Environmental health indicators (EHIs) monitor how the environment affects human health and can be defined as “an expression of the link between environment and health, targeted at an issue of specific policy or management concern and presented in a form that facilitates interpretation for effective decision-making” [1
]. When well designed and constructed, these indicators can provide valuable information for policymakers and decisionmakers to prioritise and evaluate actions, as well as raise awareness of issues [2
]. However, developing environmental health indicators can be challenging. Indicators must reflect known causal relationships between the environment and health outcomes and meet a range of selection criteria to be useful, reliable, and robust [3
]. Little guidance is available on how to develop country-specific EHIs, in particular, when numerous potential indicators meet the selection criteria and need to be prioritised to a set of core indicators.
The main guidance for developing environmental health indicators comes from documents produced by the World Health Organization (WHO) in the 1990s and early 2000s [3
], including through its work in the WHO European Region [4
] and on children’s environmental health indicators [2
]. Additionally, other substantial projects have developed environmental health indicators or health indicator sets with an environmental aspect, including the CHILD project of child health indicators for European countries [5
], and indicators for children’s health and the environment in North America [7
]. In addition to outlining a methodology, these projects developed generic sets of environmental health indicators for use in multiple countries. However, environmental health issues can differ markedly from country to country [3
] and depend on the local context. Additionally, new scientific evidence has become available on the links between the environment and health. As a result, environmental health indicators may need to be reviewed or developed by countries, using the most recent evidence and tailored to the country’s needs.
A specific methodology for developing environmental health indicators is outlined in the WHO children’s environmental health indicators report, incorporating the three stages of scoping, selection, and design [2
]. The scoping stage involves identifying the important environmental health issues, the key users and their needs, and the information needs that the indicators need to fulfil. The selection stage involves developing a conceptual framework, identifying potential indicators, applying a set of selection criteria to potential indicators, and developing a provisional indicator list. The third stage, design, involves completing detailed indicator design, including assessing data availability, and deciding analysis outputs (for example, time periods and geographic areas).
Ideally, environmental health indicators cover aspects of both environmental exposures and the related health effects to create a balanced set of indicators. Developing a conceptual framework is important for understanding the pathways through which the environment affects human health and identifying what needs to be monitored. A number of conceptual frameworks exist [8
], including the DPSEEA (driving forces, pressures, state, exposure, effects, and actions) and more recently, the MEME (multiple exposures, multiple effects) model. The MEME model is based on the DPSEEA model but combines all types of exposures (proximal and distal) and also acknowledges the many-to-many relationships between exposures and health impacts [2
]. Regardless of the conceptual model used, health indicators must be based on known (or plausible) causal relationships, that is, an implied exposure-response relationship [2
]. Potential indicators are then identified for the different parts of the conceptual framework.
Selecting indicators involves assessing potential indicators against a range of selection criteria. A number of different sets of selection criteria for indicators exist, but the WHO children’s environmental health indicators suggest the following selection criteria [2
]: being credible (based on a known linkage), sensitive to change, consistent and comparable over space and time, robust, representative, accurate (based on reliable data), and able to be used at different scales. Indicators should also have clear relevance and utility, be relevant to an issue of policy or practical concern, actionable (related to an issue that is amenable to influence or control), understandable, timely, specific (targeted at an explicit phenomenon or issue), measurable, and cost-effective to produce and use. In New Zealand, the national statistics office has published a similar set of indicator criteria for official statistics, which includes data availability, scientific validity, sensitivity, consistency, comparability, methodologically sound measurements, intelligible and easily interpreted, able to be disaggregated, and timely [9
Although a provisional list of indicators may sufficiently meet the selection criteria, the list may still need further refinement. A long list of indicators may be resource-intensive to monitor and difficult for users to use and interpret, resulting in requests for a ‘core’ set of indicators [2
]. The final set of indicators must also be balanced and represent the conceptual framework [2
] rather than place disproportionate emphasis on one part of the framework. Conversely, the indicators should not be over-simplified; they still need to measure the most important issues and be useful and informative [2
]. A further process for prioritising potential indicators may therefore be required to complete the selection phase.
Other than applying indicator selection criteria, there is limited guidance for prioritising indicators. One approach is to use working groups and expert advisory groups, such as the WHO approach, with experts screening indicators according to policy relevance, health relevance, and potential data availability [10
]. In the context of developing children’s environmental health indicators, Briggs suggested that some indicators may be prioritised if they contain more information than others, for example, by relating to several different issues or by being sufficiently generic to have general applicability [2
]. Additionally, the children’s environmental health indicators used burden of disease statistics to prioritise the health needs most important to children globally [2
]; similarly, the CHILD project included consideration of the total health burden (on individuals, families, and society) as part of the selection criteria for health indicators [6
]. However, only a portion of the full burden of disease will be attributable to environmental factors, and the burden will vary by country. These global statistics are of little help without incorporating additional information such as the health burden within a specific country and the proportion of the health burden attributable to an environmental exposure.
