Australia’s vulnerability to natural disasters is a recurring theme in Australian communities,Reference Leivesley1 with these events occurring more frequently since the 1970s.Reference Keenan, Weston and Volkova2, Reference Filkov, Ngo and Matthews3 Disasters are accompanied by destruction of property and infrastructure, the loss of wildlife, and often loss of human lives. These were familiar occurrences following the bushfires and floods that plagued Australian communities between 2019 and 2022.Reference Filkov, Ngo and Matthews3–5 Although climate variables have a role in the genesis of disasters, poverty, previous trauma experiences, building codes, and community and individual resilience are risk factors that influence the outcome of disasters.Reference Chaudhary and Piracha6 These risks are not static. The severity of events, greater urbanization, and an aging and growing population intensify the disaster risks through increased vulnerability and a reduced response capacity in disaster-affected communities.Reference Kelman, Gaillard and Lewis7–Reference Swerts, Pumain and Denis11 The worldwide trend towards urban livingReference Leeson12 is particularly evident in Australia where 89% reside in urban areasReference Maude13 with 92% of Australians predicted to live in urban communities by 2050.Reference Jiang and O’Neill14 The population drift towards major cities, established coastal centers, or regional centers exposes communities to an increased risk of disasters due to coastal or pluvial flooding or cyclonic activity.Reference Buckle and Osbaldiston15
The adverse economic, social, family, and mental health outcomes for disaster-affected Australian communities are well described.Reference Singh-Peterson and Lawrence16–Reference Crompton and McAneney18 Jurisdictions recognize the need for structured, planned, and integrated frameworks for responding to disasters. The frameworks include addressing immediate safety and infrastructure recovery and the provision of psychosocial support, although the impact of disasters on the health and well-being of communities is likely to be underestimated and may be cumulative.Reference Morrissey and Reser19–Reference Reifels, Pietrantoni and Prati23 The Australian disaster management framework includes prevention and preparation strategies and mental health strategies to address identification of at-risk populations, service accessibility, and outreach programs.24–26
However, there are gaps between the demand for services post-disaster and service availability: Post-Hurricane Katrina, those experiencing psychological problems described service access difficulties resulting in increased unmet mental health needs. The nature of the disaster and geographical factors influence service access. An Australian study demonstrated that exposure to bushfires increases the likelihood of seeking primary mental health care compared to those exposed to other disasters.Reference Reifels, Bassilios and Spittal27 Ethnicity, income, job loss, and disability affect service access and help-seeking behavior.Reference Lowe, Norris and Galea28–Reference Wang, Gruber and Powers30
Like the flood events that affected New South Wales (NSW) and Victoria in 2022-2023, the Queensland floods and cyclones (2010-2011) were associated with significant infrastructure damage, human tragedy, and psychosocial distress. A third of Queensland’s population was affected, 10 500 people were evacuated from their homes, and towns became isolated. The flood affected the capital, provincial cities and rural communities.Reference Arklay31 An evaluation estimated 314 000 people were vulnerable to emotional distress, with a predicted 1% increase in severe mental disorders and a 5% increase in mild to moderate mental disorders.32
The Queensland Mental Health Natural Disaster Recovery Plan 2011-13 (The Plan) was developed to address psychosocial recovery in the immediate and the medium to long-term.33, 34 The Plan addressed the mental health challenges associated with evacuation, damaged homes, and infrastructure and aimed to link primary health care, the non-government sector (NGOs), community-based supports, and health services. A critical feature of the strategy was addressing the needs of vulnerable populations, providing evidence-informed treatment programs, and enhancing resilience.35, 36 The Specialised Mental Health Program (SMHP) was a key element of the mental health response.22 Across Queensland, SMHP treatment teams were implemented in areas affected by floods and cyclones. The Centre for Trauma, Loss and Disaster Recovery undertook the implementation, oversight, and monitoring of the SMHP and established a database to support staff supervision and report on service activity.37, 38
The SMHP multidisciplinary teams included mental health nurse, psychologists, social workers, occupational therapists, and psychiatrists. Clinical services were provided in community facilities, general practices, or homes. The Australian Government National Disaster Relief and Recovery Arrangements (NDRRA) funded the services. The clinicians were trained in Skills for Psychological Recovery (SPR) and Trauma-Focused Therapy.Reference Crompton, Young and Shakespeare-Finch39 The program included pre-referral assessment, standardized pre-post assessments, and clinical evaluation. The treatment sessions included psychoeducation and Cognitive Behavioral Therapy (CBT) to address anxiety and depression and trauma-focused CBT (TF-CBT).
