Mental health: build predictive models to steer policy - Nature
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Comment
ALEXANDER DEMIANCHUK/TASS/GETTY
A student during a COVID-19 quarantine in St Petersburg, Russia.
Mental health: build predictive
models to steer policy
Jo-An Occhipinti, Adam Skinner, P. Murali Doraiswamy, Cameron Fox, Helen Herrman, Shekhar Saxena, Elisha London,
Yun Ju Christine Song, Ian B. Hickie, on behalf of the World Economic Forum Global Future Council on Mental Health
A
Combine economic, social crisis is growing in mental health mental health and well-being for the first time.
as the widespread impacts of the If urgent and effective action is not taken, the
and medical data to forecast COVID-19 pandemic and the eco- protracted and global scale of the pandemic
need and design services to nomic hardship it has brought bites disruption will cast a long shadow on mental
deeper. In Japan, suicides rose by 16% health, particularly that of young people.
address the growing crisis. during the second wave of the pandemic, from Decades of research suggests that the
July to October 2020, compared with the rate response must be all-encompassing and long
in previous years1. In the United States, 25% term. The fact that this is neither feasible nor
of people aged 18–24 surveyed in June 2020 affordable in many contexts gives rise to two
reported increased substance use to cope types of response. Some governments or
with pandemic-related stress2. This year, the agencies allocate available resources over too
flagship report of the United Nations chil- broad a range of evidence-based programmes
dren’s charity UNICEF, The State of the World’s and services; without the scale and inten-
Children, focused on child and adolescent sity needed, these cannot achieve real and
Nature | Vol 597 | 30 September 2021 | 633
©
2
0
2
1
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
Comment
BRIAN L. FRANK
A food-bank worker in California: economic programmes must be part of the mental-health policy toolkit.
sustained impact. An example is Australia’s were used to inform timely, decisive and effec- timing and duration of health, social and eco-
struggle, over three decades, to shift the nee- tive responses to the pandemic3. nomic policies and initiatives that will deliver
dle on many mental-health conditions. Other We argue that a similar systems-modelling the greatest impacts. Understanding which
governments and agencies take a reactive and approach should be used to tackle the combinations of interventions work best at
ad hoc approach — as exemplified by the US mental-health challenge. Drawing together which stage is key to reducing harm.
response to the synthetic-opioid overdose qualitative and quantitative evidence and data, As experts in mental health and systems
epidemic. Neither approach will be adequate models should capture changes triggered by modelling, we outline here five challenges and
to tackle today’s mental-health crisis. the pandemic — such as education loss, job loss, four priorities to ensure that models are used
Instead, policymakers must account for how domestic violence, social isolation, fear and to reliably guide policy and allocate resources.
the pandemic has fundamentally changed the uncertainty. Models should forecast demand
state of mental health across society. It is time for community mental-health services and Five challenges
to examine preconceptions about what inter- Closing the care gap. In most societies, men-
ventions are effective. “Analysing risk factors tal-health systems were underfunded and
Lessons can be learned from infectious- fragmented even pre-pandemic4. The gap
disease research. Systems models allowed
independently fails between those that need care and those that
researchers to rapidly predict the spread of to account for their receive it is widest in countries where there is
COVID-19 (see Nature 580, 316–318; 2020), interactive effects.” conflict, such as Afghanistan, and in countries
integrating contact-tracing data based on with higher rates of unemployment, unequal
commuting patterns and mobile-phone loca- wealth distribution and budget cuts, such as
tion trackers. Although imperfect, in some acute care, including emergency-department Greece. Across low-, middle- and high-income
places these models provided a virtual test- presentations and psychiatric hospitalizations, countries there are significant treatment gaps
ing ground for alternative assumptions and as well as outcomes such as suicidal behaviour between the most privileged people and those
for the timing and scale of mitigation strat- (see ‘Mental-health forecast’). who are most marginalized, such as Indige-
egies, including lockdowns, mask wearing, Credible projections of population nous populations in Canada, New Zealand, the
school closures and vaccination. The models mental-health outcomes are needed to stress- United States and Australia5.
