The Evolution of Mental Health: From DSM and ICD Diagnoses to Mechanism-Informed Mental Health Formulation

For much of the modern history of mental health care, diagnosis has been organized around a deceptively simple question:

Which disorder does this person have?

The answer has traditionally been constructed through systems such as the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Classification of Diseases (ICD). Clinicians identify patterns of symptoms, establish whether specific criteria are met, assign a diagnostic category, and use that category to guide communication, treatment, documentation, research, and—in many healthcare systems—reimbursement.

This approach has enormous practical value. A shared diagnostic language matters. The World Health Organization’s ICD-11, for example, provides a standardized framework for identifying and recording mental, behavioral, and neurodevelopmental disorders across health systems worldwide. Its clinical descriptions and diagnostic requirements were specifically developed to improve diagnostic consistency and clinical care. (World Health Organization⁠)

But a diagnosis is not necessarily an explanation.

“Major depressive disorder,” for example, describes a recognizable constellation of symptoms. It does not, by itself, tell us why a particular person’s depression developed, what biological or psychological mechanisms are maintaining it, why that person developed insomnia while another developed hypersomnia, why one responds to an SSRI while another does not, or why the same individual may experience dramatically different symptoms at different points in life.

That distinction is increasingly important.

Mental health science is gradually moving toward a more sophisticated model: from simply naming syndromes toward understanding the mechanisms that generate and maintain them.

This does not necessarily mean abandoning DSM or ICD diagnoses. Rather, it means recognizing that a diagnostic label may be the beginning of a clinical formulation rather than the endpoint.

The emerging paradigm asks a different question:

What processes are producing this person’s symptoms, and how can we intervene in those processes?

That shift—from label-based diagnosis toward mechanism-informed formulation—has the potential to change how mental disorders are researched, assessed, monitored, and treated.

The Strength—and Limitation—of Diagnostic Labels

Modern psychiatric classification was developed primarily to solve a legitimate problem: clinicians needed a common language.

Before standardized diagnostic systems, clinicians could use the same word to mean very different things. Classification systems improved reliability by establishing observable criteria that could be communicated across clinicians and institutions.

That remains important.

A diagnosis can help a clinician communicate quickly:

• the patient has a recognizable syndrome;
• certain symptoms are clinically significant;
• a particular treatment pathway may be appropriate;
• certain risks should be assessed;
• insurance or healthcare systems can recognize the condition;
• researchers can define populations for clinical studies.

The problem arises when the classification is mistaken for the underlying disease process.

Consider two people who both meet criteria for major depressive disorder.

One may have experienced prolonged childhood adversity, chronic interpersonal threat, sleep disruption, social isolation, and persistent hypervigilance.

Another may develop depression following an inflammatory illness, hormonal changes, medication exposure, severe sleep deprivation, or another physiological stressor.

A third may have a strong genetic vulnerability interacting with environmental stress.

A fourth may have a predominantly anhedonic presentation characterized by profound loss of reward sensitivity.

They can all receive the same diagnostic label.

Yet the mechanisms involved may be substantially different.

Conversely, people with entirely different diagnoses may share important mechanisms.

Anxiety, depression, trauma-related disorders, substance-use disorders, obsessive-compulsive symptoms, and some psychotic-spectrum conditions can overlap in processes involving threat detection, reward processing, cognitive control, arousal regulation, sleep, inflammation, stress physiology, and learning.

This creates an important paradox:

Different diagnoses can share mechanisms, while the same diagnosis can contain multiple mechanisms.

That is one reason symptom categories alone may be insufficient for precision mental healthcare.

From Categories to Dimensions

One of the most important conceptual shifts occurring in psychiatric research is the movement from purely categorical thinking toward dimensional and mechanistic models.

A categorical model essentially asks:

Does the person have disorder X?

A dimensional model asks:

Where does this person fall along several clinically meaningful dimensions?

For example:

• threat sensitivity
• reward sensitivity
• cognitive control
• emotional regulation
• arousal
• impulsivity
• social processing
• sleep/circadian regulation
• psychomotor activation
• attention
• interoception
• stress reactivity

These dimensions do not map neatly onto one DSM diagnosis.

That is precisely the point.

The National Institute of Mental Health’s Research Domain Criteria (RDoC) initiative was created around this broader conceptualization. RDoC encourages researchers to investigate dimensions of behavior and neurobiological functioning across traditional diagnostic boundaries, incorporating information ranging from genetics and neural circuits to behavior and self-report. (PubMed Central (PMC)⁠)

RDoC is not a replacement diagnostic manual for routine clinical practice. It is better understood as a research framework pointing toward a different way of conceptualizing psychopathology.

