Health system-scale language models are all-purpose prediction engines.

Physicians make critical time-constrained decisions every day. Clinical predictive models can help physicians and administrators make decisions by forecasting clinical and operational events. Existing structured data-based clinical predictive models have limited use in everyday practice owing to complexity in data processing, as well as model development and deployment1-3. Here we show that unstructured clinical notes from the electronic health record can enable the training of clinical language models, which can be used as all-purpose clinical predictive engines with low-resistance development and deployment. Our approach leverages recent advances in natural language processing4,5 to train a large language model for medical language (NYUTron) and subsequently fine-tune it across a wide range of clinical and operational predictive tasks. We evaluated our approach within our health system for five such tasks: 30-day all-cause readmission prediction, in-hospital mortality prediction, comorbidity index prediction, length of stay prediction, and insurance denial prediction. We show that NYUTron has an area under the curve (AUC) of 78.7-94.9%, with an improvement of 5.36-14.7% in the AUC compared with traditional models. We additionally demonstrate the benefits of pretraining with clinical text, the potential for increasing generalizability to different sites through fine-tuning and the full deployment of our system in a prospective, single-arm trial. These results show the potential for using clinical language models in medicine to read alongside physicians and provide guidance at the point of care.

Practical recommendations for using ctDNA in clinical decision making.

The continuous improvement in cancer care over the past decade has led to a gradual decrease in cancer-related deaths. This is largely attributed to improved treatment and disease management strategies. Early detection of recurrence using blood-based biomarkers such as circulating tumour DNA (ctDNA) is being increasingly used in clinical practice. Emerging real-world data shows the utility of ctDNA in detecting molecular residual disease and in treatment-response monitoring, helping clinicians to optimize treatment and surveillance strategies. Many studies have indicated ctDNA to be a sensitive and specific biomarker for recurrence. However, most of these studies are largely observational or anecdotal in nature, and peer-reviewed data regarding the use of ctDNA are mainly indication-specific. Here we provide general recommendations on the clinical utility of ctDNA and how to interpret ctDNA analysis in different treatment settings, especially in patients with solid tumours. Specifically, we provide an understanding around the implications, strengths and limitations of this novel biomarker and how to best apply the results in clinical practice.

Ultra-fast deep-learned CNS tumour classification during surgery.

Central nervous system tumours represent one of the most lethal cancer types, particularly among children1. Primary treatment includes neurosurgical resection of the tumour, in which a delicate balance must be struck between maximizing the extent of resection and minimizing risk of neurological damage and comorbidity2,3. However, surgeons have limited knowledge of the precise tumour type prior to surgery. Current standard practice relies on preoperative imaging and intraoperative histological analysis, but these are not always conclusive and occasionally wrong. Using rapid nanopore sequencing, a sparse methylation profile can be obtained during surgery4. Here we developed Sturgeon, a patient-agnostic transfer-learned neural network, to enable molecular subclassification of central nervous system tumours based on such sparse profiles. Sturgeon delivered an accurate diagnosis within 40 minutes after starting sequencing in 45 out of 50 retrospectively sequenced samples (abstaining from diagnosis of the other 5 samples). Furthermore, we demonstrated its applicability in real time during 25 surgeries, achieving a diagnostic turnaround time of less than 90 min. Of these, 18 (72%) diagnoses were correct and 7 did not reach the required confidence threshold. We conclude that machine-learned diagnosis based on low-cost intraoperative sequencing can assist neurosurgical decision-making, potentially preventing neurological comorbidity and avoiding additional surgeries.

Circulating tumor DNA in advanced solid tumors: Clinical relevance and future directions.

The application of genomic profiling assays using plasma circulating tumor DNA (ctDNA) is rapidly evolving in the management of patients with advanced solid tumors. Diverse plasma ctDNA technologies in both commercial and academic laboratories are in routine or emerging use. The increasing integration of such testing to inform treatment decision making by oncology clinicians has complexities and challenges but holds significant potential to substantially improve patient outcomes. In this review, the authors discuss the current role of plasma ctDNA assays in oncology care and provide an overview of ongoing research that may inform real-world clinical applications in the near future.

