Unplanned Hospitalization and Emergency Department Use in Patients With Head and Neck Cancer Receiving Radiotherapy

Background and Rationale Head and neck cancer (HNC) arises from the oral cavity, pharynx, larynx, nasal cavity, paranasal sinuses, and salivary glands and is the sixth most common malignancy worldwide. In China, approximately 148,000 new HNC cases and 78,000 deaths were reported in 2021, and the burden continues to increase. Most patients receive radiotherapy-often with concurrent chemotherapy or after surgery-and intensity-modulated techniques (IMRT/VMAT) are the standard of care. Radiotherapy and chemoradiotherapy cause substantial acute toxicity, including radiation-induced oral mucositis, orofacial pain, xerostomia, dysphagia, nausea, dehydration, radiation dermatitis, and myelosuppression; these toxicities commonly peak during treatment and may persist after discharge.

Unplanned hospitalization and emergency department (ED) use are important patient-centered and health-system outcomes. Among patients with cancer, the 30-day unplanned readmission rate is 8.2%-19.0%, higher than after general surgery, and the burden is greater still in HNC. A large Canadian cohort (n=3,898) found that 28%-55% of HNC patients had at least one ED visit or unplanned hospitalization within 90 days of treatment, and patients receiving concurrent chemoradiotherapy may have admission rates approaching 58%. Such events disrupt treatment-radiotherapy interruptions are associated with lower overall survival-and generate considerable cost. Recognized risk factors include older age, comorbidity, frailty, low BMI, weight loss ≥10%, low albumin, concurrent chemotherapy, feeding-tube use, and prolonged hospitalization. However, most evidence is retrospective, focused on postoperative patients, and based on single time-point measurements. The predictive value of longitudinal inflammatory-nutritional markers and patient-reported symptom burden during radiotherapy, when acute toxicity evolves rapidly, has not been adequately characterized in prospective cohorts, and no validated early-warning tool specific to HNC patients undergoing IMRT is available in China.

Objectives Primary objective: to determine the incidence of unplanned hospitalization occurring during IMRT and within 30 days after radiotherapy completion and to identify associated factors.

Secondary objectives: (1) to estimate the incidence and reasons for ED use; (2) to characterize longitudinal trajectories of patient-reported symptoms and serial inflammatory-nutritional laboratory indices; (3) to examine the associations of these trajectories with unplanned hospitalization and other acute outcomes, including severe (grade ≥3) oral mucositis, radiotherapy interruption, and significant weight loss/feeding-tube use; and (4) to develop and internally validate a multivariable risk prediction model for unplanned hospitalization, comparing conventional logistic regression with a machine-learning (random forest) approach and presenting the model as a nomogram.

Study Design and Setting This is a single-center, prospective, observational cohort study conducted at the head-and-neck oncology and radiotherapy unit of a tertiary teaching hospital in China. Eligible consecutive patients are enrolled before the start of IMRT. No study-assigned intervention is given; all patients receive standard radiotherapy and supportive care according to institutional practice. A single cohort is used; participants are not randomized and clinical treatment is not altered by the study.

Participants Adults (≥18 years) with histologically confirmed HNC who are scheduled to receive IMRT/VMAT (definitive or postoperative, with or without concurrent chemotherapy or targeted therapy) and who provide written informed consent will be included. Patients not receiving IMRT, those treated primarily for another major disease without anti-tumor therapy, those unable to complete follow-up, and those who die in hospital or forgo treatment will be excluded. Full criteria are provided in the Eligibility module.

