A Discharge Summary Prompt Reduces Postoperative Radiotherapy Delay in Head and Neck Cancer Patients

Abstract

Objectives

Delays in postoperative radiotherapy (PORT) beyond 6 weeks are associated with worse oncologic outcomes in head and neck cancer. Up to 50% of patients experience PORT delays in real-world practice. Inpatient discharge following definitive resection represents a critical handoff point where standardized interventions may reduce delay. This study evaluates the implementation of an electronic health record (EHR) discharge summary prompt as a quality improvement (QI) initiative to decrease PORT delay in patients undergoing head and neck cancer surgery and identifies additional factors associated with delayed PORT.

Methods

A retrospective review of patients receiving PORT from 2018–2023 was used as a pre-intervention baseline. An EHR-based discharge summary intervention was introduced in May 2023. The discharge summary prompt asked providers three questions: “1. Does the patient need adjuvant radiation? 2. Was a radiation oncology referral made? 3. Was a dental referral made?” Patients who underwent head and neck cancer surgery and received PORT before and after the intervention were reviewed. The primary outcome was delayed PORT initiation, defined as >42 days from surgery to PORT start. Delay rates were compared using χ² testing. Time to PORT initiation was evaluated with Kaplan–Meier analysis and log-rank testing. Univariate and multivariable logistic regression were used to identify predictors of delayed PORT, and Cox proportional hazards modeling was used to evaluate factors associated with time to PORT initiation.

Results:

A total of 391 patients were included (pre-intervention n=260; post-intervention n=131). Following implementation of standardized discharge documentation, delayed PORT initiation (>42 days) decreased from 74.2% (193/260) to 46.6% (61/131), an absolute reduction of 27.6% (χ²=28.67, p=8.59×10⁻⁸), and post-intervention status remained associated with lower odds of delay on multivariable analysis (OR=0.28, 95% CI 0.14–0.56; p=0.0004). Median time to PORT improved from 52 days (95% CI 50–56) pre-intervention to 42 days (95% CI 41–45) post-intervention (log-rank χ²=40.4, p=2.0×10⁻¹⁰), with faster initiation in the adjusted Cox model (HR=2.05, 95% CI 1.43–2.93; p=7.70×10⁻⁵). In adjusted models, higher pathologic stage (T2/T4a and N3b), receipt of radiation alone (vs chemoradiation), and postoperative complications were associated with delay, and in the Cox model uninsured/VA insurance was also associated with slower initiation.

Conclusions

A simple EHR-based discharge summary prompt reduced PORT delay in head and neck cancer patients. While PORT delay is multifactorial in etiology, this study demonstrates how a targeted QI measure may reduce PORT delay within nationally recommended guidelines.

Introduction

Head and neck cancer (HNC) is diagnosed in approximately 72,000 patients and causes 16,680 deaths per year. [1] While many HNC patients undergo surgical resection, as many as 2/3rds of patients necessitate multimodal treatment including adjuvant radiotherapy (RT) or chemoradiotherapy (CRT).[2,3] As the incidence of HNC continues to rise[1], awareness and adherence to the National Comprehensive Cancer Network’s (NCCN) guidelines[3] for post-operative radiotherapy (PORT) are increasingly important. The significance of these guidelines has been demonstrated in recent studies showing that a time to PORT (TTP) of greater than 6 weeks is associated with worse overall survival and locoregional control.[3–8] These findings led both the Commission on Cancer and American Head and Neck Society to implement a quality metric reinforcing the 6-week window for timely PORT.

Despite the clear importance of timely PORT, many patients fail to receive PORT within the 6-week window. A recent meta-analysis by Duckett et al. reported that up to 49% of patients in the US still fail to receive PORT within this timeframe.[9] Meanwhile, Divi et al estimate that the optimal rate timely PORT is approximately 80% given some unavoidable causes of PORT delay.[10] In recent years multiple studies have begun to elucidate causes of PORT delay.[9,11–14] While our understanding of the multifactorial nature of delay has improved, there is a scarcity of literature evaluating interventions for mitigating risk of PORT delay.[2,10] While a small number of institutions have described quality improvement (QI) interventions aimed at reducing PORT delay rates[2,10,15] replicable QI tools are needed if widespread adoption and enduring results are to be achieved.

In a previous single-institution study by Vivek et al, 73% of patients at our institution had PORT delay.[11] Considering these findings, this prospective cohort study was designed to evaluate the impact of a discharge summary-based QI tool aimed at decreasing TTP. While other studies have examined the implementation of complex multidisciplinary interventions[7,10], this investigation is important in assessing the utility of a less resource-intensive QI intervention which may be more easily replicated by other institutions.

Methods

Study Design

This study was approved by the Vanderbilt University Medical Center Institutional Review Board (IRB #221228). The study period was from January 1^st, 2018 – June 1^st, 2025. Our cohort was formed from all patients evaluated in the Vanderbilt University Medical Center Head and Neck clinic who received surgery for HNC and required either adjuvant radiation therapy or chemoradiotherapy. Patients received PORT at either Vanderbilt University Medical Center (VUMC) or external facilities within Tennessee, Kentucky, Alabama, or Georgia. The primary outcome of this study was a delay in PORT, defined as time to radiation initiation greater than 42 days. Secondary outcomes included time to PORT initiation (days from surgery to start of radiation therapy) and identification of additional patient, disease, and treatment-related factors associated with delayed PORT initiation (>42 days).

