Predictors of Colonoscopy Use Among Asian Indians in New York City, 2003-2016

Background: Indians are among the most under-screened Asian American subgroups, but there is limited data for this population. Methods: Using 2003-2016 the New York City Community Health Survey, we examined the association between sociodemographic and medical factors and up-to-date colonoscopy (de�ned as colonoscopy within the last 10 years) in Asian Indians using logistic regression over four time periods: 2003-2008, 2009-2012, 2013-2014, 2015-2016. Results: On multivariable analysis, only language was associated with colonoscopy uptake in more than one time period. Compared to participants who preferred English, those who preferred an “Other” language were less likely to be up-to-date on colonoscopy in 2003-2008 (OR 0.15, 95% CI 0.04-0.66), while those who preferred an Indian language were less likely to have been up-to-date in 2013-2014 (OR 0.12, 95% CI 0.02-0.66). Conclusions: Among Asian Indians living in New York City, language appeared to be an important predictor of colonoscopy use and colorectal cancer screening.


Introduction
Asians Americans are one of the few racial/ethnic groups in the US whose leading cause of death is cancer 1 .Whereas 65.6% of non-Hispanic Whites are up-to-date on colorectal cancer screening, only 51.1% of Asian Americans are 2 .Research has also shown that substantial variation in screening uptake exists among subgroups of the immensely diverse Asian American population.However, all Asian subgroups have screening rates below the Healthy People 2020 target of 70% 3 .
The California Health Interview Survey (CHIS) was the rst population-based study that compared screening rates among different Asian American subgroups 4,5 .Data from CHIS showed that Filipinos, Koreans, Paci c Islanders, and South Asians were less likely to receive colorectal cancer screening than Chinese, Japanese, and Vietnamese individuals, which subsequent studies have con rmed 5,6 .Pooled data from 2000 to 2016 from several large databases including CHIS, National Health Interview Survey (NHIS), Behavioral Risk Factors Surveillance System (BRFSS), and Health Information National Trends Survey (HINTS) showed that South Asians have one of the lowest screening rates, ranging from 39.7%-47.2%,which is consistently lower than for Whites (ranging from 46%-62%) 6 .We have previously shown that in New York City in 2014, only 45% of Asian Indians were up-to-date with colonoscopy, compared to 70% of non-Hispanic Whites 7 .A recent analysis that we conducted examined the extent to which colorectal screening campaigns have reached racial and minority communities in NYC since 2014, revealing that while colonoscopy screening rates have improved across groups, South Asians (61.1%) had among the lowest rates compared to Hispanics (71.3%),Blacks (70.2%) and Whites (68.6%).Results indicated that South Asians have not equally bene ted from campaigns to increase colonoscopy screening in NYC 8 .The studies together underscore the need to develop targeted, linguistically and culturally adapted cancer screening campaigns to reach South Asian populations, including disaggregating data by ethnic group within the larger Asian American population and South Asian category.
The population of Asian Indians, the largest South Asian group, has nearly doubled in the US from less than 2 million in 2000 to nearly 4 million in 2015, and the population consistent of both Indian-born immigrants and rst-generation Indian Americans.Data for other Asian subgroups has shown that USborn members of these communities have a higher incidence of colorectal cancer compared to foreignborn members.For example, from 1973 to 1986, the incidence of colorectal cancer among US-born Japanese men was twice that of foreign-born Japanese men 9 .Currently, there is a paucity of data regarding screening behavior in the rapidly growing Asian Indian population.Our goal was to identify predictors of colonoscopy use among Asian Indians in New York City, which contains the largest Asian Indian population in the US 10 .

