Assessing inequalities in health utility scores among cancer patients undergoing systemic and radiation therapy.

Introduction
Cancer is a leading cause of morbidity and mortality worldwide, with a disproportionate and growing impact in low- and middle-income countries (LMICs) such as Bangladesh [1, 2]. The number of cancer cases is increasing rapidly due to shifts in lifestyle, urbanisation, and environmental exposures, while in many LMICs health system capacity has not kept pace [3]. In this setting, health-related quality of life (HRQoL) is essential to evaluate outcomes beyond survival and tumour response [4, 5].
Socioeconomic status (SES), including income, education, and place of residence, shapes avoidable and unjust differences in health outcomes through later diagnosis, constrained treatment choice, reduced continuity of care, and suboptimal management of treatment-related toxicities. Studies using EuroQol instruments have consistently shown that patients from socioeconomically disadvantaged backgrounds report lower health utility scores [6, 7]. Emerging evidence suggests that cancer treatment modality may further modify these disparities, as systemic and radiation therapies have distinct toxicity profiles that can differentially affect function and wellbeing [8–11].
Systemic and radiation therapies are core components of cancer care [12], but are associated with substantial side effects that can reduce HRQoL [10, 11]. For instance, systemic therapy commonly causes generalised toxicities such as fatigue and nausea, and increased infection risk, while radiation therapy can lead to localised complications (e.g. skin reactions, mucositis, organ-specific damage) that impair physical functioning and contribute to psychological distress [11]. In LMICs, where access to high-quality, timely supportive care is uneven, these treatment-related burdens are likely to have a disproportionately greater impact on patients with lower SES. The EQ-5D-5 L instrument is widely used to measure HRQoL, with evidence from high- and low-income settings showing that patients from lower SES groups consistently report lower utility scores during active treatment [13–18]. Translating this evidence into policy requires robust, context-specific and treatment-specific estimates of inequality to guide resource allocation and supportive care strategies. Targeted interventions to reduce financial barriers, improve access to quality oncology and supportive care services, and address sociocultural determinants of care remain difficult to prioritise without local effect-size estimates [19]. Incorporating routine HRQoL assessment into oncology practice can help clinicians individualise treatment decisions, identify high-risk patients earlier, and improve patient-centred outcomes [20].
In Bangladesh, pronounced income inequalities and unequal access to quality cancer services mean many patients experience delays in diagnosis, limited treatment options and inadequate follow-up care [21]. Social and cultural factors (e.g. care-seeking norms, travel and waiting time costs, caregiving productivity losses and health burden), further underscore the need to evaluate outcomes beyond traditional clinical endpoints [9]. However, there is a scarcity of robust, treatment-specific health utility evidence from Bangladesh. Therefore, this study measures socioeconomic inequalities in EQ-5D-5 L health utility scores among Bangladeshi cancer patients receiving systemic or radiation therapy. By comparing utilities by income groups, cancer stage and other key clinical characteristics, we aim to quantify the magnitude of inequality, assess heterogeneity by treatment modality, and explore potential drivers. These estimates provide equity-relevant evidence to inform patient-centred care, targeted supportive interventions and equity-oriented distributional cost-effectiveness analysis in Bangladesh.

Methods

Study design and setting
We conducted a cross-sectional study to evaluate HRQoL in cancer patients undergoing systemic or radiotherapy. Participants were recruited from the outpatient departments of two high-volume cancer treatment facilities in Dhaka, Bangladesh: (i) The National Institute of Cancer Research & Hospital, a 300-bed government tertiary hospital that treated 83,795 new cancer patients from all eight administrative divisions between January 2018 and December 2020, and (ii) Ahsania Mission Cancer Hospital, a privately funded 50-bed hospital serving approximately 100 outpatients per day. Inclusion of both a public and a private centre enabled capture of a broad clinical case-mix and variation in care pathways within an urban referral setting.

