Syndromic surveillance and heat wave morbidity: a pilot study based on emergency departments in France

Over the last 30 years, the impact of major heat wave shocks on population morbidity and mortality has become an urgent public health concern [1‐3]. In different countries, stakeholders and health authorities are often overwhelmed by the magnitude of both the threat to public health and the social impact of heat waves. This reflects a lack of preparation and failed organization within the healthcare system. Public awareness campaigns can be oriented to promotion of the adoption of various preventive measures, such as the creation of air-conditioned areas in nursing homes and active hydration programs for high-risk groups such as the elderly or young children. Nevertheless, the first step in preventing the drastic consequences of such climatic events as heat waves is to implement systems to provide early warning based on prompt detection of excess morbidity and mortality [4]. Such an early-warning surveillance system should examine data in real time and it should be highly sensitive and cost-effective.

In August 2003, France suffered an extreme heat wave that had major public health consequences. Overall 15,000 extra deaths during a 16-day period were recorded, corresponding to a 60% increase in expected mortality [5, 6]. Since summer 2004, the French Institute for Public Health Surveillance, in close cooperation with Météo France^® defined and implemented a heat health watch warning system on the basis of biometerological indicators. The warning system operates from 1^st June to 31^st August (level 1, seasonal surveillance period). When the alert criteria are fulfilled, an awareness and action level is declared by the Préfet (level 2) who manages the département (French administrative unit). A third level (maximum mobilization) is implemented if the impacts of heat wave overwhelm the health field: power cuts, drought, management problems in the funeral centres and heavy air pollution.

The alert system aims to give the public authorities 3 days' prior warning that a heat wave may occur, in order for the National Heat Wave Plan (NHWP) measures to be put into operation. The preventive measures are aimed at modifying the behaviour of people, health institution and health authorities with regard to high summer temperatures [7, 8].

The health impact of hot weather is measured in routine during the seasonal surveillance period (from 1^st June to 31^st August) on the basis of the rough daily activities of emergency departments in hospitals (ED) and mortality data [7].

In the summer of 2006, a new heat wave occurred that lasted approximately three weeks, and generated an excess mortality of 2,100 deaths [8, 9]. This extreme heat event has been detected by the national heat/health watch warning system which has been able to identify properly hot days on the basis of 3 day prediction temperatures [9].

Using data collected during the 2006 heat wave in France, we undertook this study to evaluate the ability of a syndromic surveillance system to detect health impacts of heat wave through variations in ED activity over time and to determine relevant indicators for this kind of surveillance in this particular weather situation.

From the 1^st of June to the 31^st of August 2006, 415,862 visits were logged in the system. Among them, 82,040 (19.7%) were recorded during ONAP, including 8,580 (10.5%) patients aged 75 years and older, and 15,996 (19.5%) children aged 15 or younger. A total of 91,802 hospitalisations were recorded, 17,384 (18.9%) of them during the ONAP including 4,991 (28.7%) involving people aged 75 or older. The daily mean number of patients per day was not significantly different between the ONAP and OFAP across both age groups (4,557.7/day vs. 4,511.2/day, respectively (Table 2)). The number of venues per day was, however, significantly different for elderly people (476.7/day vs. 446.2/day, p < 0.05). Similar results were observed for the daily mean number of hospitalisations: hospitalisation outcomes including both age groups were 1,011.6/day during ONAP vs. 994.6/day during OFAP (NS); for elderly people: 257.6/day vs. 227.7/day, p < 0.05). The ratio of males to females during ONAP across both age groups was 1.2, and this number did not change significantly between ONAP and OFAP. The daily percentage of diagnosis miscoded or uncoded was not significantly different between ONAP and OFAP (23.7% vs. 24.2% respectively)

Patients presented with malaises, dehydration, hyperthermia, hyponatremia, renal colic, or renal failure, were observed at significantly higher frequency during ONAP than OFAP (Table 2). In contrast, neither the number of deaths recorded in ED nor the number of patients diagnosed with respiratory nor cerebrovascular diseases differed significantly between the two time periods (Table 2). Therefore, the group of HWDS was composed of the following diseases: malaises, dehydration, hyperthermia, hyponatremia, renal colic and renal failure (Table 2 and Table 1 for ICD-10 codes). Figure 1 displays the indicator derived from the HWDS, the national average temperature observed in France during ONAP and OFAP, and the NHWP. The figure shows that the indicator increases one day after every increase in temperature.

ED visits and hospitalisations increased significantly for the elderly during ONAP, relative to OFAP (10.5% vs. 9.9% and 28.7% vs. 25.5%, respectively, p-value for both differences < 0.001; Table 3). Higher proportions of the elderly were admitted to the ED or were hospitalized during ONAP compared to OFAP in relation to a HWDS (27.7% vs. 23.8% and 58.0% vs. 50.1%, respectively, p-value for both differences p < 0.001; Table 3).

As shown in Table 4, the proportion of visits linked to HWDS during the heat wave increased significantly for elderly people: the proportion of total visits linked to HWDS was 12.43% during ONAP compared to 8.15% during OFAP (p < 0.001).