Despite the extensive mental health response plan to the Queensland floods and cyclones of 2010-11, there were concerns regarding the program’s capacity to meet the demands for specialist mental health care due to the size of the state and the number of people affected. These concerns reflect those of other studies. Following Hurricanes Katrina and Sandy, the demand for services and the delayed emergence of adverse psychological adversely exceeded the availability of trained clinicians.Reference Lowe, Norris and Galea28, Reference Taioli, Tuminello and Lieberman-Cribbin40–Reference Ruskin, Schneider and Bevilacqua42
The mismatch between service demand and clinician availability is not unique to disasters. The development of a capacity to predict who may participate in treatment is, therefore, likely to assist in determining resource allocation and treatment planning.Reference Bone, Simmonds-Buckley and Thwaites43, Reference Uckelstam, Philips and Holmqvist44 Previous studies have highlighted witnessing injury or death as aspects of the disaster experience that increase the risk of adverse psychological outcomes, while noting indirect factors such as resilience potentially ameliorate the psychosocial effects.Reference Maguire and Hagan45–Reference Brooks, Amlôt and Rubin49 Similarly, studies have identified a relationship between optimismReference Hirsch, Wolford and LaLonde50, Reference Benight, Swift and Sanger51, perception of well-beingReference Clemens, Berry and McDermott52, and coping styleReference Benight, Ironson and Durham53–Reference Navarro, Krien and Rommel55 and the psychosocial response to disasters. Other studies recognize that psychosocial outcomes are influenced by pre-and-post disaster experiences and pre-disaster physical and general health.Reference Seplaki, Goldman and Weinstein56–Reference Lowe, McGrath and Young58
This paper reports a retrospective evaluation of the relationship between perception of optimism and resilience, disaster-related factors, physical and mental health history, family history, pre-event trauma experiences, demographics, post-disaster health, social changes, and the likelihood of participation in a specialist post-disaster mental health program (Figure 1). This study aimed to identify factors that may predict participation in a specialist program in those exposed to a natural disaster and assessed as at risk for post-traumatic stress disorder. A Classification Tree Analysis (CTA)Reference Atieh, Pang and Lian59 was utilized to identify which assessment measures predicted participation in the post-disaster SMHP.
Figure 1. Triage, intake, and discharge assessment questions.
a) If the participant experienced bereavement due to the floods or cyclones, complete CBI, and if yes referred to bereavement service.
b) Clinicians should review if self-rated questions are not answered.
c) To be completed if a participant entered the treatment program.
**) If the answer was “yes” to this question, participants were referred to the Post-disaster Bereavement Service
Method
The study evaluates data from assessments of those referred to the SMHP treatment program (n = 881) during 2012. Ethics approval was granted by Metro South Health Human Centre for Health Research Ethics Committee (HREC/14/QPAH/472) and Queensland University of Technology (Ethics approval number 1500000016) – A retrospective evaluation of the outcomes of State-wide disaster mental health programs established and delivered following the Cyclones and Floods of 2010-2011.
