Medial

Validation of LungFlag™ Prediction Model Using Electronic Medical Records (EMR) on Taiwan Data Presented at the World Conference on Lung Cancer

medialus_admin — Tue, 10 Sep 2024 07:07:00 +0000

Share your details and we'll email you our Publication

Authors:

Eran Choman, Alon Lanyado, and Eitan Israeli, PhD, from Medial EarlySign in Hod Hasharon, Israel. Nicolò Olghi, F. Hoffmann-La Roche Ltd, Basel, Switzerland. Yue Jin, PhD, . F. Hoffmann-La Roche Ltd, Santa Clara, USA. Shang-Yun Liu, PhD, F. Hoffmann-La Roche Taipei, Taiwan. Milan Obradovic, PhD, F. Hoffmann-La Roche Ltd, Basel, Switzerland. Pan-Chyr Yang, MD, PhD, National Taiwan University Hospital and College of Medicine, Taipei, Taiwan.

Introduction:

EarlySign’s efforts at validation of its LungFlag model continue to receive scientific recognition on a global scale. In the category of “Screening and Early Detection – Innovative Screening Technology,” EarlySign was selected to present a new poster at the 2024 World Conference on Lung Cancer in San Diego.

Background:

Entitled “Validation of LungFlag Prediction Model Using Electronic Medical Records (EMR) on Taiwan Data,” the study was built with the knowledge that Lung cancer (LC) is the leading cause of cancer death according to the WHO. To reduce mortality rate, starting 2022, the Ministry of Health and Welfare in Taiwan launched the LC Early Detection Program to provide biennial low-dose computed tomography (LDCT) screening for high-risk groups.

Conclusions:

In conjunction with researchers at Roche and the National Taiwan University Hospital and College of Medicine in Taipei, the conclusions of the study demonstrated that LungFlag was able to work on Taiwanese EMR data and presented superiority over criteria based selection of individuals (USPSTF) and performed better than gold standard benchmark risk model PLCOm2012 that was adapted to be used on EMR data. The results were comparable to the previous validation studies and indicates that LungFlag can immediately support Taiwan screening program in identifying elevated risk populations among ever-smokers and potentially also on never-smokers populations.

Budget impact model of LungFlag™, a predictive risk model for lung cancer screening

medialus_admin — Wed, 29 May 2024 11:52:30 +0000

Share your details and we'll email you our Publication

Authors:

Michael K. Gould,1 Eran Choman,2 Nicolò Olghi,3 Milan Obradovic,3 Sarika Ogale,4 Carolina Heuser Sanmartin3, 1Kaiser Permanente Bernard J. Tyson School of Medicine, Pasadena, CA, USA; 2Medial-EarlySign, Hod Hasharon, Israel; 3F. Hoffmann-La Roche Ltd, Basel, Switzerland; 4Genentech Inc, South San Francisco, CA, USA

Background

Lung cancer screening (LCS) programs using low-dose computed tomography (LDCT) for early detection reduce mortality and have been widely recommended. LungFlag is a machine-learning risk prediction model that uses patient-level data to identify individuals at high risk of developing non-small cell lung cancer (NSCLC), prompting their physician to recommend screening with LDCT.

Methods

A budget impact model was developed to estimate the costs associated with adoption of LungFlag as an adjunct to existing US Preventive Services Task Force (USPSTF) screening guidelines for a hypothetical US commercial health plan population of 1 million beneficiaries. The model calculates the total expected annual costs of screening for NSCLC with LDCT in scenarios with and without LungFlag, including healthcare resource utilization for detecting NSCLC and treatment of patients diagnosed with NSCLC. Incremental costs were evaluated over a 5-year period.

Results

Among 36,803 USPSTF-eligible persons, we assumed that 4600 (12.5%) had already initiated LCS, leaving 32,203 persons who were candidates for pre-screening with LungFlag. The model estimated that 17 additional NSCLC diagnoses per year would be detected by screening when using LungFlag, with most in stage 1. Over 5 years, LungFlag was estimated to result in 33 fewer patients with stage 3 or stage 4 NSCLC at diagnosis and 22 fewer NSCLC-related deaths. Use of LungFlag increased annual costs during the first 2 years and provided cost savings from Year 4 onwards (Table). Cost savings from LungFlag were attributable to reductions in the costs of advanced NSCLC treatment.

Conclusions

In a population of 1 million commercial health plan beneficiaries, the adoption of LungFlag as an adjunct to existing screening guidelines for USPSTF-eligible patients was estimated to prevent 22 additional NSCLC-related deaths, with a cost savings of $2.87 million over 5 years from a US commercial payer perspective.