The larger implication is profound:

Psychiatric disorders may be better understood as disturbances in interacting biological, psychological, behavioral, and environmental systems than as isolated disease categories.

The Brain as a Network, Not a Collection of “Chemical Imbalances”

One of the most visible developments in this transition is the growing emphasis on neural circuits and network function.

Older explanations of mental illness often emphasized individual neurotransmitters:

• serotonin
• dopamine
• norepinephrine
• GABA
• glutamate

These systems remain important, but the idea that a mental disorder can be reduced to “too little serotonin” or “too much dopamine” is increasingly inadequate.

The brain is not simply a chemical soup.

It is a dynamic network.

Mental states emerge from interactions among distributed neural systems involved in:

• threat detection
• reward and motivation
• memory
• executive control
• salience
• social cognition
• interoception
• arousal
• learning
• emotional regulation

For example, depression may involve altered communication among networks involved in reward, cognitive control, self-referential processing, and emotional salience. Anxiety may involve altered threat processing and regulation. Trauma can produce persistent changes in threat detection, memory processing, autonomic regulation, and contextual learning.

The important question therefore becomes less:

“Which neurotransmitter is responsible?”

and more:

“Which neural systems are functioning differently, under what circumstances, and why?”

That is a considerably more complicated question—but potentially a much more useful one.

Inflammation and the Brain-Body Connection

Another important development is the recognition that psychiatric symptoms cannot always be cleanly separated from systemic physiology.

The immune system, endocrine system, nervous system, metabolism, sleep, and microbiological environment continuously interact.

Inflammatory signaling can influence:

• energy
• motivation
• sleep
• cognition
• appetite
• pain
• reward processing
• mood
• psychomotor activity

This has helped revive an important idea that was sometimes obscured by the historical division between “physical” and “mental” illness:

The brain is an organ embedded within the body.

That does not mean depression is simply “inflammation,” or that every person with anxiety has an immune disorder. Such claims would go well beyond the evidence.

Rather, inflammation represents one potentially important biological pathway among many.

The same principle applies to metabolic function, endocrine signaling, mitochondrial processes, sleep physiology, autonomic regulation, and other biological systems.

A mechanism-informed formulation therefore asks whether biological processes may be contributing to the person’s presentation rather than automatically assuming that all clinically relevant information can be captured by psychiatric symptoms alone.

Genetics: Risk Is Not Destiny

Genomics offers another example of why mechanism-informed thinking is different from traditional diagnostic classification.

Psychiatric disorders clearly have genetic components, but most common psychiatric conditions are highly polygenic.

There is rarely a single “depression gene,” “schizophrenia gene,” or “anxiety gene.”

Instead, thousands of genetic variants may contribute small amounts of vulnerability, interacting with:

• development
• environment
• stress
• trauma
• sleep
• nutrition
• substance exposure
• social conditions
• other biological factors

This creates a more nuanced model of risk.

Genetics may tell us something about susceptibility, but susceptibility is not the same thing as inevitability.

A person may possess biological vulnerabilities that never become clinically significant. Another person may develop symptoms because environmental stressors interact with those vulnerabilities during a particular developmental period.

This is where genetics becomes particularly powerful when combined with other forms of information.

Rather than asking:

“Does this person have the gene for disorder X?”

the more scientifically meaningful question becomes:

“How do this person’s genetic characteristics interact with development, environment, neural circuitry, physiology, and behavior to produce vulnerability or resilience?”

That is a much more sophisticated model of human behavior.

Biomarkers: The Search for Measurable Biology

Perhaps the most anticipated development in psychiatry is the search for reliable biomarkers.

A biomarker is a measurable biological characteristic that can provide meaningful information about a physiological or pathological process.

Potential psychiatric biomarkers have been investigated across multiple domains, including:

• genetics
• epigenetics
• inflammatory markers
• hormones
• neuroimaging
• electrophysiology
• autonomic measures
• sleep architecture
• metabolic measures
• proteomics
• metabolomics
• cognitive performance

The promise is substantial.

Imagine a future in which a clinician could combine symptom reports with biological measurements to determine that two patients with superficially similar depression have fundamentally different underlying profiles.

One might show pronounced sleep/circadian disruption.

Another might show abnormalities associated with inflammatory signaling.

Another might demonstrate altered reward processing.

Another might show a predominantly anxiety-driven threat phenotype.

Treatment could eventually become more targeted as a result.

But this is where scientific caution is essential.