Methods
We analyzed 2003-2016 data from the New York City Community Health Survey (NYCCHS), an annual cross-sectional telephone survey.This survey has been administered by the New York City Department of Health and Mental Hygiene (DOHMH) since 2002.It samples approximately 10,000 randomly selected adults aged 18 years and older from all ve boroughs of New York City and collects information on chronic disease and risk factors.All data is self-reported and interviews are conducted in English, Spanish, Russian, and Chinese (Mandarin and Cantonese).For this study, we analyzed responses from Asian Indian participants aged 50 years or older.
We extracted individual-level data on a number of sociodemographic (age, sex, home language, education level, marital status, employment status, children in the household, income, insurance status, borough of residence) and health-related (primary care provider status, body mass index (BMI), diabetes, tobacco, exercise, sugar-sweetened beverage intake) variables.For insurance status, "Other" included TriCare, COBRA, or VA.For language, we used the following categories from the survey: English, Indian (including Hindi and Tamil), and Other.We examined four consecutive time periods (2003-2008, 2009-2012, 2013-2014, 2015-2016) to assess for factors associated with up-to-date colonoscopy, which is de ned as a colonoscopy performed within the last 10 years 11 .This study was approved by the NYU School of Medicine Institutional Review Board.
We performed statistical analyses in SAS Enterprise Guide, version 4.2, software (SAS Institute, Inc., Cary, North Carolina) to accommodate the complex survey design.Survey results were weighted to adjust for the probability of selection as well as a post-strati cation weight.Multi-year weights were also included for the combination of multiple years of data.Notations for data reliability followed DOHMH guidelines 12 .For each of the four time periods, we rst examined the association between variables and up-to-date colonoscopy on univariable analysis.We removed variables missing more than 30% of data and used the nal model regardless of signi cance level.For the other variables, entry into the multivariable model was based on P < 0.15 on univariable analysis.Separate models were constructed for each time period.Odds ratios and their 95% CI were estimated.Statistical signi cance was de ned as a two-sided P < 0.05.

Results
Table 1 presents the unweighted baseline demographic and health characteristics of the study population, strati ed by time period.We identi ed a total of 496 participants aged 50 years and older during the study period.Overall, 64.9% of participants were male and 68.1% were aged 50-64 years.A slight majority of the population graduated from college, had an income level > 200% above the poverty line, and had private insurance.Almost all of the study participants were born outside of the US, and the majority of the individuals resided in the borough of Queens, with smaller numbers of additional participants in the remaining boroughs.2) were considered potentially unreliable based on the relative standard error, con dence interval, or sample size.On univariable analysis, a number of variables met the P < 0.15 threshold for entry into the multivariable model.Colonoscopy was associated with age and language in three of the four time periods.Borough of residence, insurance status, in uenza vaccination status, and BMI were associated with colonoscopy use in two of the four time periods.There was no consistent relationship between a particular borough of residence and colonoscopy use; for this reason data is not shown.Individuals with an insurance status of "other/uninsured" were less likely to be up-to-date.Those with a BMI of 25 and above were less likely to be up-to-date.Individuals who had received in uenza vaccination were more likely to be up-to-date.Factors inversely associated with colonoscopy use in only one time period included having children in the household, not having a primary care provider, not having diabetes, not regularly consuming sugarsweetened beverages, and not exercising in the last 30 days.Factors positively associated with colonoscopy use in only one time period included income > 200% over the poverty line.Variables that were not associated with up-to-date colonoscopy use in any time period included sex, education level, marital status, employment status, and tobacco use.
On multivariable analysis, only language was associated with colonoscopy uptake in more than one time period.Compared to participants who preferred English, those who preferred an Other language were less likely to be up-to-date on colonoscopy in 2003-2008 (OR 0.15, 95% CI 0.04-0.66),while those who preferred an Indian language were less likely to have been up-to-date in 2013-2014 (OR 0.12, 95% CI 0.02-0.66).Risk estimates for both Indian and Other language also suggested an inverse relationship with colonoscopy in 2009-2012, but

