Development of predictive models to identify advanced-stage cancer patients in a US healthcare claims database

Productivity loss by cancer stage in patients newly diagnosed with hepatocellular carcinoma: A claims database analysis

BACKGROUND

New cancer diagnoses are associated with employment decrease, workplace absenteeism, and attributable costs to employers.

OBJECTIVE

To estimate the workplace productivity loss in the year following a new diagnosis of early-, intermediate-, or advanced-stage hepatocellular carcinoma (HCC) in commercially insured US adults.

METHODS

We conducted a retrospective cohort study using Merative MarketScan commercial claims to identify incident HCC diagnoses from 2010 to 2020. Patients were stratified into early-, intermediate-, or advanced-stage cohorts based on presence of secondary malignancy codes or first treatment received. Mean workdays lost and attributable cost in the year following a new diagnosis were calculated using the Kaplan-Meier sample averages to account for censoring. An exploratory analysis was conducted on subgroups in the early and advanced cohorts to assess productivity loss in patients with and without treatment.

RESULTS

Mean workdays lost in the year following a new HCC diagnosis among the early, intermediate, and advanced cohorts was 22.6 days (95% CI = 16.0-29.8), 17.4 days (95% CI = 11.9-23.2), and 19.5 days (95% CI = 15.6-23.6), respectively. Corresponding indirect costs were $6,031(95% CI = $4,270-$7,953), $4,644 (95% CI = $3,176-$6,192), and $5,204 (95% CI = $4,163-$6,298). Early-stage patients without a liver transplant and advanced-stage patients who received systemic therapy had 19.7 (95% CI = 12.7-27.4) and 22.0 (95% CI = 16.6-27.7) mean workdays lost, respectively.

CONCLUSIONS

Productivity loss varies by stage and appears to be higher in early-stage patients who receive more intensive treatments in the first year following a new HCC diagnosis.

Machine learning models for identifying predictors of clinical outcomes with first-line immune checkpoint inhibitor therapy in advanced non-small cell lung cancer

Immune checkpoint inhibitors (ICIs) are standard-of-care as first-line (1L) therapy for advanced non-small cell lung cancer (aNSCLC) without actionable oncogenic driver mutations. While clinical trials demonstrated benefits of ICIs over chemotherapy, variation in outcomes across patients has been observed and trial populations may not be representative of clinical practice. Predictive models can help understand heterogeneity of treatment effects, identify predictors of meaningful clinical outcomes, and may inform treatment decisions. We applied machine learning (ML)-based survival models to a real-world cohort of patients with aNSCLC who received 1L ICI therapy extracted from a US-based electronic health record database. Model performance was evaluated using metrics including concordance index (c-index), and we used explainability techniques to identify significant predictors of overall survival (OS) and progression-free survival (PFS). The ML model achieved c-indices of 0.672 and 0.612 for OS and PFS, respectively, and Kaplan-Meier survival curves showed significant differences between low- and high-risk groups for OS and PFS (both log-rank test p < 0.0001). Identified predictors were mostly consistent with the published literature and/or clinical expectations and largely overlapped for OS and PFS; Eastern Cooperative Oncology Group performance status, programmed cell death-ligand 1 expression levels, and serum albumin were among the top 5 predictors for both outcomes. Prospective and independent data set evaluation is required to confirm these results.

Incidence of safety events after immune checkpoint inhibitor initiation for advanced-stage non-small-cell lung cancer: a real-world study

Aim: To describe the incidence of safety events after immune checkpoint inhibitor (ICI) initiation for advanced-stage non-small-cell lung cancer. Methods: Retrospective cohort study using the HealthCore Integrated Research Database in the USA to examine the incidence of prespecified safety events of interest after ICI initiation (n = 5278). Results: The most common safety events after ICI initiation included malaise/fatigue (incidence rate [IR]: 70.7 per 100 person-years; 95% CI: 66.5-75.1) and nausea/vomiting (IR: 32.4; 30.0-34.8). Other potential immune-mediated events, including colitis (IR: 7.11; 6.26-8.04) and pneumonitis (IR: 5.47; 4.76-6.25), were less frequent but higher than after any systemic anti-cancer therapy. No safety event rate substantially increased 6 months after ICI initiation. Conclusion: This large real-world study reports the incidence of safety events with ICI regimens for advanced-stage non-small-cell lung cancer.