Patient Selection

We systematically reviewed all surgical cases performed by faculty head and neck surgeons and included patients who underwent oncologic resections for HNC. Cases identified as biopsies, benign tumor removals, and minor skin resections were excluded prior to chart review to ensure inclusion of only patients with an indication for PORT. All patients 18 years or older requiring adjuvant radiation or chemoradiation therapy with English language proficiency were included in this study. Patients who did not receive PORT or were lost to follow up were excluded from this study. The pre-intervention control group was from January 2018 to May 2023. A QI measure was implemented on May 15, 2023, and intervention group data was collected from May 15^th, 2023-June 1^st, 2025.

Intervention

An electronic health record (EHR)-based discharge summary intervention was introduced in May 2023 for patients undergoing head and neck cancer surgery (Figure 1). Providers completed these prompts at the time of postoperative discharge. Completion of these questions was intended to improve provider communication and facilitate timely initiation of PORT.

Figure 1. Standardized Discharge Summary Documentation Fields from Electronic Health Record.

Screenshot of the standardized discharge summary template implemented in May 2023 for patients with head and neck cancer undergoing curative-intent surgical resection. The template includes three mandatory dropdown fields to be filled out by the provider upon postoperative hospital discharge: (1) documentation of need for adjuvant radiation therapy (Yes/No/Unknown), (2) confirmation of radiation oncology referral (Yes/NA/No), and (3) confirmation of dental referral (Yes/NA/No).

Data Collection

De-identified clinical data was extracted from the Vanderbilt University Medical Center Epic Electronic Health Record (Epic Systems Corporation, Verona, WI, USA) and entered into a REDCap (Research Electronic Data Capture) database. Data abstraction was performed via manual patient chart review by trained research personnel. Data extracted included patient characteristics such as age, sex, race/ethnicity, home address and distance to radiation facility, employment status, education level, insurance status, smoking status, and alcohol use. Tumor characteristics collected include cancer site, histology, margin status, TNM stage, and extranodal extension. Treatment data included items such as surgery date, surgical procedure, postoperative medical complications, radiation oncology referral date, need for dental extraction, radiation facility (VUMC or outside radiation facility), and radiation start date.

Statistical Analysis

All statistical analyses and data visualization were performed in RStudio (Version 2024.12.1+563).

Baseline characteristics were summarized using descriptive statistics. Differences in the proportion of patients experiencing delayed PORT between the pre- and post-intervention groups were assessed using the chi-square test. Baseline categorical variables between the pre- and post-intervention groups were compared using the chi-square test or Fisher’s exact test, as appropriate. Continuous variables (eg. age, distance to radiology facility) were compared using Student’s t-test for normally distributed variables and Wilcoxon rank-sum test for non-normally distributed variables; normality was assessed using the Shapiro-Wilk test.

Univariate Regression

Univariate logistic regression was conducted to assess associations between individual variables and delayed PORT. Odds ratios (ORs), 95% confidence intervals (CIs), and p-values were estimated for each variable. For time-to-event outcomes (days to initiation of PORT), univariate Cox proportional hazards regression was performed, with hazard ratios (HRs) 95% CIs, and p-values reported. Statistical significance was defined as p < 0.05.

Multivariable Modeling

Covariates with significant associations in univariate analyses, as well as clinically relevant variables, were included in multivariable models. Variables were excluded if deemed clinically irrelevant, were collinear, or did not improve model fit. Multivariable logistic regression was used to estimate adjusted ORs for delayed PORT. A multivariable Cox proportional hazards model was constructed for time to initiation of radiation therapy. Due to violation of the proportional hazards assumption for pathologic T stage (Schoenfeld residuals test p = 0.012), the final Cox model stratified by pathologic T stage, allowing for different baseline hazards across T stage groups while estimating common covariate effects.

Model discrimination for logistic regression was assessed using the area under the receiver operating characteristic curve (AUC), and calibration was evaluated with the Hosmer-Lemeshow goodness-of-fit test. For the Cox proportional hazards model, discrimination was assessed using Harrell’s concordance index (C-index), with 95% confidence intervals calculated from the standard error, and overall model fit was assessed using the likelihood ratio test. After stratification by pathologic T stage, the proportional hazards assumption was verified using Schoenfeld residuals with the global test. Multicollinearity among covariates in both models was evaluated using variance inflation factor (VIF) for the logistic regression model and generalized variance inflation factors (GVIF) for the Cox model, with values below 5 considered acceptable. Internal validation of the Cox model was performed using bootstrap resampling with 200 iterations to assess model stability.

Survival Analysis

Kaplan-Meier survival curves were generated to estimate time to initiation of postoperative radiation therapy for included variables. Group differences in time-to-event distributions were tested using the log-rank test.

All statistical analyses and data visualization were performed in RStudio (Version 2024.12.1+563).

Results

Cohort Characteristics

A total of 391 patients were included, with 260 (66.5%) patients in the pre-intervention cohort and 131 (33.5%) in the post-intervention cohort. The overall cohort had a mean age of 62.4 years (standard deviation, SD = 12.6) and was predominantly male (n = 289, 73.7%) and white (n = 348, 89.7%). Most patients were married or partnered (n = 252, 64.3%) and had private insurance (n = 224, 57.1%). The most common overall cancer subsite was the oral cavity (n = 175, 44.6%). Most patients received concurrent chemoradiotherapy after surgery (n = 224, 57.1%) while 168 (42.9%) patients received postoperative radiotherapy alone. Postoperative medical complications occurred in 56 (14.3%) of patients and the mean distance to radiation facility was 49.6 miles (SD = 105.6). Most patients received postoperative radiation treatment at VUMC (n = 218, 55.6%). Most patients had a pathologic tumor (T) stage of T3 or higher (n = 265, 67.6%) and nodal (N) stage of N2 or higher (n = 220, 56.1%).

