**Contents**



#### vi


## **About the Editors**

#### **Emily Ying Yang Chan**

Emily Y. Y. Chan, MD, SM (Harvard), FFPH (UK) serves as a Professor and Assistant Dean at the Faculty of Medicine, Director of CCOUC and Centre for Global Health (CGH), Chinese University of Hong Kong (CUHK). She is Co-chair of the WHO Thematic Platform for Health Emergency and Disaster Risk Management (Health-EDRM) Research Network and WHO COVID-19 Research Roadmap Social Science working group, a member of the World Meteorological Organization SARS-CoV-2/COVID-19 Task Team and Asia Pacific Science Technology & Academia Advisory Group of the United Nations Office for Disaster Risk Reduction, a Visiting Professor at Oxford University Nuffield Department of Medicine, a Fellow at Harvard University FXB Center, and CEO of the GX Foundation. Her research interests cover disasters and humanitarian medicine, climate change and health, global and planetary health, human health security and Health-EDRM, remote rural health, implementation and translational science, ethnic minority health, injury and violence epidemiology, and primary care.

#### **Holly Ching Yu Lam**

Holly Lam (PhD) is an environmental epidemiologist. Having completed her PhD at the Jockey Club School of Public Health and Primary Care, Chinese University of Hong Kong (CUHK), investigating associations between meteorological factors and respiratory hospital admissions in Hong Kong, she worked on environmental health studies in the school and joined the Collaborating Centre for Oxford University and CUHK for Disaster and Medical Humanitarian Response (CCOUC) to work on research projects quantifying exposure–health–outcome associations, assessing risk perceptions and analyzing knowledge–attitude–practice for health-risk reductions in emergencies/disasters. She now works as a Medical Research Council (MRC) Early Career research fellow at the National Heart and Lung Institute, Imperial College London, focusing on environmental exposures and allergic condition associations. Her research interests cover assessing the health effects of ambient exposures and climate change and identifying related prevention strategies.

### *Editorial* **Research in Health-Emergency and Disaster Risk Management and Its Potential Implications in the Post COVID-19 World**

**Emily Ying Yang Chan 1,2,\* and Holly Ching Yu Lam <sup>3</sup>**


Health-Emergency Disaster Risk Management (Health-EDRM) is one of the latest academic and global policy paradigms that capture knowledge, research and policy shift from response to preparedness and health risk management in non-emergency times [1]. This concept encompasses risk analyses and interventions, such as accessible early warning systems, timely deployment of relief workers, provision of suitable drugs, and medical equipment to decrease the impact of disasters on people before, during, and after an event(s). The approach emphasizes the investment into disaster health risk reduction efforts which may thereby strengthening health systems and capacity to ensure community health resilience building. Health emergency disaster risk management (Health-EDRM) thus refers to the systematic analysis and management of health risks surrounding emergencies and disasters, and plays an important role in reducing hazards and vulnerability along with extending preparedness, response, and recovery measures [1].

Disasters such as earthquakes, cyclones, floods, heat waves, nuclear accidents, and large-scale pollution incidents cost human lives and incur long-term health and well-being implications. The most vulnerable population subgroups in the majority of the disasters often comprise of extreme ages, remote living areas, and endemic poverty, as well as people with low literacy. However, scientific evidence gaps remain in the published literature regarding health risks patterns and cost-effectiveness Health-EDRM risk reduction strategies to facilitate a more efficient reduction of global disaster risks through global policies and initiatives [2,3]. The first Special Issue of *IJERPH*, published in 2018–2019 with the thematic focus on Health-EDRM included 20 papers that characterized disaster risks, analysed health risk and interventions effectiveness [4]. The 2nd edition (2019–2020) further compiles 16 scientific papers that have been published in 2020. Papers included in this 2nd edition demonstrate the diverse range of health-related disaster and emergency risk management topics and research analyses that evaluate short- and long-term health impacts, associated risk factors, risk assessment methods and tools as well as multidisciplinary research methods related to program evaluation and policy analysis.

With the complexity and interconnectedness of modern living, multidisciplinary research methodologies development is crucial, as it may enhance the assessment of health risks, impacts and promote understanding of health risks, disaster and humanitarian medicine for non-health stakeholders. The study of Zhang et al. [5] shows how multidisciplinary methodologies might be applied to examine communicable disease spread among close-contacts. Ma et al. [6] proposes a landslide risk prediction approach that might be useful to prevent mortality and morbidity in at-risk communities.

In the 21st century, the human health impacts of climate change are expected to be significant globally. In this issue, five studies are published to delineate population health risks that may be associated with climate change in various contexts. The studies examine socio-demographic patterns of self-help and community bottom-up health

**Citation:** Chan, E.Y.Y.; Lam, H.C.Y. Research in Health-Emergency and Disaster Risk Management and Its Potential Implications in the Post COVID-19 World. *Int. J. Environ. Res. Public Health* **2021**, *18*, 2520. https:// doi.org/10.3390/ijerph18052520

Received: 25 February 2021 Accepted: 2 March 2021 Published: 4 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

protection strategies during extreme temperature events [7,8] (Lam et al., Liu et al.) and typhoon/hurricane [9,10] (Shang et al., Shih et al.). In addition, with the global increasing health risks associated with vector-borne diseases as a result of climate abnormalities, [11] Chan et al. present a narrative review paper of the current understanding of primary preventive Health-EDRM measures that might reduce the health risks of vector-borne disease in communities.

Another research article subset published in this edition is related to the evaluation of human health and well-being with man-made/technology-related disasters. In their case study, Genereux et al. [12] analyze how mixed-method-based need assessment might capture potential health risks and needs and subsequently enhance effectiveness and relevance response and recovery of the 2013 Lac-Megantic Rail disaster. The study of Orui et al. [13] finds an association between media information and post nuclear accident health anxiety. The findings of this study may inform policymakers and clinical practitioners about mental and health risk reduction strategies. Based on a mixed-method approach, Lorenzoni et al. [14] examines various long-term implications of disasters on public health system performance, security and health protection.

Psychosocial and mental health risks and impacts are another important research development area of the Health-EDRM. Takahashi et al. and Umeda et al. have examined the acute mental health needs [15] of disaster victims and potentially how to protect and promote the mental risks of responders [16]. Newnham et al. [17] described the activities and strategic plans of a newly established mental health network in Asia Pacific that aims to encourage and build the research agenda of Health-EDRM-related issues in the region.

Among all the major global disasters in 2020, SARS-CoV2, the biological hazard which caused the COVID-19 pandemic, once again showed how health risks may be far-reaching to cause mortality and affect lives in the 21st century. Within a year, the pandemic has accumulated over 110 million cases and 2.5 million deaths [18]. Meanwhile, research and reported experiences in Asia, as the first global region that was hit by the pandemic in early 2020, might be useful to facilitate understanding of health risks and impacts of a new disease of unknown origin. Kim et al. reports on the COVID-19 impact on mental health status of people living in Daegu, South Korea, the community with the 2nd highest rate of COVID-19 beyond Wuhan city PRC China, during the early phase of the pandemic [19]. Chan et al. [20] examines the sociodemographic predictors of health risk perception, attitudes and behavioral practices of the management COVID-19 in a high-density metropolis—HK, China during the first 2 months of the pandemic (6). The same research team also published an article that examines the health risks and situation of people with non-communicable diseases during the pandemic [21].

In the upcoming months, the global research community is expected to receive a significant amount of Health-EDRM research outputs related to the COVID-19 pandemic. With heavy emphasizes on the hierarchy of prevention and adverse disaster risk reduction [22], researchers and policy makers of Health-EDRM should be proactive in the application of the concepts and tools highlighted in the Health-EDRM paradigm and frameworks [23]. Such efforts will improve research effectiveness for strengthening practice and policy making that aim to protect human health and well-being from future epidemics and disasters.

**Author Contributions:** Writing—Original Draft Preparation, E.Y.Y.C.; Writing—Review & Editing, H.C.Y.L. Both authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** Our special thank goes to the Editorial Office and Chi Shing Wong for their facilitation and coordination in this Special Issue in Health-EDRM.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


International Journal of *Environmental Research and Public Health*

### *Article* **Infection Spread and High-Resolution Detection of Close Contact Behaviors**

**Nan Zhang <sup>1</sup> , Boni Su <sup>2</sup> , Pak-To Chan <sup>1</sup> , Te Miao <sup>1</sup> , Peihua Wang <sup>1</sup> and Yuguo Li 1,\***


Received: 20 December 2019; Accepted: 20 February 2020; Published: 24 February 2020

**Abstract:** Knowledge of human behaviors is important for improving indoor-environment design, building-energy efficiency, and productivity, and for studies of infection spread. However, such data are lacking. In this study, we designed a device for detecting and recording, second by second, the 3D indoor positioning and head and body motions of each graduate student in an office. From more than 400 person hours of data. Students spent 92.2%, 4.1%, 2.9%, and 0.8% of their time in their own office cubicles, other office cubicles, aisles, and areas near public facilities, respectively. They spent 9.7% of time in close contact, and each student averagely had 4.0 close contacts/h. Students spent long time on close contact in the office which may lead to high infection risk. The average interpersonal distance during close contact was 0.81 m. When sitting, students preferred small relative face orientation angle. Pairs of standing students preferred a face-to-face orientation during close contact which means this pattern had a lower infection risk via close contact. Probability of close contact decreased exponentially with the increasing distance between two students' cubicles. Data on human behaviour during close contact is helpful for infection risk analysis and infection control and prevention.

