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
A
Feng
Zhang
1
Emaildczhangfeng2025@163.com
RuixiangSun
1
Emailsjs20180530@163.com
Jing
Huang
1
Emailhuangjing@yjsyy.com
Zhiqing
Zhou
1
Emailzhouzhiqing@yjsyy.com
Ting
Yang
1
Yanling
Li
1
Email452745372@qq.com
Li
hong
Min
1
Email895907205@qq.com
Zuan
Yu
1
Emailyuzuan@yjsyy.com
Jiaqiang
Liu
1
Emailljq1031@hotmail.com
Huayue
Zhang
2
Lei
He
3✉
Emailhelei15830837@126.com
A
Ping
Xu
1✉
Zhang
4,5,7,8
Huang
4
Zhou
4
Email1223880286@qq.com
Ting
Last
Name
6
Yang
1
Li
6
Min
1
Email252709889@qq.com
Zuan
Last
Name
6
Yu
1
Jiaqiang
Last
Name
6
Liu
6
Xu
4
Email1315636809@qq.com
1
The First Affiliated Hospital of Wannan Medical Colleg
Zheshan West Road on the 2nd
241000
Wuhu
Anhui
China
2
The first people’s Hospital of Chuzhou City
No. 369 zuiwang West Road, Nanqiao District
239001
Chuzhou
Anhui
China
3
The First Affiliated Hospital of Soochow University
No. 899, Pinghai Road
215000
Suzhou
Jiangsu
China
4
No. 2, Zheshan West Road. City/postcode
241000
Wuhu, Country
Anhui
China
5
Department of Critical Care Medicine
241000
Wuhu, Country
Anhui
China
6
Department of Neurosurgery
241000
Wuhu, Country
Anhui
China
7
Department of Cardiovascular Medicine
No. 369 West Zuiweng Road, Nanqiao District
239001
Chuzhou City City, Chuzhou, Anhui, Country
China
8
Department of Cardiology
No. 899, Pinghai Road City/Post code
215000
Suzhou, Country
China
Feng Zhang1; RuixiangSun1;Jing Huang1;Zhiqing Zhou1; Ting Yang; Yanling Li1; Li hong Min1; Zuan Yu1; Jiaqiang Liu1; Huayue Zhang2; Lei He3,*; Ping Xu1,*
1. The First Affiliated Hospital of Wannan Medical Colleg, Zheshan West Road on the 2nd, Wuhu 241000, Anhui,China
2.The first people's Hospital of Chuzhou City, No. 369 zuiwang West Road, Nanqiao District,Chuzhou 239001, Anhui,China
3.The First Affiliated Hospital of Soochow University,No. 899, Pinghai Road,Suzhou 215000,Jiangsu,China
*Correspondence Author:
Ping Xu1 ; Lei He3
1. the First Affiliated Hospital of Wannan Medical Colleg, Zheshan West Road on the 2nd, Wuhu 241000, Anhui,China
3.The First Affiliated Hospital of Soochow University,No.899, Pinghai Road,Suzhou 215000,Jiangsu,China
Title: Deputy Chief Nurse
Name: Feng
Last name: Zhang
Email: 1223880286@qq.com
Unit name (University): the First Affiliated Hospital of Wannan Medical College
Department/laboratory: Neurosurgery
Address: No. 2, Zheshan West Road.
City/postcode: Wuhu, Anhui 241000
Country: China
Title: Nurse in charge
First Name: Ruixiang
Last Name: Sun
Email: sjs20180530@163.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory: Department of Critical Care Medicine
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Deputy Chief Nurse
Name: Jing
Last name: Huang
Email:huangjing@yjsyy.com
Unit name (University): the First Affiliated Hospital of Wannan Medical College
Department/laboratory: Pediatrics
Address: No. 2, Zheshan West Road.
City/postcode: Wuhu, Anhui 241000
Country: China
Title: chief nurse
Name: Zhiqing
Last name: Zhou
Email:zhouzhiqing@yjsyy.com
Unit name (University): the First Affiliated Hospital of Wannan Medical College
Department/laboratory: Nursing Department
Address: No. 2, Zheshan West Road.
City/postcode: Wuhu, Anhui 241000
Country: China
Fund:2024 university scientific research plan compilation project of Wannan Medical College (2024AH040381)
Title: Title: associate chief nurse
First Name: Ting
Last Name: Yang
Email: 252709889@qq.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory: Department of Neurosurgery
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Nurse in charge
First Name: Yanling
Last Name: Li
Email:452745372@qq.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory:Department of Neurosurgery
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Nurse in charge
First Name:Lihong
Last Name: Min
Email: 895907205@qq.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory: Department of Neurosurgery
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Resident physician
First Name: Zuan
Last Name: Yu
Email: yuzuan@yjsyy.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory: Department of Neurosurgery
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Attending physician
First Name:Jiaqiang
Last Name:Liu
Email: ljq1031@hotmail.com
Organization Name (University): the First Affiliated Hospital of Wannan Medical Colleg
Department/Laboratory: Department of Neurosurgery
Address: Zheshan West Road on the 2nd.
City/Post code: Wuhu, Anhui 241000
Country: China
Title: Nurse Practitioner
First Name:HuaYue
Last Name: Zhang
Email: 1315636809@qq.com
Organization Name (University): The First People's Hospital of Chuzhou
Department/Laboratory: Department of Cardiovascular Medicine
Address: No. 369 West Zuiweng Road, Nanqiao District, Chuzhou City
City/Post code: Chuzhou, Anhui, 239001
Country: China
Title: Nurse in charge
First Name: Lei
Last Name:He
Email: helei15830837@126.com
Organization Name (University): The First Affiliated Hospital of Soochow University
Department/Laboratory: Department of Cardiology
Address: No. 899, Pinghai Road
City/Post code: Suzhou, 215000
Country: China
Title: Deputy Chief Nurse
Name: Ping
Last name: Xu
Email:dczhangfeng2025@163.com
Unit name (University): the First Affiliated Hospital of Wannan Medical College
Department/laboratory: Neurology
Address: No. 2, Zheshan West Road.
City/postcode: Wuhu, Anhui 241000
Country: China
Abstract
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
A
A
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.
KEYWORDS:
neurosurgical Intensive care unit (NICU)
precision nursing
family needs
machine learning
anxiety and depression
The neurosurgical intensive care unit (NICU) serves as the primary centre for managing acute and critical neurosurgical conditions. Its primary mission is to provide care for patients at high risk of neurological impairment, such as those with severe traumatic brain injury (sTBI), aneurysmal subarachnoid haemorrhage or post-craniotomy for intracranial tumours. These patients often experience unpredictable disease progression and require continuous medical intervention. Patients in the NICU are typically in critical and unstable conditions. Studies indicate that the in-hospital mortality rate of patients with sTBI in the NICU can reach 46%, and the incidence of functional impairment remains as high as 40% at 6 months postoperatively[1–2]. Therefore, timely and effective diagnosis, treatment, and nursing care are crucial for optimising patient prognosis.
Family members act as continuous emotional supporters throughout a patient’s illness trajectory, and their psychological states evolve dynamically alongside the patient’s diagnosis and treatment. Prolonged anxiety not only endangers their own mental health but may also hinder the patient’s postoperative recovery through emotional transmission. However, the psychological needs of this group are often overlooked in clinical practice[3].With the steady increase in NICU admissions, the mental health of family members has become an important public health concern. Concurrently, as healthcare and nursing philosophies shift toward a "patient-centered" approach that also recognises family needs, the ability to provide psychological support to families has gradually become a key indicator of NICU nursing quality. This, in turn, directly influences satisfaction with medical services and public trust in the healthcare system. Therefore, research on the psychological well-being of NICU family members has become a key focus in this field.