Uncertainty in lung cancer stage for survival estimation via set-valued classification

The difficulty in identifying cancer stage in health care claims data has limited oncology quality of care and health outcomes research. We fit prediction algorithms for classifying lung cancer stage into three classes (stages I/II, stage III, and stage IV) using claims data, and then demonstrate a method for incorporating the classification uncertainty in survival estimation. Leveraging set‐valued classification and split conformal inference, we show how a fixed algorithm developed in one cohort of data may be deployed in another, while rigorously accounting for uncertainty from the initial classification step. We demonstrate this process using SEER cancer registry data linked with Medicare claims data.

Problem-solving skills training in adult cancer survivors: Bright IDEAS-AC pilot study

PURPOSE

Cancer patients experience significant distress and burden of decision-making throughout treatment and beyond. These stressors can interfere with their ability to make reasoned and timely decisions about their care and lead to low physical and social functioning and poor survival. This pilot study examined the impact of offering Problem-Solving Skills Training (PSST) to adult cancer survivors to help them and their caregivers cope more successfully with post-treatment decision-making burden and distress.

PATIENTS AND METHODS

Fifty patients who completed their definitive treatment for colorectal, breast or prostate cancer within the last 6 months and reported distress (level > 2 on the National Comprehensive Cancer Network distress thermometer) were randomly assigned to either care as usual (CAU) or 8 weekly PSST sessions. Patients were invited to include a supportive other (n = 17). Patient and caregiver assessments at baseline (T1), end of intervention or 3 months (T2), and at 6 months (T3) focused on problem-solving skills, anxiety/depression, quality of life and healthcare utilization. We compared outcomes by study arm and interviewed participants about PSST burden and skill maintenance.

RESULTS

Trial participation rate was 60%; 76% of the participants successfully completed PSST training. PSST patients reported reduction in anxiety/depression, improvement in QoL (p < 0.05) and lower use of hospital and emergency department services compared to CAU patients (p = 0.04).

CONCLUSIONS

The evidence from this pilot study indicates that a remotely delivered PSST is a feasible and potentially effective strategy to improve mood and self-management in cancer survivors in community oncology settings.

Health care resource utilization and costs associated with advanced or metastatic nonsmall cell lung cancer in the United States