**Keywords:** infection spread and control; infection risk; human behavior; close contact; sensor-based; indoor environment; indoor positioning; head and body motion; open-plan office

#### **1. Introduction**

Indoor human behaviors directly impact on indoor thermal comfort [1], energy efficiency [2], office design [3], and exposure to pollutants (e.g., infectious microbes) [4]. Indoor human behaviors in and between different environments also directly impact on the infection risk [5]. Close contacts are believed to facilitate the spread of many viral respiratory diseases such as influenza [6], SARS [7], MERS [8], and even Ebola [9].

Infection risk via close contact is influenced by interpersonal distance, respiratory activities, and movement of body parts. Interpersonal distance directly affects the risk of virus exposure due to inhalation and deposition, the so-called short-range airborne and large droplet routes, respectively [10]. A threshold distance of close contact less than 1.5 m to 2 m is generally accepted as risky [11–13]. Human respiratory activities such as breathing, talking, and coughing can generate droplets of different numbers and sizes [14–18]. Infectious pathogens are shed and exhaled by the infected during these respiratory activities, and transported by the exhaled air streams, while inhalation of fine droplets and exposure to large droplets are also affected by the inspiratory air streams and body/head/arm movement [19]. Relative face orientation (e.g., face-to-face, face-to-side) and posture are important factors in determining the cross-infection, especially over short distance [20]. Exposure of face-to-back close contact is much smaller than it of face-to-face pattern [21,22]. Posture also important in droplet

deposition, for example, droplets deposited on trousers on the thighs of a sitting person should be more than on a standing person.

Very little data exist on indoor close contact behaviors, especially data combining all of the factors mentioned above. Some human behaviors are difficult to accurately monitor with high temporal resolution [11,19]. Electronic sensors such as radiofrequency identification devices (RFIDs) are the most commonly used to collect human contact data. These devices, however, can only detect close contact by taking readings of interpersonal distance every 20 s, and only one-on-one close contact can be detected [23,24]. A temporal resolution of 20 s is not sufficient, as the median value duration of a close contact is 17 s [11]. Moreover, human respiratory activities and movement of body parts (e.g., head and body) cannot be detected by this means. Body movements impact the air flows in a room. To collect close contact data with high temporal resolution, video recordings have been used, which can be processed second-by-second [11,19]. This approach is subject to human error on the part of the video analysts, and also time-consuming.

Office are the most common form of workplace for the most of employees, with open-plan designs widely used [25,26]. In this study, we monitored and analyzed indoor human behaviors in a graduate student office using automatic devices installed on the hats and clothes of the participants. These devices overcome nearly all of the shortcomings mentioned above and automatically collect high-resolution data on indoor human behaviors. The data monitored included the indoor position, head and body motion, and posture of each individual which are important for infectious disease transmission. We collected more than 1,440,000 s of data for 49 students across two days. This study supported data of indoor human behaviors on close contact. From indoor positioning distribution, we can know that which part in the office has higher infection risk. Combining individual inhalation and exhalation patterns with head's and body's motion during close contact, high-precision quantitative risk assessment on infection spread and control could be conducted.

#### **2. Materials and Methods**

The experiments were conducted in two close-by Chinese graduate student offices on two consecutive Saturdays. The choice of two nearly identical offices was due to the need of minimising microbial interference, as the experiment also contained another part on surface microbial monitoring, which is not reported in this paper.

#### *2.1. Room Setting*

Each room was designed for a maximum of 30 students, i.e., containing 30 cubicles (Figure 1). There were 26 students (13 male and 13 female) in Room 1 on the first experiment day (day 1), and 23 students (12 male and 11 female) in Room 2 on the second experiment day (day 2). A total of 32 students participated in the experiment, and 17 of them in both experiments. Among 32 participants, 27 students were those who worked in the institute and five students were invited by some students who were in the institute. In addition, three pairs of students were romantically involved. We selected these students because they were familiar with the experimental environment, which should minimize any bias caused by environment. The length (12.2 m) and height (2.73 m) of both rooms were the same, and the layouts of the cubicles in the two rooms were symmetrical, see Figure 1.

Each room had a water dispenser and a group of lockers. Only Room 2 had a printer. Each room was divided into six cubicle regions (marked by dotted lines in Figure 1), and students in the same cubicle region were more likely to communicate with each other. All of the students were monitored by 22 video cameras (1080P) from 8:30 a.m. to 9:30 p.m. The indoor temperature during the experimental days, which controlled by central air conditionings, was between 26 and 27 degrees. There were 27 fluorescent tubes on the ceiling of the office to keep the luminance, therefore, combing with the high-resolution cameras, most human behaviors can be captured. Each camera monitored one or two office cubicles, except for two with a global view, one camera for the door, and one camera for the water dispenser.

**Figure 1.** Room settings: (**a**) Room 1; (**b**) Room 2.

#### *2.2. Detection Devices*

The sensors we developed are shown in Figure 2a. The sensors for indoor positioning and head motion were installed on the participants' hats, and those for body motion were installed on the chest of a tight shirt. An ultrawide band (UWB) radio real-time location system (RTLS) was applied to obtain the indoor positions of all participants [27,28]. The distance resolution of UWB is within 0.1 m when there is no obstruction. To ensure continuous data transmission, the UWB tag was installed on the top of each hat, and four UWB anchors were installed on the office ceiling. An inertial measurement unit (IMU), which can measure and record the position and motion of the head and body (i.e., rotation), was also installed on the top of each hat and the front of each tight shirt. A microphone was installed on the collar band of the shirt to determine when the participants were talking. To protect privacy, only the sound level was recorded. Adjustable bands on the hat and around the head were used to avoid relative movement between the hat and head. Tight clothing was worn to avoid relative movement between the shirt and body. All of the data recorded during the experiment were saved in a chip. The weight of a device on hats and shirts was 73 g, and the weight of hat with fixed accessories was 159 g. Therefore, each student worn a hat of 232 g on the head and worn a device of 73 g on the shirt. Light weight brings a much smaller impact on human behaviors.

Rotations were recorded in the form of quaternions. To obtain the absolute (relative to the ground) rotation of the head and body, all of the participants underwent calibration after wearing and before taking off the hat and the shirt. During the calibration, the participant stood still and faced the same wall while keeping his/her head and body upright for 10 s. For any IMU (head or body), the quaternions during the calibrations before and after the experiment were denoted as *qstart* and *qend*, respectively. If the difference between *qstart* and *qend* was greater than a threshold value (equivalent to 10◦ rotation), it was probable that the participant had moved the hat or shirt during the experiment, and the data were discarded. Data were also regarded as invalid if the fluctuation of rotation between 10-s calibrations was more than 5◦ . To eliminate drift errors for all sensors, the quaternions were adjusted from all valid raw data based on spherical linear quaternion interpolation (Equation (1)):

$$q = q\_{\rm raw} q\_{\rm start}^{-1} \left( q\_{\rm start} q\_{\rm end}^{-1} \right)^{\left(t - t\_{\rm start}\right) / \left(t\_{\rm end} - t\_{\rm start}\right)}\tag{1}$$

/

1 1

#### *Int. J. Environ. Res. Public Health* **2020**, *17*, 1445

where *t* is time, *tstart* and *tend* are the time of the calibrations before and after the experiment. The steering vectors (⇀ x*h* , ⇀ *yh* , ⇀ *zh* for head and ⇀ *xb* , ⇀ *yb* , ⇀ *zb* for body, relative to the ground) can be obtained from the quaternions (Figure 2b). Usually, the horizontal rotation of the head relative to the body direction is more important. The relative steering vectors of head to body ( ⇀ *xhb*, ⇀ *<sup>y</sup>hb*, and ⇀ *zhb*) can be calculated using coordinate transformation (Equation (2)): x 0 

$$
\begin{bmatrix}
\overleftarrow{\boldsymbol{x}\_{hb}} & \overleftarrow{\boldsymbol{y}\_{hb}} & \overleftarrow{\boldsymbol{z}\_{hb}}
\end{bmatrix} = 
\begin{bmatrix}
\overleftarrow{\boldsymbol{x}\_{b}} & \overleftarrow{\boldsymbol{y}\_{\boldsymbol{x}\_{b}}} & \boldsymbol{0} \\
\overleftarrow{\boldsymbol{x}\_{\widetilde{y}\_{b}}} & \boldsymbol{y}\_{\widetilde{y}\_{b}} & \boldsymbol{0} \\
\boldsymbol{0} & \boldsymbol{0} & 1
\end{bmatrix}
\begin{bmatrix}
\overleftarrow{\boldsymbol{y}\_{h}} & \overleftarrow{\boldsymbol{y}\_{h}} & \overleftarrow{\boldsymbol{z}\_{h}}
\end{bmatrix} \tag{2}
$$

where *x*⇀ *xb* and *y*⇀ *xb* are the components of ⇀ *xb* the in x and y directions and *x*⇀ *yb* and *y*⇀ *yb* are the components of ⇀ *yb* in the x and y directions, respectively. Here, the roll and pitch of the head are relative to the ground; only the yaw of the head is relative to the body. There is no yaw of the body because no relevant reference point on the waist was monitored. Therefore, data on three head motions (yaw, pitch, and roll) and two body motions (pitch and roll) were collected. 