Pre- and post-intervention cohorts were similar in regard to race, gender, age, medical complications, distance to radiation facility, and pathologic T stage. Several baseline characteristics differed significantly between the pre- and post-intervention cohorts. The post-intervention cohort had a higher proportion of patients receiving radiation at VUMC (83.6% vs 53.9%, p < 0.001) and a lower proportion receiving concurrent chemoradiotherapy (48.8% vs 63.7%, p=0.008). Cancer subsite distribution differed between groups (p=0.001), with the post-intervention cohort having more oropharynx cancers (22.9% vs 10.8%) and fewer larynx/hypopharynx cancers (4.6% vs 13.1%). Pathologic N stage distribution also differed (p=0.002), with the post-intervention cohort having more N1 disease (33.3% vs 12.4%) and less N3b disease (15.9% vs 23.2%). Marital status (p=0.025) and insurance status (p=0.008) also differed between cohorts. Baseline demographics and group comparisons are summarized in Table 1.

Table 1.

Baseline Characteristics of Study Cohort by Intervention Period.

Variable	Pre-Intervention, n (%)	Post-Intervention, n (%)	p-value
Overall	260	131
Health Insurance Status			0.008
Medicare	94 (36.6)	43 (32.8)
Medicaid	3 (1.2)	0 (0.0)
Private	140 (54.5)	84 (64.1)
Uninsured	17 (6.6)	0 (0.0)
VA	3 (1.2)	4 (3.1)
Marital Status			0.03
Single	2 (0.8)	3 (2.5)
Married/Current Partner	181 (73.6)	71 (59.7)
Separated/Divorced	21 (8.5)	19 (16.0)
Widowed	42 (17.1)	26 (21.8)
Race			0.162
White	233 (89.7)	115 (89.6)
Non-White	27 (10.3)	13 (9.9)
Cancer Subsite			0.001
Larynx/Hypopharynx	40 (10.2)	34 (13.1)
Oral Cavity	120 (46.2)	55 (42.0)
Oropharynx	28 (10.8)	30 (22.9)
Nasal Cavity/Paranasal Sinuses	18 (6.9)	3 (2.3)
Salivary	16 (6.2)	16 (12.2)
Skin/Scalp	37 (14.2)	15 (11.5)
Unknown primary	7 (2.7)	6 (4.6)
Radiation with or without CRT			0.008
Radiation alone	93 (36.3)	64 (51.2)
CRT	163 (63.7)	61 (48.8)
Dental Evaluation Prior to Radiation			<0.001
Yes	87 (35.5)	21 (17.2)
No	41 (16.7)	36 (29.5)
Unable to Determine	117 (47.8)	65 (53.3)
Distance to Radiation Facility (miles), ()	()	()	0.
Postoperative Complications			0.91
Yes	36 (14.3)	20 (15.4)
No	215 (85.7)	110 (84.6)
Radiation at VUMC			<0.001
Yes	111 (53.9)	107 (83.6)
No	95 (46.1)	21 (16.4)
Pathologic Tumor Stage			0.14
1	24 (9.4)	25 (19.4)
2	51 (20.1)	27 (20.9)
3	66 (26.0)	31 (24.0)
4	2 (0.8)	0 (0.0)
4a	96 (27.8)	38 (29.5)
4b	5 (2.0)	2 (1.6)
x	10 (3.9)	6 (4.7)
Pathologic Nodal Stage			0.002
0	67 (27.8)	33 (26.2)
1	30 (12.4)	42 (33.3)
2	4 (1.7)	1 (0.8)
2a	16 (6.6)	9 (7.1)
2b	26 (10.8)	9 (7.1)
2c	8 (3.3)	3 (2.4)
3	5 (2.1)	1 (0.8)
3a	2 (0.8)	0 (0.0)
3b	56 (23.2)	10 (15.9)
x	27 (11.2)	8 (6.3)
Alcohol Use			0.42
Current	101 (38.8)	58 (44.3)
Previous	71 (27.3)	30 (22.9)
Never	85 (32.7)	43 (32.8)
Tobacco Use			0.01
Current	47 (18.1)	17 (13.0)
Previous	125 (48.1)	48 (36.6)
Never	87 (33.5)	66 (50.4)
Employment Status			<0.001
Full-Time Paid Work	61 (23.5)	28 (21.4)
Disability	27 (10.4)	1 (0.8)
Part-Time Paid Work	3 (1.2)	2 (1.5)
Retired	68 (26.2)	7 (5.3)
Unemployed	16 (6.2)	3 (2.3)
Other	2 (0.8)	0 (0.0)
Age at Surgery, mean (SD)	62.99 (12.12)	61.37 (13.59)	0.23
Gender			0.43
Female	64 (24.6)	35 (26.7)
Male	193 (74.2)	96 (73.3)
Living Situation			<0.001
Spouse	91 (92.9)	7 (53.8)
Alone	4 (4.1)	2 (15.4)
Children	0 (0.0)	2 (15.4)
Friends	1 (1.0)	1 (7.7)
Parents	2 (2.0)	1 (7.7)

