Real-time sensing of war’s effects on wellbeing with smartphones and smartwatches

Background Modern wars have a catastrophic effect on the wellbeing of civilians. However, the nature of this effect remains unclear, with most insights gleaned from subjective, retrospective studies. Methods We prospectively monitored 954 Israelis (>40 years) from two weeks before the May 2021 Israel-Gaza war until four weeks after the ceasefire using smartwatches and a dedicated mobile application with daily questionnaires on wellbeing. This war severely affected civilians on both sides, where over 4300 rockets and missiles were launched towards Israeli cities, and 1500 aerial, land, and sea strikes were launched towards 16,500 targets in the Gaza Strip. Results We identify considerable changes in all the examined wellbeing indicators during missile attacks and throughout the war, including spikes in heart rate levels, excessive screen-on time, and a reduction in sleep duration and quality. These changes, however, fade shortly after the war, with all affected measures returning to baseline in nearly all the participants. Greater changes are observed in individuals living closer to the battlefield, women, and younger individuals. Conclusions The demonstrated ability to monitor objective and subjective wellbeing indicators during crises in real-time is pivotal for the early detection of and prompt assistance to populations in need.

Figure 1a shows the distribution of the number of questionnaires filled by the participants during the 53 days of the study.The average number of days with questionnaire answers was 16.14 out of 53 days (30.19%).Supplementary Figure 1b shows the distribution of the number of days smartwatches were worn during the study.The average number of days with smartwatch data was 33.01 and the median was 43 out of 53 days (81.13%).Supplementary Figure 1c shows the distribution of the number of days data from the smartphone sensors were collected.The average number of days with smartphone sensor data was 25.97 and the median was 25 out of 53 days (47.17%).

Supplementary Methods
The daily questionnaire included the following eight questions: 1. How is your mood today?
3. How would you define your last night sleep quality?
5. How would you describe the level of your stress during the last day?
• Very Low (-2) • Low (-1) • Medium (0) • High (1) • Very high (2) 6. Try to estimate the number of people you have been with in the last day for up to 2 meters away?

5/12
In Israel, when incoming missiles are spotted, air-raid sirens are turned on and civilians are instructed to head to shelters.The effective time to reach a shelter once an air-raid siren is turned on ranges from 15 to 180 seconds, depending on the area's distance from the closest missiles launch area.Notably, missiles against Israel can be launched from various countries, including Gaza in the south and Lebanon and Syria in the north.The rationale behind our selection of the three exposure groups was mainly guided by the effective time to reach a shelter once an air-raid siren was turned on.However, as can be seen from Supplementary Figure 3, the effective time to reach a shelter alone is insufficient to define one's exposure group.That is, participants who live far away from Gaza and were not at risk to be hit by missiles during the current war cycle (for example, participants living in the north of Israel) are also associated with a short (below 90 seconds) effective time to reach a shelter.Therefore, instead of relying purely on the effective time to reach a shelter, we used the distance from Gaza.Namely, the lower threshold (60km) was chosen since it is able to capture the short effective time to reach a shelter for areas close to Gaza, and the upper threshold (110km) was chosen based on the maximum range of missiles fired from Gaza during this war cycle.

Out of Gaza Missile Range
High Exposure Medium Exposure Low Exposure Supplementary Figure 3. Classification into exposure groups.The y-axis represents the effective time to reach a shelter and the x-axis represents the distance from Gaza.Each circle represents a set of participants sharing the same effective time to reach a shelter and distance from Gaza, where the area of the point is proportional to the size of the set.The grey area represents areas out of Gaza missile range during this war cycle.This return to baseline values remained stable also in the succeeding two weeks period (R2).This remarkable recovery is also evidenced by the data presented in Supplementary Figure 4, showing that the changes between the baseline period (B) and the first "back to routine" period (R1) are distributed roughly normally around 0, with a relatively small standard deviation, and seemingly symmetric tails.

8/12
Supplementary Figure 5 shows an example of a participant who does not seem to recover from the war effects.As can be seen from the figure, the participant presents roughly the same changes as the general population during the war period, but does not show a return to baseline values after the war in most of the examined well-being indicators.Specifically, after the war, the reported stress level continued to rise and the reported mood level, step count, reported sport time, reported sleep duration and reported sleep quality continued to drop.In contrast, number of encounters, average heart rate and awake time do seem to return to baseline values after the war.Reported sleep quality.The x-axis represents four time periods: baseline period (B), war period (W), first "back to routine" period (R1) and second "back to routine" period (R2).The y-axis represents the mean value for the examined well-being indicator.

1 .
Data quality: (a) Number of questionnaires filled during the 53 days of the study, (b) Number of days smartwatches were worn during the 53 days of the study (c) Number of days with smartphone sensor data.Supplementary

2 .
Mean daily values for: (a) reported stress level and (b) step count.The x-axis represents time and the y-axis represents the mean daily value.The red vertical lines represent the beginning and the end of the war period, and the black vertical dashed lines indicate free days.Examination of the figure suggests the existence of a weekly rhythm, where for example, free days (weekends and national holidays) exhibit lower mean daily values than work days across both indicators.

4 .
Back to routine analysis -distribution of the difference between the first "back to routine" period (R1) and the baseline period (B) for various well-being indicators: (a) Screen time in hours (b) Reported mood level, (c) Reported stress level, (d) Reported number of encounters, (e) Step count, (f) Average heart rate in beats per minute, (g) Percentage of time still (h) Reported sport time in minutes, (i) Awake time during night sleep in seconds, (j) Sleep start hour (k) Reported sleep duration in hours, and (l) Reported sleep quality.The x-axis represents the mean difference value.The y-axis represents the kernel density estimator (KDE).

5 .
An example of a participant who does not seem to recover from the war effects in most of the examined well-being indicators: (a) Screen-on time in hours (b) Reported mood level, (c) Reported stress level, (d) Reported number of encounters, (e) Step count, (f) Average heart rate in beats per minute, (g) Percentage of time still (h) Reported sport time in minutes, (i) Awake time during night sleep in seconds, (j) Sleep start hour (k) Reported sleep duration in hours, and (l)

Table 1 .
Descriptive statistics of the Mixed ANOVA tests.Number of participants, mean and standard error for each combination of well-being indicator, time period and exposure group.Each row represents a combination of a single well-being indicator and a single exposure group (or all groups together).Columns represent the number of participants considered in the analysis and the mean and standard error for each time period.