Harnessing the Power of Big Data and Simulation to Improve Colorectal Cancer Screening

Technical Considerations: The Past, Present, and Future of Simulation Modeling of Colorectal Cancer

Authors:

Siddhartha Nambiar (Presenter)
University of North Carolina

Maria Mayorga, North Carolina State University
Rachel Townsley, North Carolina State University
Kristen Hassmiller Lich, University of North Carolina
Stephanie Wheeler, University of North Carolina

Public Health Statement: Several counties within North Carolina (NC) were recently identified as one of three distinct “hot spots” in the U.S. due to excessive colorectal cancer (CRC) deaths. Simulation models can help us to understand which evidence-based interventions will have the most potential for success and inform planning and implementation efforts in NC and other states. However, the public health community may not be fully informed about the scope and technical capabilities of existing simulation tools.

Purpose: In this session, we aim to shed light on the current landscape and state-of-the-art in simulation modeling of CRC progression, screening, and treatment, and to identify areas that offer opportunities for public health intervention.

Methods/Approach: We will provide an overview of approaches to simulation, including a discussion of the benefits and caveats of cohort versus individual simulation modeling. We will describe existing simulation models for CRC progression, screening, and treatment, including a detailed description of our state-based individual simulation model. We also will summarize key technical considerations that are necessary to develop valid and reliable cancer-focused simulations for public health planning and implementation.

Results: Including individual risk factors, screening behaviors, and healthcare utilization and costs into CRC simulation models can inform public health efforts to improve CRC screening coverage through value-based, data-informed decision making.

Conclusions/Implications: Increased awareness by public health experts of the capabilities of CRC simulation modeling efforts can increase speed of implementation of evidence-based programs and practices and allow us to quantify disparities.

Using Individual-Based Simulation Modeling to Integrate Big Data and Intervention Evidence to Inform Intervention Selection, Adaptation, and Evaluation: An Example on Colorectal Cancer Screening

Authors:

Kristen Hassmiller Lich (Presenter)
University of North Carolina at Chapel Hill

Maria Mayorga, North Carolina State University
Rachel Townsley, North Carolina State University
Stephanie Wheeler, University of North Carolina at Chapel Hill
Leah Frerichs, University of North Carolina at Chapel Hill

Public Health Statement: This presentation illustrates one approach for leveraging big data (here, all-payer claims data, census data, BRFSS, state medical facilities data, and trial data) and cutting-edge individual-based simulation methods to inform decision makers’ understanding of the reach, impact, and cost requirements required for alternative screening interventions under consideration to be cost-effective. It takes into account the unique intervention context—including characteristics of the population, determinants of current care/behavior, and existing resources and processes.

Purpose: Despite evidence that colorectal cancer screening saves lives, adherence to care guidelines is suboptimal and disparities persist. The relative impact of evidence-based interventions is not fixed across contexts. We will illustrate this, and our approach to supporting decision makers’ learning and decision making by assessing the required cost and effectiveness of four evidence-based interventions under consideration in North Carolina.

Methods/Approach: We used synthetic population data, statistical models estimating the relationship between multi-level determinants and both receipt and modality of screening, and a natural history simulation model of colorectal cancer to project the effects of different intervention approaches on colorectal cancer screening and outcomes to the population level in North Carolina and for subpopulations (e.g., African American males, Medicaid enrollees). We used the model to estimate the combinations of individual intervention cost and effectiveness that result in cost-effectiveness estimates under various willingness to pay thresholds.

Results: Across the ranges of plausible willingness to pay thresholds, a statewide mass media campaign robustly dominates other alternatives. Two-dimensional data tables and visual graphics depict the combinations of intervention cost and requisite per-person impact for alternate interventions to be cost-effective in North Carolina. For all interventions considered, clear thresholds exist, determining conditions under which alternate interventions are more effective, or whether any intervention is within the range of cost-effectiveness we are willing to pay for.

Conclusions/Implications: We will illustrate how this information can be used to inform decision makers’ conversations about intervention feasibility, selection, adaptation, and evaluation.

