Depression and Anxiety Often Precede a First Schizophrenia Diagnosis, Japanese Claims Data Show

Long before psychosis announces itself, schizophrenia often whispers. A large retrospective study drawing on a nationwide Japanese health insurance claims database has found that depressive episodes and anxiety disorders are the most common psychiatric diagnoses recorded in the months and years before a person receives a first schizophrenia diagnosis, and that the pattern of those earlier diagnoses is linked to how patients fare afterward. The findings, published in BMC Psychiatry by a team led by Yoshiteru Takekita of Kansai Medical University, add weight to a long-standing clinical observation: the road to a schizophrenia diagnosis frequently begins with symptoms that look, on paper, like something else.

The research team analyzed records from more than 21,000 patients who received a first recorded schizophrenia diagnosis between January 2006 and November 2023, each with at least one year of continuous insurance enrollment before and after that index month. Of the 21,590 eligible patients, 15,955 had at least one psychiatric diagnosis recorded before the index month. Among those with prior diagnoses, the single most common earliest recorded condition was a depressive episode, accounting for 45.2 percent, followed by other anxiety disorders at 21.6 percent. Only 5,788 of these patients — roughly one in three — had schizophrenia itself as their first recorded psychiatric diagnosis.

That distribution matters because it quantifies what psychiatrists have long described anecdotally. Schizophrenia is a heterogeneous disorder, and affective symptoms — depression, anxiety, social withdrawal, sleep disturbance — commonly emerge before frank psychosis. When a patient presents with low mood or crippling worry, few clinicians would immediately suspect a psychotic illness that may not manifest for another year or more. The study found that the mean interval between the earliest recorded psychiatric diagnosis and the first recorded schizophrenia diagnosis was 12.8 months overall, but stretched to 20.2 months among patients whose earliest recorded diagnosis was something other than schizophrenia. In other words, for many patients, a full year and a half of clinical contact passes under a different diagnostic label before the picture clarifies.

To make fair comparisons between patients with and without prior psychiatric diagnoses, the researchers used inverse probability of treatment weighting, a statistical technique that reweights the two groups to mimic the balance of a randomized trial, adjusting for differences in age, sex, and other baseline characteristics. They then applied Cox proportional hazards models to estimate the risk of two downstream outcomes after the index month: psychiatric inpatient admission and physician-certified sick leave. The approach is standard in pharmacoepidemiology, but it hinges on the assumption that all relevant confounders are captured in the claims data — an assumption the authors acknowledge is difficult to fully satisfy with administrative records.

The results on hospitalization were nuanced. In the adjusted Cox model, patients with prior psychiatric diagnoses had a modestly lower risk of psychiatric inpatient admission, with a hazard ratio of 0.87 and a 95 percent confidence interval of 0.77 to 0.98, reaching statistical significance at P = 0.020. Yet when the team examined the same outcome using Kaplan–Meier survival curves and a patient-years analysis, the difference between groups disappeared. This kind of internal disagreement is not unusual in observational data. The Cox model adjusts for measured confounders, while the unadjusted analyses do not, so a protective signal that appears only after adjustment may reflect confounding rather than a true clinical effect. The authors are careful not to overinterpret it.

Sick leave told a cleaner and more troubling story. Across all three analytical approaches, patients with prior psychiatric diagnoses were at higher risk of physician-certified sick leave after their schizophrenia diagnosis. The Cox model yielded a hazard ratio of 1.19 (95 percent confidence interval, 1.11 to 1.27; P < 0.001), the Kaplan–Meier analysis showed highly significant differences by both log-rank and Wilcoxon tests, and the patient-years analysis produced a risk ratio of 1.09 (95 percent confidence interval, 1.04 to 1.15). Sick leave is a blunt but meaningful proxy for functional impairment — it captures the point at which symptoms become severe enough that a physician judges a patient unable to work. The consistent signal across methods suggests that patients whose schizophrenia was preceded by other psychiatric diagnoses may experience a more disruptive illness course in occupational terms.

Why might that be? One plausible reading is that a prior psychiatric diagnosis creates a longer pathway through the health system, during which the underlying psychotic process continues to evolve untreated. The field has a name for the delay between the onset of psychotic symptoms and the start of adequate treatment: the duration of untreated psychosis, or DUP. Longer DUP has been repeatedly associated with worse functional outcomes, including slower symptom remission and greater social and occupational disability. If depressive or anxiety symptoms delay recognition of emerging psychosis, the roughly 20-month gap documented in this study could represent a window in which earlier intervention might alter the trajectory. The authors stop short of claiming causality, but the temporal pattern is consistent with that interpretation.

Alternative explanations deserve equal attention. Depression and anxiety may genuinely be part of the prodromal phase of schizophrenia — early manifestations of the same neurodevelopmental process rather than misdiagnoses. Some patients may have true comorbid mood or anxiety disorders that persist alongside psychosis and independently worsen functional outcomes. And there are artifacts of the data itself to consider: claims databases record diagnoses that are billed, not necessarily diagnoses that are biologically correct, and diagnostic practice in Japan, as elsewhere, may favor affective labels in the early stages of unclear presentations. The retrospective design cannot disentangle these possibilities, and the study’s funding by Sumitomo Pharma, with several authors employed by the company or the data provider JMDC, is disclosed in the paper and should be weighed by readers.

The scale of the dataset is the study’s principal strength. JMDC claims records cover a large share of Japan’s employed population and their dependents, offering statistical power that individual clinical cohorts cannot match, and the 17-year observation window spans changes in diagnostic practice and treatment availability. The requirement of continuous enrollment for at least a year on either side of the index month reduces the risk that prior diagnoses were simply missed due to gaps in coverage. Still, the population skews toward working-age employed people, which may explain why sick leave was measurable at all — and it means the findings may not generalize to patients outside the formal labor market, older adults, or those with severe illness from the outset.

For clinicians, the practical message is one of sustained vigilance rather than alarm. The authors conclude that because depressive and anxiety disorders commonly precede a first recorded schizophrenia diagnosis in Japan, and because the mean interval between the earliest non-schizophrenia psychiatric diagnosis and the schizophrenia diagnosis is approximately 20 months, continued clinical assessment of patients presenting with mood or anxiety symptoms may be warranted — particularly when symptoms are atypical, treatment-refractory, or accompanied by social decline or unusual thinking. That is not a call to diagnose psychosis prematurely, which carries its own harms, but a reminder that a first psychiatric diagnosis is sometimes a waypoint rather than a destination. For researchers, the study offers a template for using claims data to map the diagnostic pathways that lead into severe mental illness, and a quantitative benchmark — 45 percent depression, 22 percent anxiety, 20 months — against which future early-detection strategies can be measured.

Subject of Research: Diagnostic patterns preceding a first schizophrenia diagnosis and their association with subsequent clinical outcomes in Japan

Article Title: Previous psychiatric diagnoses preceding a first recorded schizophrenia diagnosis and subsequent clinical outcomes: a retrospective study using a Japanese claims database

Article References: Takekita, Y., Azekawa, T., Baba, K., Nosaka, T., Tajima, T., Funakoshi, Y., Takamatsu, Y., & Kato, M. (2026). Previous psychiatric diagnoses preceding a first recorded schizophrenia diagnosis and subsequent clinical outcomes: a retrospective study using a Japanese claims database. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08687-6

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08687-6

Keywords: schizophrenia, depression, anxiety disorders, prodrome, claims database, psychiatric diagnosis, duration of untreated psychosis, sick leave, psychiatric hospitalization, retrospective cohort study, Japan, early detection