Increased antidepressant prescriptions following the 2018 Japan floods: a longitudinal analysis using the national health insurance database

ABSTRACT

Background: Natural disasters are known to affect the mental health of the victims; however, the understanding of their impact on real-world clinical practice remains insufficient.

Objective: This study aimed to evaluate the effects of the 2018 Japan floods, one of the largest disasters in Japan’s recorded history, on antidepressant prescriptions over time.

Method: Prescription data from the medical insurance claims database covering three prefectures that accounted for 90% of flood damage were analyzed for the years before and after the floods. Participants were categorized as disaster victims or nonvictims based on local government designations. A difference-in-differences analysis compared the trends in antidepressant prescriptions between victims and non-victims during the period surrounding the floods.

Results: Of 5,000,129 participants, 31,235 were disaster victims. Victims were more likely to be prescribed antidepressants after the disaster than nonvictims (p < .001). This trend peaked 2–3 months after the disaster (adjusted Ratio of Odds Ratios [ROR], 1.13; 95% confidence interval [CI] 1.07–1.20) and persisted up to 1 year later (adjusted ROR, 1.20; 95% CI 1.12–1.28). Among antidepressants, noradrenergic and specific serotonergic antidepressants (NaSSAs) and serotonin antagonist and reuptake inhibitors (adjusted ROR 1.47, 1.49; 95% CI 1.21–1.80, 1.22–1.83) were particularly prescribed more frequently among victims. When limited to those who had not used antidepressants before the disaster, NaSSAs (adjusted ROR 2.56; 95% CI 2.14–3.07, p < .001) were conspicuously more prescribed.

Conclusions: The floods were associated with an increase in antidepressant prescriptions, suggesting the development of disaster-related mental health conditions such as depression and post-traumatic stress disorder. The need for care became pronounced 2–3 months after the event and persisted for 1 year. These findings highlight the need for psychiatric drug treatment among disaster victims and emphasize the importance of identifying appropriate timing for such interventions.

HIGHLIGHTS

• Floods are hypothesized to affect mental health care for victims. We tested this hypothesis using data from the national health insurance database.

• The 2018 Japan floods were associated with increased antidepressant prescriptions among government-designated disaster victims, with the effects emerging 2–3 months after the disaster and persisting for at least 1 year.

• These findings underscore the heightened need for psychiatric interventions among flood victims. Temporal changes in this need should be considered in the planning and delivery of psychiatric care.

1. Background

Flooding is the most common natural disaster worldwide, with over 2,900 flood events occurring between 2001 and 2018, resulting in significant economic damage (Lee et al., 2020). Additionally, flooding has severe adverse effects on the mental health of affected individuals, with reports of worsened symptoms of depression, post-traumatic stress disorder (PTSD), and generalized anxiety disorder (Fernandez et al., 2015; Stanke et al., 2012). These symptoms may occur simultaneously and can persist over the long term, potentially developing into a chronic condition. Furthermore, it has been suggested that these symptoms could lead to other mental health issues, such as increased smoking, alcohol, and drug use (Turner et al., 2013). A study has raised concerns about the increased use of highly addictive medications like benzodiazepines for sleep disorders triggered by flood damage (Okazaki et al., 2022). Furthermore, symptoms of depression and PTSD are not only short-term but may persist and become chronic over time (Dai et al., 2017; Fernandez et al., 2015). Factors that can affect mental health following a flood include age, sex, low socioeconomic or educational status, interpersonal relationships, loss of family members, risks to the physical health of individuals and their families, loss of means for stress relief, inadequate compensation for flood damage to homes, disruption of social services such as electricity and water, and living conditions in shelters (Lowe et al., 2013; Tempest et al., 2017).

Although the impact of flooding on mental health has been reported, its impact on real-world clinical practice, particularly drug treatment for the victims, is not known. In addition, many existing studies that showed the relationship between flooding and mental health exhibit several methodological limitations. For example, Studies that define the disaster victims based on distance from the disaster area may treat individuals who are not directly affected by the disaster as victims. Moreover, studies that focus solely on statistical data from residents of affected areas without identifying individuals are susceptible to ecological fallacies, implying that these findings may not be applicable to individuals. Additionally, preexisting temporal trends before the disaster make it difficult to precisely determine whether the post-disaster results were directly attributable to the disaster itself. Therefore, examining the changes in trends before and after a disaster is crucial.

