Impact of the Affordable Care Act
NOTE: Due to a technical glitch in the conference room, several minutes of the audio are missing from the end of the recording.
Room: Crimson Clover
Wednesday, April 22, 2015, 10:45 AM to 12:15 PM
Moderator: Carolyn Leep, MS, MSPH
Michael Meit, MPH, MA
Impacts of the Affordable Care Act for State and Local Health Departments
Co-Investigator(s): Amy Nevel, MPH (also presenting)
Background: The Patient Protection and Affordable Care Act (ACA) includes reforms that expand insurance coverage and enhance access to clinical preventive services. To explore the impacts of the ACA on state and local health departments, the HHS Office of the Assistance Secretary for Planning and Evaluation (ASPE) contracted with NORC at the University of Chicago to conduct the study, "Implications of the Affordable Care Act for HHS Public Health Programs." Research Objective: The study seeks to identify the effect of ACA reforms on where individuals seek clinical preventive services, where services are delivered, how health departments are responding and plan to respond to ACA-related changes, anticipated future changes to public health practice, and the role of federal support for public health programs. Data Sets and Sources: Qualitative data were collected via interviews with staff from five state health departments (SHDs) and select local health departments (LHDs) in case study states. Researchers held discussions with the state health officer, program staff, Medicaid staff, and LHD staff. Study Design: Methods include an environmental scan, case studies using site visits and telephone interviews, and ongoing consultation with a Technical Advisory Group (TAG). Case study states were selected using diverse selection criteria, including governance structure, geography, rurality, and Medicaid expansion (i.e., expanding or not expanding). Analysis: The research team conducted qualitative content analysis and coding to identify themes and group findings by key topics. Qualitative findings were summarized into written case study issue briefs. Principal Findings: Findings show that health departments are seeing reductions in client volume for some public health programs (e.g., immunizations, breast and cervical cancer screening) and respondents believe it is a result of the ACA. While all case study states are expanding their capacity to bill for services, they report that billing is unlikely to replace program revenue should program funds be reduced. Many respondents discussed that some health departments may need to continue providing preventive services, especially in areas where there is an insufficient number of providers. While concerned, participants acknowledge that many of the changes that will result from the ACA are still unknown. Conclusions: Health department leaders expressed concern that reduced utilization of key public health services may make it difficult to sustain programs seen as important to maintaining the public's health. Further, they note that the resulting loss of revenue could have additional impacts on health department functions and response capacity. NORC will continue to explore these themes and key findings by conduction an additional set of five case studies. Implications for Public Health Practice and Policy: NORC is providing a voice to state and local public health leaders' concerns about the impacts of the ACA, while also capturing strategies for program planning, sustainability, and adaptation. To improve practice, health departments must be aware of the opportunities and challenges that may emerge because of the ACA, as well as strategic planning efforts states and localities are implementing to adapt to anticipated changes.
Kyle Bogaert, MPH
Variations in Levels of Clinical Service Delivery at State Health Agencies
Co-Investigator(s): Rivka Liss-Levinson, PhD & Katie Sellers, DrPH, CPH
Background: State and territorial health agencies (S/THAs) play a critical role in promoting and protecting the health of their residents by performing a wide range of functions and services. Given the breadth of services S/THAs provide their residents, their service delivery levels may vary over time due any number of factors. In particular, levels of clinical service delivery by S/THAs may currently be changing in part due to changing budgets, high priority issues, and the political climate including the implementation of the Affordable Care Act. Research Objective: This research assesses changes in levels of service delivery for clinical services by S/THAs by governance classification, region, and size of population served using the 2014 and 2015 ASTHO Forces of Change surveys. Data Sets and Sources: The Association of State and Territorial Health Officials (ASTHO) conducted an online survey of S/THAs in July 2014, entitled Forces of Change, to address emerging issues in the management, resources, activities, and finances of S/THAs, including the impact of the Affordable Care Act. The survey will be re-administered in January 2015. Study Design: The 2014 ASTHO Forces of Change survey was administered to the 50 states, D.C., and U.S. territories and freely-associated states in July 2014. The survey will be administered to the same respondents in January 2015; the study is cross-sectional. Analysis: Thirty-seven states and two freely-associated states responded to the 2014 ASTHO Forces of Change survey. Data cleaning was completed, and unusual values were clarified. Descriptive statistics were calculated using SPSS. Principal Findings: In general, most S/THAs saw little or no change in levels of service delivery for select clinical services; however, there was variability among the different services. Among S/THAs that reduced delivery of at least one of the select clinical services, a larger percentage were in the Mid-Atlantic & Midwest region, served a large population, or had a centralized governance structure. Among those S/THAs that increased delivery of at least one of the select services, a larger percentage served a medium population, were in the South, or had a decentralized governance structure. The results will be updated to include the 2015 data. Conclusions: Overall, S/THAs experienced little change in levels of service delivery for select clinical services. The overall stability suggests that factors that might impact levels of service delivery, including changes in budgets and the political climate, have not had a significant effect on levels of clinical service delivery. Further research is needed to investigate potential causes for the variations in levels of delivery across geographic regions, governance classifications, and sizes of populations served. Implications for Public Health Practice and Policy: This analysis contributes to the PHSSR research agenda by utilizing current data to assess variation in the levels of clinical service delivery by S/THAs. Further research may build upon the initial analyses to elaborate on the variations across geographic regions, size of population served, and governance classification. Additionally, future longitudinal research may assess the potential impact of the implementation of the Affordable Care Act on levels of clinical service delivery at the governmental public health level.