Study population and data collection procedures
Eligible participants were adult (≥ 18 years) with a confirmed cancer diagnosis who were actively receiving systemic therapy and/or radiation therapy at the time of survey, able to communicate and provide data related to study questions, and willing to provide informed consent. Before conducting survey, all eligible participants were informed of the study objectives, procedures, risks and benefits. If the participant agreed to continue the survey, a written consent was obtained prior to enrolment. For participants with limited or no formal education, trained interviewers read the information sheet al.oud; an adult caregiver acted as an impartial witness, the patient the patient provided a thumbprint/mark, and the witness co-signed the consent form. Data were collected via structured, face-to-face interviews between January and May 2022 by two junior physicians and three senior medical students. The data collection team received three days of training on the questionnaire, interviewing techniques, ethical conduct, confidentiality, and safety considerations for patients with cancer. A pilot with 24 patients was undertaken to assess clarity and feasibility; minor wording and ordering refinements were made prior to final survey implementation. Each day, all eligible patients visiting the outpatient departments were consecutively approached to minimise selection bias. The required sample size was n = 376, derived from a single-proportion calculation for the prevalence of ‘poor quality of life’ (EQ-5D-5 L utility < 0.70) [22], assuming p = 0.5376 (53.76%) [23], a two-sided 95% confidence level (z = 1.96), 5% absolute margin of error, and 90% power. In practice, we collected 607 cancer survivors via face-to-face interviews (response rate 86%), exceeding the requirement, thereby improving precision of prevalence estimates and supporting subgroup analyses. For robustness, using the standard single-proportion formula with d = 0.05 and p in the 0.45–0.60 range, the required sample size varies from ≈ 380 to ≈ 405. In total, 607 patients completed the survey (response rate ≈ 86%), exceeding the required sample size and thereby improving the precision of estimates and enabling subgroup analyses.

Outcome measures
The primary outcome was health utility scores, measured using the English version of the EQ-5D-5 L instrument [22]. This instrument assesses five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression, and is widely used to HRQoL in oncology populations [4, 16, 17]. These HRQoL assessments include a broad range of physical, emotional, and social dimensions of health that are particularly relevant for cancer patients, who often experience significant functional impairment and psychological distress throughout their treatment journey. These health utility scores support the calculation of composite health outcome measures in health economic evaluations, including quality-adjusted life years (QALYs) [24].
In the resource-limited setting of Bangladesh, such measures are essential for guiding patient-centred care and policy. Understanding the tones captured by these utility scores is essential for tailoring healthcare services [4], including introducing new interventions and policies that aim to improve the overall quality of life of cancer patients. We administered the official Bengali (Bangla) translation of EQ-5D-5 L in accordance with EuroQol’s standardized forward–backward translation and cognitive-debriefing procedures, which aim to ensure linguistic and conceptual equivalence to the source instrument [22, 25]. Although we did not identify a Bangladesh cancer-specific psychometric validation study, EQ-5D-5 L has been used and linguistically validated in adult Bangladeshi populations [26, 27], and EQ-5D-5 L demonstrates construct validity in oncology cohorts in the region [18]. Accordingly, we prespecified internal construct-validity checks within our dataset (known-groups validity by cancer stage, treatment facility, and education; inspection of ceiling/floor effects) to corroborate measurement performance in this sample.
This study considered the complete base case analysis with 607 patients’ data, whereas observations with missing data (< 1%) were excluded from the analysis.

Covariates
We collected a comprehensive set of potential determinants of HRQoL. Several individual-level factors were included, including sociodemographic (gender, age groups [18–35, 36–45, 46–64, and > 64 years], marital status [single, married, and widowed/divorced/separated], education [no schooling to tertiary education], residential status [urban vs. rural], household size [< 4, 4–5, or ≥ 6 members], and monthly household income [quintiles from poorest, Q1, to richest, Q5]) and lifestyle characteristics (smoking, use of smokeless tobacco, and physical activity exceeding 150 min of walking per week). Body mass index (BMI) was categorised as underweight (< 18.49 kg/m2), healthy weight (18.50–24.99 kg/m2), overweight (25–29.99 kg/m2), or obese (≥ 30 kg/m2). Clinical details such as cancer stage, primary cancer site, and type of treatment facility (public vs. private) were also collected to capture the full breadth of variables potentially influencing HRQoL.