Psychiatry does not yet have a single blood test, brain scan, or genetic test that can diagnose most mental disorders with the precision that a blood glucose measurement can diagnose diabetes.

Many proposed biomarkers are still investigational. Some findings are statistically meaningful at the population level but insufficiently specific for individual clinical decision-making.

The future of biomarkers is promising—but precision medicine requires validated, reproducible, clinically useful biomarkers, not simply interesting correlations.

Digital Phenotyping: Mental Health in Real Time

Perhaps the most transformative development is the ability to observe behavior continuously rather than relying exclusively on retrospective clinical interviews.

A traditional mental health assessment might ask:

How have you been sleeping?

The patient might answer:

“Not very well.”

A digital monitoring system could potentially provide information about:

• sleep duration
• sleep regularity
• physical activity
• heart rate
• heart-rate variability
• mobility
• communication patterns
• social activity
• voice characteristics
• typing behavior
• smartphone use
• daily routines
• self-reported mood
• medication adherence

This broader concept is sometimes called digital phenotyping.

Instead of obtaining a snapshot every few weeks, clinicians could potentially obtain a longitudinal picture of behavior and physiology.

That distinction matters.

Mental health conditions are dynamic.

A patient may appear relatively stable during a 50-minute appointment while experiencing significant changes in sleep, activity, social engagement, or mood between appointments.

Wearables and smartphones potentially allow researchers and clinicians to study those changes.

Recent research has demonstrated associations between wearable-derived measures—including sleep, physical activity, and heart rate—and depression and anxiety severity, while also showing the potential value of combining behavioral, physiological, and self-report data. (arXiv⁠)

But digital phenotyping also presents substantial challenges.

Correlation is not causation.

A reduction in movement might indicate depression—or illness, injury, weather, work schedule, medication effects, or hundreds of other factors.

Similarly, changes in smartphone use are not inherently psychiatric biomarkers.

The goal should therefore not be to turn smartphones into automated psychiatrists.

The goal is to use longitudinal data as additional context within a broader clinical formulation.

From Diagnosis to Formulation

This brings us to perhaps the most important distinction.

A diagnosis answers:

What recognizable clinical syndrome does this person meet criteria for?

A formulation asks:

Why is this person experiencing these symptoms, what maintains them, what protects against them, and what mechanisms can we target?

A mechanism-informed formulation might integrate several levels simultaneously.

Level 1: Phenotype

What is actually happening?

• depressed mood
• panic
• insomnia
• intrusive memories
• compulsive behavior
• irritability
• anhedonia
• cognitive difficulties
• substance use

Level 2: Psychological mechanisms

What processes may be contributing?

• avoidance
• threat conditioning
• rumination
• attentional bias
• impaired emotion regulation
• maladaptive beliefs
• reinforcement patterns
• attachment-related processes

Level 3: Neural mechanisms

What systems may be involved?

• threat circuitry
• reward circuitry
• executive control networks
• salience networks
• memory systems
• arousal systems

Level 4: Biological mechanisms

What physiological factors might contribute?

• sleep disruption
• inflammatory signaling
• endocrine changes
• metabolic dysfunction
• autonomic dysregulation
• medication effects
• genetic vulnerability

Level 5: Environmental mechanisms

What is happening around the person?

• chronic stress
• trauma
• relationships
• occupational demands
• socioeconomic conditions
• social isolation
• cultural context
• environmental instability

Level 6: Temporal dynamics

How does the condition change?

• What preceded the symptoms?
• What triggers worsening?
• What predicts improvement?
• What maintains the condition?
• What happens before relapse?

This final dimension may be particularly important.

Mental disorders are not static objects. They are processes unfolding over time.

The Patient Becomes More Than the Diagnosis

There is also an important human consequence to this shift.

Labels can be useful, but they can inadvertently become identities.

A person can gradually move from:

“I have depression”

to:

“I am depressed.”

Or from:

“I experienced trauma”

to:

“I am a traumatized person.”

A mechanism-informed formulation offers a different possibility.

It can frame symptoms as understandable outputs of interacting systems rather than as evidence of a defective identity.

That does not minimize suffering.

Quite the opposite.

It can make the clinical picture more precise.

Instead of saying:

“You have anxiety.”

we might eventually be able to say:

“Your threat-detection system appears highly sensitized, your sleep is fragmented, your autonomic arousal remains elevated, and avoidance is reinforcing the fear response.”

That formulation creates multiple potential treatment targets.

The label identifies the territory.

The formulation begins to explain the terrain.

What Happens to the DSM and ICD?