Genomics to select treatment for patients with metastatic breast cancer.

Cancer progression is driven in part by genomic alterations1. The genomic characterization of cancers has shown interpatient heterogeneity regarding driver alterations2, leading to the concept that generation of genomic profiling in patients with cancer could allow the selection of effective therapies3,4. Although DNA sequencing has been implemented in practice, it remains unclear how to use its results. A total of 1,462 patients with HER2-non-overexpressing metastatic breast cancer were enroled to receive genomic profiling in the SAFIR02-BREAST trial. Two hundred and thirty-eight of these patients were randomized in two trials (nos. NCT02299999 and NCT03386162) comparing the efficacy of maintenance treatment5 with a targeted therapy matched to genomic alteration. Targeted therapies matched to genomics improves progression-free survival when genomic alterations are classified as level I/II according to the ESMO Scale for Clinical Actionability of Molecular Targets (ESCAT)6 (adjusted hazards ratio (HR): 0.41, 90% confidence interval (CI): 0.27-0.61, P < 0.001), but not when alterations are unselected using ESCAT (adjusted HR: 0.77, 95% CI: 0.56-1.06, P = 0.109). No improvement in progression-free survival was observed in the targeted therapies arm (unadjusted HR: 1.15, 95% CI: 0.76-1.75) for patients presenting with ESCAT alteration beyond level I/II. Patients with germline BRCA1/2 mutations (n = 49) derived high benefit from olaparib (gBRCA1: HR = 0.36, 90% CI: 0.14-0.89; gBRCA2: HR = 0.37, 90% CI: 0.17-0.78). This trial provides evidence that the treatment decision led by genomics should be driven by a framework of target actionability in patients with metastatic breast cancer.

Multidisciplinary standards of care and recent progress in pancreatic ductal adenocarcinoma.

Despite tremendous gains in the molecular understanding of exocrine pancreatic cancer, the prognosis for this disease remains very poor, largely because of delayed disease detection and limited effectiveness of systemic therapies. Both incidence rates and mortality rates for pancreatic cancer have increased during the past decade, in contrast to most other solid tumor types. Recent improvements in multimodality care have substantially improved overall survival, local control, and metastasis-free survival for patients who have localized tumors that are amenable to surgical resection. The widening gap in prognosis between patients with resectable and unresectable or metastatic disease reinforces the importance of detecting pancreatic cancer sooner to improve outcomes. Furthermore, the developing use of therapies that target tumor-specific molecular vulnerabilities may offer improved disease control for patients with advanced disease. Finally, the substantial morbidity associated with pancreatic cancer, including wasting, fatigue, and pain, remains an under-addressed component of this disease, which powerfully affects quality of life and limits tolerance to aggressive therapies. In this article, the authors review the current multidisciplinary standards of care in pancreatic cancer with a focus on emerging concepts in pancreatic cancer detection, precision therapy, and survivorship.

Swarm Learning for decentralized and confidential clinical machine learning.

Fast and reliable detection of patients with severe and heterogeneous illnesses is a major goal of precision medicine1,2. Patients with leukaemia can be identified using machine learning on the basis of their blood transcriptomes3. However, there is an increasing divide between what is technically possible and what is allowed, because of privacy legislation4,5. Here, to facilitate the integration of any medical data from any data owner worldwide without violating privacy laws, we introduce Swarm Learning-a decentralized machine-learning approach that unites edge computing, blockchain-based peer-to-peer networking and coordination while maintaining confidentiality without the need for a central coordinator, thereby going beyond federated learning. To illustrate the feasibility of using Swarm Learning to develop disease classifiers using distributed data, we chose four use cases of heterogeneous diseases (COVID-19, tuberculosis, leukaemia and lung pathologies). With more than 16,400 blood transcriptomes derived from 127 clinical studies with non-uniform distributions of cases and controls and substantial study biases, as well as more than 95,000 chest X-ray images, we show that Swarm Learning classifiers outperform those developed at individual sites. In addition, Swarm Learning completely fulfils local confidentiality regulations by design. We believe that this approach will notably accelerate the introduction of precision medicine.

