Try a new search

Format these results:

Searched for:



Total Results:


Interpretability, Then What? Editing Machine Learning Models to Reflect Human Knowledge and Values

Chapter by: Wang, Zijie J.; Kale, Alex; Nori, Harsha; Stella, Peter; Nunnally, Mark E.; Chau, Duen Horng; Vorvoreanu, Mihaela; Wortman Vaughan, Jennifer; Caruana, Rich
in: Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining by
[S.l.] : Association for Computing Machinery, 2022
pp. 4132-4142
ISBN: 9781450393850
CID: 5329952

Validation of parsimonious prognostic models for patients infected with COVID-19

Harish, Keerthi; Zhang, Ben; Stella, Peter; Hauck, Kevin; Moussa, Marwa M; Adler, Nicole M; Horwitz, Leora I; Aphinyanaphongs, Yindalon
OBJECTIVES/OBJECTIVE:Predictive studies play important roles in the development of models informing care for patients with COVID-19. Our concern is that studies producing ill-performing models may lead to inappropriate clinical decision-making. Thus, our objective is to summarise and characterise performance of prognostic models for COVID-19 on external data. METHODS:We performed a validation of parsimonious prognostic models for patients with COVID-19 from a literature search for published and preprint articles. Ten models meeting inclusion criteria were either (a) externally validated with our data against the model variables and weights or (b) rebuilt using original features if no weights were provided. Nine studies had internally or externally validated models on cohorts of between 18 and 320 inpatients with COVID-19. One model used cross-validation. Our external validation cohort consisted of 4444 patients with COVID-19 hospitalised between 1 March and 27 May 2020. RESULTS:Most models failed validation when applied to our institution's data. Included studies reported an average validation area under the receiver-operator curve (AUROC) of 0.828. Models applied with reported features averaged an AUROC of 0.66 when validated on our data. Models rebuilt with the same features averaged an AUROC of 0.755 when validated on our data. In both cases, models did not validate against their studies' reported AUROC values. DISCUSSION/CONCLUSIONS:Published and preprint prognostic models for patients infected with COVID-19 performed substantially worse when applied to external data. Further inquiry is required to elucidate mechanisms underlying performance deviations. CONCLUSIONS:Clinicians should employ caution when applying models for clinical prediction without careful validation on local data.
PMID: 34479962
ISSN: 2632-1009
CID: 5000192

Prediction of Resuscitation for Pediatric Sepsis from Data Available at Triage

Stella, Peter; Haines, Elizabeth; Aphinyanaphongs, Yindalon
Pediatric sepsis imposes a significant burden of morbidity and mortality among children. While the speedy application of existing supportive care measures can substantially improve outcomes, further improvements in delivering that care require tools that go beyond recognizing sepsis and towards predicting its development. Machine learning techniques have great potential as predictive tools, but their application to pediatric sepsis has been stymied by several factors, particularly the relative rarity of its occurrence. We propose an alternate approach which focuses on predicting the provision of resuscitative care, rather than sepsis diagnoses or criteria themselves. Using three years of Emergency Department data from a large academic medical center, we developed a boosted tree model that predicts resuscitation within 6 hours of triage, and significantly outperforms existing rule-based sepsis alerts.
PMID: 35308977
ISSN: 1942-597x
CID: 5200392

A validated, real-time prediction model for favorable outcomes in hospitalized COVID-19 patients

Razavian, Narges; Major, Vincent J; Sudarshan, Mukund; Burk-Rafel, Jesse; Stella, Peter; Randhawa, Hardev; Bilaloglu, Seda; Chen, Ji; Nguy, Vuthy; Wang, Walter; Zhang, Hao; Reinstein, Ilan; Kudlowitz, David; Zenger, Cameron; Cao, Meng; Zhang, Ruina; Dogra, Siddhant; Harish, Keerthi B; Bosworth, Brian; Francois, Fritz; Horwitz, Leora I; Ranganath, Rajesh; Austrian, Jonathan; Aphinyanaphongs, Yindalon
The COVID-19 pandemic has challenged front-line clinical decision-making, leading to numerous published prognostic tools. However, few models have been prospectively validated and none report implementation in practice. Here, we use 3345 retrospective and 474 prospective hospitalizations to develop and validate a parsimonious model to identify patients with favorable outcomes within 96 h of a prediction, based on real-time lab values, vital signs, and oxygen support variables. In retrospective and prospective validation, the model achieves high average precision (88.6% 95% CI: [88.4-88.7] and 90.8% [90.8-90.8]) and discrimination (95.1% [95.1-95.2] and 86.8% [86.8-86.9]) respectively. We implemented and integrated the model into the EHR, achieving a positive predictive value of 93.3% with 41% sensitivity. Preliminary results suggest clinicians are adopting these scores into their clinical workflows.
PMID: 33083565
ISSN: 2398-6352
CID: 4640992