Baseline demographic, clinical, and treatment-related characteristics of 391 patients with head and neck cancer who underwent curative-intent surgical resection with pathologic indication for postoperative radiation therapy (PORT), stratified by intervention period: pre-intervention (January 2018 - May 2023, n=260) versus post-intervention (May 2023 2022 - June 2025, n=131). The intervention consisted of standardized discharge documentation with mandatory fields for adjuvant radiation indication and referral confirmation, implemented in May 2023. Categorical variables are presented as n (%), with percentages calculated within each intervention group. Continuous variables are presented as mean ± standard deviation (SD) for normally distributed variables or median (interquartile range, IQR) for non-normally distributed variables. P-values were calculated using χ² tests for categorical variables, independent t-tests for normally distributed continuous variables, and Wilcoxon rank-sum tests for non-normally distributed continuous variables. All tests two-sided with α = 0.05.

Primary Outcome: Delayed PORT by Intervention Group

Dichotomous Outcome (Delayed vs. Not Delayed):

Following implementation of the intervention, the proportion of patients experiencing delayed PORT initiation (> 42 days) decreased from 74.2% (193/260 patients) in the pre-intervention group to 46.6% (61/131 patients) in the post-intervention group. This represents an absolute reduction of 27.6% in delayed PORT, with the difference in proportions being statistically significant (χ² = 28.67, df = 1, 95% CI: 17.3% - 37.4%, p = 8.59 × 10⁻⁸).

Time-To-Event Analysis

Kaplan-Meier analysis demonstrated significantly faster PORT initiation in the post-intervention cohort compared to the pre-intervention cohort (log-rank test: χ² = 40.4, p = 2.0 × 10⁻¹⁰) (Figure 2). Median time to PORT was 52 days (95% CI: 50–56; Interquartile Range, IQR: 42–67) in the pre-intervention group (n = 260) versus 42 days (95% CI: 41–45; IQR: 35–51) in the post-intervention group (n = 131), representing a 10-day (19%) reduction in median time to treatment. The post-intervention cohort also demonstrated reduced variability in treatment timing, with a narrower interquartile range (14.5 vs. 25.2 days) and standard deviation (15.8 vs. 44.2 days).

Figure 2. Kaplan-Meier Survival Curves for Time to Postoperative Radiation Therapy Initiation by Intervention Period.

Kaplan-Meier curves showing the cumulative probability of not yet initiating postoperative radiation therapy (PORT) over time following curative-intent surgical resection for head and neck cancer, stratified by intervention period. The pre-intervention cohort (orange line, n = 260, January 2018 - May 2023) had median time to PORT of 52 days (95% CI: 50–56 days). The post-intervention cohort (blue line, n = 131, May 2023-December 2025) had median time to PORT of 42 days (95% CI: 41 – 45 days). Shaded regions represent 95% confidence intervals. The log-rank test demonstrated significantly shorter time to PORT initiation in the post-intervention period (χ² = 40.4, p < 0.001). The post-intervention period began in May 2023 with implementation of standardized discharge summary documentation including mandatory fields for adjuvant radiation need and radiation oncology referral. Numbers at risk are displayed below the curves at 20-day intervals. All patients (n = 391) had pathologic indication for adjuvant radiation therapy.

Univariate Regression Analysis

Univariate Logistic Regression

Post-intervention status was associated with significantly lower odds of PORT delay (OR = 0.31, 95% CI: 0.20–0.48, p = 1.73 × 10⁻⁷). Further, dental evaluation (OR = 2.13, 95% CI: 1.13–4.03, p = 0.02), requiring dental extraction, (OR = 2.58, 95% CI: 1.34–5.26, p = 0.006), and undergoing dental extraction (OR = 2.44, 95% CI: 1.27–5.00, p = 0.01) prior to radiation therapy were associated with increased odds of delay. Each additional day from surgery to radiation oncology referral increased the odds of delay (OR = 1.02 per day, 95% CI: 1.01–1.03, p = 0.0002), and receiving radiation at a facility other than VUMC was also associated with higher odds of delay (OR = 1.95, 95% CI: 1.21–1.32, p = 0.007). Previous smoking history (OR: 2.02, 95% CI: 1.28–3.22, p = 0.003) and lack of insurance (OR = 9.20, 95% CI: 1.79 – 168.0, p = 0.03) were also significant risk factors for delayed PORT.

Higher pathologic tumor (T) stage (T2-T4a) was associated with increased odds of PORT delay; T2 (OR = 2.90, 95% CI: 1.40–6.15, p = 0.005), T3 (OR = 2.35, 95% CI: 1.17–4.80, p = 0.02), T4a (OR = 3.53, 95% CI: 1.80–7.06, p = 0.0003) and unknown primary (“x”) (OR = 6.28, 95% CI: 1.76–30.10, p = 0.009). Stage 4b was not significantly associated (OR = 1.93, 95% CI: 0.39–10.70, p = 0.42). Pathologic nodal (N) staging was not significantly associated with PORT delay; however, stage N3b showed a trend (OR = 1.92, 95% CI: 1.00–3.75, p = 0.052). Regarding tumor type, skin squamous cell carcinoma (SCC) (OR = 2.64, 95% CI: 1.05–7.06, p = 0.045), had significantly increased odds of PORT delay compared to those with p16-positive SCC mucosa and those with p16-negative SCC mucosal histology had a non-significant trend (OR = 1.75, 95% CI: 0.82–3.30, p = 0.08).