also accounted for the changing likelihood test new policies and mitigation strategies, Allocating resources. Interest and invest-
of people complying with such measures from employment programmes to helplines ment in mental health is growing. Witness
as the pandemic persisted. Where decision and investments in education and retraining. the work of the World Health Organization
makers worked closely and cooperatively with Before allocating significant investments, (WHO), UNICEF and global civil-society
modellers, as in Australia, New Zealand (see alternative scenarios should be simulated organizations such as the World Economic
go.nature.com/3kiw79n) or Taiwan, models to reveal the combination, scale, targeting, Forum’s Global Shapers Community (see
634 | Nature | Vol 597 | 30 September 2021
©
2
0
2
1
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.go.nature.com/3tpbtrz). In May 2020, the are needed to respond effectively to the as service coverage and capacity, and medical
UN secretary-general António Guterres called mental-health crisis. What impact will that records15,16. A 2021 report by the WHO17 on 183
for more urgent action on the issue6, and the combination have across different outcomes? of its member states highlighted that 86 have
elevation of mental health and suicide pre- Will the most effective combination be similar suicide data that are considered good qual-
vention in the global development agenda in across different contexts? What targeting, tim- ity (49 are high-income countries, and 37 are
recent years will be important to the COVID-19 ing, scale, frequency and intensity of invest- LMICs). And countries including India, China
mental-health response7. Up to US$160 billion ments are necessary? Will there be rebound and the United Arab Emirates are collecting
has been committed by the World Bank Group effects when temporary mitigation strategies data that will contribute to estimating the
to help developing countries tackle the health, such as income support or eviction moratori- burden of mental ill-health through national
social and economic impacts of the pandemic, ums are removed? What are the consequences and regional surveys. These could be used as
and governments around the world are com- of delayed actions? Will unintended conse- model inputs. Hence, the advanced modelling
mitting trillions to social and economic aid quences arise from well-meant but ill-designed tools to support decisions are no longer the
packages. However, without good planning or ill-timed mitigation measures? exclusive domain of high-income countries.
tools, decision makers will continue to be chal- The scale of these challenges behoves us to During the COVID-19 crisis, the Brain and
lenged by the complexity of causal drivers, the take a more progressive research path8,12. We Mind Centre at the University of Sydney,
quagmire of the known and the unknown, and must recognize the feedback loops, threshold Australia, leveraged years of experience to
the kaleidoscope of voices and choices. effects, non-independence and non-linearity develop a series of models to inform policy
Expanding the toolkit. The predominant that characterize our subject of study13,14. and planning for the country’s regional, state
approach to research on mental illness and For example, an increase in unemployment and national mental-health systems. In youth
suicidal behaviour uses retrospective data increases both the prevalence of psychological and in the population as a whole, the models
to identify independent risk factors, such as distress (which can lead to increased substance project the likely trajectories for a range of out-
unemployment, substance abuse or child- misuse and suicidal behaviour) and domestic comes including the prevalence of psycholog-
hood trauma. Half a century of such study has violence. Domestic violence increases rates ical distress, rates of help-seeking, wait times,
brought only partial progress towards popu- of adverse childhood exposures and psycho- emergency-department presentations, self-
lation-level impact, as a recent meta-analysis logical distress, further driving up rates of harm hospitalizations, and suicide deaths as
noted8. In addition, analysing risk factors inde- substance misuse and so on14. Modelling and a result of the pandemic10,18. Such projections
pendently fails to account for their interac- simulation can help us get a handle on such provide a test bed for probing the trade-offs
tive effects. This makes robust projections of complexity. and potential synergies of economic and social
population mental-health outcomes difficult, measures, among other strategies.
if not impossible. Opportunity in crisis For example, models from the Brain and
By contrast, infectious-disease epidemiol- A paradigm shift in population mental-health Mind Centre suggested that among the smart
ogy has matured into a robust interdisciplinary research is emerging at just the right time. choices for Australia in the recovery period
field through methodological expansion that The complexity and global scope of the crisis (2021–25) were investments in childcare,
makes routine use of the analytic techniques requires it, and the computational tools are employment programmes and job creation
of complex-systems science. At the outset of sufficiently advanced and accessible to make (particularly for women), active follow-up
the COVID-19 pandemic, this enabled global it feasible. The past two decades have seen after suicide attempts, and expansion of
research teams to rapidly deploy and custom- substantial investments in the data systems digitally coordinated specialist mental-health
ize existing systems models. Researchers could of low- and middle-income countries (LMICs), services. These were forecast to prevent an
analyse and forecast transmission trajectories covering civil registration — the collection of estimated 6% of self-harm hospitalizations and
under different conditions, allowing uncer- statistics such as births, marriage, divorce, 4.1% of emergency-department presentations
tainty to be quantified. As new information cause of death — and health information such relating to mental health10. Investing in more
and data about the virus and its transmissibil- psychiatric-hospital beds, awareness cam-
ity became available, models were refined and MENTAL-HEALTH FORECAST paigns, helplines or stand-alone primary-care
provided governments with critical tools for The predictions of a model developed at the University or specialist services were projected to deliver
of Sydney, Australia, to inform public-health decisions
testing and weighing the impact of responses, were close to the suicide rate for a region of New South
little impact, despite being considered ‘evi-
from mask wearing to travel quarantines. Wales in the years before the COVID-19 pandemic. dence based’.
