About

Mary Catherine Harris, M.D., is a Professor and Chief of Neonatology and Newborn Services in the Department of Pediatrics at the Children's Hospital of Philadelphia. She completed her undergraduate education at Vassar College in 1972 and earned her M.D. from Dartmouth Medical School in 1976. Her professional focus includes neonatal care, with particular interest in sepsis recognition, bacterial meningitis, and infection management in infants. Dr. Harris's research involves the development and refinement of machine learning models for predicting infant sepsis, as well as improving clinical alignment and usability of these models through human-centered design methods. She has contributed to multiple studies on neonatal infections, sepsis prediction, and antibiotic use in neonatal intensive care units, sharing her findings through various presentations and publications.

Research topics

• Medicine
• Immunology
• Pediatrics
• Intensive care medicine
• Internal medicine

Selected publications

• Refining a Machine Learning Model for Predicting Infant Sepsis: A Multidisciplinary Team Supported by Human-Centered Design Methods.

  PubMed · 2025-08-01

  article

  Human-centered design (HCD) methods in machine learning generally focus on workflow, user interfaces, and data visualizations, but there is the potential to apply these methods to inform the model development and testing process.This study aimed to demonstrate the potential of HCD methods to support the design and testing of machine learning models developed for clinical decision-making.In preparing for formative user testing of clinician facing representations of a machine learning model for detecting sepsis in neonatal intensive care unit (NICU) patients, we discovered that interactive low fidelity mockups using real patient data revealed potential model anomalies. To further investigate these potential anomalies, we utilized the qualitative analysis of interviews with 31 NICU clinicians concerning their experience with neonatal sepsis. The review process was conducted by a multidisciplinary team with members having expertise in neonatology, informatics, data science, and human computer interaction (HCI). Anomalies identified via the mockups and interview analysis were further analyzed by inspections of patient charts and model features and code.The HCD-facilitated review revealed anomalies in three categories: (1) feature inclusion and exclusion, (2) feature importance, and (3) model stability over time. Data entry errors in the electronic health record and their impact on model output were also noted. The review resulted in 41 changes to the model.The discovery of over 41 opportunities to improve our prediction model was a serendipitous by-product of the HCD process. Our results suggest that HCD can be applied not only to model display design and measures of explainability, but to the development and evaluation of the model itself. This case report also demonstrates the need for a multidisciplinary team of clinicians, data scientists, and HCI experts in identifying and addressing issues involving machine learning model performance.

  PublisherDOI

• Aligning prediction models with clinical information needs: infant sepsis case study

  JAMIA Open · 2025-03-06 · 1 citations

  articleOpen access

  Abstract Objective Sepsis recognition among infants in the Neonatal Intensive Care Unit (NICU) is challenging and delays in recognition can result in devastating consequences. Although predictive models may improve sepsis outcomes, clinical adoption has been limited. Our focus was to align model behavior with clinician information needs by developing a machine learning (ML) pipeline with two components: (1) a model to predict baseline sepsis risk and (2) a model to detect evolving (dynamic) sepsis risk due to physiologic changes. We then compared the performance of this two-component pipeline to a single model that combines all features reflecting both baseline risk and evolving risk. Materials and Methods We developed prediction models (two-stage pipeline and a single model) using logistic regression and XGBoost trained on electronic healthcare record data of an NICU cohort (1706 observations from 1094 patients, with a 1:1 ratio of cases to controls). We used nested 10-fold cross-validation to evaluate model performance on predictions made 1 h (T−1) before actual clinical recognition. Results The single model (XGBoost) achieved the best performance with a sensitivity of 0.77 (0.74, 0.80), specificity of 0.83 (0.80, 0.85), and positive predictive value (PPV) of 0.82 (0.79, 0.84), at 1 h prior to clinical sepsis recognition (T−1). The pipeline model (XGBoost) achieved a sensitivity of 0.72 (0.69, 0.75), specificity of 0.84 (0.82, 0.87), and PPV of 0.82 (0.80, 0.85) at T−1. Discussion Our findings highlight the challenges of aligning machine learning with NICU clinical decision-making processes. The two-stage pipeline, designed to mirror clinicians’ reasoning, underperformed compared to the single model. Future work should explore integrating continuous physiological data to enhance real-time risk assessment. Conclusion Although a pipeline model that separately estimates baseline and dynamic sepsis risk aligns with clinical information needs, at similar levels of specificity the observed sensitivity of the pipeline is inferior to that of a single model. Additional research is needed to better align model outputs with clinician information needs.