Existing studies have confirmed that family members of ICU patients commonly experience psychological distress and that the incidence of anxiety and depression among NICU family members is significantly higher than that among family members of patients in general ICUs. The studies conducted a prospective observation of 84 primary family members of ICU patients and found that 52.4% exhibited anxiety symptoms meeting diagnostic criteria, with anxiety scores showing a negative correlation with the severity of neurological impairment measured by the Glasgow Coma Scale (GCS)[4]. Previous research further reported that over 70% of NICU family members experienced both anxiety and depression. Anxiety arising from concerns about the patient’s inability to regain self-care ability postoperatively accounted for 63% of cases, which was significantly higher than the proportion of family members in general ICUs[5]. However, while the studies identified limited information access, low participation in medical decision-making, and strong feelings of environmental isolation as the main causes of ICU family anxiety, their analysis combined family members from several speciality ICUs and failed to identify the core stressors unique to NICU family members[6].
Studies demonstrated that video visits and individualised information-based interventions can reduce short-term anxiety; however, the former did not address the unique needs of NICU family members for information on neurological function, while the latter only focused on short-term effects and did not consider differences in family members’ decision-making preferences[7–82]. Previous research used a neural network–based machine learning model to analyse NICU family needs, and concluded that it outperformed traditional statistical methods; however, they did not compare the accuracy of different machine learning models[9].
Long-term clinical practice has revealed that the core triggers of anxiety among NICU family members exhibit phase-specific characteristics. In the early stage, anxiety stems from a limited understanding of neurological assessment indicators and inadequate awareness of treatment plans. In the later stage, it arises from uncertainty regarding the recovery process and potential complications. However, the existing nursing protocols fail to address these phase-specific psychological needs.
Precision nursing, which centres on individualised needs assessments, aims to maximise nursing effectiveness by accurately identifying patients’ or families’ unique needs and developing targeted interventions accordingly[10]. The studies noted that the sources of anxiety among ICU family members vary widely, requiring a systematic needs assessment to identify core priorities before implementing targeted interventions[11].
Given that previous studies have insufficiently explored the psychological changes and needs of NICU family members, and that existing interventions are often limited in scope and lack specialty-specific adaptation, resulting in suboptimal outcomes, this study aimed to: ① assess the anxiety status of NICU family members using the Hospital Anxiety and Depression Scale (HADS); ② analyse the results of the Critical Care Family Needs Inventory (CCFNI) using three commonly used machine learning models—Random Forest, Support Vector Machine, and Neural Network—and compare their accuracy to determine the optimal model for ranking family members’ needs; ③ design a multi-component intervention programme tailored to the NICU contest, verifying its effectiveness using the Family Satisfaction in the Intensive Care Unit-24 (FS-ICU-24) questionnaire combined with anxiety scores. The findings of this study will provide the rationale and feasibility of needs-based targeted interventions to effectively alleviate anxiety among family members, enhance nursing satisfaction, and foster a positive emotional environment conducive to patient recovery.
A
A
Method
Study Design
This was a single-centre retrospective analysis. First, the HADS was used to screen for anxiety among the primary relatives of patients in the NICU. Subsequently, data from CCFNI were analysed using three machine learning models: Random Forest, Support Vector Machine, and Neural Network[12–15]. This approach was adopted to overcome the limitations of traditional statistical methods and to identify the optimal machine learning model for CCFNI data analysis[16].(Fig. 1)
Fig. 1
Experimental Flow Chart
Data Collection
1042 family members of NICU patients with a GCS score of < 8 were recruited as study participants. From January 2024 to December 2024 were selected in this research from Department of Neurosurgery, Yijishan Hospital of Wannan Medical College, after screening according to the inclusion and exclusion criteria, complete survey data were collected from 350 family members and included in the study.
From January 2025 to June 2025, another cohort of family members of NICU patients with a GCS score < 8 was recruited. Complete data were obtained from 160 family members after applying the same inclusion and exclusion criteria. These 160 participants were randomly assigned to one of two groups.The control group is from January to March 2025, and the intervention group is from April to June 2025.The conventional nursing group (control group) and the needs-based intervention group (intervention group), with 80 participants in each group.
Questionnaires were administered on the 7th day after admission to the NICU. Inclusion Criteria: ① Primary relatives (spouse, parent, or child) of an NICU patient, and serving as the primary decision-maker for the patient’s care. ② Having accompanied the patient for ≥ 7 days during the patient’s NICU stay, and being capable of cooperating to complete the questionnaire. ③ No history of mental illnesses, such as anxiety disorder or depression, and no cognitive impairment. Exclusion Criteria were as follows: ① Family members with severe physical illnesses (e.g., heart failure or malignant tumour) that may affect the quality of the questionnaire completion. ② Patients transferred to a general ward within 1 week of admission to the NICU. ③ Patients who died during the questionnaire survey period.
Intervention Measures
Conventional Nursing Group (Control Group)
Family members in this group received routine nursing care from NICU, which included the following: Health Education: Conducted by nurses at the time of the patient’s admission into the NICU and once a week thereafter. This was a one-way explanation session lasting approximately 20–30 min, covering basic knowledge of the patient’s disease, the diagnostic basis for the disease, and the current treatment plan. After the session, family members were given the opportunity to ask questions on the spot. Visit Management: A fixed visiting schedule was implemented, allowing each visit to last 30 min and occur once or twice a week. Before each visit, nurses informed family members of relevant precautions; during the visit, nurses also answered any questions raised by family members. Consultation Response: Family members could consult the responsible nurse either via telephone or by visiting the consultation room. Nurses provided timely responses in line with their work schedules to ensure that no important questions were left unanswered.
Needs-Based Intervention Group (Intervention Group)
Building upon conventional nursing care, family members in the intervention group received multi-dimensional precision-nursing interventions developed based on the analysis results of the CCFNI scale. The interventions included the following measures: Flexible Visit System: A combination of video and extended visits during non-peak hours was adopted. Video visits were conducted once every 2 days, for 15 min each, using a dedicated video platform to enable family members to clearly observe the patient’s condition. Extended visits were allowed during non-peak hours (16:00–17:00), lasting 1 h each time, with one to two family members permitted to enter the ward. During these visits, nurses explained the details of patient care for that day and guided family members in performing simple care procedures. Disease Explanation Assisted by Virtual Reality Technology: Based on the patient’s imaging data, virtual reality technology was used to construct a three-dimensional model of the intracranial lesion and its surrounding nerves and blood vessels. Nurses used this model to visually demonstrate the extent of neurological impairment and the relationship between the lesion and surrounding anatomical structures to family members. They also provided a detailed explanation of the treatment plan and recorded comprehensive information about the patient’s daily care, including positioning care, complication prevention measures, and daily neurological function assessment results. Each day from 15:00 to 15:30, nurses met face-to-face with family members to discuss the contents of the care diary, interpret the recorded information, answer questions, document feedback, and adjust the focus of subsequent care records according to family’s needs. Specialized Guidance on Neurological Function Information: Twice a week, specialist neurosurgical nurses delivered lectures on neurological function, presented using PowerPoint slides, and provided personalized answers to family members’ questions.(Fig. 2)
Fig. 2
Pictures of Precision Care Based on Needs
A.Nursing Diary: Nurses explain the details of the patient's daily care. B. Video Visitation: Conducted via a dedicated video platform, enabling family members to clearly observe the patient's condition. C. Application of Virtual Reality (VR) Technology: By utilizing the patient's imaging examination data such as CT and MRI, a three-dimensional model of intracranial lesions and surrounding nerves and blood vessels is constructed to assist in explaining the patient's condition.