Embracing ideas. Discourse on mitigating Similar approaches are also under way in
the mental-health impacts of COVID-19 has 100 the United States. The US National Institute
SOURCE: AUSTRALIAN INST. HEALTH AND WELFARE NATL MORTALITY DATABASE
focused largely on mental-health education of Mental Health (NIMH) recently launched
programmes, crisis helplines, improved access the Advanced Laboratories for Accelerating
to virtual services, and tinkering with existing 80 the Reach and Impact of Treatments for Youth
arrangements to enhance access to emergency and Adults with Mental Illness (ALACRITY)
care. This health-sector view fails to recognize research-centres programme. These centres
Suicides (per year)
that the most potent mental-health interven- 60
harness systems science, computational
tions can be social and economic. These could approaches, behavioural economics and dig-
include employment support, eviction mora- ital health to test methods to synthesize data.
toriums, subsidized education and training, Still, there is progress to be made on sev-
40
or increased unemployment insurance9,10. For eral fronts. Pandemic modelling benefited
Model (fitted)
instance, an income supplement that moved Model (projected)
from unprecedented levels of intergovern-
14% of Native American households out of Actual figures mental and intersectoral cooperation on the
poverty in North Carolina saw a 32% decrease 20 exchange of data and crucial information, for
in psychiatric symptoms among children of instance about movement patterns, that were
those households11. shared across telecommunication, transport
Huge questions. It is not at all clear what 0 and health agencies. Unfortunately, similar
combination of policies, initiatives or reforms 2011 2012 2013 2014 2015 2016 2017 2018 2019 cooperation is rare around the ethical sharing
Nature | Vol 597 | 30 September 2021 | 635
©
2
0
2
1
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.Comment
of data that would benefit efforts to model hospitalization and intentional-self-harm Sydney, Australia; and managing director at
mental health during COVID, for instance rates. Adjustments will need to be made Computer Simulation & Advanced Research
across health, education, social services and depending on the local context. In Australia, Technologies (CSART), Sydney, Australia.
the economic sector. for example, self-harm hospitalizations are Adam Skinner is a systems modeller at the
recorded, but not data on suicide attempts; Brain and Mind Centre, Faculty of Medicine
Four priorities in Colombia, suicide attempts are reported. and Health, University of Sydney, Australia.
Several Australian members of the World Model parameterization also draws on P. Murali Doraiswamy is professor of
Economic Forum’s Global Future Council on research evidence (including systematic psychiatry and medicine at Duke University
Mental Health (for details, see Supplementary reviews, randomized controlled trials and School of Medicine, Durham, North Carolina,
information) have spent five years applying cohort studies) and expert consensus. USA. Cameron Fox is health technologies lead
systems modelling to mental health. This Commit to co-design. Models that are cre- at the World Economic Forum, New York City,
experience has highlighted the following ated with multidisciplinary stakeholders and, USA. Helen Herrman is professor of psychiatry
four priorities for rapid and successful deploy- crucially, those with lived experience of mental at Orygen in Parkville, Australia; and at the
ment of models to improve population mental ill-health have better credibility and validity. Centre for Youth Mental Health, the University
health and prevent suicide. It is key to: A participatory approach helps to identify the of Melbourne, Australia. Shekhar Saxena is
Use a technical blueprint. To construct key referral pathways of the service system, professor of the practice of global mental
systems models in diverse contexts, a clear bottlenecks and barriers. It highlights users’ health in the Department of Global Health
picture of how social, economic and health experiences of delay, disengagement, service and Population, Harvard T. H. Chan School
factors interact to drive psychological dis- gaps and interruptions to continuity of care14. of Public Health, Boston, Massachusetts,
tress and mental-health outcomes is needed. Self-harm survivors, for example, identified USA. Elisha London is a social entrepreneur
An exemplar blueprint will be released this leading systems change in mental health,
year19, based on COVID-19 research by the Brain “What helps the based in London. Yun Ju Christine Song is
and Mind Centre and supported by the World research operations manager at the Brain and
Economic Forum. It details how to develop a
privileged might Mind Centre, Faculty of Medicine and Health,
national or regional system-dynamics model not aid the University of Sydney, Australia. Ian B. Hickie
of mental health, including key inputs, out- disadvantaged.” is co-director of health and policy at the Brain
puts and processes. The blueprint includes and Mind Centre, Faculty of Medicine and
guidelines for analysing the projections of Health, University of Sydney, Australia.
population mental-health outcomes. It also that long wait times made it more likely that e-mail: jo-an.occhipinti@sydney.edu.au
offers guidance on modelling the impact of they would stop seeking care, prolonging their
policies such as investing in jobs, childcare, distress and increasing the likelihood of sui- Supplementary information accompanies this
education or social connectedness. cidal behaviours. Qualitative data, triangu- article; see go.nature.com/3axqnhm
Build multidisciplinary teams. Expertise lated with quantitative service data, help to
in computational and systems science, epi- customize and ground models. 1. Tanaka, T. & Okamoto, S. Nature Hum. Behav. 5, 229–238
(2021).