  PublisherOA PDFDOI

• Demystifying Prolonged Antibiotic Use for Blood Culture-negative Sepsis Evaluations in the Neonatal Intensive Care Unit

  The Pediatric Infectious Disease Journal · 2025-04-28 · 2 citations

  articleSenior authorCorresponding

  OBJECTIVE: This study aimed to determine the incidence and clinical characteristics of infants evaluated and treated with a prolonged course of antibiotics for culture-negative sepsis in a quaternary Neonatal Intensive Care Unit (NICU) over a 4-year period. STUDY DESIGN: Retrospective chart review of patients in the NICU at Children's Hospital of Philadelphia who had negative blood cultures and received ≥5 days of antibiotics. Data collection included demographics, clinical and laboratory data, and underlying diagnoses. Statistical analysis included Mann-Whitney and chi-square tests, and multivariable logistic regression. RESULTS: We identified 774 culture-negative sepsis evaluations where antibiotic treatment was continued ≥5 days. While the majority were attributed to a focal etiology, 146 had negative blood cultures and no focal source. Infants with no focal source were younger at the time of sepsis evaluation, of greater gestational age, and more frequently required extracorporeal membrane oxygenation ( P < 0.001). In multivariable analysis, evaluations for early-onset disease and need for extracorporeal membrane oxygenation were increased among infants with no focal source ( P < 0.01). Although rates of invasive ventilation, and central venous catheters were similar, length of stay and mortality were significantly higher in late-onset episodes ( P < 0.001 and P = 0.029, respectively). Consultation with the infectious disease team increased during the study period ( P = 0.002). CONCLUSIONS: Although it is challenging to limit the initiation of antibiotics in infants with complex underlying disease processes with concern for sepsis, minimizing antibiotic use can be achieved by timely discontinuation when cultures are negative. A robust antimicrobial stewardship program can identify valid reasons for prolonged antibiotic administration and suggest approaches to minimize antibiotic exposure.

  PublisherDOI

• Bacterial meningitis in a quaternary NICU: A multiyear retrospective study

  Medicine · 2024-12-20

  articleOpen accessSenior authorCorresponding

  Bacterial meningitis causes significant morbidity and mortality in infants. Lumbar punctures are often deferred until the results of blood cultures are known and sometimes not considered, making this population susceptible to a missed diagnosis. There are few studies describing the epidemiology of neonatal meningitis in quaternary neonatal intensive care unit settings. We describe the epidemiology of meningitis in a level IV neonatal intensive care unit; compare pathogens and rates of concordant bacteremia between infants with and without neurosurgical (NS) devices. Retrospective review of infants < 1 year of age in the Children's Hospital of Philadelphia neonatal intensive care unit with bacterial meningitis (June 2007-October 2021). Analysis included summary statistics, Wilcoxon rank sum, Chi square, and Fisher exact tests. We identified 101 episodes of bacterial meningitis (95 infants). 9 infants died. At diagnosis, 26 infants (27%) had NS devices. Group B streptococcus (GBS) and Escherichia coli (E coli) were most common pathogens, however, coagulase-negative staphylococci and Staphylococcus aureus (S aureus) predominated among infants with NS devices. While 86% had positive blood cultures in the absence of a NS device, only 14% of episodes with NS devices had concomitant bacteremia (P < .0001). Although Group B streptococcus and E coli remain most prevalent overall, coagulase-negative staphylococci and S aureus were common pathogens in NS patients. Infants with NS devices rarely had concomitant bacteremia. Meningitis was diagnosed in the absence of a positive blood culture in 36% of episodes, underscoring the importance of developing guidance for lumbar punctures in infants evaluated for sepsis.

  PublisherDOI

• A Case of Resistant Status Asthmaticus: Resistant to Steroids and Responsive to IV Epinephrine

  International Journal of Respiratory and Pulmonary Medicine · 2024-01-19

  articleOpen accessSenior author

  Status Asthmaticus, is a medical emergency, an extreme form of asthma exacerbation characterized by hypoxic and hypercapnic respiratory failure with resistance to standard therapy including inhaled selective beta-adrenergic agonists, low dose systemic steroids and usually responds to high dose systemic steroids with continuous nebulization of bronchodilators. Epinephrine, while not significantly advantageous in cases of mild to moderate asthma compared to standard therapy, can be significantly useful in cases of resistant, life-threatening asthma.

