Machine Learning Identifies Changing Predictors of Medication Experience in Schizophrenia

A longitudinal mixed-methods study tracking clinically stable patients with schizophrenia across three distinct post-discharge stages has identified key clinical, psychological, and social factors associated with patients’ medication experiences over time. Utilizing repeated assessments alongside machine learning algorithms, the investigation revealed that the primary determinants of medication experience change dynamically as patients transition from inpatient care back into the community. Among the analytical approaches tested, random forest models consistently demonstrated superior predictive performance in identifying the most critical variables at each evaluation interval, highlighting how clinical symptoms, subjective treatment evaluations, and personal support systems interact throughout recovery.

The clinical background outlined by the study investigators emphasizes that schizophrenia requires sustained antipsychotic therapy, yet medication adherence remains suboptimal in more than 50 percent of patients. This pervasive challenge is largely driven by suboptimal medication experiences, which frequently prompt individuals to undertake self-directed dose reductions or complete treatment discontinuation. The authors noted that non-adherence substantially contributes to disease relapse, functional impairment, and heavy economic burdens on families, while recurrent psychiatric hospitalizations continue to place severe strain on broader healthcare systems, underscoring the urgent need to understand modifiable correlates of medication experiences.

To examine these factors across the post-discharge trajectory, the research team recruited a cohort of 425 clinically stable patients with schizophrenia from two tertiary psychiatric hospitals located in Heilongjiang Province, China. The investigators implemented a mixed-methods research design featuring repeated assessments that captured concurrent quantitative and qualitative data across three designated time points: prior to hospital discharge (T1), at one month post-discharge (T2), and at three months post-discharge (T3). This multi-stage timeframe allowed the researchers to observe how patient perceptions, symptoms, and social environments shifted during the critical transition from institutional care to community reintegration.

The quantitative component of the study incorporated several standardized measurement instruments to evaluate distinct clinical and psychosocial domains. Patient medication experience was measured using the Subjective Well-being under Neuroleptics scale, while medication burden was evaluated through the Living with Medicines Questionnaire. The researchers assessed medication beliefs using the Beliefs about Medicines Questionnaire, psychiatric symptomatology through the Positive and Negative Syndrome Scale, social support via the Social Support Rating Scale, and family functioning using the APGAR index. To analyze these multidimensional variables, the authors trained and compared multiple machine learning models, including logistic regression, random forest, and support vector machines. Concurrently, qualitative interviews conducted with 15 patients experiencing suboptimal medication experiences were analyzed thematically and integrated with the quantitative findings.

The resulting predictive models revealed that the primary factors influencing medication experience evolved over the follow-up period. At pre-discharge (T1), random forest models identified medication beliefs as the most critical predictor, followed by medication burden and psychotic symptoms in descending order of importance. By one month post-discharge (T2), medication burden emerged as the leading predictor, followed by medication beliefs, family functioning, and psychotic symptoms. At three months post-discharge (T3), medication beliefs regained the top ranking, followed by social support, medication burden, psychotic symptoms, and family functioning. The thematic qualitative analysis uncovered four core themes at T1 and three themes across T2 and T3, demonstrating complementary quantitative and qualitative patterns regarding how patients weigh benefit-risk perceptions and determine treatment valuation.

Based on these findings, the authors suggested that psychiatric nurses and healthcare providers consider stage-specific interventions targeting modifiable factors across recovery. During inpatient hospitalization (T1), interventions should focus on mitigating medication burdens while actively enhancing perceived treatment benefits. In the early post-discharge period (T2), clinical attention should prioritize strengthening medication beliefs and reinforcing family support systems. Finally, during community reintegration at three months (T3), clinical efforts should emphasize bolstering broader social support networks. The researchers concluded that identifying these specific intervention targets across post-discharge time points offers practical pathways to optimize medication experiences and may have meaningful implications for medication-related clinical outcomes in patients living with schizophrenia.

Subject of Research: Correlates of medication experience in patients with schizophrenia across post-discharge time points

Article Title: Factors associated with medication experience in patients with schizophrenia: a mixed-methods study across three post-discharge time points

Article References: Yao, D., Yu, H., Zhang, J., Jiang, W., Wang, Z., & Sun, Y. (2026). Factors associated with medication experience in patients with schizophrenia: a mixed-methods study across three post-discharge time points. Frontiers in Psychiatry, 17, Article 1902239. https://doi.org/10.3389/fpsyt.2026.1902239

Image Credits: AI Generated

DOI: 10.3389/fpsyt.2026.1902239

Keywords: schizophrenia, medication experience, antipsychotic adherence, random forest, medication burden, psychiatric nursing, longitudinal study, mixed-methods research, Factors, associated, medication, experience