Health-related quality of life by race, ethnicity, and country of origin among cancer survivors

Abstract

Background

Typical cancer research studies of health-related quality of life (HRQOL) in the United States do not include country of origin when examining race and ethnic group differences. This population-based, cross-sectional study used an innovative methodology to examine how self-reported racial and ethnic groups, by country of origin, report differential HRQOL experiences after adjusting for clinical and demographic characteristics, including socioeconomic status.

Methods

Recruited from 4 cancer registries in California, Louisiana, and New Jersey, cancer survivors completed Patient-Reported Outcomes Measurement Information System measures of fatigue, pain interference, anxiety, depression, sleep disturbance, physical function, ability to participate in social roles, and cognitive function. Latent profile analysis clustered survivors in HRQOL clusters based on including all the Patient-Reported Outcomes Measurement Information System domains.

Results

The 5366 participants (60% female; 40% male; average age of 59.8 years) included 17% Asian, 18% Black, 21% Hispanic, and 41% White survivors. Survivors were grouped into 4 clusters: high HRQOL (26%), average HRQOL (34%), low HRQOL (29%), and very low HRQOL (11%). Among many differences by race, ethnicity, and country of origin, Caribbean cancer survivors were more likely to be in the very low HRQOL cluster (odds ratio = 2.67, 95% confidence interval = 1.31 to 5.43) compared with non-Hispanic White survivors. Similarly, American Indian and Alaska Native, Cuban, Dominican, and Puerto Rican cancer survivors had relatively high percentages in the very low HRQOL cluster.

Conclusions

This study found statistically significant differences in HRQOL experience by race, ethnicity, and country of origin, even after adjusting for social determinants of health. These findings inform future HRQOL research to include these self-reported factors.

The estimated number of cancer survivors in the United States is projected to increase from the current 18.1 million to 22.5 million by 2032 (1,2). Thus, it is important to identify contributing factors to survivors’ overall well-being (3). Cancer survivors indicate optimizing physical, mental, and social aspects of their health-related quality of life (HRQOL) is important during and after cancer treatment (4). HRQOL is influenced by treatment and factors at the level of the individual (health behaviors, cultural identity, spirituality), interpersonal (social support), community (access to specialists, rural or urban setting), and society (systemic barriers to care, health-care laws) (3,5).

Robust evidence has identified racial and ethnic health disparities across the cancer survivorship continuum (1,6-9). Recent studies and a 2019 National Cancer Institute Survivorship Workshop have called for research to elucidate multi-level influences on HRQOL outcomes among survivors from diverse races, ethnicities, and countries of origin (10-14). To gain insight into diverse HRQOL experiences, large samples are needed to allow subgroup analysis by demographic characteristics, including country of origin, within larger subpopulations of race and ethnic groups. For example, are there differences in HRQOL over the survivorship trajectory between Black survivors born in the Caribbean compared with Black survivors born in the United States (15)?

We are unaware of studies that have examined HRQOL clusters in a large cohort of adult cancer survivors across the United States from diverse countries of origin who had received treatment for 1 of 7 distinct invasive cancers. The goal of our study was to examine how race, ethnicity, and country of origin vary in relation to cancer survivors’ HRQOL experiences when accounting for clinical and demographic characteristics, including social determinants of health (SDOH).

Methods

Participants and data collection procedures

Adult participants were enrolled in the Measuring Your Health Study, a large, cohort-based study from 4 population-based Surveillance, Epidemiology, and End Results (SEER) cancer registries in 3 states (California, 2 registries; Louisiana; New Jersey). Participants’ diagnoses included 1 of 7 different cancer types: prostate, colorectal, non-small cell lung, non-Hodgkin’s lymphoma, female breast, uterine, or cervical. Participants were initially surveyed between 6 and 13 months after their incident cancer diagnosis [details in Jensen et al. (16)]. The study received approval by institutional review boards at all collaborating sites.

Participants received a mailed baseline survey to be completed in either English, Spanish, or Mandarin Chinese. They returned the survey in a stamped, preaddressed envelope. Individuals who did not respond were contacted by the study team and invited to complete the survey by telephone in English, Spanish, or Mandarin Chinese. Stratified sampling was used to achieve balanced representation across 3 age groups (21-49 years, 50-64 years, 65-84 years) and 4 racial and ethnic groups (Hispanic or Latino, non-Hispanic Asian or Pacific Islander, non-Hispanic Black, and non-Hispanic White). Participants received a $30 incentive. Data were collected between 2010 and 2012.

