Limited information is available about the representativeness of survivors engaging in patient-centered research, despite the potential for threats to generalizability. We thus aimed to assess the representativeness of survivors engaged or interested in research development.
We used data from the Health Information National Trends Survey, a nationally representative survey, to identify survivors of adult cancers. Our outcomes of interest were based on responses to questions about engagement or interest in developing patient-centered research. We estimated the ratio of relative frequencies (RRF) and corresponding 95% confidence limits (CL) of sociodemographic and survivorship characteristics between survivors engaged or interested in patient-centered research and the overall survivor population.
Our study population comprised 934 survivors, of whom 5% reported being engaged in patient-centered research and 26% reported an interest in participating. Relative frequencies of characteristics were discordant for engaged survivors but largely similar for interested survivors compared with all survivors. In particular, engaged survivors had a higher relative frequency of individuals ages 50 to 64 years (RRF = 1.7; 95% CL, 1.1–2.5), Hispanic (RRF = 2.9; 95% CL, 1.2–6.9), non-Hispanic Black (RRF = 2.9; 95% CL, 1.1–2.5), and unemployment (RRF = 4.7; 95% CL, 1.4–16).
We observed several meaningful differences in the characteristics of survivors engaged in patient-centered research compared with all survivors, which raises concerns about the generalizability of findings from such studies.
Patient-centered research may not benefit the broader survivor community if survivors engaging in research development are not representative of all survivors. Greater attention to recruiting mechanisms is necessary to avoid creating disparities.
The number of cancer survivors increased from 9.6 million in 2000 to 16.9 million in 2019 (1). Research pertaining to survivors or survivorship also increased during this period. For example, the number of articles about survivorship published in PubMed-indexed journals increased from 493 in 2000 to 3,169 in 2018. This research generated considerable evidence about cancer survivorship (2–5), but substantial gaps remain in our understanding of the survivorship experience. Research to address these gaps may benefit from engaging survivors during study development (6–8).
The conventional approach to research emphasizes investigator-identified priorities (9), but this approach has been criticized because of disconnect with patients' priorities. Patient-centered research aims to address this disconnect by engaging patients through all stages of research, particularly study development, to ensure responsiveness to patients' priorities (10–15). Consequently, more studies are engaging survivors in research development (16–19), but little is known about the representativeness of survivors who are engaging in this process. If engaged survivors are systematically different from unengaged survivors, then patient-centered research may not sufficiently generalize or benefit the wider survivor community. For example, certain patient-reported outcomes may be emphasized for collection over others because of characteristics of survivors engaged in research development. Therefore, we aimed to assess the distribution of characteristics among survivors in two contexts: (i) engagement in patient-centered research development and (ii) interest in patient-centered research development.
Materials and Methods
We used data from the Health Information National Trends Survey (HINTS), which is a nationally representative cross-sectional survey administered by the US National Cancer Institute that aims to monitor changes in health communication (20), Each cycle of the HINTS survey consists of a two-stage design. In the first stage, a stratified random sample of U.S. addresses was selected. In the second stage, one adult was selected on the basis of the next birthday rule within the sampled household (20). The sampling strata include areas with high minority population concentrations, areas with low minority population concentrations, and counties comprising Central Appalachia (20). Poststratification replicate weights in HINTS are calibrated to distributions of population characteristics based on the American Community Survey. Relevant to our study, the survey ascertains information from individuals aged ≥18 years about cancer information access and usage (20). Individuals eligible for our study were aged ≥18 years when diagnosed with cancer and participated in HINTS 4 cycles 3 or 4 (2013 or 2014). Eligibility was limited to these cycles because the questions relevant to our study were not asked in other cycles (Fig. 1).
Our outcomes of interest were engagement in patient-centered research or interest in patient-centered research, which are different from research participation. HINTS includes a question which prefaces, “More and more, people are getting involved in research in new ways beyond being a research subject. They are partnering with medical researchers to help decide what research is done and how it is done. For example, people can suggest important topics to study or how to report results to the public. This is sometimes called “patient engagement” in research.” We defined engagement or interest based on whether participants responded “Yes” to the subsequent HINTS questions, “Have you ever engaged in medical research in this way?” and “Would you ever be interested in engaging in research in this way?” We included sociodemographic characteristics such as age (18–34, 35–49, 50–64, 65–74, and 75 years or older), gender (male/female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic other), education (high school graduate or less, some college, and college graduate or more), occupation (employed, unemployed, homemaker, retired, disabled, and other), household income (<$20,000, $20,000 to <$34,999, $35,000–<$49,999, $50,000–<$74,999, and $75,000 or more), and insurance coverage (yes/no). Survivorship characteristics included years since diagnosed with cancer (≤5, 6–10, 11–20, >20 years), type of cancer (top 5 prevalent cancers among the survivors, i.e., breast, cervical, melanoma, other skin cancers, prostate cancer, all other cancers, and survivors with multiple cancers), and general health status (excellent, very good, good, fair, and poor).