Michael Preston, PhD, MPH
Insurance Coverage Mandates: Impact of Physician Utilization in Moderating Colorectal Cancer Screening Rates
Co-Investigator(s): Billy Thomas, MD, MPH; Jonathan Laryea, MD; Zoran Bursac, PhD, MPH; Glen Mays, PhD, MPH; & J. Mick Tilford, PhD
Background: Colorectal cancer (CRC) is the third most common cancer found in men and women in the United States. In 2014, the American Cancer Society estimated as many as 136,000 new cases of colorectal cancer and approximately 50,000 deaths. Health care reform was introduced in 2010 and became the cornerstone for Americans seeking change in the health care system. Health care reform is a critical factor in increasing CRC screenings by increasing coverage rates for all Americans. The primary purpose of the Affordable Care Act is to decrease the number of uninsured Americans and reduce the overall cost of health care by reducing the amount of out-of-pocket expenses for preventive services such as colorectal cancer screenings. Physician visits can moderate preventive service utilization for CRC when out-of-pocket expenses are reduced. Research Objective: To examine insurance coverage mandate variations and the effect of physician utilization in moderating colorectal cancer screening rates. Data Sets and Sources: Secondary data were analyzed from the Behavioral Risk Factor Surveillance System and National Cancer Institute State Cancer Legislative Database from 1997-2012. The target population was a sample of U.S. adults age 50 to 74 that lived in an insurance coverage mandate or non-mandate states before and after health care reform. Study Design: Retrospective cohort study using a strong quasi-experimental design to examine insurance coverage mandate variations and the effect of physician utilization in moderating CRC screening rates from 1997-2012. Analysis: A time-series analysis using a difference-in-difference-in-differences (DDD) approach was used to examine the effect of health care reform on non-mandate states. Key variables of interest are insurance coverage mandates, colorectal cancer screening status, and physician utilization. Principal Findings: The adjusted average marginal effects from the difference-in-difference-in-differences model indicates that physician utilization increased the probability of being “up-to-date” relative to being non-compliant by 1.5 percentage points, suggesting that an estimated 1.27 million additional age-eligible persons would receive a CRC screening after health care reform annually. Our findings are robust to different model specifications. Conclusions: Health care reform that lowers out-of-pocket costs is an effective approach that increases colorectal cancer screenings when physician utilization is considered as a mechanism. Implications for Public Health Practice and Policy: With the introduction of the Affordable Care Act, responsive public health systems require strategies to determine which policies, systems, and administrative strategies are most effective in increasing preventive service utilization and reducing health disparities. This research demonstrates that insurance mandates that reduce out-of-pocket costs increased colorectal cancer screenings when physician visits are considered as a moderator. Future health care reforms that increase access to preventive services, such as CRC screening, are likely with low out-of-pocket costs and will increase the number of people who are considered “up-to-date”.
Eoghan Brady, MS
Machine Learning Algorithms to Code State Public Health Spending Accounts
Co-Investigator(s): Jonathon Leider, PhD; David Bishai, MD, PhD, MPH, BA, FAAP; Beth Resnick, PhD; & Jennifer Le, MPH
Background: The US census of local governments has collected detailed spending information from state health departments since 2000. Spending data mixes individual clinical services with public health spending. Labels for what is spent vary widely across states. Research Objective: To compare human and machine learning approaches to coding public health spending data. Data Sets and Sources: Census data on state’s non-hospital health spending (code 32) from 2000-2012 were examined. There were 1.8 million separate entries of spending data overall and roughly 150,000 entries per year. A single state report comprised between 350 to 13,000 entries of reported expenses using up to six ledger fields. There were only 56,291 unique descriptions of spending items in the ledgers. Study Design: Teams of human coders examined all labels for spending, reached consensus, and then taught a machine to code the same way. Analysis: Two teams of human coders examined the 56,291 unique entries coding each as “Yes”, “No” or “Maybe” public health spending. They met weekly for 6 months to share decisions about frequently encountered items. Individuals recoded each others files and examined discrepant cases until each of the two teams reached consensus. RTextTools was used for document classification and text mining. Nine models were fit to the manually coded data, namely Support Vector Model (SVM), GLMnet, Maximum Entropy, SLDA, Boosting, Bagging, Random Forests, NNEt, Tree. The performance of each of the nine models was compared. Precision, recall and composite F-scores are reported. Results of all algorithms are combined to measure ensemble performance based on agreement of predictions. Principal Findings: Precision, which is the proportion of predictions that are correctly assigned to a class that actually belong to that class ranged from 0.74 to 0.91 across the 9 models. Recall, the proportion of each class that is correctly assigned by that algorithm, ranged from 0.79 to 0.93. An F-score is calculated as a composite measure of overall performance and ranged from 0.77 to 0.92. Requiring 6 or more algorithms to be in agreement still achieved 88% recall with 92% coverage of observations. This meets the usual standard for inter-rater agreement of 90%. Conclusions: We have shown the potential for computer algorithms to closely replicate the choices made by human coders in deciding how much states are spending on public health. Validated uniform data on how states have been spending on public health can help practitioners compare their spending to others and to outcomes achieved. Implications for Public Health Practice and Policy.: Census data on public health spending holds great potential for PHSSR, if only it can be properly coded. Until the US embraces a uniform chart of accounts, each year of census data will generate an additional 150,000 lines of state spending data that will have to be recoded by human coders.