To further prioritise potential indicators, two complementary approaches are helpful. Health impact assessment tools are used to assess the potential health impacts of new policies, programmes, and projects, and they consider the overall public health impacts, as well as health equity, as part of the assessment. Additionally, environmental burden of disease studies (or ‘attributable burden of disease’ studies) use robust evidence about the causal links between environmental exposures and health effects and provide information on the deaths, hospitalisations, and/or healthy years of life lost (measured in disability-adjusted life years (DALYs)) that could have been prevented if people were not exposed to a specific risk factor. These two methodologies were included in a review of potential conceptual frameworks for climate change EHIs [8
]. They were rejected as conceptual frameworks, as they were not designed to develop environmental health indicators, and they did not describe the exposure pathway sufficiently [8
]. However, we consider that these methodologies have great merit and utility in helping to prioritise indicators as part of a wider indicator development process, including a focus on health equity.
During this project, we developed a new approach for prioritising potential environmental health indicators, based on their public health impact and incorporating an environmental burden of disease approach. This approach was applied during the development of indoor environment EHIs for New Zealand, which covered aspects of housing quality and indoor air quality.
2. The Context
The New Zealand Environmental Health Indicators programme is funded by the Ministry of Health to achieve the following:
to monitor trends in the state of the environment;
to monitor trends in health outcomes linked to environmental hazards and exposures;
to compare the environmental health status of geographic areas and population groups;
to monitor the effectiveness of policies and other interventions on environmental health;
to help raise awareness about environmental health issues; and
to help initiate further investigations into links between the environment and health.
As at 2018, the indicator programme monitored over 60 indicators in nine domains: air quality, drinking-water quality, recreational water quality, ultraviolet (UV) exposure, transport, indoor environment, hazardous substances, climate change, border health, and population vulnerability. The indicators are updated annually where possible, and the indicators, factsheets, metadata, and background information are published on the Environmental Health Indicators website (www.ehinz.ac.nz
), with additional supporting data about environmental health and the health status of New Zealanders published on an online atlas (www.healthspace.ac.nz
For this project, we focused on developing indicators for the indoor environment. The indoor environment can affect health in a number of ways, including household crowding, second-hand smoke exposure, cold, damp, and mouldy housing, unflued gas heaters and open fires, safety hazards, lead-based paint flakes, asbestos, and pest infestations [11
]. In New Zealand, a key issue is housing availability and affordability, with the recent lack of housing availability and affordability leading to reported overcrowding and homelessness [12
]. Additionally, many houses (particularly older houses) have historically been cold and damp, in part due to lack of or poor insulation [13
Initially, our indoor environment domain only included indicators on household crowding, second-hand smoke exposure, and lack of home heating. A wide range of health outcomes had been identified as being linked to these exposures, but an indicator development process was required to develop a full set of indicators of environmental exposures and related health outcomes.
5.1. The Importance of Using a ‘Public Health Impact’ Approach to Develop Environmental Health Indicators
This project has demonstrated the importance and value of using a public health impact approach when developing environmental health indicators and has provided a framework and methodology that can be used for other topics and in other countries to prioritise and select environmental health indicators. Using public health impact criteria and an attributable burden of disease approach during the indicator selection process ensures that environmental health indicators are based on robust, scientific evidence about the health burden caused by the environmental exposure and therefore focus on the most important health needs of the country. A focus on policy relevance, vulnerable populations, and the number of people affected ensures that indicators are relevant and useful to decisionmakers and policymakers, who can use policies, programmes, and projects to address these environmental health issues.
This work also demonstrates how prevalence data about exposure in the population can be used to estimate the proportion of health burden attributable to an environmental exposure, through the population attributable fraction (PAF). In this paper, we demonstrated this method by estimating the PAF for children’s asthma caused by damp and mouldy housing in New Zealand at 15%. Calculating the PAF allows the development of ‘attributable burden indicators’, which have been suggested as the most robust form of environmental health indicators [Briggs, personal communication], as they link specifically between the environmental exposure and the related health effects.