Discussion
Among Asian Indians residing in New York City, home language was the most consistent predictor of colonoscopy uptake between 2003 and 2016.Age, income, recent exercise, and in uenza vaccination were also associated with colonoscopy use in single time periods.Our study population was similar to the overall Asian Indian population in the US, although it included a higher proportion of foreign-born individuals (99% vs. 69%) and a lower proportion of college graduates (58% vs. 70%) 10 .
Existing literature supports the role of language as a predictor of colorectal cancer screening in the Asian Indian community 13,14 .Indeed, language may be especially important because other factors frequently associated with screening in other groups-such as education and income-have not been consistent predictors for Asian Indians in several national surveys 15,16 .In a similar study population of foreign-born South Asians living in New York and New Jersey, English language was associated with being up-to-date on screening even after adjusting for other demographic variables 13 .In the South Asian Health Descriptor Study, which focused on a largely rst-generation immigrant community with low educational attainment in Chicago, participants with better English language abilities were three times more likely to have completed endoscopic screening 14 .Our results support these ndings, as participants who preferred non-English languages had lower colonoscopy uptake in two time periods.However, the NYCCHS questionnaire does not permit evaluation of whether preferred language represents a measure of acculturation or an increased ability to access care, follow medical recommendations or more comfortably communicate with health care providers.
Another measure of acculturation that has been cited frequently in the literature is length of time living in the US.Data from CHIS showed that those who lived in the US for less than 15 years were half as likely to have undergone or be up-to-date on screening 4 .The NHIS showed a similar pattern, with Asian Indians who had lived in the US for 10 years or more having a higher odds of screening 17 .Unfortunately, we lacked power to evaluate this variable because only 6% of our cohort had lived in the US for fewer than 10 years.
In uenza vaccination was associated with screening uptake in a single time period, which may represent overall engagement with preventive healthcare.Interestingly, in uenza vaccination rates did not differ between South Asians who did not speak English and English-speaking White participants in the CHIS 18 .This indicates that the same population may avoid colorectal cancer screening while still engaging in other preventive health services.Coupling fecal immunochemical testing to in uenza vaccination has been shown to increase screening rates 19 .
Several study limitations should be noted.First, some variables were only assessed in the NYCCHS for a portion of the study period and weights were only available to pool certain combinations of years for analysis.We divided the overall study into four time period to maximize data capture, but this also reduced the sample size and data reliability in each period.Nevertheless, to our knowledge this is the largest dataset to examine predictors of colonoscopy use among Asian Indians in New York City, which has the highest concentration of Asian Indians in the US.Second, similar to other population-based surveys 4 , the NYCCHS was not administered in South Asian languages during the study period.This may lead to a selection bias favoring participants with higher English pro ciency and level of acculturation, thereby overestimating colonoscopy use in this community.Furthermore, a number of languages were aggregated into the Indian and Other categories, which precluded more in-depth analysis.Third, the NYCCHS did not inquire about other screening modalities such as stool-based testing, so our results underestimate the overall up-to-date screening rate.However, colonoscopy is by far the predominant screening modality used in the US.Lastly, a survey study does not permit more nuanced examination of the patient-physician relationship.Recent data has shown that South Asian patients cite physician recommendation as a leading reason to undergo screening 20,13 .For this study, we were only able to assess whether participants had a primary care provider, but not whether the providers recommended screening or engaged in language-concordant care.
In summary, language appears to be an important predictor of colonoscopy use among Asian Indians in New York City.Given the low level of colorectal cancer screening in this population, further examining the precise meaning of language-as a measure of acculturation or a communication barrier-may identify the need for targeted and translated and in-language interventions to improve uptake.Additional qualitative research may provide important contextual data to supplement and expand these ndings.

Declarations
Authors' contributions -NN and PL were the primary authors on this manuscript.All authors: KL, MG, SN, CT-S, AT, SK, YX. reviewed the manuscript.YX completed the data analysis.SK and CT-S were involved in the primary grant submissions.

Table 1
Baseline demographic and health behavior characteristics among Asian Indians.
this was not statistically signi cant.Individuals older than 65 years were more likely than those aged 50-64 years to have received colonoscopy only in 2009-2012 (OR 3.92, 95% CI 1.49-10.26),although the risk estimates were also consistently positive in the other three time periods.In 2009-2012, compared to those who exercised in the last 30 days, individuals who did not were less likely to have had colonoscopy (OR 0.31, 95% CI 0.10-0.98).In 2015-2016, income > 200% above the poverty line (OR 20.11, 95% CI 5.18-78.06)and in uenza vaccination (OR 8.15, 95% CI 2.78-23.92)were both associated with increased uptake.In uenza vaccination also trended toward signi cance in 2013-2014 (OR 3.37, 95% CI 0.94-12.13).