We estimated the relative frequencies (i.e., proportions) of sociodemographic and survivorship characteristics for survivors engaged or interested in patient-centered research and the overall survivor population. We combined HINTS 4 cycles 3 and 4 data and incorporated the sample weights, which accounted for the survey design, nonresponse, and noncoverage (20). The combined surveys allowed for a larger sample size and reduced variance, but the weighted number of survivors in this scenario represents the combined total survivor population in the United States for each survey iteration (i.e., effectively doubled) and should not be interpreted as a single estimate of the number of survivors in the United States. Nevertheless, this weighting does not affect the validity of the estimated frequencies of characteristics. In addition, we used 100 replicate weights to obtain the standard errors of the relative frequencies (20). We used these relative frequencies to estimate the ratio of relative frequencies (RRF), which compared the frequency of sociodemographic and survivorship characteristics between survivors engaged or interested in patient-centered research development and all survivors using the formulae:
We estimated 95% confidence limits (CL) for RRFs using an approximate formula for the SE (21, 22). The numerator of the RRF represents the relative frequency of the characteristic among survivors engaged or interested in research development and the denominator is the relative frequency of the characteristic among all eligible survivors. An RRF ≠1 indicates that the distribution of the characteristic is different between survivors engaged or interested in patient-centered research development and cancer survivors overall. This measure thus represents the magnitude of selection bias (i.e., representativeness).
Our eligible population comprised 990 cancer survivors, of whom 934 were included in the analysis (n = 56 with missing age at diagnosis; Fig. 1), which represented an estimated total population of 35,943,723 cancer survivors. Survivors were largely aged 50 years or older (81%), 60% were female, 71% were non-Hispanic White, and 43% survived ≥10 years from cancer diagnosis at the time of survey. Among all cancer survivors, 16% reported having heard of patient engagement in medical research, 5% reported being engaged in patient-centered research development, and 26% reported that they would be interested in participating.
Figure 2 and Supplementary Table S1 summarize the RRFs of sociodemographic and survivorship characteristics among survivors engaged in patient-centered research development compared with all cancer survivors. The frequency distributions among survivors who reported engagement in patient-centered research development were different from survivors overall. For example, compared with all survivors, engaged survivors had a higher relative frequency aged 50–64 years (RRF = 1.7; 95% CL: 1.1–2.5), non-Hispanic Black (RRF = 3.4; 95% CL: 1.4–8.6), or Hispanic (RRF = 2.9; 95% CL: 1.2–6.9). In contrast, engaged survivors had a lower relative frequency of individuals aged 35–49 years (RRF = 0.12; 95% CL: 0.02–0.88), an annual household income of $50,000–$74,999 (RRF = 0.42; 95% CL: 0.18–0.96), and survivors of skin cancers other than melanoma (RRF = 0.30; 95% CL: 0.10–0.91).
Figure 3 and Supplementary Table S2 summarize the RRFs of sociodemographic and survivorship characteristics among survivors interested in patient-centered research development compared with all cancer survivors. We observed fewer differences in frequency distributions between survivors interested in patient-centered research development compared with all survivors. For example, compared with all survivors, interested survivors had a higher relative frequency of Hispanics (RRF = 1.7; 95% CL: 1.0–2.7) and survivors aged 50–64 years (RRF = 1.3; 95% CL: 1.1–1.5). In contrast, interested survivors had a lower relative frequency of non-Hispanic others (RRF = 0.24; 95% CL: 0.11–0.55), survivors aged ≥75 years (RRF = 0.59; 95% CL: 0.39–0.89), and prostate cancer survivors (RRF = 0.43; 95% CL: 0.27–0.70).
The characteristics of cancer survivors engaged in patient-centered research development have received limited attention despite potential consequences for selection bias and generalizability. We used data from a nationally representative sample of cancer survivors to address this gap in the evidence. Our results suggest meaningful differences in the relative frequencies of sociodemographic and survivorship characteristics among survivors engaged or interested in patient-centered research compared with the overall survivor population, which raises questions about generalizing patient-centered cancer survivorship research. In addition, few survivors were engaged in patient-centered research (5%), whereas a larger proportion reported an interest in engaging patient-centered research (26%).
