Friday, March 17, 2017

How Should We Measure The Distribution Of Health In A Population?

Population health has been defined as "the health outcomes of a group of individuals, including the distribution of such outcomes within the group." Measuring population health and its distribution can unite groups across sectors around a set of clear, defined goals. However, no one metric can capture the intricate and complex nature of population health. Instead, we need a matrix of indicators to gain a full picture of health and how it is changing. (see Figure 1). For example, rather than measuring only end-of-the-line health outcomes such as mortality, we need to measure a range of metrics across the health pathway, including the determinants of health, risk factors, prevention and treatment.

Figure 1: A framework for measuring population health

In addition, to understand the distribution of health in a population and address inequalities, we need to measure health across different subpopulations.

Yet there is little evidence on which sub-population groups should be considered. Commonly used segmentations are based on socioeconomic status, geography, gender and ethnicity. However, population health can also be explored across different disease or age groups. In addition, risk factors play an important part in determining population health, and could provide a basis to segment the population. There also exist specific societal or clinical groups that carry particular relevance to policymakers, such as employees, prisoners, homeless people, disabled people or people with drug dependencies.

While all these population groups are important to population health, it is not practically possible, or desirable, to measure and present health outcomes across all possible dimensions. Therefore, we conducted an expert Delphi study, which uses several rounds of questionnaires, where the results from earlier rounds feed into the next in order to reach a consensus among participants. Our goal was to prioritize population segmentation approaches, and guide both the collection and presentation of population health data.

The Delphi Study

On July 8 and July 9, 2016, Imperial College's Institute of Global Health Innovation (IGHI) and GE's Healthymagination brought together 30 experts from health care, academia, industry and government to explore what it takes to create healthy populations in the 21st century. Ahead of the event, the experts were invited to respond to two rounds of questionnaires to prioritize segmentation approaches. The online questionnaire received an 89 percent response rate.

Round one of the Delphi presented panelists with a long-list of population segmentation approaches, from which they were asked to select the ones they considered important for measuring the distribution of health in a population. Some population segmentation approaches (e.g. socioeconomic groups) were followed by a list of subcategories (e.g. income level, education level, etc.). Participants were given the option to add to the initial list of options during the first round of questions. Items which at least 70 percent of participants considered important were carried over to the second round, where participants were asked to rank them.

There was a high level of consensus around which population segmentation approaches to include in the second round of consideration, with all ten options meeting the 70 percent cut-off (see Figure 2).

Figure 2: First round results on high-level groupings

In the second round, the delegates ranked socioeconomic status as the most important population segmentation to understand and measure population health (see Figure 3). This is interesting considering that other segmentation approaches are more directly linked to health, such as clinical groups, age or long-term conditions. This emphasizes the importance of going beyond traditional health care factors and considering the wider determinants of health. (No categories were consistently considered less important, demonstrating that there exists a wide range of population segmentation options that should be considered.)

Figure 3: Ranking of population groups for measuring population health

The ranking of the subgroups resulted in varying degrees of consensus (see Figure 4). Within age groups, geographical areas, socioeconomic and risk groups, there were no clear favorite subgroups.

Subcategories for specific long-term conditions and societal groups did exhibit a clear order of preference. Diabetes received an average ranking of 1.6 out of 5, making it the top ranked disease to consider when measuring population health. The most important societal group was unemployed people, which received a 1.8 average rank out of 4. Finally, the various clinical groups received similar rankings ranging between 3.0 and 4.6 out of 6. Interestingly, here the newly added subcategory 'multiple chronic conditions' received the highest ranking.

Figure 4: Ranking of subcategories

Implications For Policymakers

There exists a clear consensus among health care experts around the need for population segmentation in order to measure population health and health equity. However, there is no single way to do this. All ten population segmentation approaches were considered important by the panel. These results highlight the value of considering the wide range of different population groups that may influence health outcomes.

The results of this study can help researchers and policymakers prioritize the way they analyze and present population health data. In addition, these results should guide the collection of data. For example, the panel considered socioeconomic status and risk factors to be very important, but administrative datasets collect information on these issues in different ways and according to different definitions. Standardizing the collection of segmentation variables would allow population-wide analysis of the distribution of health.

Policymakers should also consider using a data-driven approach to identify population segments, rather than a priori defined population groups. Big data and data mining techniques can help quantify the distribution of outcomes in a population and identify the factors driving these differences.

It is important to note that measuring the distribution of health is only one of the many steps we can and should take to create healthier populations. We must continue to explore how population health will benefit from emerging innovations in technology, service model design and big data and analytics.

Authors' Note

GE Healthymagination and IGHI would like to thank the delegates for their contribution to the event.



from Health Affairs BlogHealth Affairs Blog http://ift.tt/2mauYtp

No comments:

Post a Comment