Why mental disorders cannot be reduced to a single test

Psychiatry is in a complex transitional period. On the one hand, it remains a clinical specialty in which diagnosis is largely based on conversation, observation, medical history, assessment of behavior, complaints, symptom duration and the effect of the disorder on the patient’s life. On the other hand, medicine is increasingly searching for objective signs of mental disorders: biomarkers, neuroimaging patterns, digital indicators of sleep and activity, speech characteristics, cognitive parameters, genetic and inflammatory signals. This search reflects the desire to make diagnostics more precise, earlier and more individualized.

The main problem in modern psychiatry is the high heterogeneity of diagnoses. Two patients with the same diagnosis of depression may have different symptoms, causes, disease course and response to treatment. In one patient, insomnia, anxiety and loss of appetite may dominate; in another, hypersomnia, psychomotor slowing and weight gain may be more prominent. In one case, the episode may be associated with inflammatory disease; in another, with chronic stress, bipolar spectrum features, traumatic experience or circadian rhythm disruption. Therefore, one clinical label may unite different biological and psychological mechanisms.

This is why the idea of precision psychiatry emerged. Its goal is not to replace clinical thinking with a laboratory test, but to divide broad diagnostic categories into more understandable subtypes. If the physician can identify in advance which patient has a higher risk of chronic course, suicidal behavior, psychotic episode, adverse effects or poor response to a specific drug, treatment can become more personalized. The field is developing actively, but it still requires stricter validation, standardization and clinical applicability.

A biomarker in psychiatry does not have to be only a blood test. It may be a genetic variant, a protein, a hormonal marker, an inflammatory marker, a metabolite, a sleep characteristic, brain structure, functional connectivity between brain regions, a speech pattern, a cognitive profile or a digital trace of behavior. However, the presence of a statistical association with a disorder does not yet make a marker a diagnostic test. For clinical use, it must be shown that the biomarker works reliably in different patient groups, provides additional information compared with ordinary assessment and truly helps make a medical decision.

Neuroimaging has long attracted interest in psychiatry. MRI, functional MRI, PET and other methods allow researchers to study brain structure and activity, neural networks, volumes of specific regions and connections between the limbic system, prefrontal cortex, thalamus and other structures. Numerous group differences have been described in depression, schizophrenia, bipolar disorder, post-traumatic stress disorder and autism spectrum disorders. However, a group difference is not the same as individual diagnosis. If a group of patients shows an average difference in the activity of a certain network, this does not mean that a diagnosis can be reliably made from the scan of one person.

Prediction of treatment response is considered especially promising, perhaps more than diagnosis from a single scan. Neuroimaging and electrophysiological markers are being studied to help select an antidepressant, psychotherapy, transcranial magnetic stimulation or other interventions. This task is more realistic: it is not necessary to prove that a biomarker “sees depression,” but enough to show that it helps choose a more effective strategy. Even here, caution is needed. Many models work well in a research sample, but lose accuracy when transferred to another clinic, another population or another data collection system.

Blood biomarkers are also actively studied. Researchers evaluate inflammatory cytokines, stress hormones, neurotrophic factors, metabolites, lipids, microRNAs, gene expression and other indicators. In depression, signs of chronic low-grade inflammation and metabolic disturbance are of interest. In schizophrenia and bipolar disorder, immune, neurodevelopmental and transcriptomic markers are being studied. Such tests may eventually help in differential diagnosis or prognosis, but they are not yet a universal replacement for clinical assessment.

It is important to emphasize the limitation: in ordinary practice, a psychiatric diagnosis is not currently made based on a single blood test. Even a potentially useful biomarker must be applied as an addition to clinical evaluation. Symptoms of mental disorders depend on context: sleep, medications, substance use, somatic diseases, life events, neurological disorders and social environment. A test may help in differential diagnosis or prognosis, but it cannot fully replace understanding the patient as a person with a history, behavior and dynamic condition.

