Oncology Electronic Health Record Research Articles

Cancer screening and first-time visits to oncology services during and after COVID-19 pandemic in Uruguay.

e23016 Background: On March 13, 2020, Uruguay confirmed the first cases of COVID-19 and the same day the government declared a Public Health Emergency that was extended until April 4, 2022. The measures implemented to mitigate the transmission included, among others, postponing screening studies and in-person consultations. Added to these measures was the fear of going to health centers and the consequences of the economic crisis, with the increasing difficulties in traveling and facing the co-payments necessary to access health services. This study aims to evaluate the impact of the COVID-19 pandemic on cancer screening and first time- visits to oncology services during the public health emergency and following its lifting. Methods: We compare the number of PAP test, mammograms and fecal immunochemical test (FIT) reported by the same comprehensive health providers during 3 periods: pre-pandemic (2018-2019), public health emergency period (March 13, 2020- April 4, 2022) and post-public health emergency (April 5 2022-June 30,2023). Cancer screening data was obtained from the screening registry of the Honorary Commission to Fight Cancer (CHLCC) and first oncology visits from the Oncology Electronic Health Record (HCEO). Health providers that report PAP, mammograms, and FIT cover around 90%, 30%, and 40% of health system users, respectively, and those that use HCEO, 40%. Results: After two years of COVID-19 public health emergency, the number of reported PAP smears test, mammography and fecal occult blood test, decreased by 26%, 46% and 33% respectively (Table 1). The reduction was greater during the first months of the pandemic and during the peak of the first wave (April-June 2021: PAP -50%, mammography -58%, occult blood test: -70%). Likewise, the number of first-time visits to oncology services decreased by 37%, reaching a 56% decline during the first wave. As shown in Table 1, these declines continued after the health emergency was lifted. Conclusions: Our results show an important decline of cancer screening and first-time visits to oncology services during the pandemic in Uruguay. After emergency was lifted, reported PAP test showed a partial recovery but mammography, occult fecal blood test and first-time visits to oncology services remain well below pre-pandemic levels. [Table: see text]

Representativeness of race, ethnicity and social determinants of health in patients with advanced non-small cell lung cancer via linkage of real-world electronic health records, administrative claims, and consumer financial data.

e13559 Background: Social determinants of health (SDOH) are both drivers of health disparities and barriers to clinical research participation, particularly among historically underserved or marginalized populations. Using real-world data (RWD), diversity plans in cancer clinical trials frequently aim to reduce disparities and improve representation. Our aim was to describe the representativeness of patients with advanced non-small cell lung cancer (aNSCLC) with respect to race, ethnicity and person-level SDOH in a curated electronic health record (EHR) database. Methods: This was a retrospective cohort study of individuals diagnosed with aNSCLC between 2016–2022 in a US-based oncology EHR database (ConcertAI). Information on person-level SDOH was derived from linkage of curated EHR data to administrative claims and consumer financial data. These data were also compared to a sample of patients diagnosed with first primary NSCLC from the Surveillance, Epidemiology, and End Results (SEER) Program and American Community Survey (ACS), where possible. The Kaplan-Meier product limit estimator was used to compare time from advanced diagnosis to treatment initiation. Results: From an overall sample of 17,772 patients with aNSCLC, a total of 10,390 patients had 1 or more SDOH characteristics of interest documented. Compared with patients without linked SDOH data, a higher proportion of patients with linked person-level SDOH data were non-Hispanic (NH) White patients (79% vs 71%) and lower proportions were NH Black (10% vs 13%), NH Asian/Pacific Islander (2% vs 4%) and Hispanic patients (2% vs 4%). Patients with linked SDOH were also more likely to have aNSCLC progressed from an earlier stage and received biomarker testing for targetable alterations. Overall, 54% of patients were married, 38% completed a bachelor’s degree or higher, 14% worked in a blue-collar occupation, and 61% had an individual income of ≤$35,000. All 4 of these SDOH characteristics differed significantly by race/ethnicity. Differences in time-to-treatment initiation were statistically significant by marital status (median of 50 [married] vs 55 [not married] days; P=0.02) and individual income (median of 50 [>$60,000] vs 55 [≤$30,000] days; P=0.04). Patients with aNSCLC in these data differed from patients in SEER and ACS estimates with respect to greater representation of patients who are NH White, have lower levels of education, and lower individual income. Conclusions: This analysis demonstrated the feasibility of measuring person-level SDOH using EHR from patients with aNSCLC treated at oncology clinics across the US. RWD with linked SDOH characteristics should be used cautiously to address the representativeness gap in evidence from oncology trials given their potential inherent selection biases.

Elranatamab efficacy in MagnetisMM-3 compared with real-world control arms in triple-class refractory multiple myeloma.

Elranatamab efficacy in the single-arm, registrational MagnetisMM-3 trial (NCT04649359) was compared with that of physician's choice of treatment (PCT) for triple-class refractory multiple myeloma. MagnestisMM-3 eligibility criteria were applied to two USA-based oncology electronic health record databases, COTA and Flatiron Health (FH), to identify cohorts for this study (NCT05932290). Applied statistical techniques accounted for cohort imbalances. MagnetisMM-3 (BCMA-naive; n=123) outcomes were compared with those from COTA (n=239) and FH (n=152). Elranatamab was associated with a significantly higher objective response rate (risk ratios, 1.88-2.25), significantly longer progression-free survival (hazard ratios [HRs], 0.37-0.57), and, across most analyses, significantly longer overall survival (HRs, 0.46-0.66) versus PCT. BCMA-naive patients who were treated with elranatamab exhibited significantly better outcomes than patients treated in real-world clinical practice.