Relationships, Data, and Quality Improvement: Critical Factors When Accountable Care Organizations and Primary Care Practices Collaborate to Increase Colorectal Cancer Screening in Medicaid Members

Authors:

Melinda Davis, PHD (Presenter)
Oregon Health & Science University

Rose Gunn, MA, Oregon Health & Science University
Robyn Pham, Oregon Health & Science University
Kristen Hassmiller Lich, PhD, MHSA, University of North Carolina at Chapel Hill
Stephanie B. Wheeler, PhD, MPH, University of North Carolina at Chapel Hill

Public Health Statement: Research to describe how big data can inform tailored implementation in real world, dynamic contexts is needed.

Purpose: Health system stakeholders are increasingly aligning as Accountable Care Organizations (ACOs) to achieve the triple aim. Research is needed to describe how ACOs interface with primary care clinics to implement evidence-based interventions. Our study explores how Oregon’s 16 Coordinated Care Organizations (CCO)—which are the single point of accountability for health care access, quality, and outcomes of Medicaid members—are working with clinics to increase colorectal cancer (CRC) screening, one of 18 CCO quality incentive metrics.

Methods/Approach: Observational cross case comparative study of Oregon’s CCOs using public document review, key informant interviews, and field notes from technical assistance consults with CCO leaders. Data was collected between February 2016–August 2016, transferred to Atlas.ti and analyzed by a multidisciplinary team using a data-driven, emergent approach.

Results: We engaged 14 CCOs and 26 key informants who represented state innovator agents (n=4), CCO leadership (n=16), and primary care practices (n=6). Over 30% of the informants (n=8) worked with more than 1 CCO. CCOs were implementing multiple interventions to improve CRC screening, including efforts designed to reduce structural barriers (e.g., direct mail programs), facilitate provider and patient behaviors (e.g., education, incentives, reminders), and increase the capacity of practices to implement desired changes (e.g., staffing, data management). CCOs addressed three key dimensions as they sought to improve CRC screening in partnership with regional primary care clinics: 1) establishing relationships, 2) producing and sharing data, and 3) developing a process and infrastructure to support quality improvement.

Conclusions/Implications: Our research shows that CCOs/ACOs need to consider relationships, data, and quality improvement infrastructure when working with clinics to implement CRC screening interventions. Health system and policy leaders must attend to these factors when moving from big data to support implementation of population health initiatives across diverse multilevel payer and practice settings.

Data-powered Decision Making: One State’s Approach to Improving Colorectal Cancer Screening in Underserved Populations

Authors:

Stephanie Wheeler (Presenter)
UNC at Chapel Hill

Maria Mayorga, North Carolina State University
Melinda Davis, Oregon Health & Science University
Leah Frerichs, UNC at Chapel Hill
Michael Pignone, University of Texas Austin
Florence Tangka, Centers for Disease Control and Prevention
Lisa Richardson, Centers for Disease Control and Prevention
Kristen Hassmiller Lich, UNC at Chapel Hill

Public Health Statement: Despite major advances in care delivery, substantial variation in cancer-specific mortality persists, with the largest gaps by geography, race/ethnicity, and income occurring in preventable and curable cancers, like colorectal cancer (CRC). Big data coupled with simulation can enable more informed decision-making about which CRC-focused efforts are expected to yield the greatest value and reduce disparities.

Purpose: We describe our approach to improving CRC screening in North Carolina, focusing on data we have assembled and used for simulation. We also describe how these efforts have been translated into CRC-focused quality improvement efforts locally.

Methods/Approach: We employed health insurance claims data from Medicare, Medicaid, and commercially available health plans, as well as cancer registry data, to pinpoint regions and sub-populations with relatively low CRC screening rates and high CRC mortality. We engaged local stakeholders and underserved patients to understand context-specific considerations and preferences using in-depth interviews and discrete choice experiments. We assembled these and other relevant contextual data to develop data-powered decision models to anticipate impact and cost-effectiveness of evidence-based interventions.

Results: Claims and cancer registry analyses revealed counties and sub-populations with the lowest rates of CRC screening and highest CRC mortality. Findings were used to launch a mailed reminder +/- FIT campaign in the Medicaid population within one urban county with low CRC screening rates.

Conclusions/Implications: Once the multilevel determinants of CRC outcomes are better understood locally, evidence-based approaches to increase CRC screening can be systematically evaluated to ensure that decision-makers select, adapt, and implement interventions optimally.