Academic Appointments

All Publications

  • Leukostasis-induced digital ischemia. EJHaem Hasty, A., Joshi, M., Lee, D., Mannis, G. N. 2023; 4 (2): 497-498

    View details for DOI 10.1002/jha2.666

    View details for PubMedID 37206295

    View details for PubMedCentralID PMC10188469

  • Association of clinical prediction scores with hospital mortality in an adult medical and surgical intensive care unit in Kenya. Frontiers in medicine Brotherton, B. J., Joshi, M., Otieno, G., Wandia, S., Gitura, H., Mueller, A., Nguyen, T., Letchford, S., Riviello, E. D., Karanja, E., Rudd, K. E. 2023; 10: 1127672


    Mortality prediction among critically ill patients in resource limited settings is difficult. Identifying the best mortality prediction tool is important for counseling patients and families, benchmarking quality improvement efforts, and defining severity of illness for clinical research studies.Compare predictive capacity of the Modified Early Warning Score (MEWS), Universal Vital Assessment (UVA), Tropical Intensive Care Score (TropICS), Rwanda Mortality Probability Model (R-MPM), and quick Sequential Organ Failure Assessment (qSOFA) for hospital mortality among adults admitted to a medical-surgical intensive care unit (ICU) in rural Kenya. We performed a pre-planned subgroup analysis among ICU patients with suspected infection.Prospective single-center cohort study at a tertiary care, academic hospital in Kenya. All adults 18 years and older admitted to the ICU January 2018-June 2019 were included.The primary outcome was association of clinical prediction tool score with hospital mortality, as defined by area under the receiver operating characteristic curve (AUROC). Demographic, physiologic, laboratory, therapeutic, and mortality data were collected. 338 patients were included, none were excluded. Median age was 42 years (IQR 33-62) and 61% (n = 207) were male. Fifty-nine percent (n = 199) required mechanical ventilation and 35% (n = 118) received vasopressors upon ICU admission. Overall hospital mortality was 31% (n = 104). 323 patients had all component variables recorded for R-MPM, 261 for MEWS, and 253 for UVA. The AUROC was highest for MEWS (0.76), followed by R-MPM (0.75), qSOFA (0.70), and UVA (0.69) (p < 0.001). Predictive capacity was similar among patients with suspected infection.All tools had acceptable predictive capacity for hospital mortality, with variable observed availability of the component data. R-MPM and MEWS had high rates of variable availability as well as good AUROC, suggesting these tools may prove useful in low resource ICUs.

    View details for DOI 10.3389/fmed.2023.1127672

    View details for PubMedID 37089585

    View details for PubMedCentralID PMC10113620

  • Current and Future Applications of Artificial Intelligence in Cardiac CT. Current cardiology reports Joshi, M., Melo, D. P., Ouyang, D., Slomka, P. J., Williams, M. C., Dey, D. 2023


    PURPOSE OF REVIEW: In this review, we aim to summarize state-of-the-art artificial intelligence (AI) approaches applied to cardiovascular CT and their future implications.RECENT FINDINGS: Recent studies have shown that deep learning networks can be applied for rapid automated segmentation of coronary plaque from coronary CT angiography, with AI-enabled measurement of total plaque volume predicting future heart attack. AI has also been applied to automate assessment of coronary artery calciumon cardiac and ungated chest CT and to automate the measurement of epicardial fat. Additionally, AI-based prediction models integrating clinical and imaging parameters have been shown to improve prediction of cardiac events compared to traditional risk scores. Artificial intelligence applications have been applied in all aspects of cardiovascular CT - in image acquisition, reconstruction and denoising, segmentation and quantitative analysis, diagnosis and decision assistance and to integrate prognostic risk from clinical data and images. Further incorporation of artificial intelligence in cardiovascular imaging holds important promise to enhance cardiovascular CT as a precision medicine tool.

    View details for DOI 10.1007/s11886-022-01837-8

    View details for PubMedID 36708505

  • Implementation approaches and barriers for rule-based and machine learning-based sepsis risk prediction tools: a qualitative study. JAMIA open Joshi, M., Mecklai, K., Rozenblum, R., Samal, L. 2022; 5 (2): ooac022


