SKSanjat Kanjilal

Sanjat Kanjilal

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Sanjat Kanjilal has a total of 433 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Predicting persistent fever in patients with cancer receiving antibiotics: a machine learning approach

    by Gernot Pucher, Sanjat Kanjilal, Kevin Kopp, Aman Deep, Till Rostalski, Martin Schuler, Felix Nensa, Jens Kleesiek, H. Christian Reinhardt, Christopher M. Sauer

    Persistent fever after initiating intravenous antibiotics is common in hospitalized patients with cancer and often leads to additional imaging and antimicrobial escalation. Early identification of patients unlikely to remain febrile could provide additional information during clinical reassessment. We developed and externally validated machine learning models to predict persistent fever 48–72 h after antibiotic initiation using electronic health records from two academic hospitals. Adult cancer patients with fever within 24–48 h of treatment were included. Models incorporated demographics, comorbidities, laboratory results, vital signs, and temperature patterns. The best-performing model achieved an AUROC of 0.82 (95% CI 0.79–0.84) internally and 0.78 (95% CI 0.77–0.79) externally. Temperature-related variables were the dominant predictors. Decision curve analysis demonstrated positive net benefit across much of the evaluated threshold range. Retrospective assessment suggested that approximately one quarter of antibiotic escalations and chest CT scans occurred in patients correctly predicted not to remain febrile, highlighting the potential contextual value of fever predictions. These findings indicate that short-term fever persistence can be predicted with clinically meaningful accuracy using routine data. However, clinical use should be limited to an adjunctive role rather than standalone rule-out decision-making, and prospective studies are required to evaluate safety, clinician uptake, and real-world impact.

  • CSF Testing for Neuroinvasive West Nile Virus and Measures to Improve Guideline Adherence

    by Carson M. Quinn, Karan Singh, Arjun Singh, Sarah E. Turbett, Sanjat Kanjilal, George Kyle Harrold, Isaac H. Solomon, Nagagopal Venna, Daniel M. Pastula, Lakshmi Chauhan, et al.

    Background and ObjectivesDiagnosis of West Nile virus (WNV) neuroinvasive disease, an important cause of neurologic disability in endemic areas, requires appropriate testing given its often nonspec...

  • Open Access

    Evaluation of four large language models on complex, infectious disease case scenarios

    by Alyssa Pradhan, Bennett Waxse, Wilfredo Matias, Sarah F Mercaldo, Kathryn Bowman, Cameron Nutt, Sanjat Kanjilal, James M Hillis

    Objectives: Large language models (LLMs) are increasingly used in medicine, but evaluation is often on multiple choice questions and management of common conditions. Infectious diseases (ID) can present complex scenarios that require considerations beyond guideline- based responses. We assessed LLM performance in these situations including with ID- specific criteria to consider infection control or antimicrobial stewardship (AMS). Methods: We evaluated four LLMs (Claude 3.5 Sonnet, GPT-4o, GPT-o1, and a local instance of Llama 3.1 8B) in October 2024, on five complex ID vignettes. The LLM responses were each evaluated for 18 items by two board-certified ID clinicians and pairwise comparisons were performed between LLMs. Results: There was no significant difference between performance of GPT-o1, GPT-4o and Claude Sonnet on general medical criteria, and were comparable with respect to how often they provided an unsafe response (GPT-o1 30%, GPT-4o 40%, Claude 37%) and contained a critical omission (GPT-o1 27%, GPT-4o 43%, Claude 47%). Llama 3.1 8B had significantly decreased performance for most criteria. On ID-specific criteria, GPT-o1 outperformed other models and all models significantly outperformed Llama for interpreting microbiology results, AMS principles, appropriate antimicrobial spectrum and infection control considerations. Performance was poor in secondary prevention and management of risk factors. Conclusions: On complex ID scenarios, LLM responses were variable. The open-source, smaller Llama 3.1 8B model performed poorly and large, non-reasoning models varied, but more than 30% of responses containing a risk of harm or critical omission. These findings suggest caution is required when deploying these models in ID domains without specialist oversight. Summary: This evaluation of LLM responses to complex ID scenarios, demonstrates variable performance between large and small models, with more than 30% of responses containing risk of harm or critical omission. Caution is required when using models without ID-specialist oversight.