Measuring inequalities in assessing health utility scores
To quantify socioeconomic inequality in health utility scores, we calculated relative inequality (rich-poor ratio) and absolute inequality (rich-poor difference). We then applied the Erreygers-bounded concentration index [28], which is designed for bounded outcome measures such as utility scores, by ranking participants according to income quintiles. This index is preferred over the standard concentration index (CI) because it adjusts for the fixed range of the dependent variable, ensuring that values lie within [–1, + 1] and that key properties such as the mirror condition and level independence, are satisfied.
Specifically, for a health outcome (health utility scores) y ∈ [a, b], the Erreygers concentration index is given by:
where is an Erreygers index of y, µ is the mean of y, CI is a standard concentration index of y; a, b are the lower and upper bounds of y.
The factor 4µ is not arbitrary; it arises from the normalisation in the Erreygers formula, which rescales the inequality measure to the range of the outcome. This adjustment is essential when decomposing the index, as it ensures that contributions from individual covariates remain comparable and interpretable across variables with different distributions [29].
We estimated associations between utility scores and covariates using a generalised linear model (GLM) with a gamma distribution and log link, appropriate for positively skewed outcomes such as health utility scores [29]. From each fitted model, we obtained regression model outputs such as coefficients and standard errors, z-statistics, and 95% confidence intervals (CIs). Statistical significance was determined when 95% CI excluded zero (equivalently, |z| >1.96; α = 0.05).
In the inequality decomposition, elasticity quantifies the responsiveness of the outcome to a given factor, controlling for other variables: a positive elasticity indicates that higher values of the factor are associated with higher utility, whereas a negative elasticity indicates the opposite. A factor’s contribution to overall inequality is determined by two components: (i) its elasticity (how strongly the factor shifts utility) and (ii) its concentration across income groups (who tends to have more of the factor). Intuitively, contribution ≈ (responsiveness: how much the factor changes utility) × (who tends to have the factor). Thus, a factor that improves utility (positive elasticity) and is more common among higher-income patients (positive concentration) will increase pro-rich inequality. A factor that reduces (lower) utility (negative elasticity) and is more common among lower-income patients (negative concentration) will also increase pro-rich inequality (a negative × a negative gives a positive pro-rich contribution). For clarity and readability: elasticity = how much the utility score changes when a factor changes; concentration = whether a factor is more common among higher- or lower-income groups; contribution to inequality = the combined effect of responsiveness and distribution.
In decomposing this index, the partial contribution of each determinant is expressed as:
This form (Eq. 2) is algebraically equivalent to:
Thus, the 4µ term is not arbitrary, but a direct consequence of decomposing the Erreygers index to reflect bounded outcomes like health utility. This formulation ensures comparability and interpretability of inequality contributions across different explanatory variables. The 4µ term here follows directly from the Erreygers normalization for bounded outcomes, ensuring that elasticity is correctly scaled when multiplied by the partial concentration index of each covariate. The standard error of elasticity was obtained using the delta method.
We computed elasticity, defined as the percentage change in the dependent variable associated with a 1% change in the independent variable, holding other factors constant as four times the product of the GLM coefficient and the covariate’s sample mean (Eqs. 3 and 4), following a standard approach for decomposing continuous outcomes [30, 31]. The standard error of the elasticity was derived via the delta method. To calculate each covariate’s partial concentration index, we applied ‘conindex’ command in Stata, ranking individuals by income quintile but substituting each covariate in place of the outcome. The resulting concentration index (CI) indicates how that covariate is distributed across the income ranking. Multiplying the elasticity by this partial concentration index yielded the contribution of each covariate to overall inequality in utility scores. We then expressed this contribution as a percentage of the total CI by dividing the covariate-specific contribution by the overall CI and multiplying by 100 [32–34].
In the decomposition framework, the contribution of each variable k to the total CI is given by:
And the percentage contribution is computed as:
However, percentage contributions can exceed ± 100% when the total CI is small, and contributions from individual variables are large or offsetting. The signs of contributions reflect both the direction of association (via the elasticity) and the distributional concentration of the explanatory variable. All analyses were performed using Stata version 15/SE (StataCorp, College Station, TX, USA).