Stratified Communication on Rehabilitation Expectations: Within 1 week of the patient’s admission, the attending physician and rehabilitation therapist jointly conducted two specialised communication sessions on rehabilitation expectations, each lasting 30 min.
A
A
All methods were carried out in accordance with relevant guidelines and regulations. The study was carried out respecting the Declaration of Helsinki in its current version. A
A
The study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Wannan Medical Colleg (No.2018043), and written informed consent was obtained from all the participants.Observation Indicators
General Clinical Data
Patient-related data included the time of admission and the type of disease. Family member-related data included sex, age, educational level, ethnicity, religion, relationship with the patient, average monthly household income (in RMB), occupation, and previous experience accompanying hospitalised patients .
Hospital Anxiety and Depression Scale[17]
This HADS is used to screen and assess anxiety and depression in both the general population and specific patient groups. It is suitable for the rapid evaluation of psychological state in non-psychiatric clinical settings. The scale consists of 14 items divided into two independent dimensions: "Anxiety" and "Depression" (7 items each). Each item is scored on a 0–3 scale. The total score for the anxiety dimension ranges from 0 to 21 and is classified as follows: no anxiety symptoms (0–7 points), mild anxiety (8–10 points), moderate anxiety (11–14 points), and severe anxiety (15–21 points). This HADS enables quantitative assessment of anxiety severity and is recognised as a concise, efficient, reliable, and valid tool for clinical evaluation.
Critical Care Family Needs Inventory[18]
The CCFNI is a specialised tool for assessing the needs of family members of critically ill patients in settings such as ICUs. It systematically identifies family members’ needs across multiple dimensions, including information acquisition, emotional support, participation in medical decision-making, and environmental security. It provides a targeted basis for the formulation of clinical care plans and demonstrates excellent reliability, validity, and cross-cultural applicability. In addition, it can accurately capture both the explicit and potential needs of family members and is widely used in clinical practice and research involving family members of critically ill patients.
Family Satisfaction in the Intensive Care Unit-24 (FS-ICU-24) Scale[16]
The FS-ICU-24 is a specialised tool used to assess the satisfaction of family members of ICU patients.
A
It quantitatively evaluates family members’ overall satisfaction with medical and nursing services and is particularly suitable for examining the effects of family care on acute and critical illnesses. The scale contains 24 items covering four core dimensions: information communication, doctor-patient interaction, quality of care, and participation rights. Each item is rated on a 1–5 scoring scale, giving a total score range of 24 to 120. Cross-cultural studies have confirmed the scale’s reliability and validity. The FS-ICU-24 not only directly reflects family members’ evaluation of ICU services but also shows a significant negative correlation with their anxiety levels, enabling simultaneous assessment of the dual effects of intervention measures to improve satisfaction and alleviate anxiety. It is widely used to evaluate ICU nursing quality and related clinical research.Statistical Analysis
This study combined traditional statistical analysis and machine learning models for data processing.
Traditional Statistical Analysis
IBM SPSS Statistics for Windows, version 26.0 (IBM Corp., Armonk, N.Y., USA), was used for the traditional statistical analysis. After testing the normality of continuous data, variables conforming to a normal distribution are described as mean ± standard deviation (x ± s), and inter-group comparison was performed using the independent samples t-test. Data that did not conform to a normal distribution are expressed as median (interquartile range) [M (Q1, Q3)], and inter-group comparison was conducted using the Mann-Whitney U test. Categorical data are expressed as count (percentage) [n (%)], and the chi-square (χ²) test was used for inter-group comparison. All statistical tests were two-tailed, with a significance level (α) of 0.05. Statistical significance was set at P-value < 0.05. significant.
Machine Learning
Model Construction and Comparison
Three classic machine learning algorithms: Random Forest, Support Vector Machine, and Artificial Neural Network, were used to construct models predicting anxiety states. The aim was to identify the optimal prediction model.
Data Preprocessing
The anxiety state was defined as the target variable, where family members with an anxiety scale score > 7 points were classified as the anxious group, and those with scores of ≤ 7 as the non-anxious group. The analytical variables included 15 psychometric indicators, from CCFNI1 to CCFNI15. All continuous variables were standardised to eliminate the influence of dimensions. The dataset was randomly divided into training and test sets in a 7:3 ratio to ensure a balanced distribution of anxiety states between the two sets.
Model Construction and Parameter Setting
Random Forest Model
An ensemble-learning strategy was adopted. The number of decision trees was set to 500, and the number of randomly selected features for each tree split was the square root of the total number of features. Two indicators—Mean Decrease Accuracy and Mean Decrease Gini coefficient—were used simultaneously to assess variable importance.
Support Vector Machine model
The radial basis function was selected as the kernel function. The penalty parameter (C) and kernel function parameter (γ) were optimised using grid search. A probability estimation function was used to obtain the probability output of the samples belonging to each category.
Artificial Neural Network Model
A feed-forward neural network with a single hidden layer was constructed. The number of hidden layer nodes was determined through cross-validation. A backpropagation algorithm was used for weight optimisation, a sigmoid function served as the activation function, and the learning rate was adaptively adjusted.
Model Evaluation and Comparison
Multiple indicators were used to comprehensively evaluate the model performance: Area Under the Receiver Operating Characteristic Curve (AUC), accuracy, sensitivity, and specificity. These indicators were used to assess each model’s ability to distinguish between anxious and non-anxious states.
Machine Learning Analysis Tools
All analyses were performed using R (version 4.1.0), mainly employing software packages such as randomForest, e1071, nnet, caret, and pROC. Statistical significance level was set at p = 0.05.
