Identifying predictors of COVID‐related delays in cancer‐specific medical care

Abstract Purpose Evidence of the impact of the COVID‐19 pandemic on cancer prevention and control is growing, but little is known about patient‐level factors associated with delayed care. We analyzed data from a survey focused on Iowan cancer patients' COVID‐19 experiences in the early part of the pandemic. Methods Participants were recruited from the University of Iowa Holden Comprehensive Cancer Center's Patients Enhancing Research Collaborations at Holden (PERCH) program. We surveyed respondents on demographic characteristics, COVID‐19 experiences and reactions, and delays in any cancer‐related health care appointment, or cancer‐related treatment appointments. Two‐sided significance tests assessed differences in COVID‐19 experiences and reactions between those who experienced delays and those who did not. Results There were 780 respondents (26% response), with breast, prostate, kidney, skin, and colorectal cancers representing the majority of respondents. Delays in cancer care were reported by 29% of respondents. In multivariable‐adjusted models, rural residents (OR 1.47; 95% CI 1.03, 2.11) and those experiencing feelings of isolation (OR 2.18; 95% CI 1.37, 3.47) were more likely to report any delay, where experiencing financial difficulties predicted delays in treatment appointments (OR 5.72; 95% CI 1.96, 16.67). Health insurance coverage and concern about the pandemic were not statistically significantly associated with delays. Conclusion These findings may inform cancer care delivery during periods of instability when treatment may be disrupted by informing clinicians about concerns that patients have during the treatment process. Future research should assess whether delays in cancer care impact long‐term cancer outcomes and whether delays exacerbate existing disparities in cancer outcomes.


| INTRODUCTION
The impact of COVID-19 on US healthcare systems was widespread, affecting both general medical care and specialty care. 1,211][12]16 Care was disrupted due to facility limitations, and patients canceling appointments out of fear of getting COVID-19. 5,6,9To determine the long-term impacts of the COVID-19 pandemic on cancer patients and prepare the healthcare system for any future disruptions, it is important to identify factors that contributed to delays in cancer care.
To address this knowledge gap, we conducted a survey focused on Iowan cancer patients' COVID-19 experiences in the early part of the pandemic.The purpose of this study was to assess whether cancer-related health care was delayed in Iowa due to the COVID-19 pandemic and to identify factors that contributed to delays in cancer care.We expected that survey respondents would report substantial delays in cancer-related health care.We hypothesized that patients would report more delays in routine check-ups and surgery than other cancer care types and that health insurance coverage and rurality would be among the strongest predictors of delays in cancer-related health care.

| Data source
Data for this analysis were collected as part of the PERCH (Patients Enhancing Research Collaborations at Holden) Cancer and COVID-19 Survey at the University of Iowa College of Public Health.Full details on the PERCH Cancer and COVID-19 survey are being published elsewhere [paper in review].Briefly, the PERCH Cancer and COVID-19 Survey was a cross-sectional emailed Qualtrics survey conducted between September 9, 2020, and October 30, 2020, through partnership with the University of Iowa Holden Comprehensive Cancer Center.A reminder email was sent 2 weeks after initial invitation reminding participants of the survey.Approval for this study was provided by the University of Iowa Institutional Review Board (#202006014).

| Study population
PERCH is an opt-in program for individuals receiving cancer diagnostic testing at the University of Iowa Hospitals and Clinics who are interested in participating in cancer research.Individuals eligible for inclusion in this study were those who enrolled in PERCH after February 1, 2018, with an e-mail address on file and who were 18 years of age or older.Those excluded from our analytic dataset were those who opted into PERCH after a potential cancer diagnosis and were found to have a negative biopsy result, meaning they did not have cancer.These respondents were identified by a survey item asking patients to indicate whether they ultimately received a cancer diagnosis.

