In this section, we present the methodological framework to assess the annual flows of benefits and costs of a HWS, combining information on Tcrit and projected maximum daily temperatures Tmaxt over the period 2020–2040. We apply the framework to the city of Madrid.
HWS benefits from avoided mortality
The assessment is conducted as follows: (a) estimation of the preventable mortality, (b) choice of the monetary metric and adjustment of economic values from literature to study site, and (c) projections of flows of benefits over time.
Estimation of preventable mortality
The economic benefits of heat alert systems can be valued on the basis of potentially avoided mortality impacts (AM), in terms of the percentage of preventable deaths or years of life lost, given the expected effectiveness (E) of the HWS:
$$ {AM}_t={M}_t\bullet E $$
(7)
Given the absence of studies evaluating the effectiveness of HWS specifically for Madrid, we refer to Fouillet et al. (2008) which is one of the most comprehensive studies in this respect in Europe. They evaluated the effectiveness of the alert system and prevention plan adopted in France in 2006 by comparing observed excess deaths with those that would have been expected to occur in the absence of preventive measures. The evaluated plan comprises a wide set of measures in line with those typically introduced in Europe and North America (McGregor et al. 2015), ranging from the setup of a warning system to real-time surveillance of health data, and emergency plans for vulnerable people with visits and care offer. Following the Fouillet et al. (2008) analysis, we assume E varies between 60 and 78%, with a central value of 65%. The range is in line with the findings of Ebi et al. (2004) and Benmarhnia et al. (2016) for North American cities.
Choice of monetary metric to value mortality
For mortality valuation, two approaches have been used in the literature and by governmental agencies (European Commission 1999): the VSL and the VOLY. Both values provide a monetary metric to estimate the benefits of measures that deliver a risk reduction on health (OECD 2001). The VSL is based on the individual willingness to pay (WTP) for small reductions in the risk of dying. The VOLY is derived from the willingness to pay for increasing life expectancy by one additional year.
Although decision making in public health continuously involves comparing incremental costs with expected benefits, the monetization of mortality has been criticized in the literature on the basis that human life is inestimable and cannot be priced (Ackerman and Heinzerling 2004; Patokos 2010). However, even when explicit values are not provided, mortality risk reductions will receive an implicit valuation through the decision to either implement or not to implement a particular policy. Using unambiguous values of VSL or VOLY is therefore preferred to support more transparent and consistent decision-making process (OECD 2011).
In the context of air pollution, OECD (2008, 2011) recommends using VSL as a general rule. European Commission (1999), the Clean Air for Europe (CAFE) program (Hurley et al. 2005; Watkiss et al. 2005), and the APHEKOM project (Chanel 2011) suggest VOLY as the preferred approach for short-term effects on mortality from air pollution. Hurley et al. (2005) have argued that assigning a full statistical life for short-term exposure to air pollution with consequential minor changes in life expectancy might be misleading.
The VSL has been commonly used to assess the mortality impact of heat waves in previous studies leading to very high benefits of heat warning systems (Ebi et al. 2004; EPA 2015). However, the same debate on air pollution applies in the context of heat waves, where existent studies have estimated a few days to about 1-month change in life expectancy for displaced mortality (Saha et al. 2014; Hajat et al. 2005). So, in the context of displaced deaths, heat stress is a contributing cause of death, but not the main underlying mortality risk factor. Therefore, assigning a full statistical life for such a short reduction of life expectancy might exaggerate the economic impact or conversely exaggerate the benefits of an averted heat death. To shed further insights on heat wave mortality valuation, we propose to compare three approaches (Table 2). Option 1 is the traditional approach where VSL is applied to both displaced and premature mortality. Option 2 is an intermediate view, in which we use VSL for premature deaths, and VOLY for displaced mortality, considering the appropriate loss of life. Option 3 is a conservative approach where both premature and displaced deaths are calculated in terms of loss of life expectancy and valued as VOLYs.
Table 2 Monetary approaches used for mortality valuation Table 3 summarizes the mortality costs selected for this study. For VSL, we rely on the global meta-analysis of stated preference studies conducted by Lindheim et al. (2011) in the context of environmental, health, and transport policies. For EU27, the recommended VSL is 3.6 million dollars (at 2005 prices).
Table 3 Valuation of heat-related mortality in year 2020 (base costs €2013) For displaced mortality, we have taken the air pollution VOLY estimated by Chilton et al. (2004) as a proxy for “displaced VOLY” in the heat wave context, given the lack of specific valuation studies. Using mortality valuation estimates from air pollution context can be a reasonable first approximation given that health outcomes (cardiovascular and respiratory diseases) are similar across the two risk factors (air pollution and heat waves). The estimate by Chilton et al. (2004) was determined using a WTP survey conducted in the UK to estimate the monetary equivalent of a small change in life expectancy under the context of short-term mortality associated with air pollution exposure. More specifically, this corresponds to a gain of 1 month in life expectancy for someone currently in poor health. The study of Chilton et al. (2004) revealed that people attributed a lower value to a gain in life expectancy when faced with poor health compared to the preference of a healthy individual. For premature deaths, as a proxy for “premature VOLY” in the heat wave context, we use the value of de Ayala and Spadaro (2014), who recommend an estimate for the EU-27 considering the available literature on long-term mortality in the context of air pollution.