Results
Baseline Characteristics of 350 NICU Patients and Their Family Members
The mean age of the patients was 55.68 ± 14.94 years, with males comprising the majority (183, 52.3%). The most common diagnoses were craniocerebral injury (140 cases, 40.0%) and hypertensive intracerebral haemorrhage (105 cases, 30.0%). The median time from onset to hospital admission was 12.40 (6.78–17.60) h, and the median ICU length of stay was 10.40 (5.40–15.02) days. The most common type of medical insurance was employee insurance (171 cases, 48.9%). Most patients were the primary source of family income (216 patients, 61.7%).The mean age of the family members was 48.38 ± 18.11 years, with females being the majority (190 cases, 54.3%). The educational level was mainly junior high school (103 cases, 29.4%) and senior high school (104 cases, 29.7%). Most family members were either spouses (154 cases, 44.0%) or children (123 cases, 35.1%). The average monthly household income was between 3,000 and 5,000 RMB (184 cases, 52.6%), and nearly half of the family members were employed (172 cases, 49.1%). Approximately half of the family members (185 cases, 52.9%) had no previous experience accompanying a hospitalised patient. (Table 1)
|
Indicator
|
Result
|
|---|---|
|
Gender
|
55.68 ± 14.94
|
|
Age
|
|
|
Male(%)
|
183(52.3)
|
|
Female(%)
|
167(47.7)
|
|
Diagnosis
|
|
|
Traumatic brain injury(%)
|
140(40.0)
|
|
Hypertensive intracerebral hemorrhage (surgical treatment)(%)
|
105(30.0)
|
|
Intracranial aneurysm(%)
|
70(20.0)
|
|
Intracranial tumor(%)
|
35(10.0)
|
|
Time from onset to hospital admission (h)
|
12.40(6.78, 17.60)
|
|
NICU length of stay(d)
|
10.40(5.40, 15.02)
|
|
Medical insurance type
|
|
|
Self-payment (%)
|
37(10.6)
|
|
Resident medical insurance(%)
|
101(28.9)
|
|
Employee medical insurance(%)
|
171(48.9)
|
|
Third-party insurance(%)
|
41(11.7)
|
|
Whether the patient is the main family breadwinner
|
|
|
Yes(%)
|
216(61.7)
|
|
No(%)
|
134(38.3)
|
|
Family member's gender
|
|
|
Male(%)
|
160(45.7)
|
|
Female(%)
|
190(54.3)
|
|
Family member's age
|
48.38 ± 18.11
|
|
Family member's educational level
|
|
|
Primary school and below(%)
|
80(22.9)
|
|
Junior high school(%)
|
103(29.4)
|
|
Senior high school(%)
|
104(29.7)
|
|
College degree and above(%)
|
63(18.0)
|
|
Relationship between family member and patient
|
|
|
Patient's children(%)
|
123(35.1)
|
|
Patient's parent(s)(%)
|
41(11.7)
|
|
Patient's spouse(%)
|
154(44.0)
|
|
Patient's Sibling(s) (%)
|
32(9.1)
|
|
Marital status
|
|
|
Married (%)
|
232(66.3)
|
|
Unmarried(%)
|
46(13.1)
|
|
Divorced or widowed (%)
|
72(20.6)
|
|
Average monthly household income per person
|
|
|
<3000CNY(%)
|
103(29.4)
|
|
3000−5000CNY(%)
|
184(52.6)
|
|
>5000CNY(%)
|
63(18.0)
|
|
Occupation
|
|
|
Unemployed(%)
|
25(7.1)
|
|
Employed(%)
|
172(49.1)
|
|
Farmer/Worker(%)
|
47(13.4)
|
|
Self-employed(%)
|
36(10.3)
|
|
Retired(%)
|
70(20)
|
|
Whether having had accompanying experience
|
|
|
Yes(%)
|
165(47.1)
|
|
No(%)
|
185(52.9)
|
Assessment Based on the Hospital Anxiety and Depression Scale
Among the 350 family members, 125 (35.7%) had no anxiety symptoms (0–7 points), 89 (25.4%) had mild anxiety (8–10 points), 56 (16.0%) had moderate anxiety (11–14 points), and 80 (22.9%) had severe anxiety (15–21 points). The overall incidence of anxiety was 64.3% (225 cases). (Table 2)
|
Clinical Stratification of Anxiety
|
Result
|
|---|---|
|
No symptoms (0–7 points)
|
125(35.7)
|
|
Mild (8–10 points)
|
89(25.4)
|
|
Moderate (11–14 points)
|
56(16.0)
|
|
Severe (15–21 points)
|
80(22.9)
|
|
Anxiety Status (Presence/Absence)
|
|
|
Yes(%)
|
225(64.3)
|
|
No(%)
|
125(35.7)
|
Comparison of CCFNI Data Analysis Results using Random Forest, Support Vector Machine, and Neural Network in Machine Learning
Based on findings, The Random Forest model efficiently identifies their core needs.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 algorithm was selected as the analytical method for subsequent analyses. (Table 3)
|
Model
|
Accuracy
|
Sensitivity
|
Specificity
|
Precision
|
F1
|
AUC
|
|---|---|---|---|---|---|---|
|
Random Forest
|
0.9904
|
0.9851
|
1.0000
|
1.0000
|
0.9925
|
1.0000
|
|
Support Vector Machine
|
0.9712
|
0.9851
|
0.9459
|
0.9706
|
0.9778
|
0.9968
|
|
Artificial Neural Network
|
0.9231
|
0.9552
|
0.8649
|
0.9275
|
0.9412
|
0.9794
|
|
Item
|
High-Quality Nursing Group (N = 80)
|
Traditional Nursing Group (N = 80)
|
t/χ2/Z
|
P Value
|
|---|---|---|---|---|
|
Age (y)
|
54.14 ± 13.31
|
54.39 ± 14.49
|
0.114
|
0.910
|
|
Gender (%)
|
||||
|
Male
|
36(45.0)
|
48(60.0)
|
3.609
|
0.057
|
|
Female
|
44(55.0)
|
32(40.0)
|
||
|
Diagnosis
|
||||
|
Traumatic brain injury(%)
|
30(37.5)
|
34(42.5)
|
1.244
|
0.743
|
|
Hypertensive intracerebral hemorrhage (surgical treatment)(%)
|
27(33.8)
|
24(30.0)
|
||
|
Intracranial aneurysm(%)
|
15(18.8)
|
17(21.3)
|
||
|
Intracranial tumor(%)
|
8(10.0)
|
5(6.3)
|
||
|
Time from onset to hospital admission (h)
|
11.15(6.02, 18.22)
|
10.65(5.68, 17.18)
|
0.570
|
0.569
|
|
ICU length of stay (d)
|
11.65(6.50, 16.48)
|
12.25(6.48, 16.30)
|
0.032
|
0.974
|
|
Medical insurance type
|
||||
|
Self-payment (%)
|
6(7.5)
|
5(6.3)
|
0.513
|
0.916
|
|
Resident medical insurance(%)
|
25(31;3)
|
29(36.3)
|
||
|
Employee medical insurance(%)
|
42(52.5)
|
40(50.0)
|
||
|
Third-party insurance(%)
|
7(8.8)
|
6(7.5)
|
||
|
Whether the patient is the main family breadwinner
|
||||
|
Yes(%)
|
52(65.0)
|
46(57.5)
|
0.948
|
0.330
|
|
No(%)
|
28(35.0)
|
34(42.5)
|
||
|
Family member's gender
|
||||
|
Male(%)
|
44(55.0)
|
37(46.2)
|
1.225
|
0.268
|
|
Female(%)
|
36(45.0)
|
43(53.8)
|
||
|
Family member's age
|
46.48 ± 16.51
|
45.45 ± 19.00
|
0.364
|
0.717
|
|
Family member's educational level
|
||||
|
Primary school and below(%)
|
20(25.0)
|
18(22.5)
|
2.465
|
0.482
|
|
Junior high school(%)
|
18(22.5)
|
25(31.3)
|
||
|
Senior high school(%)
|
24(30.0)
|
25(31.3)
|
||
|
College degree and above(%)
|
18(22.5)
|
12(15.0)
|
||
|
Relationship between family member and patient
|
||||
|
Patient's children(%)
|
32(40.0)
|
37(46.3)
|
1.362
|
0.714
|
|
Patient's parent(s)(%)
|
5(6.3)
|
4(5.0)
|
||
|
Patient's spouse(%)
|
32(40.0)
|
32(40.0)
|
||
|
Patient's Sibling(s) (%)
|
11(13.8)
|
7(8.8)
|
||
|
Marital status
|
||||
|
Married (%)
|
57(71.3)
|
49(61.3)
|
2.365
|
0.306
|
|
Unmarried(%)
|
8(10.0)
|
14(17.5)
|
||
|
Divorced or widowed (%)
|
15(18.8)
|
17(21.3)
|
||
|
Average monthly household income per person
|
||||
|
<3000CNY(%)
|
21(26.3)
|
28(35.0)
|
3.039
|
0.219
|
|
3000−5000CNY(%)
|
48(60.0)
|
37(46.3)
|
||
|
>5000CNY(%)
|
11(13.8)
|
15(18.8)
|
||
|
Occupation
|
||||
|
Unemployed(%)
|
8(10.0)
|
11(13.8)
|
0.676
|
0.954
|
|
Employed(%)
|
42(52.5)
|
39(48.8)
|
||
|
Farmer/Worker(%)
|
9(11.3)
|
8(10.0)
|
||
|
Self-employed(%)
|
6(7.5)
|
6(7.5)
|
||
|
Retired(%)
|
15(18.8)
|
16(20.0)
|
||
|
Whether having had accompanying experience
|
||||
|
Yes(%)
|
33(41.3)
|
42(52.5)
|
2.033
|
0.154
|
|
No(%)
|
47(58.7)
|
38(47.5)
|
Random Forest Analysis Method and Results
Random Forest Analysis Method
The Random Forest algorithm was used for the feature importance analysis and classification prediction. The main parameters were as follows: 500 decision trees were constructed, 15 features were randomly selected for each tree split, and the Gini impurity was used as the splitting criterion. Model performance was internally validated using 10-fold cross-validation and evaluated using accuracy and AUC values. Variable importance was quantified using two methods: (1) Mean Decrease Accuracy, which was calculated by measuring the average reduction in model accuracy after random permutation of feature values, and (2) Mean Decrease Gini, which was based on the average reduction in Gini impurity contributed by each feature across all tree node splits.