| Measures
Data in this analysis were collected from the PERCH survey and subsequent linkage to electronic medical records at the University of Iowa Hospitals and Clinics for collection of clinical data.Survey items were created using a combination of previously validated and newly developed pre-tested Likert-type and multiple-choice items.Skip patterns were implemented for specific questions to reduce the likelihood of erroneously reported data.
Our outcome, delays in cancer care, was measured using a survey item "Were you scheduled for any cancerrelated medical care, surgical care, or follow-up/surveillance that you had to cancel or reschedule during the COVID-19 restrictions?" with yes or no response options.Respondents could indicate the types of appointments that were delayed using an item "What did you have to cancel or reschedule?" with response options including routine appointment, screening or cancer follow-up test, blood test, surgery, chemotherapy, radiation therapy, occupational/physical therapy, and other (please specify).Respondents were able to select any and all that applied.Demographic characteristics measured by survey items included age in years, sex assigned at birth (male or female, prefer not to answer), race (White, Black, American Indian or Alaska Native, Asian, Native Hawaiian/Pacific Islander, and others), ethnicity (Mexican, Chicano/a, Puerto Rican, Cuban, other Hispanic/Latinx/Spanish origin, or not of Hispanic/Latinx/Spanish origin), household income (categorized into nine levels), education level (less than 8th grade, 8th-11th grade, high school diploma or GED, vocational or technical school, some college, college graduate, and postgraduate), health insurance coverage (employer-provided, private coverage, Medicare, Medicaid, TRICARE/Veterans Affairs/Military, Alaska Native/Indian Health Service/Tribal Health Services, and others), and marital status (single/never married, married, not married but living with a romantic partner, separated, divorced, widowed).Rurality was measured using Rural-Urban Continuum Codes (RUCCs) and categorizing counties as either rural (codes 4-9) or urban (codes 1-3).
We surveyed respondents on the impacts of and experiences related to the COVID-19 pandemic, including thoughts and behaviors related to social distancing, emotional reactions to the pandemic, and ability to find information on health topics.Thoughts and behaviors about COVID-19 and social distancing included concern about COVID-19 (not at all, a little, somewhat, very), receiving emotional, physical, and/or material support in the past 2 weeks (yes/no), whether they lost their job or primary source of income (yes/no), if they had a pay decrease (yes/ no), if they lost insurance coverage (yes/no), whether they think social distancing is important (not at all, a little, somewhat, very), whether they felt their cancer history impacted social distancing measures (a great deal, a moderate amount, a little, none at all), whether COVID-19 impacted their cancer care, whether COVID-19 disrupted their care, felt social distancing kept them safe, and felt social distancing was easy to do (all of which had response options strongly disagree, disagree, neutral, agree, strongly agree).
Emotional reactions to the pandemic included items on whether the respondent felt their cancer history impacted social distancing measures, COVID-19 impacted their cancer care, COVID-19 disrupted their care, felt social distancing kept them safe, felt social distancing was easy to do, were nervous/anxious/on edge, were anxious about getting COVID-19, felt more vulnerable to COVID-19 due to their cancer history, were worried about dying from COVID-19, were worried about friends/family dying from COVID-19, experienced feelings of social isolation or loneliness, had difficulty or inability to work as usual, experienced financial difficulties, and felt their healthcare team shared adequate information on prevention, protection, or care for COVID-19 (all of which had response options strongly disagree, disagree, neutral, agree, strongly agree).Items based on their most recent search for information asked about effort needed to find information, frustration when searching for information, concern about the quality of information, and thinking that the information was hard to understand (strongly disagree, disagree, agree, strongly agree).

| Statistical analysis
All data were analyzed using SAS 9.4 software (SAS Institute, Cary, NC).We calculated frequencies for all variables and used Chi-square tests to assess differences in demographic characteristics between those who reported delays in care compared to those who reported no delays in care.We used univariate logistic regression to individually assess potential predictors of delays in cancer care.Multivariable-adjusted logistic regression models were used to determine independent predictors of delays in cancer care.Variable selection for predictive modeling was based on a combination of univariate model significance, correlation between variables (i.e., choosing between variables with high correlation), and scientific reasoning for variables that could feasibly result in care delays.
Outcomes in these models were delays in care, either reported for any type of appointment (routine appointment, screening, blood test, surgery, chemotherapy, radiation therapy, physical/occupational therapy, and others) or specific to cancer treatment (surgery, chemotherapy, radiation therapy).Covariates (predictors) in the models included age (<65 year [referent] vs. ≥65 year), marital status (married [referent] vs. unmarried), income ($50,000 or more per year [referent] vs. $49,999 or less per year), insurance (three variables for private insurance [referent] vs. Medicare, Medicaid/TRICARE/VA/Military, and no insurance), race (White [referent] vs. non-White), education (two variables for college degree or more [referent]  vs. some college/technical/trade school, high school degree or less), and impact and emotional response variables (disagree/strongly disagree/neutral [referent] vs. agree/ strongly agree).All significance tests were two-sided with p < 0.05 considered statistically significant in all analyses.

| RESULTS
The survey yielded a 26% response with 780 respondents.Of these, 123 were found to not have cancer and were excluded, resulting in 657 respondents in our analytic sample of patients who opted into the PERCH registry and had an email on file after February 2018.Among the analytic sample, 130 patients were diagnosed in 2003-2017 and 443 were diagnosed in 2018-2020, the mean age was 62.7 years, 56% were female, 79% were married or living with a partner, 54% were rural, 99% were non-Hispanic White, 59% had private insurance coverage, 3% were uninsured, and 59% had college degree or higher educational attainment (Table 1).The majority of respondents were concerned about COVID-19 (92%), felt their cancer history impacted their social distancing measures (83%), thought social distancing was easy (70%), keeping them safe (78%), and important (98%) (Table 2).Nearly two-thirds of respondents were worried about dying from COVID-19, but less than half of respondents were nervous/anxious/on edge (34%), were anxious about getting COVID-19 (45%), felt isolated (23%), had difficulty or inability to work as usual (22%), and experienced financial difficulties (14%).

