Comparison of multiple machine learning methods to identify the needs of family members of patients in the neurosurgical ICU and analyse the efficacy of precision nursing in improving patients’ family anxiety

Background The neurosurgical intensive care unit (NICU) is the specialised unit for managing acute and critical neurosurgical diseases. Patients in the NICU face a high risk of neurological impairment and unpredictable disease progression. As constant emotional supporters, family members experience psychological distress that impairs their mental health and even affects the patient’s recovery. Consequently, the mental health of family members has become an important public health concern.To assess the prevalence of anxiety and depression among family members of NICU patients, identify their core needs, and evaluate whether a needs-oriented precision nursing model could alleviate their anxiety and improve satisfaction. Methods A total of 350 family members of NICU patients (with patient’s NICU stay ≥ 1 week) were surveyed, and 160 patients participated in the subsequent randomised controlled trial.Patients' family members were assessed using the Hospital Anxiety and Depression Scale (HADS) and the Critical Care Family Needs Inventory (CCFNI) to evaluate their anxiety and family needs. Three machine learning models—Random Forest, Support Vector Machine, and Artificial Neural Network—were compared to determine the optimal model for identifying core family needs. Based on the CCFNI analysis, precision nursing interventions were developed and evaluated using Family Satisfaction in the Intensive Care Unit-24 (FS-ICU-24) and HADS. Results Among the 350 family members, 64.3% (225 cases) experienced anxiety. Among the three machine learning models, the Random Forest model exhibited superior performance: accuracy = 0.9904, sensitivity = 0.9851, specificity = 1.0000, precision = 1.0000, F1-score = 0.9925, and AUC = 1.0000. The Random Forest model identified the top three core needs as follows: "I need to be with the patient as much as possible", "I need a clear explanation of the patient’s condition", and "I need to understand the purpose of each treatment". These informed the development of individualised precision nursing interventions for 160 patients. Before intervention, anxiety scores did not differ significantly between groups (12.95 ± 2.83 vs. 12.93 ± 2.87, t = 0.057, P = 0.954). After intervention, anxiety score decreased significantly in the intervention group (8.39 ± 2.88), but not in the control group (12.78 ± 3.08), with a significant difference between groups (t = 9.312, P < 0.001 ). Results of the FS-ICU-24 questionnaire showed that the total score of the intervention group (112.60 ± 2.12) was significantly higher than that of the control group (81.60 ± 2.40) (t = 67.737, P < 0.001 ). Scores of care satisfaction, decision-making satisfaction, participation in treatment decisions, and understanding of medical recommendations were all significantly higher in the intervention group than in the control group (all P < 0.001 ). Conclusions Family members of NICU patients commonly experience anxiety. The Random Forest model efficiently identifies their core needs, and precision nursing intervention based on these needs can significantly alleviate family members’ anxiety and improve their satisfaction with nursing services.