REVIEW
Mental health implications for older adults after natural
disasters a systematic review and meta-analysis
...........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
ABSTRACT
Background: Natural disasters affect the health and well-being of adults throughout the world. There is some
debate in the literature as to whether older persons have increased risk of mental health outcomes after exposure
to natural disasters when compared with younger adults. To date, no systematic review has evaluated this.
We aimed to synthesize the available evidence on the impact of natural disasters on the mental health and
psychological distress experienced by older adults.
Design: A meta-analysis was conducted on papers identified through a systematic review. The primary
outcomes measured were post-traumatic stress disorder (PTSD), depression, anxiety disorders, adjustment
disorder, and psychological distress.
Results: We identified six papers with sufficient data for a random effects meta-analysis. Older adults were
2.11 times more likely to experience PTSD symptoms and 1.73 more likely to develop adjustment disorder
when exposed to natural disasters when compared with younger adults.
Conclusions: Given the global rise in the number of older adults affected by natural disasters, mental health
services need to be prepared to meet their needs following natural disasters, particularly around the early
detection and management of PTSD.
Key words: mental health, post-traumatic stress disorder, older persons, natural disaster, meta-analysis
during a disaster (Yang et al., 2003; Jia et al., 2010). inform the planning of age-specific services and
This is possibly because older adults are often in resources from the context of disaster.
worse health before a disaster, and less able to
seek assistance after a natural disaster, with increase
in resulting morbidity and mortality (Ticehurst Methodology
et al., 1996; Jia et al., 2010). The older adults also
appear to have a greater sense of loss, with even Search and selection strategy
a minor injury provoking the fear of loss of self- We conducted a systematic literature review of
reliance (Goldstraw et al., 2012). By contrast, other the following databases from database inception
studies have reported no age-related differences to 24 March 2014 to identify peer-reviewed
in mental health outcomes for older adults post studies reporting on age cohort differences
disaster (Knight et al., 2000). for mental health outcomes in response to a
There have been four hypotheses for older adults natural disaster: MEDLINE, CINAHL, PubMed,
varying psychological responses to natural disasters. PsycINFO, ANZRC, and Health Source Nursing
The inoculation hypothesis suggests that prior Academic Edition, Scopus, Embase, LILAC,
experiences of disaster result in decreased emotional PILOTS, and Cochrane databases. The search used
reactivity to subsequent disaster exposure, placing the following key terms: natural disaster, aged,
younger adults at high risk of adverse mental health mental Health, and well-being. The list of
outcomes (Thompson et al., 1993). The maturation comprehensive search terms is provided in Sup-
hypothesis suggests that older adults develop an plementary Table S1 (available as supplementary
adaptive coping style with greater resilience to material attached to the electronic version of this
psychological distress than younger adults (Knight paper at www.journals.cambridge.org/jid_IPG).
et al., 2000). The burden hypothesis suggests that Only publications in peer-reviewed journals and
it is the middle-aged adults, rather than the older English language were considered.
adults, who are at the highest risk of psychological
distress because of greater social and financial Inclusion and exclusion criteria
responsibilities in the wake of a natural disaster
Inclusion and exclusion criteria were established
(Thompson et al., 1993; Cherry, 2009). Lastly,
(Supplementary Table S1), and three authors
the exposure hypothesis proposes that older adults
reviewed the papers (Georgina Parker, David Lie,
are at the greatest risk of mental health outcomes
and Dan Siskind). Studies were included if they
because of a lower likelihood of receiving advanced
presented a clear hypothesis, including comparison
warnings, reluctance to evacuate, disruption to
of age cohorts as a factor for mental health outcomes
accustomed patterns of life, and a sense of
after natural disaster exposure. In addition, the
deprivation resulting from their personal, physical,
study had to clearly define age cohorts when assess-
and social losses post disaster (Ticehurst et al.,
ing the impact of a natural disaster. The literature
1996; Yang et al., 2003; Jia et al., 2010).
review included studies that used definition of age
Similarly, when attempting to define what
as more than 60 to 65 years as the cut-off for
constitutes an older adult, no one clear definition
older adults. Studies were excluded if they had
of older age is available. The most commonly cited
insufficient data, either as published in the paper or
cut-off for older adults in the literature is the age
following attempts to contact the authors, or if they
of 6065 years (Tinker, 1993). However, there
were of insufficient quality (see section on Quality
is variability within the literature. Cross-culturally,
Evaluation). As the study goal was to examine
there are differences in the cultural loading of age,
impact of natural disasters on mental health, we
with frailty and disability contributing to whether an
excluded subtypes of disasters that were secondary
individual is considered an older adult. For the
to human error, terrorism, or technological failure.