| Cancer care delays
Almost one-third (29%, n = 170) of respondents reported delays in their cancer care due to the COVID-19 pandemic.Among those that reported delays, 38% reported delays for routine appointments, 23% for screenings, 15% for blood tests, 6% for surgery, 3% for chemotherapy, 3% for physical/occupational therapy, and 1% for radiation therapy.Ten percent of delayed appointments were unable to be attributed to one of these categories and were listed as "other" (Table 3).Respondents who experienced delays were younger (p = 0.01), had a higher proportion of natal sex females (64% vs. 53%, p = 0.04), and had a higher proportion of rural residents (p = 0.04) compared to those who did not experience delays.Differences were not observed for race, ethnicity, health insurance coverage, educational attainment, income, or marital status.Among respondents who experienced delays, the mean age was 61.1 years, 64% were female, 77% were married or living with a partner, 59% were rural, 99% were non-Hispanic White, none identified as Hispanic, 57% had private insurance coverage, 3% were uninsured, and 59% had college degree or higher educational attainment.Among respondents with no delays, the mean age was 63.4 years, 53% were female, 80% were married or living with a partner, 49% were rural, 98% were non-Hispanic White, 50% had private insurance coverage, 3% were uninsured, and 59% had college degree or higher educational attainment.

| Predictors of cancer care delays
In univariate models, statistically significant predictors of any care delays were living in a rural area (  4).

