Jesse D. Raffa

Jesse D. Raffa

Auto-generated profile, based on publicly available information

Jesse D. Raffa

Massachusetts Institute of Technology, Cambridge, MA, USA

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
20 December 2016 in Journal of Medical Internet Research
Verified
Journal Article
Open access
LCLeo Anthony Celi
Leo Anthony Celi
GDGuido Alejandro Davidzon
Guido Alejandro Davidzon
AJAlistair Ew Johnson
Alistair Ew Johnson
MKMatthieu Komorowski
Matthieu Komorowski
DMDominic C Marshall
Dominic C Marshall
DSDavid J Stone
David J Stone
Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

1 January 2024 in Critical Care Science
Verified
Journal Article
Open access
PDF file
BLBarbara D. Lam
Barbara D. Lam
TSTristan Struja
Tristan Struja
YLYanran Li
Yanran Li
João MatosJoão Matos
João Matos
ZCZiyue Chen
Ziyue Chen
JRJesse Raffa
Jesse Raffa
Objective: Determine how each organ component of the SOFA score differs in its contribution to mortality risk and how that contribution may change over time. Methods: We performed multivariate logistic regression analysis to assess the contribution of each organ component to mortality risk on Days 1 and 7 of an intensive care unit stay. We used data from two publicly available datasets, eICU Collaborative Research Database (eICU-CRD) (208 hospitals) and Medical Information Mart for Intensive Care IV (MIMIC-IV) (1 hospital). The odds ratio of each SOFA component that contributed to mortality was calculated. Mortality was defined as death either in the intensive care unit or within 72 hours of discharge from the intensive care unit. Results: A total of 7,871 intensive care unit stays from eICU-CRD and 4,926 intensive care unit stays from MIMIC-IV were included. Liver dysfunction was most predictive of mortality on Day 1 in both cohorts (OR 1.3; 95%CI 1.2 - 1.4; OR 1.3; 95%CI 1.2 - 1.4, respectively). In the eICU-CRD cohort, central nervous system dysfunction was most predictive of mortality on Day 7 (OR 1.4; 95%CI 1.4 - 1.5). In the MIMIC-IV cohort, respiratory dysfunction (OR 1.4; 95%CI 1.3 - 1.5) and cardiovascular dysfunction (OR 1.4; 95%CI 1.3 - 1.5) were most predictive of mortality on Day 7. Conclusion: The SOFA score may be an oversimplification of how dysfunction of different organ systems contributes to mortality over time. Further research at a more granular timescale is needed to explore how the SOFA score can evolve and be ameliorated.

18 October 2016 in BMC Proceedings
Verified
Journal Article
Open access
PDF file
EBElizabeth Blue
Elizabeth Blue
LBLisa A. Brown
Lisa A. Brown
MCMatthew P. Conomos
Matthew P. Conomos
JKJennifer L. Kirk
Jennifer L. Kirk
ANAlejandro Q. Nato
Alejandro Q. Nato
TTTimothy Thornton
Timothy Thornton
Estimating relationships among subjects in a sample, within family structures or caused by population substructure, is complicated in admixed populations. Inaccurate allele frequencies can bias both kinship estimates and tests for association between subjects and a phenotype. We analyzed the simulated and real family data from Genetic Analysis Workshop 19, and were aware of the simulation model. We found that kinship estimation is more accurate when marker data include common variants whose frequencies are less variable across populations. Estimates of heritability and association vary with age for longitudinally measured traits. Accounting for local ancestry identified different true associations than those identified by a traditional approach. Principal components aid kinship estimation and tests for association, but their utility is influenced by the frequency of the markers used to generate them. Admixed families can provide a powerful resource for detecting disease loci, as well as analytical challenges. Allele frequencies, although difficult to adequately estimate in admixed populations, have a strong impact on the estimation of kinship, ancestry, and association with phenotypes. Approaches that acknowledge population structure in admixed families outperform those which ignore it.