purpose of this study, we used the current United
Nations cut-off for older population, i.e. age more
than 60 years (WHO, 2014). Statistical analysis
We undertook a systematic review and meta- The data from the included papers were extracted
analysis to better determine the evidence of any of by Georgina Parker, and validated by Steve Kisely
the four competing hypotheses. In particular, we and Dan Siskind. Specific mental health outcomes
examined whether older adults are at an increased were included in the meta-analysis if they were
risk of adverse mental health outcomes, including reported in at least two included studies. We
post-traumatic stress disorder (PTSD) symptoms, used Review Manager Version 5.3 to calculate a
psychological distress, and depressive symptoms pooled estimate of the association of age cohorts
after exposure to a natural disaster compared with and mental health outcomes reported post disaster.
younger adults. This information could help to Meta-analyses were undertaken for rates of PTSD,
Mental health of older adults post natural disaster 13
depression, anxiety disorders, and adjustment dis- bias control, methodology, data extraction, and
orders. Further, sensitivity analyses were performed suitability of analysis.
to examine the effect of sudden impact (earthquakes
and tsunamis) versus advanced warning events (cyc-
lones, floods, and storms) and quality of research. Results
To check for heterogeneity between studies, both
the Q and I2 tests of heterogeneity were used as In the initial electronic searches, 1,216 citations
well as visual inspection of forest plots. The Q and were identified, of which 92 papers were potentially
I2 statistics provide an estimate of the percentage of relevant and assessed for eligibility (Figure 1). We
variability due to heterogeneity rather than chance excluded 80 papers that did not meet inclusion
alone. We used the random effects model for all the criteria, leaving 12 papers for formal review. A
analyses given the heterogeneity of the data. review of the reference lists of these 12 papers and
For any outcomes where there were at least a hand search of additional papers identified no
ten studies, we tested for publication bias using additional papers. The 12 papers were reviewed
funnel plot asymmetry where low p-values suggest for quality. Six were appraised as high, one as
publication bias. moderate, and five as of low quality. The six low
and moderate quality papers were excluded and
only the six studies assessed to be of high quality
Quality evaluation were included in the meta-analysis. Consensus was
As there is no agreed quality assessment tool reached for all the ratings.
for epidemiological prevalence studies, we adapted Studies included retrospective cohort, prospect-
a modified assessment tool based on published ive cohort, and case-control designs. The samples
STROBE and JHNEBP guidelines (von Elm et al., came from five countries, including two high-
2008; Dearholt et al., 2013) provided in Sup- income (Australia and Taiwan), one upper-middle
plementary Table S2 (available as supplementary income (China), and two lower-middle income
material attached to the electronic version of this (India and Indonesia) countries as defined by
paper at www.journals.cambridge.org/jid_IPG). the World Bank (2013). Four studies were on
The evaluation conducted by Georgina Parker, earthquakes (Ticehurst et al., 1996; Yang et al.,
David Lie, and Dan Siskind allowed for an overall 2003; Jia et al., 2010; Kun et al., 2013) and two
rating of high, moderate, or low for each study based were on tsunami (Viswanath et al., 2012; Musa et al.,
on key objectives. These included design, sampling, 2014).