  PublisherOA PDFDOI

• Sepsis Huddles in the Neonatal Intensive Care Unit: A Retrospective Cohort Study of Late-onset Infection Recognition and Severity Assessment

  The Journal of Pediatrics · 2024-05-28 · 2 citations

  article
  PublisherDOI

• Use of an Escape Room to Enhance Skills and Self-efficacy Utilizing the Nutrition Care Process

  Journal of the Academy of Nutrition and Dietetics · 2023-08-17

  articleSenior authorCorresponding
  PublisherDOI

• Using a Sociotechnical Model to Understand Challenges with Sepsis Recognition among Critically Ill Infants

  ACI Open · 2022-07-01 · 4 citations

  articleOpen access

  Abstract Objective The aim of the study is to apply a sociotechnical model to the requirements phase of implementing a machine learning algorithm-based system to support sepsis recognition in the neonatal intensive care unit. Methods We incorporated components from the sociotechnical model, Safety in Engineering for Patient Safety 2.0, in three requirements phase activities: (1) semi-structured interviews, (2) user profiles, and (3) system use cases. Results Thirty-one neonatal intensive care unit clinicians participated in semi-structured interviews (11 nurses, 10 front line ordering clinician, five fellows, and five attending physician). Interview transcripts were coded and then compiled into themes deductively based on components from the sociotechnical model (persons, environment, organization, tasks, tools and technology, collaboration, and outcomes). The interview analysis was used to create four user profiles defining responsibilities in sepsis recognition, team collaboration, and attributes relevant to sepsis recognition. Two user profiles (nurse, front line ordering clinician) included variants based on experience relevant to sepsis recognition. The interview analysis was used to develop three system use cases representing clinical sepsis scenarios. Each use case defines the precondition, actors, and high-level sequence of actions, and includes variants based on sociotechnical works system factors that can complicate sepsis recognition. The interview analysis, user profiles, and use cases serve as the foundation for supporting sociotechnical design to all subsequent human-centered design methods including subject recruitment, formative design, summative user testing, and simulation testing. Conclusion Integration of the sociotechnical model-guided requirements gathering activities, analysis, and deliverables by framing a range of sociotechnical components and the interconnectedness of these components in the broader work system. Applying the sociotechnical model resulted in discovering work system, process, and outcome requirements that would otherwise be difficult to capture, or missed entirely, using traditional requirements gathering methods or approaches to clinical decision support design.

  PublisherOA PDFDOI

• Dual-site blood culture yield and time to positivity in neonatal late-onset sepsis

  Archives of Disease in Childhood Fetal & Neonatal · 2021-11-09 · 8 citations

  article

  Objective To determine whether culture yield and time to positivity (TTP) differ between peripheral and central vascular catheter-derived blood cultures (BCx) in neonatal intensive care unit (NICU) patients evaluated for late-onset sepsis. Design Single-centre, retrospective, observational study. Setting Level IV NICU. Participants The study included infants &gt;72 hours old admitted to NICU in 2007–2019 with culture-confirmed bacteraemia. All episodes had simultaneous BCx drawn from a peripheral site and a vascular catheter (‘catheter culture’). Main outcome measures Dual-site culture yield and TTP. Results Among 179 episodes of late-onset bacteraemia (among 167 infants) with concurrently drawn peripheral and catheter BCx, the majority (67%, 120 of 179) were positive from both sites, compared with 17% (30 of 179) with positive catheter cultures only and 16% (29 of 179) with positive peripheral cultures only. 66% (19 of 29) of episodes with only positive peripheral BCx grew coagulase-negative Staphylococcus , while 34% (10 of 29) were recognised bacterial pathogens. Among 120 episodes with both peripheral and catheter BCx growth, catheter cultures demonstrated bacterial growth prior to paired peripheral cultures in 78% of episodes (93 of 120, p&lt;0.001). The median TTP was significantly shorter in catheter compared with peripheral cultures (15.0 hours vs 16.8 hours, p&lt;0.001). The median elapsed time between paired catheter and peripheral culture growth was 1.3 hours. Conclusion Concurrently drawn peripheral and catheter BCx had similar yield. While a majority of episodes demonstrated dual-site BCx growth, a small but important minority of episodes grew virulent pathogens from either culture site alone. While dual-site culture practices may be useful, clinicians should balance the gain in sensitivity of bacteraemia detection against additive contamination risk.

  PublisherDOI

• Derivation of a metabolic signature associated with bacterial meningitis in infants

  Pediatric Research · 2020-03-02 · 16 citations

  articleOpen accessSenior author
  PublisherOA PDFDOI

Recent grants

• NIH Grant M01RR000240

  NIH · $6.4M · 2006

Frequent coauthors

• Lakshmi Srinivasan

  University of Pennsylvania

  66 shared

• Richard A. Polin

  Research Network (United States)

  58 shared

• Steven D. Douglas

  53 shared

• Laurie E. Kilpatrick

  Temple University

  50 shared

• Paul R. Gallagher

  Children's Hospital of Philadelphia

  39 shared

• Carl T. D’Angio

  33 shared

• Robert W. Grundmeier

  Children's Hospital of Philadelphia

  28 shared

• David Kaufman

  27 shared

Labs

• Mary Catherine Harris LabPI

Education

• M.D., Medicine

  Dartmouth Medical School

  1976