To address these methodological issues and accurately measure the impact of flooding on mental health care, this study compared trends in antidepressant prescriptions among individuals living in flood-affected areas of western Japan during the 2018 flooding disaster. We compared the data of individuals officially recognized as flood victims by local governments to the data of non-victims. We hypothesized that sudden changes in living conditions caused by floods would exacerbate mental health issues, leading to a gradual increase in antidepressant use over time in the affected group. From the perspective of pharmacoepidemiology, this study aimed to provide a detailed examination of the relationship between flooding and mental health.

2. Methods

2.1. Study design

This study was a retrospective cohort study.

2.2. Universal health insurance coverage

The Japanese universal health insurance system stipulates that the insured person’s co-payment rate is generally 30%. However, for children under the age of 6 years and older individuals between the ages of 70 and 74 years, the co-payment rate is reduced to 20%. For those aged 75 years and older, the co-payment rate is further reduced to 10%. Nevertheless, individuals aged 70 years and above who are classified as ‘those with income equivalent to that of working-age individuals’ are required to pay a co-payment of 30%. Many prefectures and municipalities have established their own medical expense assistance programmes, which reduce or fully cover the co-payment amounts for children. Additionally, when a policyholder’s monthly out-of-pocket expenses exceed a predetermined threshold, the excess amount is reimbursed under the high-cost medical expense reimbursement system.

2.3. Data source

This study was conducted using the National Health Insurance Claims Database (NDB), a Japanese administrative claims database. The database includes age categories, gender, and the medical institutions used by the patients. Disease diagnoses, socioeconomic statuses, and patient addresses were not included. The NDB contains almost all information on medical services in Japan but excludes those who receive public income support and those who are covered by liability insurance or workers’ compensation insurance. The exclusion rate is minuscule at <2% and does not significantly affect the study results.

2.4. Setting

The disasters targeted in this study were heavy rainfall, subsequent flooding, and landslides that occurred primarily in western Japan from June 28 to July 8, 2018. As of January 2019, there were 237 confirmed deaths, 8 missing persons, 432 injuries, 6,767 completely destroyed homes, and 11,243 partially destroyed homes (Cabinet Office, Government of Japan, 2019). This event was recorded as one of the largest disasters in Japan’s history.

Individuals directly affected by flooding were tagged as disaster victims by their local municipality and assigned a victim code if they met one of the following criteria: (1) their homes were completely or partially destroyed, burned, or flooded, and (2) a family member who had financially supported the victim was killed, injured or missing, or had no income. Although it is difficult to determine the amount of damage suffered by individual survivors, several previous studies (Joseph et al., 2014; Keya et al., 2023) suggest that worsening economic conditions have a negative impact on the mental health of survivors. From this, it can be inferred that those who meet this criterion may have experienced economic losses severe enough to affect their mental health. Those recognized as victims were exempt from medical service costs, and this relief measure was effective until June 2020.

The observation period for this study was from July 2017, 1 year before the disaster, to June 2019, 1 year after the disaster. Participants included adults aged ≥20 years who sought care or were hospitalized in the Hiroshima, Okayama, and Ehime Prefectures, which were heavily affected by the 2018 flooding. Children were excluded due to different sources of mental health impact and prevalence of mental disorders compared with adults (Stanke et al., 2012). Individuals with a victim code were assigned to the victim group, whereas others were assigned to the non-victim group. Data extracted from the NDB included gender, age categories (20–39, 40–59, 60–79, and ≥80 years), names of prescribed antidepressants, the number of patients receiving prescriptions, and the number of days of prescriptions.

2.5. Antidepressant prescriptions

All antidepressants prescribed under Japan’s health insurance system are registered in the NDB. Antidepressants were categorized as depicted in Table 1.

Table 1.

List of drug category in antidepressants.

Category [N06A; antidepressant]	Name of antidepressant
tricyclic antidepressants	Amitriptyline, Amoxapine, Clomipramine, Dosulepin, Imipramine, Lofepramine, Nortriptyline, Trimipramine
tetracyclic antidepressants	Maprotiline, Mianserin, Setiptiline†
SSRIs	Escitalopram, Fluvoxamine, Paroxetine, Sertraline
SNRIs	Duloxetine, Milnacipran, Venlafaxine
NaSSAs	Mirtazapine
SARIs	Trazodone

Abbreviations: SSRIs = Selective Serotonin Reuptake Inhibitor, SNRIs = Serotonin-Norepinephrine Reuptake Inhibitors, NaSSAs = Noradrenergic and Specific Serotonergic Antidepressants, SARIs = Serotonin Antagonist and Reuptake Inhibitors.