Abstract 4967: Predicting progression-free survival (PFS) in first-line (1L) immune checkpoint inhibitor (ICI)-treated patients (pts) with advanced non-small cell lung cancer (aNSCLC): Machine learning (ML) application in real-world data

Abstract Introduction: Anti-programmed cell death (ligand)-1 ICIs are the standard of care 1L treatment for pts with aNSCLC without actionable oncogenic driver mutations. Heterogeneity in ICI utilization and outcomes is observed in real-world pts. By leveraging ML in electronic health records (EHRs), we identified predictors influencing key clinical outcomes, and developed a nomogram to predict likelihood of median PFS in 1L ICI-treated aNSCLC pts for potential risk stratification for interventions. Methods: 1L ICI-treated aNSCLC pts without ALK/ROS1/KRAS/BRAF/EGFR alterations were identified in a US oncology EHR database (ConcertAI; Jan 2015-Feb 2023). Survival ML models were trained on 112 clinical and demographic features with 5-fold nested cross-validation. The top predictors, determined by the best-performing ML model through SHapley Additive exPlanations and clinical judgement, were used to create a Cox proportional hazard (CPH) nomogram of median PFS. All models were evaluated using the concordance index (c-index). Patients were categorized as having high and low risk of progression/death at median PFS according to median risk predicted by the nomogram. Results: The study cohort had 4668 pts (median PFS: 6.1 months; 3811 events). The CPH nomogram predicting 6-month PFS had a c-index of 0.60 with the top 10 predictors identified using the XGBoost model (c-index: 0.62). Nomogram predictors included increased number of metastases, ECOG PS, cough suppressant use, WBC counts, and NLR (Table). Median probability of 6-month PFS for low- and high-risk groups was 35.9% and 19.4%, respectively. Conclusions: Future research should validate these findings and evaluate the opportunity to guide clinical practice to optimize outcomes in aNSCLC pts treated with 1L ICIs. Summary of top 10 clinical predictors at baseline Mean HR (95% CI) Max nomogram points Total metastases, n 1.15 (1.11-1.19) 100 ECOG PS, ordinal 1.14 (1.10-1.19) 55 Basophils, % 0.81 (0.73-0.90) 52 Albumin, g/dL 0.81 (0.75-0.87) 46 Chloride, mmol/L 0.98 (0.97-0.99) 46 Cough suppressants/expectorants, n 1.08 (1.03-1.14) 42 NLR 1.03 (1.02-1.05) 33 WBC counts, 103 cells/mL 1.02 (1.01-1.03) 29 PD-L1 expression, % 1.00 (1.00-1.00) 28 Presence of liver metastases 1.26 (1.15-1.39) 24 ECOG PS, Eastern Cooperative Oncology Group performance status; NLR, neutrophils-to-lymphocytes ratio; PD-L1, programmed cell death-ligand 1; WBC, white blood cell. Citation Format: Deep K. Hathi, James Harnett, Ning Wu, Zhenxing Xu, Weishen Pan, T G. Hager, Ruben G. Quek, Shivani Aggarwal, Young Kim, Petra Rietschel, Frank Seebach, Fei Wang, Ying Li. Predicting progression-free survival (PFS) in first-line (1L) immune checkpoint inhibitor (ICI)-treated patients (pts) with advanced non-small cell lung cancer (aNSCLC): Machine learning (ML) application in real-world data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4967.

An Indirect Comparison of Elranatamab's Progression-Free Survival and Overall Survival from Magnetismm-3 Versus Real-World External Control Arms in Triple-Class Refractory Multiple Myeloma

BACKGROUND Elranatamab(ELRA) is a BCMAxCD3 bispecific antibody being investigated for the treatment of relapsed/refractory multiple myeloma (MM). The phase 2 MagnetisMM-3 (MM-3; NCT04649359) trial was single-armed; the aim of this study was to contextualize the efficacy data from MM-3 with two real-world (RW) external control arms. METHODS We conducteda retrospective cohort study to indirectly compare the efficacy observed in MM-3 Cohort A (BCMA-naïve; N=123) from the March 2023 data cut (approximately 15 months of follow-up) with two US-based oncology electronic health record databases, COTA and Flatiron Health (FH), as external controls. MM-3 inclusion (eg, prior MM diagnosis, ECOG≤2, refractory to ≥1 PI, ≥1 IMiD, and ≥1 anti-CD38) and exclusion (I/E) criteria (eg, plasma cell leukemia, smoldering MM) were applied to each RW database to obtain comparable patient populations across sources. After imposing MM-3 I/E criteria, we conducted comparisons between data sources on progression-free survival (PFS) and overall survival (OS) using Kaplan-Meier estimators or Cox proportional hazard models (“unweighted analyses”) and restricted mean survival time (RMST) analyses in instances where the proportional hazards assumption was violated. We used additional Cox proportional hazard models, Kaplan-Meier estimators, or semi-parametric models implementing inverse probability of treatment weighting (IPTW), doubly robust, and propensity score (PS) matching analyses to adjust for any remaining imbalances across cohorts on confounding variables (eg, age, comorbidities, Eastern Cooperative Oncology Group (ECOG) score, International Staging System (ISS), prior lines/refractoriness, cytogenetic risk, extramedullary disease [COTA only], lab values). RESULTS N=123 patients from MM-3 Cohort A were compared with the 239 and 152 patients identified from the COTA and FH databases, respectively. Treatment regimens in the RWD sources included various combinations of PIs (carfilzomib- and bortezomib-based regimens were each used by 43% and 15% of patients, respectively), IMiDs (lenalidomide- and pomalidomide-based regimens were used by 9% and 41% of patients, respectively), and mAbs (daratumumab- and isatuximab-based regimens were used by 32% and 1% of patients, respectively), among other agents (eg, selinexor-based regimens were used by 5% of patients). Across unweighted, IPTW, doubly robust, and PS matching analyses, ELRA was associated with significantly longer PFS than RW physicians' choice of treatment from both COTA and FH (hazard ratios [HRs] ranged from 0.37 to 0.57; see Table 1). Similarly, across unweighted, IPTW, doubly robust, and PS matching analyses, ELRA was associated with significantly longer OS than RW physicians' choice from COTA (HRs ranging from 0.46 to 0.65) and longer OS across unweighted (RMST difference at 15 months = 1.70) and PS matching analyses (HR = 0.60) than RW physicians' choice from FH. IPTW and doubly robust methods indicated only directionally longer OS for ELRA than RW physician's choice from FH (RMST difference at 15 months = 0.72 and HR = 0.66, respectively). CONCLUSIONS Among BCMA-naïve patients who resemble those enrolled in the MM-3 trial, those treated with ELRA exhibit significantly longer PFS and OS compared with real-world treatments.

