The impact of an artificial intelligence (AI)-assisted education program on mental health, social support and quality of life in older individuals with head and neck cancer: study protocol of a randomized controlled trial

Abstract

Background

Head and neck cancer (HNC) is a common malignant tumor, and its treatment often leads to functional impairments in speech, swallowing, and appearance, severely affecting patients’ quality of life. Older individuals with HNC, due to the combined stress of aging and disease, face heightened mental health challenges. This study aims to evaluate the effect of AI-driven personalized health education on mental health, social support and quality of life in older patients after HNC surgery.

Methods

A single-center, two-group, randomized controlled trial will be conducted. One hundred of postoperative HNC patients aged ≥ 60 years will be randomly assigned to the intervention group (n = 50) or the control group (n = 50). The intervention group will receive 12 months of personalized and phased health education through the “Kangkang” AI assistant, including video/graphic and real-time AI Q&A. The control group received standardized SMS health education at the same frequency. The outcomes will include Perceived Stress Scale, Barthel Index, Patient Assessment of Constipation Quality of Life, Nutritional Risk Screening 2002, Numerical Rating Scale for Pain, Rosenberg Self-Esteem Scale, Pittsburgh Sleep Quality Index, UCLA Loneliness Scale (3rd edition), Fear of Cancer Progression Questionnaire-Short Form, World Health Organization Quality of Life Assessment for Older Adults, Generalized Anxiety Disorder-7, Patient Health Questionnaire-9, Morse Fall Scale. All assessments will be performed 5 times at baseline (preoperative) and 1, 3, 6, and 12 months postoperatively. Statistical analysis will be conducted using intention-to-treat analysis. Linear mixed models with maximum likelihood estimation will be used to analyze the continues variables and manage missing data.

Discussion

This will provide evidence on whether AI-driven, personalized health education can improve mental health, social support and quality of life in older patients after HNC surgery. The single-center design and reliance on self-reported outcomes may limit generalizability, and future multicenter studies are warranted.

Trial registration

The trial has been prospectively registered in the China Clinical Trials Registry on December 5, 2025, with registration number of ChiCTR2500114052.

Background

Head and Neck Cancer (HNC) is one of the most common malignant tumors worldwide, with more than 660,000 new cases and more than 325,000 deaths each year, posing a significant public health burden [1]. In China, the burden of HNC is particularly high, with approximately 148,000 new cases and 79,000 deaths reported in 2022, positioning it among the top high-risk cancers nationally [2]. The impact of HNC extends beyond physical symptoms, significantly affecting psychological and social well-being. Systematic reviews indicate that over 25% of HNC patients experience clinically significant depression and anxiety during diagnosis and treatment, with peak distress occurring post-treatment [3]. Furthermore, issues related to body image distress, social isolation, and financial toxicity are prevalent and closely linked to lower quality of life [4, 5].

China is experiencing the world’s fastest population ageing, with an increasingly large group of older cancer patients and increasingly prominent health problems [6]. Older people with HNC are especially vulnerable due to age-related declines in physiological reserve, higher rates of comorbidities, and reduced tolerance to aggressive treatments. Studies specific to the Chinese older HNC population reveal that frailty, malnutrition, and psychosocial factors such as stigma and diminished social support independently contribute to impaired quality of life, irrespective of tumor stage or treatment modality [7, 8].

Despite the clear multidimensional burden, current post-treatment support and health education models often fail to meet the complex, continuous, and personalized needs of older HNC patients. These conventional approaches such as traditional discharge education and regular telephone follow-up, have limitations including homogeneous content, discontinuous information transmission, and unfriendly forms for older people (e.g., text-intensive, lack of interaction) [2, 9]. This makes it difficult for patients, especially older patients, to fully understand, remember, and apply this health information, and the adherence and ultimate benefits of the intervention are greatly reduced. Digital health interventions, including mobile applications, have emerged as potential solutions. However, the components are highly heterogeneous and lack structured and generalizable evidence-based intervention models that target the specific population of older HNC patients with the core goal of improving their multidimensional quality of life [10, 11].

Recently, the rapid development of artificial intelligence (AI) technology offers a revolutionary tool to address these shortcomings. AI-assisted education systems can provide highly personalized, systematic, and 24-hour responses to health education and psychosocial support based on preset rules, patient personal characteristics, and disease stages. This study innovatively introduces the “Kangkang” AI intelligent assistant, aiming to build a continuous support platform integrating “regular content engine”, “phased disease management”, “age-friendly interaction design”, and “health behavior monitoring and feedback” for older HNC patients. This study hopes to evaluate the effectiveness and feasibility of the AI-assisted education program in improving the mental health and overall quality of life of older HNC patients through a rigorous randomized controlled trial, to fill the gap in academic and clinical practice in this field, and promote the clinical translation and application of AI in the field of cancer rehabilitation and psychosocial oncology.