Measures

Patient characteristics

Clinical data were abstracted from collaborating cancer registry databases including cancer diagnosis date, cancer type and stage, and receipt of initial surgery and radiation therapy. On the survey, participants were asked to self-report if they were of “Hispanic, Latino, or Spanish origin” and, if “yes,” were asked to indicate if they were “Cuban,” “Dominican,” “Mexican, Mexican American, Chicano,” “Puerto Rican,” or “another Hispanic, Latino, or Spanish origin.” Participants also self-reported their race, marking all that applied, for categories of “American Indian or Alaska Native,” “Asian Indian,” “Black, African American, or Negro,” “Chinese,” “Filipino,” “Guamanian or Chamorro,” “Japanese,” “Korean,” “Native Hawaiian,” “Samoan,” “Vietnamese,” “White,” “Other Asian,” “Other Pacific Islander,” or “Some other race” (allowing write-in of their race). In addition, participants self-reported country of origin, number of years in the United States, age at time of immigration, education status, sex, marital status, financial variables (eg, health-care coverage, financial well-being) (17), health behaviors (eg, smoking status, physical activity, body mass index), social support (ability to find companionship when needed), spirituality (18,19), and history of comorbid conditions. Participants self-reported receipt of chemotherapy or hormonal therapy because this information is incomplete in most cancer registries.

For the primary analyses, we required each racial and ethnic group to have more than 100 individuals, resulting in non-Hispanic Black, non-Hispanic White, Mexican, Caribbean, other Hispanic, Chinese, Filipino, other Asian, and other non-Hispanic and non-Asian groups (see Table 1). Subgroup analyses summarized HRQOL findings in smaller groups that were combined, including American Indian or Alaska Native, Caribbean (Puerto Rican, Cuban, Dominican), other Asian (Japanese, Korean, Vietnamese, and Indian), and Pacific. The other Hispanic group included a mix of countries with sample sizes smaller than 25, including Argentina, Brazil, Chile, Colombia, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, and Peru. Racial and ethnicity data were self-reported and recognized as social constructs (20-22).

Table 1.