Development and Validation of a Machine-learning Prediction Model to Improve Abdominal Aortic Aneurysm Screening

medialus_admin — Wed, 17 Jan 2024 14:55:00 +0000

Share your details and we'll email you our Publication

Authors:

Gregory G. Salzler, MD, Evan J. Ryer, MD, Robert W. Abdu, DO, Alon Lanyado, Bsc, Tal Sagiv, Bsc, Eran N. Choman, Msc, Abdul A. Tariq, PhD, Jim Urick, MS, Elliot G. Mitchell, PhD, Rebecca M. Maff, BS, Grant DeLong, BA, Stacey L. Shriner, BS, and James R. Elmore, MD

ABSTRACT

Objective: Despite recommendations by the United States Preventive Services Task Force and the Society for Vascular Surgery, adoption of screening for abdominal aortic aneurysms (AAAs) remains low. One challenge is the low prevalence of AAAs in the unscreened population, and therefore a low detection rate for AAA screenings. We sought to use machine learning to identify factors associated with the presence of AAAs and create a model to identify individuals at highest risk for AAAs, with the aim of increasing the detection rate of AAA screenings.

Methods

A machine-learning model was trained using longitudinal medical records containing lab results, medications, and other data from our institutional database. A retrospective cohort study was performed identifying current or past smoking in patients aged 65 to 75 years and stratifying the patients by sex and smoking status as well as determining which patients had a confirmed diagnosis of AAA. The model was then adjusted to maximize fairness between sexes without significantly reducing precision and validated using six-fold cross validation.

Results

Validation of the algorithm on the single-center institutional data utilized 18,660 selected patients over 2 years and identified 314 AAAs. There were 41 factors identified in the medical record included in the machine-learning algorithm, with several factors never having been previously identified to be associated with AAAs. With an estimated 100 screening ultrasounds completed monthly, detection of AAAs is increased with a lift of 200% using the algorithm as compared with screening based on guidelines. The increased detection of AAAs in the model-selected individuals is statistically significant across all cutoff points.

Conclusions

By utilizing a machine-learning model, we created a novel algorithm to detect patients who are at high risk for AAAs. By selecting individuals at greatest risk for targeted screening, this algorithm resulted in a 200% lift in the detection of AAAs when compared with standard screening guidelines. Using machine learning, we also identified several new factors associated with the presence of AAAs. This automated process has been integrated into our current workflows to improve screening rates and yield of high-risk individuals for AAAs. (J Vasc Surg 2024;79:776-83.)

Use of the Predictive Risk Model, LungFlag™ for Lung Cancer Screening in a Spanish Reference Center: A Cost-effectiveness Analysis.

medialus_admin — Tue, 24 Oct 2023 15:08:00 +0000

Share your details and we'll email you our Publication

Authors:

María Eugenia Olmedo, Luis Gorospe, Luis Miguel Seijo, Virginia Pajares, Mercè Marzo, Joan B. Soriano, Juan Carlos Trujillo, Natalia Arrabal, Andrés Flores, Fran García, María Crespo, David Carcedo, Carolina Heuser, Olghi Nicolò, Eran Choman, Oliver Higuera

BACKGROUND

Several risk prediction models have been developed to select high-risk individuals for lung cancer screening. These allow the calculation of personalized risk as an alternative to standard criteria based on age and cumulative smoking exposure.

LungFlag is an artificial intelligence-based risk prediction model effective in the selection of high-risk individuals by evaluating routine clinical and laboratory.
In Spain, there is no national lung cancer screening program, and only a few pilot programs have been developed.
The aim of this analysis is to assess the cost-effectiveness of LungFlag for the identification of high-risk individuals for enrolment in a NSCLC screening programme in a hypothetical Spanish reference center.

Conclusions

The implementation of LungFlag as a risk model for NSCLC screening in a hypothetical Spanish reference center would be cost-effective compared to no-screening for the 2 hypothetical cohorts analyzed, providing savings and a higher clinical benefit. Narrowing the screening to patients who meet USPSTF criteria seems to optimise the benefits of using LungFlag.

Cost-effectiveness of a machine learning risk prediction model (LungFlag™) in the selection of high-risk individuals for non-small cell lung cancer screening in Spain

medialus_admin — Tue, 24 Oct 2023 15:03:00 +0000

Share your details and we'll email you our Publication

Authors:

Pajares, Soriano JB, Seijo LM, Trujillo JC, Marzo M, Olmedo ME, Higuera O, Arrabal N, Flores A, García JF, Crespo M, Carcedo D, Heuser C, Olghi N, Choman E, Gorospe L

BACKGROUND AND OBJECTIVES

Risk Prediction models (RPM) based on individual risk have proved to effectively identify individuals at high risk of lung cancer for screening in comparison to selection criteria based on age and pack-years alone.
LungFlag is a machine learning-based risk prediction model designed to improve selection of individuals at risk for enrollment in NSCLC screening programs with low dose computed tomography (LDCT), by evaluating routine clinical and laboratory data.
The aim of this analysis is to assess the cost-effectiveness of implementing LungFlag in Spain.