  • Open Access

    Prevalence and genetics of cefepime resistance among the AmpC-producing organisms Citrobacter freundii complex, Enterobacter cloacae complex, and Klebsiella aerogenes

    by Sarah M. Schrader, Zachary Pearson, Samantha Taffner, Sanjat Kanjilal, Nicole D. Pecora

    Cefepime is generally recommended for treatment of clinically significant AmpC producers, including Citrobacter freundii complex ( Cfre ), Enterobacter cloacae complex ( Eclo ), and Klebsiella aerogenes ( Kaer ). Clinicians increasingly use the results of multiplex PCR panels to make early treatment decisions, with cefepime often chosen when an AmpC producer is detected without a bla CTX-M extended-spectrum beta-lactamase (ESBL) or carbapenemase. To understand the safety of this approach, we evaluated the prevalence of cefepime resistance among 4,687 Cfre ; 9,878 Eclo ; and 4,800 Kaer clinical isolates collected in our hospital network in New England, USA, from 2015 to 2024. We additionally reviewed Blood Culture Identification 2 (BCID2) (bioMérieux) results for 136 blood cultures and the sequenced genomes of 184 isolates (14 cefepime-susceptible dose-dependent [SDD], 34 cefepime-resistant). Cefepime-SDD and cefepime-resistant isolates made up 0.4% (18/4,687) and 1.3% (59/4,687) of Cfre , 2.0% (194/9,879) and 3.0% (299/9,879) of Eclo , and 0.3% (16/4,800) and 0.6% (28/4,800) of Kaer . Of 117 Eclo and Kaer identified in blood cultures on the BCID2, 4.3% (5/117) were cefepime-SDD or cefepime-resistant and negative for an ESBL or carbapenemase target. Among sequenced cefepime-SDD isolates, 21% (3/14) carried an ESBL. Cefepime-resistant isolates carried diverse beta-lactamase genes, including ESBL genes, carbapenemase genes, and plasmid-borne ampC . While reduced cefepime susceptibility was generally associated with carriage of more beta-lactamase genes, 57% of cefepime-SDD and 35% of cefepime-resistant isolates encoded only chromosomal ampC . Our results highlight variable cefepime susceptibility rates for AmpC producers across clinical settings and demonstrate diverse mechanisms underlying reduced cefepime susceptibility.


  • Tailored care for patients with possible pneumonia who test positive for viruses

    by Brett Biebelberg, Tom Chen, Cara McKenna, Sanjat Kanjilal, Claire Shappell, Chanu Rhee, Michael Klompas

  • Open Access

    Using electronic health records to assess the relationship between colonization pressure and nosocomial pathogen acquisition

    by Luke Sagers, Ziming Wei, Caroline McKenna, Christina Chan, Anna A. Agan, Theodore R. Pak, Chanu Rhee, Michael Klompas, Sanjat Kanjilal

    Hospitalized patients are at risk for developing hospital acquired infections. Active surveillance for bacterial colonization is effective at preventing infection but is resource-intensive and limited to high-risk units and a subset of high-risk pathogens. Colonization pressure, defined as the prevalence of an organism among ward co-occupants, has been associated with hospital acquired infection and can be calculated in real-time using data in the electronic health record, but its use by infection control programs has been limited. As a proof-of-concept study, we built an open source informatics tool to model the impact of colonization pressure on nosocomial acquisition for a wide range of drug susceptible and resistant pathogens using electronic health record data from a large integrated health system in the Northeast United States, collected between May 2015 and July 2024. Using a matched case-control design, we show a consistent positive association between colonization pressure for an organism and nosocomial acquisition of that organism, regardless of whether that organism was drug susceptible or resistant. We also observed significant positive relationships between disparate organisms (e.g., colonization pressure from vancomycin resistant Enterococcus faecium and hospital acquisition of extended-spectrum beta-lactamase producing Klebsiella pneumoniae ) as well as negative associations, primarily between organisms that inhabit distinct niches, such as colonization pressure from drug resistant Pseudomonas aeruginosa and hospital acquisition of vancomycin susceptible Enterococcus faecalis . Our results suggest nosocomial transmission of potential pathogens is widespread in a tertiary care hospital system with advanced infection control programs. We have made the software and our patient-level dataset publicly available to support future research and infection control interventions.


