Demographic Factors Impacting Time to Diagnosis of Diffuse Large B-Cell Lymphoma

Purpose: Diffuse large B-cell lymphoma (DLBCL) is the most common form of non-Hodgkin’s lymphoma, and patients are sometimes asymptomatic until the presentation of an oncological emergency, such as end organ complications, tumor lysis syndrome, or spinal cord compression. We aimed to investigate demographic and clinical factors associated with time from initial symptom onset until diagnosis of DLBCL. Methods: We performed a retrospective cohort study of 1076 patients with DLBCL diagnosed in the Monte�ore Health System from 2005 to 2022. Logistic regression models, adjusted for race and preferred language, estimated the associations between demographic factors and increased time to diagnosis (TTD), de�ned as a time to diagnosis greater than the median for non-Hispanic White patients (58.5 days).


Introduction
Diffuse large B-cell lymphoma (DLBCL) is the most common high-grade form of non-Hodgkin lymphoma [1]. Prolonged time to diagnosis (TTD) of DLBCL can lead to increased risk of morbidities, with those diagnosed later at risk for more advanced disease at presentation, and tumor related complications including spinal cord compression, which can lead to irreversible de cits, such as paralysis or loss of sphincter control [2]. Prolonged TTD can also lead to other oncologic emergencies or end organ damage, such as acute respiratory failure due to pulmonary spread of the lymphoma or acute and chronic kidney disease due to tumor lysis syndrome (TLS) [3]. While the majority of patients diagnosed with DLBCL have a good prognosis with favorable life expectancy outcomes with current treatment methods, the morbidity associated with both undiagnosed disease and treatment can also be extensive. Undiagnosed spinal cord compression was associated with a 2.5 relative risk of loss of sphincter function in lymphoma patients [4]. Even in patients who eventually died of lymphoma, continued control of sphincter function enabled better quality of life for these patients [4]. Similarly, a hospital-based study of patients with clinical TLS found that while a majority of patients survived, 58% of patients with TLS had resultant renal failure [4].
The extent of adverse effects of treatment is dependent in part upon the stage of the tumor at diagnosis, with those with delayed diagnoses at risk for signi cant impacts on treatment-related morbidities [6].
Patients of different racial or ethnic groups may be more likely to have delayed diagnoses. In previous studies evaluating hematologic malignancy patients in the Bronx, non-Hispanic Black patients were 10% more likely than non-Hispanic White patients to present with advanced stage disease, despite a mean age at diagnosis which was approximately eight years younger [7]. Here we seek to identify demographic factors associated with time to diagnosis of DLBCL.  positive patients were excluded given that lymphomas in this population tend to have signi cantly more interaction with the healthcare system with disease at a more advanced stage, potentially confounding time to diagnosis data [8,9]. Race and ethnicity were categorized as non-Hispanic White (NHW), non-Hispanic Black (NHB), and Hispanic/Latino. Asian and Paci c Islander, American Indian and Alaskan Native patients were excluded from this study due to small sample size (n = 20), as were patients with a follow-up time of zero (n = 32). Time to diagnosis was calculated by the date of rst encounter within the EMR of MMCHMC and the diagnosis date.

Methods
Pearson's Chi-square tests and t-tests were used to evaluate variables normally distributed after visual inspection with non-parametric testing used for variables not normally distributed. Time from symptom onset to diagnosis (or time to diagnosis, TTD) was compared by insurance type, preferred language, biologic sex, vital status, socioeconomic status (SES), and race with log rank testing and Wilcoxon testing. SES was based upon the patient's census block group and is reported as a summary Z-score relative to the New York State mean using six variables. Additional details in reference [10]. Kaplan Meier curves were used to compare the TTD by race as well. Analyses strati ed patients by insurance status, preferred language, and spread of malignancy at the time of presentation, de ned by evidence of extranodal spread, to determine whether there were differences among racial/ethnic groups by insurance status, language, or spread of disease at initial presentation. Further analysis was performed using multivariable logistic regression to identify risk factors associated with prolonged time to diagnosis longer than the median 58.5 days within Non-Hispanic White patients. SES, sex, race/ethnicity, preferred language, vital status, insurance type, age at diagnosis in years, and spread at presentation were compared using univariate analysis. A priori modeling with race/ethnicity and all variables with p < 0.2 in univariate modeling were included, with stepwise backward elimination for nal model building. All analyses were conducted using Stata, version 17. This study was approved by the Albert Einstein College of Medicine Institutional Review Board.