Random Forest Graphical Results
Random Forest Variable Importance Ranking Based on Mean Decrease Accuracy
The results showed that CCFNI_4 had the highest predictive importance, indicating its strongly discriminative ability in distinguishing between anxious and non-anxious states. The subsequent variables include CCFNI_7, CCFNI_9, and CCFNI_1, each with decreasing importance scores, collectively forming a hierarchical structure of feature contributions within the classification model. (Fig. 3)
Fig. 3
Variables are arranged on the vertical axis in descending order of importance, and the horizontal axis represents the corresponding Mean Decrease Accuracy.
Variable Importance Ranking Based on Mean Decrease Gini
This indicator measures each features’s contribution to reducing Gini impurity across all decision tree node splits. The results showed that the variable importance ranking was consistent with the ranking based on the Mean Decrease Accuracy, with variables such as CCFNI_4, CCFNI_7, and CCFNI_9 remaining the top predictors, indicating the robustness of the findings. Gini importance focuses on measuring the ability of a feature to improve data purity during splitting, whereas accuracy importance measures its effect on prediction accuracy. The consistency of the two indicators further verifies the reliability of the key feature selection. (Fig. 4 )
Fig. 4
The figure quantifies the contribution of each feature to the reduction of Gini impurity across all decision tree node splits.
Confusion Matrix Visualisation of the Random Forest Model on the Test Set
The model correctly predicted 37 non-anxious and 67 anxious samples, with no misclassification of non-anxious samples as anxious (false positives) or anxious samples as non-anxious (false negatives). These findings confirm the effectiveness and practical applicability of the Random Forest algorithm for this dataset. (Fig. 5)
Fig. 5
The diagonal elements of the matrix represent the number of samples correctly classified by the model, while the off-diagonal elements represent the number of misclassified samples.
Variation Trend of Error Rate with the Increase in the Number of Decision Trees in the Random Forest Model
As the number of trees increased, all types of errors gradually decreased and tended to stabilise, indicating a good convergence of the model and stability of the error curve.(Fig. 6)
Fig. 6
The three curves represent the out-of-bag error, anxious class error, and non-anxious class error, respectively.
Probability Density Distribution of Samples in the Test Set Predicted as Anxious
The predicted probabilities of the non-anxious samples are concentrated in the low-value region, whereas those of anxious samples are distributed in the high-value region. Although some overlap existed between the two distributions, the results showed good discriminative ability. (Fig. 7 )
Fig. 7
The two curves correspond to the actual non-anxious and anxious sample groups, respectively. The morphological analysis of the predicted probability distribution helps assess the discriminative ability of the model and determine the optimal classification threshold.
Baseline Characteristics of NICU Patients and Their Family Members in the Intervention Group and Control Group (Table 4)
|
Item
|
High-Quality Nursing Group (N = 80)
|
Traditional Nursing Group (N = 80)
|
t/χ2/Z
|
P Value
|
|---|---|---|---|---|
|
Age (y)
|
54.14 ± 13.31
|
54.39 ± 14.49
|
0.114
|
0.910
|
|
Gender (%)
|
||||
|
Male
|
36(45.0)
|
48(60.0)
|
3.609
|
0.057
|
|
Female
|
44(55.0)
|
32(40.0)
|
||
|
Diagnosis
|
||||
|
Traumatic brain injury(%)
|
30(37.5)
|
34(42.5)
|
1.244
|
0.743
|
|
Hypertensive intracerebral hemorrhage (surgical treatment)(%)
|
27(33.8)
|
24(30.0)
|
||
|
Intracranial aneurysm(%)
|
15(18.8)
|
17(21.3)
|
||
|
Intracranial tumor(%)
|
8(10.0)
|
5(6.3)
|
||
|
Time from onset to hospital admission (h)
|
11.15(6.02, 18.22)
|
10.65(5.68, 17.18)
|
0.570
|
0.569
|
|
ICU length of stay (d)
|
11.65(6.50, 16.48)
|
12.25(6.48, 16.30)
|
0.032
|
0.974
|
|
Medical insurance type
|
||||
|
Self-payment (%)
|
6(7.5)
|
5(6.3)
|
0.513
|
0.916
|
|
Resident medical insurance(%)
|
25(31;3)
|
29(36.3)
|
||
|
Employee medical insurance(%)
|
42(52.5)
|
40(50.0)
|
||
|
Third-party insurance(%)
|
7(8.8)
|
6(7.5)
|
||
|
Whether the patient is the main family breadwinner
|
||||
|
Yes(%)
|
52(65.0)
|
46(57.5)
|
0.948
|
0.330
|
|
No(%)
|
28(35.0)
|
34(42.5)
|
||
|
Family member's gender
|
||||
|
Male(%)
|
44(55.0)
|
37(46.2)
|
1.225
|
0.268
|
|
Female(%)
|
36(45.0)
|
43(53.8)
|
||
|
Family member's age
|
46.48 ± 16.51
|
45.45 ± 19.00
|
0.364
|
0.717
|
|
Family member's educational level
|
||||
|
Primary school and below(%)
|
20(25.0)
|
18(22.5)
|
2.465
|
0.482
|
|
Junior high school(%)
|
18(22.5)
|
25(31.3)
|
||
|
Senior high school(%)
|
24(30.0)
|
25(31.3)
|
||
|
College degree and above(%)
|
18(22.5)
|
12(15.0)
|
||
|
Relationship between family member and patient
|
||||
|
Patient's children(%)
|
32(40.0)
|
37(46.3)
|
1.362
|
0.714
|
|
Patient's parent(s)(%)
|
5(6.3)
|
4(5.0)
|
||
|
Patient's spouse(%)
|
32(40.0)
|
32(40.0)
|
||
|
Patient's Sibling(s) (%)
|
11(13.8)
|
7(8.8)
|
||
|
Marital status
|
||||
|
Married (%)
|
57(71.3)
|
49(61.3)
|
2.365
|
0.306
|
|
Unmarried(%)
|
8(10.0)
|
14(17.5)
|
||
|
Divorced or widowed (%)
|
15(18.8)
|
17(21.3)
|
||
|
Average monthly household income per person
|
||||
|
<3000CNY(%)
|
21(26.3)
|
28(35.0)
|
3.039
|
0.219
|
|
3000−5000CNY(%)
|
48(60.0)
|
37(46.3)
|
||
|
>5000CNY(%)
|
11(13.8)
|
15(18.8)
|
||
|
Occupation
|
||||
|
Unemployed(%)
|
8(10.0)
|
11(13.8)
|
0.676
|
0.954
|
|
Employed(%)
|
42(52.5)
|
39(48.8)
|
||
|
Farmer/Worker(%)
|
9(11.3)
|
8(10.0)
|
||
|
Self-employed(%)
|
6(7.5)
|
6(7.5)
|
||
|
Retired(%)
|
15(18.8)
|
16(20.0)
|
||
|
Whether having had accompanying experience
|
||||
|
Yes(%)
|
33(41.3)
|
42(52.5)
|
2.033
|
0.154
|
|
No(%)
|
47(58.7)
|
38(47.5)
|
|
Model
|
Accuracy
|
Sensitivity
|
Specificity
|
Precision
|
F1
|
AUC
|
|---|---|---|---|---|---|---|
|
Random Forest
|
0.9904
|
0.9851
|
1.0000
|
1.0000
|
0.9925
|
1.0000
|
|
Support Vector Machine
|
0.9712
|
0.9851
|
0.9459
|
0.9706
|
0.9778
|
0.9968
|
|
Artificial Neural Network
|
0.9231
|
0.9552
|
0.8649
|
0.9275
|
0.9412
|
0.9794
|
Comparison of HADS Anxiety Scores Between the Intervention Group and Control Group Before and After Intervention
Before the intervention, no significant difference was observed in anxiety scores between the two groups (intervention group: 12.95 ± 2.83 vs. control group: 12.93 ± 2.87, t = 0.057, P = 0.954). After the intervention, the anxiety score of the intervention group decreased significantly to 8.39 ± 2.88, while that of the control group showed no significant change (12.78 ± 3.08). The difference between the two groups was statistically significant (t = 9.312, P < 0.001). Within-group comparisons showed a significant reduction in the intervention group’s anxiety score before and after the intervention (t = 18.234, P < 0.001), while the change in the control group was not statistically significant (P = 0.653). (Table 5)