Trial of an Intervention to Improve Acute Heart Failure Outcomes.

BACKGROUND: Patients with acute heart failure are frequently or systematically hospitalized, often because the risk of adverse events is uncertain and the options for rapid follow-up are inadequate. Whether the use of a strategy to support clinicians in making decisions about discharging or admitting patients, coupled with rapid follow-up in an outpatient clinic, would affect outcomes remains uncertain. METHODS: In a stepped-wedge, cluster-randomized trial conducted in Ontario, Canada, we randomly assigned 10 hospitals to staggered start dates for one-way crossover from the control phase (usual care) to the intervention phase, which involved the use of a point-of-care algorithm to stratify patients with acute heart failure according to the risk of death. During the intervention phase, low-risk patients were discharged early (in ≤3 days) and received standardized outpatient care, and high-risk patients were admitted to the hospital. The coprimary outcomes were a composite of death from any cause or hospitalization for cardiovascular causes within 30 days after presentation and the composite outcome within 20 months. RESULTS: A total of 5452 patients were enrolled in the trial (2972 during the control phase and 2480 during the intervention phase). Within 30 days, death from any cause or hospitalization for cardiovascular causes occurred in 301 patients (12.1%) who were enrolled during the intervention phase and in 430 patients (14.5%) who were enrolled during the control phase (adjusted hazard ratio, 0.88; 95% confidence interval [CI], 0.78 to 0.99; P = 0.04). Within 20 months, the cumulative incidence of primary-outcome events was 54.4% (95% CI, 48.6 to 59.9) among patients who were enrolled during the intervention phase and 56.2% (95% CI, 54.2 to 58.1) among patients who were enrolled during the control phase (adjusted hazard ratio, 0.95; 95% CI, 0.92 to 0.99). Fewer than six deaths or hospitalizations for any cause occurred in low- or intermediate-risk patients before the first outpatient visit within 30 days after discharge. CONCLUSIONS: Among patients with acute heart failure who were seeking emergency care, the use of a hospital-based strategy to support clinical decision making and rapid follow-up led to a lower risk of the composite of death from any cause or hospitalization for cardiovascular causes within 30 days than usual care. (Funded by the Ontario SPOR Support Unit and others; COACH ClinicalTrials.gov number, NCT02674438.).

Using measures of race to make clinical predictions: Decision making, patient health, and fairness.

The use of race measures in clinical prediction models is contentious. We seek to inform the discourse by evaluating the inclusion of race in probabilistic predictions of illness that support clinical decision making. Adopting a static utilitarian framework to formalize social welfare, we show that patients of all races benefit when clinical decisions are jointly guided by patient race and other observable covariates. Similar conclusions emerge when the model is extended to a two-period setting where prevention activities target systemic drivers of disease. We also discuss non-utilitarian concepts that have been proposed to guide allocation of health care resources.

Experimental evidence for structured information-sharing networks reducing medical errors.