We categorized each antidepressant based on the Anatomical Therapeutic Chemical classification system.

†This drug is not listed in the Anatomical Therapeutic Chemical code because it was approved in Japan and is primarily used within that country.

Furthermore, participants were classified into two groups – non-previous users and existing users – based on whether they used antidepressants during the year prior to the disaster, from July 2017 to June 2018. Non-previous users were defined as those who were never prescribed antidepressants during this period, whereas existing users were defined as those who received at least one prescription for any antidepressant. This study was based on Japan's medical insurance database, in which antidepressants are available only through prescriptions. Therefore, our study tracked antidepressant use in affected regions, allowing for an accurate assessment of the impact of floods on mental health.

2.6. Ethics

Owing to the high level of anonymity of the NDB data, the need for informed consent was waived. This study was approved by the Hiroshima University Epidemiology Research Ethics Committee (Approval Number E-1688) and received permission to use the data from the Ministry of Health, Labour and Welfare (Approval Number 1223-2).

2.7. Outcomes

2.7.1. Primary outcome

The primary outcome was the monthly antidepressant prescriptions for all participants during the observation period.

2.7.2. Secondary outcome

The secondary outcomes included: (1) new antidepressant prescriptions among non-previous users after the disaster, and (2) changes in prescription patterns for existing users.

New prescriptions for non-previous users were defined as any instance of an antidepressant prescription occurring at least once after a disaster. Prescription counts were recorded monthly, and any month with two or more prescriptions was counted as one occurrence. If two or more types of antidepressants were simultaneously prescribed, each type was counted separately.

2.7.3. Data analysis

We assessed how much higher the monthly rates of antidepressant prescriptions were among disaster victims compared that in non-victims. Continuous variables are expressed as medians (interquartile ranges), while categorical variables are presented as counts (%). Chi-squared tests were used to compare categorical variables between victims and non-victims. The odds ratios (OR) indicate the ratio of the occurrence probability of antidepressant prescriptions. OR and 95% confidence intervals (CI) were calculated using a multivariate logistic regression model adjusted for age and sex as confounding factors.

Additionally, we examined whether the prescribed dosage of antidepressants adhered to the dosages indicated in the product labelling approved in Japan for each drug. Previous studies (Cherrie et al., 2020) have employed a similar approach, and we investigated the proportion of prescriptions in line with the labelled dosage. A comparison between disaster victims and non-victims was made using chi-square tests.

To investigate whether the disaster genuinely affected the antidepressant prescription rates, we conducted a difference-in-differences (DID) analysis. This analysis compared the pre- and post-disaster prescription rates for victims and non-victims using a quasi-experimental approach. This method accounts for baseline differences between the groups and variations over time. In our DID analysis, we evaluated existing users and all participants, adjusting for age and sex. Generally, disaster victims may have higher disease rates before the disaster due to socioeconomic factors such as low income. Additionally, disease occurrence can change over time regardless of the disaster status. Therefore, adjusting for these confounders was essential, and we used a DID analysis to eliminate the effects of other background factors. Thus, we compared the pre-disaster prescription trend of victims with the post-disaster trend to investigate any increase and then compared this with the pre- and post-disaster trend among non-victims to quantify the true impact of the disaster. In this study's DID analysis, the first comparison was made among all participants who had utilized medical services at the target institution. The second step involved examining those who were already using antidepressants in the year prior to the disaster. The DID analysis assumes that the trends in the year before the disaster for both groups (those who later became disaster victims and those who remained non-victims) were parallel. Therefore, the parallel trends assumption was tested before conducting the analysis using the STATA logistic command with clustered robust standard errors to account for repeated measures between individuals. The model for the DID analysis is as follows:

$$\log \left({\displaystyle \frac{p}{1-p}}\right)={\beta }_0+{\beta }_1\mathrm{age}+{\beta }_2\mathrm{sex}+{\beta }_3\mathrm{victim}+{\beta }_4\mathrm{month}+{\beta }_5(\mathrm{victim}\times \mathrm{month})$$

In addition, we assumed that the differences between victims and non-victims in a non-disaster context were consistent, visually confirming the parallel trends in antidepressant prescriptions in the year before the disaster. In this analysis, ROR refers to the ratio of the odds ratios between the post-disaster and pre-disaster periods. We compared the occurrence of antidepressant prescriptions in the months following the disaster for both victims and non-victims using July 2017 as the baseline to present adjusted RORs with 95% CIs. Thus, the adjusted ROR in each month indicates an increase relative to the adjusted ROR from July 2017.