    Many options are currently available for sepsis surveillance clinical decision support (CDS) from electronic medical record (EMR) vendors, third party, and homegrown models drawing on rule-based (RB) and machine learning (ML) algorithms. This study explores sepsis CDS implementation from the perspective of implementation leads by describing the motivations, tool choices, and implementation experiences of a diverse group of implementers.Semi-structured interviews were conducted with and a questionnaire was administered to 21 hospital leaders overseeing CDS implementation at 15 US medical centers. Participants were recruited via convenience sampling. Responses were coded by 2 coders with consensus approach and inductively analyzed for themes.Use of sepsis CDS is motivated in part by quality metrics for sepsis patients. Choice of tool is driven by ease of integration, customization capability, and perceived predictive potential. Implementation processes for these CDS tools are complex, time-consuming, interdisciplinary undertakings resulting in heterogeneous choice of tools and workflow integration. To improve clinician acceptance, implementers addressed both optimization of the alerts as well as clinician understanding and buy in. More distrust and confusion was reported for ML models, as compared to RB models. Respondents described a variety of approaches to overcome implementation barriers; these approaches related to alert firing, content, integration, and buy-in.While there are shared socio-technical challenges of implementing CDS for both RB and ML models, attention to user education, support, expectation management, and dissemination of effective practices may improve feasibility and effectiveness of ML models in quality improvement efforts.Further implementation science research is needed to determine real world efficacy of these tools. Clinician acceptance is a significant barrier to sepsis CDS implementation. Successful implementation of less clinically intuitive ML models may require additional attention to user confusion and distrust.

    View details for DOI 10.1093/jamiaopen/ooac022

    View details for PubMedID 35474719

    View details for PubMedCentralID PMC9030109

  • Training Internal Medicine Residents in Difficult Diagnosis: A Novel Diagnostic Second Opinion Clinic Experience. Journal of medical education and curricular development Testa, S., Joshi, M., Lotfi, J., Lin, B., Artandi, M., Chiang, K. F., Chang, K., Singh, B., Geng, L. N. 2022; 9: 23821205221091036


    Background: In primary care clinics, time constraints and lack of exposure to highly complex cases may limit the breadth and depth of learning for internal medicine residents. To address these issues, we piloted a novel experience for residents to evaluate patients with puzzling symptoms referred by another clinician.Objective: To increase internal medicine residents' exposure to patients with perplexing presentations and foster a team-based approach to solving diagnostically challenging cases.Methods: During the academic year 2020-2021, residents participating in their 2-week primary care "block" rotation were given protected time to evaluate 1-2 patients from the Stanford Consultative Medicine clinic, an internist-led diagnostic second opinion service, and present their patients at the case conference. We assessed the educational value of the program with resident surveys including 5-point Lickert scale and open-ended questions.Results: 21 residents participated in the pilot with a survey response rate of 66.6% (14/21). Both the educational value and overall quality of the experience were rated as 4.8 out of 5 (SD 0.4, range 4-5; 1:"very poor"; 5:"excellent"). Residents learned about new diagnostic tools as well as how to approach complex presentations and diagnostic dilemmas. Residents valued the increased time devoted to patient care, the team-based approach to tackling difficult cases, and the intellectual challenge of these cases. Barriers to implementation include patient case volume, time, and faculty engagement.Conclusions: Evaluation of diagnostically challenging cases in a structured format is a highly valuable experience that offers a framework to enhance outpatient training in internal medicine.

    View details for DOI 10.1177/23821205221091036

    View details for PubMedID 35372696

  • Patient factors predict complicationsafter partial nephrectomy: Validation and calibration of the PREP (Preoperative Risk Evaluation for Partial Nephrectomy) score. BJU international Huynh, M. J., Wang, Y., Joshi, M., Krasnow, R., Yu, A. X., Mossanen, M., Chung, B. I., Chang, S. L. 2020


    OBJECTIVES: To develop and validate the PREP (Preoperative Risk Evaluation for Partial Nephrectomy) score to predict the probability of major postoperative complications following partial nephrectomy (PN) based on patient comorbidities.PATIENTS AND METHODS: The Premier Healthcare Database was used to identify patients who had undergone elective PN. Through review of ICD-9 codes, we identified patient comorbidities and major surgical complications (Clavien grade 3-5). Multivariable logistic regression was used to identify predictors of major complications. We used half of the set as the training cohort to develop our risk score and the other half as a validation cohort.RESULTS: From 2003-2015, 25,451 PN were performed. The overall rate of major complications was 4.9%. The final risk score consisted of 10 predictors: age, sex, CHF, CAD, COPD, CKD, diabetes, hypertension, obesity, smoking. In the training cohort, the area under the receiver-operator characteristic curve (AUC) was 0.75 (95% CI 0.73-0.78), while the AUC for the validation cohort was 0.73 (95% CI 0.70-0.75). The predicted probabilities of major complication in the low risk (≤10 points), intermediate risk (11-20 points), high risk (21-30 points), and very high risk (>30 points) categories were 3% (95% CI 2.6-3.2), 8% (95% CI 7.2-9.2), 24% (95% CI 20.5-27.8), and 41% (95% CI 34.5-47.8) respectively.CONCLUSIONS: We developed and validated the PREP score to predict the risk of complications following PN based on patient characteristics. Calculation of the PREP score can help providers select treatment options for patients with a cT1a renal mass and enhance the informed consent process for patients planning to undergo PN.

    View details for DOI 10.1111/bju.15240

    View details for PubMedID 32920933