Cross-sectional survey of factors influencing treatment adherence in elderly cancer patients in rural mountainous Guangyuan, China

Abstract

BACKGROUND

In the rural mountainous areas of Guangyuan, elderly patients with malignant tumors are poorly managed in terms of toxic side effects after comprehensive antitumor therapy, severely affecting their treatment adherence.

AIM

To explore the factors influencing treatment adherence in elderly cancer patients in rural mountainous Guangyuan and clarify the relative importance of each factor.

METHODS

This multicenter cross-sectional survey assessed 103 elderly patients with malignant tumors in rural mountainous areas of Guangyuan who completed core scales. Research tools included the Supportive Care Needs Survey Short Form, the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events, the Comprehensive Score for financial Toxicity, psychological status scales (Hospital Anxiety and Depression Scale/Distress Thermometer/Social Support Rating Scale), and the Medication Adherence Scale. Descriptive statistics, correlation analysis, multiple linear regression, and relative weight analysis were used for data processing.

RESULTS

The average Medication Adherence Scale score was 4.12 ± 1.85 points, with 53.9% of patients having poor treatment adherence. Univariate and correlation analyses showed that supportive care needs, symptom severity, financial toxicity, anxiety/depression, and lack of social support were all significantly correlated with treatment adherence (all P < 0.05). On multivariate analysis, financial toxicity was the primary associated factor (relative weight 28.40%), with direct medical costs having the strongest negative independent association (β = -0.312, P = 0.001). Symptom burden, especially fatigue (β = -0.254, P = 0.002), and psychological status (β = 0.234, P = 0.005) were the second and third most influential factors, respectively, whereas social support was a protective factor. Supportive care needs also had an independent influence. The comprehensive model explained 51.8% of the variance in treatment adherence (R² = 0.518, P < 0.001).

CONCLUSION

The treatment adherence of elderly patients with malignant tumors in the rural mountainous areas of Guangyuan is jointly influenced by multidimensional factors. Financial toxicity had the strongest association with treatment adherence, while symptom management and psychological support serve as key supporting elements. This study provides a clear direction for the development of precise medical management and policies for this special population and the construction of a comprehensive intervention model.

Key Words: Elderly cancer patients; Treatment adherence; Symptom burden; Cross-sectional survey; Supportive care needs; Psychological factors

Core Tip: This study, through a cross-sectional survey, found that the treatment adherence of elderly cancer patients in rural mountainous areas of Guangyuan was generally poor. Economic toxicity was the strongest factor associated with adherence, with direct medical costs having the greatest impact, followed by fatigue symptoms. Social support served as a significant protective factor. The findings provide a basis for developing targeted intervention strategies for patients in this region.

INTRODUCTION

In light of the aging population of China, malignant tumors have become one of the major chronic diseases that seriously threaten the health of the elderly[1]. This can be additionally difficult for those in certain areas such as Guangyuan city, located in the Qinba mountainous area[2], due to its remote geographical location, inconvenient transportation, and uneven distribution of medical resources. Elderly patients with malignant tumors residing in these rural areas face many challenges during comprehensive antitumor treatment, especially in terms of managing toxic side effects, treatment adherence, psychological support, and economic burden[3]. They often experience interrupted treatment or have poor adherence due to the insufficient capacity of primary medical institutions to manage toxic side effects and significant economic pressure, thus severely affecting their quality of life[4]. Therefore, constructing a comprehensive management model suitable for such patients residing in mountainous rural areas is of great practical significance and holds immense policy-making value for improving their treatment adherence and quality of life. Unfortunately, there is a scarcity of systematic research on this special group of elderly patients with malignant tumors in mountainous rural areas. Existing studies indicate that treatment adherence is influenced by multiple factors such as symptom burden, psychological status, financial toxicity, and supportive care needs[5]. However, the interaction mechanisms of these factors among elderly patients in mountainous rural areas remain unclear, and there is a lack of multidimensional comparative analysis. This study aims to fill this knowledge gap through a cross-sectional survey and multifactor analysis.

Financial toxicity is one of the key factors that affect treatment adherence, especially in economically underdeveloped areas[6]. Symptom burden (e.g., fatigue, nausea, and pain) not only directly affects patients’ quality of life but also influences their willingness to cooperate with treatment[7]. Our preliminary investigations revealed that elderly tumor patients in the rural mountainous areas of Guangyuan have poor treatment adherence and high unmet supportive care needs, especially in terms of health information, psychological support, physiology, and daily living. Additionally, factors such as financial toxicity (e.g., direct medical cost burden), and symptom burden (e.g., fatigue, nausea) significantly affect patients’ willingness and ability to continue treatment. Therefore, systematically analyzing the multidimensional associated factors of treatment adherence can greatly help in the development of targeted comprehensive intervention strategies. In patients with tumors, treatment adherence depends on the combined action of multiple factors[8]. Psychological status and social support levels have a significant moderating effect on patients’ treatment attitudes and behaviors[9,10], and the degree to which supportive care needs are met is closely related to treatment adherence[11].