For individuals who had not used antidepressants prior to the disaster, we could not assess the parallelism of the pre-disaster trends. As a result, standard logistic regression was used instead of the DID analysis to calculate the OR for the entire year. The presence or absence of antidepressant prescriptions within 1 year after the disaster was the outcome of interest, and the impact of the disaster was examined using logistic regression, adjusting for gender and age.

Furthermore, as a sensitivity analysis, we examined only the prescriptions that adhered to the dosages specified in the product labelling as the outcome. A DID analysis was then performed on all participants to assess the disaster's impact, and we confirmed the consistency of the results’ directionality.

All statistical analyses were performed using STATA/MP version 16 (StataCorp, 2019), and a two-tailed p-value of <.05 was considered statistically significant.

3. Results

3.1. Overall characteristics

Of the 6,176,299 individuals registered in the NDB, 5,000,129 met the inclusion criteria, and 31, 235 (0.6%) were classified as disaster victims. The baseline characteristics of the study participants are shown in Table 2. The victim group had a higher proportion of individuals aged ≥60 years and women (56.5%). Among the participants, 4.0% were existing users of antidepressants during the year prior to the disaster. Before the disaster, the victim group tended to receive antidepressant prescriptions more often than the non-victim group (4.7% vs. 4.0%, p < .01). Regarding the types of antidepressants, before the disaster, the victim group was more likely to be prescribed Selective Serotonin Reuptake Inhibitors (SSRIs) (2.2% vs. 1.8%, p < .01) and Serotonin–Norepinephrine Reuptake Inhibitors (SNRIs) (1.4% vs. 1.1%, p < .01) compared to the non-victim group. This trend continued after the disaster, with an increase in the prescriptions of tetracyclic antidepressants (0.3% vs. 0.2%, p < .01), Noradrenergic and Specific Serotonergic Antidepressants (NaSSAs) (0.8% vs. 0.5%, p < .01), and Serotonin Antagonist and Reuptake Inhibitors (SARIs) (1.1% vs. 0.7%, p < .01). Overall, while the rate of antidepressant prescriptions increased after the disaster, the increase was more pronounced in the victim group (5.9%–4.7% = 1.2%) than in the non-victim group (4.2%–4.0% = 0.2%; p < .01).

Table 2.

Baseline characteristics of participants.

	Victims N = 31,235	Non-victims N = 4,968,894	p value
Male sex, n(%)	13,583(43.5)	2,304,674(46.4)	<.01
Age group, n(%)			<.01
20–39 years	4,654(14.9)	1,302,881(26.2)
40–59 years	7,741(24.8)	1,513,019(30.5)
60–79 years	13,681(43.8)	1,540,583(31.0)
80- years	5,159(16.5)	612,411(12.3)
Antidepressant user status, n(%)			<.01
Non-user	29,778(95.3)	4,770,079(96.0)
Continuous user	1,457(4.7)	198,815(4.0)
Antidepressant prescriptions, n(%)
Tricyclic antidepressants
Pre disaster	180(0.6)	29,329(0.6)	.75
Post disaster	202(0.7)	27,761(0.6)	.04
Tetracyclic antidepressants
Pre disaster	72(0.2)	8,312(0.2)	.01
Post disaster	79(0.3)	7,908(0.2)	<.01
SSRIs
Pre disaster	675(2.2)	91,327(1.8)	<.01
Post disaster	740(2.4)	89,950(1.8)	<.01
SNRIs
Pre disaster	448(1.4)	56,368(1.1)	<.01
Post disaster	595(1.9)	61,957(1.3)	<.01
NaSSAs
Pre disaster	154(0.5)	24,902(0.5)	.84
Post disaster	255(0.8)	25,279(0.5)	<.01
SARIs
Pre disaster	198(0.6)	29,465(0.6)	.35
Post disaster	336(1.1)	34,474(0.7)	<.01
Total
Pre disaster	1,457(4.7)	198,815(4.0)	<.01
Post disaster	1,847(5.9)	206,405(4.2)	<.01

Abbreviations: SSRIs = Selective Serotonin Reuptake Inhibitor, SNRIs = Serotonin-Norepinephrine Reuptake Inhibitors, NaSSAs = Noradrenergic and Specific Serotonergic Antidepressants, SARIs = Serotonin Antagonist and Reuptake Inhibitors.