Demographic and clinical characteristics of cancer survivors

		Asian cancer survivors			Hispanic cancer survivors			Non-Hispanic cancer survivors
	Total	Chinese	Filipino	Other Asiana	Caribbeanb	Mexican	Other Hispanic	Black	White	Other racec
Variable	(n = 5366)	(n = 306)	(n = 268)	(n = 315)	(n = 139)	(n = 688)	(n = 309)	(n = 977)	(n = 2216)	(n = 148)
Mean age (SD), y	59.8 (12.5)	58.4 (12.7)	58.0 (11.3)	58.2 (12.8)	56.6 (12.3)	57.7 (12.8)	58.1 (12.8)	59.2 (11.4)	62.0 (12.4)	57.7 (13.7)
Sex, %
Female	59.8	64.1	72.4	63.8	60.4	60.0	63.1	56.2	58.1	60.1
Male	40.2	35.9	27.6	36.2	39.6	40.0	36.9	43.8	41.9	39.9
Born in United States, %
Yes	71.3	21.0	13.5	27.2	36.8	57.3	28.0	94.9	94.6	58.5
Marital status, %
Married or living with partner	62.5	75.3	71.2	70.5	58.2	57.0	58.8	46.3	68.4	58.6
Never married, separated, divorced, or widowed	37.5	24.7	28.8	29.5	41.8	43.0	41.2	53.7	31.6	41.4
Education level, %
Less than HS degree	17.1	18.1	5.7	10.9	34.5	41.4	28.3	19.8	7.7	18.8
HS degree	19.5	10.1	10.6	11.6	18.7	20.3	17.7	21.7	22.1	20.8
Some college	19.3	8.7	14.4	14.1	18.0	17.4	16.3	23.7	21.5	16.0
College degree	32.0	36.2	57.6	39.9	22.3	17.3	28.7	27.3	34.9	29.9
Graduate degree	12.1	26.8	11.7	23.5	6.5	3.6	9.0	7.4	13.8	14.6
Employment status, %
Full-time	26.8	29.3	34.3	28.6	29.2	22.7	22.3	26.1	27.6	21.4
Part-time	8.6	10.0	8.3	8.0	7.3	7.5	12.0	5.9	9.7	9.0
Full-time homemaker, family caregiver, or student	7.7	7.3	7.2	9.6	8.0	11.5	13.3	5.3	6.8	5.5
Unemployed but seeking employment	6.6	7.3	7.9	7.4	11.7	8.3	7.6	8.6	4.1	10.3
Retired	37.3	35.3	29.4	36.3	25.5	32.3	30.6	37.2	41.8	38.6
Other	13.0	10.7	12.8	10.0	18.2	17.6	14.3	16.9	9.9	15.2
Insurance type, %
Private/private + government	68.1	69.9	67.3	71.1	53.6	54.1	51.2	59.2	78.9	65.7
Government/no insurance	31.9	30.1	32.7	28.9	46.4	45.9	48.8	40.8	21.1	34.3
Smoker status, %
Current	10.4	3.7	4.9	3.9	15.9	9.4	9.1	16.6	10.3	13.2
Former	33.8	18.5	18.3	19.9	29.7	26.3	25.6	29.7	45.6	30.6
Never	55.7	77.8	76.8	76.2	54.3	64.3	65.3	53.8	44.2	56.3
Cancer type, %
Breast	30.5	43.5	47.4	41.9	40.3	27.2	33.3	28.6	25.9	32.4
Cervix	2.7	0.7	4.1	1.3	3.6	5.4	3.2	1.9	2.2	4.7
Colorectal	16.9	19.9	14.2	21.3	12.2	18.8	16.5	17.7	15.7	16.2
Lung	12.7	3.6	4.9	7.0	4.3	5.5	6.8	11.6	19.9	9.5
Non-Hodgkins lymphoma	8.3	6.5	4.5	7.6	7.2	8.7	8.1	4.9	10.4	10.8
Prostate	21.7	18.0	17.5	15.2	26.6	23.7	21.4	28.5	19.8	20.3
Uterus	7.2	7.8	7.5	5.7	5.8	10.8	10.7	6.9	6.1	6.1
Cancer stage, %
In situ/stage I	38.3	42.1	40.2	37.5	46.0	35.4	39.6	34.3	39.9	34.3
Stage II	32.6	35.5	33.7	35.5	32.5	34.6	32.8	36.1	29.3	34.3
Stage III	17.5	12.7	18.4	17.1	12.7	18.0	17.4	17.8	18.0	17.5
Stage IV	11.7	9.7	7.7	9.9	8.7	12.0	10.2	11.8	12.8	14.0
Treatment type, %
Local therapy only: surgery and/or RT	35.6	34.3	32.5	30.8	32.4	33.9	34.3	37.6	37.0	34.5
Adjuvant systemic: surgery and CT, with or without HT	15.7	17.0	22.4	16.5	18.7	16.1	16.5	16.7	13.9	12.2
Adjuvant systemic: RT and CT, with or without HT	8.2	2.6	5.2	5.4	7.2	9.3	8.4	7.6	9.8	6.8
Adjuvant systemic: surgery, RT, CT, with or without HT	16.7	21.6	19.0	20.3	19.4	18.3	21.4	16.1	13.9	19.6
Adjuvant systemic: surgery and/or RT, with HT alone	10.8	15.4	10.8	14.6	10.8	9.4	8.4	9.3	11.0	12.2
Systemic therapy only: CT and/or HT	8.2	5.6	5.6	6.7	7.9	8.4	8.1	7.7	9.3	8.8
No therapy	4.8	3.6	4.5	5.7	3.6	4.5	2.9	5.1	5.1	6.1
SEER region, %
Greater California	33.9	17.3	44.4	42.9	11.5	59.2	35.9	23.2	30.9	45.3
Greater San Francisco Bay	22.7	73.9	42.2	39.4	5.0	24.0	27.5	15.5	13.6	31.1
Louisiana	19.8	0.3	0.4	2.2	2.2	10.5	6.1	40.5	25.0	5.4
New Jersey	23.6	8.5	13.1	15.6	81.3	6.4	30.4	20.8	30.5	18.2
Heart-related condition (heart attack, heart failure, or stroke), %
Yes	13.0	5.6	9.3	8.6	10.1	11.5	13.9	15.1	14.4	17.6
Lung-related condition (asthma or COPD), %
Yes	19.8	9.2	14.6	9.2	30.9	16.9	17.2	19.7	23.8	21.6
Mental health-related condition (depression or anxiety), %
Yes	26.3	13.1	20.5	20.0	41.7	30.7	29.1	24.0	27.8	29.1
Sleep disturbance condition, %
Yes	14.5	6.9	11.9	15.2	18.7	18.0	20.7	13.4	13.9	15.5
Companionship, %
Never/rarely	11.2	19.6	7.6	11.1	14.9	13.7	14.3	15.3	7.0	16.7
Sometimes	16.9	17.9	15.5	15.7	22.4	17.8	23.3	22.9	13.1	17.4
Often/always	71.8	62.5	76.9	73.2	62.7	68.5	62.3	61.8	79.9	66.0
Financial well-being, mean (SD)	59.6 (26.4)	60.7 (24.2)	8.5 (25.9)	62.4 (24.7)	49.4 (24.9)	5.3 (25.6)	52.9 (26.2)	56.7 (28.0)	63.9 (25.8)	50.6 (25.8)
Spiritual well-being, mean (SD)	37.2 (9.4)	34.4 (9.9)	40.0 (8.4)	36.0 (9.9)	36.9 (9.9)	37.7 (8.9)	38.6 (9.0)	39.5 (8.7)	36.1 (9.7)	36.7 (9.4)

^aOther Asian = Japanese (n = 83), Korean (n = 38), Vietnamese (n = 60), Asian Indian (n = 104), Other Asian (n = 52). CT = chemotherapy; COPD = chronic obstructive pulmonary disease; HS = high school; HT = hormonal therapy; RT = radiation therapy; SEER = Surveillance, Epidemiology, and End Results Program.

^bCaribbean = Puerto Rican (n = 86), Cuban (n = 25), Dominican (n = 28).

^cOther = Other non-Hispanic race or multiple race (American Indian or Alaska Native [n = 52], Pacific islander [n = 34], some other race, or multiple race).