Final efficacy results from B-F1RST, a prospective phase II trial evaluating blood-based tumour mutational burden (bTMB) as a predictive biomarker for atezolizumab (atezo) in 1L non-small cell lung cancer (NSCLC) [Meeting Abstract]

Socinski, M; Velcheti, V; Mekhail, T; Chae, Y K; Leal, T A; Dowell, J E; Tsai, M L; Dakhil, C S; Stella, P; Shen, V; Hu, S; Paul, S M; Shames, D S; Schleifman, E; Fabrizio, D A; Nowicki, M; Yun, C; Phan, S; Kim, E S
Background: bTMB assays determine TMB using a noninvasive blood test. B-F1RST (ITT, n=152) is the first prospective trial to evaluate bTMB as a biomarker to predict benefit of 1L atezo monotherapy in advanced NSCLC. bTMB high (score of >=16; >= 14.5 mut/Mb) predicted better ORR with atezo vs bTMB low (< 16; 28.6% vs 4.4%) in the biomarker evaluable population (BEP) with >= 6 mo follow up in the primary analysis. Numerical benefit for bTMB high was seen in median (m)PFS and mOS. Here we report the B-F1RST final analysis.
Method(s): Eligibility criteria included untreated stage IIIB-IVB NSCLC and ECOG PS 0/1. Pts received atezo 1200 mg IV q3w until PD, intolerance or loss of benefit. Co-primary endpoints were investigator assessed ORR for efficacy (ITT) and PFS for biomarker analysis (BEP) at a prespecified bTMB cutoff of 16 for high (>= 16) vs low (< 16). PFS and OS, a secondary endpoint, were further evaluated at various bTMB cutoffs. Serum C-reactive protein (CRP), an inflammation marker in cancer, was evaluated as a surrogate biomarker, ratio of CRP at C3D1 to CRP at screening, to predict PFS and OS.
Result(s): With >= 18 mo follow up (data cutoff, 14 May 2019) in ITT pts, ORR was 17% (95% CI: 12, 24), mPFS was 4.1 mo (95% CI: 2.8, 4.9) and mOS was 14.8mo (95% CI: 12.7, 21.3). In bTMB >= 16 vs<16, mPFS was 5.0 vs 3.5 mo and mOS was 23.9 vs 13.4 mo (Table). For CRP ratio <0.5 vs >=0.5, mPFS was 14.1 vs 4.6 mo (HR, 0.43 [90% CI: 0.24, 0.77]), and mOS was NE vs 15.9 mo (HR, 0.30 [90% CI: 0.13, 0.72]). 14% of pts had treatment-related (TR) serious AEs, and 20% had Gr 3-4 TRAEs. 18% of pts had AEs that led to discontinuation.
Conclusion(s): B-F1RST shows the clinical utility of bTMB as a predictive biomarker for pts receiving 1L atezo monotherapy. The final analysis confirmed that pts with bTMB >= 16 had numerical benefit for PFS and OS. Decrease in serum CRP over 6 wk predicted PFS and OS benefit. No new safety signals were seen. (Table Presented)
ISSN: 1569-8041
CID: 4286072

Pharmaceutical pricing, cost containment and new treatments for rare diseases in children

Stella, Peter; Gold-von Simson, Gabrielle
Cost-containment in healthcare spending has become a central issue in public policy and healthcare reform, especially as the affordable care act adds millions of people to public and private insurance rolls. In this climate, longstanding criticism of pharmaceutical pricing has grown sharper, and many in both policy and medicine have characterized the costs of newly developed drugs as both exorbitant and wasteful of scarce healthcare resources. At the same time, pharmaceutical research and development pipeline costs are increasing exponentially.Price resistance poses a significant threat to the development of drugs to treat rare pediatric diseases, where exceptionally high prices are a sine qua non of commercial viability. This article examines the trends in public discussion of high cost drugs and the potential consequences for orphan drug development. We conclude that despite growing public hostility towards high unit costs, drugs that treat rare diseases in children are likely to remain well-compensated and commercially viable.
PMID: 25348640
ISSN: 1750-1172
CID: 1322682