Digital phenotyping has become another major direction. It uses data from smartphones, wearable devices and digital behavior: activity, sleep, geolocation patterns, typing speed, communication frequency, voice characteristics, phone use rhythm, physical mobility and social engagement. The idea is that mental state manifests not only in the physician’s office, but also in everyday life. For example, worsening depression may be accompanied by reduced activity, sleep changes, decreased communication and disruption of routine. A relapse of psychosis may be preceded by changes in behavior, sleep and digital activity.

Digital biomarkers are especially attractive for early warning of relapses. In bipolar disorder, depression or psychotic disorders, deterioration often develops gradually. If a system notices changes in sleep, activity, speech or social patterns before a severe crisis, the physician may intervene earlier: adjust therapy, increase monitoring, schedule a consultation or involve support. But such a model requires patient consent, data protection and clear rules. Continuous monitoring of mental state may be useful, but it also carries the risk of intruding into private life.

Speech and voice analysis is a separate area. Speech rate, pauses, intonation, voice variability, coherence of expression, vocabulary patterns and acoustic features are studied in depression, mania, psychosis, anxiety and cognitive impairment. Such methods are potentially convenient because speech is a natural part of clinical contact. However, attempts to create diagnostic artificial intelligence based on voice face technical and regulatory difficulties. A model may show interesting results compared with questionnaires, but clinical use requires evidence of safety, reliability, resistance to error, performance across languages and populations, and a clear plan for algorithm updates.

This example is important because it shows the difference between research accuracy and a medical device. A model may demonstrate promising results in a controlled dataset, but for clinical implementation it must prove that it works safely in real patients. Psychiatry is especially sensitive to false-positive and false-negative results. An incorrect indication of depression may cause anxiety and unnecessary visits, while a missed high-risk state may have serious consequences.

Artificial intelligence in psychiatry creates both hope and concern. It may help structure data, track symptom dynamics, analyze scales, identify relapse risk and support the physician in decision-making. But the use of chatbots as independent therapy remains problematic. Generative systems may provide responses that appear supportive, but they do not necessarily follow clinical guidelines, do not always recognize risk, and do not replace professional help. This is particularly important for patients with suicidal thoughts, psychosis, severe depression, mania or complex trauma.

The regulatory environment is gradually adapting to digital psychiatry. Digital medical devices, software as a medical device, artificial intelligence, machine learning and digital therapeutics require clinical evidence and safety evaluation. This shows that digital psychiatry cannot develop as an unregulated market of applications if it claims medical significance. A tool that only supports general well-being has one level of responsibility, while a system that claims to detect disease, predict relapse or recommend treatment requires a much stricter level of assessment.

One of the main scientific difficulties is reproducibility. Psychiatric biomarker studies often suffer from small samples, differences in diagnostic criteria, medication effects, comorbidities, age, sex, sleep, stress, substance use and social factors. A model built on one sample may not work in another. Generalizable biomarkers require new strategies that account for design errors, confounders, multimodal data integration and the mismatch between biological and clinical heterogeneity.

The future of precision psychiatry will probably be multimodal. One blood test, one brain scan or one digital indicator is unlikely to reliably explain a complex mental disorder. A more realistic model is one in which the physician combines clinical presentation, family history, cognitive profile, sleep data, activity, laboratory markers, neuroimaging and dynamics of treatment response. Such a model may help not only clarify diagnosis, but also select therapy, determine relapse risk and adapt monitoring.

The main significance of biomarkers in psychiatry is not the rejection of clinical conversation, but its strengthening. Mental disorders manifest through subjective experience, behavior, the body, the brain and social context at the same time. Therefore, precision psychiatry must remain humane and medically responsible. Technologies can provide the physician with new data, but they cannot replace trust, observation, ethical judgment and understanding of the patient’s individual history. Psychiatry of the future will become more precise not because it finds one universal test, but because it learns to combine biological, digital and clinical signs into a more reliable picture of the person’s condition.