Using a public health impact criterion and sub-criteria extends previously published processes for developing environmental health indicators [2
]. Researchers and developers of indicators may have intuitively used a similar process in the past for prioritising environmental health indicators. For example, a WHO working group screened potential indicators based on their ‘policy relevance, health relevance and potential data availability’ [43
]. The public health impact criteria are also similar to criteria suggested in the children’s environmental health indicators (‘relevant to an issue of policy or practical concern’ and ‘actionable’) [2
]. However, our approach goes further than these criteria, as it includes an assessment of the health impact from the environmental exposure for the country of interest and gives researchers a solid and systematic framework for carrying out an assessment. Having a set of specific public health impact criteria, against which to assess indicators, moves decision making from intuition to a transparent and informed decision.
We have also highlighted some modifications to the WHO approach for developing environmental health indicators, which are useful when developing indicators for a specific country. Most importantly, it is beneficial to identify data sources before applying the indicator selection criteria, rather than in the design stage, to ensure that the proposed indicators meet selection criteria such as data availability, timeliness, consistency, comparability, and methodologically-sound measurement. Additionally, identifying causal relationships in the selection stage ensures that the conceptual framework is based on robust evidence and, furthermore, that both health and exposure indicators monitor outcomes for the most important age groups. A further addition to the WHO approach was to identify vulnerable population groups in the scoping stage and to consider these groups in both the selection and design stages of indicator development.
Overall, we found that using a public health impact approach was valuable for identifying a ‘core set’ of indicators for regular monitoring. While our approach has only been applied to health indicators at this stage, the approach could also be adapted for exposure variables as well. We also identified attributable burden indicators for potential development: the burden of disease attributable to second-hand smoke exposure, household crowding, and damp and mouldy housing. These indicators may require more work than other indicators and so would be updated less regularly than the core indicators. However, the attributable burden indicators are beneficial for informing action, as they estimate the potential health impacts that could be gained from removing the environmental exposure. For example, eliminating household crowding in New Zealand would prevent an estimated 1300 hospitalisations for infectious diseases each year. Thus, attributable burden indicators show how improving a specific aspect of the environment would have substantial health benefits.
5.2. Findings from the Indoor Environment Indicators
Our indoor environment indicators showed that the indoor environment has a sizeable impact on health in New Zealand. In particular, damp and mouldy housing has a relatively large burden on children’s health in New Zealand, accounting for an approximate 500 hospitalisations for asthma each year. Future work and upcoming Census data releases will help to refine this estimate, as well as provide a better estimate of the percentage of children living in damp and mouldy houses and the population attributable fraction. Household crowding affected almost one in six (16%) children in New Zealand in 2013, increasing their risk of infectious diseases; the recent housing affordability issues in New Zealand [12
] may well have impacted on this figure, and upcoming data from the 2018 Census may shed more light on this issue. By comparison, second-hand smoke exposure affected 5% of children and 3.7% of non-smoking adults in New Zealand; nonetheless, the health impacts from second-hand smoke exposure were significant, particularly for babies exposed to second-hand smoke from their mother in their first year, which increases the risk of SUDI.
Focusing on vulnerable population groups was an important part of this indicator development process. Potentially vulnerable populations were identified at the scoping stage and were considered as part of the public health impact assessment. This information then informed how indicator results were output to ensure monitoring of high priority population groups. In particular, our indicators showed that some ethnic groups were disproportionately affected by the indoor environment in New Zealand. For example, household crowding affected 43% of Pacific children and 25% of Māori children in 2013. These high exposure rates were mirrored by higher rates of lower respiratory tract infection hospitalisations and meningococcal disease notifications for children in these ethnic groups. This example demonstrates the power of having several health effect indicators relating to one environmental exposure to allow triangulation of health impacts. Additionally, focusing on vulnerable population groups provides evidence to inform targeted actions to help reduce inequalities.
5.3. Limitations of Our Approach
A limitation of this study was the comparison of published attributable burdens using different measurement units (attributable deaths, hospitalisations, and DALYs). DALYs tend to be the most helpful measurement, as they combine both fatal outcomes (years of life lost to premature death) and non-fatal outcomes (years of life lived in disability or ill-health). While it is not ideal to compare across exposures using different measurements, we were able to use a consistent measurement within each individual exposure, which allowed the most important health effects for each exposure to be identified. Additionally, the attributable burden data were only used as a guide for prioritising indicators and using the published statistics was a pragmatic decision given limited resources. A helpful next step could be converting the various attributable burdens into the same measurement, for example, by applying PAFs to country-specific DALYs sourced from the Global Burden of Disease Study (http://ghdx.healthdata.org/
) or a similar national study. The Global Burden of Disease Study is also a useful data source for deaths and DALYs at the country level when other data sources or research results are unavailable.