Other clinical and demographic variables, including age, gender, living situation, radiation modality, distance to treatment, cancer subsite, postoperative medical complications, marital status, surgical and reconstructive procedure, and margin status, were not associated with PORT delay on univariate regression analysis. Complete results for all covariates in this univariate logistic regression analysis are presented in a supplemental table.

Univariate Cox Regression

In a Cox proportional hazards model, patients in the post-intervention group experienced a significantly shorter time to PORT initiation compared to the pre-intervention group (HR = 2.01, 95% CI: 1.62–2.50, p = 3.54 × 10⁻¹⁰). Uninsured patients (HR = 0.41, 95% CI: 0.25–0.70, p = 0.0009) had a slower rate of PORT initiation compared to Medicare. Pairwise comparisons confirmed that both Medicare (p = 0.008) and private insurance (p = 0.01) were associated with faster PORT initiation compared to uninsured patients. Tumor subsite was also significantly associated with PORT initiation (p = 0.005). Compared to oral cavity tumors, oropharyngeal (HR = 1.46, 95% CI: 1.08–1.97, p = 0.01) and salivary gland tumors (HR = 1.75, 95% CI: 1.20–2.57, p = 0.004) were associated with faster initiation. Other subsites did not differ significantly. Further, histology was overall significantly associated (p = 0.03) with time to PORT initiation, with mucoepidermoid carcinoma displaying a longer time to initiation (HR = 5.12, 95% CI: 1.23–21.33, p = 0.03), while p16-negative mucosal SCC had shorter time (HR = 0.64, 95% CI: 0.47–0.88, p = 0.006). Other histologic subtypes were not significant.

Undergoing dental evaluation (HR = 0.61, 95% CI: 0.46–0.83, p = 0.001), requiring a dental extraction (HR = 0.58, 95% CI: 0.43–0.77, p = 0.0002), and having a dental extraction performed (HR = 0.62, 95% CI: 0.46–0.83, p = 0.001) before the start of radiation were significantly associated with a longer time to PORT initiation. Additionally, receiving radiation at a facility other than VUMC was associated with slower PORT initiation (HR = 0.75, 95% CI: 0.60–0.94, p = 0.01). Radiation modality (radiation alone vs. chemoradiation) and distance to radiation facility were not significantly associated.

Employment status was not broadly associated with slower PORT initiation, but patients on disability had slower PORT initiation compared to those with full-time paid work (HR = 0.64, 95% CI: 0.42–0.98, p = 0.04) and unemployed patients trended towards significantly slower PORT initiation (HR = 0.51, 95% CI: 0.37–1.007, p = 0.054).

Medical complications after surgery (HR = 0.65, 95% CI: 0.48–0.87, p = 0.004) and post-operative complications requiring return to OR (HR = 0.66, 95% CI: 0.49–0.87, p = 0.004) were associated with delayed PORT initiation. Surgical procedure was not associated with time to PORT initiation; however, undergoing parotidectomy trended towards faster PORT initiation (HR = 1.25, 95% CI: 0.97–1.62, p = 0.08). Not undergoing a reconstructive procedure (HR = 1.24, 95% CI: 1.001–1.52, p = 0.049) was associated with a faster time to PORT initiation while undergoing a fasciocutaneous free flap (HR = 0.81, 95% CI: 0.65–1.003, p = 0.053) trended towards slower initiation.

Compared to patients with pathologic T1 disease, those with T2 (HR = 0.69, 95% CI: 0.48–0.99, p = 0.04), T3 (HR = 0.66, 95% CI: 0.47–0.93, p = 0.02), T4a (HR = 0.57, 95% CI: 0.41–0.79, p = 0.0007), and unknown (“x”) stage (HR = 0.52, 95% CI: 0.29–0.91, p = 0.02) experienced significantly greater delays in PORT initiation. Patients with pathologic N1 disease experienced faster PORT initiation compared to patients with N0 disease (HR = 1.36, 95% CI: 1.002–1.85, p = 0.049).

Age, race, sex, marital status, living situation, tobacco/alcohol use, prior cancer treatment modality, margin status, HPV status, extranodal extension, and most other employment and clinical categories were not associated with time to PORT initiation (all p > 0.05).

Complete results for all covariates in this univariate cox regression analysis are presented in a supplemental table.

Multivariable Regression Analysis

Multivariable Logistic Regression

On multivariable logistic regression analysis (Table 2), patients in the post-intervention cohort had significantly lower odds of PORT delay (OR = 0.28, 95% CI: 0.14 – 0.56, p = 0.0004). Increased odds of delay were further observed in patients with T2 (OR = 4.19, 95% CI: 1.48 – 12.59, p = 0.008), T4a (OR = 3.43, 95% CI: 1.17 – 10.44, p = 0.03), and N3b (OR = 2.96, 95% CI: 1.16 – 7.90, p = 0.03) disease compared to T1 and N0 disease, respectively. Patients who received radiation alone had increased odds of delay compared to those who received chemoradiation (OR = 2.92, 95% CI: 1.48 – 5.98, p = 0.003).

Table 2.

Univariate and Multivariable Logistic Regression Analysis for Delayed PORT Initiation.

