A single-center retrospective controlled study design was employed, with data obtained from the electronic medical records of the psychiatry department of Liupanshui Shuikuang hospital. A total of 168 patients who received standardized treatment with aripiprazole and completed baseline and follow-up assessments from May 2023 to June 2025 were included. Patients were naturally divided into the combined treatment group (aripiprazole + CBT) and the monotherapy group (aripiprazole only), with 84 patients in each group. To minimize selection bias, all eligible cases were consecutively included. To reduce baseline bias related to the retrospective and non-randomized design, propensity score matching (PSM) was performed (8). Receipt of combination therapy was treated as the dependent variable, and a logistic regression model was used to estimate the propensity score for each patient. Based on clinical relevance and potential confounding effects, the following baseline covariates were included: age, sex, disease duration, baseline PANSS total score, PANSS positive subscale score, PANSS negative subscale score, PANSS general psychopathology subscale score, and baseline antipsychotic dosage. A 1:1 nearest-neighbor matching method without replacement was applied. The caliper width was set at 0.2 of the standard deviation of the logit of the propensity score. After matching, balance between the two groups was assessed using standardized mean differences (SMD). An SMD value < 0.1 was considered indicative of adequate balance for both continuous and categorical variables. The matched cohort was used for subsequent efficacy analyses.

The baseline time point (T0) was defined as the most recent complete evaluation before the start of stable aripiprazole treatment. Follow-up data was collected at 3 months (T1) and 6 months (T2) to assess changes in clinical symptoms, cognitive function, and emotional state over time. The data used were from previously stored records without additional interventions.

The study subjects were consecutively enrolled schizophrenia patients, with all cases retrieved from the electronic medical record system. Since this was a retrospective analysis, the study subjects were based on historical medical record data and no additional interventions were applied. The Ethics Committee of Shuikuang Hospital approved this study (Ethics Approval No. skyy-II-20251209), which did not require patients to sign informed consent personally. However, the personal information of all patients was kept strictly confidential and processed in a de-identified manner to ensure compliance with data protection regulations. The inclusion criteria were: (1) meeting the schizophrenia diagnostic criteria in the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5) (1), as clearly recorded by a psychiatrist in the medical records; (2) aged 18–65 years; (3) complete and traceable electronic medical record data. The exclusion criteria were: (1) clear organic brain diseases, such as post-traumatic syndrome, brain tumors, epileptic encephalopathy, or severe cerebrovascular lesions; (2) previously diagnosed neurodegenerative diseases, such as Alzheimer’s disease or frontotemporal dementia; (3) significant alcohol or drug dependence (except smoking or occasional drinking); (4) severe physical diseases (such as cardiovascular diseases, diabetes, liver or kidney dysfunction) that could interfere with the assessment of cognitive, emotional, and treatment outcomes; (5) use of other psychiatric medications (such as antidepressants or other antipsychotics) during treatment; (6) missing key variables or poor quality of scale data in the medical records; (7) previous mild cognitive impairment or other psychiatric comorbidities (such as severe depression or anxiety disorders). To ensure the representativeness and reproducibility of the sample, eligible cases were included in the analysis sequentially. The case screening process is shown in Figure 1.

Monotherapy Group (Aripiprazole only): Patients in this group only received aripiprazole (Shanghai Shiyuan Pharmaceutical Co., Ltd.; National Drug Standard H20041506; 5mg tablets). The starting dose was set at 10 mg per day, once daily, regardless of meals. If patients tolerated the drug well and did not show significant clinical improvement, the dose was gradually increased to 15–30 mg/day after 2 weeks, with a maximum dose of 30 mg/day, and this dose was maintained until the end of the study. During maintenance treatment, the treating physician decided whether to continue within the 15–30 mg dose range based on the patient’s clinical response and tolerance. This dose-time design follows the recommended dosing regimen for aripiprazole in the treatment of schizophrenia (1). During this period, patients did not receive any form of CBT or other psychological interventions, and the treatment solely relied on the pharmacological effects of aripiprazole. Patients were clinically followed up monthly to assess drug efficacy and side effects, and dose adjustments were made when necessary.