Results
A total of 607 patients undergoing cancer treatment were included in this study. The majority (75%) were in early-stage disease (stages I/II), reflecting a prevalent clinical catchment for newly diagnosed patients. Advanced stages (III/IV) accounted for approximately 22% of the sample, indicating a substantial subset with more progressed disease at the time of treatment. In terms of cancer type, the higher proportion (28%) fell into the “others” category, followed by female organs (20%) and breast (18%), reflecting a broad spectrum of cancer diagnoses. Most participants (88%) received care in a public hospital, highlighting the central role of public healthcare in oncological services. Demographically, slightly more than half (55%) were female, and the predominant age range was 46–64 years (44%), consistent with higher cancer incidence in middle to older adulthood. Notably, 70% of the sample reported engaging in regular physical activity (Supplementary document S1).
Figure 1 illustrates the distribution of severe or extreme health problems across five EQ-5D-5L dimensions by income quintile. Across all dimensions, a higher proportion of low-income patients reported severe or extreme problems compared to their higher-income counterparts. For example, in the mobility dimension, 5.77% of the lowest-income group reported severe or extreme problems, versus 3.8% in the highest-income group. Similar disparities were noted in pain/discomfort (11.54% vs. 7.3%) and anxiety/depression (27.4% vs. 19%), underscoring the disproportionate burden experienced by economically disadvantaged individuals. Table 1 presents the mean utility scores derived from the EQ-5D-5L instrument, revealing pronounced socioeconomic inequalities (Fig. 1).

The wealthiest income quintile consistently had higher utility scores compared to the lowest quintile, resulting in a relative inequality (rich-poor ratio) of 1.10 and an absolute difference of 0.07. The concentration index (CI = 0.025; SE = 0.019) further quantified this socioeconomic gradient. By conventional benchmarks, these represent small, aggregate inequalities; therefore, we interpret them as a modest overall gradient and focus on domain-specific and subgroup patterns where disparities are more pronounced. When stratified by cancer stage, patients with stage-zero disease showed the highest mean utility score (0.76), whereas early-stage (I/II) and advanced-stage (III/IV) groups both averaged 0.67 and 0.66, respectively. Despite similar relative inequalities (1.18) for early and advanced stages, the absolute gap was more pronounced in advanced disease (0.027 vs. 0.017). As visualised in Fig. 2 (concentration curve), suggesting that wealthier patients consistently concentrated higher health utility scores across all cancer stages. Subgroup analyses were exploratory and were interpreted with caution. Several estimates had wide confidence intervals, with some overlapping the null, indicating imprecision and limited power for definitive subgroup inferences. These results were therefore hypothesis-generating and were presented to inform future research rather than to support confirmatory claims (Fig. 2).

Variations in health utility also emerged when stratifying by cancer type. Breast cancer patients recorded relatively high overall utility scores (rich-poor ratio = 1.12; absolute difference = 0.011), whereas lung cancer patients demonstrated lower mean utility scores but a slightly larger income-based disparity (rich-poor ratio = 1.15; absolute difference = 0.012), although this difference did not reach statistical significance. Cancers involving the cervix, vagina, vulva, and ovary presented more pronounced inequalities (rich-poor ratio = 1.62; absolute difference = 0.039), indicating a stronger association between income and quality of life in these groups. Physical activity emerged as another critical correlate: individuals who exercised regularly reported higher mean utility across all quintiles (relative inequality = 1.18; absolute difference = 0.015) compared to inactive patients (relative inequality = 2.03; absolute difference = 0.042). Gender-specific comparisons suggested slightly higher overall scores among females, although the patterns of inequality were largely similar between sexes. Age-stratified analyses showed that middle-aged individuals (36–45 and 46–64 years) had modestly higher utility scores and somewhat larger relative differences than younger and older groups (Table 1).

Table 2 presents the relative risk ratios (RRRs) of experiencing ‘slight/moderate’ or ‘severe/extreme’ problems (versus ‘no problems’) across the five EQ-5D-5L dimensions by income quintile. Low-income groups consistently demonstrated higher risks of having severe/extreme health problems relative to the wealthiest group; however, these association sometimes attenuated when adjusted for age, gender, education, BMI, cancer stage, and treatment facility. Certain dimensions continued to exhibit strong disparities even after adjustment. For example, in the self-care dimension, lower-income individuals had a markedly elevated risk of severe/extreme problems compared to the wealthiest group (RRR = 3.2; 95% CI: 2.55, 68.26), while the RRR for usual activities remained high (RRR = 6.63; 95% CI: 2.35–18.69). These findings confirm that patients in the poorest quintile bear a disproportionate burden of severe or extreme health limitations (Table 2).