demiology, psychology, psychiatry, social 2. Czeisler, M. É. et al. Morb. Mortal. Wkly Rep. 69,
science, policy and economics will ensure that Recognize limitations 1049–1057 (2020).
models have valid, robust, interdisciplinary Of course modelling is no policy panacea — just 3. Chang, S. L., Harding, N., Zachreson, C., Cliff, O. M. &
Prokopenko, M. Nature Commun. 11, 5710 (2020).
underpinnings, and limit bias. The COVID-19 look at the snail’s pace of climate-change mit- 4. World Health Organization. The Impact of COVID-19 on
international modelling consortium (CoMo) is igation. There is necessary scepticism about Mental, Neurological and Substance Use Services: Results
a notable example, drawing on a broad range the gulf between generalizations and granular of a Rapid Assessment (WHO, 2020).
5. Collins, P. Y. & Saxena, S. Nature 532, 25–27, (2016).
of disciplines and contextual knowledge to detail. Useful forecasts in high-income coun- 6. United Nations. Policy Brief: COVID-19 and the Need for
provide relevant, trusted and tailored tools tries might not work in low-income ones; and Action on Mental Health (United Nations, 2020).
to more than 30 countries20. what helps the privileged might not aid the 7. Arensman, E. Crisis 38, 1–6 (2017).
8. Franklin, J. C. et al. Psychol. Bull. 143, 187–232 (2017).
Strengthen data systems. Improving the disadvantaged. Critics point out that if the 9. Lund, C. et al. Lancet Psychiatry 5, 357–369 (2018).
range, quality and timeliness of data col- social determinants of physical health aren’t 10. Occhipinti, J., Ho, N., Skinner, A., Song, Y. C. &
lected and continuously reappraisng models being addressed — as has been so obvious Hickie, I. B. Closing the Gender Gap: Supporting the
Mental Health of Australian Women During the COVID-19
as fresh data emerge will reduce uncertainty from disparate COVID-19 death and disability Era (Brain and Mind Centre, University of Sydney, Sydney,
bounds, expand the insights that can come rates — it is naive to imagine they will suddenly Australia, 2021).
from systems modelling and improve decision be addressed for mental health once a good 11. Costello, E. J., Compton, S. N., Keeler, G. & Angold, A.
JAMA 290, 2023–2029 (2003).
making. In addition, a commitment to trans- model is available. Critics also note that out- 12. Atkinson, J. et al. Front. Psychiatry https://doi.org/10.3389/
parency and multidisciplinary co-design, and puts are only as good as inputs, cautioning fpsyt.2020.606035 (2020).
presenting the results of alternative assump- about a paucity of data and the perils of incor- 13. Atkinson, J. et al. Front. Psychiatry https://doi.org/10.3389/
fpsyt.2019.00448 (2019).
tions helps to guard against the politicization rect assumptions. 14. Occhipinti, J. et al. BMC Med. 19, 61 (2021).
of model outputs for public policy. Nonetheless, as the pandemic rages on, 15. Hung, Y. W., Hoxha, K., Irwin, B. R., Law, M. R. &
The data that are needed depends on the alongside so many other global challenges, Grépin, K. A. BMC Health Serv. Res. 20, 790 (2020).
16. Lopez, A. D., McLaughlin, D. & Richards, N. BMC Med. 18,
scope of the model and the outcome indica- there has never been a more important time 58 (2020).
tors of interest. These should be determined to strive for what we describe here. 17. World Health Organization. Suicide Worldwide in 2019:
by national priorities, stakeholder inputs and Global Health Estimates (WHO, 2021).
18. Atkinson, J., Skinner, A., Lawson, K., Song, Y. C. & Hickie,
the research question being posed. Time-se- I. B. Road to Recovery: Restoring Australia’s Mental Wealth
ries data used for model calibration range from The authors (Brain and Mind Centre, University of Sydney, 2020).
demographic and labour-force statistics such 19. Occhipinti, J. et al. Front. Psychiatry
https://doi.org/10.3389/fpsyt.2021.759343 (2021).
as births, mortality, migration and unemploy- Jo-An Occhipinti is co-director of the
20. Aguas, R. et al. BMJ Glob. Health 5, e003126 (2020).
ment rates, to estimates of the prevalence of Mental Wealth Initiative and head of systems
distress and mental disorder and related data modelling, simulation and data science at The authors declare competing interests; see go.nature.
such as emergency-department presentation, the Brain and Mind Centre, University of com/3axqnhm for details.
636 | Nature | Vol 597 | 30 September 2021
©
2
0
2
1
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.You can also read