Real-World Evaluation of Treatment Patterns and Clinical Outcomes Among Patients with Chronic Myeloid Leukemia in Chronic Phase (CML-CP) Treated with Asciminib in US Clinical Practice

Introduction: Tyrosine kinase inhibitors (TKIs) are the mainstay of treatment for CML-CP. Asciminib, a first-in-class inhibitor to Specifically Target the ABL Myristoyl Pocket (STAMP), was FDA-approved on 10/29/2021 for the treatment of adult patients with Philadelphia chromosome-positive (Ph+) CML-CP, previously treated with two or more TKIs or Ph+ CML-CP with the T315I mutation. This study described treatment patterns and real-world clinical outcomes of asciminib in the US. Methods: Data through 11/30/2022 from adult patients with Ph+ CML-CP who initiated asciminib were obtained from the Flatiron Health oncology electronic health record (EHR)-derived de-identified database. Patients who initiated asciminib after ≥2 previous TKIs, without a stem-cell transplant prior to asciminib, and without the T315I mutation, were included in the overall cohort. Time-to-treatment discontinuation and molecular response (MR; time-to-BCR::ABL1 ≤0.1% and time-to-BCR::ABL1 ≤1%, separately) were evaluated from asciminib initiation (index date) using Kaplan-Meier analyses overall, and by the number of TKIs used pre-index (2 previous TKIs, and ≥3 previous TKIs). Patients were required to have ≥1 MR test post-index to be included in the MR assessment; a sensitivity analysis was conducted excluding patients with BCR::ABL1 ≤0.1% or better (≤1% or better) for time-to-BCR::ABL1 ≤0.1% (time-to-BCR:ABL1 ≤1%), as evaluated in the ASCEMBL trial. Results: Overall,97 patients with Ph+ CML-CP initiated asciminib (median age: 63 years, 50.5% female, 64.9% White, 84.8% ECOG 0-1, 61.9% from community-based practices, median follow-up of 6 months post-index). Other malignancies (19.6%), chronic pulmonary disease (18.6%), renal disease (18.6%), and congestive heart failure (16.5%) were the most prevalent pre-index comorbidities. The distribution of last MR in the 3 months pre-index was: 24.7% BCR::ABL1 >10%, 18.6% BCR::ABL1 ≤10% and >1%, 12.4% BCR::ABL1 ≤1% and >0.1%, 12.4% BCR::ABL1 ≤0.1% and >0.01%, and 8.2% BCR::ABL1 ≤0.01% (23.7% not tested/documented). Forty-six patients (47.4%) initiated asciminib after 2 previous TKIs, 24 (24.7%) after 3 TKIs, and the remainder after ≥4 previous TKIs (27.8%). Most patients (78.4%) were prescribed 80 mg (40.2% once daily, 38.1% 40 mg twice a day) at treatment initiation. Use of TKIs prior to asciminib included dasatinib (85.6%), imatinib (63.9%), bosutinib (61.9%), nilotinib (53.6%), and ponatinib (15.5%). Treatment sequences among patients who initiated asciminib after 2 previous TKIs (n=46) are presented in Figure 1. Overall (n=97), persistence rates were 85.7% by 12-weeks (2 previous TKIs: 89.8%; ≥3 previous TKIs: 82.5%) and 78.1% by 24-weeks (2 previous TKIs: 81.8%; ≥3 previous TKIs: 75.3%) post-index. Among patients with ≥1 MR test post-index (n=70), 31.3% (2 previous TKIs: 43.3%; ≥3 previous TKIs: 21.6%) and 49.7% (2 previous TKIs: 71.5%; ≥3 previous TKIs: 33.4%) achieved or maintained BCR::ABL1 ≤0.1% by 12- and 24-weeks, respectively. In addition, 51.3% (2 previous TKIs: 68.7%; ≥3 previous TKIs: 37.2%) and 64.2% (2 previous TKIs: 79.7%; ≥3 previous TKIs: 51.4%) achieved or maintained BCR::ABL1 ≤1% by 12- and 24-weeks, respectively. For patients without BCR::ABL1 ≤0.1% prior to asciminib (n=52; Figure 2), 14.6% (2 previous TKIs: 20.6%; ≥3 previous TKIs: 10.4%) and 32.9% (2 previous TKIs: 56.7%; ≥3 previous TKIs: 17.9%) achieved BCR::ABL1 ≤0.1% by 12- and 24-weeks, respectively, with a median time of 30.7 weeks (2 previous TKIs: 23.0 weeks; ≥3 previous TKIs: 38.1 weeks). For patients without BCR::ABL1 ≤1% prior to asciminib (n=43), 35.6% (2 previous TKIs: 61.9%; ≥3 previous TKIs: 21.9%) and 46.3% (2 previous TKIs: 61.9%; ≥3 previous TKIs: 37.2%) achieved BCR::ABL1 ≤1% by 12- and 24-weeks, respectively, with a median time of 24.6 weeks (2 previous TKIs: 11.3 weeks; ≥3 previous TKIs: 26.9 weeks). Conclusion: This is the first real-world study describing treatment patterns and clinical outcomes among patients with Ph+ CML-CP treated with asciminib in the US. The findings of real-world treatment efficacy based on achieving BCR::ABL1 ≤0.1% were consistent with the ASCEMBL trial. Patients who initiated asciminib after 2 previous TKIs had higher persistence and treatment response rates than those initiated on asciminib after ≥3 previous TKIs. Further studies are warranted to continuously demonstrate the effectiveness of asciminib in a real-world setting.

Machine Learning Approach to Understand Real-World Treatment in Patients with Higher-Risk Myelodysplastic Syndromes