|
Item
|
Observation group(n = 80)
|
Control group(n = 80)
|
t Value
|
P Value
|
|---|---|---|---|---|
|
Before intervention
|
12.95 ± 2.83
|
12.93 ± 2.87
|
0.057
|
0.954
|
|
After intervention
|
8.39 ± 2.88
|
12.78 ± 3.08
|
9.312
|
< 0.001
|
|
t Value
|
18.234
|
0.452
|
||
|
P Value
|
< 0.001
|
0.653
|
|
Item
|
Observation group(n = 80)
|
Control group(n = 80)
|
t Value
|
P Value
|
|---|---|---|---|---|
|
Before intervention
|
12.95 ± 2.83
|
12.93 ± 2.87
|
0.057
|
0.954
|
|
After intervention
|
8.39 ± 2.88
|
12.78 ± 3.08
|
9.312
|
< 0.001
|
|
t Value
|
18.234
|
0.452
|
||
|
P Value
|
< 0.001
|
0.653
|
|
Item
|
Observation group(n = 80)
|
Control group(n = 80)
|
t Value
|
P Value
|
|---|---|---|---|---|
|
Total score
|
112.60 ± 2.12
|
81.60 ± 2.40
|
67.737
|
< 0.001
|
|
Care satisfaction score
|
65.62 ± 1.70
|
47.54 ± 1.68
|
55.295
|
< 0.001
|
|
Decision-making satisfaction score
|
46.98 ± 1.39
|
34.06 ± 1.56
|
86.571
|
< 0.001
|
|
Participation in treatment decision-making
|
4.66 ± 0.48
|
3.43 ± 0.50
|
16.079
|
< 0.001
|
|
Understanding of medical staff's suggestions
|
9.31 ± 0.59
|
6.84 ± 0.70
|
24.206
|
< 0.001
|
Family Satisfaction in the Intensive Care Unit-24 (FS-ICU-24) Questionnaire
The FS-ICU-24 Questionnaire consists of 24 items, divided into two core dimensions: "Care Satisfaction" (Items 1–14) and "Decision-Making Satisfaction" (Items 15–24). Among these, items such as "Participation in Treatment Decisions" (Item 15) and "Understanding of Medical and Nursing Recommendations" (Items 1 and 18) indirectly reflect family members’ compliance. This table compares the total scores for each dimension of the FS-ICU-24 Questionnaire between the intervention and control groups. 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). Among all dimensions, the intervention group scored significantly higher than the control group in care satisfaction (65.62 ± 1.70 vs. 47.54 ± 1.68), decision-making satisfaction (46.98 ± 1.39 vs. 34.06 ± 1.56), participation in treatment decisions (4.66 ± 0.48 vs. 3.43 ± 0.50), and understanding of medical and nursing recommendations (9.31 ± 0.59 vs. 6.84 ± 0.70) (all P < 0.001).(Table 6)
|
Item
|
Observation group(n = 80)
|
Control group(n = 80)
|
t Value
|
P Value
|
|---|---|---|---|---|
|
Total score
|
112.60 ± 2.12
|
81.60 ± 2.40
|
67.737
|
< 0.001
|
|
Care satisfaction score
|
65.62 ± 1.70
|
47.54 ± 1.68
|
55.295
|
< 0.001
|
|
Decision-making satisfaction score
|
46.98 ± 1.39
|
34.06 ± 1.56
|
86.571
|
< 0.001
|
|
Participation in treatment decision-making
|
4.66 ± 0.48
|
3.43 ± 0.50
|
16.079
|
< 0.001
|
|
Understanding of medical staff's suggestions
|
9.31 ± 0.59
|
6.84 ± 0.70
|
24.206
|
< 0.001
|
Discussion
The NICU is the core unit for the intensive treatment of patients with acute and critical neurosurgical conditions. Owing to the high mortality and disability rates, as well as the dynamic progression of disease among admitted patients, healthcare providers primarily focus on ensuring patient safety. This focus, however, often leads to the unintentional neglect of family members’ emotional needs[19]. Family members often experience intense anxiety due to the critical condition of their loved ones and uncertainty surrounding their prognosis. In this study, a survey of 350 family members of NICU patients with a hospital stay of ≥ 1 week using the HADS revealed that the overall incidence of anxiety was 64.3% (225 out of 350 cases). This finding further confirms the high prevalence of anxiety in this group and highlights the urgent need for clinical intervention, which is consistent with the conclusions of previous studies[8].
Precision nursing, a novel nursing model centred on "individual needs", is founded on the core principles of developing individualised intervention plans based on systematic needs assessment. This model provides a scientific framework for addressing the long-neglected issue of anxiety among NICU family members. To overcome the limitation that homogeneous interventions fail to account for individual differences among family members, this study first assessed their needs using the internationally recognised CCFNI. This scale has been validated for use among ICU family members during the pandemic, owing to its accuracy in identifying potential needs and its strong cross-cultural applicability, thereby providing a scientific foundation for the needs assessment in this study[20–21].
To overcome the limitations of traditional statistical methods in exploring in-depth correlations of needs, this study introduced machine learning models and compared the accuracy of three commonly used methods: Random Forest, Support Vector Machine, and Neural Network. The results revealed that the Random Forest model had optimal performance and most accurately reflected the needs of family members. Analysis of CCFNI data using this model the four most important core needs of family members in the following order: "I need to be with the patient as much as possible (CCFNI_4)", "I need a clear explanation of the patient’s condition (CCFNI_7)", "I need to understand the purpose of each treatment (CCFNI_9)", and "I need to know that my loved one is receiving the best possible care (CCFNI_1)". Among these, CCFNI_7 and CCFNI_9 both essentially reflect the need for "acquisition of disease and treatment information". This ranking provides a clear basis for designing subsequent intervention plans.
Based on the above needs analysis, an intervention plan was developed around these core needs and aligned with the multiple causes of family members’ anxiety. The studies noted that the severity of a patient’s condition and the frequency of medical information acquisition are independent risk factors for anxiety among ICU family members[11].Previous research further observed that the psychological characteristic of family members influence their susceptibility to anxiety, and that the high mortality and disability rates associated with neurosurgical diseases, together with the closed environment of the ICU, may further amplify negative emotions, creating an anxiety chain of "worry about condition – lack of information – environmental isolation"[22]. Therefore, the multi-dimensional intervention plan designed in this study, which is consistent with the studies conclusion that "multi-dimensional intervention is superior to single intervention", aims to break this anxiety cycle at its root[23].