Through a cross-sectional survey, this study aims to systematically analyze the multidimensional associated factors of treatment adherence in elderly patients with malignant tumors in rural mountainous areas of Guangyuan. These factors include symptom burden, psychological status, financial toxicity, and supportive care needs. This study seeks to clarify the independent effects and relative importance of each factor, as well as construct an optimized comprehensive management strategy tailored to this group. This will provide a basis for primary medical institutions to develop targeted intervention measures and offer empiric evidence for local governments and health administrative departments to develop cancer prevention and treatment policies for the elderly in mountainous areas.

MATERIALS AND METHODS

Study subjects

This study targeted elderly patients with malignant tumors residing in the rural mountainous areas of Guangyuan city.

Inclusion and exclusion criteria: The inclusion criteria were as follows: (1) Age ≥ 55 years; (2) Diagnosed with a malignant tumor via pathology or imaging; (3) Currently receiving or planning to receive comprehensive antitumor therapy (e.g., chemotherapy, targeted therapy, immunotherapy, or radiotherapy); (4) Permanent residence in rural mountainous areas of Guangyuan city; and (5) Clear consciousness, basic communication skills, and provided signed informed consent. The exclusion criteria were as follows: (1) Severe cognitive impairment or mental illness, resulting in inability to complete the questionnaire survey; (2) Terminal-stage disease, with an expected survival period < 3 months; and (3) Simultaneous participation in other interventional clinical trials that may interfere with the results of this study.

Sample size calculation and sampling method: The following sample size estimation formula for the cross-sectional survey was used: n = Z² × P(1-P)/d², where n is the sample size, Z is the Z-value of the standard normal distribution (Z = 1.96 when α = 0.05), P is the estimated proportion of poor treatment adherence, and d is the permissible error.

Setting α = 0.05 (Z = 1.96), permissible error d = 0.1, and estimated proportion of poor treatment adherence P = 0.50 (to obtain the maximum sample size), the minimum sample size was calculated to be 96 cases. Considering an estimated 20% non-response rate, the final target sample size was determined to be 100-120 cases. The final sample of 103 cases meets this requirement. The final regression model included 9 independent variables, yielding an events-per-variable ratio of approximately 11.4, which exceeds the generally recommended minimum threshold of 10. A multicenter, convenience sampling method was used. From June 2024 to June 2025, eligible subjects were consecutively recruited from the Guangyuan Hospital of Traditional Chinese Medicine and multiple cooperating township health centers.

Study population and sample characteristics: Initially, 130 patients were recruited. After excluding cases with missing key information and those who did not complete the core scales, the core analysis sample included 103 patients who completed all core scales [Comprehensive Score for financial Toxicity (COST), Supportive Care Needs Survey Short Form (SCNS-SF34), 8-item Morisky Medication Adherence Scale (MMAS-8)]. The study population included 58 males (56.3%) and 45 females (43.7%) with an average age of 68.2 ± 7.8 years. The main treatment method was chemotherapy (n = 65, 63.1%). The sample was generally consistent with the regional tumor epidemiological characteristics in terms of gender and age distribution, indicating good representativeness. The participant screening and enrollment process is illustrated in Figure 1.

Figure 1 Of 130 eligible participants initially assessed, 5 refused participation, and 125 provided informed consent. A total of 103 participants completed the core scales [Comprehensive Score for Financial Toxicity (COST), Supportive Care Needs Survey-Short Form 34 (SCNS-SF34), and 8item Morisky Medication Adherence Scale (MMAS-8)] and constituted the final core analysis sample; the remaining 22 did not complete these scales for various reasons. Within this core sample, instrument-specific completion rates varied: SCNS-SF34 was completed by 106 participants (including 3 who did not complete COST), COST by 103, MMAS-8 by 102 (1 missing), PRO-CTCAE by 71-81 depending on the symptom, and Hospital Anxiety and Depression Scale/Distress Thermometer/Social Support Rating Scale by 49 participants. Analyses of psychological status (Hospital Anxiety and Depression Scale/Distress Thermometer/Social Support Rating Scale) were exploratory and limited to these 49 participants. Note: Sample sizes differ across additional instruments due to variable completion rates. SCNS-SF34 and COST were the core predictor scales; MMAS-8 was the primary outcome. COST: Comprehensive Score for Financial Toxicity; SCNS-SF34: Supportive Care Needs Survey-Short Form 34; MMAS-8: 8-item Morisky Medication Adherence Scale; PRO-CTCAE: Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events; DT: Distress Thermometer; HADS: Hospital Anxiety and Depression Scale; SSRS: Social Support Rating Scale.

Survey design

This study adopted a multicenter cross-sectional survey design. To control information bias, all data collectors followed uniform survey standards, and standardized instructions were used during the survey process. Questionnaires were completed by the participants themselves. For those with reading difficulties, investigators asked questions item by item and marked the answers on their behalf to ensure that patients fully understood the questions.

Ethics approval and consent to participate

This study was conducted following the ethical principles of the Declaration of Helsinki. The study protocol was reviewed and approved by the Medical Ethics Committee of Guangyuan Hospital of Traditional Chinese Medicine, No. 2024028. Written informed consent was obtained from all study participants.