Variable	Multivariable Odds Ratio (95% CI)	p-value
Intervention (Ref = Pre-Intervention)
Post-Intervention	0.28 (0.14 – 0.56)	<0.001*
Health Insurance Status (Ref = Medicare)
Medicaid	0.23 (0.007 – 7.31)	0.35
Private	1.13 (0.59 – 2.18)	0.71
Uninsured	12.46 (1.50 – 329.72)	0.051
VA	5.56 (0.35 – 161.77)	0.24
Marital Status (Ref= Single)
Married/Current -Partner	1.47 (0.66 – 3.26)	0.34
Separated/divorced	2.21 (0.71 – 7.18)	0.18
Widowed	1.22 (0.30 – 5.31)	0.78
Race (Ref = White)
Non-White	0.64 (0.23 – 1.75)	0.38
Cancer Subsite (Ref = Oral Cavity)
Larynx/Hypopharynx	0.54 (0.19 – 1.57)	0.26
Oropharynx	2.15 (0.69 – 7.00)	0.19
Nasal Cavity/Paranasal sinuses	0.69 (0.18 – 2.78)	0.60
Salivary	1.26 (0.40 – 4.11)	0.69
Skin/Scalp	1.93 (0.69 – 5.70)	0.22
Unknown primary	1.64 (0.22 – 12.30)	0.61
Radiation with or without CRT (Ref = CRT)
Radiation alone	2.92 (1.48 – 5.98)	0.003*
Dental Evaluation Prior to Radiation (Ref = No)
Yes	1.68 (0.68 – 4.21)	0.27
Distance to Radiation Facility (miles)	1.002 (0.99 – 1.01)	0.18
Postoperative Complications (Ref = No)
Yes	2.47 (1.006 – 6.52)	0.056
Radiation at VUMC (Ref = Yes)
No	1.65 (0.83 – 3.31)	0.16
Pathologic Tumor Stage (Ref = Stage 1)
2	4.19 (1.48 – 12.59)	0.008*
3	2.07 (0.68 – 6.50)	0.20
4†	-	-
4a	3.43 (1.17 – 10.44)	0.03*
4b	6.00 (0.039 – 11.75)	0.72
x	3.69 (0.51 – 37.32)	0.22
Pathologic Nodal Stage (Ref = Stage 0)
1	1.34 (0.53 – 3.45)	0.54
2	0.73 (0.041 – 21.11)	0.83
2a	2.20 (0.60 – 9.13)	0.25
2b	1.26 (0.40 – 4.09)	0.68
2c	7.99 (0.13 – 5.50)	0.81
3†	-	-
3a†	-	-
3b	2.96 (1.16 – 7.90)	0.03*
x	1.06 (0.32 – 3.56)	0.93

Odds ratios (OR) and 95% confidence intervals (CI) from logistic regression models evaluating predictors of delayed postoperative radiation therapy (PORT) initiation (>42 days from surgery) following curative-intent surgery for head and neck cancer (n=390). Univariate analyses examined each variable separately. The multivariable model was stratified by pathologic T stage and adjusted for intervention period, health insurance status, marital status, race, cancer subsite, treatment modality, dental evaluation, distance to radiation facility, postoperative complications, pathologic N stage, and radiation facility location. The intervention consisted of standardized discharge documentation implemented in May 2023. ORs >1 indicate increased odds of delayed PORT; ORs <1 indicate decreased odds of delay (protective effect). All patients had pathologic indication for adjuvant radiation therapy and received treatment between January 2018 and December 2023.

Health insurance status and postoperative medical complications were not significantly associated with PORT delay, though trends were observed for uninsured patients (OR = 12.46, p = 0.051) and those who experienced complications postoperatively (OR = 2.47, p = 0.056). Other variables, including race, marital status, cancer subsite, distance to radiation facility, and receiving radiation at VUMC, were not independently associated with PORT delay.

The multivariate logistic regression model demonstrated good discrimination (AUC = 0.79, 95% CI: 0.74 – 0.85) and excellent calibration (Hosmer-Lemeshow χ² = 1.32, df = 8, p = 0.995). No significant multicollinearity was detected among predictors (all VIFs < 5, range: 1. 00–3.47).

Multivariable Cox Regression

In the adjusted multivariate Cox proportional hazards model (n = 256) (Table 3), patients treated in the post-intervention period experienced significantly faster PORT initiation compared to the pre-intervention group (HR = 2.05, 95% CI: 1.43–2.93, p = 7.70 × 10⁻⁵). Uninsured patients (HR = 0.46, 95% CI: 0.22–0.96, p = 0.04) and those with Veterans Affairs insurance (HR = 0.22, 95% CI: 0.05–0.87, p = 0.03) had longer delays to PORT initiation, whereas no significant initiation delays were observed for patients with Medicaid or private insurance. Additionally, patients receiving PORT only (vs. concurrent chemoradiotherapy) (HR = 0.65, 95% CI: 0.47–0.90, p = 0.01) was associated with delayed PORT initiation. Postoperative medical complications also lengthened the time to PORT start (HR = 0.51, 95% CI: 0.33–0.79, p = 0.002).

Table 3.

Cox Proportional Hazards Regression Analysis of Factors Associated with Time to Postoperative Radiation Therapy Initiation.