Procedures and Data Collection

Data are collected at standardized time points using a structured case report form and extracted from the electronic medical record:

• Baseline (before radiotherapy): demographic and socioeconomic characteristics (age, sex, marital status, education, income, employment); behavioral exposures with dose information (cigarette smoking, alcohol, betel-nut use); performance status (ECOG); nutritional risk (NRS-2002), height, weight and BMI; anxiety/depression screening; comorbidity quantified by the (age-adjusted) Charlson Comorbidity Index and a frailty assessment; tumor site, histology, TNM stage (and HPV/p16 status where available); surgery within the preceding 3 months; and planned radiotherapy parameters (technique, target doses, number of fractions) and systemic therapy.
• During radiotherapy (at a scheduled mid-treatment assessment): repeat blood tests, weight, clinician-rated toxicity (CTCAE), and patient-reported symptoms.
• Before discharge / at treatment completion: repeat blood tests, weight, toxicity, symptoms, frailty, feeding-tube status, and treatment interruption.
• Serial laboratory tests include complete blood count parameters (leukocytes, neutrophils, lymphocytes, monocytes, platelets, hemoglobin), serum albumin, electrolytes and related measures. From these, inflammatory-nutritional indices are derived at each time point: Prognostic Nutritional Index (PNI = 10 × albumin [g/dL] + 0.005 × total lymphocyte count [/µL]), Neutrophil-to-Lymphocyte Ratio (NLR), Platelet-to-Lymphocyte Ratio (PLR), Lymphocyte-to-Monocyte Ratio (LMR), and Systemic Immune-Inflammation Index (SII).
• Patient-reported symptoms are assessed with a validated head-and-neck symptom instrument (MDASI-HN), capturing symptom severity and interference on 0-10 scales; oral care behaviors (oral cleaning, mouthwash frequency) are also recorded.

Outcome Ascertainment and Follow-up Participants are followed from the first fraction of IMRT through 30 days after radiotherapy completion/discharge. Trained research nurses perform weekly follow-up via telephone and medical-record review (including admissions to other hospitals) and document every acute-care event with its date and reason.

• Unplanned hospitalization is defined as any unscheduled inpatient admission during the observation window, encompassing (i) transfer from the outpatient radiotherapy clinic to an inpatient ward during treatment, (ii) admission initiated through the ED, and (iii) readmission after treatment discharge; planned elective admissions are excluded.
• ED use is defined as any unscheduled emergency department visit, whether or not it leads to admission.
• Additional acute outcomes recorded include severe (grade ≥3) radiation-induced oral mucositis and other CTCAE toxicities, radiotherapy interruption and its duration, weight loss ≥10%/feeding-tube placement, and mortality.

Statistical Analysis Analyses will be performed using SPSS and R. (a) The incidence of unplanned hospitalization and ED use will be reported with 95% confidence intervals, together with the reasons and timing of events. (b) Baseline and longitudinal characteristics will be compared between participants with and without events using independent-samples t tests or Mann-Whitney U tests and chi-square or Fisher exact tests as appropriate. (c) Multivariable logistic regression will identify independent associated factors; longitudinal symptom and laboratory trajectories will be modeled with mixed-effects models and/or group-based trajectory modeling, and their association with outcomes assessed. (d) A risk prediction model will be developed using logistic regression (with LASSO or prespecified variable selection as required) and presented as a nomogram; a random forest classifier will be trained for comparison, with variable-importance ranking. (e) Model performance will be assessed for discrimination (AUC/C-statistic), calibration (Hosmer-Lemeshow test, calibration plot, Brier score), and clinical utility (decision-curve analysis), with internal validation by bootstrap and/or k-fold cross-validation, and temporal split-sample validation where available. Missing data will be addressed by complete-case analysis with sensitivity analyses and multiple imputation where appropriate. Sample size is guided by a criterion of at least 10 outcome events per candidate predictor; the target enrollment is specified in the Study Design module.

Quality Control Investigators will be trained on standardized definitions and data extraction. Data will be double-entered and cross-checked, with source verification against the medical record, and regular team meetings and periodic ethics review will be conducted.

Ethics, Consent and Data Governance The study has been approved by the institutional ethics committee (approval number 2025101711). All participants provide written informed consent before enrollment and may withdraw at any time without affecting their treatment or care. The study involves minimal risk and no additional invasive procedures; only data generated during routine care and questionnaire responses are used. Data are de-identified for analysis, and an individual participant data sharing statement is provided separately.