22 October 2024 in medRxiv
Verified
Preprint
Open access
PDF file
HHHarry Hochheiser
Harry Hochheiser
JKJesse Klug
Jesse Klug
TMThomas Mathie
Thomas Mathie
TPTom J. Pollard
Tom J. Pollard
JRJesse D. Raffa
Jesse D. Raffa
CHChristopher Horvat
Christopher Horvat
Objective: To challenge clinicians and informaticians to learn about potential sources of bias in medical machine learning models through investigation of data and predictions from an open-source severity of illness score. Methods: Over a two-day period (total elapsed time approximately 28 hours), we conducted a datathon that challenged interdisciplinary teams to investigate potential sources of bias in the Global Open Source Severity of Illness Score. Teams were invited to develop hypotheses, to use tools of their choosing to identify potential sources of bias, and to provide a final report. Results: Five teams participated, three of which included both informaticians and clinicians. Most (4/5) used Python for analyses, the remaining team used R. Common analysis themes included relationship of the GOSSIS-1 prediction score with demographics and care related variables; relationships between demographics and outcomes; calibration and factors related to the context of care; and the impact of missingness. Representativeness of the population, differences in calibration and model performance among groups, and differences in performance across hospital settings were identified as possible sources of bias. Discussion: Datathons are a promising approach for challenging developers and users to explore questions relating to unrecognized biases in medical machine learning algorithms. Author summary: Disadvantaged groups are at risk of being adversely impacted by biased medical machine learning models. To avoid these undesirable outcomes, developers and users must understand the challenges involved in identifying potential biases. We conducted a datathon aimed at challenging a diverse group of participants to explore an open-source patient severity model for potential biases. Five groups of clinicians and informaticians used tools of their choosing to evaluate possible sources of biases, applying a range of analytic techniques and exploring multiple features. By engaging diverse participants with hands-on data experience with meaningful data, datathons have the potential to raise awareness of potential biases and promote best practices in developing fair and equitable medical machine learning models.

14 January 2026 in The Lancet Regional Health - Americas
Verified
Journal Article
Open access
MLMaximin Lange
Maximin Lange
LCLeo Anthony Celi
Leo Anthony Celi
BCBen Carter
Ben Carter
JRJesse D. Raffa
Jesse D. Raffa
SOSharon C. O'DOnoghue
Sharon C. O'DOnoghue
TPTom J. Pollard
Tom J. Pollard
Physical restraints are widely used in intensive care units (ICUs) despite uncertain clinical benefit and risks. We aimed to characterise patterns of restraint use, demographic and clinical predictors, and temporal trends before and after introduction of federal restraint-related reporting requirements. We conducted a retrospective cross-sectional study of 51,838 adults admitted to ICUs at Beth Israel Deaconess Medical Center, Boston, MA, USA, between 2008 and 2022, using data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) electronic health record repository. Primary outcome was the proportion of ICU days with documented physical restraint use. Associations between restraint use and demographic and clinical factors were estimated using a binomial generalised linear model with a logit link. Propensity score matching compared Black and White patients under varying adjustment specifications. Among 51,838 patients (mean age 63.8 years; 57% male), 21,091 (40.7%) experienced restraint. Use increased from 36.9% in 2008–10 to 44.0% in 2020–22 (p < 0.0001). Asian (aOR 0.84, 95% CI 0.79–0.89) and Hispanic/Latino patients (aOR 0.87, 95% CI 0.83–0.92) had lower odds of restraint than White patients. Propensity score matching between Black and White patients revealed ethnic patterns were highly sensitive to model specification: excluding demographic characteristics revealed significant disparities, which were attenuated when psychiatric diagnoses were also excluded. Matched White patients were not representative of all White ICU patients but rather a subset resembling Black patients on observed characteristics. Restraint practices appear to vary with patient acuity, institutional factors, and communication barriers. The sensitivity of ethnic disparities to psychiatric diagnosis adjustment suggests these diagnoses may function as mediators rather than confounders, potentially reflecting systematic differences in clinical assessment along the causal pathway between ethnicity and restraint decisions. The non-representativeness of matched cohorts underscores that disparities depend on which patient subgroups are compared. Prospective multisite studies with standardized assessment protocols are needed to validate findings, disentangle true clinical variation from systematic bias and provide a more comprehensive understanding of restraint practices across US ICU settings. No study-specific funding was received.

25 April 2025 in medRxiv
Verified
Preprint
Open access
PDF file
MLMaximin Lange
Maximin Lange
LCLeo Anthony Celi
Leo Anthony Celi
BCBen Carter
Ben Carter
JRJesse D Raffa
Jesse D Raffa
SOSharon C O’donoghue
Sharon C O’donoghue
TPTom J Pollard
Tom J Pollard
Objective: To investigate disparities in physical restraint use in a U.S. intensive care unit (ICU) setting, focusing on the influence of demographic factors (ethnicity, sex, age), mental health diagnoses, intubation status, and ICU type. The study also examines trends before and after policy changes in 2014. Methods: This retrospective cross-sectional study uses MIMIC-IV data from adult ICU patients (2008–2022) at Beth Israel Deaconess Medical Center. The primary outcome is the proportion of ICU days with physical restraint. A binomial Generalized Linear Model (GLM) with a logit link function will be used to estimate associations between patient factors and the proportion of ICU time spent in restraints, modeling the number of days with restraint as a binomial outcome with the number of trials equal to the total ICU length of stay. Results will be reported as adjusted odds ratios with 95% confidence intervals. Temporal trends will be evaluated across predefined three-year intervals. Secondary analyses include binary restraint use (yes/no), death within 24 hours of restraint use (yes/no), interaction effects, and multiple sensitivity analyses.