**Figure 2.** Device design and motions of head and body. (**a**) Device for indoor positioning, and head and body motion detection; (**b**) current and base vectors for head and body.

All of the sensors were calibrated prior to the experiments. We first installed the IMUs on a large plate, and performed specific rotations with different angles (e.g., 45◦ , 90◦ , 135◦ , 180◦ ) in three directions. The IMU was regarded as well-calibrated when the difference between all measured values and real values was less than 2◦ . The UWB sensors were also calibrated in the experimental rooms before the experiments. We chose five indoor points at which to perform calibration to reduce the error of indoor positioning to no more than 10 cm.

#### *2.3. Close Contact Behavior*

As motions relevant to the inhalation/exhalation flows, we considered individual head and upper body motions, human respiratory activities (e.g., coughing, sneezing, and speaking), and the features of relative position (e.g., interpersonal distance and relative face orientation) of two people. The definitions of head and body motions can be found in Zhang et al. [11].

Close contact was defined as any full or partial face-to-face interaction within 2 m [23,29–31]. A face-to-face interaction can occur with or without conversation, including when two individuals

read a book or watch a computer screen together. An event was not counted as close contact if the distance between the two students was shorter than 2 m but there was no interaction between them; for example, if two students used their own computers in their own cubicles. If any close contact lasted for more than 1 s, it was counted as a single close contact. If the two students were separated (more than 2 m apart or with no interaction) by more than 1 s, the individual students' close contact behavior (e.g., with another student at that distance) was counted separately. In this study, interpersonal distance was defined as the distance between the sensors on the two participants other than those on their faces. In addition to interpersonal distance, the relative face orientation angle of the two participants was also obtained. This is the angle between the normal of the two students and ranges from 0◦ to 180◦ [11].

#### *2.4. Data Processing*

During the two experimental days, 1,440,492 s of human behavior data were collected. The first author processed all video episodes second by second, recording all visible close contacts between each pair of students in the office. The collected data include posture (standing, sitting, and squatting), whether two students were in close contact, the identities of those in close contact, and the start and end time of each episode of close contact. The indoor position and head and body motions were monitored by sensors. The indoor position data on *day 1* were unfortunately missed, which means that all of the valid indoor position data were from *day 2*. Out of the 23 participants, 21 had indoor position data during *day 2*, and in total 717,168 s of indoor position data were collected. Among these data, 10,827 s (1.5%) was lost or disrupted, and linear interpolation was used to approximate these data. To maintain the accuracy of all recorded data on head and body motions, we discarded the data for which the difference between two calibrations was more than 10◦ or the fluctuation during each calibration was more than 5◦ . After data filtering, 541,200 s of data on head and body motions were valid. Moreover, a total of 1,250,392 s of valid data on sound levels were recorded by the microphones over the two days. All of the results reported below are based on these valid data.

#### **3. Results**

#### *3.1. General Human Behavior Data*

While indoors, the students spent 5.3% of their time standing, 94.6% sitting, and only 0.1% squatting. We divided the office into functional areas, i.e., occupied office cubicles, vacant office cubicles, aisle, and areas near public facilities (Figure 3a). Figure 3b shows the distribution of indoor positions during *day 2*, where red indicates the highest coverage rate (≥40 times per 25 cm<sup>2</sup> ). Occupied office cubicles had the highest coverage rate, mainly due to occupation by their owner (Figure 3b). The students spent on average 92.2% of their time staying in their own cubicles. For the remaining 7.8% of the time, the students occupied other places. The results show that they spent most of the latter time in the aisle close to the doorway, the area close to the door, and areas close to the printer and the water dispenser (Figure 3c). Some students had a particularly close relationship with certain others or even a romantic relationship (e.g., boy-/girlfriend), and their office cubicles had a higher probability of being occupied by others. Most vacant office cubicles had a very low occupied percentage. In general, when outside their own cubicle, the students spent 47.9%, 4.5%, 37.4%, and 10.2% of their time in occupied office cubicles, vacant office cubicles, the aisle, and areas near public facilities.

Based on 541,200 valid data of head and body motions, we obtained the general characteristics of these motions for the overall group of students. Figure 4a illustrates the head and body motions. The circle shows the probability distribution of face orientation in the form of a projected half sphere; the reader can imagine that the eyes are located at the center of the sphere. This half sphere is divided into 1296 sectors, each of them 5◦ × 5 ◦ . The top, bottom, left, and right correspond to the participant raising, lowering, left-turning, and right-turning the head by 90◦ . The center indicates that the student is looking almost exactly forward.

**Figure 3.** Indoor positioning in Room 2 (*day 2*). (**a**) Functional area; (**b**) Distribution of indoor positions; (**c**) Distribution of indoor positions by functional area during the time spent in other places (i.e., students outside their own cubicles). (Indoor positioning data for students 3-4 and 3-5 were lost).

This provides an intuitive visualization of the students' preferred head motions. In the office, the students preferred to look towards the red and orange grids in the circle (i.e., lower their head), and had a very low probability of raising their head (blue grids). To characterize the head motions, we considered the movement of the head in three directions independently. In the horizontal direction, the average degree of yaw was 4.1◦ , which means that the students on average slightly turned their head to the right by 4.1◦ .

**Figure 4.** Indoor human behavior in terms of head and body motion. (**a**) Head motion; (**b**) Body motion. (Yaw of head is relative to the body; pitch and roll of head and all body motions are relative to the ground).

The students had almost equal probability of turning their head left and right, i.e., the probability distribution of yaw was symmetric. The average degree of pitch was 23.0◦ , which means that the

students preferred to lower their head by 23.0◦ on average. Indeed, the head was lowered during 93.5% of indoor time. The students on average tilted (rolled) the head 2.8◦ to the left. The probability distributions of left and right tilts were almost the same. The students spent 86.1% and 97.4% of time tilting their head within 15◦ and 30◦ , respectively.

Figure 4b shows the body motions during indoor time. The students on average lowered their bodies (pitch) by 23.2◦ , and spent 85.6% of indoor time bending their bodies forward. The average degree of body roll was 3.7◦ , which means that the students on average slightly tilted their bodies to the right. The probability distribution of roll was almost symmetric. The students spent 86.7% and 98.2% of time tilting their bodies within 15◦ and 30◦ , respectively.

Three postures were considered in this study: standing, sitting, and squatting. Here we only analyzed head and body motions during standing and sitting because the students spent very little time squatting (0.1%). Figure 5 lists the probability distributions of head and body motions in three directions for all students during indoor time. The rolling of the head and body showed little difference between standing and sitting postures, and the participants had only a slightly higher probability of bending their head during sitting than during standing. However, the characteristics of yaw and pitch of the head, and pitch of the body, differed by posture. The average angles of head yaw during standing and sitting were −3.0◦ and 4.5◦ , respectively. The students on average lowered their heads by 30.9◦ and 24.1◦ while standing and sitting, respectively. The pitch of the body was strongly linked to posture. The average angles of pitch of the body during standing and sitting were 21.2◦ and 24.7◦ , respectively. However, the students had a higher probability of keeping their body bent forward at a large angle during standing (β*<sup>b</sup>* > 60◦ during 8.1% of the time) than during sitting (β*<sup>b</sup>* > 60◦ during 0.7% of the time). The most common forward-bending angle of the body during standing was between 10◦ and 15◦ , while that during sitting was 35◦ to 40◦ . − *β β*

**Figure 5.** Probability distribution of head and body motions by posture.

#### *3.2. Indoor Behavior During Close Contact*

The students spent more than 9.7% of their time in close contact and each student had on average 4.0 close contact episodes per hour. The probability distribution of close contact fitted a log-normal distribution (Figure 6). The average and median durations of close contact were 54.5 and 15 s, respectively. Close contacts with duration between 8 and 16 s had the highest frequency. The durations of 38.2%, 68.8%, and 82.8% of close contacts were no more than 10, 30, and 60 s, respectively. From the microphone data, at least one student was speaking during 68.6% of the time during close contact. One-on-one close contacts accounted for more than 90% of close contact time, while 7.8%, 1.5%, and 0.6% of close contact time involved three, four, and five students simultaneously. Six-student conversations only accounted for 0.06% of close contact time.

