Medication and Therapy Clues Predict Which Teens With Mood Disorders Need Hospital Care

When a teenager with a mood disorder spirals into crisis, one of the most consequential decisions a psychiatric team can make is whether hospitalization is necessary. Admission to an inpatient psychiatric unit is among the most intensive and disruptive interventions in youth mental health care, yet clinicians have long lacked clear, real-world evidence about which young patients are most likely to end up hospitalized. A new study drawing on electronic medical records from a major Taiwanese medical center now offers a detailed portrait of the factors that track with psychiatric admission among children and adolescents treated for mood-related conditions, and its findings challenge some common assumptions about who is at highest risk.

The research, published in BMC Psychiatry by a team led by Shu-Yu Wu and Chia-Chien Liu of Taichung Veterans General Hospital, took advantage of a rich but underused resource: the structured electronic medical record, or EMR, data generated routinely during clinical care. Rather than relying on small, highly selected research samples or broad insurance claims, the investigators examined patient-level records from a tertiary medical center covering the period from January 1, 2017 to March 31, 2023. This retrospective, period-based observational design allowed them to capture the messy reality of clinical practice, where diagnoses evolve, medications change, and service use unfolds unevenly over time.

The analytic sample comprised 511 patients aged 18 years or younger who carried eligible mood disorder diagnoses, along with a small subgroup diagnosed with personality disorders. The cohort skewed female, with 355 patients, or 69.5 percent, being girls, and the mean age was 15.7 years. Within this group, 76 patients, or 14.9 percent, experienced psychiatric hospitalization during the observation window. That figure provides a useful benchmark for clinicians and health services researchers alike: in a tertiary-care, mood-disorder-focused youth population, roughly one in seven young people required inpatient psychiatric care.

To identify the characteristics associated with hospitalization, the researchers employed multivariable logistic regression, a statistical technique that estimates the independent contribution of each variable while holding the others constant. Their models incorporated demographic factors such as age and sex, broad diagnostic categories, medication exposures recorded at the index qualifying encounter, psychotherapy-related service exposure, and the duration of observation. This approach matters because naive comparisons can be deeply misleading. For example, if older adolescents are more likely to receive certain medications and also more likely to be hospitalized, a simple analysis might wrongly implicate the medication itself. Multivariable modeling helps disentangle such overlapping signals, although it cannot eliminate them entirely.

The results were striking in one central respect: treatment characteristics at the index encounter, rather than demographic or diagnostic categories, were the variables most consistently associated with hospitalization. Patients whose records showed exposure to antipsychotic medications at the index encounter had roughly three times the odds of hospitalization compared with those without such exposure, with an odds ratio of 3.03 and a 95 percent confidence interval spanning 1.65 to 5.57. Exposure to anticonvulsant medications used as mood stabilizers was similarly associated, with an odds ratio of 3.25 and a confidence interval of 1.16 to 9.09. Most dramatic of all was lithium: the 28 patients, or 5.5 percent of the cohort, with recorded lithium exposure at the index encounter showed an odds ratio of 10.64 for hospitalization, with a confidence interval of 3.76 to 30.12.

Before those numbers spark alarm about the medications themselves, the authors are emphatic about how they should be read. Lithium, antipsychotics, and anticonvulsant mood stabilizers are typically reserved for the most severe, complex, or treatment-resistant presentations in child and adolescent psychiatry. A teenager prescribed lithium for bipolar-spectrum illness is, almost by definition, a young person whose illness has already proved serious enough to warrant a potent agent with demanding monitoring requirements. In other words, these medication exposures function as markers of clinical severity and treatment complexity, not as causes of hospitalization. The study’s design, which records exposure at a single index encounter and observes hospitalization over a subsequent period, cannot establish temporal ordering or causal direction, and the authors explicitly caution against interpreting the odds ratios as medication effects.

Psychotherapy-related service exposure told a parallel story. Young patients whose records showed contact with psychotherapy services had seven times the odds of hospitalization, with an odds ratio of 7.00 and a confidence interval of 3.65 to 13.42. Again, the most plausible interpretation is not that therapy drives admission, but that both therapy and hospitalization reflect a higher intensity of care. Patients who are engaged in structured psychotherapy are often those with more severe symptoms, greater functional impairment, or more vigilant clinical follow-up, all of which increase the likelihood that a deterioration will be detected and escalated to inpatient treatment. The association captures care intensity, the authors note, without any established sequence of events linking the two.

Equally informative were the variables that failed to show significant associations. Age, sex, broad diagnostic category, and exposure to benzodiazepine anxiolytics were not significantly linked to hospitalization once other factors were accounted for. This null finding carries real weight. It suggests that, within a cohort already focused on mood disorders, knowing a patient’s age, sex, or coarse diagnosis adds little predictive value compared with knowing what treatments that patient is actually receiving. For risk assessment and service planning, the granular details of the clinical record may matter far more than the demographic and diagnostic boxes that often dominate intake summaries.

The study also examined all-cause emergency department utilization, though the authors frame this as descriptive context rather than an independent psychiatric outcome. Only 14 patients, or 2.7 percent of the cohort, had records of emergency department visits for any cause during the observation period, and 12 of those 14 also experienced psychiatric hospitalization. The tight overlap between emergency contact and inpatient admission suggests that, in this Taiwanese tertiary-care setting, the emergency department functioned less as a routine entry point for youth mental health crises and more as a way station on the path to admission. This pattern contrasts with settings where emergency departments serve as the primary access route for acute psychiatric care, and it offers a comparative data point for health systems designing crisis pathways for young people.

Beyond its specific findings, the study illustrates the growing power of real-world electronic medical record research in psychiatry. Tertiary-care EMR cohorts can capture service utilization patterns that randomized trials and survey-based studies miss, particularly in Asian healthcare settings where such data have historically been scarce. The authors conducted sensitivity analyses, including one restricted to outpatient-index encounters and another excluding patients with personality disorder diagnoses, to probe the robustness of their results, and the study received institutional review board approval with the informed consent requirement waived given the retrospective use of de-identified data. The practical implications point toward a shift in how clinicians and health systems think about risk: rather than relying on demographic profiles or diagnostic labels to anticipate which young patients may need hospitalization, attention should turn to the treatment trajectory itself. A prescription for lithium, an antipsychotic, or an anticonvulsant mood stabilizer, or active involvement in psychotherapy, signals a young person already navigating a severe and complex illness, and that signal may be the most actionable early warning a clinical team has. As electronic records grow richer and analytical methods mature, studies of this kind could help psychiatric services worldwide move from reactive crisis management toward genuinely anticipatory care for the most vulnerable children and adolescents.

Subject of Research: Correlates of psychiatric hospitalization among youth with mood disorders using electronic medical record data

Article Title: Correlates of psychiatric hospitalization in a mood-disorder-focused youth psychiatric cohort: a real-world electronic medical record study

Article References: Wu, S.-Y., Chen, I.-C., Lan, C.-C., Hu, Y.-W., & Liu, C.-C. (2026). Correlates of psychiatric hospitalization in a mood-disorder-focused youth psychiatric cohort: a real-world electronic medical record study. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08699-2

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08699-2

Keywords: child and adolescent psychiatry, psychiatric hospitalization, mood disorders, electronic medical records, mental health service utilization, lithium, antipsychotics, psychotherapy, emergency department utilization, Taiwan, logistic regression, bipolar disorder