A standardized process was used to assess all referrals (Figure 1). A panel of experts chose, by consensus and informed by the literature, the various assessment questionnaires used to evaluate those referred to the SMHP. The assessment measures reflected the known relationship between disaster exposure and psychosocial outcomes, such as Post-traumatic Stress Disorder (PTSD), alcohol use, and intimate partner violence (IPV), and aspects such as resilience, psychological coping strategies, life history, prior trauma experience, perception of self-efficacy, mental health history, and demographic and socioeconomic factors.Reference North, Kawasaki and Spitznagel60–Reference Stewart67
Clinicians conducted the pre-treatment screening by telephone. The primary care post-traumatic stress disorder scale (PC-PTSD) was used to screen for PTSD. This measure has good test-retest reliability, with the 4 items reflective of the PTSD construct. A score of >2 on the scale indicates a person is at risk for PTSD.Reference Ouimette, Wade and Prins68 Individuals who scored >2 on the PC-PTSD were further assessed (Figure 1) in relation to their experience of the natural disaster of 2010-11. The screening assessment included a narrative description of their disaster experience and measures that focused on their perception of wellbeing using a question from the public health computerized-assisted telephone interview program (CATI)Reference Fox and Eyeson-Annan69, an individual’s perception of optimism that Abdel-KhalekReference Abdel-Khalek70, Reference Abdel-Khalek71 ascertained as identifying a relationship with coping and health outcomes and the Connor Davidson Resilience Scale (CR-2) that has been demonstrated to reliably discriminate for resilience.Reference Waddimba, Baker and Pogue72
The pre-treatment assessment included questions related to alcohol consumption, as detailed in questions 1 and 2 of the Alcohol Use Disorders Identification Test (AUDIT),Reference Reinert and Allen73 gambling behavior, and individual or family Intimate Partner Violence (IPV). An affirmative response to these questions at pre-treatment screening resulted in a more detailed evaluation during the initial assessment, which also included a clinical history (Figure 1). Alcohol use was assessed using the initial 6 questions of the Alcohol Use Disorders Identification TestReference Reinert and Allen73, with the assessment of gambling behavior (NODS) Reference Volberg, Munck and Petry74,Reference Toce-Gerstein, Gerstein and Volberg75 and STaT measure for recent partner violenceReference Paranjape, Rask and Liebschutz76 measures sensitive to identification of problem gamblers (79%) and recent IPV (94.9%), respectively. The presence or absence of suicidal ideations was assessed during screening and further explored in the clinical history during the initial assessment.
Clinicians utilized an electronic clinical record. Deidentified data were collated and entered for analysis using IBM SPPS (v23). The data were grouped for analysis into 1) demographic variables (age, gender, income source, marital status, education, accommodation), 2) exposure variables, 3) financial and property impact, 4) stress impact health, relationship, and behavior variables, 5) screening measures, 6) pre-disaster mental health, and 7) chronic disease variables (Tables 2-5).
The evaluation adopted a Classification Tree Analysis (CTA) model to evaluate which factors predict those most likely to enter the post-disaster specialist mental health treatment program. CTA optimally seeks to discriminate between 2 or more groups using data with discrete values. The sensitivity across groups will vary from 0% discrimination accuracy (chance) to 100% accuracy. The CTA model uses multiple discriminate analyses.Reference Linden and Yarnold77 Several authors have highlighted that CTA not only lends itself to easy interpretation but also provides evidence of causal mechanisms when assessing health care data. Additionally, CTA obtains P values at each node (study variables).Reference Camdeviren, Yazici and Akkus78, Reference Demir79
The chi-square automatic interaction detection (CHAID) algorithm was chosen to construct the classification treeReference Kass80, Reference Hill, Delaney and Roncal81. The CHAID method analyzes the relationship between the decision to enter treatment or not participate in the SMHP and variables that may influence the decision. CHAID technique uses the most significant factor to divide the study group into 2, and then subdivide it by the next most significant factor. The process continues stepwise until no more significant factors are identified. The method enables the identification of the most statistically significant factors that divide, in the case of this study, those who enter treatment versus those who do not.Reference Steadman, Silver and Monahan82
The study analysis aimed to ascertain if CTA can identify which questions may predict entry into the treatment program. The CTA was conducted using all variables (Figure 1 and Tables 2–5). The level of significance was set at P < 0.05. The minimum number of cases in the “parent,” or first, node was 100, and the second, or “child,” node was 50. The maximum depth of the tree was 3. Cross-validation and re-substitution evaluations were undertaken to estimate the risk of misclassification of a classifier.Reference Braga-Neto, Hashimoto and Dougherty83, Reference Braga-Neto and Dougherty84
Results
Descriptive Analyses
In 2012, 881 people were assessed by the SMHP. The mean PC-PTSD was 2.14 (SE 0.029, 95%, CI: 2.08;2.20). The treatment group (TG) (n = 215), in contrast to the non-treatment group (NTG) (n = 666), were more likely to describe their life was threatened by floods or cyclones (85.1% vs 8.1%), a fear of dying (78.6% vs 11%), and fear for the lives of others (81.9% vs 9.9%).