Three diseases increased dramatically during ONAP for this specific age group: hyperthermia (11.5-fold increase), hyponatremia (2.8-fold), and dehydration (3.0-fold). For younger adults, the proportion of total visits related to HWDS was 3.78% during ONAP compared to 2.87% during OFAP (p < 0.001). Only diagnoses presented in Table 4 increased significantly between the two time periods.

This study documents the first use of syndromic surveillance based on hospital emergency department data to observe and analyze in detail the public health burden of a major heat wave. The results show that a limited number of indicators are possibly useful for daily monitoring of the health impact of a heat wave on a population using syndromic surveillance. The results also show that the elderly are more severely affected by hot weather than young people. This system provides qualitative and quantitative elements for understanding the immediate public health danger and to evaluate it in real time. The parameters monitored by this system of environmental health surveillance are crucial components because weather modifications require permanent state of adaptability of public health responses, as shown in Figure 1[19, 20].

During the OFAP we observed a correlation between the HWDS indicator and temperature peaks in mid June and the beginning of July. But during the ONAP the health effects of the heat wave only appear 3–4 days after the heat wave has begun. This lag may be due to physiological adaptation and/or preventive measures stipulated by the national heat wave plan and activated during the heat wave [7]. The non increase in our indicator during the first days of the "official" beginning of the heat wave made possible such interpretation. This hypothesis should be confirmed by new studies conducted in similar context: a heat wave occurring while a heat wave plan is activated.

During the 'on alert' period, the percentage of patients admitted to hospital emergency departments who were elderly was three times greater than the percentage who were younger adults. Nevertheless, the number of ED admissions across the nation during the heat wave did not change significantly from normal levels. This finding is consistent with other studies, which suggests the necessity of monitoring ED activities for different age groups and of collecting medical diagnosis in real time [21]. It is known that cardiovascular mortality increased during a heat wave [2, 22]. Also in agreement with other findings, our results show that hot weather does not affect cardiovascular morbidity [14, 23]. This point is discussed in the literature and an impact on cardiovascular morbidity has been shown by others in case of extreme temperatures [24]. In our study, we did not find any significant increase in several other diseases frequently associated with hot weather, such as respiratory diseases, asthma (despite that several cities announced air pollution alerts during the heat wave), urinary infections, hypoglycaemia, or cerebrovascular diseases [1, 14].

In terms of number of visits and hospitalisations, elderly patients (≥ 75 years) were the only group to show a significant increase during the 'on alert' period compared to the 'off alert' period. As a result, we think that effort to monitor the health burden of a heat wave should focus on elderly people and on the following diseases in that age group: malaise, hyperthermia, hyponatremia, and dehydration. An increase in one of these parameters, in case of hot weather, is likely to indicate the onset of health effects of a high temperature and it therefore could constitute a signal to initiate a rapid public health response.

This study used several ICD-10 codes to define syndromes relevant to detecting the health impacts of heat waves. This approach was motivated by the concern that the codes are too fine-grained for detection of syndromes related to heat waves. Development of the nationwide surveillance network required accurate coding, inter-coder reliability, and minimal variation in coding over time. By clustering codes, we hope to include all codings that may conceivably be applied to a patient exhibiting some relatively common syndromes related to high temperature [25].

Our study is limited in several ways. The geographical coverage of the network during the study was limited to large cities and so is not entirely representative of the French population. Because most heat effects are more significant in cities with their own "island heat effect areas", similar effect might be expected in other cities. For other areas (i.e., rural area), the impact of heat wave on population could be different and complementary analysis should be necessary. Since the complete automation of data transmission is crucial, not all emergency departments record all of the information per visit which creates an information deficit in the health surveillance and alert system. Given their pivotal role in public health surveillance, ED health care professionals should be trained in collecting and validating data in ways that maximize the quality of surveillance. Our results are supported by the analysis of a single heat wave and should be confirmed by other studies in the same weather conditions. An over estimation of our findings is possible due to over-diagnosis by doctors aware that a heat wave has been declared. In such situation ED physicians are more attentive to heat effects and may code first heat related syndromes. This effect should be taken in account and might explain a part of the increase in some diagnoses.

It is now well accepted that ED activity reflects the overall state of health of a population [26]. ED activity should therefore be considered as a public health 'sentinel' indicator, and health surveillance of EDs should be used to design appropriate public health interventions [27]. Because this method of surveillance uses routinely generated data and does not require the health care professionals to enter additional information, there would be minimal disruption to the work patterns of the data providers [28].

Until now in case of heat wave, weather forecast allowed understanding what would happen in the following days, the health impact of a heat was estimated on the basis of the number of deaths observed and the rough number of daily ED visits for the main city of each French département. This system, made possible the follow up by age groups targeted by preventive measures (elderly) or not (younger) and their adaptation in real time if necessary. It gives to public health authorities and stakeholders information near real time about the possible impact of temperature on population. In this circumstance, syndromic surveillance based on EDs, which is already accepted as a validated tool for public health surveillance in the field of infectious diseases [29], might occupy a key position to propose extending syndromic surveillance to managing environmental health concerns [27]. If our results are validated by new studies, applying syndromic surveillance to environmental health concerns may be considered by the public health community. Further studies are now necessary to evaluate the efficacy of this type of surveillance in preventing the public health effects of heat waves.