Research instruments

For data collection, this study used several standardized scales which were translated into Chinese and demonstrated good reliability and validity. The SCNS-SF34[11] covers five dimensions of health information, psychological, physical, and daily living, patient care and support, and sexuality. Answers are scored on a 4-point Likert scale (1 = no need, 4 = high need), with higher total scores indicating higher unmet supportive care needs. The Cronbach’s α coefficient for this scale in this study was 0.926.

The Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE)[12] is used to assess the frequency, severity, and interference with daily life of common toxic side effects. This study focused on 12 core symptoms, including nausea, vomiting, diarrhea, constipation, and fatigue, each graded 1-5 according to National Cancer Institute standards, with higher grades indicating more severe symptoms.

The COST[13] contains 10 items, including direct medical costs, indirect medical costs, medication costs, and treatment interruption situations. The total score is calculated according to the preset scoring criteria. In this study, the COST items were reverse-coded such that higher scores indicate more severe financial toxicity. The total score ranges from 0 to 44. To facilitate clinical interpretation, the COST scores were categorized based on the score distribution observed in this study as either low (0-10 points), moderate (11-20 points), high (21-30 points), and severe toxicity (> 30 points)[13,14].

Various dimensions of psychological status were assessed using the Hospital Anxiety and Depression Scale (HADS), Distress Thermometer (DT), and Social Support Rating Scale (SSRS). The HADS[15] includes two subscales of anxiety (HADS-A) and depression (HADS-D), each including 7 items. Scores are interpreted as having no symptoms (0-7 points), possible presence (8-10 points), and definite presence (11-21 points). The Cronbach’s α coefficient for the total scale was 0.881. In the DT[16], patients rate the degree of impact of specific life events in the past, using a 1-5 scale. The SSRS[17] includes three dimensions of objective support, subjective support, and utilization of social support, wherein higher total scores indicate higher levels of social support.

The MMAS-8[18] serves as the main dependent variable of this study. The total score of 0-8 is interpreted as either poor adherence (< 6 points), medium adherence (6-7.9 points), or good adherence (8 points). The Cronbach’s α coefficient for this scale in this study was 0.802. All patients in this study were receiving oral antitumor medications (e.g., oral chemotherapeutic agents, targeted therapies, and immunotherapies) as part of a comprehensive treatment regimen. Medication adherence as measured by the MMAS-8 serves as a proxy for overall treatment adherence in this context.

Data collection and quality control

Data was collected through questionnaire surveys, structured interviews, and a review of medical records. Before the survey, all investigators underwent unified training to ensure consistency in survey procedures and instructions. After questionnaire collection, quality control staff checked the completeness and logic of each questionnaire, excluding incomplete survey reports. The following survey data was collected: 106 SCNS-SF34 forms, 103 COST forms, 102 MMAS-8 forms, 71-81 symptom assessment forms (varies by symptom), and 49 psychological status scale forms. Since the HADS/DT/SSRS psychological scale had only 49 valid answers, we conducted an exploratory and complete case analysis only. Due to the lack of non-random data, no interpolation processing was performed.

Variable definition and measurement

The dependent variable was treatment adherence, measured using the MMAS-8 total score (0-8 points). For logistic regression analysis, this was transformed into a three-category ordinal variable (poor, medium, and good) and a binary variable (poor and good/medium). The independent variables were as follows: Supportive care needs (continuous variable, represented by SCNS-SF34 total score and dimension scores), symptom burden (ordinal multi-categorical variable, represented by the severity level[1-5] of each symptom in the PRO-CTCAE), financial toxicity (continuous variable, represented by the COST total score), and psychological status (continuous variable, that incorporates the HADS anxiety/depression scores, DT life event impact score, and SSRS social support total score). Demographic and disease characteristics were analyzed as categorical variables.

Statistical analysis

Data analysis was performed using the SPSS 26.0 and R 4.2.0 software. All statistical tests were two-sided, with a significance level of α = 0.05. For the descriptive analysis, normally distributed data were described using the mean ± SD, whereas non-normally distributed data were described using the median (interquartile range). Count data were described using the frequency (percentage). The univariate analysis involved t-test, analysis of variance, Mann-Whitney U test, or Kruskal-Wallis H test. The relationship between categorical variables and adherence levels was analyzed using the χ² test or Fisher’s exact test. For the correlation analysis, Pearson correlation or Spearman rank correlation analysis was used to determine the relationship between continuous/ordinal variables and treatment adherence. Multiple linear regression used the MMAS-8 total score (continuous variable) as the dependent variable and incorporated significant variables on univariate analysis (P < 0.1) to analyze the independent associations of each factor. Stepwise regression was used for variable selection.

Ordinal logistic regression was used as a sensitivity analysis to confirm the consistency of the findings (data not shown). Additionally, an exploratory mediation analysis was conducted using the subsample of 49 patients who completed the psychological scales. The significance of indirect effects was tested using the bootstrap method with 5000 resamples. Due to the limited sample size, this analysis is considered exploratory and its findings should be interpreted with caution.