The TG response to the CATI question differed from the NTG; 15.3% of the TG reported “Life was good,” whereas 62.5% of the NTG described life as “good,” The TG recorded an altered level of optimism (TG 74.2% vs NTG 33%). The CR-2 scores differed; 66% of the TG and 2.7% of the NTG perceived an inability to adapt to change and bounce back after adversity. The TG considered themselves as “less optimistic in uncertain times and less likely to look on the bright side of life” compared to the NTG (70.2% and 6.9%, respectively). Thoughts of self-harm were uncommon (TG 9.77% and NTG nil) (Table 1).
Table 1. Pre-treatment screening assessment: narrative history of disaster exposure, CATI question, optimism, resilience, and thoughts of self-harm
The majority of those assessed were aged 20-49 (76.6%). Queensland’s 2011 population data indicates 41.6% were aged 20-49.85 Social security was the primary income source for the TG (63.3%). Almost 50% of the NTG had full-time employment. The marital status of those assessed differed from Queensland 2011 ABS: married/de-facto 19.5% vs 59.6%, divorced 23.7% vs 9.1%, and separated 31.6% vs 3.3%. Flooding affected the majority of those assessed (TG 83.3%, NTG 89.9%).
The TG, compared to the NTG, more often reported major property damage compared to none, minor, or moderate damage (43.3% vs 3.8%; P<0.05), relocation from home (32.6% vs 8.3%; P<0.05), personal loss (53.5% vs 24.3%; P<0.01), protracted insurance claims (25.6% vs 12.6%; P<0.01), or litigation 10.2% vs 0.5%; P<0.01) (Table 2). Changes to physical health, relationships, tobacco and alcohol use, gambling, and the type of stresses individuals experienced experiencing are detailed in Table 3. The impact on physical health was similar for the TG and NTG (47.4% vs 41.6%); relationship deterioration was more common in the TG (28.4% vs 0%; P<0.01,) while increased alcohol and drug use, tobacco consumption, and gambling were apparent in the TG (20% vs 8.1%, 16.7% vs 4.2%, and 3.3% vs 0%, respectively; all significant P<0.01). The TG reported more anxiety/depression (40.9% vs 16.2%; P<0.01). Interestingly, the NTG more often reported increased social stressors (TG 4.7% vs NTG 27.5%; P<0.01).
Table 2. Property and financial intake (economic) variables
Table 3. Stress impact variables
Personal and family history of physical and psychological health and previous disaster experience are detailed in Tables 4-5. The TG more frequently described a history of trauma. Family violence was reported in 16.7% of the TG compared to 11.9% in the NTG (P < 0.01). Childhood abuse occurred in 16.7% of the TG (NTG 5.7%, P<0.01), and sexual abuse in 11.6% of the TG and 6.5% of the NTG (P < 0.01). Prior exposure to disasters was more common in the TG (10.7% vs NTG 6.8%, P<0.01). The TG more often reported a family history of mental illness, a history of complex grief and suicidal thoughts, and personal history of mental illness and treatment (17.7% vs 11.3%, 14.4% vs 9.3%, 16.3% vs 2.4%, 27.9% vs 6.6%, and 31.6% vs 2.0%, respectively [all significant P<0.01]) (Table 4). The TG in comparison with the NTG more often experienced chronic illness (20% vs 6.4%) and was more likely to take prescribed medications (34% vs 9.6%; P<0.01) (Table 5).