Outcome measures

The Patient-Reported Outcomes Measurement Information System (PROMIS) measures included the short-form versions of fatigue, pain interference, anxiety, depression, sleep disturbance, physical function, ability to participate in social roles (social function), and cognitive function (23). These HRQOL domains were selected by investigators based on their relevance and impact on cancer patients. All PROMIS T-scores are calibrated to a general US population with a mean of 50 and an SD of 10 points (24). The exception is sleep disturbance, which is calibrated on a mixed sample of the US population and those with clinical conditions. Higher PROMIS function T-scores represent better functioning, and higher PROMIS symptom T-scores reflect greater symptom burden. Differences of 5 points (1/2 SD) or more are considered a clinically meaningful difference between groups (25,26).

Statistical analyses

We used latent profile analysis (LPA) to identify distinct clusters of cancer patients with similar mean scores across the HRQOL indicators (27-29). We progressively increased the number of clusters of survivors (starting with 2 clusters) and selected the ideal number of clusters based on model goodness-of-fit statistics, at least 10% survivors in each cluster, and clinical interpretability (28,30). Goodness-of-fit statistics included Akaike information criterion, Bayesian information criterion (BIC), adjusted BIC, Vuong-Lo-Mendell-Rubin likelihood ratio test, Lo-Mendell-Rubin adjusted likelihood ratio test, and entropy. Lower Akaike information criterion, BIC, and adjusted BIC values indicate better fit. Both the Vuong-Lo-Mendell-Rubin likelihood ratio and the Lo-Mendell-Rubin adjusted likelihood ratio tests compare 1 nested LPA model (eg, 3 clusters) with another LPA model with 1 additional cluster (eg, 4 clusters); statistically significant improvement (P <.05) suggests the model with more clusters reflects better fit (31). Entropy is a measure of uncertainty in the posterior classifications of the model, with higher entropy values reflecting less uncertainty (32). Clinical interpretability was conducted among the research team (including an oncology clinical nurse specialist and a clinical health psychologist) based on a review of the profiles. In the full sample, only 7 missing data patterns were present and deemed negligible for the LPA covariance coverage among the 8 PROMIS HRQOL indicators. LPA was performed with M-plus (version 8.2).

We used a multinomial logistic regression model to determine whether the demographic and clinical factors were associated with HRQOL cluster membership. The model included race and ethnic group, born in the United States (no/yes), survey language, age at diagnosis, education, employment status, insurance status, marital status, smoking status, cancer type, SEER site, Derived American Joint Committee on Cancer stage, treatment modality, heart-related condition, lung-related condition, mental health-related condition, sleep disturbance condition, companionship, financial capability, and spirituality. We also examined models with race and ethnicity group interacting with other key factors, including born in the United States, survey language, age at diagnosis, education, employment status, Derived American Joint Committee on Cancer stage, heart-related condition, and financial capability. Initially, an omnibus likelihood ratio test was used to determine if all the interactions together were jointly statistically significant. If the omnibus test showed statistical significance, a multinomial logistic model was fit to the data with the statistically significant interaction terms, and the final model included all the main effects and interaction terms that were statistically significant. Results from both the main effects model and the final model with statistically significant interactions were reported. We summarized the results using model-adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Approximately 17% of all observations had some missing data for explanatory variables. Missing at random was assumed, and we used all available data to estimate the models using full information maximum likelihood. All statistical analyses, except LPA, were conducted using SAS (version 9.4).

Results

Participants

The participation rate among those contacted was 53%, and response rate was 36%. The enrolled cohort was representative of all diagnosed cancer cases in the participating SEER regions, with slightly lower response rates for patients over age 75 years, males, and Blacks (16). Table 1 provides the demographic and clinical characteristics for the 5366 cancer survivors by key racial and ethnicity groups. The total sample included approximately 17% Asian, 21% Hispanic, 18% non-Hispanic Black, and 41% non-Hispanic White cancer patients. The sample was 60% female, 40% male, were a median of 9 months from diagnosis, and ranged from 21 to 84 years of age (mean 59.8 [SD 12.5] years), with 60% of survivors younger than 65 years of age. Supplementary Table 1 provides the average PROMIS HRQOL T-scores across the race and ethnic groups.

HRQOL clusters

The model fit indices of LPA favored more than 4 HRQOL clusters; however, this resulted in one or more clusters with less than 10% of the total sample. The 4-cluster solution (high, average, low, very low) provided a parsimonious, clinically interpretable, and accurately separable solution (entropy = 0.85). As seen in Figure 1, the high-HRQOL cluster included approximately 26% of the sample and reported high functioning and low symptom burden, approximately 10 points (1 SD) better than the US general population. The average-HRQOL cluster included approximately 34% of the sample with functioning and symptom burden levels similar to the US general population. The low-HRQOL cluster included approximately 29% of the sample, with functioning and symptom burden levels approximately 5 points worse than the general US population. The very low-HRQOL cluster included approximately 11% of the sample, with functioning and symptom burden levels approximately 15 points (1.5 SD) worse than the general US population.