Based on the above research background and actual needs, this study will conduct a single-center randomized controlled trial to evaluate the feasibility and effectiveness of an AI-assisted education intervention program for improving metnal health and quality of life among older people with HNC. We hypothesize that patients in the intervention group who receive personalized support from the “Kangkang” AI assistant for 12 months will show more significant improvements in mental health and quality of life compared to the control group receiving standardized text message education.

Methods

Study design

This study will be a single-center, assessor-blinded, parallel-group randomized controlled trial. The intervention will be a 12-month personalized, AI-assisted education program (“Kangkang” AI assistant) delivering twice-weekly multimedia content and daily real-time interactive support. The primary outcome, the Chinese version of the Perceived Stress Scale (PSS), focusing on the stress, which is one risk factor for mental health problems. Secondary outcomes encompassed Barthel Index (BI), Short version of the Fear of Disorder Progression Questionnaire (Fop-Q-SF), Generalized Anxiety Disorder Scale-7 items (GAD-7), Morse Fall Scale (MFS), Numeric Pain Rating Scale (NRS), Nutritional Risk Screening 2002 (NRS2002), Quality of Life Scale for Patients with Constipation (PAC-QOL), Patient Health Questionnaire-9 items (PHQ-9), Pittsburgh Sleep Quality Index (PSQI), Rosenberg Self-Esteem Scale (RSES), World Health Organization Quality of Life Scale for Older Adults (WHOQOL-OLD). This study has been registered in the China Clinical Trials Registry on December 5, 2025 (#ChiCTR2500114052). This protocol was written in accordance with the SPIRIT statement with details shown in Table 1.

Table 1.

Timetable of activities planned during the study

Abbreviation: BI Barthel Index, Fop-Q-SF Short version of the Fear of Disorder Progression Questionnaire, GAD-7 Generalized Anxiety Disorder Scale-7 items, MFS Morse Fall Scale, MTh month, NRS Numeric Pain Rating Scale, NRS2002 Nutritional Risk Screening 2002, PAC-QOL Quality of Life Scale for Patients with Constipation, PHQ-9 Patient Health Questionnaire-9 items, PSQI Pittsburgh Sleep Quality Index, PSS Perceived Stress Scale, RSES Rosenberg Self-Esteem Scale, WHOQOL-OLD World Health Organization Quality of Life Scale for Older Adults

Participant recruitment and eligibility criteria

Recruitment location will be in the Department of Head and Neck Surgery, Shenzhen Hospital, Cancer Hospital, Chinese Academy of Medical Sciences. Inclusion criteria are (i) Age ≥ 60 years; (ii) Pathologically diagnosed with primary HNC (including oral cavity, oropharynx, larynx, hypopharynx, etc.), and planned to undergo surgical treatment; (iii) Presence of mild or above psychological distress before operation, defined as a score of PHQ-9 ≥ 5 and/or a score of GAD-7 ≥ 5, to maximize clinical relevance and avoid floor effects; (iv) Have basic smartphone operation skills or have a caregiver available to assist in using it; (v) Voluntarily participate and sign the informed consent form. Exclusion criteria are (i) Presence of severe cognitive impairment with a score of Mini-Cog ≦ 2 points; (ii) Combined with other end-stage diseases (e.g., advanced heart failure, renal failure) or life expectancy < 6 months; (iii) Suffering from severe mental illness (e.g., schizophrenia, acute phase of bipolar disorder); (iv) Inability to communicate effectively in Chinese.

Sample size calculation was based on prior pre-experimental data and relevant literature [12], assuming a moderate intervention effect on the primary outcome (difference in PSS scores between groups; Cohen’s d = 0.64). Using G^*Power 3.1.9.2 with a two-sided α of 0.05 and 80% power (1-β), an independent-samples t-test indicated that at least 40 participants per group were required. Allowing for an anticipated 20% attrition rate (e.g., loss to follow-up, withdrawal, or death), the total planned sample size was 100 participants, with 50 in the intervention group and 50 in the control group. The Department of Head and Neck Surgery has a sufficiently high patient volume to reach the target sample size of 100 participants within the planned recruitment period.