RESULTS

Descriptive characteristics of the study population

This study included 103 elderly patients with malignant tumors from the rural mountainous areas of Guangyuan; they mainly received chemotherapy and commonly experienced a multiple-symptom burden with fatigue at its core. The gender and age distributions of the study population were consistent with regional tumor epidemiological characteristics and covered various treatment methods at different treatment stages, demonstrating good representativeness. As shown in Tables 1 and 2, the sample exhibited distinct group characteristics in terms of demographics, disease and treatment features, and symptom burden.

Table 1 Distribution of demographic characteristics, n (%).

Characteristics	n = 103
Median age, years, mean ± SD	68.2 ± 7.8
Age
60-69	54 (52.4)
70-79	44 (42.7)
≥ 80	5 (4.9)
Gender
Male	58 (56.3)
Female	45 (43.7)
Treatment modality
Chemotherapy	65 (63.1)
Targeted therapy	22 (21.4)
Immunotherapy	11 (10.7)
Radiotherapy	5 (4.9)

Table 2 Patient symptom burden characteristics.

Symptom category	Symptom	Grade 1 (n)	Grade 2 (n)	Grade 3 (n)	Grade 4 (n)	Total (n)	Moderate-to-severe symptom proportion (%)
Gastrointestinal symptoms	Nausea	38	19	7	2	66	13.6
	Vomiting	44	14	3	1	62	6.5
	Diarrhea	49	5	2	3	59	8.5
	Constipation	33	25	4	0	62	6.5
Systemic symptoms	Fatigue	27	18	17	6	71	32.4
	Fever	49	10	0	1	60	1.7
Skin/mucosal symptoms	Oral mucositis	41	6	3	3	53	11.3
	Hand-foot syndrome	39	11	4	0	54	7.4
Neurological symptoms	Peripheral neuropathy	39	10	6	1	56	12.5
	Headache	28	16	6	3	55	16.4

Demographic characteristics: As shown in Table 1, the study population (n = 103) included 58 males (56.3%) and 45 females (43.7%), showing a gender distribution consistent with that of the regional tumor incidence. The age range was 60-85 years, with an average age of 68.2 ± 7.8 years. Specifically, 54 (52.4%), 44 (42.7%), and 5 (4.9%) patients belonged to the age groups of 60-69, 70-79, and ≥ 80 years, respectively.

Disease and treatment characteristics: Regarding treatment modality (Table 1), most patients (n = 65, 63.1%) received chemotherapy as the main antitumor regimen. The next most frequent modalities were targeted therapy (n = 22, 21.4%) and immunotherapy (n = 11 cases, 10.7%), whereas radiotherapy was the least common (n = 5, 4.9%).

Symptom burden characteristics: During the assessment of patient-reported symptoms, the study population generally experienced a heavy symptom burden (Table 2). Out of 71 patients reporting fatigue, 32.4% had moderate-to-severe (grade 3-5) fatigue. The proportions of moderate-to-severe headache (16.4%), nausea (13.6%), peripheral neuropathy (12.5%), and oral mucositis (11.3%) all exceeded 10%. Gastrointestinal symptoms (nausea, vomiting, diarrhea, constipation) were highly prevalent toxic side effects of antitumor therapy, but the proportion of moderate-to-severe cases was relatively low (6.5%-13.6%).

Current status of treatment adherence

Out of 102 patients who answered the Chinese version of the MMAS-8, there was substantially poor treatment adherence. The average MMAS-8 score was 4.12 ± 1.85 points, which is far below the threshold of 6 points for medium adherence. As shown in Table 3, more than half (n = 55, 53.9%) had poor adherence, while 35 patients (34.3%) had medium adherence. Only 12 patients (11.8%) could fully comply with medical advice and had good adherence. This clearly indicates the highly prevalent and serious problem of poor medication adherence in this study population, which urgently requires comprehensive intervention.

Table 3 Classification of patient treatment adherence levels.

Adherence level	Score range	n (%)	Cumulative percentage (%)	Evaluation result
Good adherence	8 points	12 (11.80)	11.80	Fully compliant
Medium adherence	6-7.9 points	35 (34.30)	46.10	Partially compliant, needs intervention
Poor adherence	< 6 points	55 (53.90)	100.00	Severely non-compliant, urgent intervention needed
Total	0-8 points	102 (100.00)	100.00	Average score: 4.12 ± 1.85 points

Specific manifestations of treatment status: Up to 53.9% of patients found it difficult to adhere to the long-term treatment plan. Simultaneously, 35.3% of patients would self-discontinue medication upon feeling improvement of their symptoms, whereas 21.6% would stop medication if they felt worsening of their symptoms. These findings suggest that many patients lack scientific understanding of the continuity, standardization, and necessity of treatment and are prone to arbitrarily interrupt treatment based on subjective feelings. Regarding memory problems, 39.2% of patients had issues remembering to take their medication, while 42.2% forgot to carry medication when going out. This may be closely related to the declining cognitive function among the elderly and the complexity of their medication regimens. Meanwhile, 31.4% of patients admitted to stopping medication due to excessive drug side effects in the past two weeks, indicating that the symptom burden borne by patients is also a major cause of treatment interruption. These findings are shown in Supplementary Table 1.