Background: Higher-risk myelodysplastic syndromes (HR-MDS) is a severe and aggressive form of MDS that represents a group of rare and heterogeneous hematologic disorders that primarily affect the bone marrow and blood cells. Treatment typically involves a combination of supportive care and active treatment with targeted therapies, hypomethylating agents, chemotherapy, and hematopoietic stem cell transplant. In this study, we have evaluated real-world treatment patterns in patients with HR-MDS with machine-learning (ML) models to identify clinical and demographic features that may impact decisions about which patients receive systemic therapies and when the therapies are implemented. Methods: We identified 821 patients aged ≥18 years with HR-MDS diagnosed between January 2015 and April 2022, using ConcertAI's RWD360 database linked with open claims data. RWD360 consists of structured records from US-based oncology electronic health record systems. Demographic and clinical data up to 1 year prior to the date of HR-MDS diagnosis were used to develop a family of XGBoost-based ML models to investigate drivers of treatment. We created five models: a time-to-therapy model that predicts patient-level hazard ratios and four classification models that predict likelihood of therapy in the first 2, 3, 6, and 12 months, respectively, following HR-MDS diagnosis (therapy yes/no models). Model performance was evaluated using Harrell's concordance index (C-index) and Akaike information criterion (AIC) for the time-to-therapy model and area under curve (AUC) for the therapy yes/no models. Key features included laboratory values, Charlson comorbidity index (CCI), and prescribed medications. To better capture the temporal characteristics of a patient's disease state evolution, clinical features were organized into time windows alongside the variations within them. We used recursive feature elimination to bring the total features used in the model to 50 to enhance the interpretability and compactness. Five-fold cross-validation (data was split 5 ways; for each fold, a different 20% is left out for validation, while 80% of the data was used to train the model) on the full 821-patient cohort was performed to train and evaluate the model. Results: The median age in this patient cohort was 73 years (range = 21-88 years); approximately 60% were male, and 66% patients were treated in community oncology practices. Only 42.5% patients were recorded to ever receive systemic active therapy, and of those, the median time to therapy was 4.2 months (Figure 1). Patients had a median hemoglobin concentration of 9.4 g/dL (range = 4.4-18.9 g/dL), median CCI of 1 (range = 0-10), and median polypharmacy of 10 (computed on patients who had at least 1 concomitant medication (49%); range = 1-93) around time of diagnosis (assessed 1 year prior, and up to 30 days after diagnosis date). Model performance for the time-to-therapy model (C-index = 0.78; AIC = 3314.2) and for the therapy yes/no models (AUC = 0.7-0.74) were evaluated. A key predictive feature across models was the minimum laboratory value of white blood cell (WBC) count near the point of diagnosis; low WBC values around time of diagnosis was consistently a strong predictor for receiving systemic therapy and receiving it earlier. Two other top predictors of systemic therapy and its timing were the rate of fall in measurements of blood glucose and the rate of fall of hemoglobin during the 180 to 30 days before diagnosis. Table 1 details model performance and the top 3 features impacting time and likelihood of therapy for all 5 models. Conclusion: We have built a family of data-driven ML models for patients with HR-MDS, based on data collected in routine clinical practice, towards predicting time-to and probability-of first-line systemic therapy with reasonable accuracy driven by patient characteristics. A majority of patients diagnosed with HR-MDS in this study appeared to receive delayed or no systemic therapy in their treatment journey. To bridge gaps in care and facilitate timely treatment in eligible patients, data-driven predictive modeling that take into account detailed patient-level factors is critical. Further analysis on outcomes of untreated patients linked with their key clinical characteristics identified using these models may provide opportunities to improve clinical care by implementing strategies for treatments based on patient-level factors.

Evaluation of US oncology electronic health record real-world data to reduce uncertainty in health technology appraisals: a retrospective cohort study

ObjectivesExamine whether data from early access to medicines in the USA can be used to inform National Institute for Health and Care Excellence (NICE) health technology assessments (HTA) in oncology.DesignRetrospective...

An indirect comparison of elranatamab’s (ELRA) objective response rate (ORR) from MagnetisMM-3 (MM-3) vs real-world external control arms in triple-class refractory (TCR) multiple myeloma (MM).

6618 Background: ELRA is a BCMAxCD3 bispecific antibody being investigated for the treatment of R/R MM. The phase 2 MM-3 trial was single-armed; the aim of this study (NCT05565391) was to contextualize the efficacy data from MM-3 with two real-world (RW) external control arms. Methods: A retrospective cohort study was conducted to indirectly compare the efficacy observed in MM-3 Cohort A (BCMA-naïve; N=123) from the 9-month data cut with two US-based oncology electronic health record databases, Flatiron Health (FH) and COTA, as external controls. MM-3 inclusion (eg, prior MM diagnosis, ECOG≤2, refractory to ≥1 PI, ≥1 IMiD, and ≥1 anti-CD38) and exclusion (I/E) criteria (eg, plasma cell leukemia, smoldering MM) were applied to each RW database to obtain comparable patient populations across sources. After imposing MM-3 I/E criteria, comparisons between data sources on ORR were conducted by estimating rate ratios (RRs) using log-binomial regression models (unweighted analysis). RRs were also estimated using both inverse probability treatment weighting (IPTW) and augmented IPTW (doubly robust) analyses to account for key covariates (eg, age, comorbidities, ECOG, ISS, prior lines/refractoriness, cytogenetic risk, extramedullary disease, lab values). Results: The 123 patients from MM-3 Cohort A were compared with the 152 and 233 patients identified from the FH and COTA databases, respectively. Treatment regimens in the RWD sources included various combinations of PIs, IMiDs, and mAbs, among other agents (eg, selinexor). Across unweighted, IPTW, and doubly robust analyses, the ORR for ELRA was significantly higher than treatments used for TCR MM patients from RWD sources (Table; all p<.05). Conclusions: Among TCR MM patients who resemble those of the MM-3 trial, ELRA showed improved ORR compared with treatments currently used in clinical practice. [Table: see text]

Abstract 851: Machine learning (ML) model for prediction of pneumonitis using real-world data (RWD) on immune checkpoint inhibitor (ICI)-treated advanced non-small cell lung cancer (aNSCLC)