| DISCUSSION
In this study, only 29% (170 of 585) of Iowa cancer patients reported they experienced delays to their cancer care in the early part of the pandemic, with routine appointments being the most frequently reported type of appointment cancelation (38% of appointments canceled).In multivariable-adjusted models, factors associated with any cancer-related appointment delays were rurality and experiencing feelings of loneliness; the only statistically significant factor associated with treatment delays was experiencing financial difficulties.We did not find that health insurance coverage, concern about the COVID-19 pandemic, and/or anxiety about catching or dying from COVID-19 were statistically significantly associated with healthcare delays in multivariable models.These findings indicate that while most Iowans did not experience during non-pandemic times, suggesting that the pandemic may have widened already existing disparities.
Our findings are consistent with other studies on cancer treatment delays in the early part of the pandemic.In two qualitative studies exploring the experience of cancer patients during the pandemic, delays in care presented as main themes. 23,24The frequency of individuals reporting delays ranged from 18% to 47% in two quantitative studies conducted in single centers. 25,26In studies stratifying delays by treatment type, delays in follow-up and patient management ranged from 44% to 74% while individuals reporting delays in surgery were around 14%. 7,12 Delays in care related to the COVID-19 pandemic were not specific to the U.S. healthcare experience: our results were also consistent with studies outside of the U.S. [10][11][12][13][14][15]21,[27][28][29][30][31][32][33] Two studies in the Philippines and Kenya reported between 33% and 42% of patients experienced general delays in care. 27 In stdies that stratified by treatment type, studies from Germany, Austria, Spain, and the Netherlands reported between 21% and 31% of patients experiencing delays for follow-up visits; 15,[28][29][30] studies from India, Germany, and China reported 9% to 52% patients experiencing cancer surgery delays; 9,10,29,32 and, studies from Spain, Italy, India, and China reported between 23% to 38% patients reporting delays in radiotherapy.8,28,32,33 Additionally, a majority of patients in three studies reported differences in median time to surgery during the pandemic compared to before the pandemic in the U.K. and Greece.34,35 This consistency demonstrated in national and global findings indicates the importance of addressing cancer care delays and in opportunities for learning from solutions identified in different global contexts.
In our study, those who experienced delays were typically younger in age, more often female (perhaps due to the large number of breast cancer patients present in our sample), more often rural, and experienced greater financial difficulties.2][43] There is a need for long-term monitoring of cancer surveillance data to track whether disparities maintain or continue to widen for individuals living in rural areas and of low socioeconomic status.This long-term monitoring should also prioritize a more racially diverse sample, as there was a lack of racial diversity in our study due to our mostly non-Hispanic White sampling frame.
This study has several limitations that warrant consideration when interpreting its findings.This study was crosssectional; therefore, we were unable to assess temporality in the relationship between COVID-19 patient experiences and care delays.Future studies temporally assessing delays to cancer care due to large-scale events are important for identifying mechanisms for delays and opportunities to intervene and minimize said delays.Additionally, our 26% response rate increases the potential of results being impacted by selection bias; those who responded may have different demographic and clinical characteristics than those who did not.However, the goal of this study was to assess COVID-19 delays at a particular academic medical center.Due to only living cancer patients being able to participate in the survey, there is a risk for results being impacted by survivor bias; however, it is unlikely that delays in patients who survived to participate in the survey would be different than delays among those who did not survive until the survey.
The sample of those invited to participate in the survey were those who were diagnosed or treated at an academic medical center, which may minimize the generalizability of these results.As respondents in this survey were diagnosed and treated at this one location, we could not examine variability in delays by facility, but rather we examined personal or social factors that may be associated with healthcare delays.Prior studies suggest that facility characteristics, specifically organizational-level factors such as centralization, patient outreach systems, and facility-level factors such as reductions in provider/service availability, are important predictors of cancer care delays and, thus, important to intervene on when attempting to minimize disparities. 11,22his study illuminates the experiences of patients at a Midwest cancer center during the early part of COVID-19.Since cancer treatment and follow-up appointments are integral parts of preventing poor cancer outcomes, these findings important insight into the experiences of cancer patients during the part of the pandemic that may have impacted longer-term outcomes among Iowa patients.We found that rurality and experiencing social isolation were the factors most strongly associated with delays in any type of cancer care, and experiencing financial difficulties was most strongly associated with delays in cancer treatment appointments.These findings can inform cancer care delivery in future destabilizing events that may impact cancer care, and inform potential areas for improvement during times when health systems are not stressed, with the goal of alleviating impacts during such destabilizing events.Future research should also assess whether delays in cancer care impact long-term cancer outcomes.
Demographic characteristics of University of Iowa Holden Comprehensive Cancer Center PERCH survey respondents, stratified by those who experienced and who did not experience cancer care delays (n = 657), 2020.Differences in responses to and impacts of the COVID-19 pandemic between those who did and did not experience treatment delays, University of Iowa Holden Comprehensive Center PERCH survey respondents (n = 657), 2020.
T A B L E 1Note: Rows with # were censored due to small cell counts (<5).Bold values are two-sided with p < 0.05 considered statistically significant.aOthercancer sites include liver and intrahepatic bile duct, urinary bladder, lymphoma, uterine, ovarian, leukemia, oral cavity and pharynx, soft tissue (including heart), lung, brain, esophagus, eye and orbit, stomach, bone and joints, small intestine, testicular, uteter, penile, and thyroid.bPercentageslisted are among all cancer sites, not just the top five reported sites.disturbances to their care, those that did were more impacted by proximity to care and financial concerns than concerns about COVID-19 itself.We were unable to determine whether this lack of concern was attributable to healthcare practices successfully mitigating fears (e.g., mandatory mask-wearing, restricting visitors to encourage physical distancing) or a general lack of concern about the COVID-19 pandemic in this mostly rural sample.Nevertheless, our findings indicate that barriers to care during the COVID-19 pandemic mirrored those facedT A B L E 2Note: Bold values are two-sided with p < 0.05 considered statistically significant.
Types of cancer care appointments delayed among University of Iowa Holden Comprehensive Cancer Center PERCH Survey respondents that identified cancer-related health care delays during COVID-19 restrictions (N = 170), 2020.Responses in the "other" category are those that were unable to be attributed to other appointment type categories.Univariate logistic regression models identifying predictors of delays in any cancer-related health care appointments and for cancer treatment appointments among University of Iowa Holden Comprehensive Cancer Center PERCH Survey respondents (n = 657), 2020.
bVariables included in the "treatment delays" variable.cNote:Variables with "-"were unable to be used in multivariable models due to convergence issues.Bold values are two-sided with p < 0.05 considered statistically significant.a Treatment delays include postponed or canceled surgery, chemotherapy, and radiation therapy appointments.

of any cancer care delays Predictors of treatment delays a
Multivariable-adjusted logistic regression models identifying predictors of delays in any cancer-related health care appointments and for cancer treatment appointments among University of Iowa Holden Comprehensive Cancer Center PERCH Survey respondents (n = 657), 2020.Variables with "-"were unable to be used in multivariable models due to convergence issues.Bold values are two-sided with p < 0.05 considered statistically significant.; the University of Iowa Holden Comprehensive Cancer Center P30 and P30 supplement grant funded the creation and distribution of the survey, and the NIH/ NCI grant funded the effort of research staff.We would also like to extend our gratitude to the University of Iowa Holden Comprehensive Cancer Center Community Advisory Board as well as the Community, Outreach, and Engagement group for their feedback throughout the survey creation process.
T A B L E 5Note: a Treatment delays include postponed or canceled surgery, chemotherapy, and radiation therapy appointments.publication