Correlation analysis between supportive care needs and treatment adherence

Correlation and multiple regression analyses were conducted between the scores of each dimension of the SCNS-SF34 and the total score of the MMAS-8 among 103 patients who completed the SCNS-SF34.

Correlation analysis reveals a significant negative association: As shown in Table 4, the Spearman correlation analysis revealed a significant negative correlation of moderate strength between the total score of the supportive care needs scale and the MMAS-8 total score (r = -0.336, P < 0.001). Higher supportive care needs were associated with lower treatment adherence levels. Physical and daily living needs had the strongest negative correlation with adherence (r = -0.351, P < 0.001), suggesting the importance of factoring in the decline in daily living ability associated with symptoms such as fatigue and pain. Health information needs (r = -0.324, P = 0.001) and care and support needs (r = -0.298, P = 0.003) were also significantly negatively correlated with adherence, suggesting that treatment adherence is influenced by patients’ insufficient understanding of their disease, treatment plans, and cost details as well as information about medical processes and resource access. Psychological needs (r = -0.287, P = 0.004) also showed a significant negative association. Emotional issues, such as anxiety, depression, and fear of disease recurrence and death, were significantly associated with treatment willingness and motivation. No statistically significant association was noted between sexual needs and adherence (r = -0.158, P = 0.116). This could be because the surveyed population pays relatively low attention to this issue, or the prevailing sociocultural norms and attitudes make them less likely to disclose such concerns.

Table 4 Correlation analysis between Supportive Care Needs Survey-Short Form 34 dimension scores and 8-item Morisky Medication Adherence Scale total score.

Need dimension	Correlation coefficient (r) with MMAS-8 total	P value	Correlation strength	Correlation evaluation
Health information needs	-0.324	0.001	Moderate negative	Higher needs, lower adherence
Psychological needs	-0.287	0.004	Moderate negative	Higher needs, lower adherence
Physical and daily living needs	-0.351	< 0.001	Moderate negative	Higher needs, lower adherence
Care and support needs	-0.298	0.003	Moderate negative	Higher needs, lower adherence
Sexuality needs	-0.158	0.116	Weak negative	Not statistically significant
Overall supportive care needs	-0.336	< 0.001	Moderate negative	Higher needs, lower adherence

MMAS-8: 8-item Morisky Medication Adherence Scale.

Multiple linear regression analysis results: As shown in Table 5, the multiple linear regression analysis showed that patients’ physical and daily living needs (β = -0.183, P = 0.016) had the largest absolute value of the standardized regression coefficient, suggesting the importance of symptom management and function maintenance in treatment adherence. Physical discomfort and functional limitations may be directly associated with lower patient treatment adherence. Regarding health information needs (β = -0.162, P = 0.035), patient education and information provision had positive significance in improving adherence. The inability to obtain understandable and sufficient disease and treatment information was associated with non-adherent behaviors. The data regarding psychological needs (β = -0.143, P = 0.049) suggested the importance of psychological support as a necessary component of a comprehensive intervention plan.

Table 5 Multiple linear regression analysis results (dependent variable: 8-item Morisky Medication Adherence Scale total score).

Independent variables	Standardized coefficient (β)	t-value	P value	95%CI
Supportive care needs (n = 103, adjusted R2 = 0.142, F = 4.128, P = 0.001)
Physical and daily living needs	-0.183	-2.456	0.016	-0.942 to -0.100
Health information needs	-0.162	-2.134	0.035	-0.815 to -0.031
Psychological needs	-0.143	-1.987	0.049	-0.768 to -0.002
Care and support needs	-0.102	-1.423	0.158	-0.712 to 0.116
Sexuality needs	-0.052	-0.724	0.471	-0.501 to 0.233
Symptom burden (n = 71-81, adjusted R2 = 0.168, F = 4.782, P < 0.001)
Fatigue severity	-0.254	-3.128	0.002	-0.412 to -0.096
Nausea severity	-0.186	-2.245	0.027	-0.351 to -0.021
Oral mucositis severity	-0.172	-2.078	0.040	-0.337 to -0.007
Vomiting severity	-0.134	-1.618	0.108	-0.298 to 0.030
Diarrhea severity	-0.098	-1.182	0.240	-0.262 to 0.066
Financial toxicity (n = 103, adjusted R2 = 0.243, F = 6.892, P < 0.001)
Direct medical costs	-0.312	-3.456	0.001	-0.498 to -0.126
Indirect medical costs	-0.268	-2.978	0.004	-0.445 to -0.091
Medication costs	-0.234	-2.598	0.011	-0.415 to -0.053
Treatment interruption	-0.198	-2.198	0.030	-0.376 to -0.020
Debt situation	-0.175	-1.943	0.055	-0.354 to 0.004
Psychological status (n = 49, adjusted R2 = 0.268, F = 5.784, P < 0.001)
Social support (SSRS)	0.312	3.102	0.003	0.109-0.515
Anxiety symptoms (HADS-A)	-0.286	-2.845	0.007	-0.489 to -0.083
Depression symptoms (HADS-D)	-0.234	-2.326	0.025	-0.438 to -0.030
Life event impact (DT)	-0.198	-1.968	0.056	-0.402 to 0.006

β is standardized regression coefficient. All models were checked for multicollinearity (all variance inflation factors < 3.0), residual normality (Shapiro-Wilk test P > 0.05), and influential observations (Cook’s distance < 0.5 for all cases). Variable selection used stepwise regression with entry criterion P < 0.10 and removal criterion P > 0.15.