**Figure 6.** Probability distribution of duration per close contact.

During close contact, 22.9%, 76.3%, and 0.8% of students stood, sat, and squatted, respectively (Table 1). During 59.6% of the close contact time, both students were sitting. The pattern of one standing and one sitting was adopted during 33.5% of the close contact time. The students almost never chatted with each other while one stood and one squatted (0.1% of close contact time).


**Table 1.** Probability distribution of individual postures and posture pattern of students during close contact.

Figure 7 shows the characteristics of close contact. As can be seen, 68.2% of the close contacts were between students in the same region, and most were between adjacent students. Only 13.2% of the close contacts were between students in remote regions (Figure 7a). Influenced by the circulation of people, students near the aisle and the door had a higher probability of being contacted (Figure 7b). From Figure 7c, the greater the distance between the office cubicles of two students was, the lower was the probability of a close contact. The average cubicle distance between two students for close contact was 1.25 areas (for the area distribution refer to Figure 3a), and 77.2% of close contacts occurred between two students with a cubicle distance of no more than three areas. The student who had the closest contacts had 50.5 episodes per hour (Figure 7d). On average each student had 4.0 close contact episodes per hour, and no one had zero close contacts during either day. During the two days, 12.2% (total pairs of contact: 141; possible pairs of contact: 1156 = 26 × 25 + 23 × 22) of all possible pairs of students had close contact. Each pair of students had close contact on average 11.3 times per day, and the pair with the highest frequency of close contacts had 100 episodes per day (Figure 7e). As shown in Figure 7f, 22.4%, 40.8%, 59.2%, and 85.7% of the students spent no more than 1%, 5%, 10%, and 20% of their time in close contacts, respectively. The most sociable student spent 34% of her indoor time in

close contact. The average and median ratios of close contact time to total indoor time were 9.7% and 6.4%, respectively. During the two days, the students had close contacts with an average of 8.4 and 4.2 students, respectively. The most sociable student had close contacts with 10 and 20 students during *day 1* and *day 2*, respectively (Figure 7g).

**Figure 7.** Characteristics of close contact. (**a**) Probability distribution by relative position of students' office cubicles (same region means two students are in the same region but not adjacent or back to back; percentage shows the episodes of close contact occurred in different relative positions); (**b**) Probability of area occupancy by area during close contact between remote students (students in different regions). The colour bar shows the ln values; (**c**) Probability distribution by distance between work cubicles of two students (distance is calculated as number of functional areas between cubicles of the two students, as illustrated in Figure 3a. For example, distance = 1 for adjacent or back-to-back office cubicles, and distance = 5 between *cubicles 2-2* and *3-2* (see Figures 1b and 3a)); (**d**) Cumulative probability distribution by frequency of close contact (episodes/hour); (**e**) Cumulative probability distribution by total episodes of close contact per day between each pair of participants (episodes/day); (**f**) Cumulative probability distribution by ratio of close contact time to total indoor time (%); (**g**) Cumulative probability distribution by number of contacted students per day; and (**h**) Distribution of number of students who stayed in the room and had close contact during *day 1* and *day 2* (black and blue points are total number of indoor students and total number of students in close contact, respectively).

From Figure 7h, the times of day with the most students in the office were between 9:30 and 11:30 and between 13:00 and 17:30. The peak frequency of close contact was between 11:00 and 12:30 and after 16:00.

Figure 8 shows the interpersonal distance by posture and gender. The average interpersonal distance during close contact was 0.81 m. The average interpersonal distances between sitting-sitting, standing-standing, and standing-sitting students were 0.74 m, 0.93 m, and 0.88 m, respectively. There were three peaks of interpersonal distance during close contact between two students who were sitting. As shown in Figure 8a, the first peak (0.1–0.3 m) was caused by pairs of participants with a very close relationship (e.g., boy-/girlfriend), the second peak (0.6–0.7 m) was caused by pairs of students who sat back-to-back, and the third peak (1.1–1.2 m) corresponded to the distance between two adjacent students. The probability distributions of interpersonal distance for sitting-standing students and standing-standing students accorded with log-normal distributions, and the most frequent interpersonal distances were 0.5 and 0.7 m, respectively. As 60% of close contact was between two sitting students, the overall probability distribution of interpersonal distance (black line) was similar to that of two sitting students.

From Figure 8b, 21.4%, 22.8%, 38.1%, and 17.7% of close contacts were between two male students (M-M), two female students (F-F), a male and a female student (non-couple) (M-F\_NC), and couples (M-F\_C). The contact rates (ratios of actual relationships with close contact to total possible relationships) for M-M, M-F, and F-F were 15.7%, 25.9%, and 10.3%, respectively. The average episodes of close contact per day between each pair of M-M, M-F\_NC, M-F\_C, and F-F who had contact with each other were 6.9, 8.2, 70.5, and 12.1, respectively. The average durations per close contact between each pair of M-M, M-F\_NC, M-F\_C, and F-F were 68.3, 48.7, 53.6, and 58.5 s, respectively. The average interpersonal distances during close contact between these four groups were 0.88 m, 0.70 m, 0.96 m, and 0.71 m, respectively. There was no interpersonal distance shorter than 0.2 m during close contact between two male students. Two female students preferred close contact at a short distance (0.1–0.4 m). Students of different genders (non-couple) had two peaks of interpersonal distance, which were related to their seating position. Couples had a normal distribution of interpersonal distance, and preferred distances between 0.4 and 0.6 m.

**Figure 8.** Probability distribution of interpersonal distance by: (**a**) Posture; (**b**) Gender.

There were three major posture patterns during close contacts in the office: sitting-sitting (both students sitting), standing-standing (both students standing), and sitting-standing (one student sitting, the other standing). The head and body motions of students and relative face orientation angles between students under different patterns of posture during close contact are shown in Figure 9. When two students were both standing or sitting, they preferred to look slightly downward. The average pitch angles of the head under sitting-sitting and standing-standing were 14.7◦ and 20.1◦ , respectively (Figure 9a,b). The average pitch angles of the body under these two conditions were 24.3◦ and 17.8◦ , respectively. However, under the sitting-standing condition (Figure 9c), the eye direction of the standing student was much lower than that of the sitting student.

**Figure 9.** Head and body motions during close contact by posture of the two students: (**a**) Sitting-sitting; (**b**) Standing-standing; (**c**) Sitting-standing. (The circle shows the face orientation of the student; see Figure 4a).

The average pitch angles of the head of the sitting and the standing students in a sitting-standing pattern were 11.3◦ and 34.0◦ , respectively, and the average pitch angles of the body were 22.2◦ and 33.9◦ , respectively. The relatively low probability of looking downward for the sitting student implies that the standing student was usually located at the side of the sitting student rather than face-to-face. Standing students had a very high probability of facing downward.

Two sitting students had a high probability of only a slight relative angle of face orientation (5◦ to 25◦ ) during close contact, which means that they usually faced in similar directions. Two standing students preferred face-to-face close contact with a relative face orientation angle between 150◦ and 170◦ . There was no obvious preference regarding relative face orientation angle under the sitting-standing pattern.

#### **4. Discussion**

#### *4.1. Automatically Collected Indoor Human Behavior*

In this study, we provided the first comprehensive dataset combining the indoor position, head and body motions, and posture of students at the same time which are helpful for infection risk assessment via close contact route. Indoor human behavior is strongly dependent on the type of indoor environment. In a hospital ward, patients usually lie in their beds, while health care workers walk between rooms and beds [32]. In an aircraft cabin or a cruise ship, passengers usually sit in their seats, while crew members walk through the aisle to provide service and food [33–35]. Both passengers in an aircraft cabin or a cruise ship and students in an office spend most of their time in their own seats or in aisles. Children in nurseries have been found to spend more time standing than sitting during indoor free-play time [36]. In a primary school, pupils have been found to spend 2.1 times more time sitting than standing during school hours [37]. However, in this study of a graduate student office, the students spent 94.6% of their time sitting. Therefore, people may tend to prefer sitting with increasing age, although the type of environment is also highly influential.