Figure 1.

PROMIS (Patient-Reported Outcomes Measurement Information System) T-scores by health-related quality of life (HRQOL) cluster.

Racial and ethnicity group distribution by HRQOL cluster

Table 2 presents the distribution of survivors by largest race and ethnicity groups across the 4 HRQOL clusters and identifies if a group is overrepresented or underrepresented in the HRQOL cluster relative to the reference group of non-Hispanic White survivors in the high-HRQOL group. If not flagged as over- or underrepresented in Table 2, then the percentages by racial and ethnic group within an HRQOL cluster were similar to the reference group after adjusting for demographic and clinical characteristics. Black (OR = 0.48, 95% CI = 0.35 to 0.65), Caribbean (OR = 0.46, 95% CI = 0.22-0.96), Chinese (OR = 0.47, 95% CI = 0.28 to 0.79), Filipino (OR = 0.33, 95% CI = 0.20 to 0.57), Mexican (OR = 0.38, 95% CI = 0.26 to 0.56), and Other Hispanic (OR = 0.32, 95% CI = 0.19 to 0.54) cancer survivors were less likely to be in the average-HRQOL cluster than White cancer survivors. Black (OR = 0.73, 95% CI = 0.57 to 0.94), Chinese (OR = 0.64, 95% CI = 0.42 to 0.99), Filipino (OR = 0.56, 95% CI = 0.36 to 0.88), Mexican (OR = 0.38, 95% CI = 0.26 to 0.56), and Other Hispanic (OR = 0.52, 95% CI = 0.34 to 0.79) cancer survivors were less likely to be in the low-HRQOL cluster than White cancer survivors. Caribbean cancer survivors were more likely to be in the very low-HRQOL profile (OR = 2.67, 95% CI = 1.31 to 5.43) compared with White cancer survivors.

Table 2.

Distribution of major racial and ethnic groups across HRQOL clusters including tests for over- or underrepresentation in each cluster after adjusting for demographic and clinical characteristics

Race and ethnic group	High HRQOL No. (%)	Average HRQOL No. (%)	Low HRQOL No. (%)	Very low HRQOL No. (%)
Asian cancer survivors
Chinese	82 (26.8)	113 (36.9)f	91 (29.7)f	20 (6.5)
Filipino	67 (25.0)	104 (38.8)f	69 (25.8)f	28 (10.5)
Other Asiana	91 (28.9)	122 (38.7)	73 (23.2)	29 (9.2)
Hispanic cancer survivors
Caribbeanb	25 (18.0)	34 (24.5)f	42 (30.2)	38 (27.3)g
Mexican	140 (20.4)	218 (31.7)f	233 (33.9)f	97 (14.1)
Other Hispanicc	63 (20.4)	97 (31.4)f	112 (36.3)f	37 (12.0)
Non-Hispanic cancer survivors
Black	237 (24.3)	319 (32.7)f	303 (31.0)f	118 (12.1)
Whited	650 (29.3)	781 (35.2)	584 (26.4)	201 (9.1)
Other racee	29 (19.6)	48 (32.4)f	51 (34.5)	20 (13.5)

^aOther Asian included Japan, Korea, Vietnam, India, and other Asian countries. HRQOL = health-related quality of life.

^bCaribbean included Puerto Rico, Cuba, and Dominican Republic.

^cOther Hispanic group included a mix of countries with sample sizes less than 25, including Argentina, Brazil, Chile, Colombia, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, and Peru.

^dNon-Hispanic Whites are the reference group for the regression model. The model adjusted for age, born in the United States, survey language, education, employment, insurance, marital status, smoking status, cancer type, stage of disease, treatment, Surveillance, Epidemiology, and End Results site, heart-related condition, lung-related condition, mental health-related condition, sleep condition, companionship, financial well-being, and spiritual well-being.

^eOther race included Native American Indian, Alaska Natives, Pacific islands, and other or multiple races.

^fThis racial group is less likely to be represented in the HRQOL profile compared with high HRQOL for Non-Hispanic Whites (reference group).

^gThis racial group is more likely to be represented in the HRQOL profile compared with high HRQOL for Non-Hispanic Whites (reference group).

The omnibus test of interaction terms was statistically significant, with race and ethnicity group interacting both with birth in the United States (P < .01) and with financial well-being (P < .01). However, the financial well-being interaction was statistically non-significant (P = .06) when the other statistically non-significant interaction terms were removed from the model. In the final model, the interaction of race and ethnicity group with being born in the United States was included and statistically significant (P = .03). Black survivors not born in the United States were less likely to be in the average-HRQOL cluster (OR = 0.21, 95% CI = 0.08 to 0.57) and low-HRQOL cluster (OR = 0.26, 95% CI = 0.11 to 0.65) compared with Black survivors born in the United States (average: OR = 0.50, 95% CI = 0.37 to 0.69; low: OR = 0.78, 95% CI = 0.61 to 1.02). Chinese survivors not born in the United States were less likely to be in the very low-HRQOL cluster (OR = 0.78, 95% CI = 0.32 to 1.86) than Chinese survivors born in the United States (OR = 1.96, 95% CI = 0.63 to 6.09). Other Hispanic groups not born in United States were less likely to be in the average-HROQL cluster (OR = 0.27, 95% CI = 0.15 to 0.51) than Other Hispanic groups born in the United States (OR = 0.82, 95% CI = 0.36 to 1.83).