Errors in clinical decision-making are disturbingly common. Recent studies have found that 10 to 15% of all clinical decisions regarding diagnoses and treatment are inaccurate. Here, we experimentally study the ability of structured information-sharing networks among clinicians to improve clinicians' diagnostic accuracy and treatment decisions. We use a pool of 2,941 practicing clinicians recruited from around the United States to conduct 84 independent group-level trials, ranging across seven different clinical vignettes for topics known to exhibit high rates of diagnostic or treatment error (e.g., acute cardiac events, geriatric care, low back pain, and diabetes-related cardiovascular illness prevention). We compare collective performance in structured information-sharing networks to collective performance in independent control groups, and find that networks significantly reduce clinical errors, and improve treatment recommendations, as compared to control groups of independent clinicians engaged in isolated reflection. Our results show that these improvements are not a result of simple regression to the group mean. Instead, we find that within structured information-sharing networks, the worst clinicians improved significantly while the best clinicians did not decrease in quality. These findings offer implications for the use of social network technologies to reduce errors among clinicians.

Collaborative privacy-preserving analysis of oncological data using multiparty homomorphic encryption.

Real-world healthcare data sharing is instrumental in constructing broader-based and larger clinical datasets that may improve clinical decision-making research and outcomes. Stakeholders are frequently reluctant to share their data without guaranteed patient privacy, proper protection of their datasets, and control over the usage of their data. Fully homomorphic encryption (FHE) is a cryptographic capability that can address these issues by enabling computation on encrypted data without intermediate decryptions, so the analytics results are obtained without revealing the raw data. This work presents a toolset for collaborative privacy-preserving analysis of oncological data using multiparty FHE. Our toolset supports survival analysis, logistic regression training, and several common descriptive statistics. We demonstrate using oncological datasets that the toolset achieves high accuracy and practical performance, which scales well to larger datasets. As part of this work, we propose a cryptographic protocol for interactive bootstrapping in multiparty FHE, which is of independent interest. The toolset we develop is general-purpose and can be applied to other collaborative medical and healthcare application domains.

AGA Clinical Practice Guideline on Endoscopic Eradication Therapy of Barrett's Esophagus and Related Neoplasia.

BACKGROUND & AIMS: Barrett's esophagus (BE) is the precursor to esophageal adenocarcinoma (EAC). Endoscopic eradication therapy (EET) can be effective in eradicating BE and related neoplasia and has greater risk of harms and resource use than surveillance endoscopy. This clinical practice guideline aims to inform clinicians and patients by providing evidence-based practice recommendations for the use of EET in BE and related neoplasia. METHODS: The Grading of Recommendations Assessment, Development and Evaluation framework was used to assess evidence and make recommendations. The panel prioritized clinical questions and outcomes according to their importance for clinicians and patients, conducted an evidence review, and used the Evidence-to-Decision Framework to develop recommendations regarding the use of EET in patients with BE under the following scenarios: presence of (1) high-grade dysplasia, (2) low-grade dysplasia, (3) no dysplasia, and (4) choice of stepwise endoscopic mucosal resection (EMR) or focal EMR plus ablation, and (5) endoscopic submucosal dissection vs EMR. Clinical recommendations were based on the balance between desirable and undesirable effects, patient values, costs, and health equity considerations. RESULTS: The panel agreed on 5 recommendations for the use of EET in BE and related neoplasia. Based on the available evidence, the panel made a strong recommendation in favor of EET in patients with BE high-grade dysplasia and conditional recommendation against EET in BE without dysplasia. The panel made a conditional recommendation in favor of EET in BE low-grade dysplasia; patients with BE low-grade dysplasia who place a higher value on the potential harms and lower value on the benefits (which are uncertain) regarding reduction of esophageal cancer mortality could reasonably select surveillance endoscopy. In patients with visible lesions, a conditional recommendation was made in favor of focal EMR plus ablation over stepwise EMR. In patients with visible neoplastic lesions undergoing resection, the use of either endoscopic mucosal resection or endoscopic submucosal dissection was suggested based on lesion characteristics. CONCLUSIONS: This document provides a comprehensive outline of the indications for EET in the management of BE and related neoplasia. Guidance is also provided regarding the considerations surrounding implementation of EET. Providers should engage in shared decision making based on patient preferences. Limitations and gaps in the evidence are highlighted to guide future research opportunities.