CI: Confidence interval; SSRS: Social Support Rating Scale; HADS-A: Hospital Anxiety and Depression Scale-Anxiety; HADS-D: Hospital Anxiety and Depression Scale-Depression; DT: Distress Thermometer.

Univariate associations between symptom burden and treatment adherence

The patient-reported symptom assessment (PRO-CTCAE) was used to investigate the relationship between the severity of treatment toxicities and treatment adherence, among 71-81 patients who completed the survey (sample size varied by symptom). Symptom burden emerged as another key factor influencing adherence, independent of economic and psychological factors.

Univariate association between patient symptoms burden and adherence: This study assessed 12 common symptoms, including fatigue, nausea, and vomiting. The severity of multiple symptoms was significantly correlated with treatment adherence (MMAS-8 score). As shown in Table 6, patients with moderate-to-severe symptoms had significantly lower treatment adherence than those with mild or no symptoms. Among all symptoms, fatigue was the most critical. Across 71 patients with fatigue, the average MMAS-8 score was 3.85 ± 1.78 points, with 32.4% having moderate-to-severe fatigue. Their adherence score was significantly lower compared to those without fatigue (P = 0.008). Moreover, the severities of nausea (P = 0.023), vomiting (P = 0.045), and oral mucositis (P = 0.015) were all negatively correlated with patient treatment adherence.

Table 6 Association analysis between symptom severity and treatment adherence (8-item Morisky Medication Adherence Scale score), mean ± SD.

Symptom type	Sample size	Moderate-to-severe symptom proportion (%)	Average MMAS-8 score of patients with symptom	P value vs patients without symptom
Fatigue	71	32.4	3.85 ± 1.78	0.008
Nausea	66	13.6	4.02 ± 1.82	0.023
Vomiting	62	6.5	4.15 ± 1.79	0.045
Diarrhea	59	8.5	4.28 ± 1.76	0.087
Constipation	62	6.5	4.32 ± 1.74	0.124
Oral mucositis	53	11.3	3.91 ± 1.81	0.015

MMAS-8: 8-item Morisky Medication Adherence Scale.

Independent influence of symptom burden and core symptom identification: On multiple linear regression analysis (Table 5), after controlling for other symptoms, the severities of fatigue (β = -0.254, P = 0.002), nausea (β = -0.186, P = 0.027), and oral mucositis (β = -0.172, P = 0.040) were all independently and negatively associated with treatment adherence. Fatigue had the largest absolute standardized coefficient, indicating that it was the strongest correlate of treatment adherence.

Distribution of medical costs and univariate association with adherence

The COST score was used to analyze the spectrum of economic burden experienced by elderly patients with malignant tumors in rural mountainous areas of Guangyuan (n = 103).

Assessment of patient financial toxicity levels and correlation with treatment adherence: Financial toxicity was quantified using the COST score, a comprehensive scoring method. As shown in Table 7, patients’ total COST scores were widely distributed, indicating significant individual differences in economic burden. After grading based on total score, only 18 patients (17.5%) had a low toxicity level. Meanwhile, 45.6% (n = 47), 28.2% (n = 29), and 8.7% (n = 9) of patients had moderate, high, and severe financial toxicity, respectively. Collectively, 82.5% of patients experienced moderate-to-severe financial toxicity. The impact of financial toxicity on treatment adherence showed a clear dose-response relationship. As the financial toxicity score increased from those with low to severe toxicity, there was a sharp drop in patients’ average MMAS-8 score (5.42 to 2.18, respectively) and a rapid increase in the proportion of those with poor adherence increased rapidly (22.2% to 88.9%, respectively). Thus, economic burden is strongly associated with treatment adherence.

Table 7 Relationship between financial toxicity grade and treatment adherence, mean ± SD.

Financial toxicity score grade	Number of patients	Percentage (%)	Average MMAS-8 score	Proportion with poor adherence (%)
0-10 points (low toxicity)	18	17.5	5.42 ± 1.23	22.2
11-20 points (moderate toxicity)	47	45.6	4.18 ± 1.45	48.9
21-30 points (high toxicity)	29	28.2	3.25 ± 1.68	75.9
> 30 points (severe toxicity)	9	8.7	2.18 ± 1.32	88.9

MMAS-8: 8-item Morisky Medication Adherence Scale.