Randomization and blinding

In this single-center trial, eligible postoperative HNC patients will be recruited by trained research staff and randomly allocated in a 1:1 ratio to the intervention or control group using a computer-generated sequence prepared and implemented by an independent coordinator, with allocation concealment ensured via sequentially numbered opaque sealed envelopes. The intervention will be delivered by designated nurses and the “Kangkang” platform team (FZ, HHZ, XMY), who are blind to the assessments. All outcome assessments will be conducted by trained assessors (YHP) who will remain blinded to group allocation and intervention. Both participants and intervention delivers will be unblind to the allocation. To minimized the potential biases non-blinding, we will standardize the assessments, separate the assessors from intervention dilivers. Unblinding of the blinded assessors is permitted only in a medical emergency when knowledge of group allocation is essential for participant safety. The procedure is as follows: the attending physician requests the independent coordinator to open the sealed envelope and reveal the allocation.

Interventions

AI-assisted education intervention group. Participants will be enrolled in a 12-month “Kangkang” AI assistant-based intervention program after surgery. The program will be developed using a rule-based engine and will cover perioperative knowledge, postoperative functional rehabilitation (e.g., swallowing and speech training), nutritional guidance, pain and symptom management, psychological support, follow-up reminders, and medication guidance. The intervention will include two components: (i) scheduled content delivery, in which personalized video clips and graphic guides will be delivered twice weekly according to the patient’s treatment stage (e.g., preoperative preparation, acute recovery, and long-term rehabilitation); and (ii) real-time interaction, in which patients will be able to consult the AI assistant via text or voice from 08:00 to 20:00 daily to receive immediate responses, including informational support and supportive messaging. All interaction logs will be recorded by the AI assistant and will be accessible to the research team for monitoring and quality oversight.

Control group: participants will receive standardized SMS-based health education for 12 months. The messages will provide concise and general guidance for perioperative care and recovery, delivered twice weekly. No personalized interactive question-and-answer function will be provided in the control group. At enrollment, the research team will instruct control-group participants to follow the predefined paper-based materials and SMS education plan throughout the study period. To minimized the differential attention bias, the check-ins will be standardized for both group. The details of intervention is presented in Table 2.

Table 2.

The details of the intervention

Item	AI group	Control group
Intervention period and schedule	Full coverage from admission to postoperative 1, 3, 6, and 12 months; messages pushed twice weekly (Tue & Sat); real-time AI Q&A available daily 08:00–20:00 (after-hours queries replied to next day)	Full coverage from admission to postoperative 1, 3, 6, and 12 months; standardized SMS pushed twice weekly (Tue & Sat)
Phase-specific procedures	Preoperative: Day 2: pre-op education video + infographic checklist; Day 1: anxiety self-help tips; nurses check learning completion; bedside coaching if needed. Postoperative: A set of short video libraries with a total duration of approximately 35 min, divided by themes (each segment is 2–5 min, covering wound care, pain management, rehabilitation training, etc.); Day 3 starts check-in tasks with automated personalized content; nurses review risk list daily and strengthen guidance for high-risk patients. Post-discharge: themed education twice weekly (rehab, nutrition, emotion regulation, etc.); scheduled check-in screening; high-risk patients receive phone follow-up within 48 h; on-demand consultation with triage and closed-loop management	Standardized SMS content at the same frequency (core education points, diet guidance, rehab reminders, follow-up visit notifications); questions handled through routine clinical pathways

The backend system of the “Kangkang” AI assistant platform will automatically and objectively record the following multi-dimensional usage data in AI group. For the control group, the number of active consultations will be recorded by researchers. To improve the adherence, in the AI group, completion of scheduled tasks will be incentivized with points that can be redeemed for low-cost small gifts (e.g. anti-noise earplugs, portable pill boxes, grip balls, portable toiletry bags, etc.). For participants with low adherence, a stepwise support strategy will be implemented, including automated reminders, nurse-led guidance, engagement of family caregivers, and bedside re-education when applicable. In the control group, adherence support will follow usual care procedures, without AI-based incentives. For controlling contamination between groups, four measures will be implemented, inlcuding distinct intervention modes lowering shareability, ethical informed consent guidance, separated blinded assessment and intervention, and natural clinical isolation from dispersed residence and home-based recovery.

The intervention may be discontinued or modified if: (1) the participant requests withdrawal; (2) a serious adverse event or significant worsening of clinical condition occurs (e.g., severe postoperative complications, disease progression); (3) the intervention causes or exacerbates psychological distress; or (4) the participant achieves sustained clinical improvement. All modification or discontinuation decisions will be documented with reasons.