3.2. Trends of participants’ antidepressant prescriptions

To examine monthly trends in antidepressant prescription rates among all participants before and after the disaster, we conducted a DID analysis. The rates were visually parallel in both groups before the disaster. However, after the disaster, only the victim group showed an upward trend in prescription rates, with the most significant increase occurring 2–3 months post-disaster. Although the increase slowed, it continued for up to 1 year (Figure 1). When classified by the type of antidepressant, the two groups also appeared visually parallel before the disaster, confirming the assumption for the DID analysis (Figure 2). Consistency of direction was also confirmed in the results of the DID analysis for all participants in the sensitivity analysis with each antidepressant prescription in compliance with dosage and administration as the outcome (Supplemental Figures 1 and 2).

Figure 1.

Percentage of subjects who were prescribed antidepressants each month between July 2017 and June 2019. Footnote: The numbers on the horizontal axis indicate the number of months that have elapsed before and after the disaster.

Figure 2.

Percentage of subjects who were prescribed each type of antidepressant each month. SSRIs; Selective Serotonin Reuptake Inhibitor. SNRIs; Serotonin-Norepinephrine Reuptake Inhibitors. NaSSAs; Noradrenergic and Specific Serotonergic Antidepressants. SARIs; Serotonin Antagonist and Reuptake Inhibitors.

3.3. Participants and existing users’ antidepressant prescriptions

To assess pre- and post-disaster monthly trends in antidepressant prescriptions among all participants and existing users, a DID analysis was conducted. The results are shown in Figure 3 and Supplemental Tables 1 and 2. For all participants, the rate of antidepressant prescriptions for the victim group compared to the non-victim group increased most rapidly 2–3 months after the disaster (adjusted ROR 1.13; 95% CI 1.07–1.20) and continued to rise gradually to the highest value 1 year post-disaster (adjusted ROR 1.20; 95% CI 1.12–1.28). In contrast, while existing users showed a similar trend after the disaster, the increase was not statistically significant compared to the pre-disaster rates.

Figure 3.

Adjusted ratio of odds ratios of monthly antidepressant prescriptions for victims vs. non-victims in the period surrounding the flood: a difference-in-difference analysis.

3.4. Types of prescribed antidepressants

To compare the types of prescribed antidepressants before and after the disaster, a DID analysis was conducted. The results are shown in Figure 4. Both participants and existing users demonstrated a significant increase in the prescription rates of NaSSAs and SARIs in the post-disaster victim group. This suggests that prescription patterns vary by medication type and may reflect the mental state and symptoms associated with the disaster.

Figure 4.

Adjusted ratio of odds ratios of each type of antidepressant for victims vs. non-victims: a difference- in-difference analysis. SSRIs; Selective Serotonin Reuptake Inhibitors. SNRIs; Serotonin-Norepinephrine Reuptake Inhibitors. NaSSAs; Noradrenergic and Specific Serotonergic Antidepressants. SARIs; Serotonin Antagonist and Reuptake Inhibitors.

3.5. Non-previous users’ antidepressant prescriptions

To compare trends in antidepressant prescriptions between victims and non-victims among non-previous users, a logistic regression analysis was performed to calculate ORs. The results are shown in Figure 5. The results indicate that all types of antidepressants were prescribed at significantly higher rates in the disaster victims, with NaSSAs showing the highest point estimate (adjusted ROR 2.56; 95% CI 2.14–3.07, p < .001). The disaster victims also tended to use antidepressants for longer periods after the disaster (Table 3).

Figure 5.

Odds ratios of each antidepressant for victims vs. non-victims among non-previous users: a logistic regression analysis.

Table 3.

Number of days each antidepressant was prescribed among new users after the disaster.