Conclusions

Using LungFlag for the selection of high-risk individuals for NSCLC screening in Spain, would be cost-effective compared to a hypothetical scenario screening all individuals meeting USPSTF criteria. This strategy would be dominant versus the current situation of no-screening.

Improved Efficiency with LungFlag™ vs. Opportunistic Selection in a Theoretical East Asian Lung Cancer Screening Program

medialus_admin — Sat, 09 Sep 2023 15:12:00 +0000

Share your details and we'll email you our Publication

Authors:

Eran Choman, Michael K. Gould, MD, MS, Pan-Chyr Yang, MD, PhD, David Morgenstern, PhD

BACKGROUND

Asia has the highest incidence of lung cancer and lung cancer mortality in the world and is estimated to have the largest increase in lung cancer incidence of all regions by 2040 (more than Europe and NA combined). The incidence of lung cancer among non-smokers in Asia (particularly East Asia) is substantially higher than in Europe and NA, and based on this difference, it would be intriguing to include never smokers in population-based screening algorithms currently under exploration.

Conclusions

The model demonstrated a highly increased odds ratio compared to opportunistic selection in ever smokers (x5) and never smokers (x6) subpopulations, suggesting a single model can be used as a pre-screening tool to identify elevated risk populations. Furthermore, the model showed increased performance in detection of earlier-stage cancers (>50% improvement).

LungFlag™, a Machine-Learning (ML) Personalized Tool for Assessing Pulmonary Complications a Community Setting, Demonstrates Comparable Performance in Flagging Non-Small Cell Lung Cancer.

medialus_admin — Fri, 02 Jun 2023 15:16:00 +0000

Share your details and we'll email you our Publication

Authors:

Eran N. Choman, David Morgenstern, PhD

BACKGROUND

Clinical trials targeting heavy smokers have shown to reduce mortality and paved the way for the USPSTF to recommend lung cancer screening based on a combination of age, smoking history and current or recent smoking status. However, less than 27% of Americans diagnosed with lung cancer meet the original USPSTF criteria for screening, leaving many high-risk individuals unable to access screening.

Using a risk-based selection score resulted in higher sensitivities compared to criteria using dichotomized age and smoking history. Application of ML-based risk models to cancer screening cohorts has shown increase in screening efficiency expressed by increased compliance, resources yield and increased detection of early stages in particular. The classifiers developed for that purpose used training data that are random samples of the screen-eligible population, with same distribution of the key data elements.

Therefore, ML-based models may be vulnerable to sex-and race-based bias arising from historical bias in access to healthcare as well as biased training data. Demonstrating fairness in the predictions of ML-based models is a prerequisite to their acceptance by clinicians, patients and policy-makers. We assessed the clinical performance of LungFlag based on sex and race, two key demographic subgroups at risk for disparate outcomes due to bias.

Conclusions

We determined that in a large, community-based retrospective data set the LungFlag model demonstrated fairness with respect to sex and race, showing similar clinical sensitivity while themPLCOm 2012 model demonstrated statistically significant differences between the sub-populations. Additionally, LungFlag demonstrated statistically significant improvement over mPLCO2012 (41%-78%) for the sex-and race-based sub-populations. Further assessment in prospective studies and in additional racial sub-populations is recommended to support this conclusion.

Computer-assisted Flagging of Never Smokers at High Risk of NSCLC in a Large US-based HMO using the LungFlag Model

medialus_admin — Thu, 27 Oct 2022 06:55:54 +0000

Share your details and we'll email you our Publication

Authors:

Eran N. Choman

Alon Lanyado

Title

Computer-assisted Flagging of Never Smokers at High Risk of NSCLC in a Large US-based HMO using the LungFlag Model

Background & Aims

Smoking is considered to be the major cause of lung cancer, but lung cancer is not just a smokers’ disease. The prevalence of lung cancer in never-smokers is gradually rising with around 20% of lung cancers in the UK and US occurring in people who have never smoked[1][2]. This figure rises to around 50% in some Asian countries[3][4].

Methods

A machine-learning algorithm based on routine EHR and laboratory data, previously developed and validated on an ever-smoker population[5] was used to evaluate the accuracy in detection of Non-Small Cell Lung Cancer (NSCLC) among individuals who never smoked. The cohort was from a large US health system and included 509 case patients with NSCLC and 50,001 contemporaneous NSCLC-free controls. We compared the performance of two risk prediction models, LungFlag and the PLCOm2012 model adapted to EHR data (mPLCOm2012).