Close contact is an important activity in daily life, and also plays a critical role in infectious disease transmission. The duration of close contacts directly determines the exposure to viruses. Many researchers have reported the distribution of the duration of close contact (Figure 10) in different types of indoor environment. However, RFIDs or wireless sensors can only detect close contacts at intervals of 20 s [24,31,38–41], an insufficient temporal resolution for close contacts, which have a median duration of only 17 s. Although video observations can reach a temporal resolution of 1 s [11], the subjectivity of video analysts and the huge workload are two major shortcomings. As summarised in Figure 10, all types of indoor environment show similar values of the cumulative probability distribution (CDF) of the duration of close contacts. Brief close contacts (<20 s) are dominant, and prolonged close contacts (>300 s) are rare. Conferences and museums have a higher rate of long close contacts, while in hospitals and congress buildings shorter close contacts are more common. In a video observation study of a graduate student office, the average and median duration of close contact were 53.8 s and 17 s [11], respectively. Our study, meanwhile, found the average and median duration to be 54.5 s and 15 s, respectively.

**Figure 10.** Cumulative probability distribution (CDF) of duration per close contact.

#### *4.2. Close Contact Behavior*

We collected and analyzed three types of data during close contact: indoor position, head and body motion/movement, and posture. These three factors are important to infectious disease transmission. Indoor position can help calculating the interpersonal distance of people during close contact. The infection risk decreases sharply with the increase of the interpersonal distance [12,20,42,43]. Posture, and head and body movement influence the body plume during close contact. For example, short-range exposure can be affected strongly by body plumes [44]. Frequent movement of the head and body during conversation can change not only the orientations of the exhaled/inhaled airflows, but also the patterns of body convective flows and the thermal plume. The exhaled airflows of two people also interact with and affect each other [12]. Various gestures involving small movements of the hands, palms, legs, eyebrows and other small-scale facial features may not significantly affect the body plume or exhaled flows [45,46]. Posture also important in droplet deposition. For example, more droplets may deposit on thighs of a sitting person if he/she talks with an infected. People have high probability to touch their thighs and legs with the frequency of more than 30 times per hour [19]. It may lead to a high infection risk because people also have high touch frequency on mucous membranes [47]. Relative face orientation is a critical factor for exposure during close contact, and it could be calculated by body and head motion. Previous studies found that the exposure during face-to-face close contact is the most, followed by face-to-side pattern, while face-to-back pattern had the lowest exposure [21,22].

As no such data on indoor human behavior in an office had previously been published, we mainly compared the sensor-collected data from this study with those of our previous experiment based on video observation [11] (Table 2). As head and body motions cannot easily be recorded using observation methods, the accuracy is difficult to guarantee. This may be the cause of the disagreements between the results in Table 2. In addition, in our previous experiment, all participants were students whose cubicles were in the studied office. In this study, however, some students were from the next room and some were invited into the office under study by students who worked there. Therefore, the various relationship networks may have caused the difference in the characteristics of close contact and of indoor human behaviors during close contact.


**Table 2.** Comparison of indoor human behaviors during close contact in a graduate student office between the previous observation study and the present sensor-based study.

<sup>1</sup> Observation study: all data obtained through observation of video tapes by video analysts.11 <sup>2</sup> Sensor-based study (this study): all data obtained by sensor detection. <sup>3</sup> ID: interpersonal distance.

Our previous experiment showed that students spent 9.9% of their time in close contact interactions, while in this study that percentage was 9.7% ± 8.8%. The probability distributions of the students' posture during close contact differed between the two experiments, although sitting was always the most common posture, followed by standing. Almost none of the students had close contact when squatting. Sitting-sitting and sitting-standing were the most common posture patterns during close contact. In this study, the sitting-sitting pattern was much more dominant than in the previous experiment because all students chose their office cubicles before the experiment. Students with the closest relationships sat next to each other, and could chat without standing up or walking. During both experiments, one-on-one close contact accounted for almost 90% of all close contacts. The number of participants per close contacts depends on the type of indoor environment. For example, more people will participate in a close contact in a group discussion, while one-on-one close contact is very highly probable in a doctor's consulting room.

Preferred interpersonal distances differ by culture, ethnicity, relationship, personal habits, age, gender, and ambient environment [48]. Closer interpersonal distance means higher virus inhalation which increases infection risk via close contact route [42]. The average personal distances for acquaintances and intimately close persons in China are 83.6 and 57.6 cm, respectively [49]. However, there is no clear definition of interpersonal distance. Our previous study defined the interpersonal distance as that between two mouths, while this study defined it as the distance between two sensors worn on the top of the head. The previous study showed the average interpersonal distance in the office to be 0.67 m. In this study, the average interpersonal distance (0.81 m) was larger because of the different definitions of this parameter. Both experiments found the same effect of gender on interpersonal distance: two female students adopted the shortest distance while students of different genders (non-couples) were farthest apart. As an office is a public area, intimate couples do not interact at such close distance as they would otherwise. Posture also influences the interpersonal distance. Both experiments showed that two sitting students had the shortest interpersonal distance. In this study, more than 50% of close contacts were between adjacent or back-to-back students, and the probability distribution of interpersonal distance was strongly influenced by their cubicle positions. A compact indoor design would lead to shorter interpersonal distance during close contact.

The circulation of people around the office strongly influenced the probability of close contacts. Students at long distances apart had low probability of contact with each other. Therefore, a linear office

cubicle design would lead to a lower close contact frequency compared with a matrix design because of longer average distance. People are generally disinclined to make close contact with someone a long distance away if there is no critical information to communicate. In this study, most close contacts were between students sitting in the same region because students with good relationships had chosen to sit together. Most students in this office were acquaintances, and had high probability of a brief close contact when encountering each other. Therefore, the students near the aisles, especially those near the door, had a high close-contact frequency with remote students (Figure 7b) as a result of office circulation. There was no close contact between 88% of the possible pairs of students (connection rate = 12%). In general, this value is strongly related to the level of intimacy between people indoors. For example, the connection rate tends to be very high in a home, but low in public environments such as conference rooms and aircraft cabins. This may explain why homes have a higher infection risk than other environments during an infectious outbreak [29].

This sensor-based study recorded higher pitch values of the head and body, which showed that students preferred to lower their head and lean their body forward. When sitting, they were usually looking at a computer monitor (mostly laptop computers) or reading a book from a desk, and therefore both the torso and head leaned forward. Over the long term, such a sitting posture may lead to a high prevalence of kyphosis among students [50]. Research has also showed that tilting the head forward by 15◦ places about 27 pounds of force on the neck [51]. This increases to 40 pounds at 30◦ , 49 pounds at 45◦ , and 60 pounds at 60◦ [51]. The damage caused by untreated 'text neck' can be of similar severity to occupational overuse syndrome or repetitive stress/strain injury. The office management team could be encouraged to educate their workers and provide ergonomic office equipment such as laptop stands to promote proper posture.

In simulations of infectious disease transmission via the close contact route, face to face is the most common orientation assumed, and all heads are assumed to be at the same height [12,20]. However, in our analysis, the students spent relatively little of their close contact time in a face-to-face position. Staring at the same screen, reading the same paper, or chatting without making eye contact were also common. Before this study, no data had been published on the percentage of time spent speaking during close contact. We found that students in close contact on average spent 68.6% of their time speaking. Studies have shown that talking for 5 min can generate the same number of droplet nuclei as a cough, i.e., some 3000 droplet nuclei [10], and speaking usually involves prolonged exhalation [52]. Our data may provide a reference for the simulation of infectious disease transmission via speaking during close contact. Moreover, combined with indoor surface touching behaviors [19], a comprehensive simulation of infection spread in an office considering the long-range airborne, fomite, and close contact routes could be performed [29]. The results would provide support for infection risk analysis on a large scale such as a city [5,53].

#### *4.3. Limitations*

This study has several limitations. The presence of the cameras might have had a psychological impact on the students' behaviors. The experimental hats and shirts showed slight relative movement when the participants moved their heads and bodies. Although we discarded some data with very large fluctuations, some error remained (resolution: 5◦ ). Still, this error was much smaller than that of observation methods. Another limitation is that the interpersonal distance was defined as the distance between two sensors installed on the top of the experimental hats rather than between two mouths or noses. The interpersonal distances obtained in this study were longer than those between two mouths/noses, which are normally used to simulate close contact between two persons. Our experiment collected more than 1440,000 s of indoor data and more than 139,000 s of close contact data over 2 days. This data volume is large, but still insufficient to represent all indoor human behaviors in graduate student offices. Moreover, the close contact behaviors presents the characteristics in the office with around 25 students. Building type and total indoor population will influence the human behaviors.

### **5. Conclusions**

Students spent long time on close contact (9.7%) in the office, which may explain the importance of the close contact route for many respiratory infections. The probability of close contact decreased exponentially with the increasing distance between two students' cubicles. Therefore, students who sit closer to the infected student, have much higher infection risk via close contact route than students who sit further. Comparing with pairs of sitting students, pairs of standing students had lower infection risk via close contact route because they did not prefer a face-to-face talk. The fact that standing students much preferred lower their head and body than sitting students may lead to a shorter distance of exhalation jet than we thought.

