us Research Program, the authors looked at Electronic Health Records to understand how fragmented patient data was across a variety of health provider organizations, using a Privacy-Preserving Record Linkage tool in participating sites to generate a unique set of keyed encrypted hashes. Of the 5,831,238 individuals, 458,680 patients had data at more than one institution. Patients with some care fragmentation were almost 10 times as likely to have conflicting or inconsistent demographic data.

Jason H. Raad, Elizabeth Tarlov, Abel N. Kho, and Dustin D. French
The VA does not capture health encounters that occur outside its facilities. Therefore, any analyses of health-care utilization will miss certain data; this is particularly important for at-risk populations, where understanding utilization will enable better and less fragmented care.  VA data was linked with Chicago’s HealthLNK Data Repository to find that of the 13,948 veterans who were homeless or at risk of becoming homeless, 17 percent of those veterans received some or all of their care in the community.

Melanie Canterberry, Alan F. Kaul, Satyender Goel, Pi-I Debby Lin, Jason P. Block, Vinit P. Nair, Qianli Ma, and Thomas W. Carton
The Patient-Centered Outcomes Research Network (PCORnet) connects health systems, health plans, EHR data, and claims data to facilitate real-world clinical research. Datavant powers the linking of PCORnet. In this study, the authors looked to identify and understand the extent to which a clinical data research network and a health plan network contained consistent data on the patient level. Of the 549 patients from the EHR record who had claims with the health plan, sixty percent had consistent antibiotic exposure while 43.1% had a matched claims record.

Nikhilesh R. Mazumder, Kofi Atiemo, Amna Daud, Abel Kho, Michael Abecassis, Josh Levitsky, and Daniela P. Ladner
Most patients with cirrhosis have low Model for End-Stage Liver Disease-Sodium (MELD-Na) scores. The ability for the MELD-Na score to predict patient outcomes at low scores remains unclear. The researchers were able to link data from across Chicago to understand what patients might be eligible; from there, 7922 patients were identified, with 3999 patients having low MELD-Na scores. Further analyses were performed to show that, though patients with high MELD-Na scores had higher rates of mortality and were more likely to die of liver-related causes, even patients with low MELD-Na scores had high rates of liver-related mortality.

Qinli Ma, Haechung Chung, Sonali Shambhu, Matthew Roe, Mark Cziraky, W. Schuyler Jones, and Kevin Haynes
This study examined a retrospective cohort as part of the ADAPTABLE trial (Aspirin Dosing: A Patient-centric Trial Assessing Benefits and Long-Term Effectiveness). The trial was designed to compare the efficacy of different doses of Aspirin for the prevention of atherosclerotic cardiovascular disease events. By linking claims data from Anthem with data from the Patient-Centered Outcomes Research Network (PCORnet), the investigators were able to follow participants and understand their health outcomes without direct follow-up or bringing them into a trial site.