	Victims N = (29,778)	Non-victims N = (4,770,079)	p value
Antidepressant prescriptions, n(%)
Tricyclic antidepressants			<0.001
No prescriptions	29,699(99.7)	4,762,485(99.8)
1–30 days	33(0.11)	3,501(0.07)
31–120 days	23(0.08)	2.434(0.05)
121- days	23(0.08)	1,659(0.03)
Tetracyclic antidepressants			<0.001
No prescriptions	29,754(99,9)	4,768,606(99.9)
1–30 days	11(0.04)	789(0.02)
31–120 days	*(*)	410(0.01)
121- days	*(*)	274(0.01)
SSRIs			<0.001
No prescriptions	29,544(99.2)	4,747,203(99.5)
1–30 days	97(0.33)	9,290(0.19)
31–120 days	70(0.24)	7,480(0.16)
121- days	67(0.22)	6,106(0.13)
SNRIs			<0.001
No prescriptions	29,486(99.0)	4,744,525(99.5)
1–30 days	136(0.46)	11,384(0.24)
31–120 days	101(0.34)	8,324(0.17)
121- days	55(0.18)	5,846(0.12)
NaSSAs			<0.001
No prescriptions	29,658(99.6)	4,763,173(99.9)
1–30 days	54(0.18)	3,197(0.07)
31–120 days	42(0.14)	2,196(0.05)
121- days	24(0.08)	1,513(0.03)
SARIs			<0.001
No prescriptions	29,623(99.5)	4,756,298(99.7)
1–30 days	69(0.23)	6,735(0.14)
31–120 days	52(0.17)	4,173(0.09)
121- days	34(0.11)	2,873(0.06)
Total			<0.001
No prescriptions	28,999(97.4)	4,700,601(98.5)
1–30 days	322(1.08)	28,678(0.60)
31–120 days	242(0.81)	22,403(0.47)
121- days	215(0.72)	18,397(0.39)

Abbreviations: SSRIs = Selective Serotonin Reuptake Inhibitor, SNRIs = Serotonin-Norepinephrine Reuptake Inhibitors, NaSSAs = Noradrenergic and Specific Serotonergic Antidepressants, SARIs = Serotonin Antagonist and Reuptake Inhibitors.

*Values less than 10 that were masked owing to the NDB Japan rules.

4. Discussion

This study demonstrated that following the 2018 Japan Floods, disaster victims were more likely to be prescribed antidepressants than non-victims. This effect remained significant even after accounting for pre-disaster antidepressant usage trends. Moreover, this trend did not emerge immediately after the disaster but began to increase 2–3 months after the event, persisting for at least 1 year.

In this study, it was observed that a significant proportion of the affected population consisted of elderly individuals, women, and existing users of antidepressants. Previous research has suggested that women may have lower resilience to recovery after a disaster compared to men (Goldmann & Galea, 2014), and elderly individuals are more prone to developing symptoms such as PTSD, adjustment disorders, and sleep-related problems when exposed to natural disasters compared to younger adults (Balikuddembe et al., 2024). According to previous research (Lowe et al., 2013), characteristics of flood victims include lower income levels and social status, which are also recognized as risk factors for depression. Since areas prone to flooding, such as rural regions and impoverished communities, are inherently more vulnerable to disasters, the higher proportion of existing SSRI users among the victim group may be associated with these environmental and socioeconomic factors.

The types of antidepressants prescribed differed between before and after the disaster. Among existing users, SSRIs and SNRIs were more commonly prescribed before the disaster, whereas after the disaster, there was an increase in the prescription of NaSSAs and SARIs in addition to SSRIs and SNRIs. An analysis limited to non-previous users revealed that NaSSAs were the most frequently prescribed antidepressants. NaSSAs stimulate the release of both norepinephrine and serotonin, thereby exerting antidepressant effects and reducing anxiety. Additionally, they can cause sedation and appetite stimulation, making them useful for treating insomnia and appetite loss associated with depression and anxiety disorders. Compared to traditional SSRIs and SNRIs, NaSSAs have the advantages of a shorter onset of action and fewer gastrointestinal symptoms and sexual dysfunction (Fawcett & Barkin, 1998). Although SARIs are commonly used to treat delirium, they are also used to reduce anxiety and treat insomnia in patients with depression (Fagiolini et al., 2023; Wada et al., 2018). The differing prescription trends before and after the disaster likely reflect the emergence of mental health symptoms, such as insomnia and anxiety, among disaster victims. Additionally, people with a history of exposure to trauma, such as previous natural disasters, are more likely to experience worsening mental health issues such as depression, PTSD, and anxiety disorders when they are affected by subsequent natural disasters (Johnson et al., 2024; Lowe et al., 2019). It has also been suggested that a history of mental illness can serve as a risk factor for the development of PTSD (Cameron et al., 2006; Mason et al., 2002). Therefore, in the case of existing users, the psychological vulnerability associated with pre-existing conditions may have contributed to the increase in antidepressant prescriptions.