The reference values obtained from the literature (Table 3) have been adjusted to Euro 2013 accounting for income differences, as well as for changes in the cost of living over time. For this, we follow the recommendations of OECD (2011) and economic indicators from the World Bank (http://data.worldbank.org/indicator): (i) conversion to the national currency using the Purchasing Power Parity (PPP) adjusted exchange rate, (ii) adjustment to current prices using the Consumer Price Index (CPI), and (iii) adjustment for variation of real income in time and space using the PPP-adjusted gross domestic product (GDP) per capita. The reference values have been further adjusted for GDP per capita growth from the baseline year 2013 to year 2020.
Projections of benefits over time
As the mortality benefits occur over time, economic values have been adjusted to reflect the real purchase power of the population in future years according to SSP2 and SSP5 projections. The total benefits are then obtained by summing the annual flow of benefits:
$$ TB=\sum \limits_t\frac{AM_t\bullet {V}_t}{{\left(1+d\right)}^t} $$
(8)
TB is the present value of the total benefits for the chosen time period 2020–2040 (t = 0 to 20); AMt are the preventable deaths or YLL in each year t; Vt = [Vt − 1 ∙ (1 + βggdp)] is the VSL or VOLY adjusted for variation of real income in time, where V0 is the value estimated for Spain in Table 3 for t = 0. ggdp is the growth rate of the PPP-adjusted GDP per capita for Spain according to the SSPs scenarios, β = 1 is the income elasticity over time (OECD 2011), and d is the discount rate. In the case of HWS, where costs and benefits are occurring at the same time, the impact of discounting on the benefit-cost ratio is not significant, so that we take d = 0.
HWS operational costs
The assessment is conducted as follows: (a) type of actions included, (b) choice of the monetary values, and (c) projections of flows of costs over time.
Type of actions
A heat warning system must provide advice on protective behavior to the general public, as well as to targeted vulnerable groups (the elderly, children, people with serious chronic illnesses and mobility problems, low-income groups, outdoor workers) (PHE 2013). It is active during a specified operational timeframe which depends on the period when heat-related health burdens are expected to increase anomalously. These systems use information on weather forecasting and threshold temperatures to release watch alerts and warning messages. Specific operational conditions and actions foreseen to reduce health risks vary according to the geographical location. A HWS is usually part of a wider Heat-Health Action Plan (HHAP) (McGregor et al. 2015), which comprises a number of actions at the community level, when the pre-identified alert threshold is reached. These are public health actions which, in Europe and the USA, generally include media announcements to the general public that suggest behavioral recommendations, a bulletin or web page directed to professionals or for the overall population, leaflets for home care managers, dedicated telephone heat lines, and alerts going out to hospital emergency services and health centers, among others (McGregor et al. 2015). Following the 2003 heat wave, the Health Department in Madrid has put in place a warning system whose actions are aligned with those of other countries in Europe and the USA (Comunidad de Madrid 2015).
Choice of monetary values
The cost assessment within a benefit-cost analysis of any intervention in public health usually involves a number of steps, including the identification of specific actions, the categorization of costs, and the quantification of personnel and time required as well as projections and discounting (Hutton and Rehfuess 2006). In practical terms, difficulties arise when estimating the costs of HWS in specific urban contexts due to the lack of a specific dedicated budget for planning, limited accessibility of quantitative information on personnel, and time dedication per type of action from implementing agencies, as well as lack of studies providing a transparent and comprehensive analysis of resource costs. Given these limitations, a basic procedure has been followed for the Madrid case study that relies on information from Ebi et al. (2004). We assume a lower bound estimate of direct costs per day, referring to basic activities such as the additional labor required to maintain a heat line and emergency medical services, and an upper bound for a wider set of actions, which additionally include dissemination campaigns, media announcements, and alerts to nursing homes and other facilities that provide extra care to vulnerable people through community outreach programs. The estimates in Table 4 are adjusted to year 2013 for national currency, changes in real income per capita (in time and space), and population size at risk, and further corrected for GDP per capita and population growth from the baseline year 2013 to year 2020.
Table 4 Projected operational costs of HWS in year 2020 (base costs €2013) Projections of costs over time
Annual costs are projected to 2040 using PPP-adjusted GDP per capita to reflect increased labor wages over time in future scenarios (SSP2 and SSP5), population growth, and taking into account the number of days in which the HWS is activated, when the maximum daily temperature is above Tcrit. Income elasticity is applied in the same way as for the valuation of benefits in Section 4.1. Annual costs are summed to obtain the total cost:
$$ TC=\sum \limits_t\frac{Ndays_t\bullet {DC}_t}{{\left(1+d\right)}^t} $$
(9)
TC is the present value of the yearly flow of costs for the period 2020–2040 with t = 0 to 20 years; Ndayst is the number of days in which the maximum daily temperature is above Tcrit in year t; the operational costs per heat wave day are DCt = [DCt − 1 ∙ (1 + βggdp + gpop)] and DC0 are the costs reported in Table 4 for t = 0. ggdp and gpop are, respectively, the growth rate of the PPP-adjusted GDP per capita and the growth rate of the population for Spain according to the SSPs scenarios, β = 1 is the income elasticity over time, and d the discount rate (where we have chosen d = 0).