Variable	Multivariable Hazard Ratio (95% CI)	p-value
Intervention (Ref = Pre-Intervention)
Post-Intervention	2.05 (1.43 – 2.93)	<0.001*
Health Insurance Status (Ref = Medicare)
Medicaid	2.87 (0.61 – 13.52)	0.18
Private	0.92 (0.68 – 1.26)	0.60
Uninsured	0.46 (0.22– 0.96)	0.04*
VA	0.22 (0.05 – 0.87)	0.03*
Marital Status (Ref= Single)
Married/Current -Partner	1.11 (0.34 – 3.56)	0.87
Separated/Divorced	0.86 (0.25 – 2.91)	0.80
Widowed	1.20 (0.36 – 4.02)	0.77
Race (Ref = White)
Non-White	0.96 (0.60 – 1.54)	0.86
Cancer Subsite (Ref = Oral Cavity)
Larynx/Hypopharynx	0.94 (0.57 – 1.55)	0.81
Oropharynx	1.06 (0.61 – 1.84)	0.85
Nasal Cavity/Paranasal Sinuses	1.09 (0.55 – 2.16)	0.81
Salivary	1.32 (0.76 – 2.30)	0.33
Skin/Scalp	1.26 (0.78 – 2.04)	0.34
Unknown primary	1.79 (0.66 – 4.89)	0.26
Radiation with or without CRT (Ref = CRT)
Radiation alone	0.65 (0.47 – 0.90)	0.01*
Dental Evaluation Prior to Radiation (Ref = No)
Yes	0.73 (0.48 – 1.10)	0.13
Distance to Radiation Facility (miles)	0.9995 (0.998 – 1.001)	0.59
Postoperative Complications (Ref = No)
Yes	0.51 (0.33 – 0.79)	0.002*
Radiation at VUMC (Ref = Yes)
No	0.84 (0.60 – 1.18)	0.31
Pathologic Tumor Stage‡ (Ref = Stage 1)
2	-	-
3	-	-
4	-	-
4a	-	-
4b	-	-
x	-	-
Pathologic Nodal Stage (Ref = Stage 0)
1	0.91 (0.57 – 1.46)	0.70
2	0.48 (0.12 – 1.96)	0.30
2a	0.98 (0.54 – 1.78)	0.94
2b	0.85 (0.50 – 1.46)	0.57
2c	1.90 (0.69 – 5.25)	0.22
3	1.14 (0.22 – 5.94)	0.88
3a	1.35 (0.11 – 16.70)	0.81
3b	0.82 (0.53 – 1.27)	0.37
x	1.09 (0.62 – 1.94)	0.76

Hazard ratios (HR) and 95% confidence intervals (CI) from Cox regression models evaluating predictors of time to postoperative radiation therapy (PORT) initiation following curative-intent surgery for head and neck cancer (n=256). Univariate analyses examined each variable separately. The multivariable model was stratified by pathologic T stage and adjusted for intervention period, health insurance status, marital status, race, cancer subsite, treatment modality, dental evaluation, distance to radiation facility, postoperative complications, pathologic N stage, and radiation facility location. The intervention consisted of standardized discharge documentation implemented in May 2023. HRs >1 indicate shorter time to PORT initiation; HRs <1 indicate longer delays. All patients had pathologic indication for adjuvant radiation therapy and received treatment between January 2018 and December 2023.

In contrast, race, marital status, cancer subsite, dental evaluation, pathologic nodal staging, radiation facility, and distance to the radiation facility did not show statistically significant associations with time to PORT initiation in the adjusted Cox proportional hazards model (all p > 0.05).

The model demonstrated acceptable discrimination (C-index = 0.66, 95% CI: 0.62–0.69) and was highly significant overall (likelihood ratio test: χ² = 64.09, df = 30, p < 0.001). After stratification by pathologic T stage to address non-proportional hazards, the proportional hazards assumption was verified using Schoenfeld residuals (global test: χ² = 42.49, df = 30, p = 0.065), confirming no significant violations. Multicollinearity assessment using generalized variance inflation factors confirmed independent predictor variables (all GVIF < 1.3). Internal validation via bootstrap resampling (200 iterations) demonstrated stable model performance.

Discussion

Timely PORT initiation after head and neck cancer surgery remains challenging despite national quality metrics targeting treatment within 6 weeks.[3] In this single-institution cohort (n = 391), an EHR discharge summary prompt was associated with a marked reduction in delayed PORT initiation (> 42 days) from 74.2% to 46.6% (absolute reduction, 27.6%; adjusted OR 0.28, 95% CI: 0.14 – 0.56, p = 0.0004). Time to PORT similarly improved (median 52 to 42 days; log-rank p < 0.001), with faster initiation in adjusted time-to-event analysis (HR 2.05, 95% CI: 1.43–2.93, p < 0.001). Although baseline characteristics differed between periods (including radiation facility and adjuvant treatment modality), the association between the intervention period and timeliness persisted after multivariable adjustment. These findings suggest that a standardized discharge handoff may be a practical tool to accelerate the transition from surgery to adjuvant radiation

Prior work suggests PORT delay is multifactorial, reflecting patient-level barriers (including social determinants of health), as well as system-level failures in care coordination, communication, and fragmentation of surgical and radiation oncology workflows.[4,8,9,11–14,16,17] Multi-level interventions have improved timeliness, often by adding patient navigators and redesigning referral and tracking processes. Graboyes et al. reported improved care coordination and high guideline-adherent PORT initiation (86% overall) using the NDURE model, which incorporates patient education, travel support, PORT care plans, and referral tracking and follow-up.[2] Similarly, Divi et al. implemented a multidisciplinary workflow intervention targeting dental and radiation oncology workflows, decreasing delayed PORT initiation from 38% to 27%.[10] While effective, these approaches can require substantial staffing and process redesign that may be difficult to replicate in smaller academic or community hospital settings. Notably, even less resource-intensive efforts, such as an additional pre-operative visit to address treatment-related questions, have been associated with reduced odds of delay (adjusted OR ~0.30, p = .026).[15] In this context, our findings extend prior work by demonstrating that a discharge-based, EHR-embedded prompt requiring documentation of adjuvant RT need and referral actions was associated with a substantial improvement in timeliness (delayed PORT > 42 days decreased from 74.2% to 46.6%), while also highlighting bottlenecks that may require more targeted interventions.