Table 4. Pre-disaster mental health variables
Table 5. Chronic disease variables
The intake assessment was generally completed in full for demographic data and the impact of the disaster. In contrast, the family history, history of trauma, and personal history of mental and physical illness were often omitted, particularly for the NTG.
Classification Tree Analysis
The initial CTA (Tree 1) included all independent variables. The analysis identified the resilience measures as the initial (node 0) distinguishing feature (P < 0.001) between the TG and NTG, with property damage, financial losses, and threat to life (nodes 1, 2, and 3, respectively – all P<0.001) as the next factors that distinguished between the TG and NTG. Insurance claims (P < 0.001) linked to the perception that one’s life was threatened was the only other feature that distinctly predicted the decision to enter or not enter treatment (Figure 2 and Table 6). The CTA prediction accuracy for the TG was 90.3% and NTG 96.7%.
Figure 2. Classification Tree Analysis (CTA); treatment group (TG) vs non-treatment group (NTG).
Table 6. CHAID treatment group (TG) vs non-treatment group (NTG) model predicts 90.3% entering TG and 96.7% of NTG
CHAID was separately used to assess the relevance of (a) demographic factors (Tree 2), (b) narrative questions and psychological measures (Tree 3), (c) property damage and insurance claims (Tree 4), (d) physical health, behavior changes and stressors (Tree 5), (e) previous mental health history and family history (Tree 6), and (e) and (f) chronic disease variables (Tree 7). The CTA prediction accuracy for the NTG varied from 91.9%-100%. In contrast, prediction was less accurate for the TG (28.6%-84.7%) (Table 7).
Table 7. CHAID treatment group (TG) vs non-treatment group (NTG) model prediction CTA
The CTA (Table 7) indicates that prior mental health factors, the post-disaster perception of stress, property damage and losses, whether someone believed their life was threatened, coping, and the presence of chronic disease predicted non-participation (>90%). In contrast, the factors in Tree 3 (84.7%) were the only variables that predicted participation in treatment with greater than 80% accuracy.
The CHAID methodology identifies 5 items that distinguish between the TG and NTG. The resilience questions (TG vs NTG P<0.001), severity of property damage (TG vs NTG P<0.001), financial losses (TG vs NTG P<0.001), the belief one’s life was threatened (TG vs NTG P<0.001), and ongoing insurance claims (TG vs NTG P<0.003) identified those who entered treatment (90.3%) and the NTG (96.7%) (Table 6).
Discussion
Disasters place significant demands on responders and services. The need for services may extend beyond the timeframes adopted by governments and occur in an environment challenged by limited clinical resources and demand for services.Reference van den Berg, Wong and van der Velden86, Reference Fitzgerald, Toloo and Baniahmadi87 The naturalistic study reported in this paper relates to people affected by floods or cyclones 9-22 months before assessment. Identifying those with psychological symptoms that may require treatment and those more likely to enter treatment aids in resource management and prioritizing services to those more likely to participate in a treatment program.
The CHAID evaluation identified 5 variables that predict entry and non-entry into the SMHP in over 90% of people. The most parsimonious questions to predict program participation and, conversely, non-participation, were questions regarding resilience, severity of property damage, financial losses, ongoing insurance claims, and the perception one’s life was threatened. These findings reflect those of other studies that evaluated factors linked to adverse psychological outcomes after a disaster. Several authors have identified links between psychological distress, a person’s coping strategies, sociodemographic characteristics, health status, proximity to and disaster severity, risk to life, and difficulties with housing reconstruction.Reference Tunstall, Tapsell and Green63, Reference Mason, Andrews and Upton88–Reference Norris, Murphy and Baker90 Studies also show a relationship between resilience and psychological outcomes.Reference Nishi, Kawashima and Noguchi91, Reference Saja, Goonetilleke and Teo92
This study emphasizes that the disaster experience is not the only factor influencing participation in treatment. Clinical assessment should inquire about the degree of property damage, financial impacts, if insurance claims are resolved,Reference Eriksen, McKinnon and de Vet93 previous trauma exposure, and the personal and family history of mental health care and chronic illness. Other relevant factors noted in this study include changes in physical health and demographic factors such as age, gender, marital status, and employment.