14 G. Parker et al.
Jia et al., 2008 China, Cross- PTSD: 15 276 (46.4%) Random Young 1859
2010 Sichuan sectional PCL-C sampling (50%
earth- Distress: affected Old >60
quake GHQ-12 geographic (50%)
sub-districts
Kun et al., 2008 China, Cross- PTSD: HTQ 3 1,820 (48.1%) Random Young 1534
2013 Sichuan sectional and sampling (19.3%)
earth- DSM-IV affected Middle 3564
quake PTSD items geographic (58.4%)
sub-districts Old >65
(22.3%)
Musa et al., 2009 West Cross- MDD: >24 137 (69.5%) Stratified Young 1825
2014 Sumatra sectional DASS-21 cluster (26.5%)
and Aceh GAD: sampling Middle 2640
tsunami DASS-21 (27.9%)
Stress: and 4155;
DASS-21 (24.5%)
Old 5676
(21.1%)
Ticehurst 1989 Prospective PTSD: IES 624 3,007 (57.4%) Interview Young <65
et al. Australia, cohort Distress: personal (78.8%)
(1996) Newcastle GHQ-12 and neigh- Old >65
earth- borhood (22.2%)
quake damage,
#earth-
quakes
experienced
Viswanath 2004 India Cross- PTSD: >1 438 (NS) Geographic Young 1959
et al., tsunami sectional ICD-10; displace- (83%)
2012 MDD: ment Old >60
ICD-10; evacuee (17%)
GAD: shelter
ICD-10;
Adjustment
Dx
Yang et al., 1999 Cross- PTSD: 34 663 (58.4%) Case series Young <44
2003 Taiwan sectional PTSD-IV- exposed (32.4%)
earth- CL areas Middle 4565
quake Distress: (31.3%)
CHQ Old >65
(36.3%)
Notes: GAD = generalised anxiety disorder; PTSD = post-traumatic stress disorder; MDD = major depressive disorder; GHQ-12 =
General Health Questionnaire; CHQ = Chinese Health Questionnaire; PTSD-IV-CL = PTSD DSM-IV checklist; ICD-10 =
International Classification of Disease, version 10; DSM IV = Diagnostic and Statistical Manual, version IV; HTQ = Harvard Trauma
Questionnaire; IES Impact of Event Scale; PCL-C = PTSD Checklist, civilian version; DASS-21 = Depression, Anxiety and Stress
Scale 21; PD = post-disaster time between initial exposure and administration of questionnaires Where multiple times are provided, the
study administered questionnaires at those times; NS = not stated. All included studies were of high quality.
The available papers reported on disasters 12 was utilized in both validated English and
occurring between 1984 and 2009. Mental Chinese versions. Multiple measures of symptoms
health outcomes included PTSD, generalized of PTSD, and Major Depressive Disorder (MDD),
anxiety disorder (GAD), depressive symptoms, GAD were used, including both questionnaire and
and psychological distress as measured by the structured clinical interview diagnosis, as identified
General Health Questionnaire (GHQ). The GHQ- in Table 1. There was a mix of both structured
Mental health of older adults post natural disaster 15
Figure 2. (Colour online) Forest plot of mental health outcomes for young and old adults.
gender skew among older people, with female A possible explanation is that the review did not
gender a known risk factor for PTSD (Brewin et al., directly compare age groups and was not restricted
2000). to natural disasters, with the vast majority of papers
Further, the prevalence of PTSD most likely covering motor vehicle accidents, violent crime, or
differs from age cohort, country to country, and war. This points to the need to consider each type
region to region. Some research on the baseline of disaster separately and is one of the reasons that
prevalence of PTSD in non-disaster community the present study was restricted to natural disasters.
samples has identified that as age increases beyond
55 years, the prevalence of PTSD reduces (Creamer
et al., 2001). This suggests that the higher rate
Strengths and limitations
PTSD among older adults following a natural
disaster found in this review is due to the effects of Given the aging of the population, it is increasingly
the natural disaster, and not due to a baseline higher important to understand the effect of natural
prevalence when compared with younger adults. disasters on older adults. To our knowledge, this
Our findings are at odds with a previous meta- is the first published systematic review focusing on
analysis, where younger age was associated with the mental health sequelae for older adults exposed
greater PTSD symptoms (Brewin et al., 2000). to natural disaster. Although this result is obtained
Mental health of older adults post natural disaster 17
from only small number of papers, the outcome A variety of measures were used for diagnostic
is from an accumulated sample size of over 5,700 case finding. One study (Jia et al., 2010) used a
participants. published symptom checklist, the PTSD Checklist
There are several limitations to our study. First, (Blanchard et al., 1996). Three studies used
there are constraints common to epidemiological unpublished checklists, with two (Yang et al., 2003;
studies of disasters in general. Disaster research is Kun et al., 2013) based on Diagnostic and Statistical
a relatively new phenomenon with a relative lack Manual, 4th Edition (DSM-IV) (American
of standardized methodologies (Rodriguez et al., Psychiatric Association (APA), 2000) and one
2006). This diversity in study design was reflected (Viswanath et al., 2012) based on International
in the studies we found in our search, particularly Classification of Diseases, version 10 (ICD-10)
in relation to differences in the type and severity of (WHO, 2004). Another (Ticehurst et al., 1996)
disaster, level of dose exposure, intervals between used a self-report scale of distress (Horowitz et al.,
exposure and assessment, sampling methodology, 1979). There are psychometric issues to consider
and outcome measures. The review identified a when using symptom checklists as surrogates for
need for standardized validated scales that can be diagnostic case-finding, the gold standard being a
used not only cross culturally, such as the GHQ, semi-structured diagnostic interview.