Deep transfer learning for clinical decision-making based on high-throughput data: comprehensive survey with benchmark results.

The rapid growth of omics-based data has revolutionized biomedical research and precision medicine, allowing machine learning models to be developed for cutting-edge performance. However, despite the wealth of high-throughput data available, the performance of these models is hindered by the lack of sufficient training data, particularly in clinical research (in vivo experiments). As a result, translating this knowledge into clinical practice, such as predicting drug responses, remains a challenging task. Transfer learning is a promising tool that bridges the gap between data domains by transferring knowledge from the source to the target domain. Researchers have proposed transfer learning to predict clinical outcomes by leveraging pre-clinical data (mouse, zebrafish), highlighting its vast potential. In this work, we present a comprehensive literature review of deep transfer learning methods for health informatics and clinical decision-making, focusing on high-throughput molecular data. Previous reviews mostly covered image-based transfer learning works, while we present a more detailed analysis of transfer learning papers. Furthermore, we evaluated original studies based on different evaluation settings across cross-validations, data splits and model architectures. The result shows that those transfer learning methods have great potential; high-throughput sequencing data and state-of-the-art deep learning models lead to significant insights and conclusions. Additionally, we explored various datasets in transfer learning papers with statistics and visualization.

Clinical decision-making and treatment of myelodysplastic syndromes.

ABSTRACT: The myelodysplastic syndromes (MDSs) constitute a profoundly heterogeneous myeloid malignancy with a common origin in the hemopoietic stem cell compartment. Consequently, patient management and treatment are as heterogeneous. Decision-making includes identifying risk, symptoms, and options for an individual and conducting a risk-benefit analysis. The only potential cure is allogeneic stem cell transplantation, and albeit the fraction of patients with MDS who undergo transplant increase over time because of better management and increased donor availability, a majority are not eligible for this intervention. Current challenges encompass to decrease the relapse risk, the main cause of hematopoietic stem cell transplantation failure. Hypomethylating agents (HMAs) constitute firstline treatment for higher-risk MDSs. Combinations with other drugs as firstline treatment has, to date, not proven more efficacious than monotherapy, although combinations approved for acute myeloid leukemia, including venetoclax, are under evaluation and often used as rescue treatment. The treatment goal for lower-risk MDS is to improve cytopenia, mainly anemia, quality of life, and, possibly, overall survival. Erythropoiesis-stimulating agents (ESAs) constitute firstline treatment for anemia and have better and more durable responses if initiated before the onset of a permanent transfusion need. Treatment in case of ESA failure or ineligibility should be tailored to the main disease mechanism: immunosuppression for hypoplastic MDS without high-risk genetics, lenalidomide for low-risk del(5q) MDS, and luspatercept for MDS with ring sideroblasts. Approved therapeutic options are still scarcer for MDS than for most other hematologic malignancies. Better tools to match disease biology with treatment, that is, applied precision medicines are needed to improve patient outcome.

Limitation of life sustaining therapy in disorders of consciousness: ethics and practice.

Clinical conversations surrounding the continuation or limitation of life-sustaining therapies (LLST) are both challenging and tragically necessary for patients with disorders of consciousness (DoC) following severe brain injury. Divergent cultural, philosophical and religious perspectives contribute to vast heterogeneity in clinical approaches to LLST-as reflected in regional differences and inter-clinician variability. Here we provide an ethical analysis of factors that inform LLST decisions among patients with DoC. We begin by introducing the clinical and ethical challenge and clarifying the distinction between withdrawing and withholding life-sustaining therapy. We then describe relevant factors that influence LLST decision-making including diagnostic and prognostic uncertainty, perception of pain, defining a 'good' outcome, and the role of clinicians. In concluding sections, we explore global variation in LLST practices as they pertain to patients with DoC and examine the impact of cultural and religious perspectives on approaches to LLST. Understanding and respecting the cultural and religious perspectives of patients and surrogates is essential to protecting patient autonomy and advancing goal-concordant care during critical moments of medical decision-making involving patients with DoC.