**Author Contributions:** Conceptualization, N.Z., Y.L. and P.W.; methodology, N.Z., B.S. AND P.W.; software, B.S.; formal analysis, N.Z. and P.-T.C.; investigation, N.Z., P.-T.C., T.M., B.S. and P.W.; resources, N.Z.; data curation, N.Z. and Y.L.; writing—original draft preparation, N.Z.; writing—review and editing, N.Z., Y.L., P.-T.C., T.M., B.S. and P.W.; visualization, B.S. and P.-T.C.; supervision, Y.L.; project administration, Y.L.; funding acquisition, Y.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by a General Research Fund provided by the Research Grants Council of Hong Kong, grant number 17202719; a Collaborative Research Fund provided by the Research Grants Council of Hong Kong, grant number C7025-16G; an HKU ZIRI seed fund, grant number 04004.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

International Journal of *Environmental Research and Public Health*

### *Article* **A Salutogenic Approach to Disaster Recovery: The Case of the Lac-Mégantic Rail Disaster**

### **Mélissa Généreux 1,2,\* , Mathieu Roy 1,3 , Tracey O'Sullivan <sup>4</sup> and Danielle Maltais <sup>5</sup>**


Received: 29 January 2020; Accepted: 20 February 2020; Published: 25 February 2020

**Abstract:** In July 2013, a train carrying crude oil derailed in Lac-Mégantic (Canada). This disaster provoked a major fire, 47 deaths, the destruction of 44 buildings, a massive evacuation, and an unparalleled oil spill. Since 2013, Public Health has undertaken several actions to address this challenging situation, using both quantitative and qualitative methods. Community-based surveys were conducted in Lac-Mégantic in 2014, 2015 and 2018. The first two surveys showed persistent and widespread health needs. Inspired by a salutogenic approach, Public Health has shifted its focus from health protection to health promotion. In 2016, a Day of Reflection was organized during which a map of community assets and an action plan for the community recovery were co-constructed with local stakeholders. The creation of an Outreach Team is an important outcome of this collective reflection. This team aims to enhance resilience and adaptive capacity. Several promising initiatives arose from the action plan—all of which greatly contributed to mobilize the community. Interestingly, the 2018 survey suggests that the situation is now evolving positively. This case study stresses the importance of recognizing community members as assets, rather than victims, and seeking a better balance between health protection and health promotion approaches.

**Keywords:** disaster; psychosocial impacts; community resilience

### **1. Motivation**

### *1.1. The Lac-Mégantic Train Derailment Disaster*

On 6 July 2013, in the middle of the night, a train carrying crude oil derailed in the heart of Lac-Mégantic (Quebec, QC, Canada). This small town, situated on a lakeshore in the Estrie region of the province, has a population of 6000 residents.

The train, with no engineer at the controls, spontaneously rolled downhill from its night stop location 11 kilometres away, toward the town of Lac-Mégantic. With a relatively constant downhill slope, the train's descent accelerated to almost 100 kph by the time the locomotives encountered a sharp curve in downtown Lac-Mégantic and most of the trailing cars derailed. As they derailed, 63 tank cars ruptured and escaping crude oil ignited, leading to a succession of powerful explosions and a major conflagration. The fire spread rapidly to nearby structures, destroying 44 buildings. The derailment, the explosions and the subsequent fire resulted in 47 deaths and necessitated mass evacuation of 2000 persons, equivalent to one-third of the town's population.

With the coupling of human suffering and environmental degradation, the Lac-Mégantic derailment caused serious psychosocial and economic consequences, including the relocation of many families forced to leave their homes, loss of many jobs, and closure of many local businesses for weeks before relocating elsewhere in town [1]. Given the impact of this technological disaster, the involvement of public health personnel and resources was critical throughout the emergency response operations. The Public Health Department (PHD) for the Estrie region responded immediately to provide direct services needed to protect the citizens of Lac-Mégantic and on-site responders from several health hazards. The priority at that time was to assess, communicate, and manage immediate risks to public health associated with exposure to chemical, physical and biological agents [1].

In the face of disasters, it is important to recognize that the operational domain of public health in affected communities extends beyond health protection and disease prevention to include promotion of health and well-being. It is with this in mind that Estrie PHD, in collaboration with researchers in the field of psychosocial recovery, conducted a population health survey entitled "Enquête de santé populationnelle estrienne" (ESPE) in 2014. Unknowingly, this was the first of a long series of promising initiatives to mobilize the local community in this post-disaster landscape.

#### *1.2. The Salutogenic Approach*

The approach and orientation of the Theory of Salutogenesis are now well established in health sciences and used in various health promotion settings and contexts [2]. Unlike traditional preventive approaches aimed at identifying risk factors, limitations or diseases, asset-based approaches are used to identify factors fostering well-being, resources, or abilities [3]. According to the scientific literature, a greater stock of health assets empowers individuals and communities and helps to improve health and well-being. This is true both directly (i.e., health assets are associated with better health outcomes [4]) and indirectly (i.e., health assets moderate the relationships between a disadvantaged social position and negative health outcomes [5,6]). Examining the role of salutogenic factors in disaster contexts, however, needs further exploration as the field of health emergency and disaster risk management places much more focus on hazards, risks factors, vulnerability, and short- and long-term adverse health outcomes.

#### *1.3. Objectives*

Through the case of the Lac-Mégantic train derailment tragedy, we aim to discuss how salutogenesis can be used as a relevant and complementary framework in disaster settings, and how it can be incorporated into post-disaster recovery strategies to promote resilience. More specifically, the objectives of this case study are to: (1) describe the salutogenic approach applied to the Lac-Mégantic train derailment, (2) present the long-term psychosocial outcomes of this disaster, and (3) discuss some benefits observed from applying a salutogenic approach in a post-disaster landscape.

#### **2. Approach**

#### *2.1. Quantitative Approach in Addressing a Challenging Environment*

Any collective trauma, including technological or natural disasters, is likely to lead to adverse health impacts among survivors and the wider community. Due to the experience of extensive stress and loss, people exposed to large-scale disasters like the Lac-Mégantic train derailment are subject to long-term adverse outcomes [1]. There is now a solid evidence base for the substantial effects of such a tragic event on psychological health in directly affected communities, which may persist over time in the absence of adequate support. Interestingly, disasters may also result in positive psychological outcomes in some exposed persons [7–12].

Among actions that can be performed by public health agencies to bring support to local communities following disasters, monitoring long-term psychosocial outcomes (both positive and negative ones) is certainly relevant. Monitoring helps tailor interventions aimed at supporting affected individuals and communities, by promoting their resilience and recovery processes. The Estrie PHD, in close collaboration with the "Université du Québec à Chicoutimi" (UQAC) and the "University of Sherbrooke", has therefore spent the first years following the event tracking the health needs and assets of those living in the Granit area using repeated cross-sectional community-based surveys.

In 2014, one year following the rail disaster in Lac-Mégantic, the PHD conducted a first health survey using a community-based random sample of 811 adults from the Granit area and additional 8000 adults residing elsewhere in the Estrie region. This representative sample responded to a telephone or web survey covering a variety of physical and psychological health outcomes. The second phase of the ESPE was carried out in the fall of 2015 and sought to better understand the local population's health and well-being, along with its possible link to the July 2013 railway disaster. In total, 1600 adults were recruited randomly in 2015 to take part in this large-scale telephone survey. These included 800 from the Granit area, and 800 from elsewhere in the Estrie region. In the fall of 2018, a third, similar, survey was conducted and is referred to as phase 3. Each of these three studies is composed of a separate sample of adults residing in the Granit area or elsewhere in Estrie; the original sample of participants was not monitored across time. While an additional study was conducted in 2016 by UQAC, a different sampling strategy was used; therefore, it is not used for comparison with the other surveys [13,14].

The adults who agreed to participate in these studies were asked to answer an anonymous questionnaire, which took approximately 30 min to complete. A number of questions were identical across all three surveys, allowing for the comparability of results over time (years 1 to 5 following the tragedy). Various psychosocial outcomes were examined, including adverse effects of disasters (e.g., psychological distress, depressive episodes, signs of post-traumatic stress, diagnosed anxiety or mood disorders, social worker or psychologist consultation, anxiolytic drug use, alcohol abuse), but also positive ones (e.g., resilience, positive mental health, sense of coherence, sense of community belonging, social support). The following outcomes, all self-reported, were examined in at least one of the three cross-sectional surveys. These outcomes have been described more thoroughly elsewhere [8,15].