Multivariate regression analysis indicates financial toxicity as a core factor: This study included various economic burden indicators in a multiple linear regression model (F = 6.892, P < 0.001; Table 5). Direct medical costs (β = -0.312, P = 0.001) had the strongest negative independent association, with an influence exceeding other factors such as fatigue. Indirect medical costs (β = -0.268, P = 0.004) and medication costs (β = -0.234, P = 0.011) also had significant independent negative effects on adherence. This trend of treatment interruption (β = -0.198, P = 0.03) as a consequence of financial toxicity suggests a link between economic pressure, treatment interruption, and poor adherence.

Basic situation of various dimensions of psychological status (exploratory subgroup analysis, n = 49)

Psychological status was assessed across various dimensions using the HADS, DT, and SSRS. Elevated distress levels were significantly correlated with poorer treatment adherence. In this context, social support emerged as an important protective factor.

Patient psychological status assessment results: Only 49 out of 103 patients (47.6% of the core sample) completed the psychological scales (i.e., HADS, DT, and SSRS), and thus this analysis is considered exploratory. For the HADS scale (Table 8), patients had an average anxiety score (HADS-A) of 8.62 ± 3.45 points and depression score (HADS-D) of 7.89 ± 3.28 points. According to the cutoff score standard (≥ 8 points), 42.9% and 38.8% of patients had scores indicating probable or definite anxiety and depressive symptoms, respectively. On the DT scale assessment, the average level of distress caused by life events reported by patients was mild to moderate (2.78 ± 1.12 points), whereas 51% of patients scored at least 3 points (moderate impact or above). Meanwhile, the average total social support score was 32.45 ± 6.78 points, but 24.5% of patients had low social support levels. Upon further analysis, social support demonstrated a significant positive correlation with treatment adherence (MMAS-8).

Table 8 Patient psychological status scale assessment results (n = 49), mean ± SD.

Psychological status dimension	Average score	Abnormal proportion (%)	Score range
Anxiety symptoms (HADS-A)	8.62 ± 3.45	42.9	3-16 points
Depression symptoms (HADS-D)	7.89 ± 3.28	38.8	2-15 points
Life event impact (DT)	2.78 ± 1.12	51	1-5 points
Total social support (SSRS)	32.45 ± 6.78	24.5	18-45 points
Treatment adherence (MMAS-8)	4.23 ± 1.82	53.1	1.5-7.5 points

HADS-A: Hospital Anxiety and Depression Scale-Anxiety; HADS-D: Hospital Anxiety and Depression Scale-Depression; DT: Distress Thermometer; SSRS: Social Support Rating Scale; MMAS-8: 8-item Morisky Medication Adherence Scale.

Multiple linear regression analysis of the relationship between psychological factors and treatment adherence: On multiple linear regression analysis (Table 5), anxiety (β = -0.286, P = 0.007) had the strongest negative association among the psychological variables, with its harm exceeding depressive symptoms. Depression (β = -0.234, P = 0.025) was also independently and negatively associated with adherence. Meanwhile, social support (β = 0.312, P = 0.003) served as a protective factor, and its positive promoting effect was sufficient to partially offset the negative impact of anxiety and depression. In the multiple linear regression model, social support was the only psychological factor that had an independent positive association with treatment adherence (β = 0.312, P = 0.003).

Exploratory analysis of the relationship between psychological factors and treatment adherence: Upon further exploratory mediation analysis, social support was not only associated with adherence but also played a potential contributory role in the pathways from anxiety/depression and life event impact to treatment adherence. Bootstrap method tests (5000 resamples) indicated that the proportion of the mediating effect of social support to the total effect ranged from 29.5% to 30.5%. These exploratory findings suggest an association between psychosocial factors and adherence, although causal pathways cannot be definitively determined due to the small sample size and cross-sectional design.

Multivariate comprehensive analysis

Independently associated factors were identified through multivariate statistical analysis methods, and the relative importance of each factor was compared across 98 patients who completed all core scales and were included in the final regression model.

Results of the multivariate multiple linear regression analysis: The multiple linear regression model used the total MMAS-8 score as the dependent variable and included variables from other dimensions that showed a significant correlation (Table 9). The multiple linear regression model was statistically significant (F = 14.267, P < 0.001) and identified 8 independent associated factors. Among them, direct medical costs and fatigue severity had the strongest negative association, whereas social support was the only protective factor.

Table 9 Relationship between multiple factors and patient treatment adherence in linear regression analysis.

Influencing factor dimension	Specific variable	Standardized coefficient (β)	t-value	P value	Independent contribution rank
Financial toxicity	Direct medical costs	-0.312	-3.456	0.001	1
	Indirect medical costs	-0.268	-2.978	0.004	4
	Medication costs	-0.234	-2.598	0.011	5
Symptom burden	Fatigue severity	-0.254	-3.128	0.002	2
	Nausea severity	-0.186	-2.245	0.027	6
Psychological status	Social support	0.234	2.895	0.005	3
	Anxiety symptoms	-0.158	-2.015	0.047	8
Supportive care needs	Physical and daily living needs	-0.183	-2.456	0.016	7
	Health information needs	-0.162	-2.134	0.035	9

Model statistics: R² = 0.518, adjusted R² = 0.482, F = 14.267, P < 0.001.