Combined Treatment Group (Aripiprazole + CBT intervention): In addition to monotherapy, patients in this group also received structured CBT. Cognitive behavioral therapy was delivered by psychiatrists or psychotherapists who had received formal CBT training and had relevant clinical experience. All therapists held a master’s degree or higher in psychology and had at least two years of clinical practice. Before study initiation, all therapists completed a unified training program on the treatment protocol. Interventions were conducted according to a standardized CBT manual. Core components included cognitive restructuring, emotional regulation, and behavioral interventions. These measures were used to ensure consistency in treatment procedures and key session elements. During the study period, therapists completed standardized treatment records after each CBT session. The research team held case review meetings every two weeks. During these meetings, a random sample of cases was selected for review of treatment records. The implementation of core intervention modules and adherence to the predefined treatment protocol were evaluated. When deviations were identified, feedback and protocol calibration were provided during the discussions to minimize inter-therapist variability. The treatment included the following four parts:

All scales and data used in this study were derived from the electronic medical record system. Baseline and follow-up data for all patients were recorded and rated by trained physicians and psychological evaluators according to standard procedures. The details are as follows:

To assess the impact of different treatment intensities on efficacy, the study performed a dose-response analysis using the actual maintenance doses of aripiprazole and CBT treatment frequencies as one of the main independent variables. Treatment frequencies were divided into low frequency (once a week), medium frequency (twice a week), and high frequency (three times a week). The doses of aripiprazole were classified as low dose (10–20 mg/day), standard dose (20–30 mg/day), and high dose (30 mg/day). The study analyzed changes in PANSS scores, MoCA scores, PHQ-9, and GAD-7 scores as dependent variables using linear regression models to explore the relationship between treatment intensity and treatment outcomes.

To explore how baseline characteristics of patients influenced treatment outcomes, the study divided patients into subgroups based on age, disease duration, cognitive level (MoCA score), and anxiety/depression levels (PHQ-9, GAD-7 scores). Specifically, the groups were divided as follows: younger group (<40 years), middle-aged group (40–60 years), older group (>60 years); early disease duration group (<5 years) and late disease duration group (>5 years); as well as different cognitive abilities and emotional levels (grouped based on median MoCA, PHQ-9, and GAD-7 scores). Multivariate regression analysis was then used to examine whether these baseline characteristics moderated the treatment outcomes, particularly the differences in PANSS total score, MoCA improvement, and emotional symptom ratings.

All data were analyzed using SPSS. First, for data that did not meet normal distribution, the Mann–Whitney U test was used for non-parametric comparisons. Continuous variables were expressed as mean ± standard deviation or median (interquartile range) and compared between groups using independent sample t-tests or Mann–Whitney U tests. Categorical variables were expressed as frequencies and percentages, and the χ² test was used for comparisons.

In multivariate regression analysis, in addition to the basic covariates (such as age, sex, disease duration), we also included factors such as lifestyle (e.g., smoking, drinking habits, physical activity), social support (e.g., family support, friend support), education level, and income, to further control for potential confounders. In addition, sensitivity analysis was performed to verify the robustness of the regression model under different combinations of covariates. The results showed that treatment effects remained stable under different covariate conditions, further supporting the effectiveness of CBT.

For missing data, the study used multiple imputation (Multiple Imputation) to generate complete datasets through repeated imputations based on existing data, which reduced potential bias due to missing data. During the imputation process, multiple covariates and dependencies were considered to ensure the accuracy and robustness of the imputed results. The imputed dataset was used for regression analysis to further validate the robustness of the results. All analyses were performed with two-sided tests, and a P value < 0.05 was considered statistically significant.

After propensity score matching, the two groups showed comparable distributions in baseline demographic and clinical characteristics. These included age, sex composition, disease duration, educational level, living arrangement, smoking status, alcohol consumption, marital status, and major physical comorbidities. With respect to disease severity indicators, good balance was also observed between groups in baseline PANSS total score and in the positive, negative, and general psychopathology subscale scores. All baseline variables demonstrated standardized mean differences below 0.1, indicating satisfactory post-matching comparability between the two groups. (Supplementary Table 1).