Deficit-based outcomes: Perception of fair or poor general health, excessive drinking episodes (at least once a week), finding most of the days stressful, psychological distress in the past month, based on the 6 item Kessler Scale (K6, ≥7; [16]), signs of post-traumatic stress in the last week (specific to the train derailment) based on the 15 item Impact of Event Scale (score ≥ 26 [17,18]), diagnosed anxiety disorders, diagnosed mood disorders, social worker or psychologist consultation in the past year, and perception of insecurity in the neighbourhood.

Asset-based outcomes: Resilience in the past month, based on the 10 item Connor–Davidson resilience scale (score ≥ 30; [19]), positive mental health, in the past month, based on the 14 item Mental Health Continuum-Short Form questionnaire [20,21], sense of community belonging, sense of coherence, based on the short version (3 items) of the sense of coherence (score ≥ 5; [22]), social support, based on the Multidimensional Scale of Perceived Support (score ≥ 69; [23]).

#### *2.2. Qualitative Approach in Addressing a Challenging Environment*

The release of the ESPE 2015 data (i.e., in February 2016) stimulated the emergence of health promotion and advocacy interventions for the local population in Lac-Mégantic. Given the magnitude of the tragedy, it was necessary to take a step back to understand the situation in relation to the normal process of community recovery. It was in this context that in March 2016, the Estrie PHD intensified its work with community partners, first by organizing a day of collective reflection. The purpose of this initiative was to work together to gain understanding of the situation and reverse the cycle. During this day, no fewer than 50 key actors (decision makers, stakeholders, citizens and experts) gathered. The reflection day was divided into two parts: (1) conference and workshops on resilience and lessons learned from the past and (2) conference and workshops on levers for long-term recovery and priorities for the future.

A defining moment during the Day of Reflection occurred during an asset-mapping activity through which participants were invited to construct together an historical timeline that traces key milestones in recovery of their community and to recognize the progress made (Figure 1). More precisely, they were first divided in subgroups, where they had to highlight good moves, or successful interventions and initiatives implemented by local partners and citizens since the tragedy. Then, subgroups had to share their respective thoughts to the larger group in order to collectively construct the timeline. By doing so, the large group was able to identify a wide and diversified range of local assets, including physical, cultural, economic, social and spiritual ones, that all created positive effects on the community.

**Figure 1.** Historical timeline tracing key milestones in recovery of Lac-Mégantic community (March 2016).

Throughout the Day of Reflection, a common vision of the desired future emerged and priorities for action and research were identified, leading to the co-construction of what would become the "Plan for the Recovery and Development of a Healthy Community in Lac-Mégantic and the Granit area". This plan pursues the following objectives:


In the weeks following the elaboration of the plan (i.e., April 2016), PHD advocated for additional funding to support its implementation. In June 2016, the "Ministère de la Santé et des Services sociaux" and the Canadian Red Cross announced substantial investments that would serve as financial levers to implement the adopted action plan. The ESPE data was an important contribution supporting an informed decision, based on understanding of the long-term psychosocial impacts of the tragedy.

In sum, holding such a Day of Reflection, which brought together key players from the community, contributed to the development of a common vision of solutions and the transmission of a clear, coherent and positive message to decision-makers and the community.

*"Building a project together is really motivating. Especially since everyone feels involved: from citizens to elected o*ffi*cials. It was a very inspiring day!"* —A participant of the collective reflection day.

This positive experience supports existing knowledge that beyond traditional surveys, qualitative methods are valuable for listening to, learning from, and engaging local partners and high-risk citizens. Through inclusive and empowering approaches, public health practitioners and researchers can better integrate members of the community as assets rather than victims and take into considerations their capacities in addition to their needs [24].

#### **3. Results**

#### *3.1. Observations from the Community-Based Surveys (Quantitative Approach)*

#### 3.1.1. The First Years Following the Disaster

In 2014, some differences were observed in the prevalence of deficit-based and asset-based psychosocial outcomes as a function of residential location (Lac-Mégantic, elsewhere in the Granit area, or elsewhere in the Estrie region) (Table 1). Many of these "psychosocial gradients" were stronger in 2015 [15]. Anxiety disorders, for instance, were twice as high in Lac-Mégantic residents as in other residents of the Estrie region in 2015 (14.1% vs. 7.2%, *p* = 0.003). In the same vein, adults in Lac-Mégantic, as opposed to those living elsewhere in the Estrie region, were less likely to report a high level of resilience in 2015 (47.8% vs. 63.3%, *p* < 0.0005), while this was not the case the year before. Similar observations were made for optimal mental health.

Significant time trends from year one to year two post-disaster were also observed. While most psychosocial outcomes did not show any statistically significant improvement among adults, the use of psychosocial services decreased by half among adults residing in Lac-Mégantic between 2014 and 2015 [15].

Some deficit-based (e.g., post-traumatic stress) and asset-based outcomes (e.g., sense of coherence) were only examined as from 2015 (Table 2). Findings from the second wave revealed seven in ten adults in Lac-Mégantic showed moderate to severe signs of post-traumatic stress two years after the disaster. On another note, a strong sense of coherence was observed among 48.2% of adults residing in Lac-Mégantic, regardless of residential location, and this proportion was significantly lower than that observed elsewhere in the Estrie region (61.1%). These findings suggest that the stock of health assets can weaken with time among people directly impacted by a disaster, especially in the absence of adequate support and services [8].


**Table 1.** Deficit- and asset-based outcomes among a community-based sample of adults according to residential location, two years and five years post-disaster (e.g., Lac-Mégantic train derailment, 6 July 2013), Estrie region, 2014 and 2015.

#### 3.1.2. Long-Term Trends in Psychosocial Outcomes Following the Disaster

Given the increased efforts to support recovery in Lac-Mégantic in recent years, has there been any progress in terms of psychosocial outcomes? With regards to all the data collected from our three surveys, major findings emerge. First, the adverse psychosocial impacts observed in the years following the Lac-Mégantic rail tragedy in 2013 seem to be receding. For example, after reaching a peak in 2015, the proportions of adults reporting an anxiety disorder diagnosed by a doctor stabilized in 2018 in Lac-Mégantic. On the other hand, these proportions increased significantly elsewhere in Estrie from 2014 to 2018. In other words, the gap that had developed between Lac-Mégantic and the rest of Estrie in the first two years after the tragedy is no longer, in many respects [25].

Second, there was still a very high prevalence of signs of post-traumatic stress in 2018 (71.9%). Despite a gradual adaptation of citizens to the losses and stressors experienced during and after the 2013 tragedy, the local community has been deeply affected by the traumatic event and its aftermath. These markers could persist for many years, despite an outward appearance of adaptive functioning of individuals and their community. Finally, protective factors are increasingly observed in Lac-Mégantic, particularly social support and sense of belonging to the community that were especially strong in 2018 [25]. These factors may act as powerful moderators of the adverse effects of primary and secondary stressors typically arising from large-scale disasters.


**Table 2.** Deficit- and asset-based outcomes among a community-based sample of adults according to residential location, two years and five years post-disaster (e.g., Lac-Mégantic train derailment, 6 July 2013), Estrie region, 2015 and 2018.

(+): Significant increase from 2015 to 2018. (−): Significant decrease from 2015 to 2018.

#### *3.2. Observations from the Field (Qualitative Approach)*

#### 3.2.1. The Outreach Team

Following the day of reflection, in 2016, Estrie PHD created a permanent community Outreach Team in Lac-Mégantic. Located outside formal clinical settings (i.e., in the downtown area), this multidisciplinary team has focused on bringing psychosocial services closer to the population. Four full-time professionals (two social workers, one outreach worker and two community organizers), and two part-time professionals (a kinesiologist and a nutritionist) comprise the team.

The following principles guided the entire Lac-Mégantic outreach initiative: global health, prevention, scientific rigour, a strengths-based approach, empowerment, inter-organizational and intersectoral collaboration, and inclusion. Citizen participation and community development were at the heart of this approach. A wide range of services are offered, ranging from daily interactions with citizens and local organizations (in the form of psychosocial support, response to service requests, rapid detection and response to emerging needs, collaboration with the organization of activities, etc.) to involvement in various projects emerging from the action plan [25].

#### 3.2.2. Promising Initiatives to Mobilize the Local Community

The EnRiCH (Enhancing Resilience and Capacity for Health) Community Resilience Framework for High-Risk Populations [24] inspired the strategies developed within this community to promote community resilience, health and well-being [26,27]. Based on qualitative research conducted in five Canadian communities and a review of scientific literature, this framework provides an asset-based integrated upstream and downstream approach to disaster risk. With the development and use of adaptive capacities as a central element, it advocates three pillars and four areas of intervention, as described in Table 3, all in a cultural context and working with the complexity specific to disasters.



Source [24].