The mechanism most consistently implicated in PORT delay is breakdowns in care coordination during the transition from surgery to adjuvant therapy. In recent prospective work, poor care coordination was the primary reason for delay in 41.3% of patients who did not achieve timely PORT, and perioperative adverse effects markedly reduced the likelihood of on-time initiation.[18] Our findings align with this framework: postoperative complications were independently associated with slower PORT initiation in adjusted time-to-event analysis (aHR 0.51, p=0.002) and trended toward increased odds of delay in the adjusted logistic regression model (aOR 2.47, p=0.056), suggesting that clinical recovery and disruptions to referral and scheduling workflows remain key bottlenecks even when discharge documentation is standardized. Dental clearance is another persistent barrier. Prior studies demonstrate that delays in dental referral and time from referral to clearance are strongly linked to radiation delays, particularly among patients requiring extractions and those with Medicaid or no insurance.[8,10,19] Although dental evaluation and extraction variables were associated with delay in unadjusted analyses in our cohort, they were not independently associated with time to PORT in the adjusted Cox model, suggesting that dental processes may be correlated with broader factors (e.g., postoperative course, insurance barriers, and downstream scheduling) rather than functioning as an isolated cause of delay. Beyond these patient- and clinical-level factors, disease complexity and treatment course may also influence timeliness. In our adjusted logistic model, higher pathologic stage (T2/T4a and N3b) and receipt of radiation alone (vs CRT) were independently associated with delay, which may reflect more complex postoperative recovery and less tightly coordinated multidisciplinary scheduling. Finally, system-level fragmentation has been repeatedly associated with delayed adjuvant therapy, including structural barriers such as insurance insecurity and neighborhood-level disadvantage.[14,20] In our cohort, insurance type remained associated with timeliness in the adjusted Cox model, and uninsured status showed a borderline association with delay in the adjusted logistic model (aOR 12.45, p=0.051), reinforcing that documentation-based interventions may need to be paired with targeted supports for high-risk patients. Nonetheless, our intervention likely functions by standardizing documentation and communication at a known failure point in care transitions: hospital discharge. By prompting clinicians to explicitly document adjuvant RT need and referral actions, it may improve relay of crucial clinical information to downstream teams, consistent with intervention approaches emphasizing reminders and improved clinical data transfer.[21] Together, these findings link a pragmatic, EHR-based intervention with timely PORT initiation toward the 42-day benchmark while identifying persistent barriers that likely require additional targeted solutions.

The findings of this study should be interpreted in the context of its limitations. First, the pre/post design limits causal inference, and any other time-related changes such as staffing, scheduling capacity, or referral practices may have influenced PORT timeliness. However, the improvement in PORT timeliness was observed across multiple analytic approaches (delay >42 days and time-to-event analyses) and remained significant after multivariable adjustment. Second, some baseline characteristics differed between cohorts (including radiation facility and postoperative treatment modality), raising the possibility of confounding. To address this, we incorporated clinically relevant covariates into adjusted logistic and Cox models and assessed model performance and assumptions (logistic AUC 0.79; Hosmer–Lemeshow p = 0.995; Cox proportional hazards verified after pathologic T-stage stratification with Schoenfeld global p = 0.065). Third, the multivariable Cox model relied on complete-case analysis (n = 256) and missing covariate data could introduce selection bias. However, the primary outcome analysis included the full cohort and showed a similarly large, directionally consistent improvement, making it less likely that the findings are explained solely by the reduced Cox sample. Finally, we did not directly measure completion of the discharge prompt or downstream referral completion, limiting our ability to determine which components drove improvement. Nevertheless, the intervention was intentionally simple and embedded in routine discharge workflow, supporting its real-world feasibility.

Improving the timeliness of PORT is a clear opportunity to raise adherence to an established quality metric, and our findings suggest that an EHR-based discharge prompt can shorten the postoperative interval to adjuvant radiation treatment. For patients, a more reliable transition from surgery to adjuvant therapy may reduce uncertainty after discharge and potentially improve oncologic outcomes. For clinicians, this work supports standardizing the discharge care-transition as a key step in the PORT workflow. The discharge prompt could further be converted into an automated order set that incorporates radiation oncology and dental referrals, and communicates confirmation when consults are scheduled. Furthermore, improved PORT timeliness aligns with national benchmarks and may avert downstream costs associated with recurrence and treatment intensification, supporting adoption of a scalable EHR workflow.[22–24] Moving forward, combining this low-resource intervention with targeted solutions for higher-risk patients (e.g., those with postoperative complications, insurance insecurity, or advanced pathologic stage) offers a practical pathway to further reduce residual delays and improve equitable access to timely adjuvant therapy.

Conclusion

A simple EHR-based discharge summary prompt was associated with a substantial reduction in delayed PORT after head and neck cancer surgery and shifted treatment timing toward the 6-week benchmark. Although PORT delay remains multifactorial, standardizing the discharge handoff represents a practical quality improvement strategy in head and neck cancer care.