Discussion
We observed an increased likelihood of increased TTD for DLBCL in Spanish speaking and "other" speaking patients. These ndings were also associated with an increased TTD for Hispanic speaking patients in general relative to non-Hispanic White patients; it is likely that the increased TTD in Hispanic patients is due to the higher proportion of patients dealing with a language barrier. If Spanish speaking patients are getting delays in treatment, steps taken to reduce language barriers in treatment and increase the resources for Spanish speaking patients could potentially aid in the earlier detection of lymphoma (i.e., diagnosing the disease before extranodal spread.) Of note as well is that patients who speak a language other than English or Spanish are even more likely to experienced an increased TTD; resources for patients who prefer other common languages would be helpful in this population as well.
Previous research has indicated that patients who are not pro cient in English are twice as likely to revisit the emergency department within 72 hours of a discharge [11]. When patients with limited English pro ciency present to the ED, staff were more likely to interact in English for any given interaction, which happened 29.1% of the time [12]. While this study found that virtually 100% of patients with limited English pro ciency were able to use their preferred language during triage and an initial assessment, fewer than 25% of patients consistently were able to utilize an interpreter or a pro cient staff member during ongoing evaluation and treatment [12]. Differences observed in TTD based on language may be related to providers underutilizing interpretation services at later points in the diagnosis and treatment process.
We observed differences in TTD by insurance status and language among different ethnicities. A crosssectional study of cancer survivors found that publicly insured individuals were twice as likely to use the emergency department as their privately insured peers [13]; this may be responsible for a faster initial diagnosis; however, there are data indicating that privately insured individuals are more likely to get all of their healthcare needs met in the long run [14]. While publicly insured children were more likely to have their healthcare needs met than uninsured children, privately insured children are 2.5 time more likely to receive all of their specialty referrals (a proxy measure of met healthcare needs) compared to publicly insured children [14]. While we found that privately insured patients were more likely to experience an increased TTD, it is important to consider how private insurance speci cally impacts the overall care a patient receives. Future research could investigate how demographic factors such as insurance status and preferred language impact patient care beyond disparities in TTD.
For DLBCL speci cally, it is also important to consider how evolving treatment and diagnostic tools will impact disparities in TTD. Disparities based on insurance status and language may be partly related to the extent to which patients can access care: currently, the gold standard method of diagnosis is lymph node biopsy or needle core biopsy, a relatively invasive and expensive procedure [1]. Recent research, however, has investigated the prospect of using cell-free DNA (cfDNA) to both diagnose DLBCL and track clonal evolution of resistant tumor cells, which is comparatively less invasive for patients and more economical [15]. As our research showed an increased time to diagnosis for patients with low socioeconomic status, more practical and cost-effective methods of tracking tumors may help reduce socioeconomic disparities in TTD. Similarly, non-surgical tools to diagnose DLBCL could ameliorate language barrier-based disparities in TTD as consenting a patient for a procedure is more time consuming and logistically complicated when an interpreter must be used.
Our cohort of 1,076 patients was 24.7% non-Hispanic White, 26.5% non-Hispanic Black, and 39.86% Hispanic: the diversity of the patient population studied was a major strength in our analysis. A potential limitation of this study was the extraction of preferred language, race, and other demographic information from electronic health records, which do not always accurately re ect patient selfidentity. However, there are data suggesting that EMR based data extraction methods for language preference and race are accurate [17]. Additionally, it is important to note that using EMR enabled us not only to collect data about the preferred languages of individual patients, but also to use the data of patients who were not pro cient in English or Spanish. Non-English and non-Spanish speakers are often excluded from clinical research due to the di culty associated with the language barrier inherent to working with this population. Finally, it is important to note that these data come from an inner city hospital system serving a high proportion of patients with low SES and a high proportion of non-English speaking patients. Given that the Monte ore system has extensive support systems to reduce language and nancial obstacles to patient healthcare and well-being, and that we still observed a difference, our research suggests that these differences may be even more pronounced in hospital systems which are less equipped to serve patients facing linguistic or socio-economic obstacles to timely diagnosis of liquid malignancies.
In conclusion, our study found a relationship between language barriers and TTD of DLBCL. Speci cally, we observed a longer TTD for patients who do not primarily speak English at home. Future research into the impact of a language barrier on all steps of care could potentially optimize strategies for increasing patient access to care. In particular, increased access to literature on hematological malignancies and patient services in Spanish speci cally could potentially reduce disparities in time to diagnosis based on language di culties in diverse patient populations.

Declarations
Funding: This project is funded by the Einstein-Monte ore research fellowship.