An additional limitation was the lack of published PAFs for the injury indicator, which meant we were unable to estimate the proportion of injuries attributable to an unsafe indoor environment. The total number of injuries occurring in the home for 0–4-year-olds was used instead in the public health impact assessment; while this may be an over-estimate of the attributable burden, it still provided useful information about the total injury burden occurring in the indoor environment to aid indicator selection.
One of the drawbacks of the prioritisation process was that the final core set of indicators did not include some health conditions with large attributable burdens, for example, ischaemic heart disease and stroke, which mainly affect older adults. These indicators were ruled out based on the indicator selection criteria, including difficulties in monitoring these conditions in non-smokers, lag times of 1–5 years between exposure and health impact, and a small PAF of 2%. In particular, a small PAF indicates that only a small fraction of the overall health burden is attributable to the environmental exposure, resulting in the potential indicator being less sensitive to change and therefore less valid. Nonetheless, these health conditions, and other potential indicators identified through the selection process, could be considered worthy of monitoring. There are two potential ways of addressing these limitations. Firstly, having attributable burden indicators ensure that all health conditions and age groups are included in the assessment of health impact, as part of the environmental burden of disease methodology. Secondly, a set of supplementary indicators can be developed as part of regular monitoring. This would allow lower priority health indicators to still be monitored, although not to the same level of detail or with the same amount of commentary as the core set of indicators. These supplementary indicators would also be informative for other areas of environmental health monitoring, such as outdoor air quality. For the indoor environment, potential supplementary indicators could include ischaemic heart disease, stroke, lung cancer, gastroenteritis hospitalisations, otitis media hospitalisations (acute admissions, and waiting list for grommet insertion), bronchiolitis from respiratory syncytial virus (RSV) infection, hospitalisations for sequelae of H. pylori infection, and tuberculosis notifications. Additionally, health indicators with Level 2 evidence of causality (such as upper respiratory tract infections) could also be included as supplementary indicators if desired. These supplementary indicators could also be output by different population groups to enable monitoring of vulnerable population groups (such as children and older adults).
More generally, one limitation of using a public health impact approach is that it does not provide the answer to what indicators to select. However, this approach still helps inform the final decision and provides a set of criteria to guide the decision-making process. In particular, this approach allows the indicator developers to communicate and discuss the merits of each indicator and moves the decision-making to a transparent and more systematic process.
Our assessment of indoor environment indicators for New Zealand was limited to four main exposures but could have included a wide range of additional exposures. These include indoor cold (separate from dampness and mould), traffic noise, and exposure to substances, such as lead, carbon monoxide, and formaldehyde [11
]. Other aspects of the indoor environment, such as indoor radon and indoor cooking on open wood or coal fires, have little relevance in New Zealand and are not considered national environmental health issues of concern.
This study has demonstrated a novel and valuable approach for prioritising and selecting environmental health indicators based on their ‘public health impact’. The public health impact of potential indicators was assessed based on five sub-criteria: the number of people affected (based on environmental burden of disease statistics); severity of health impact; whether vulnerable populations were affected and/or large population inequalities were apparent; whether the indicators relate to multiple environmental exposures; and relevance to policy. The approach can be used when numerous possible health indicators have been identified and need to be reduced to a ‘core’ set of indicators. In this way, this approach fills a gap by describing how to move from the results of applying indicator selection criteria to a final list of indicators. Moreover, we propose some modifications to the indicator development method described in the children’s environmental health indicators to tailor the process for a specific country.
After following the indicator development process, the following core indicators were selected to monitor the indoor environment in New Zealand: living in crowded households, exposure to second-hand smoke, maternal smoking at two weeks after birth, asthma prevalence, asthma hospitalisations, lower respiratory tract infection hospitalisations, meningococcal disease notifications, and sudden unexpected death in infancy (SUDI). Additionally, indicators on living in damp and mouldy housing and children’s injuries in the home were identified for further development, subject to data availability and quality. Furthermore, we identified indicators of the attributable health burden due to household crowding, second-hand smoke exposure, and damp and mouldy housing.
These indicators, selected using a public health impact approach, showed that housing quality and availability has a sizeable impact on the health of New Zealanders, particularly for children and Māori and Pacific peoples. Addressing these environmental issues would have substantial health benefits for New Zealand and reduce health inequalities for some population groups. Using public health impact criteria and an environmental burden of disease approach proved valuable in identifying and prioritizing the most important environmental health impacts to monitor.