Results

Data were analyzed using the area under the receiver operating characteristic curve (AUC), positive predictive value (PPV), and diagnostic odds ratio (OR) as measures of model performance for the age group 40 and above. The risk predictors were calculated for multiple time windows prior to the diagnosis date (Dx) using cut-offs yielding specificities of 90%, 95%, 97% or 99%. (Details in PDF)

Conclusions

By using available information existing in the EHR, the model demonstrated high accuracy (OR>6) in early detection of NSCLC among never-smokers with data going back up to 24 months before diagnosis. Furthermore, LungFlag creates an opportunity to carry out case finding in a population with growing rates of lung cancer that is currently not offered any screening, yet additional local validations are recommended.

Flagging high-risk individuals with an ML model improves NSCLC early detection in a USPSTF-eligible population

medialus_admin — Sun, 25 Sep 2022 06:10:00 +0000

Share your details and we'll email you our Publication

Authors:

Eran N. Choman

Alon Lanyado

Prof. Michael K. Gould

MD

MS

Title

Flagging high-risk individuals with a ML model improves NSCLC early detection in a USPSTF-eligible population

Background & Aims

The USPSTF recommends annual lung cancer screening with LDCT in adults aged 50 to 80 years who have a ≥20 pack- year smoking history and currently smoke or have quit within the past 15 years. Risk prediction models are an alternative approach to identify high-risk individuals for screening that may have advantages compared to age and smoking history- based selection. We compared the performance of two risk prediction models, LungFlag and adapted to EHR data PLCOm2012 (mPLCOm2012).

Methods

Data from a large US health system including 6,505 case patients with non-small cell lung cancer (NSCLC) and 189,597 contemporaneous NSCLC-free controls were used to evaluate the performance of an optimized version of a previously published machine-learning model (LungFlag) to detect NSCLC among individuals who meet the USPSTF criteria compared to the performance of mPLCOm2012. The model used existing routine out-patient lab measurements, smoking history, comorbidities, and demographic data.

Results

Data were analyzed using the area under the receiver operating characteristic curve (AUC), and diagnostic sensitivity on the USPSTF screen-eligible population (Tables 1-3 in PDF) and Ever Smokers ages 50-80 population (Table 4 in PDF). The risk predictor was calculated for a 3-12-month window prior to the diagnosis date (Dx) using cut-offs yielding specificity levels of 97%, 95% or 90%.

Conclusions

By using available information existed in the EHR, the LungFlag model was more accurate for early diagnosis of NSCLC than mPLCOm2012, demonstrating the potential to help prevent lung cancer deaths through early detection among the sub- group of USPSTF as well as the Ever Smokers population.

Prediction of influenza complications: Development and validation of a machine learning prediction model to improve and expand the identification of vaccine-hesitant patients at risk of severe influenza complications

medialus_admin — Tue, 26 Jul 2022 06:23:00 +0000

Share your details and we'll email you our Publication

Authors:

Donna N. Walk

Alon Lanyado

Ann Marie Tice

Maheen Shermohammed

Yaron Kinar

Amir Goren

Christopher F. Chabris

Michelle N. Meyer

Avi Shoshan

Vida Abedi

Title

Prediction of influenza complications: Development and validation of a machine learning prediction model to improve and expand the identification of vaccine-hesitant patients at risk of severe influenza complications.

Background & Aims

Influenza vaccinations are recommended for high-risk individuals, but few population-based strategies exist to identify individual risks.

Methods

Patient-level data from unvaccinated individuals, stratified into retrospective cases (n = 111,022) and controls (n = 2,207,714), informed a machine learning model designed to create an influenza risk score; the model was called the Geisinger Flu-Complications Flag (GFlu-CxFlag). The flag was created and validated on a cohort of 604,389 unique individuals. Risk scores were generated for influenza cases; the complication rate for individuals without influenza was estimated to adjust for unrelated complications. Shapley values were used to examine the model’s correctness and demonstrate its dependence on different features. Bias was assessed for race and sex. Inverse propensity weighting was used in the derivation stage to correct for biases. The GFlu-CxFlag model was compared to the pre-existing Medial EarlySign Flu Algomarker and existing risk guidelines that describe high-risk patients who would benefit from influenza vaccination.

Results

The GFlu-CxFlag outperformed other traditional risk-based models; the area under curve (AUC) was 0.786 [0.783–0.789], compared with 0.694 [0.690–0.698] (p-value < 0.00001). The presence of acute and chronic respiratory diseases, age, and previous emergency department visits contributed most to the GFlu-CxFlag model’s prediction. When higher numerical scores were assigned to more severe complications, the GFlu-CxFlag AUC increased to 0.828 [0.823–0.833], with excellent discrimination in the final model used to perform the risk stratification of the population.

Conclusions

The GFlu-CxFlag can better identify high-risk individuals than existing models based on vaccination guidelines, thus creating a population-based risk stratification for individual risk assessment and deployment in vaccine hesitancy reduction programs in our health system.