Decision-Making for Preventive Interventions in Asymptomatic Patients.

The decision to treat an incidental finding in an asymptomatic patient results from careful risk-benefit consideration and is often challenging. One of the main aspects is after how many years the group who underwent the intervention and faced the immediate treatment complications will gain a treatment benefit over the conservatively managed group, which maintains a lower but ongoing risk. We identify a common error in decision-making. We illustrate how a risk-based approach using the classical break-even point at the Kaplan-Meier curves can be misleading and advocate for using an outcome-based approach, counting the cumulative number of lost quality-adjusted life years instead. In clinical practice, we often add together the yearly risk of the natural course up to the time point where the number equals the risk of the intervention and assume that the patient will benefit from an intervention beyond this point in time. It corresponds to the crossing of the Kaplan-Meier curves. However, because treatment-related poor outcome occurs at the time of the intervention, while the poor outcome in the conservative group occurs over a given time period, the true benefit of retaining more quality-adjusted life years in the interventional group emerges at a much later time. To avoid overtreatment of patients with asymptomatic diseases, decision-making should be outcome-based with counting the cumulative loss of quality-adjusted life years, rather than risk-based, comparing the interventional risk with the ongoing yearly risk of the natural course.

Faster clinical decisions in B-cell acute lymphoblastic leukaemia: A single flow cytometric 12-colour tube improves diagnosis and minimal residual disease follow-up.

Assessing minimal residual disease (MRD) in B-cell precursor acute lymphoblastic leukaemia (BCP-ALL) is essential for adjusting therapeutic strategies and predicting relapse. Quantitative polymerase chain reaction (qPCR) is the gold standard for MRD. Alternatively, flow cytometry is a quicker and cost-effective method that typically uses leukaemia-associated immunophenotype (LAIP) or different-from-normal (DFN) approaches for MRD assessment. This study describes an optimized 12-colour flow cytometry antibody panel designed for BCP-ALL diagnosis and MRD monitoring in a single tube. This method robustly differentiated hematogones and BCP-ALL cells using two specific markers: CD43 and CD81. These and other markers (e.g. CD73, CD66c and CD49f) enhanced the specificity of BCP-ALL cell detection. This innovative approach, based on a dual DFN/LAIP strategy with a principal component analysis method, can be used for all patients and enables MRD analysis even in the absence of a diagnostic sample. The robustness of our method for MRD monitoring was confirmed by the strong correlation (r = 0.87) with the qPCR results. Moreover, it simplifies and accelerates the preanalytical process through the use of a stain/lysis/wash method within a single tube (<2 h). Our flow cytometry-based methodology improves the BCP-ALL diagnosis efficiency and MRD management, offering a complementary method with considerable benefits for clinical laboratories.

The value of virtual molecular tumor boards for informed clinical decision-making.

Genomic profiling and other new technologies have increased the volume and complexity of information available for guiding clinical decision-making in precision oncology. Consequently, there is a need for multidisciplinary expert teams, in the form of molecular tumor boards (MTBs), who can translate this information into a therapeutic plan, including matching patients to suitable clinical trials. Virtual MTBs (vMTBs) can help to overcome many of the challenges associated with in-person MTBs, such as limited time availability, access to appropriate experts or datasets, or interactions between institutions. However, real-world experience from vMTBs is lacking. Here, we describe oncologists' vMTB experiences and the value of working with multicenter and/or multinational vMTBs. We also address knowledge gaps and barriers that could affect the implementation of vMTBs in routine clinical practice. Case studies from Argentina, Turkey, and Portugal illustrate the value of informed clinical decision-making by vMTBs, including expansion of therapeutic options for patients, faster time to treatment, and the resulting improvement in patient outcomes or impact of vMTB discussions on patients. With the uptake of comprehensive genomic profiling and the evolution of some cancers now being conceptualized as a collection of rare diseases with small patient populations based on molecular profiling, the importance of MTBs has increased in modern cancer management. However, an adjustment in clinical decision-making by healthcare professionals is required and evidence of the added value of vMTBs is lacking. Existing vMTBs and recommendations from participating oncologists could point toward a structured evaluation and analysis of this new platform.