For the model that included the race and ethnicity group interaction term with being born in the United States, Table 3 displays the covariates statistically significantly associated with HRQOL cluster membership. Covariates that were important to HRQOL cluster membership that could be potential targets for screening or intervention (eg, identifying those survivors at greatest risk for very low-HRQOL profiles) include prior history of a mental health condition (OR = 2.57, 95% CI = 1.97 to 3.35), sleep disturbance (OR = 2.10, 95% CI = 1.58 to 2.97), lower compared with greater financial well-being (OR = 0.94, 95% CI = 0.88 to 0.99), and lower compared with greater spiritual well-being (OR = 0.81, 95% CI = 0.76 to 0.86).

Table 3.

Demographic and clinical covariates associated with HRQOL cluster

Covariatea	High HRQOL [Ref]	Average HRQOL OR (95% CI)	Low HRQOL OR (95% CI)	Very low HRQOL OR (95% CI)
Survey: Spanish/Chinese vs English	—	—	—	—
Education: graduate vs college	—	—	1.48 (1.10 to 1.98)	—
Education: high school vs college	—	—	—	—
Education: less high school vs college	—	—	—	1.77 (1.19 to 2.62)
Education: some college vs college	—	—	—	—
Employment: work at home vs full time	—	0.53 (0.35 to 0.81)	0.53 (0.38 to 0.75)	—
Employment: other vs full time	—	0.20 (0.13 to 0.32)	0.44 (0.32 to 0.60)	2.48 (1.60 to 3.83)
Employment: retired vs full time	—	0.59 (0.43 to 0.80)	0.61 (0.47 to 0.79)	—
Unemployed vs full time	—	—	—	—
Employment: part time vs full time	—	—	—	—
Insurance: government/none vs private	—	0.75 (0.58 to 0.96)	—	—
Not married vs married/partner	—	—	—	—
Smoking status	—	—	—	—
Breast vs colorectal	—	0.55 (0.39 to 0.78)	—	—
Cervix vs colorectal	—	—	—	2.84 (1.37 to 5.85)
Lung vs colorectal	—	0.53 (0.34 to 0.82)	—	—
Non-Hodgkin’s lymphoma vs colorectal	—	—	—	—
Prostate vs colorectal	—	2.08 (1.42 to 3.03)	—	—
Uterus vs colorectal	—	—	—	—
DAJCC stage: II vs I/in situ	—	—	—	1.57 (1.12 to 2.20)
DAJCC stage: III vs I/in situ	—	0.65 (0.47 to 0.92)	—	—
DAJCC stage: IV vs I/in situ	—	0.61 (0.40 to 0.94)	—	1.83 (1.15 to 2.91)
Adjuvant systemic (radiation and chemo) vs local therapy	—	0.31 (0.19 to 0.51)	0.49 (0.33 to 0.71)	1.72 (1.06 to 2.79)
Adjuvant systemic (surgery and chemo) vs local therapy	—	0.26 (0.18 to 0.37)	0.53 (0.39 to 0.71)	—
Adjuvant systemic (surgery and/or radiation, hormonal) vs local therapy	—	0.64 (0.45 to 0.91)	—	—
Adjuvant systemic (surgery, radiation, chemo) vs local therapy	—	0.28 (0.19 to 0.40)	0.53 (0.39 to 0.71)	—
No therapy vs local therapy	—	—	—	—
Systemic therapy vs local therapy	—	0.39 (0.24 to 0.64)	0.43 (0.29 to 0.64)	—
SEER site	—	—	—	—
Heart-related condition (yes vs no)	—	0.37 (0.27 to 0.52)	0.63 (0.49 to 0.80)	—
Lung-related condition (yes vs no)	—	0.52 (0.39 to 0.68)	0.70 (0.56 to 0.87)	—
Mental health condition (yes vs no)	—	0.29 (0.22 to 0.38)	0.51 (0.41 to 0.62)	2.57 (1.97 to 3.35)
Sleep disturbance (yes vs no)	—	0.33 (0.23 to 0.47)	0.73 (0.56 to 0.93)	2.10 (1.58 to 2.97)
Companionship (never/rarely vs often/always)	—	0.44 (0.29 to 0.65)	0.55 (0.41 to 0.74)	—
Companionship (sometimes vs often/always)	—	0.23 (0.16 to 0.32)	0.55 (0.44 to 0.68)	—
Age at diagnosis	—	—	1.05 (1.00 to 1.10)	—
Financial well-beingb	—	1.25 (1.18 to 1.32)	1.12 (1.07 to 1.17)	0.94 (0.88 to 0.99)
Spiritual well-beingb	—	1.82 (1.71 to 1.95)	1.38 (1.32 to 1.45)	0.81 (0.76 to 0.86)