Despite reports of underserved populations being underrepresented in clinical cancer research (23, 24), we observed a higher frequency (i.e., overrepresentation) of engagement for several characteristics associated with underserved populations. For example, the frequency of non-Hispanic Black, Hispanic, uninsured, and unemployed survivors was higher among engaged survivors compared with all survivors. Such overrepresentation may partly reflect targeted efforts to engage underserved or underrepresented groups (17–19, 25), but the consequences of targeted overrepresentation are unknown for the broader survivor population and require further investigation. Nevertheless, our findings more broadly reflect the growing body of literature refuting the notion that underserved populations are resistant to participating in research (26–29) and reinforces the feasibility of engaging perceived “difficult to reach” populations in survivorship research.
Several characteristics (e.g., age, gender, health status) associated with participation in survivorship research also extend to engagement in patient-centered research. For example, the frequencies of older adults (>65 years) and females were lower among survivors engaged in patient-centered research compared with all survivors. This underrepresentation is a concern because these groups are under-represented as participants in survivorship research, which limits the evidence base (30–32). In addition, survivors who reported poor health status were underrepresented in engagement. This finding is not particularly surprising given the potential physical limitations of survivors with compromised health but warrants potential solutions for engaging this under-represented group.
Our findings about interest in patient-centered research development are promising for investigators who seek to engage survivors. Other than older survivors and survivors who reported poor health status, we observed generally similar distributions of characteristics for survivors who are interested in patient-centered research development. This contrast between interest and engagement is consistent with previous reports of participation in research being related to logistical factors (24, 33). Despite differences between engaging in research development and being a participant in a study, engaging in research development also involves a substantial commitment that may not be overcome by interest alone. For example, survivor engagement may be driven by feasibility factors such as, work schedule and flexibility, familial responsibilities, transportation, and the availability of social resources to balance competing demands (34, 35). The consequence is that recruiting survivors to engage in developing research may be based on convenience (i.e., nonprobability or purposive) sampling (15, 36), which may not represent the perspectives of all interested survivors and could induce selection bias in the development of patient-centered research.
Certain limitations should be considered when interpreting our findings. Our frequency estimates may be sensitive to misclassification of engagement or interest in patient-centered research. The questions in HINTS explicitly describe the role of the patient in research (e.g., “…partnering with medical researchers to help decide what research is done and how it is done…”), but potential confusion between engagement and participation is possible. Some survivors could be engaged in multiple studies, which could increase the possibility of over-representing certain characteristics while underrepresenting others, but details about the extent of engagement were unavailable. Knowledge of patient engagement in research was reported by only 16% of survivors. In addition, our analysis was limited to the sociodemographic and survivor characteristics available in HINTS. Other factors such as types of research in which patients were engaged, geographic location, mental health, and personality traits (e.g., altruism), may provide further insight about survivors engaged or interested in engaging in patient-centered research. Finally, despite HINTS being representative of the U.S. population, representation is limited to the civilian noninstitutionalized population. Consequently, survivors outside these parameters were not represented in our analysis.
Assuming that the limitations of our analysis did not materially affect our results, our findings may have practical implications. Our results suggest low overall engagement of cancer survivors in patient-centered research development. Several characteristics that we examined were either over- or under-represented among survivors engaged in research compared with the overall survivor population. The over-representation of underserved populations among engaged survivors is encouraging because of recent efforts to target involvement of historically underrepresented survivors with an increased risk of health disparities. Engaging underserved survivors allows overcoming knowledge deficits and lack of generalizability to these populations. Nevertheless, the overall lack of representativeness of characteristics among engaged survivors compared with all survivors emphasizes the need for greater attention to strategies used to recruit survivors in patient-centered research development. The negative consequences of suboptimal representativeness of participation in research (37–39) provide evidence that considerations about representativeness of engagement are not trivial. The well-intentioned premise of engaging patients in the research process (i.e., patient-centered research) may be compromised by insufficiently harnessing the breadth of patient diversity. Certain research priorities or outcomes may be over- or underemphasized to the detriment of other survivors, which could amplify disparities. Therefore, common strategies for recruiting survivors for engagement should be reconsidered. For example, convenience sampling should be replaced by probability sampling, which may improve generalizability and ultimately decrease recently identified knowledge gaps (31, 40). Nevertheless, best practices have not been established for recruiting patients to engage in the research process (41). Greater attention to these issues is imperative to optimize patient-centered survivorship research.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M.M. Lubas, T.M. Brinkman, R.P. Ojha
Development of methodology: B. Ghabach, R.P. Ojha
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Lu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Lu, K. Narra, R.P. Ojha
Writing, review, and/or revision of the manuscript: M.M. Lubas, Y. Lu, A.W. Gehr, B. Ghabach, B. Tanna, K. Narra, T.M. Brinkman, R.P. Ojha
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Lu
Study supervision: R.P. Ojha
This work was partially supported by the NCI Training in Pediatric Cancer Care Survivorship award 5T32CA225590 (to M. Lubas).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.