The emergence of mechanism-informed psychiatry does not mean that DSM and ICD suddenly become obsolete.

They serve important functions.

The ICD remains the global standard for classification and reporting of diseases and health conditions, while its ICD-11 clinical manual provides clinicians with structured guidance for identifying mental, behavioral, and neurodevelopmental disorders. (World Health Organization⁠)

Diagnostic categories also remain necessary for:

• clinical communication
• epidemiology
• healthcare administration
• research recruitment
• treatment guidelines
• disability determinations
• insurance systems
• public health planning

The more realistic future is therefore probably not diagnosis versus formulation.

It is:

diagnosis plus formulation.

The diagnostic label becomes one layer of information rather than the entire clinical model.

A clinician might document a conventional diagnosis while simultaneously describing:

• symptom dimensions
• developmental history
• relevant neural and psychological mechanisms
• biological contributors
• environmental stressors
• protective factors
• treatment response
• longitudinal changes
• measurable outcomes

This is not a rejection of diagnosis.

It is an expansion of what diagnosis means.

The Risk of Going Too Far in the Other Direction

There is also a danger in romanticizing biological psychiatry.

A brain scan cannot explain a marriage.

A genetic profile cannot explain a person’s grief.

An inflammatory marker cannot capture the meaning of trauma.

A wearable cannot understand why someone stopped exercising.

Human beings are biological organisms, but they are also psychological, relational, cultural, social, and meaning-making organisms.

A genuinely sophisticated mental health model therefore cannot reduce psychiatry to neuroscience.

The future should not be:

DSM → brain scan.

It should be:

symptoms + behavior + development + relationships + environment + physiology + genetics + neural systems + longitudinal data.

In other words, the goal is not to replace one reductionism with another.

It is to build a more integrated model.

Toward Precision Mental Health

The long-term objective of mechanism-informed formulation is ultimately precision.

Two people with the same diagnostic label may need different interventions.

Two people with different diagnoses may benefit from targeting the same underlying mechanism.

And the same person may require different interventions at different stages of illness.

Imagine a future clinical record that does not simply say:

Major depressive disorder.

Instead, it might describe a multidimensional profile:

Phenotype: anhedonia, insomnia, cognitive slowing
Neurobehavioral: diminished reward responsiveness and cognitive-control difficulties
Physiological: circadian disruption and altered autonomic regulation
Biological: relevant metabolic or inflammatory abnormalities
Genetic: elevated polygenic vulnerability
Psychological: rumination and behavioral withdrawal
Environmental: occupational stress and social isolation
Digital trajectory: declining activity and increasingly irregular sleep over six weeks
Treatment response: improved sleep preceding improvement in mood

That is far closer to a clinical map than a diagnostic label.

And maps are useful because they can change.

If the person’s sleep improves, the map changes.

If inflammation resolves, the map changes.

If trauma is processed, avoidance decreases, or social connection improves, the map changes.

If a medication causes adverse effects, the map changes.

Mental health assessment therefore becomes an ongoing process of measurement, hypothesis, intervention, and reassessment.

The Emerging Model

The most promising future of mental healthcare may not involve choosing between psychology and biology, or between diagnosis and formulation.

It may involve integrating them.

The trajectory looks something like this:

From categories → dimensions

From symptoms → mechanisms

From snapshots → longitudinal trajectories

From single disorders → interacting systems

From retrospective reporting → continuous measurement

From trial-and-error treatment → increasingly targeted intervention

From diagnosis as endpoint → diagnosis as starting point

The DSM and ICD gave mental healthcare something it desperately needed: a common language.

The next stage may require something more ambitious: a common model of how psychological suffering actually emerges and changes.

Circuits, genetics, inflammation, biomarkers, behavior, environment, development, relationships, and digital monitoring are not competing explanations. They are different levels of the same human system.

The challenge is learning how to integrate those levels without reducing a person to any one of them.

That is ultimately what mechanism-informed mental healthcare should mean.

Not replacing the human being with a biological profile.

Not replacing diagnosis with technology.

And not pretending that every psychological experience can be reduced to a brain circuit or biomarker.

Instead, it means moving toward a model in which the diagnostic label tells us what we are seeing, while formulation helps us understand why it is happening, what is maintaining it, and where intervention might make a difference.

The future of mental healthcare may therefore be less about asking, “What disorder does this person have?”

And increasingly about asking:

“What is happening in this person, across multiple levels of their biology, brain, behavior, relationships, environment, and lived experience—and what can we change?”

That is a much harder question.

But it may also be the question that finally moves mental healthcare closer to genuine precision medicine.