Our #DataSmart Resources series curates links and examples for those looking for an introduction to a particular civic data topic.
A few years ago, Ellen, an 82 year-old widow with a history of heart problems living alone in Anaheim, got on her scale and noticed she was a few pounds heavier than normal. While this was of no great concern to Ellen, her scale was connected via Internet to the local clinic, where it would send regular readings. An employee at the clinic saw Ellen’s readings and immediately wanted to follow up. Considering Ellen’s history of congestive heart failure, a significant weight gain in a short amount of time was a concerning sign, an indication of possible fluid buildup in the lungs and increasing pressure on an already stressed heart. The clinic requested that Ellen come in, and found that her heart was indeed under duress, requiring treatment starting that very morning. Had the warning signs not been spotted so early, Ellen may have experienced a long and painful hospitalization, if not worse.
This example illustrates the potential for data to transform healthcare. While governments may not have the resources nor the jurisdiction to respond to individual cases like Ellen’s, aggregations of many residents’ health data can help governments target public health interventions to those areas in greatest need, much as Ellen’s clinic was able tailor its care based on her health data. This article offers a host of curated resources to aid governments in developing data-driven public health programs.
The National Committee on Vital and Health Statistics (NCVHS) in the U.S. Department of Health and Human Services (HHS) published a Toolkit for Communities Using Health Data, which offers a framework for stewardship of health information collection, use, and disclosure. Many communities already possess troves of actionable data, but the ways they use and protect this data will determine its ultimate value. This toolkit explains in detail eight critical elements of data stewardship: accountability; openness, transparency, and choice; community and individual engagement/participation; purpose specification; quality and integrity; security; and de-identification of data.
The Big Data in Health Working Group released a report providing data-driven public health recommendations specifically for areas outside the complex hospital systems of developed nations. The report offers a taxonomy for managing health data, endorsing a balanced approach that releases unrestricted private data in open commons while building privacy and security capacity around personal and restricted government data. In order to support the development of data-driven health systems, the authors recommend encouraging public-private partnerships, ensuring data access by patients and healthcare providers, and creating centers of excellence to train behavioral and health scientists in using data. Applying these considerations to developing nations with currently limited data-driven capacity can produce results that can transform their health systems.
In an effort to make public health data more accessible and actionable, the National Cancer Institute designed a workbook called Making Data Talk. The resource introduces a number of communication concepts, proposes a framework for communicating data, and applies that framework to two public health scenarios: crisis situations and advocacy.
On the more technical side, the National Association of Health Data Organizations (NAHDO) published a set of public health data dissemination guidelines. The publication is intended to assist data managers, epidemiologists, and analysts in public health effectively release public health information online. Recommendations are intended to facilitate the release of meaningful statistics while reducing of risk of inappropriate disclosure of sensitive information and ensuring security of data. Specific guidelines will address statistical approaches for releasing public health data on web-based dissemination systems.
Courtesy of the Agency for Healthcare Research and Quality, A Robust Health Data Infrastructure offers a possible software architecture for governments to exchange health information. The authors propose an architecture that can manage reasonable deviations in data formats, protocols, and interfaces. This architecture conforms to a number of core principles, including operability with all data sources, encryption of data in rest and in transit, and inclusion of corresponding metadata, context, and provenance information.
Locating Health Data
The first step in implementing health data analysis programs is finding good sources of data to analyze. A number of universities and organizations devoted to public health have compiled lists of sources offering relevant datasets. Partners in Information Access for the Public Health Workforce keeps a running list of county and local health datasets. Both the University of Minnesota and George Washington University have also published collections of public health datasets, among them many datasets particular to city public health.
Particularly useful among the datasets cited in these lists are a couple sets recently collected by non-profits on city health. The Big Cities Health Coalition has published data from the 28 largest cities in the U.S. on a number of specific health indicators, ranging from heart disease mortality rate, to unintended opioid-related death rates, and smoking rates for teens. In partnership with the Centers for Disease Control (CDC) and the CDC Foundation, the Robert Wood Johnson Foundation unveiled the 500 Cities project. For this project, the foundation released city- and census tract-level area estimates for chronic disease risk factors, health outcomes, and clinical preventive service use for the largest 500 cities in the United States.
Other public health experts have identified alternative data sources for public health data. An article from the Journal of the American Medical Association recommends turning to data from the health care system in order to supplement the often-outdated data provided by public sources. While many hospitals and physician offices already use health care data in order to measure their quality of care, cities could use this data in order to understand the health of their community. For example, data on preventable admissions for asthma could help public health officials identify hot spots of poor access to care and environmental conditions that lead to symptoms. However, in order to take advantage of these opportunities, cities must find ways to ensure cooperation between competing hospitals and physicians, and create a privacy framework that is amenable to residents.
A paper from The Lancet calls for the use of research data to improve public health. Public health research uncovers insights into how interactions between behavior and environment determine causation and variation in health and disease, knowledge that is underused by governments. In order to take better advantage of such research, governments and research institutions need to develop closer relationships that facilitate data sharing and collaboration. However, in establishing these relationships, both parties must be cautious not to favor researchers working in resource-rich areas who may have better access to advanced analytical tools. Moreover, data should only be shared if adequate safeguards are in place to protect the privacy of research participants. And finally, in order to justify the substantial cost of data sharing, researchers must focus on projects that produce useful and actionable data.
Social media may be another source for health data that can be leveraged by public health officials. The Journal of Internet Medical Research published a paper analyzing Facebook likes as potential indicators of health outcomes that can complment traditional public health surveillance systems and provide data at a local level. The study found that Facebook likes add significant value in predicting health outcomes. A strategy of social media mining could be used to access such data in an anonymized, but geo-located way in order to understand health outcomes across cities or in different areas of a city.
Navigating the Challenges of Data-Driven Public Health
Programs that use data to improve public health must put a premium on results for real people, anticipating unintended consequences and risks and putting measures in place to navigate challenges. PLOS Computational Biology released an article that analyzed the ethical challenges of using data to inform public health, emphasizing the need to engage with these questions while the field is still at an early stage of evolution. The authors outline three basic categories of ethical issues: context sensitivity issues about privacy, transparency, and equitable application; issues at the nexus of ethics and methodology, including the false identification of outbreaks and pressure to mobilize resources based on unvalidated predictions; and issues around legitimacy requirements, regarding shared codes of practice for responding to inaccuracies, redressing harms, and managing public expectations about outbreaks. It is vital that engagement with these challenges comes to be seen as part of the development of data-driven public health itself, not as some extrinsic constraint.
And finally, an article from BioMed Central Public Health focuses on the challenges associated with sharing data in public health. The article identifies twenty potential barriers in six categories: technical, motivational, economic, political, legal, and ethical. While the first three categories are rooted in well-known challenges associated with health information systems for which solutions have yet to be found, the latter three have solutions that lie in an international dialogue aimed at generating consensus on policies and instruments for data sharing. In order to accelerate the use of valuable public health information, governments should keep these potential barriers in mind when developing data sharing strategies.
To stay updated on public health news, be sure to follow our public health stories on Data-Smart.