To address the "need for companionship", video visits were introduced to supplement traditional in-person visits. This approach not only reduces the risk of hospital-acquired infections but also alleviates "isolation anxiety" by increasing visiting opportunities. In addition, the duration of visits during non-peak hours was extended, which is consistent with the practical benefit observed from ICU video visits during the COVID-19 pandemic. To meet the "need for disease explanation"[24], virtual reality technology was incorporated into traditional health education to intuitively illustrate the anatomical relationship between intracranial lesions and symptom-related areas[25–26]. This method overcomes the limitations of traditional two-dimensional materials and verbal explanations, reducing information discrepancies. Virtual reality technology has been widely used in neurosurgical diagnoses, surgical planning, and rehabilitation training. To satisfy the "need for confirmation of care quality", a "nurse diary" was used for daily communication with family members, allowing them to intuitively understand details such as position care and complication prevention. This finding is consistent with the conclusion of previous studies indicating that "nurse diaries can enhance family trust and reduce anxiety"[27–28].
The internationally recognised Family Satisfaction in the Intensive Care Unit-24 (FS-ICU-24) questionnaire was used to evaluate the effectiveness of the intervention. The findings indicated that the anxiety scores of the intervention group changed significantly before and after the intervention (t = 18.234, P < 0.001), while no statistically significant difference was observed in the control group (P = 0.653).
A
Family members in the intervention group reported significantly higher satisfaction than those in the control group, with the most significant improvement observed in the dimensions of "participation in treatment decisions" and "understanding of medical and nursing recommendations". These findings not only directly confirm the effectiveness of precision nursing in improving family satisfaction but also indirectly verify its effect in alleviating anxiety, as evidenced by the observed "negative correlation between satisfaction and anxiety". This outcome is consistent with the design principle of "multi-dimensional intervention matching multiple causal factors". For family members, the intervention effectively reduced anxiety levels and helped prevent the progression of short-term anxiety into long-term mental illness. For patients, the emotional stability of family members can reduce the transmission of negative emotions and create a positive environment for recovery. For medical institutions, the intervention not only reduces the risk of potential medical disputes and improves service reputation but also forms a standardised process of "needs assessment - precision intervention - effect verification", providing replicable experience for family care in ICUs of other specialities.Conclusions and Directions for Future Research
This sudy focused on the anxiety experienced by family members of patients in the NICU and established a comprehensive research framework of "problem identification, needs exploration, intervention implementation, and effect verification". Precision nursing serves as the core logical thread throughout this framework. The current status of family members’ anxiety and core needs was identified using the HADS and CCFNI scales. Machine learning models were used, and Random Forest was verified to be the most accurate method for needs analysis. Based on these findings, a targeted multi-dimensional intervention plan was designed, and its effectiveness was subsequently verified using the FS-ICU-24 questionnaire. The findings confirmed that need-oriented precision nursing can effectively alleviate the anxiety of family members of patients in NICU and improve their satisfaction. This provides robust evidence supporting the high-quality development of family care in the NICU and expanding new ideas for the application of precision nursing in the field of acute and critical illnesses.
Abbreviations
NICU Neurosurgical intensive care unit
HADS Hospital anxiety and depression scale
CCFNI Critical care family needs inventory
FS-ICU-24 Family satisfaction in the Intensive Care Unit-24
sTBI severe traumatic brain injury
GCS Glasgow coma scale
CT Computed tomography
MRI Magnetic resonance imaging
Acknowledgements
Not applicable.
A
Author Contribution
Feng Zhang were involved in data curation, formal analysis, and methodology, and also authored the initial draft. Ruixiang Sun and Jing Huang and Zhiqing Zhou contributed to data curation as well, with additional responsibilities in conceptualization, project administration, and drafting the manuscript. Ting Yang and Yanling Li and Lihong Min concentrated on conceptualization and project administration, while also participating in the manuscript’s writing, review, and editing processes. Zuan Yu and Jiaqiang Liu and Huayue Zhang Lei He and Ping Xu encompassed conceptualization and methodology, along with contributing to the writing, review, and editing of the manuscript. All authors critically reviewed and refined the manuscript.
A
Funding
This research was supported by Yijishan hospital level service management innovation project (cx2024008); School level key humanities and Social Sciences project of Wannan Medical College (wk2024szd01); The First Affiliated Hospital of Soochow University Zhou's”Nursing Research Project (HLYJ-Z-202517).
A
Data Availability
No datasets were generated or analysed during the current study.
Declarations
Ethics approval and consent to participate
All methods were carried out in accordance with relevant guidelines and regulations. The study was carried out respecting the Declaration of Helsinki in its current version.
A
A
The study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Wannan Medical College (No.2018043), and written informed consent was obtained from all the participants.Consent for publication
All patients included in this study had signed informed consent.
Competing interests
The authors declare no competing interests.
Author details
1 The First Affiliated Hospital of Wannan Medical Colleg, Zheshan West Road on the 2nd, Wuhu 241000, Anhui,China
2 The first people's Hospital of Chuzhou City, No. 369 zuiwang West Road, Nanqiao District,Chuzhou 239001, Anhui,China
3 The First Affiliated Hospital of Soochow University,No. 899, Pinghai Road,Suzhou 215000,Jiangsu,China
References
1.
Andreassen JS, Thorsen K, Søreide K, Werner D, Weber C. Is there a weekend effect on mortality rate and outcome for moderate and severe traumatic brain injury? A population-based, observational cohort study. Brain Spine.2022.1;2:101699.
2.
Sakr Y, Dünisch P, Santos C, Matthes L, Zeidan M, Reinhart K, et al. Poor outcome is associated with less negative fluid balance in patients with aneurysmal subarachnoid hemorrhage treated with prophylactic vasopressor-induced hypertension. Ann Intensive Care. 2016;6(1):25.
3.
Linjun Y, Xi L, Yanmei Y, Wulajing L, Linjie F, Jingping K et al. Perioperative real-time information sharing and its impact on family members' anxiety in patients undergoing elective thoracoscopic lobectomy: a single-center randomized controlled trial. Perioper Med (Lond). 2025.11;14(1):19.
4.
Gurbuz H, Demir N. Anxiety and Depression Symptoms of Family Members of Intensive Care Unit Patients: A Prospective Observational Study and the Lived Experiences of the Family Members. Avicenna J Med. 2023;16(2):89–96.
5.
Białek K, Sadowski M. Stress, anxiety, depression and basic hope in family members of patients hospitalised in intensive care units - preliminary report. Anaesthesiol Intensive Ther. 2021;53(2):134–40.
6.
Blok AC, Valley TS, Weston LE, Miller J, Lipman K, Krein SL. Factors Affecting Psychological Distress in Family Caregivers of Critically Ill Patients: A Qualitative Study. Am J Crit Care. 2023;32(1):21–30.
A
7.Seren H, Altun Uğraş G. Çam Yanik T.Determining the needs of relatives of patients with a COVID-19 diagnosis in the intensive care unit. Aust Crit Care. 2025;38(1):101085.
8.
Krishnamoorthy D, Senthilkumar, Venkataraman R, Ramakrishnan N, Vignesh C. Decision-making Preferences and Levels of Anxiety and Depression in Family Members of Patients Admitted to the ICU. Indian J Crit Care Med. 2025;29(2):155–63.
9.
Miller J, Garcia R, Williams S. Application of Neural Network in Analyzing ICU Family Needs. IEEE J Biomedical Health Inf. 2022;26(8):3890–8.
10.
Ma J, Ma DW. Advancements in the application of precision nursing model on hemodialysis for diabetic nephropathy: A review. Med (Baltimore) 2024 20;103(51):e40952.
11.