Abstract Background: Immune-related pneumonitis (IRP) is a rare adverse event (AE) associated with ICIs that may result in treatment interruption or death if not managed properly. IRP prediction is critical to identify at-risk patients and estimate background IRP risk for single-arm trials. ML modeling was used on electronic health records (EHR) to predict IRP risk in patients with aNSCLC using various IRP definitions. Methods: Patients with curated aNSCLC treated with ICIs between Jan 2015 - May 2022 were identified in a US-based oncology EHR database (ConcertAI®, Cambridge, MA). Survival XGBoost models were trained with 4-fold cross-validation. Due to lack of consensus on the IRP definition using RWD, putative IRPs were derived from a mix of human-curated AEs and ICD diagnosis (dx) codes (preexisting [<90 days of ICI start] codes excluded). Three IRP definitions were used: expert-curated (IRP1, n=237 IRP events), literature-derived (IRP2, n=445), and a broad definition (IRP3, n=1213). Risk factors were identified using SHapley Additive exPlanations (SHAP) values and marginal hazard ratios. Results: The study cohort had 8928 patients. Validation cumulative/dynamic area under the ROC curve for IRP1-3 was 0.68, 0.70, and 0.71, respectively. Common across IRP definitions were predictors related to underlying lung conditions, comorbidities, polypharmacy, systemic inflammation, and de novo metastases (Table 1). Chemoradiation was a risk factor for the narrow definition (IRP1), while labs suggestive of organ dysfunction were predictive for broader definitions (IRP2-3). Unsupervised clustering of SHAP resulted in 4 patient clusters. Higher-risk clusters were driven by polypharmacy (IRP1-3), reduced platelets (IRP1), and comorbidities (IRP2-3). Conclusions: ML on different IRP definitions from RWD differentiated risk factors of pneumonitis from those inherent to lung disease among aNSCLC patients. Table 1. Risk factors for the 3 different IRP definitions as identified by ML models IRP1: Mean HR (95% CI) IRP2: Mean HR (95% CI) IRP3: Mean HR (95% CI) Comorbidities: Number of comorbidities ≥4 (vs 0) 1.17 (1.00, 1.49) 1.51 (1.30, 1.77) 1.73 (1.36, 2.22) Comorbidities: BMI ≥29.07 (vs ≤21.95) 1.25 (1.13, 1.45) 1.27 (1.17, 1.38) 1.27 (1.06, 1.53) Polypharmacy: Number of concomitant meds ≥13 (vs ≤5) 1.45 (1.17, 1.84) 1.62 (1.19, 2.11) 1.29 (1.06, 1.69) Polypharmacy: Local anesthesia (vs no) 1.12 (1.0, 1.23) 1.02 (1.00, 1.06) 1.13 (1.04, 1.21) Polypharmacy: Anti-arrhythmia agents (vs no) 1.07 (0.99, 1.14) 1.00 (1.00, 1.03) 1.00 (0.97, 1.02) Polypharmacy: Anticoagulants (vs no) 1.02 (1.00, 1.10) 1.12 (1.06, 1.21) 1.19 (1.03, 1.34) History of chemoradiation: Durvalumab line of therapy (vs no) 1.18 (1.00, 1.41) 1.00 (1.00, 1.02) 0.98 (0.89, 1.00) History of chemoradiation: Baseline thoracic radiation (vs no) 1.08 (1.00, 1.35) 1.01 (1.00, 1.03) 1.00 (0.97, 1.00) History of chemoradiation: Prior chemotherapy (vs no) 1.06 (0.99, 1.19) 1.00 (0.99, 1.01) 0.91 (0.86, 0.98) History of chemoradiation: Platelets count (103/mL) ≥343 (vs ≤203) 0.79 (0.67, 0.90) 0.97 (0.91, 1.06) 0.96 (0.86, 1.07) De novo metastasis diagnosis: M0 stage at initial diagnosis (vs M1) 1.08 (1.00, 1.21) 1.11 (1.00, 1.24) 1.08 (1.02, 1.15) De novo metastasis diagnosis: Lactate dehydrogenase (IU/L) ≥323.5 (vs ≤165) 0.87 (0.73, 0.96) 0.88 (0.75, 1.00) 0.88 (0.73, 1.02) Underlying lung conditions: Oxygen saturation (%) ≥98 (vs ≤94) 0.89 (0.72, 1.0) 1.00 (0.93, 1.05) 0.96 (0.83, 1.05) Underlying lung conditions: Respiratory rate (breaths/min) ≥18 (vs ≤16) 1.09 (0.97, 1.31) 1.08 (1.02, 1.17) 1.12 (0.99, 1.32) Organ dysfunction: Blood urea nitrogen (mg/dL) ≥20 (vs ≤11.7) 1.01 (0.86, 1.19) 1.13 (0.9, 1.58) 1.11 (0.96, 1.46) Organ dysfunction: Albumin (g/dL) ≥4 (vs ≤3.35) 0.97 (0.85, 1.11) 0.95 (0.90, 0.99) 0.83 (0.72, 0.94) Organ dysfunction: Serum protein (g/dL) ≥7.4 (vs ≤6.5) 1.06 (0.95, 1.18) 1.09 (1.02, 1.17) 1.22 (1.02, 1.44) Systemic inflammation: Neutrophils-lymphocytes ratio ≥8.59 (vs ≤3.12) 1.13 (0.99, 1.07) 1.18 (1.04, 1.30) 1.20 (1.00, 1.44) Systemic inflammation: WBC counts (103/mL) ≥10.88 (vs ≤6.10) 1.01 (0.88, 1.14) 1.05 (0.99, 1.15) 1.10 (0.95, 1.35) Footnote: BMI, body mass index; CI, confidence interval; HR, hazard ratio; IRP1 expert-curated definition, n=237; IRP2, literature-derived definition, n=445; IRP3, a broad IRP definition, n=1213; WBC, white blood cell. Citation Format: Deep K. Hathi, Heather Hall, Judith Mueller, Mona H. Cai, Yangyang Liu, Alexander Liede, Rahul K. Das. Machine learning (ML) model for prediction of pneumonitis using real-world data (RWD) on immune checkpoint inhibitor (ICI)-treated advanced non-small cell lung cancer (aNSCLC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 851.