^aFor each covariate, we provide the comparison group (note “vs”) when the variable is categorical. Racial and ethnic group interaction with “born in the United States” variables were included in the model and reported in the text. Only statistically significant associations are provided in the table. The high-HRQOL cluster was the reference group (“Ref”) for comparisons. CI = confidence interval; DAJCC = Derived American Joint Committee on Cancer stage; HRQOL = health-related quality of life; OR = model-adjusted odds ratio; SEER = Surveillance, Epidemiology, and End Results.

^bPer half SD.

Table 4 provides the distribution of specific racial and ethnicity subgroups by HRQOL cluster that are often not reported individually. American Indian and Alaska Native, Cuban, Dominican, and Puerto Rican cancer survivors have relatively high percentages in the very low-HRQOL cluster. Survivors who identified as being from India, Japan, and Vietnam have relatively high percentages in the low-HRQOL cluster.

Table 4.

Distribution of racial and ethnic groups across HRQOL clusters

Racial-ethnic group	Total No.	High HRQOL No. (%)	Average HRQOL No. (%)	Low HRQOL No. (%)	Very low HRQOL No. (%)
Asian cancer survivors
Chinese	312	91 (29.2)	86 (27.6)	115 (36.9)	20 (6.4)
Filipino	288	78 (27.1)	68 (23.6)	111 (38.5)	31 (10.8)
Indian	104	22 (21.2)	28 (26.9)	42 (40.4)	12 (11.5)
Japanese	83	17 (20.5)	29 (35.0)	35 (42.2)	2 (2.4)
Korean	38	12 (31.6)	11 (29.0)	13 (34.2)	2 (5.3)
Vietnamese	60	15 (25.0)	13 (21.7)	24 (40.0)	8 (13.3)
Other Asian	52	15 (28.9)	12 (23.1)	17 (32.7)	8 (15.4)
Hispanic cancer survivors
Cuban	25	6 (24.0)	8 (32.0)	5 (20.0)	6 (24.0)
Dominican	28	10 (35.7)	5 (17.8)	6 (21.4)	7 (25.0)
Mexican	560	190 (33.9)	117 (20.9)	176 (31.4)	77 (13.8)
Puerto Rican	86	26 (30.2)	12 (14.0)	23 (26.7)	25 (29.1)
Other Hispanica	309	112 (36.3)	63 (20.4)	97 (31.4)	37 (12.0)
Non-Hispanic cancer survivors
American Indian or Alaska Native	52	13 (25.0)	8 (15.4)	18 (34.6)	13 (25.0)
Non-Hispanic Black	977	303 (31.0)	237 (24.3)	319 (32.7)	118 (12.1)
Pacific Islander	34	13 (38.2)	5 (14.7)	13 (38.2)	3 (8.8)
Non-Hispanic White	2216	584 (26.4)	650 (29.3)	781 (35.2)	201 (9.1)

^aOther Hispanic group included a mix of countries with sample sizes less than 25, including Argentina, Brazil, Chile, Colombia, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, and Peru. HRQOL = health-related quality of life.

Discussion

These findings uniquely contribute to the cancer survivorship literature by showing representation of diverse racial and ethnicity groups, further identified by country of origin, across different HRQOL clusters in a large cohort that included 7 cancer types and younger and older cancer survivors. These groups are typically underrepresented in cancer survivorship research. We are aware of only a few previous studies that examined HRQOL and survivorship outcomes among individuals living in the United States who are of American Indian or Alaskan Native, Caribbean, or Philippine identities (33). In contrast to other studies, we modeled multiple HRQOL domains simultaneously using the LPA approach.

Within the racial and ethnic subgroup findings, individuals who identified as Alaskan Native, American Indian, or Caribbean reported worse HRQOL. This supports the importance of our analysis of detailed self-reported racial and ethnic group identification, inclusive of smaller demographic groups that are typically not examined separately but rather placed into the “other” race category (1). The poor HRQOL outcomes among these groups highlight the need for oncology providers to screen for risk factors of low HRQOL and provide referrals to appropriate support services.

We also explored factors associated with HRQOL clusters that may be amenable to screening or intervention or that may identify at-risk subgroups for tailoring interventions (22). Consistent with previous studies, we identified factors that are often categorized as SDOH, including financial well-being (34), insurance type, employment status, and level of education. We observed no interaction of racial and ethnic group with these mutable factors, demonstrating that both racial and ethnic group membership and SDOH independently influence HRQOL in the presence of each other and that SDOH operates as an important means of stratifying at-risk subgroups in the survivor population across all racial and ethnic groups. HRQOL research with cancer survivors needs to be cautious about conflating race and ethnicity with SDOH (35).