Mental health screening in primary care has focused on case finding, with the sensitivity and specificity of the questions relevant to case identification and the provision of treatment.Reference Christensen, Toft and Frostholm94 This study used a well-recognized screening measure (PC-PTSD scale) and sought to identify factors that predicted participation in an SMHP. The results suggest that a limited number of screening questions may provide a guide regarding acceptance or non-acceptance of treatment. A screening strategy will enable clinicians to focus on those more likely to enter therapy while bearing in mind the need for alternative approaches to assist people screened as “at-risk” of psychological disorders such as PTSD but deciding not to participate in a treatment program. The questions identified by the CTA may also guide public health communications with “simple” media messages, like advertisements regarding driving and flood watersReference Fanham95 and changing health behaviors.Reference Abroms, Whittaker and Free96
The importance of post-disaster screening and informing the public was noted by Vardoulakis et al. (2022), who reported the key role of mental health services following disasters. However, the demand for mental health care may also exceed service capacity. Identifying and addressing individual and community mental health needs post-disaster is well recognized and supported by recommendations of the NSW Flood Inquiry (2022).Reference Vardoulakis, Matthews and Bailie97, Reference Keys98 However, there remains a risk that the learnings from the recent floods and those from the 2010-11 disasters may go unheeded.99
The findings of this study point to the importance of clinicians assessing factors such as resilience, the disaster experience, personal and property losses, and ongoing stressors such as insurance claims. The analysis also highlights the importance of clinicians assessing the post-disaster impact on physical health and the effect on relationships, substance use, and behavioral changes, such as increased gambling.
Further evaluation is required to assess the utility of these measures in other disaster settings (e.g., fires) and other countries, their potential for use in media campaigns that focus on encouraging help-seeking behavior, and how they may be used in post-disaster resource planning and training and as a strategy to screen those who present for psychological assistance following a disaster.
Limitations
Missing data imposed limitations on the study findings and raised questions regarding a clinician’s decision to ask (or not) what may be a difficult question, particularly if a person has decided not to progress with treatment. The study does not explore why questions such as those related to abuse are not asked or answered.
Strengths
The study evaluates data from disaster-affected people across a State with an area of 1.72 million kmReference Keenan, Weston and Volkova2.100 Those assessed had experienced symptoms for over 6 months. The data relates to 881 people aged 18 and over referred for assessment because of their psychological symptoms following the 2010-11 floods and cyclones. The data consisted of self-report and narrative questions about the disasters and addressed demographics, coping styles, personal history, physical health, and psychosocial impact factors.
Conclusion
This retrospective naturalistic study identified 5 factors that predicted the likelihood of participation in a Specialist Mental Health Program for those affected by the natural disasters that affected Queensland in 2010-11 (1. whether you perceived your life was threatened; 2. self-perception of resilience; 3. the degree of property damage; 4. the level of financial loss; 5. ongoing insurance claims).
Acknowledgements
The authors acknowledge the assistance of the late Professor Beverley Raphael AM who encouraged the development of the Post-disaster Specialist Mental Health Program and Dr Aaron Groves the Director Mental Health and Alcohol and Drug Services (Queensland) supported the implementation of the 2010-11 Post-disaster Mental Health Disaster Plan. Dr. David Phair provided advice and guidance in the data analysis.
The paper forms part of the PhD submission of David Crompton OAM. Professors R. Young and J. Shakespeare-Finch and Emeritus Professor G. FitzGerald are the PhD supervisors. There are no other conflict of interests.
The Program was funded by the Commonwealth Government Natural Disaster Relief and Recovery Arrangements.
Open access