Abstract P1-07-04: Claims-derived comorbidities associations with real-world overall survival (rwOS) among patients with metastatic breast cancer in an electronic health record-claims linked dataset

Abstract Introduction: Comorbid conditions can significantly impact various aspects of cancer care and therefore are often utilized as covariates to control for selection bias in observational studies. Oncology based electronic health record (EHR) data is often used as a source for comorbidity information but may be subject to missingness due to differences in clinician documentation. Utilizing additional data sources, such as health care claims data, may allow for greater capture of comorbidities. This study used a claims data linked to oncology EHR to identify comorbidities and related patient characteristics, as well as the association between a claims-derived comorbidity index and real-world overall survival (rwOS) among outpatient metastatic breast cancer (mBC) patients, to characterize the comorbidities data from this data source. Methods: We selected patients diagnosed with mBC between 2013 and 2021 from a linkage of mBC de-identified EHR-derived Flatiron Health Research Database (FHRD) and Komodo Health claims data. Charlson Comorbidity Index (CCI) were identified based on the presence of a pre-specified International Classification of Diseases (ICD) codes within 12 months preceding cancer diagnosis. To understand comorbidity data within this linkage, patient characteristics were assessed after the study cohort was stratified by the presence of any comorbidity assessed as part of CCI. We calculated the CCI for the study cohort as well as the prevalence of comorbid conditions that are used as part of the CCI. Kaplan-Meier estimates and Cox proportional hazards model were used to estimate median survival time and hazard ratios (HRs) of rwOS among patients with a CCI score of 1 and 2+. Results: The study cohort had 3,213 patients with a median age of 63 [IQR(52, 76)] and 98.7% were female. White patients accounted for 60.1% of the study population while Black and Hispanic/Latino patients accounted for 12.5% and 7.4% of the study cohort, respectively. Patients with comorbidities were older (mean age of 65 vs 61), had a higher (2+) ECOG performance status (PS) score (14.0% vs 9.7%), and were more likely to be Black (16.0% vs 11.0%) or Hispanic/Latino 10.8% vs 6.3%). Twenty percent of the patients had diabetes, 8.2% had peripheral vascular disease, 7.5% had mild liver disease, and 6.3% had congestive heart failure. According to index values, 21.0% of the study cohort had a CCI score of 1, and 18.0% had a score of 2 and above. The median survival times in years for patients with a CCI score of 1 and 2+ were 2.51 [95% CI 2.29,2.81] and 2.05 [95% CI 1.83,2.42], respectively. Based on univariate analysis, patients with a CCI score greater than zero had a higher risk of mortality (CCI 1: HR = 1.25 [95%CI 1.10, 1.41, p< 0.01], CCI 2+: HR = 1.54 [95%CI 1.35, 1.76, p< 0.01]). In the multivariate analysis, adjusted for age, ethnicity/race, and baseline ECOG PS score, the HR for the CCI 1 group was not statistically significant (HR=1.12 [95%CI 0.99, 1.27, p=0.08]). However, the risk of mortality was statistically significant for the CCI 2+ group (HR=1.31 [95%CI 1.14, 1.50], p < 0.01) in the adjusted analysis. Conclusion: Claims data linked to EHR can be used to identify comorbidities and describe patient characteristics. As anticipated, generally, the presence of more comorbid conditions was associated with worse rwOS for patients with mBC. This finding supports use of this linked dataset for similar future studies. Citation Format: Mustafa S. Ascha, Alemseged A. Asfaw, Prakirthi Yerram, Samantha N. Reiss, Sarah D. Brake, Niquelle B. Wadé. Claims-derived comorbidities associations with real-world overall survival (rwOS) among patients with metastatic breast cancer in an electronic health record-claims linked dataset [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P1-07-04.

Use of FOLFOXIRI plus bevacizumab in metastatic colorectal cancer (mCRC): An age-stratified analysis.

64 Background: As the incidence of early-onset mCRC continues to rise, evidence is emerging for the use of more aggressive treatment in the front-line setting in patients aged < 50 years (yrs). Although the superior efficacy of the multiagent chemotherapy FOLFOXIRI+bevacizumab (triplet+bev) over doublet regimens (FOLFOX or FOLFIRI+bev) is well documented, it is commonly associated with increased toxicity. Contemporary data are needed on triplet+bev use by age group in the US, which may help tailor treatment in patients with early-onset mCRC. Methods: This was a retrospective, observational cohort study using Optum Oncology Electronic Health Records data between 2010 and 2021. Use of triplet+bev and subsequent treatments were reported by age group (18–49, 50–64, ≥65 yrs) among newly diagnosed patients with mCRC. Baseline demographic and clinical characteristics data for patients receiving triplet+bev and regorafenib as a subsequent treatment were collected. Results: Among 36,053 eligible patients with mCRC diagnosed between 2010 and 2021, 13.8% were aged 18–49 yrs, 36.4% 50–64 yrs, and 49.8% ≥65 yrs at the time of diagnosis. The proportion of patients receiving triplet+bev was higher in age groups 18–49 yrs (4.0%) and 50–64 yrs (3.4%) vs ≥65 yrs (1.5%). Between 2010 and 2020, the proportion of patients receiving triplet+bev changed from 1.5% to 4.2% in age group 18–49 yrs, from 3.1% to 2.6% in 50–64 yrs, and from 1.1% to 1.2% in ≥65 yrs. Over a median 20-month follow-up period, 58.0% of patients in age groups 18–49 and 50–64 yrs, and 53.7% of patients in age group ≥65 yrs, received a new treatment after triplet+bev. In the overall cohort, the most common new treatment after triplet+bev was regorafenib (20.7%), received by 18.2% of patients in age group 18–49 yrs, 22.7% in 50–64 yrs, and 19.2% in ≥65 yrs. Other new subsequent treatments in the overall cohort included trifluridine/tipiracil (17.1%) and cetuximab (16.3%), with a similar trend seen among all age groups. Across age groups, triplet+bev recipients were more likely to be male, obese, resident in the Midwest or Northeast, and have distant metastases, and less likely to have colorectal surgical resection prior to mCRC diagnosis. Patients in age group 18–49 yrs subsequently receiving regorafenib were more likely to be obese and less likely to have Charlson Comorbidity Index ≥3, distant metastases, and colorectal surgical resection before mCRC diagnosis vs those who did not receive regorafenib. Conclusions: Our study shows that, recently, triplet+bev use has increased in patients with mCRC, particularly in those aged < 50 yrs. Chemotherapy-free treatments may be an option for these patients beyond the first-line setting. While regorafenib is typically used in third or later lines, these real-world data suggest it is often a preferred option following triplet+bev treatment.

Re-envisioning the Paradigm for Oncology Electronic Health Record Documentation by Paying for What Matters for Patients, Quality, and Research

This Viewpoint discusses re-envisioning and incentivizing a unique approach to oncology electronic health record documentation.

RWD49 Bias Characterization of Real-World Patients With and Without Imaging in a Community Oncology Electronic Health Records Database

RWD49 Bias Characterization of Real-World Patients With and Without Imaging in a Community Oncology Electronic Health Records Database

Oncologist perspectives on tumor boards: Virtual meetings and EHR integration.