Finally, the decomposition analysis in Table 3.1 identifies the clinical and sociodemographic factors influencing inequality in health utility scores. For readability, positive contributions indicate that a factor, given its association with utility and its distribution across income groups, increases pro-rich inequality; negative contributions indicate a pro-poor effect. Earlier cancer stages (particularly stage I) contributed substantially to reducing overall inequality (-22.34%), whereas stage III demonstrated a positive contribution (14.24%), aligning with the notion that advanced disease often portends worse outcomes. Cancer type also played a role, with lung cancer exhibiting a small positive contribution (1.46%) to inequality and oral cavity cancers showing a marginal negative effect (-0.30%). Moreover, treatment in public facilities exerted a large negative contribution (234.56%), emphasising lower utility scores among those receiving care in public hospitals (Table 3.1). In addition, physical activity likewise emerged as a critically important factor, as inactivity substantially contributed to poorer HRQoL (-42.92%), underscoring the potential benefits of exercise promotion in this population. To improve readability, we presented a summary decomposition in Table 3.2 reporting the overall Erreygers concentration index (E-CI) and the largest positive and negative contributors to inequality, accompanied by a plain-language interpretation. A plain-language interpretation of the full decomposition for all covariates, including coefficients, elasticities, standard errors, and confidence intervals, is provided in Supplementary Document S2. Estimates were interpreted as associations rather than effects, consistent with the study design (Table 3.2).

Discussion
This study identified socioeconomic inequalities in health utility scores among 607 cancer patients, modest in aggregate inequalities but more pronounced in specific EQ-5D-5L domains and patient subgroups (e.g., advanced-stage disease, selected cancer types such as cervical and ovarian cancers, and among those treated in public hospital, whereas physical activity was associated with narrower gaps.
These findings align with previous evidence that economic disadvantage strongly correlates with poorer HRQoL outcomes among oncology populations. While the overall effect sizes are small, their population-level relevance may still be meaningful in systems with high cancer incidence; nonetheless, claims about clinical significance should be cautious, and we highlight domain-specific gaps as the primary targets for intervention. Several studies, including those conducted in both high-income and low- to middle-income countries, have underscored that financial constraints often impede access to timely diagnostic services and optimal treatment modalities, culminating in worse patient-reported outcomes [35, 36]. The current study’s observation of pronounced disparities in self-care and usual activities among the lowest-income group resonates with earlier reports suggesting that socioeconomic deprivation may compound functional limitations and impede recovery [35]. Moreover, the high rates of anxiety/depression in economically disadvantaged patients are consistent with established research indicating that financial stressors exacerbate emotional distress and adversely affect cancer coping strategies [37, 38] .
The protective influence of physical activity on HRQoL revealed here mirrors existing literature demonstrating the myriad benefits of regular exercise for oncology patients [39]. In addition to improving functional capacity and reducing fatigue, structured exercise programs have been shown to alleviate distress and enhance self-efficacy among cancer survivors, thereby contributing to better overall quality of life. Nevertheless, this study’s evidence of heightened income-based disparities in physically inactive groups extends prior findings by highlighting the compounding effect of socioeconomic deprivation and sedentary behaviours. Similarly, the observation that certain cancer types, including the cervix, vagina, vulva, and ovary, exhibit more pronounced income disparities builds on earlier work positing that the social determinants of health play a pivotal role in disease incidence, treatment access, and survivorship in gynaecologic oncology [19, 20]. Such insights suggest that targeted screening programs, financial counselling, and navigation services to reduce barriers to care.
To reduce the observed disparities in cancer outcomes associated with socioeconomic status and public facility use, targeted screening programs must be accompanied by context-specific supportive care interventions which address the complex barriers faced by underserved populations. For instance, transportation assistance and geographic navigation support can help patients overcome travel burdens to distant tertiary cancer centres, particularly in rural or low-income settings. Financial navigation services, including counselling on available subsidies and reducing out-of-pocket expenses, may prevent catastrophic health expenditure, especially where public systems are underfunded and private services are unaffordable. Additionally, culturally and linguistically tailored psychosocial care, such as peer navigation or community health worker programs, can reduce stigma, improve trust in health institutions, and support informed decision-making. These needs are especially critical in the context of gynaecologic cancers, where stigma, lack of symptom awareness, limited specialist access, and weak referral systems often delay diagnosis and treatment initiation. Women from socioeconomically disadvantaged backgrounds may be disproportionately affected due to lower health literacy and limited autonomy in care-seeking. Strengthening public-sector capacity for early diagnosis and integrating outreach services, such as mobile clinics and decentralised screening, are essential to ensure that improvements in immunotherapy and precision diagnostics translate into equitable health gains. These structural reforms must be underpinned by health system investments and equity-focused policy strategies.
The large contribution of public facility use to observed inequality is likely driven by structural disparities in the way health services are accessed and delivered across socioeconomic groups. In many low- and middle-income countries, including Bangladesh, public facilities are the default option for lower-income populations, who often cannot afford private care [40]. These facilities frequently experience chronic underfunding, workforce shortages, supply chain disruptions, and overcrowding, which can limit the quality, timeliness, and continuity of care [41]. Patients in public facilities may also face longer travel distances and indirect costs such as time away from work, disproportionately affecting the poor [42]. In contrast, wealthier individuals more often seek care in private settings, which may provide faster service, more diagnostics, and improved doctor–patient communication. This dual-track health system reinforces a pattern where public facility use, though essential for universal access, can inadvertently perpetuate or widen inequalities in outcomes when not adequately resourced.