Comparison of the relative importance of each factor: This study evaluated the relative importance and incremental explanatory power of each dimension on treatment adherence through relative weight analysis and hierarchical regression (Table 10). Financial toxicity was the most important dimension affecting treatment adherence (relative weight: 28.4%), followed by symptom burden (24.7%) and psychological status (22.5%). After controlling for demographic factors, hierarchical regression showed that financial toxicity contributed the largest incremental explanation (ΔR² = 15.7%), suggesting a strong association with patient adherence.

Table 10 Relative importance and incremental explanatory power of each dimension on treatment adherence.

Influencing factor dimension	Relative weight (%)	Importance rank	Hierarchical regression ΔR2	Cumulative explanatory power (%)
Financial toxicity	28.40	1	0.157	37.00
Symptom burden	24.70	2	0.127	21.30
Psychological status	22.50	3	0.082	45.20
Supportive care needs	13.40	4	0.066	51.80
Demographic and disease characteristics	11.00	5	0.086	8.60

DISCUSSION

Through a cross-sectional survey, this study systematically explored the multidimensional associated factors of treatment adherence in elderly patients with malignant tumors residing in the rural mountainous areas of Guangyuan. This population had a severe problem of treatment adherence that is multifactorial in nature, involving financial toxicity, symptom burden, psychological status, and supportive care needs. However, due to the cross-sectional design of this study, no causal inference can be made. All reported associations can be considered evidence of correlation, which can serve as hypotheses for future research rather than direct evidence of causal relationships.

The multiple regression and relative weight analyses in this study revealed that financial toxicity had the strongest impact on treatment adherence in this population. Unlike previous studies that focused on clinical symptoms or psychological factors, this study comprehensively assessed the comprehensive burden of illness faced by patients. Direct medical costs had the strongest negative predictive power in the comprehensive model (β = -0.312) and the largest relative weight (28.40%), highlighting that economic status is a basic prerequisite associated with patient treatment adherence. Furthermore, the severe symptom burden (especially fatigue), unmet physical and informational needs, and anxiety suggest the interrelatedness of symptoms and psychological needs in lowering patient adherence. Conversely, social support was associated with better patient adherence.

The results of this study have both commonalities and uniqueness compared to domestic and international research in general tumor populations[19,20]. Previous studies in general patient populations have similarly confirmed that symptom burden[21], psychological stress[22], and supportive care needs[23] are key factors in adherence. Our findings are consistent with previous studies which show that financial toxicity is prevalent among Chinese cancer patients and is significantly associated with treatment nonadherence. Our study extends these observations by quantifying the relative importance of financial toxicity (28.4% relative weight) compared to other adherence-related factors, and it is one of the few studies to do this in a population of patients residing in a rural mountainous region in China. However, this claim should be interpreted cautiously, as similar quantitative approaches have been used in other populations.

This profoundly reflects the concrete impact of China’s current macro-background on individual health behaviors, specifically in the context of unevenly distributed healthcare resources and regional economic disparities. For patients in the mountainous rural areas of Guangyuan, factors including their “left-behind” family structure, lower household income, poorer medical security, and longer travel distance for medical care all jointly intensify the economic burden of healthcare on families. Notably, economic factors are more strongly associated with treatment adherence than purely clinical symptoms alone. On the other hand, social support exerted a mediating effect (29.5%-30.5%), providing a basis for explaining how psychological stress transforms into behavioral problems. Thus, primary healthcare professionals need to view treatment adherence as a “clinical sign” that must be incorporated into the comprehensive management of each patient. During consultations, a brief “adherence risk screening” should be conducted to systematically assess the patient’s economic status, core symptoms (e.g., fatigue), psychological stress levels, and social support system.

The limitations of this study are raised herein. Although the sample size of this study meets the general minimum requirements for regression analysis, the possibility of overfitting cannot be completely ruled out. Therefore, the results of this study need further validation in a larger independent cohort. The psychological scales had a low response rate (49/103, 47.6%), potentially introducing selection bias. Patients with greater psychological distress may have been less willing to complete these additional scales as well, potentially underestimating the true prevalence of anxiety and depression. Therefore, these psychological findings should be considered as exploratory. Another important point is that the MMAS-8 primarily measures adherence to oral medications. Even if oral antitumor agents constitute the main treatment modality for our study population, these findings may not be fully generalizable to adherence to intravenous therapies or radiotherapy. Additionally, the reliance on self-reported data may introduce recall and social desirability biases. Finally, the single-region design restricts generalizability beyond Guangyuan. Future multicenter, prospective studies with larger samples and objective adherence measures are warranted to validate these findings.

CONCLUSION

There is a severe problem regarding treatment adherence among elderly patients with malignant tumors residing in the rural mountainous areas of Guangyuan. The factors associated with this issue have clear multidimensionality and hierarchy. Financial toxicity had the strongest association with treatment adherence in this population. Symptom burden and psychologic factors are key secondary factors in adherence, while social support can serve as a protective resource. Addressing the treatment adherence issues of elderly patients with malignant tumors in rural Guangyuan must start with addressing patients’ economic problems while effectively managing symptom burden and psychosocial needs.