46 Background: Tumor boards (TBs) are considered a key component for ensuring high quality cancer care. However, little is known about TB operational aspects that impact effectiveness. We aimed to understand how TBs leverage Electronic Health Record (EHR) functionalities and how pandemic-driven transitions from in-person TBs to virtual TBs impacted TB quality and effectiveness. Methods: The National Comprehensive Cancer Network (NCCN) Oncology EHR Advisory Group formed a Workgroup to assess the current state of TBs, obtain providers’ views on ideal TB state, and develop recommendations for improving TB quality and effectiveness. The Workgroup surveyed providers in various specialties at 31 NCCN Member Institutions in March 2022. Results: 846 providers from 27 institutions responded to the survey (n = 76 respondents were excluded from analysis because they reported not participating in TBs). Almost all (n = 764/767, 99%) indicated that one or more participants always or frequently attended TB meetings virtually. When comparing (newly established) virtual to (pre-pandemic) in-person TBs, 69% indicated increased attendance at virtual TBs. 65% reported no change in case volume, and 56% reported unchanged discussion quality while 19% reported increased discussion quality when comparing virtual to in-person TB meetings. In regards to EHR integration, several gaps between current and ideal state were noted. For instance, 40% always or frequently added patients to the TB roster via the EHR, but 57% reported this functionality to be very important or important for TBs. Additionally, 53% always or frequently document TB recommendations within the EHR, while 87% indicated EHR documentation of recommendations as very important or important. Respondents reported that barriers to documentation include resource constraints, concerns with patient access to documentation, legal and liability concerns, and poor EHR functionality. Conclusions: Overall, the majority of responding providers reported virtual TBs to be at least similar in quality to in-person TB meetings. A major gap identified by this study is the lack of documentation of TB recommendations in the EHR. Workflows and EHR functionalities could be improved to fully optimize the EHR for TB management and documentation.[Table: see text]

Standardization of toxicity related dose modification in oncology electronic health record (EHR).

e18616 Background: Inappropriate antineoplastic drug dosing can cause excessive toxicities or impaired efficacy. Chemotherapy dose delays and reduced dose intensity are common in community oncology practices[1]. Electronic tools can improve quality and safety of chemotherapy. Methods: A quality improvement project to standardize toxicity related dose modifications was initiated in January 2021. A dose modification section was created in BEACON (Epic Systems, Verona, WI) chemotherapy protocols. Dose modifications used in primary source studies were added to the protocols (Table, illustrative). An embedded link to relevant section of the study protocol allowed detailed review. Similarly, a link to the relevant CTCAE grading was embedded for reference. A user experience survey was conducted in August 2021 to evaluate adoption and perceived utility. Results: Dose modifications recommendations have been introduced in 104 protocols to date. Of the 70 survey respondents, there were 18 physicians and 5 advance practice providers, remainder being oncology nurses and pharmacists. Pre-implementation, 29% (8/27) of respondents whose role required performing dose modifications indicated using personal clinical judgement for hematological toxicity related dose modifications. 34% (9/26) did the same for non-hematological toxicities. Post-implementation, 87% respondents indicated awareness of the recommendations. 47% indicated that recommendations saved timed and 60% found these useful. 7.2% felt recommendations were impractical for clinical use. Conclusions: Embedding evidence-based dose modification recommendations into EHR based treatment protocols is feasible with widespread user adoption. References Denduluri, N., et al., Dose Delays, Dose Reductions, and Relative Dose Intensity in Patients With Cancer Who Received Adjuvant or Neoadjuvant Chemotherapy in Community Oncology Practices. J Natl Compr Canc Netw, 2015. 13(11): p. 1383-93. Illustrative Dose Modification Table. Reference Links 1. Grothey et al. Dose Modifications 2. CTCAE Grading. Refer to Reference Link for Detailed Instructions. Decrease Oxaliplatin and 5 FU one dose level for ANC<1000, platelets <50,000, Grade 4 diarrhea Decrease 5 FU one dose level for Grade 3 diarrhea, Grade 3 or 4 mucositis Decrease Oxaliplatin one dose level for Grade 2 neurotoxicity persisting between treatments, Grade 3 neurotoxicity resolving to </= Grade 2 between treatments Discontinue Oxaliplatin for Grade 3 neurotoxicity persisting between treatments, Grade 4 neurotoxicity.[Table: see text]

Electronic capture of cancer-related distress in a community oncology program.

1577 Background: Cancer related distress can be seen in as many as one in two patients. Many organizations such as the Commission on Cancer require distress screening. National Comprehensive Cancer Network (NCCN) distress thermometer is a common tool used for screening. We studied feasibility of embedding the NCCN thermometer into oncology electronic health record (EHR) for routine patient care. We concurrently studied feasibility of using a mobile health tool for serial evaluation of cancer related distress. Methods: A flowsheet containing NCCN distress thermometer questions was created in Epic EHR (Epic Systems, Verona, WI). Oncology nurses used the flowsheet for routine patient assessment. Patients with distress level ≥ 4 answered additional questions. Ancillary care providers such as palliative care nurses or social workers addressed the identified needs by providing appropriate ‘services’. A field to capture these interventions was created. We also adapted our previously reported web based mobile tool [1] for monitoring cancer distress. Patients rated their distress on a 1-10 scale and highlighted the distress domain (physical, emotional, practical, family). Distress level ≥ 4 generated a color-coded flag for provider review and intervention. Data from both electronic tools was periodically analyzed to inform patient care and quality improvement. Results: Between January and December 2021, Epic EHR based distress flowsheet collected 28,594 distinct responses in 911 patients. 57.4%, 14.9%, 14.1%, 9.4% and 2.3% of the responses were in the physical, emotional, practical, familial and spiritual domains respectively. ‘Other’ responses were 1.4%. Cumulative frequency of non-physical problems was 42.5%. 1819 ‘services’ were provided with 357 emotional, 351 work, 351 housing, 350 transportation and 350 financial need-based interventions. Of 1231 patients who used the distress scale and provided additional comments, 315 (25.5%) had distress levels ≥ 4. The mobile tool captured 849 unique patient responses to the distress question between April 2020 and February 2022. Distress level ≥ 4 was flagged by 281 unique patients (33.09%). Average distress level was 2.7. Emotional domain problems generated the greatest distress level followed by family, physical and practical problems in decreasing order. Conclusions: We demonstrate feasibility of electronic capture of cancer related distress to facilitate holistic patient care in a community-based oncology program. EHR based and mobile tool distress evaluations generated concordant results. Distress caused by emotional, practical and familial domain problems was nearly as frequent and often more severe than distress caused by physical problems, underscoring the need for comprehensive cancer care.