Is the CTS5 a helpful decision-making tool in the extended adjuvant therapy setting?

PURPOSE: The Clinical Treatment Score post-5 years (CTS5) is an easy-to-use tool estimating the late distant recurrence (LDR) risk in patients with hormone receptor-positive breast cancer after 5 years of endocrine therapy (ET). Apart from evaluating the prognostic value and calibration accuracy of CTS5, the aim of this study is to clarify if this score is able to identify patients at higher risk for LDR who will benefit from extended ET. METHODS: Prognostic power, calibration, and predictive value of the CTS5 was tested in patients of the prospective ABCSG-06 and -06a trials (n = 1254 and 860 patients, respectively). Time to LDR was analyzed with Cox regression models. RESULTS: Higher rates of LDR in the years five to ten were observed in high- and intermediate-risk patients compared to low-risk patients (HR 4.02, 95%CI 2.26-7.15, p < 0.001 and HR 1.93, 95%CI 1.05-3.56, p = 0.035). An increasing continuous CTS5 was associated with increasing LDR risk (HR 2.23, 95% CI 1.74-2.85, p < 0.001). Miscalibration of CTS5 in high-risk patients could be observed. Although not reaching significance, high-risk patients benefitted the most from prolonged ET with an absolute reduction of the estimated 5-year LDR of - 6.1% (95%CI - 14.4 to 2.3). CONCLUSION: The CTS5 is a reliable prognostic tool that is well calibrated in the lower and intermediate risk groups with a substantial difference of expected versus observed LDR rates in high-risk patients. While a numerical trend in favoring prolonged ET for patients with a higher CTS5 was found, a significantly predictive value for the score could not be confirmed. CLINICAL TRIAL REGISTRATION: ABCSG-06 trial (NCT00309491), ABCSG-06A7 1033AU/0001 (NCT00300508).

Performance of a novel spectroscopy-based tool for adjuvant therapy decision-making in hormone receptor-positive breast cancer: a validation study.

PURPOSE: Digistain Index (DI), measured using an inexpensive mid-infrared spectrometer, reflects the level of aneuploidy in unstained tissue sections and correlates with tumor grade. We investigated whether incorporating DI with other clinicopathological variables could predict outcomes in patients with early breast cancer. METHODS: DI was calculated in 801 patients with hormone receptor-positive, HER2-negative primary breast cancer and ≤ 3 positive lymph nodes. All patients were treated with systemic endocrine therapy and no chemotherapy. Multivariable proportional hazards modeling was used to incorporate DI with clinicopathological variables to generate the Digistain Prognostic Score (DPS). DPS was assessed for prediction of 5- and 10-year outcomes (recurrence, recurrence-free survival [RFS] and overall survival [OS]) using receiver operating characteristics and Cox proportional hazards regression models. Kaplan-Meier analysis evaluated the ability of DPS to stratify risk. RESULTS: DPS was consistently highly accurate and had negative predictive values for all three outcomes, ranging from 0.96 to 0.99 at 5 years and 0.84 to 0.95 at 10 years. DPS demonstrated statistically significant prognostic ability with significant hazard ratios (95% CI) for low- versus high-risk classification for RFS, recurrence and OS (1.80 [CI 1.31-2.48], 1.83 [1.32-2.52] and 1.77 [1.28-2.43], respectively; all P < 0.001). CONCLUSION: DPS showed high accuracy and predictive performance, was able to stratify patients into low or high-risk, and considering its cost and rapidity, has the potential to offer clinical utility.