In line with this reference framework, several promising initiatives have been implemented in recent years within the Lac-Mégantic community to activate community resilience, social cohesion and citizen participation in a post-disaster setting. Committed to keeping track of local innovations and sharing them in formats that are suitable for both experts and practitioners, a synthesis of some of these promising initiatives has been produced and updated on an annual basis by the Outreach Team since 2017 [28]. These initiatives (e.g., social animation, Photovoice, Greeters, walking club) all contributed significantly to empowering citizens and mobilizing the community of Lac-Mégantic and surrounding areas. These initiatives also appear to have had a positive impact on the mental health and well-being of the citizens of this community.

As is generally known, organizing community projects or collective events, increasing opportunities to become involved as citizens, as well as other elements that strengthen social capital, contribute to building resilience in a post-disaster context. The data collected in this regard from 800 adults in MRC du Granit in the framework of ESPE 2018 provides additional support for this knowledge (Figure 2 [25]).

**Figure 2.** Elements that have significantly improved personal well-being over the past 12 months, "Enquête de santé populationnelle estrienne" (ESPE) 2018 (Granit area, 800 adults).

#### Photovoice

In 2017, in collaboration with the University of Ottawa EnRiCH research team and PHD of Estrie, the citizens of Lac-Mégantic took part in a Photovoice Initiative to map the assets of their community and develop a positive campaign and vision for the community looking forward to 2025. Over a 6 month, period the Lac-Megantic Photovoice Group met monthly to take photos of community assets and ideas to support their vision for the community. They met to discuss their photos with the group and share their ideas around issues that matter to them. The Lac-Mégantic Photovoice Group planned and hosted two exhibitions to facilitate knowledge mobilization and foster dialogue with decision-makers in Lac-Mégantic and Ottawa, including local and federal politicians. The Photovoice Initiative was highlighted as an inspirational example of community engagement in resilience initiatives in a report by the World Health Organization [29].

*"We could express our sadness, our emotions openly because we were welcomed, without criticism. At first it was quite emotional, but over the meetings, this overflow was transformed into something lighter. It did me good. It made a big di*ff*erence."* —A participant of the Photovoice Initiative.

#### Ephemeral Place

The population is struggling to reclaim the downtown area of Lac-Mégantic, which was largely destroyed during the railway tragedy of 2013. Being under reconstruction, this new place, full of meaning and memories, is a constant reminder of loss. At the same time, there is a desire among citizens to get involved and to revitalize their living environment. In 2018, Ephemeral Place in the heart of the city was created in response to this desire; it is a space to promote social activities, networking and gatherings. This outdoor venue, under the responsibility of the Outreach Team, encourages the involvement of citizens of all ages and all horizons to foster social participation. Since these are temporary installations, it is an opportunity to experiment with concepts or ideas, while creating positive experiences. Through its free and varied leisure activities offered to citizens (5 to 7 with musicians, barbecues, outdoor film screening with popcorn, laughter yoga, intergenerational karaoke, etc.) and its unique approach, Ephemeral Place undeniably supports the long-term recovery of the Lac-Mégantic community [30].

#### Lessons Learned from a Citizen's Perspective

Inspired from a similar initiative following the bushfires which swept across Victoria, Australia, in 2009 [31], the idea behind this project was to collect statements from people who experienced the tragedy through one-on-one and/or group interviews and to identify overriding themes. Driven by the Outreach Team, this initiative provided a voice and brought together people who wished to contribute in this way, naming what could be changed or improved as a way of managing future disasters. Through their experience, citizens could make recommendations to the different bodies with which they interacted during the rail tragedy of July 2013 but also during the months and years that followed it. A semi-structured interview guide was developed, based on the CHAMPSS Functional Capabilities Framework [32]. The acronym stands for the following categories of functional capabilities: Communication, Awareness, Mobility/Transportation, Psychosocial, Self-Care and Daily Tasks, and Safety and Security. Approximately 12 interviews were conducted with citizens that would not have been reached otherwise, in order to make their voice heard. Data collected through these interviews was then pooled and analyzed to draw emerging themes that were sent to participants for a further validation. By being inclusive and recognizing the various experience lived, this project gave citizens a different opportunity to contribute following the tragedy. All this led to the writing of a document sharing post-disaster good practices, according to the perspectives of citizens with a unique field expertise in the matter. This document could then be submitted to the bodies concerned, upon approval from the group.

#### **4. Recommendations**

To our knowledge, this study is among the first to report how a salutogenic approach can contribute to improve health and well-being in the aftermath of a disaster. Only few studies used positive approaches in such type of setting. Leadbeater and colleagues [33] described how community leadership facilitated the social recovery process in the community of Strathewen (Australia), following the 2009 Victorian bushfires. Van Kessel [34], for its part, explained how promoting resilience mitigated impact on mental health after the 2010/11 Victorian floods and the 2009 Victorian bushfires. It is noteworthy that a recent systematic review highlighted a gap in the evidence relating to specific interventions targeting the resilience of adults who have experienced a disaster. Authors from this review call for more studies exploring the ability of interventions to build the intrinsic capacity of a community to adapt to disasters [35]. Despite the paucity of "real-world research" and knowledge on effective asset-based approaches in a post-disaster landscape, many theoretical or conceptual papers support the key role of community resilience in promoting population health in such settings [36–42].

In our case study, the objective was to discuss how a salutogenic approach was used to help the community of Lac-Mégantic in its recovery from a profound tragedy. Our various community-based surveys, combined with continuous on-the-ground presence of the PHD, have provided situational awareness about how the psychosocial impacts resulting from the 2013 rail tragedy decreased over time. Although this tragedy has left its mark, the local community is gradually adapting to its new reality. The asset-based approach used in recovery seems to have contributed to this "new reality" and emphasizes the importance of social capital to activate individual and community resilience in post-disaster contexts.

Many lessons have been identified from this unique and informative experience. First, long-term monitoring of psychosocial impacts through repeated community-based surveys is relevant, if not essential. Such surveys serve as powerful tools for health promotion initiatives and advocacy on behalf of the local population. Such survey support priority setting (e.g., targeting most at-risk populations) and promote risk-informed decision making.

Second, the voices of various groups, including those at heightened-risk, should be heard to take account of their specific needs and capacities. It is important to take time to listen and learn from citizens and consider all members of the community as assets rather than victims. This is critical to promote concrete social measures and psychosocial support tailored to their needs.

Third, regardless of the extent of the problems observed in the field, public health must seek a balance between a health protection (focused on hazards and risk factors) and health promotion (focused on protective factors, local strengths and resources).

Fourth, public health practitioners, academics and leaders must collaborate closely with local organizations and citizen groups. This is fundamental for a successful recovery. Putting citizens at the heart of all considerations helps to make sense out of a chaotic situation and contribute to the recovery of the community.

Fifth, public health organizations should capitalize on existing knowledge to develop and apply strategies and interventions in a post-disaster context. As part of their recovery operations, they should also share their own knowledge and experiences (e.g., lessons learned, tools and resources).

Finally, this rich experience in Granit over the last six years enabled us to identify three key success factors in supporting the psychosocial recovery and social reconstruction following a disaster:


#### **5. Conclusions**

This case study gives a concrete example of how asset-based approaches can be fruitful for enhancing community resilience and improving the health and well-being of a community in a post-disaster landscape. The positive evolution of the psychosocial situation in Lac-Mégantic, assessed both quantitatively and qualitatively, demonstrates the importance of developing a common understanding of risks and working together in finding solutions.

#### **6. Future Work**

Let us recall the importance of understanding, preventing and reducing psychosocial risks in the months and years following a disaster, whether natural, technological or intentional. In any case, concerted action to promote community resilience is required during, after, and ideally before the occurrence of such an event. As advocated by the United Nations, we must move from a disaster management logic to a risk management logic associated with these events, in partnership—rather than in silos—for the good of the community [43].

Disaster risk reduction, which is closely associated with climate change adaptation (due to the increasing number and intensity of recent disasters), is a pressing field of action for decision-makers, practitioners and researchers to promote health and well-being of communities and to increase their resilience for coping with multi-origin hazards.

While much is known about interventions targeting health needs following disasters (i.e., deficit-based models), less is known about what could foster resilience. Our case study was limited to a very simple design (e.g., post-disaster cross-sectional surveys and field observations). In order to generate stronger evidence-base intervention, future work in this research field should be based on high-quality studies (i.e., randomized or prospective cohort studies).

**Author Contributions:** Conceptualization, M.G.; methodology, M.G., M.R., T.O., and D.M.; formal analysis, M.G.; writing—original draft preparation, M.G., M.R., T.O., and D.M.; Writing—review and editing, M.G., M.R., T.O., and D.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors wish to acknowledge the financial support of the "Ministère de la Santé et des Services sociaux" and the Canadian Red Cross for the implementation of the Outreach Team and the multiple initiatives to mobilize the local community in a post-disaster landscape.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

International Journal of *Environmental Research and Public Health*