A Scalable Quality Assurance Process for Curating Oncology Electronic Health Records: The Project GENIE Biopharma Collaborative Approach.

PURPOSEThe American Association for Cancer Research Project Genomics Evidence Neoplasia Information Exchange Biopharma Collaborative is a multi-institution effort to build a pan-cancer repository of genomic and clinical data curated from the electronic health record. For the research community to be confident that data extracted from electronic health record text are reliable, transparency of the approach used to ensure data quality is essential.MATERIALS AND METHODSFour institutions participating in AACR's Project GENIE created an observational cohort of patients with cancer for whom tumor molecular profiling data, therapeutic exposures, and treatment outcomes are available and will be shared publicly with the research community. A comprehensive approach to quality assurance included assessments of (1) feasibility of the curation model through pressure test cases; (2) accuracy through programmatic queries and comparison with source data; and (3) reproducibility via double curation and code review.RESULTSAssessments of feasibility resulted in critical modifications to the curation directives. Queries and comparison with source data identified errors that were rectified via data correction and curator retraining. Assessment of intercurator reliability indicated a reliable curation model.CONCLUSIONThe transparent quality assurance processes for the GENIE BPC data ensure that the data can be used for analyses that support clinical decision making and advances in precision oncology.

Development of a generalizable natural language processing pipeline to extract physician-reported pain from clinical reports: Generated using publicly-available datasets and tested on institutional clinical reports for cancer patients with bone metastases

Objective:The majority of cancer patients suffer from severe pain at the advanced stage of their illness. In most cases, cancer pain is underestimated by clinical staff and is not properly managed until it reaches a critical stage. Therefore, detecting and addressing cancer pain early can potentially improve the quality of life of cancer patients.The objective of this research project was to develop a generalizable Natural Language Processing (NLP) pipeline to find and classify physician-reported pain in the radiation oncology consultation notes of cancer patients with bone metastases. Materials and Methods:The texts of 1249 publicly-available hospital discharge notes in the i2b2 database were used as a training and validation set. The MetaMap and NegEx algorithms were implemented for medical terms extraction. Sets of NLP rules were developed to score pain terms in each note. By averaging pain scores, each note was assigned to one of the three verbally-declared pain (VDP) labels, including no pain, pain, and no mention of pain. Without further training, the generalizability of our pipeline in scoring individual pain terms was tested independently using 30 hospital discharge notes from the MIMIC-III database and 30 consultation notes of cancer patients with bone metastasis from our institution’s radiation oncology electronic health record. Finally, 150 notes from our institution were used to assess the pipeline’s performance at assigning VDP. Results:Our NLP pipeline successfully detected and quantified pain in the i2b2 summary notes with 93% overall precision and 92% overall recall. Testing on the MIMIC-III database achieved precision and recall of 91% and 86% respectively. The pipeline successfully detected pain with 89% precision and 82% recall on our institutional radiation oncology corpus. Finally, our pipeline assigned a VDP to each note in our institutional corpus with 84% and 82% precision and recall, respectively. Conclusion:Our NLP pipeline enables the detection and classification of physician-reported pain in our radiation oncology corpus. This portable and ready-to-use pipeline can be used to automatically extract and classify physician-reported pain from clinical notes where the pain is not otherwise documented through structured data entry.

Impact of the COVID-19 pandemic on health care activities at a Uruguayan mastology teaching unit: Preliminary results.

e12502 Background: After the government declared a health emergency due to COVID-19 on March, 2020, the Mastology Teaching Unit (UDAM) providing care to breast cancer patients tried to ensure adequate oncological care, and to protect patients from the virus infection and serious complications due to a possible state of immunosuppression. Towards this objective, the Department of Clinical Oncology developed guidelines with some treatment modifications. Objective: To assess the health care activities of the UDAM during the period considered as the “peak” of the pandemic, since its beginning to June 30, 2020. Methods: This is an observational study that collected data from the electronic clinical record system called Oncology Electronic Health Record (HCEO) during the aforementioned period. Results: There werea total of 293 medical appointments (221 in person), through which 131 patients were attended to. The number of medical appointments decreased by 16.7% compared to the same period in 2019 (352 appointments). Of the patients who were attended to during the “peak” period, 109 (83.2%) were on systemic onco-specific treatment and 22 (16.8%) were attended to for disease control. The medical appointments were scheduled to evaluate the continuity of treatment and treatment modifications if necessary (95 patients; 72.5%), for disease control (17; 12.9%), for first-time consultation (12; 9.1%) and to assess paraclinical studies (7; 5.3%). The patients were on hormone therapy (HT) (81 patients; 74%), chemotherapy (CT) (21; 19%), and anti-HER2 therapies (9; 8%). A total of 20 treatments were initiated, 14 with HT and 6 with CT. Of the 21 patients on CT, 14 (66.6%) were on adjuvant/neoadjuvant therapy. Of these, 9 (64.3%) continued with the same regimen with the addition of prophylactic granulocyte colony stimulating factors (G-CSF), and 5 (35.7%), who were receiving weekly paclitaxel, continued the treatment with no changes. The remaining 7 of 21 patients (33.3%) were on palliative CT. Of these, 2 (28.5%) continued the treatment with the addition of G-CSF, 3 (42.8%) continued with weekly capecitabine or paclitaxel with no treatment changes, and 2 (28.5%) changed their treatment regimen (a less myelosuppressive regimen was selected for one of them and this decision was due to the progression of the disease in the other patient). The 90 patients who were receiving adjuvant, neoadjuvant, or palliative criteria HT and/or anti-HER2 therapies, continued the treatment with no changes Conclusions: Although these are preliminary results, the available evidence suggests that, although medical appointments decreased by approximately 17%, the UDAM was able to maintain its healthcare activities and continued most of the treatments. The most modified treatment was CT, with the addition of prophylactic G-CSF, to avoid myelosuppression and potential complications from COVID-19 infection.