BACKGROUND: The treatment landscape for advanced nonsmall cell lung cancer (NSCLC) has evolved from 2015 onward, since the introduction of immune checkpoint inhibitors (ICIs). Considering this shift, there have been limited prior analyses that assess the economic burden of NSCLC within the current treatment landscape. OBJECTIVE: To present an analysis of health care resource utilization (HCRU) and costs associated with the treatment of patients with advanced or metastatic NSCLC in the United States between 2010 and 2019. METHODS: Patients with locally advanced or metastatic NSCLC who initiated first-line (1L) systemic treatment between January 1, 2010, and June 30, 2019, were included from the HealthCore Integrated Research Database using a previously developed claims-based predictive model algorithm. Mean total HCRU and costs and mean per-person-per-year (PPPY) HCRU and costs were estimated for 2 follow-up periods: the time during the entire follow-up period and the time during the 1L treatment period. Distribution of treatment classes (defined as chemotherapy, ICIs, targeted therapies, and others) were also analyzed by index year. RESULTS: 27,257 patients met the eligibility criteria and were included in the analysis. The mean duration of follow-up for all patients was 16.6 months (median 10.6 months), and the median time to discontinuation of 1L treatment was 2.8 months. The number of outpatient visits accounted for the majority of HCRU across the entire study follow-up (mean 97.7 in total and 147.1 PPPY) and for the 1L treatment period (mean 46.3 in total and 167.5 PPPY). The total mean cost across the entire study follow-up was $158,908 ($250,942 PPPY). For the 1L treatment period, the total mean cost was $72,760 ($271,590 PPPY). Total mean outpatient costs for systemic anticancer treatment were $61,797 for the entire study follow-up ($85,609 PPPY) and $27,138 during the 1L treatment period ($92,412 PPPY). Total costs increased over the study duration, which were mainly due to increasing outpatient costs for systemic therapy. In both follow-up periods, inpatient costs, other outpatient costs (nonsystemic therapy-related costs), and pharmacy costs remained relatively stable but still accounted for more than 60% of the total costs. Analysis of treatment classes over time showed that chemotherapy was the most frequently used treatment, regardless of line of therapy. A trend was observed for increased ICI use from 2015 onward. CONCLUSIONS: Despite the improvement in treatment options, a high economic burden associated with the treatment of NSCLC still exists. The total costs have been increasing, mainly driven by outpatient costs for systemic therapy, which might reflect the greater use of ICIs for advanced NSCLC. Costs for inpatient services, other outpatient services, and pharmacy services remained stable but still accounted for the majority of the economic burden. Further studies are required to assess the impact of innovative treatments on the disease management costs of advanced NSCLC. DISCLOSURES: This study was funded by the healthcare business of Merck KGaA, Darmstadt, Germany (CrossRef Funder ID: 10.13039/100009945) as part of an alliance between the healthcare business of Merck KGaA, Darmstadt, Germany, and Pfizer. Zhang, Liu, and Yang are employees of EMD Serono. Beachler, Dinh, and Jamal-Allial are employees of HealthCore Inc., which received funding from the healthcare business of Merck KGaA, Darmstadt, Germany, and Pfizer for the implementation of this study. Masters and Kolitsopoulos are employees of Pfizer. Lamy was an employee of the healthcare business of Merck KGaA, Darmstadt, Germany, at the time this study was conducted.

Development and validation of a predictive model algorithm to identify anaphylaxis in adults with type 2 diabetes in U.S. administrative claims data

PURPOSE

To use medical record adjudication and predictive modeling methods to develop and validate an algorithm to identify anaphylaxis among adults with type 2 diabetes (T2D) in administrative claims.

METHODS

A conventional screening algorithm that prioritized sensitivity to identify potential anaphylaxis cases was developed and consisted of diagnosis codes for anaphylaxis or relevant signs and symptoms. This algorithm was applied to adults with T2D in the HealthCore Integrated Research Database (HIRD) from 2016 to 2018. Clinical experts adjudicated anaphylaxis case status from redacted medical records. We used confirmed case status as an outcome for predictive models developed using lasso regression with 10-fold cross-validation to identify predictors and estimate the probability of confirmed anaphylaxis.

RESULTS

Clinical adjudicators reviewed medical records with sufficient information from 272 adults identified by the anaphylaxis screening algorithm, which had an estimated Positive Predictive Value (PPV) of 65% (95% confidence interval [CI]: 60%-71%). The predictive model algorithm had a c-statistic of 0.95. The model's probability threshold of 0.60 excluded 89% (84/94) of false positives identified by the screening algorithm, with a PPV of 94% (95% CI: 91%-98%). The model excluded very few true positives (15 of 178), and identified 92% (95% CI: 87%-96%) of the cases selected by the screening algorithm.

CONCLUSIONS

Predictive modeling techniques yielded an accurate algorithm with high PPV and sensitivity for identifying anaphylaxis in administrative claims. This algorithm could be considered in future safety studies using similar claims data to reduce potential outcome misclassification.

Novel prospective umbrella-type lung cancer registry study for clarifying clinical practice patterns: CS-Lung-003 study protocol

INTRODUCTION

Conventional cancer registries are suitable for simple surveillance of cancer patients, including disease frequency and distribution, demographics, and prognosis; however, the collected data are inadequate to clarify comprehensively diverse clinical questions in daily practice.

METHODS

We constructed an umbrella-type lung cancer patient registry (CS-Lung-003) integrating multiple related prospective observational studies (linked studies) that reflect clinical questions about lung cancer treatment. The primary endpoint of this registry is to clarify daily clinical practice patterns in lung cancer treatment; a key inclusion criterion is pathologically diagnosed lung cancer. Under this registry, indispensable clinical items are detected in advance across all active linked studies and gathered prospectively and systematically to avoid excessive or insufficient data collection. Researchers are to input information mutually, irrespective of the relevance to each researcher's own study. Linked studies under the umbrella of the CS-Lung-003 registry will be updated annually with newly raised clinical questions; some linked studies will be newly created, while others will be deleted after the completion of the analysis. Enrollment began in July 2017.