Outcome variables

Perceived stress is the primary outcome of this study, which will be measured by the Chinese version of the Perceived Stress Scale (PSS) with good reliability and validity in cancer populations [13].

The key secondary outcomes are Fear of Cancer Progression (FoP-Q-S), WHOQOL-OLD, and UCLA Loneliness Scale (3rd edition)). The exploratory secondary outcomes include anxiety (GAD-7), depression (PHQ-9), self-efficacy/self-esteem (RSES), activities of daily living (BI), nutritional risk (NRS2002), pain (NRS), constipation-related quality of life (PAC-QOL), sleep quality (PSQI), fall risk (MFS). The secondary outcomes will be assessed as follows:

• Physiological function and self-care ability will be assessed using the BI [14]. Quality of life will be comprehensively evaluated using WHOQOL-OLD, a tool developed for older adults with sound psychometric properties [15]. In addition, supportive-care-related domains will be assessed using PAC-QOL [16], NRS2002 [17], NRS, and PSQI [18].

• Mental health will be assessed using the UCLA Loneliness Scale (3rd edition) to capture emotional loneliness and social isolation, which will have predictive value for identifying depressive symptoms in older adults [19], and RSES to evaluate self-worth, which will be particularly relevant for patients with HNC experiencing appearance- or function-related impairments after treatment [20]. Fear of cancer progression will be quantified using FoP-Q-SF, a core psychological factor affecting long-term quality of life among cancer survivors [21]. Additionally, secondary outcomes will also include fall risk assessed with the MFS [22], cognitive function screened using the Mini-Cog [23], anxiety symptoms assessed using GAD-7, and depressive symptoms assessed using PHQ-9 [24].

Data management and monitoring

Data management

All study data will be entered and managed through a password-protected electronic data capture system. Data will be double-entered and logically checked to ensure quality. All information that identifies the participant will be removed and replaced with a unique anonymized identifier. After the end of the study, all data will be encrypted for at least 3 years.

Safety monitoring

This study is a behavioral intervention trial and is not expected to result in serious direct physical harm. Therefore, no independent data monitoring committee is established. The health status of all participants will be noted at the follow-up assessment by the post-training research assistant. As AI assistant identifies the severe mental health problems, the notification will be immediately sent to the platform management backend. The primary investigator (Yunwei Sun) will log into the platform management backend at least twice a week to actively screen all interaction records and questionnaire scores to ensure no omissions. The red-flag case will be managed according to the emergency response plans and operational pathways to ensure patient safety and compliance with ethical standards. At the same time, the research team will provide telephone technical support and concise operation guides for technology use barriers that older patients may encounter (such as unskilled operation).

Interim analyses and stopping guidelines

No interim analyses are planned due to the low-risk nature of the intervention. No formal stopping guidelines are established. The principal investigator, in consultation with the ethics committee, has the authority to terminate the trial if multiple safety concerns arise.

Statistical analysis

Descriptive statistics will be used to summarize baseline demographic and clinical characteristics. Between-group comparability will be assessed using independent-samples t tests for continuous variables and chi-square tests for categorical variables at baseline. Intervention effects will be evaluated primarily under the intention-to-treat analysis. Linear mixed models will be used for analysing continues variables and maximum likelihood estimation will be used to manage missing data. The covriates will include age, tumor site/stage and baseline values of outcomes. The Bonferroni correction will be used for multiplicity adjustment. All analyses will be performed by SPSS (Version 29.0), with two-sided P < 0.05 considered statistically significant.

Discussion

This study aims to evaluate the effectiveness and feasibility of a 12-month AI-based personalized education support program to improve multidimensional outcomes in older patients with HNC through a rigorous randomized controlled trial. We hypothesize that compared with the control group receiving traditional standardized text message education, patients in the intervention group who are involved in the “Kangkang” AI assistant will not only have better maintenance or improvement in mental health and quality of life, but will also show the positive changes in social support domain.

There are some strengths of this study. First, at the level of clinical practice, it will provide an empirically supported, operable, and relatively low-cost model of continuous social support for the underserved group of older HNC patients. AI assistants can overcome human resource constraints and achieve 24/7 “presence” support, which is uniquely valuable for alleviating sudden anxiety or information needs during off-hours. Secondly, at the level of intervention science, this study explores how to combine structured health information with resilient emotional support through “regularized + personalized” content push and real-time interaction, which provides a design reference for the development of more complex digital behavioral interventions in the future. Furthermore, at the health policy level, the results of the study can provide an evidence-based basis for integrating similar AI-assisted support programs into the routine follow-up care pathway of cancer patients, especially in the context of China’s accelerated aging and uneven distribution of medical resources.