Existing mental health conditions may have worsened as well, leading to the addition of new prescriptions. Furthermore, as it takes several weeks for PTSD or depression to be formally diagnosed (Du et al., 2022), the acute phase of a disaster may not immediately warrant the prescription of traditional medications. The increased prescription of all types of antidepressants among non-previous users further supports previous research suggesting that flooding can trigger the onset of mental health disorders, including depression, PTSD, and anxiety disorders. Thus, the findings of this study not only confirm the established fact that flooding worsens mental health and precipitates the onset of psychiatric disorders but also suggest that certain symptoms are more likely to emerge or worsen following such disasters, contributing to the onset or exacerbation of both new and pre-existing mental health conditions.

Another notable finding in this study was the timing of the increase in antidepressant prescriptions. During the acute phase of a disaster, individuals may experience acute stress disorder (ASD), which typically lasts for more than 3 days after exposure to a stressor and generally improves within a month. However, in cases where improvement is limited, ASD may progress to PTSD (Bryant et al., 2011). The onset of PTSD is not specific but typically emerges around 3 months following a traumatic event (Koren et al., 1999; Visser et al., 2017). The adaptive response to stress following a disaster exhibits temporal changes, as described by the general adaptation syndrome and typical disaster phases (Raphael, 2008; Selye, 1955). During the initial alarm reaction phase immediately after the disaster, individuals experience intense shock, confusion, anxiety, and fear. In this phase, behavioural abnormalities such as hyperactivity are more likely to occur than are mood disorders (Mellsop et al., 2010). Consequently, there is often an increase in the prescription of low-dose antipsychotic medications to address behavioural disturbances, agitation, and sleep disorders caused by acute stress disorder during such events (Rossi et al., 2011; Trifirò et al., 2013). Another study focusing on the same flood disaster showed an increase in benzodiazepine prescriptions immediately after the event, suggesting an association with sleep disturbances (Okazaki et al., 2022). Subsequently, the body and mind enter a phase of resistance as they attempt to adapt to the stress; however, prolonged exposure to stress eventually leads to exhaustion and a breaking point is reached, resulting in symptoms such as anxiety, depression, and apathy. This phase corresponds to the mid-to-late post-disaster period, during which depressive symptoms and PTSD may become chronic, necessitating antidepressant prescriptions. Based on these findings, one possible explanation for the delayed increase in antidepressant prescriptions in this study is that prescriptions of hypnotic sedatives for acute stress reactions, sleep disturbances, and behavioural abnormalities may have increased immediately after the disaster, while prescriptions of antidepressants may have increased 2–3 months after the disaster when the acute symptoms did not improve and became chronic with depression and PTSD.

Natural disasters other than flooding, such as wildfires, earthquakes, and cyclones, have also been associated with higher rates of PTSD, anxiety disorders, and depressive symptoms, along with an increase in antidepressant prescriptions (Kim et al., 2023; Usher et al., 2012; Wettstein & Vaidyanathan, 2024). However, studies evaluating the mental health impact of disasters, particularly from the perspective of prescribed medications, in flood-affected areas are rare. One study identified disaster victims based on the addresses of patients registered at clinics in affected regions and used a segmented time-series design to assess the relationship between antidepressant prescriptions and mental health outcomes following flood exposure (Milojevic et al., 2017). However, this study classified participants into two groups based on proximity to the flood-affected area rather than on whether they were directly impacted by the disaster. Additionally, the study did not account for prescription trends prior to the disaster, thereby limiting its ability to capture changes in prescription patterns both before and after the disaster. In response to these challenges and based on the assumption that time-series data before and after a flood share a common trend, we employed DID analysis to address selection and time-related biases. This allowed us to continuously assess the mental health impact of flood exposure through changes in antidepressant prescription patterns before and after flooding.