Strengths and limitations
This analysis offers several strengths. First, the sample size of 607 patients provides a reasonable basis for drawing conclusions about diverse cancer types, stages, and socioeconomic profiles. Second, the use of both relative and absolute measures of inequality, along with decomposition analysis, yielded a robust understanding of the factors driving inequities in HRQoL. Third, stratification by key variables such as cancer stage, type, and physical activity status allows for nuanced insights that inform policy and practice. Nonetheless, certain limitations merit consideration. This study was a cross-sectional, observational study and therefore does not permit causal inferences. Associations between facility type, education, physical activity, and health utility could have been influenced by residual confounding (e.g., unmeasured disease severity, social support), selection processes (who accessed particular services), and potential reverse causation (poorer health limiting activity). As such, the findings should be interpreted as associations and hypothesis-generating; any policy actions suggested here would require prospective, adequately powered evaluations to establish effects. Because the study did not establish causality, the proposed service changes (e.g. financial navigation, transportation support, plain-language communication) should be viewed as testable strategies rather than proven remedies; their effects and costs require prospective, equity-sensitive evaluation. Socioeconomic status was based on self-reported income, which can be prone to misclassification or reporting bias. Although the EQ-5D-5L is widely used for evaluating HRQoL, it might not capture all dimensions of well-being important to cancer patients, such as social support or stigma. Additionally, reliance on self-reported data for utility measures may introduce recall and reporting biases. These limitations underscore the need for longitudinal studies to better elucidate the causal pathways linking income and HRQoL.
Another key limitation is the absence of a Bangladesh-specific psychometric validation of EQ-5D-5 L in cancer patients. We mitigated this by following EuroQol guidance for administering the Bengali instrument [22, 25], citing Bangladeshi applications of EQ-5D [26], and conducting internal construct-validity checks (stage, treatment facility, education, and ceiling effects), which aligned with theory and regional oncology literature [4, 18, 27]. Nevertheless, a formal local validation, ideally evaluating reliability, responsiveness, and factor structure in Bangladeshi cancer cohorts and comparing English versus Bangla administration, remains a priority for future research [21].
Additionally, the generalisability of our findings is supported by similar patterns observed in South Asian and LMIC contexts [6, 16–18], although caution is needed when applying these findings to other regions with differing health systems or sociocultural norms.
Some effect estimates, particularly in subgroup analyses, yielded wide 95% confidence intervals, suggesting possible imprecision due to small sample sizes or sparse data. Our subgroup observations were interpreted cautiously. The wide confidence intervals and occasional overlap with the null indicated that apparent differences might have reflected sampling variability rather than true heterogeneity. We therefore treated these findings as exploratory and did not draw strong clinical or policy conclusions from subgroup contrasts. The primary inference remained with the overall estimates.
These results should be interpreted with caution, and future studies with larger and more diverse cohorts are needed to validate these associations. Finally, the study was conducted in two hospitals in an urban capital city, potentially limiting the generalisability of findings to more rural settings or smaller facilities.

