We adhered to the STROBE guidelines to ensure high-quality reporting in our observational cross-sectional study (Table S1).
Study Site
This study was conducted at the National Institute of Cancer Research and Hospital (NICRH), the only public facility in Bangladesh dedicated exclusively to cancer treatment. Public hospitals, like NICRH, primarily serve individuals from economically disadvantaged and lower-middle-class backgrounds, as the costs of treatment in private facilities are often unaffordable for most Bangladeshis [11]. Currently, Bangladesh lacks organized breast cancer screening programs, leading to almost all cases being diagnosed through clinical evaluation rather than early detection. Additionally, unlike in developed countries, there is no systematic referral system in place, and medical record-keeping is inadequate.
Patients
The study focused on women over 18 years old who presented with suspected breast cancer or had been diagnosed with the disease. Only those patients were enrolled whose initial cancer stage was documented in their medical records or, in cases where staging was unavailable, if the initial diagnosis occurred no more than six months prior to staging at our study center.
Questionnaire
We adapted a structured questionnaire from previous studies [5, 16], which is included as an additional file. The questionnaire comprised sections on sociodemographic variables, including age, education level, marital status, residence, and access to media and electronic devices. It also collected clinical history regarding breast cancer symptoms, capturing the type of initial symptoms (e.g., lump, breast pain, nipple discharge), the date of first symptom recognition, and participants’ perceptions of their symptoms’ severity.
Furthermore, the questionnaire explored barriers to seeking care, encompassing emotional factors (e.g., fear, embarrassment), practical constraints (e.g., financial limitations, time constraints), and health service-related issues (e.g., challenges in accessing healthcare, arranging transportation, or scheduling appointments). Participants were asked about their healthcare utilization, including the type of medical facility they first visited and any alternative treatments sought prior to diagnosis. The survey also assessed family support by gathering information on initial discussions about health concerns, recommendations to seek medical attention, and the level of support received after diagnosis. Knowledge and practices related to early detection were evaluated, focusing on breast self-examinations, prior clinical breast examinations, and awareness of mammography. Clinical variables, including tumor size and cancer stage classified by the tumor, node, and metastasis (TNM) system, were recorded. The data collected will be used to analyze associations between these variables and delays in diagnosis, offering insights into factors contributing to late-stage detection and their potential impact on treatment outcomes. The questionnaire was finalized for data collection after piloting it with five patients.
Data Collection
Face-to-face interviews using the structured questionnaire were conducted by trained interviewers—undergraduate students—who had no involvement in the clinical management of the patients. Given the conservative societal context, all interviewers were female, working under the direct supervision of our team’s oncologist at NICRH.
For patients unable to provide specific dates for their symptoms or first medical visit, they were asked to indicate a month or a range of months and the year. If a single month was indicated, the date was estimated as the 15th; for a month range, the midpoint between the 15th of those months was used. If only the year was provided, the date was coded as June 30th of that year. Cancer staging was analyzed by the oncologist based on available health records. Out of 355 cases, determining the cancer stages was not possible due to inadequate medical records.
Outcome Variables
In this study, delay is defined as the time interval experienced by women in the diagnostic and treatment processes.
Patient delay, refers to the time between the onset of symptoms and the first medical consultation, with a commonly accepted threshold for defining this delay being three months [5].
Provider delay, or system delay, refers to the time that passes between the initial medical consultation and the final diagnosis or treatment, with a commonly accepted threshold of one month [5].
Total delay encompasses the entire duration from the patient's first recognition of symptoms to the start of definitive treatment, integrating both patient and provider delays. In our study, total delay is considered significant when it exceeds four months.
To quantify our outcome variables related to delays, we categorized each patient as "1" or "Yes" if they experienced patient delay, provider delay, or total delay that surpassed the thresholds of three months, one month, and four months, respectively. Conversely, patients who did not meet these criteria were recorded as having no delays, designated by "0" or "No."
Possible factors
To find out potential risk factors associated with various types of delays, we examined a range of socioeconomic factors and the medical history of the patients as independent variables. These included the patient's age, geographic location (division), residency (urban or rural), educational attainment (illiterate, primary, and secondary) of both the patient and their spouse, household monthly income, access to portable electronic devices, exposure to mass media, lump breast pain, nipple discharge, skin changes, bone pain, breast self-examination, and family history of breast cancer.
Statistical analyses
We conducted descriptive statistics and also differences between delays associated with other factors tested by Chi-square tests and Fisher's exact test (in case of low frequency). Univariable (unadjusted) and multivariable (adjusted) logistic regression were utilized to identify factors that are associated with patient delay, provider delay, and total delay. In the univariable analysis, variables were individually added to the logistic regression model. We used an arbitrary p-value of ≤ 0.20 as a criterion for including covariates in the multivariable models [17] from univariable model. In this study, three models were utilized to identify associated risk factors of patient delay, provider delay, and total delay, designated as Model 1, Model 2, and Model 3, respectively. Results were reported as unadjusted/crude odds ratios (COR) and adjusted odds ratios (AOR) with their respective 95% confidence intervals. We considered a p-value of less than 0.05 to be statistically significant, indicating a 5% level of significance for interpreting our results. Additionally, we assessed multicollinearity in the final model using a cut-off value of 4.00 for the variance inflation factor (VIF) analysis [18]. At this stage, all variables were incorporated into the model since the VIF values for each variable were below 4.00. All analyses were performed using R software.