4

Ideas for Improving Data Sources, Methodology, and Research

Individual presenters and participants at the 2013 and 2014 workshops offered their views on (1) principles for work on premature mortality and risk factors, (2) ways to improve existing data sources, (3) suggestions for improving methodology, and (4) future research needed to further understanding on these topics.

PRINCIPLES FOR FUTURE WORK

During his presentation, Michael McGinnis suggested a list of principles that, in his view, could guide future policy and research on premature mortality and risk:

• Clarify the analytic agenda—the present workshop is a beginning.
• Keep the focus on the most important elements.
• Avoid “risk factor envy” whereby researchers focused on one topic compete against those studying another topic—the whole picture is more important than any single element.
• Avoid the inadvertent double standard with a higher burden of proof for prevention over treatment.
• Don’t let the perfect be the enemy of the good but know what the imperfections are.

• Consider the determinants of lifetime health and not just early death.

IMPROVING DATA SOURCES

Throughout both workshops, many participants recognized the need to improve consistency, measurement, and comparability of data on mortality and risk factors (especially cause-specific mortality data, social network measures, and physical activity measures) across countries, states, and counties. In addition, participants discussed two key ways that data sources used to understand premature mortality and risk factors could be improved—incorporation of social and economic factors among the risk factors for premature mortality, and more precise measurements of behavior at the county level. First, more work to incorporate social risk factors both at the individual and aggregate level into research on the determinants of premature mortality is needed, according to a number of participants. Christopher Murray noted that he and his colleagues working on the Global Burden of Disease (GBD) Study plan to broaden the set of risk factors they analyze to incorporate social risk factors, a process that will take at least 2 years. Researchers will discuss how to develop comparable ways to quantify social risk factors and their effects on population health outcomes. Christine Bachrach added that more data and research are needed to strengthen the basis for communicating about the importance of social factors because they are deeply intertwined with and shape behaviors that are important to health outcomes. Therefore, a focus only on behaviors as health determinants is seriously misleading, according to Bachrach. More work is needed to determine which social and economic elements, ranging from community social engagement to prevalence of crime in an area, should be incorporated, others noted. Eileen Crimmins observed that measuring social determinants of premature mortality is particularly challenging in an international context. Murray noted Norway is beginning to create a national dataset incorporating educational attainment among other social factors, but that country is an exception.

Samuel Preston also indicated social and economic factors are important to study, but noted an interstate factor analysis of causes of death points to the importance of behavior more than poverty or medical care factors. Further, Preston shared that research on the impact of educational change on health and mortality indicates compulsory school laws can have an effect. However, current frameworks used to examine the

proximate determinants may not be suitable for examining socioeconomic differences.

Second, many participants at the 2014 meeting noted the importance of more precise measurements of behavior at the county level. Specifically, Ali Mokdad indicated that he would like to see improvements made to the Behavioral Risk Factor Surveillance System. He explained that a national sample with physical measurements in every state would be an efficient way to accomplish “integrated surveillance.” In addition, more county-level data with pooled cohorts might improve imputation, as John Haaga, Murray, and others discussed.

IMPROVING METHODOLOGY

One of the primary methodological issues the group addressed for future efforts was the significance of measuring disability adjusted life years (DALY) versus years of life lost (YLL) and other measures of premature death. As noted during the presentations, behavior explains a smaller percentage of DALY than deaths. Some suggested that the array of risk factors that explain disability outcomes might differ in important ways from those that explain early deaths, requiring an expansion of risk factors considered or a separate effort. Preston noted the two concepts of DALY and YLL are related in that the combination of these two indices equals the years lost to a particular condition, and that the GBD currently captures both. He suggested the choice of examining DALY versus YLL is more of a conceptual choice to be made. Crimmins and others discussed whether considering the determinants of disability requires weighting disability life years, a particularly difficult policy issue, and how closely related they are to YLL.

Majid Ezzati raised a second methodological issue that he and his colleagues will be addressing in the future. He stated that over the coming year, they will examine issues around ranges of key values of risk factors related to premature mortality, including obesity, as well as changes in effect sizes and mediation using a worldwide cohort of over 10 million participants.¹ He noted biases will continue to exist at some level because the data are observational. However, he added this and other large studies will help to address uncertainty in findings about particular risks such as obesity and diabetes.

_____________

¹ Some of this work has been published since the workshop: Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects) (2014).

INDIVIDUAL IDEAS FOR FUTURE RESEARCH

Several presenters were asked to share their ideas about future research needed in this topic area. Suggestions, and the people who offered them, included the following:

• application of the complete GBD framework to measure risk at the local level especially for the United States (Murray, Haidong Wang);
• more intense study of geographic differences by urban, suburban, rural, or other units of geographic analysis beyond the county (Connie Citro);
• greater study to identify determinants of musculoskeletal disorder, mental health outcomes, and other disability conditions (Murray);
• quantification of distal socioeconomic determinants of specific diseases related to premature mortality and addressing the problems of limited generalizability for all-cause mortality findings (Murray);
• research on causes of early mortality that include lack of vaccination, lack of access to care, and other factors related to medical care (Theo Vos);
• quantification and development of a stronger evidence base about severity distributions for diseases and long-term disabilities and measure how these are affected by treatments (Vos).