DISCUSSION

We successfully launched the umbrella-type CS-Lung-003 registry. Under this single registry, researchers collaborate on patient registration and data provision for their own and other studies. Thus, the registry will produce results for multiple domains of study, providing answers to questions about lung cancer treatment raised by other researchers. Through such analysis of each linked study, this registry will contribute to the comprehensive elucidation of actual daily practice patterns in lung cancer treatment.

KEY POINTS

CS-Lung-003 registry directly integrates multiple linked studies created under the umbrella of this cancer registry to solve various clinical questions regarding daily practice patterns of lung cancer treatment.

A Systematic Review of Estimating Breast Cancer Recurrence at the Population Level With Administrative Data

BACKGROUND

Exact numbers of breast cancer recurrences are currently unknown at the population level, because they are challenging to actively collect. Previously, real-world data such as administrative claims have been used within expert- or data-driven (machine learning) algorithms for estimating cancer recurrence. We present the first systematic review and meta-analysis, to our knowledge, of publications estimating breast cancer recurrence at the population level using algorithms based on administrative data.

METHODS

The systematic literature search followed Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. We evaluated and compared sensitivity, specificity, positive predictive value, negative predictive value, and overall accuracy of algorithms. A random-effects meta-analysis was performed using a generalized linear mixed model to obtain a pooled estimate of accuracy.

RESULTS

Seventeen articles met the inclusion criteria. Most articles used information from medical files as the gold standard, defined as any recurrence. Two studies included bone metastases only in the definition of recurrence. Fewer studies used a model-based approach (decision trees or logistic regression) (41.2%) compared with studies using detection rules without specified model (58.8%). The generalized linear mixed model for all recurrence types reported an accuracy of 92.2% (95% confidence interval = 88.4% to 94.8%).

CONCLUSIONS

Publications reporting algorithms for detecting breast cancer recurrence are limited in number and heterogeneous. A thorough analysis of the existing algorithms demonstrated the need for more standardization and validation. The meta-analysis reported a high accuracy overall, which indicates algorithms as promising tools to identify breast cancer recurrence at the population level. The rule-based approach combined with emerging machine learning algorithms could be interesting to explore in the future.

A real-world study on characteristics, treatments and outcomes in US patients with advanced stage ovarian cancer

Background

Detailed epidemiologic descriptions of large populations of advanced stage ovarian cancer patients have been lacking to date. This study aimed to describe the patient characteristics, treatment patterns, survival, and incidence rates of health outcomes of interest (HOI) in a large cohort of advanced stage ovarian cancer patients in the United States (US).

Methods

This cohort study identified incident advanced stage (III/IV) ovarian cancer patients in the US diagnosed from 2010 to 2018 in the HealthCore Integrated Research Database (HIRD) using a validated predictive model algorithm. Descriptive characteristics were presented overall and by treatment line. The incidence rates and 95% confidence intervals for pre-specified HOIs were evaluated after advanced stage diagnosis. Overall survival, time to treatment discontinuation or death (TTD), and time to next treatment or death (TTNT) were defined using treatment information in claims and linkage with the National Death Index.

Results

We identified 12,659 patients with incident advanced stage ovarian cancer during the study period. Most patients undergoing treatment received platinum agents (75%) and/or taxanes (70%). The most common HOIs (> 24 per 100 person-years) included abdominal pain, nausea and vomiting, anemia, and serious infections. The median overall survival from diagnosis was 4.5 years, while approximately half of the treated cohort had a first-line time to treatment discontinuation or death (TTD) within the first 4 months, and a time to next treatment or death (TTNT) from first to second-line of about 6 months.

Conclusions

This study describes commercially insured US patients with advanced stage ovarian cancer from 2010 to 2018, and observed diverse treatment patterns, incidence of numerous HOIs, and limited survival in this population.