Compared to T0, both groups showed significant reductions in PANSS total score and subscale scores at T1 and T2 (within-group comparisons, P < 0.05). At the T2 assessment, the combined treatment group had a lower PANSS total score than the monotherapy group, and there was a significant difference in the improvement between the two groups (improvement value of −28.3 ± 9.5 for the combined treatment group, and −21.9 ± 10.1 for the monotherapy group, P = 0.008). The trends for improvement in positive, negative, and general psychopathology subscales were consistent with the total score. At T2, the improvement in all subscales was greater in the combined treatment group compared to the monotherapy group (P < 0.01 for all) (Figure 2, Supplementary Table 2).

Both groups showed significant improvements in MoCA scores at T1 and T2 compared to baseline (within-group comparisons, P < 0.05). At T2, the combined treatment group had a higher MoCA score than the monotherapy group (25.8 ± 3.2 vs. 24.0 ± 3.3, P = 0.001). Further comparison of the improvement from T0 to T2 showed that the combined treatment group had a greater improvement in MoCA (4.4 ± 2.6 vs. 2.9 ± 2.5, P = 0.002) (Table 1).

Depression and anxiety scores in both groups decreased with treatment time (PHQ-9 and GAD-7 within-group comparisons, P < 0.05). At T2, the combined treatment group had lower PHQ-9 scores compared to the monotherapy group (6.2 ± 3.1 vs. 8.0 ± 3.4, P = 0.003), and a similar difference was observed in GAD-7 scores (5.8 ± 2.9 vs. 7.4 ± 3.2, P = 0.004). When comparing the change from T0 to T2, the combined treatment group showed a larger reduction in depression and anxiety scores than the monotherapy group (PHQ-9: -5.3 ± 2.9 vs. -3.3 ± 2.7;GAD-7:-4.8 ± 2.7 vs. -3.2 ± 2.6; P values all < 0.01) (Supplementary Table 3).

At both T1 and T2, both groups showed a decrease in their SQLS total scores compared to baseline (within-group comparisons, P < 0.05). At T2, the combined treatment group had a significantly lower SQLS total score than the monotherapy group (42.5 ± 11.3 vs. 48.7 ± 12.1, P = 0.006). When analyzing the subscales, the combined treatment group also showed lower scores in the motivation and vitality and somatic-psychological dimensions (motivation and vitality: 13.2 ± 4.8 vs. 15.6 ± 5.1, P = 0.010; somatic-psychological: 14.0 ± 5.0 vs. 16.3 ± 5.4, P = 0.018), while the difference in the symptom discomfort dimension was not significant (P > 0.05) (Table 2).

Changes in PANSS total score and MoCA score were used as dependent variables in a multivariate linear regression model. Whether patients received CBT was included as the main independent variable, and covariates such as age, gender, disease duration, baseline PANSS total score, and comorbidity count were also included. The results showed that CBT combined treatment was significantly associated with greater reductions in PANSS total score and greater improvements in MoCA score (PANSS: β = −6.2, 95% CI −9.1 to −3.3, P < 0.001; MoCA: β = 1.3, 95% CI 0.6 to 2.0, P < 0.001). Further adjustment for baseline depression and anxiety scores confirmed the significance of CBT effects, indicating that CBT independently contributed to improvements in psychiatric symptoms and cognitive function (Supplementary Table 4).

The study results showed that increased treatment intensity significantly improved patients’ symptoms and cognitive function. The high-dose, high-frequency CBT treatment regimen led to the greatest improvements in PANSS total scores and MoCA scores, with PHQ-9 scores improving by -6.3 points and GAD-7 scores improving by -4.5 points (P < 0.05). In contrast, the low-dose, low-frequency CBT group showed much smaller improvements, with a PANSS total score improvement of 15.6 points (P = 0.002) and a MoCA score improvement of 2.4 points (P = 0.049). This result indicates that both the dosage and frequency of treatment are positively correlated with efficacy (Figure 3a).

There were significant differences in treatment effects across different patient subgroups. The younger patient group (<40 years) showed significantly greater improvements in PANSS and MoCA scores compared to the older patient group (>60 years) (P < 0.01). Additionally, patients with early disease duration (<5 years) showed better PANSS improvement (25.8 points vs. 21.1 points, P = 0.009), and greater reductions in PHQ-9 and GAD-7 scores (-5.8 points and -4.3 points, respectively, P < 0.05). Patients with low baseline cognitive levels showed significantly better improvement in PANSS, MoCA, PHQ-9, and GAD-7 scores compared to those with high baseline cognitive levels (P < 0.05) (Figure 3b).