Lobato CT, Camões J, Carvalho D, Vales C, Dias CC, Gomes E, et al. Risk factors associated with post-intensive care syndrome in family members (PICS-F): A prospective observational study. J Intensive Care Soc. 2023;24(3):247–57.
12.
Rosa RG, Pellegrini JAS, Moraes RB, Prieb RGG, Sganzerla D, Schneider D, et al. Mechanism of a Flexible ICU Visiting Policy for Anxiety Symptoms Among Family Members in Brazil: A Path Mediation Analysis in a Cluster-Randomized Clinical Trial. Crit Care Med. 2021;49(1):1504–12.
13.
Forsberg T, Isaksson M, Schelin C, Lyngå P, Schandl A. Family members' experiences of COVID-19 visiting restrictions in the intensive care unit-A qualitative study. J Clin Nurs. 2024;33(1):215–23.
14.
Zerwic L, Mohan A, Riley E, Byeman C, Ashwath R. The impact of 3D printed vs. 3D virtual congenital heart models on patient and family knowledge. Front Pediatr. 2025;14:13:1525549.
15.
Bosco V, Mercuri C, Giordano V, Froio AM, Commisso D, Nocerino R, et al. Enhancing ICU care with nurse-written diaries. Nurs Crit Care. 2024;29(6):1355–62.
16.
Pochard F, Azoulay E, Chevret S, Adrie C. Development and psychometric evaluation of the Family Satisfaction in the ICU (FS-ICU) instrument. Crit Care Med. 2006;34(4):960–7.
17.
Hatta H, Higashi A, Yashiro H, Ozawa K, Hayashi K, Kiyota K. A Valida-tion of the hospital anxiety and depression scale. Japanese J Psychosom Med. 1998;38:309–15.
18.
Gómez-Martíinez S, Arnal RB, Juliá BG. El Cuestionario de Necesidades de los Familiares de Pacientes de Cuidados Intensivos (CCFNI) versión breve: adaptación y validación en población españ The short version of Critical Care Family Needs Inventory (CCFNI): adaptation and validation for a spanish sample. Sist Sanit Navar. 2011;34(3):349–61. Spanish.
19.
Ramsburg H, Fischer AG, MacKenzie Greenle M, Fehnel CR. Care of the Patient Nearing the End of Life in the Neurointensive Care Unit. Neurocrit Care. 2024;41(3):749–59.
20.
Ma J, Ma DW. Advancements in the application of precision nursing model on hemodialysis for diabetic nephropathy: A review. Med (Baltim). 2024;20(51):e40952.
21.
Hinton L, Aggarwal N, Iosif AM, Weiss M, Paralikar V, Deshpande S et al. Perspectives of family members participating in cultural assessment of psychiatric disorders: findings from the DSM-5 International Field Trial. Int Rev Psychiatry 201527(1):3–10.
22.
Friligou A, Sidiropoulou M, Kavga A, Gerogianni G, Kourakos M, Vasilopoulos G, et al. Factors Affecting State and Trait Anxiety of Relatives of Hospitalized Patients. J Caring Sci. 2021;10(1):9–14. 10.34172/jcs.2021.009.
23.
Davidson N. The outcome of a multidimensional intervention strategy for the management of generalized anxiety disorder in an internal medicine setting. Evid - based Pract Project Rep, 2022.183.
24.
Forsberg T, Isaksson M, Schelin C, Lyngå P, Schandl A. Family members' experiences of COVID-19 visiting restrictions in the intensive care unit-A qualitative study. J Clin Nurs. 2024;33(1):215–23.
25.
Zawy Alsofy S, Sakellaropoulou I, Nakamura M, Ewelt C, Salma A, et al. Impact of Virtual Reality in Arterial Anatomy Detection and Surgical Planning in Patients with Unruptured Anterior Communicating Artery Aneurysms. Brain Sci. 2020;10(12):963.
26.
Mishra R, Lee K, Brown T. Virtual reality in neurosurgery: Beyond neurosurgical planning. Int J Environ Res Public Health. 2022;19(3):1719.
27.
Hackenberger A. Intensive Care Unit Diaries: A Nurse-Led Program. Crit Care Nurse. 2023;43(1):20–30.
28.
Högvall LM, Herling SF, Egerod I, Petosic A, Danielsen MER, Rüdiger U, et al. The patient experience of a nurse-written ICU-diary intervention: A cross sectional survey. Intensive Crit Care Nurs. 2025;86:103846.
Table 1 Analysis of General Data of 350 Patients in the Neonatal Intensive Care Unit (NICU) and Their Family Members
|
Indicator
|
Result
|
|---|---|
|
Gender
|
55.68 ± 14.94
|
|
Age
|
|
|
Male(%)
|
183(52.3)
|
|
Female(%)
|
167(47.7)
|
|
Diagnosis
|
|
|
Traumatic brain injury(%)
|
140(40.0)
|
|
Hypertensive intracerebral hemorrhage (surgical treatment)(%)
|
105(30.0)
|
|
Intracranial aneurysm(%)
|
70(20.0)
|
|
Intracranial tumor(%)
|
35(10.0)
|
|
Time from onset to hospital admission (h)
|
12.40(6.78, 17.60)
|
|
NICU length of stay(d)
|
10.40(5.40, 15.02)
|
|
Medical insurance type
|
|
|
Self-payment (%)
|
37(10.6)
|
|
Resident medical insurance(%)
|
101(28.9)
|
|
Employee medical insurance(%)
|
171(48.9)
|
|
Third-party insurance(%)
|
41(11.7)
|
|
Whether the patient is the main family breadwinner
|
|
|
Yes(%)
|
216(61.7)
|
|
No(%)
|
134(38.3)
|
|
Family member's gender
|
|
|
Male(%)
|
160(45.7)
|
|
Female(%)
|
190(54.3)
|
|
Family member's age
|
48.38 ± 18.11
|
|
Family member's educational level
|
|
|
Primary school and below(%)
|
80(22.9)
|
|
Junior high school(%)
|
103(29.4)
|
|
Senior high school(%)
|
104(29.7)
|
|
College degree and above(%)
|
63(18.0)
|
|
Relationship between family member and patient
|
|
|
Patient's children(%)
|
123(35.1)
|
|
Patient's parent(s)(%)
|
41(11.7)
|
|
Patient's spouse(%)
|
154(44.0)
|
|
Patient's Sibling(s) (%)
|
32(9.1)
|
|
Marital status
|
|
|
Married (%)
|
232(66.3)
|
|
Unmarried(%)
|
46(13.1)
|
|
Divorced or widowed (%)
|
72(20.6)
|
|
Average monthly household income per person
|
|
|
<3000CNY(%)
|
103(29.4)
|
|
3000−5000CNY(%)
|
184(52.6)
|
|
>5000CNY(%)
|
63(18.0)
|
|
Occupation
|
|
|
Unemployed(%)
|
25(7.1)
|
|
Employed(%)
|
172(49.1)
|
|
Farmer/Worker(%)
|
47(13.4)
|
|
Self-employed(%)
|
36(10.3)
|
|
Retired(%)
|
70(20)
|
|
Whether having had accompanying experience
|
|
|
Yes(%)
|
165(47.1)
|
|
No(%)
|
185(52.9)
|
Table 2 Anxiety Status of Family Members of 350 NICU Patients
|
Clinical Stratification of Anxiety
|
Result
|
|---|---|
|
No symptoms (0–7 points)
|
125(35.7)
|
|
Mild (8–10 points)
|
89(25.4)
|
|
Moderate (11–14 points)
|
56(16.0)
|
|
Severe (15–21 points)
|
80(22.9)
|
|
Anxiety Status (Presence/Absence)
|
|
|
Yes(%)
|
225(64.3)
|
|
No(%)
|
125(35.7)
|