It is worth noting that this study has certain limitations. First, older adults differ substantially in smartphone use and digital literacy, which may influence engagement with the AI-based education and introduce heterogeneity in intervention effects. Therefore, standardized face-to-face onboarding training, user-friendly materials, and an age-friendly interface (e.g., larger fonts and voice input) should be provided. Second, although the “Kangkang” AI assistant can deliver real-time informational and supportive responses, it cannot fully replace the depth of professional psychological counseling. Thus, it will be positioned as an educational and basic psychosocial support tool rather than a substitute for psychotherapy. Third, the 12-month follow-up may increase attrition, potentially affecting data completeness. To mitigate this, multiple contact methods will be used in the AI Group, might induce the difference in attention leading to the expectancy effect. Finally, as a single-center trial, the risk of cross-group contamination cannot be fully excluded, even with the adoption of a multimodal intervention strategy. Nevertheless, since the core of this intervention lies in highly personalized, interactive experiences rather than universally replicable content, any attenuating impact of such contamination on study outcomes is expected to be minimal. Moreover, as a single-center study, the generalizability of our findings may be constrained. Future multicenter trials are therefore warranted to further validate the effectiveness of this intervention model.

In conclusion, this study addresses the urgent demand for continuous, personalized support among older patients with HNC. The findings of this trial will provide novel evidence for the practical value of AI technology in improving mental health, social support, and overall quality of life in older HNC survivors, which would enhance tumor rehabilitation outcomes among this specific population.

Abbreviations

AI

Artificial Intelligence

BI

Barthel Index

FoP-Q-SF

Short version of the Fear of Progression Questionnaire

GAD-7

Generalized Anxiety Disorder 7-item scale

HNC

Head and Neck Cancer

MFS

Morse Fall Scale

MTh

Month

NRS

Numeric Pain Rating Scale

NRS2002

Nutritional Risk Screening 2002

PAC-QOL

Patient Assessment of Constipation Quality of Life

PHQ-9

Patient Health Questionnaire 9-item scale

PSQI

Pittsburgh Sleep Quality Index

PSS

Perceived Stress Scale

Q&A

Questions and Answers

RSES

Rosenberg Self-Esteem Scale

SMS

Short Message Service

SPIRIT

Standard Protocol Items: Interventional Trial Recommendations

SPSS

Statistical Package for the Social Sciences

WHOQOL-OLD

World Health Organization Quality of Life Scale for Older Adults

Authors’ contributions

This study protocol was completed in collaboration with all authors. YWS proposed the project idea and served as the head of the single-center study. HM, WSZ, XHF, LY, YHP, and (all project leaders and principal investigators) further refined the research content. All authors were involved in the design of the study protocol. ZF, HHZ, and XMY developed the intervention and reviewed it by all authors. YWS wrote the first draft of the paper. MYX calculated the sample size and wrote the statistical section based on the data analysis plan. XHF, YHP, ZF, HHZ, XMY, and MYX participated in the writing of the paper and made important contributions to the revision of the paper and finally reviewed and approved the final version of the paper.

Funding

This study was Sponsored by Cancer Hospital Chinese Academy of Medical Sciences, Shenzhen Center (#E010425004) to support its scientific conception. Trial data will be analyzed independently of the trial sponsor. The funder will not be involved in the study design, data analysis, outcome reporting, or decision to publish the paper. The research proposal has not been peer-reviewed by the funding agency.

Data availability

No datasets were generated or analysed during the current study.

Declarations

This study has been approved by the Ethics Committee of Shenzhen Hospital, Cancer Hospital of the Chinese Academy of Medical Sciences (registration number JS2025-12-1). Any protocol amendment will be approved by the IRB first, then communicated to investigators (by email), participants (updated consent if needed), trial registry (ChiCTR2500114052), journal (upon publication), and regulators (as required). All changes will be versioned and dated. This study protocol follows the feasibility study principle. All participants and their legal guardians will be informed of the purpose of the study. All participants or their legal guardians must sign a written informed consent form in accordance with the Declaration of Helsinki before enrollment. Participants and their relatives or legal guardians may withdraw consent at any time. All participant information and data will be stored securely and identified only by a coded ID number to ensure participant privacy. No specific ancillary care is planned beyond standard postoperative care. Post-trial, participants return to standard hospital protocols. Any harm directly related to trial participation will be compensated by the study sponsor in accordance with local regulations.

Consent for publication

The personal information of the study participants will not be released.

Competing interests

The authors declare no competing interests.