Other results necessitating further investigation include the portion of survivors (13%) who reported “other” for their employment status, with a large number of write-in participant responses indicating survivors were on disability or medical leave. Future work can explore how these employment categories affect distinct HRQOL clusters in cancer survivors (36,37).

Several important clinical implications for HRQOL assessment were found. Early, accurate identification of cancer survivors at risk for impaired HRQOL is needed over their disease trajectory to promote appropriate, timely, and patient-specific interventions. Thus, valid and reliable HRQOL assessment and prediction tools for cancer patients are needed that have clinical utility in diverse inpatient and ambulatory oncology clinical settings, including community settings (38). Risk prediction tools, based on key factors from our model, can help to place an individual with likelihood in high-, average-, low-, and very low-HRQOL profiles. Future research is needed to test the sensitivity and specificity of HRQOL risk prediction tools produced by this LPA methodology. The success for integration of these HRQOL screeners and prediction tools relies on effective communication among the primary care provider and the oncology multidisciplinary team to ensure that HRQOL data are collected per protocol or per standard care, do not place undue burden on the patient, and promote appropriate referral to palliative care or psychological services as needed (39).

The science will be advanced by pivoting from studies designed to provide description of disparities to intervention studies that will build the evidence base for strategies to mitigate and, ultimately, eliminate health disparities in HRQOL in cancer survivors. Targeted research studies are needed to elucidate the complex relationships between and among reported components of HRQOL, SDOH, and cancer patients’ race, ethnicity, country of origin, age, and language preference (6,22). Potential interventions include routine use of self-report HRQOL screeners and prediction models to identify cancer survivors at risk for poor HRQOL (38). Treatments like cognitive behavioral therapy focused on distress, pain, and sleep may affect other HRQOL domains for cancer survivors (40-43).

Although we were able to conduct subgroup analyses for a number of specific populations, limitations of the study include the small samples for certain race and ethnic groups. Survivors self-reported receipt of chemotherapy and hormonal therapy. We do not have data regarding where patients received treatment, but the SEER cancer registries record treatments delivered outside their own geographic catchment area, even if in another state or country. The US survey was available only in English, Mandarin Chinese, and Spanish languages. The employment status findings showed a large number of participants selecting the response “other” and indicating that they were disabled and/or on medical leave.

Importantly, our study did not capture other types of care-related and health-care system factors that might also influence HRQOL differences among cancer survivors. Although we did observe that general financial well-being was associated with outcomes, we lacked other details on this topic, such as financial limitations specifically related to receipt of palliative care services like pain management or psychosocial care. The study also did not capture the systemic impact of institutional racism, cultural racism, and interpersonal discrimination in the United States that may lead to inadequate access to care and treatment and worse patient outcomes (22,44-54). Lastly, our data were collected between 2010 and 2012, and new cancer treatments have become available. However, multiple influences on HRQOL impact, including SDOH, likely still remain.

In conclusion, we used rigorous analytic techniques to identify HRQOL clusters in a large, heterogenous cohort of individuals diagnosed with 1 of 7 cancers. Routine screening of patient-reported outcomes can identify patients at high risk for poor HRQOL throughout active treatment and into longer-term survivorship (55,56). The more than 18 million US cancer survivors make it imperative to focus on sustained research efforts that will yield effective and scalable interventions to improve HRQOL.

Funding

This work was supported by the following grants from the National Institutes of Health (NIH): R01 NR018841 and U01 AR057971. This research was also supported by the Survivorship Research Initiative of the Georgetown Lombardi Comprehensive Cancer Center (P30-CA051008).

Notes

Role of the funder: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Author disclosures: The authors declare no competing interests.

Author contributions: Conceptualization: BBR, KDG, ALP, JMFD. Data curation: LL, ALP. Formal Analysis: LL, BBR, JA, WP. Funding Acquisition: ALP. Investigation: BBR, KDG, LL, ALP, JA, DMH, WP, JMFD. Methodology: BBR, LL, ALP, JA, WP. Project Administration: DMH. Supervision: BBR, ALP, JMFD. Writing—original draft: BBR, KDG, ALP, JMFD. Writing—review and editing: BBR, KDG, LL, ALP, JA, DMH, WP, JMFD.

Acknowledgements: We thank Ms Felice Yang, MPH and Ms Tania Lobo, MS of the Survey, Recruitment, and Biospecimen Collection Shared Resource (SRBSR) of the Georgetown University Medical Center for their help with data management. We are grateful to Liz Wing, MA, who provided invaluable editorial support. We also thank the four participating SEER cancer registries who collaborated on the MY-Health study: the Cancer Registry of Greater California, the Greater Bay Area Cancer Registry, the Louisiana Tumor Registry, and the New Jersey State Cancer Registry.

Data availability

The data analyzed in this study are available in the Harvard Dataverse for HealthMeasures at: https://doi.org/10.7910/DVN/XD1A6B.