Implications of findings
From a policy and clinical perspective, these results underscore several crucial directions. The evident socioeconomic gradient in HRQoL highlights the need for urgent policy action, including strengthening the National Cancer Control Program to promote early detection, ensure equitable access to oncology services, enhance targeted community outreach, expand financial assistance and implement educational initiatives. The local non-governmental organisations and community-based interventions can play a critical role in bridging the services gap, raising awareness and providing social and financial support to communities. On the other hand, the large negative contribution of public hospital treatment to overall HRQoL inequality underlines potential inefficiencies or resource constraints in public healthcare systems that warrant policy attention. Strengthening infrastructure, human resources, and care pathways in public hospitals could mitigate some of these observed disparities. Additionally, promoting physical activity through tailored exercise interventions or rehabilitative services at the community level may provide an accessible and cost-effective strategy for improving quality of life, particularly among economically vulnerable populations. Although some percentage contributions were large (e.g., 234% for public health facility), this reflected the arithmetic of component shares when the overall inequality (E-CI = 0.025) was modest; these percentages should be interpreted as diagnostic indicators of where inequity is generated, not as clinical effect sizes. Three variables in our decomposition suggest actionable levers: (1) Ongoing treatment facility (public vs. private): the strong pro-rich contribution associated with public-facility care points to remediable gaps (e.g., supportive-care pathways, patient navigation, psycho-oncology, social work/financial counselling, timely diagnostics). Health services could track income-stratified key performance indicators (HRQoL/utility, unplanned admissions, time to diagnostic completion) following targeted strengthening in public settings. (2) Regular physical activity (no vs. yes): the pro-poor contribution linked to inactivity supports embedding “exercise oncology” and social prescribing in lower-income communities (supervised programs, transport vouchers, brief advice in routine visits), with monitoring of referral completion and utility gains by income group. (3) Educational background (no education vs. tertiary): the pro-rich contribution from no formal education highlights the need for plain-language communication, teach-back, and tailored decision aids; services should audit readability and patient comprehension and link these to equity-sensitive outcomes. Additional contributors (e.g., residential status: rural vs. urban; BMI categories) were smaller but align with targeted access and prevention supports. To operationalise these levers within public facilities, we specify two pragmatic programs: (a) financial navigation—screen for financial strain at intake, deliver a documented relief plan (fee waivers/subsidies, transport grants), and review at 14 and 42 days; monitor equity-stratified key performance indicators: reach, days to relief, monthly out-of-pocket spend, missed appointments, unplanned admissions, and change in EQ-5D-5L utility; (b) transportation support, issue capped transport vouchers/rideshare credits or shuttle scheduling for patients with repeated non-attendance or long travel; pair with exercise-referral and reminders; track adherence, time to diagnostic completion, exercise-referral completion, per-patient transport cost, and short-term utility change. Evaluation can use stepped-wedge or difference-in-differences designs with prespecified equity-sensitive endpoints. Interpretation notes: to avoid over-interpretation, percentage contributions indicate where to intervene, not the magnitude of expected benefit; associations of the above strategies require prospective, equity-focused evaluation.

Conclusion
This study provides compelling evidence that socioeconomic status, as measured by income quintiles, is closely associated with HRQoL among cancer patients. Although overall inequalities were reasonable, the consistent patterns across various subgroups emphasise the urgent need for targeted policy and clinical interventions. While overall inequality was modest, the decomposition pinpointed public-facility care, physical-activity support for inactive patients, and health-literacy strategies for those with no formal education as priority fronts, offering a policy roadmap for equity-oriented service improvement and future evaluation. We observed modest, income-related inequalities in health utility, with these factors associated with the gradient; these observations are hypothesis-generating and warrant prospective evaluation before causal or policy-effect claims are made. These findings translate into implementable programs, financial navigation and transportation support, specified with eligibility criteria, workflows, and equity-stratified key performance indicators, and suitable for stepped-wedge or difference-in-differences evaluation in public facilities. The overall results were robust, however, subgroup findings were exploratory and imprecise, and should not be overinterpreted. Future work with larger samples and pre-specified subgroup hypotheses was needed to confirm or refute these patterns. Research should prioritise longitudinal designs to strengthen causal inference and to test the effectiveness and cost-effectiveness of the interventions, with the goal of advancing a more equitable care landscape.

Supplementary Information
Below is the link to the electronic supplementary material.