  • Associations between antibiotic use and outcomes in patients hospitalized with community-acquired pneumonia and positive respiratory viral assays

    by Brett Biebelberg, Tom Chen, Cara McKenna, Sanjat Kanjilal, Claire Shappell, Chanu Rhee, Michael Klompas

    Background: Newly released community-acquired pneumonia (CAP) guidelines include a conditional recommendation to treat all hospitalized patients with positive respiratory virus assays with antibacterials. We assessed the frequency, duration, and outcomes of antibacterial prescribing in this population. Methods: We retrospectively identified all hospitalized patients with possible CAP and a positive respiratory virus test at five hospitals, June 2015-December 2024. We used detailed clinical data to propensity-weight patients treated with 0-2 vs 5-7 days of antibacterials and compared outcomes overall and for different viruses. Results: Among 6779 patients with possible CAP and a respiratory virus, 3269 were treated with 0-2 days and 1560 with 5-7 days of antibacterials. After propensity-weighting 2614 patients (1720 treated 0-2 days, 894 treated 5-7 days), there were no significant differences in hospital length of stay (11.7 days vs 11.1 days; OR 1.05, 95% CI 0.97-1.15), ICU admission after 48 hours (28.3% vs 28.2%; OR 1.01, 95% CI 0.86-1.18), in-hospital mortality (9.5% vs 9.8%; OR 0.97, 95% CI 0.74-1.27), or 30-day hospital-free days (16.9 days vs 17.0 days; OR 0.99, 95% CI 0.95-1.03). Results were consistent when restricted to non-SARS-CoV-2 viruses and to influenza alone, when comparing 0 vs 5-7 days of antibacterials, and when restricting to patients with ICD-10 codes for pneumonia present on admission. Conclusions: Antibacterial use for patients with possible CAP and respiratory viruses is highly variable but outcomes are similar with 0-2 vs 5-7 days of antibacterials. This suggests antibiotics are not beneficial in most CAP patients who test positive for respiratory viruses.


  • Open Access

    Nowcasting epidemic trends using hospital- and community-based virologic test data

    by Tse Yang Lim, Sanjat Kanjilal, Shira Doron, Jessica A. Penney, Meredith Haddix, Tae Hee Koo, Phoebe Danza, Rebecca Fisher, Yonatan H. Grad, James A. Hay

    Population viral loads measured by reverse transcription quantitative polymerase chain reaction (RT-qPCR) cycle threshold (Ct) values are an alternative to case counts and hospitalizations for tracking epidemic trends, but their strengths, limitations, and statistical power under various real-world conditions have not been explored. Here, we used SARS-CoV-2 RT-qPCR results from hospital testing in Massachusetts, USA, municipal testing in California, USA, and a combination of theory and simulation analysis to quantify biological and logistical factors impacting Ct-based epidemic nowcasting accuracy. We found that changes to peak viral load, viral growth and clearance rates, and sampling approach and delays all affect the relationship between growth rates and Ct values. We fitted generalized additive models to predict the growth rate and direction of SARS-CoV-2 incidence using time-varying Ct value distributions and assessed nowcasting accuracy over two-week windows. The model predicted epidemic growth rates and direction well from ideal synthetic data (growth rate root-mean-squared error (RMSE) of 0.0192; epidemic direction area under the receiver operating characteristic curve (AUC) of 0.910) but showed modest accuracy with real-world data (RMSE of 0.039-0.052; AUC of 0.72-0.80). Predictions were robust to testing regimes and sample sizes, and trimming outliers improved performance. Our results elucidate the possibilities and limitations of Ct value-based epidemic surveillance, highlighting where they may complement traditional incidence metrics.


  • Open Access

    Antibiotic Resistance Microbiology Dataset (ARMD): A Resource for Antimicrobial Resistance from EHRs

    by Fateme Nateghi Haredasht, Fatemeh Amrollahi, Manoj V. Maddali, Nicholas Marshall, Stephen P. Ma, Lauren N. Cooper, Andrew O. Johnson, Ziming Wei, Richard J. Medford, Sanjat Kanjilal, et al.

    The Antibiotic Resistance Microbiology Dataset (ARMD) is a de-identified resource derived from electronic health records (EHR) that facilitates research in antimicrobial resistance (AMR). ARMD encompasses big data from adult patients collected from over 15 years at two academic-affiliated hospitals, focusing on microbiological cultures, antibiotic susceptibilities, and associated clinical and demographic features. Key attributes include organism identification, susceptibility patterns for 55 antibiotics, implied susceptibility rules, and de-identified patient information. This dataset supports studies on antimicrobial stewardship, causal inference, and clinical decision-making. ARMD is designed to be reusable and interoperable, promoting collaboration and innovation in combating AMR. This paper describes the dataset’s acquisition, structure, and utility while detailing its de-identification process.