Patients (N=105) who received systemic therapy (eg, chemotherapy or immunotherapy) within the last year for NHL or CRC were recruited between November 2022 and April 2023 from 3 clinics at MedStar Georgetown University Hospital, MedStar Washington Hospital Center, and John Theurer Cancer Center as part of a larger cross-sectional study about cancer survivorship experiences. The study’s methods were previously published, and the current analysis represents a secondary analysis of these data [18]. NHL and CRC were selected to represent both solid and hematological cancers and due to their high prevalence at collaborating clinics, frequent use of systemic therapies, and prevalence across age and gender groups (versus, eg, breast or prostate cancer) [19]. Participants reported DHL on the validated eHealth Literacy Scale (eHEALS) by rating their ability or confidence to complete digital health tasks on 8 Likert-scale items (score range: 8-40) [20,21]. Scores ≤20 indicate low DHL [22-24]. Cronbach α was 0.96 in this sample. Participants reported their age, gender, race, ethnicity, education, general health literacy (with the 3-item Health Literacy Screening questionnaire [25]), and disease- and treatment history–related clinical factors. Surveys were mailed or emailed to the patient and were completed on paper, online, or over the phone.

Statistical analysis used RStudio (version 2023.12.0+369) and GraphPad-Prism (version 10.3.1). We computed descriptive statistics, then used independent-sample t tests, χ² tests, and correlations (as appropriate) to test associations between sociodemographic or clinical factors and DHL, with a 2-tailed α level of .05. We used simultaneous multiple linear regression to evaluate whether associations between clinical factors and DHL remained when adjusting for covariates (age, education level, and general health literacy; these were selected according to prior work or significant relationships with DHL in this dataset at P<.05) [13,15,26,27].

To identify how single items performed to identify low DHL, we generated receiver operating characteristic (ROC) curves for each eHEALS item using a previously described approach [28,29]. We examined the area under the curve (AUC) for each ROC curve to compare the overall performance of each item for predicting low DHL.

The Georgetown-MedStar Joint Oncology Institutional Review Board (STUDY00005586) approved the study. All participants provided informed consent. Study procedures included standard practices to prioritize participants’ privacy and confidentiality, including labeling surveys with participant IDs rather than identifiers, secure data storage, deidentification prior to analysis, and aggregate reporting. A US $20 gift card was offered to all participants.

Among the 105 participants (Table 1), 50 (48%) had NHL and 55 (52%) had CRC. Participants were primarily middle-aged to older (mean age 60, SD 13.2, range 22‐84 years), male (n=57, 54%), identified as White (n=65, 62%), and had a college-level education or higher (n=57, 54%). The mean eHEALS score was 26.9 (SD 8.24, range 8‐40), and 24% (n=25) had low DHL (score ≤20).

Those with lower education levels and lower general health literacy had lower DHL (Spearman ρ=0.29; P=.004 and r=0.25; P=.009, respectively). DHL did not vary by age, gender, race/ethnicity, or household income (all P>.21).

Patients with NHL reported lower DHL than patients with CRC (t₁₀₃=2.72; P=.008; Figure 1A). This pattern persisted when controlling for age, education, and general health literacy (unstandardized b coefficient=−3.07, SE=1.65; P=.07). The patient subset who participated in a clinical trial (n=10) as part of their cancer treatment had lower DHL than patients who did not participate in a trial (t₁₀₀=3.08; P=.003; Figure 1B). In the adjusted model, patients reporting trial participation still showed lower DHL when controlling for age, education, general health literacy, or disease type (b=−10.45; SE=3.20; P=.002; Table 2). DHL did not vary by time since diagnosis or treatment type (all P>.23).

In an ROC analysis predicting low DHL (eHEALs ≤20), individual items’ AUC values ranged from 0.895 to 0.975 (Table 3, Figure S1 in Multimedia Appendix 1). eHEALS item 4, “I know where to find helpful health resources on the Internet,” had the highest AUC (0.975, 95% CI 0.949‐1.00; P<.001; Figure 2A). Responses to item 4 were strongly correlated to overall eHEALS scores (r=0.926; P<.001; Figure 2B). We determined the optimal threshold score for this item to predict high versus low DHL based on accuracy (Table S1 in Multimedia Appendix 1). Using the Likert-scale options (ranging 1-5), a score less than 3 predicted low DHL with an accuracy of 96.2%, a sensitivity of 94.94% (95% CI 87.69‐98.01) and a specificity of 96.15% (95% CI 81.11‐99.80).

DHL allows patients to benefit from technological advances in oncology care. However, little is known about which patients with cancer are more likely to experience low DHL, and clinically feasible ways to identify patients with low DHL are limited. Here, we examined sociodemographic and clinical factors associated with DHL and explored whether single items could identify patients with low DHL. Patients with a lower education level and general health literacy had lower DHL. Patients with NHL also had lower DHL than those with CRC. Interestingly, those who reported participating in a clinical trial had lower DHL than those who did not. ROC analysis showed that agreement with a single item from the widely used eHEALS (“I know where to find helpful health resources on the Internet”) was strongly predictive of low DHL; this may be a brief yet sensitive way to screen for low DHL.

Links between DHL and sociodemographic factors have previously been reported [13,15,26,27], but clinical correlates have been less frequently explored. We observed notable differences in DHL between patients with CRC and NHL that remained after adjusting for covariates. Future studies should identify drivers of DHL differences by cancer type in larger samples. We observed a novel, unexpected association between lower DHL and clinical trial participation, contrasting with prior findings that clinical trial participants used the internet first, before physicians, for clinical trial information [30]. On the other hand, it is possible that patients with worse DHL have poorer outcomes, thus necessitating later lines of therapy and enrollment in clinical trials, which could help to explain our findings. The complex relationship between clinical trial participation and DHL requires further investigation in prospective studies with electronic medical record–documented trial participation.

Rapidly and effectively identifying those with low DHL is critical to advancing this research area. Short-form screeners are well-suited for time-limited oncology clinics [29,31]. Our findings indicate that disagreeing or strongly disagreeing with the statement “I know where to find helpful health resources on the Internet” is a candidate screening criterion for low DHL. If validated in larger samples and other cancers, the item could be included in previsit paperwork, and identified patients with low DHL could receive additional support (eg, for telehealth and patient portals) or alternative modalities (eg, print resources). Future studies could also use this item to quickly screen for digital health needs in study candidates.

This study’s small convenience sample may limit its statistical power to identify associations and may restrict generalizability. Although we included people with solid and hematological cancers, correlates and best screening items for DHL may differ across other cancer types. Nonetheless, the observed links to sociodemographic factors are consistent with prior literature, and the full range of possible eHEALS scores (8-40) was represented in our sample. Future studies should include larger samples, more clinical trial participants, and more diverse sociodemographic and cancer groups. Additionally, future studies should include family members and other care partners, who have an important role in helping patients with lower DHL navigate their care.

DHL is an emerging and actionable SDOH impacting health outcomes for people with cancer. These results provide insight into how DHL varies across cancer patients and suggest a highly predictive single-item screener that may be used to identify patients with low DHL. Using this item, if it is further validated, could increase opportunities for clinical support and developing DHL interventions better suited to patient needs.