4.1. Limitations

First, the study relied on prescription data and did not evaluate the detailed medical records, diagnoses, traumatic experiences or comorbidities for each individual. Therefore, we cannot arrive at a definite conclusion on the increase in incidence or worsening of certain mental conditions. Furthermore, we cannot exclude the confounding effect of comorbidities. Previous research has suggested that physical conditions such as infections and skin diseases (Tunstall et al., 2006). Also, antidepressants are sometimes prescribed for non-mental conditions, including certain types of pain. Research utilizing the NDB carries certain limitations; however, it is considered an effective approach to addressing critical issues in national healthcare policy, particularly regarding the appropriate use of pharmaceuticals, healthcare quality, and medical economics (Matsuda, 2019; Ohtera et al., 2024). As demonstrated in this study, analyzing antidepressant prescription patterns and types allows for an indirect evaluation of whether high-dependency or side-effect-prone medications are being inappropriately prescribed to disaster victims, and whether appropriate pharmacological treatments are being administered for each specific symptom. This approach can assist in identifying patient groups requiring pharmacological treatment and may contribute to the improvement of drug usage patterns (Hennessy, 2006).

Second, it is important to consider that the increased likelihood of antidepressant prescriptions may be influenced by the fact that disaster victims were no longer required to pay for treatment due to their designation as victims. However, Japan's healthcare costs are relatively low compared to those of other developed countries, and the national health insurance system significantly reduces out-of-pocket expenses for individuals. The price elasticity of outpatient medical services in Japan is generally low, reportedly ranging from −0.125 to −0.076, suggesting that the relationship between an increase in antidepressant prescriptions and full exemption from treatment costs is likely negligible (Sawamo, 2000).

Third, the timing of victim certification and the exemption or reduction of medical payments varied across municipalities. Receiving financial support is presumed to represent a clear benefit for the victims. Given that the system was based on self-reporting and allowed for retrospective certification and fee reductions, many patients likely had their payments exempted shortly after the disaster. Although the exact timing of individual victim certification could not be determined, the potential influence on healthcare utilization is likely to be limited due to the retrospective and self-reporting nature of the certification process, which allowed patients to receive exemptions shortly after the disaster. Additionally, regarding the impact of the disaster on medical institutions themselves, the possibility remains that issues such as the decline in the quality of post-disaster medical services and accessibility challenges have not been fully resolved. However, this disaster had a widespread impact, but the damage from river swelling and landslides was concentrated in specific local areas. In other words, the affected population was likely concentrated within certain regions, and the medical institutions they visited also showed similar trends. Although it was difficult to consider the specific damage to the individual medical institutions due to the nature of the data, the impact of the disaster on the affected individuals also implicitly included access to these medical institutions. Hence, it does not appear to be a major impediment to the analysis.

5. Conclusion

During the 2018 Japan Floods, an increase in antidepressant prescriptions was observed among disaster victims. This increase did not occur immediately after the disaster but rather 2–3 months later and persisted for up to 1 year post-disaster. Additionally, among the antidepressants, the prescription of NaSSAs and SARIs notably increased after the disaster, which may reflect temporal changes in the mental health status of the victims. These findings show the increased need for psychiatric drug treatment among disaster victims and provide insights into the appropriate care in post-disaster clinical practice.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Author contributions

All authors approved the final manuscript as submitted and agreed to be accountable for all aspects of the work. MH conceptualized and designed the study, carried out the initial analyses, drafted the initial manuscript, and reviewed and revised the manuscript. SY carried out the initial analyses and reviewed and revised the manuscript. SO conceptualized and designed the study and critically reviewed and revised the manuscript. NS critically reviewed and revised the manuscript. MM supervised the whole study by contributing to the study design, date collection, interpretation of date and reviewing of the manuscript.

Date sharing

This data of this research was obtained from the National Database of Health Insurance Claims and was permitted to be used by the Ministry of Health, Labor and Welfare.

Ethics approval

The requirement for informed consent was waived because the anonymous data of National Database of Health Insurance Claims was used in this study. And this study was approved by the Hiroshima University Epidemiology Research Ethics Committee (Approval Number E-1688).

Data availability statement

The data that support the findings of this study are not publicly available due to restrictions stipulated by the Ministry of Health, Labour and Welfare.

Supplemental Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/20008066.2025.2537547.