but have been also validated for older populations, One study used ICD-10, while four others used
such as the Geriatric Anxiety Inventory (Pachana DSM-IV criteria. As these criteria are divergent,
et al., 2007). there is a risk of discrepancies on case inclusion
This analysis chose to examine specifically the between studies (Peters et al., 1999). The majority
relationship between older adults and natural of the scales used in the included studies were
disasters. In the aftermath of 9/11 there have not validated, which may increase the risk of false
been significant research investments examin- positive or false negative cases of PTSD. Even for
ing the response of older adults to human- validated scales, scales that have been validated in
made/technological disasters (Bonanno et al., working-age adults may not necessarily be suitable
2007). Promising results have emerged from this for older individuals. This highlights the importance
research supportive of middle-age adult vulnerab- of considering scales that have been validated not
ilities and older adult resilience. However, it has only cross-culturally but also to account for age-
been documented that human-made/technological related variability, as has been done for the Geriatric
disasters may have different and more marked con- Anxiety Inventory (Pachana et al., 2007).
sequences than natural disasters on mental health It is possible that the discrepancies between case
outcomes (Norris et al., 2002a). For this reason, this finding instruments may have contributed to the
analysis excluded research from significant human- heterogeneity we observed in the meta-analysis.
made or technological disaster events in an attempt Although we attempted to take heterogeneity into
to capture the differential vulnerability of older account using a random-effects model, our results
adults with natural disasters alone. Our group is should be treated with caution.
currently undertaking a meta-analysis of the effects Two studies reported a three to four months time
of human-made disaster on older persons. gap between the traumatic event and the screening
In common with gerontic research in general, for PTSD (Yang et al., 2003; Viswanath et al.,
there was a lack of standardized age range and 2012), while another reported a 15-month time
common understanding of biological, social, and gap (Jia et al., 2010). The other three studies did
psychological factors that contribute to the concept not report the time gap. DSM-IV criteria specify a
of old age (Roebuck, 1979). It has been observed minimum six-month delay between the traumatic
that effects of age on post-disaster outcomes may event and a diagnosis of PTSD (APA, 2000). Any
vary across cultures even for the same type of diagnosis prior to this time would be an acute stress
disaster (Norris et al., 2002b), as what constitutes reaction. Based on these diagnostic criteria, it is
midlife in one cultural settings may be old in possible that at least two studies were reporting
another. There is no one definition for older age acute stress reaction rather than PTSD. For studies
in some settings as it has been shown to be more with a longer time gap between traumatic event
appropriate to use lower cut-off ranges for defining and screening, there is the possibility that some
an older adult to account for cultural variability, cases of PTSD may have spontaneously remitted
frailty, cognitive, and physical disability differences (Chapman et al., 2012). However, one study (Musa
(WHO, 2014). Several studies were excluded due et al., 2014) reported on rates of depression, anxiety,
to lower age cut-off for older adults ranging from and stress, but not PTSD, in two geographically
4555 years. This was done with the intention to distinct regions that were exposed to separate
capture the response of older adults rather than natural disasters five years apart from each other.
the adult population alone. Rates of reported mental health disorders were not
18 G. Parker et al.
significantly different in the two regions. These validated scales that can be utilized cross-culturally
differences in time gap may have also contributed to and have been validated for age-specific differences.
the heterogeneity of our results, which again should Given the potential confounding effects of gender,
be treated with caution, even with the use of a disability, prior trauma, and bereavement, these
random-effects model. will need to be controlled for when analyzing the
A final limitation is that we were only able to contribution of older age to the epidemiology of
identify a small number of studies, and these were disaster-related PTSD.
all of sudden impact disasters. This meant that Further research should aim to understand
we were unable to examine whether other types of that there may be age-specific differences in the
natural disaster of slower impact, such as flooding, prevalence of PTSD and other target disorders
had a different effect on mental health outcomes. prior to exposure to natural disasters. Better studies
In addition, we were not able to assess the grade of examining the impact of age among those exposed
exposure to the traumatic event as only one study to disasters could look to include a control sample
(Kun et al., 2013) provided data on exposure by of non-exposed participants from the same country
age. We were unable to test for gender effects as no of relevant ages or provide details of the known
studies reported data on age by gender. The small prevalence data for that country. Studies could also
number of studies also meant that we were unable look to link some baseline data for participants in
to test for publication bias. each age cohort, i.e. those enrolled in other studies
prior to affect to assess post-disaster outcomes.