Characteristics of cancer hospitals with written language access policies

Abstract

Background:

Patients with limited English proficiency receive worse care due to communication barriers. Little is known about which cancer hospitals have written language access policies addressing bilingual clinicians.

Methods:

We conducted a cross-sectional survey of healthcare organizations, matching survey data to American Hospital Association Survey and American Community Survey data. We analyzed characteristics associated with hospitals having bilingual clinician policies.

Results:

The response rate was 71% (127/178). Many hospitals (53 [42%]) did not have written policies on bilingual clinicians. Having bilingual clinicians available at the hospital was associated with having a written policy on bilingual clinicians, while being an NCORP site was associated with not having a written policy on bilingual clinicians. Patient demographic characteristics were not associated with hospitals having written policies on bilingual clinicians.

Discussion:

A substantial proportion of cancer hospitals do not have policies that cover language use by bilingual clinicians, particularly at NCORP sites. Having written policies on bilingual clinicians has the potential to mitigate cancer disparities by facilitating accountability, improving communication, and reducing errors.

INTRODUCTION

About 26 million people in the US speak English “less than very well,” or are limited English proficient (LEP) (1), a rapidly growing population (2). Due to communication barriers, LEP individuals are at greater risk of adverse events (3), increased frequency of medical errors (4), and longer hospital length-of-stay (5,6) than English-proficient patients. Specifically, LEP cancer patients participate in complex discussions with multiple clinicians during their care (7), increasing risk for impaired communication (8–11) and consequent disparities in cancer treatment completion (12,13). However, little is known of how many and which cancer hospitals have language access plans. To mitigate language access disparities, we need to better characterize which cancer hospitals have written language access plans, an important proxy for equitable language service provision.

Inadequate language access results in negative health outcomes (14,15) and legal and financial consequences (16); however, hospitals do not sufficiently provide language services to LEP patients. Federal regulations require hospital language services provision. Implementation guidance is outlined in the National Standards for Culturally and Linguistically Appropriate Services in Healthcare (CLAS Standards) (17), which describe how to comply with Title VI of the Civil Rights Act (18). More recently, Section 1557 of the Affordable Care Act strengthened patients’ rights to language services (19).

However, the existence of national or state-level policies does not guarantee their implementation. One national study found that only 13% of hospitals met all four of the CLAS Standards that are considered mandatory (18,20). In 2016, only 56% of hospitals reported offering linguistic or translation services (21). Hospital language services are also inadequate in geographic areas of demonstrated need, with one-fourth of hospitals with large LEP populations lacking language services (22). In Massachusetts, for example, the use of professional interpreters remained low despite a well-publicized state law mandating all healthcare settings to provide professional interpreter access (23). Lack of interpreter use is especially important in cancer care, which often involves numerous clinicians and treatment modalities (7) and necessitates complex and emotional discussions.

Little is known of characteristics associated with hospitals having written language access plans. Such knowledge could help gauge cancer hospital compliance to language service requirements by using the presence of a written policy as a proxy for having language services. Moreover, written hospital policies can guide language service use and are necessary for accountability. Despite knowledge that professional interpreters improve care for LEP patients (14), research has shown that clinicians underuse language services in favor of their own, often limited, non-English language skills (24,25). Thus, having written bilingual clinician policies can help assess, monitor, and ensure quality and safety.

Our primary aim was to assess whether hospitals serving cancer patients have written interpreter and bilingual clinician policies. We anticipated that most healthcare organizations serving LEP cancer patients would have policies on interpreter use but not explicit rules about interpreter use for bilingual clinicians. Our secondary aim was to identify characteristics associated with cancer hospitals having bilingual clinician policies. To do so, we examined hospitals’ geographic, hospital, patient demographic, and workforce characteristics. We hypothesized that patient demographics, more than other characteristics, would be associated with having bilingual clinician policies. We surveyed 178 cancer hospitals on their written language access policies, assessed hospital characteristics using multiple data sources, and analyzed characteristics associated with having bilingual clinician policies.

METHODS

Design.

We surveyed 178 healthcare organizations, including all 60 National Cancer Institute- (NCI-) designated Cancer Centers, which represent the gold standard of US cancer treatment. However, their patient populations may not be nationally representative. Thus, we included 106 healthcare organizations in the 61 NCI Community Oncology Research Program (NCORP) sites, which deliver care in a variety of community settings and perform cancer disparities research. NCORP sites are a collection of clinics in geographic proximity, often associated with an academic hospital, and intended to increase clinical trial access to communities previously underrepresented in research. Because it is possible that practices grouped as one NCORP site may have different policies, we randomly selected one to two practices within each NCORP site to survey. We also included 12 New York City safety-net hospitals, providing cancer care to underserved patients, increasing the proportion of institutions serving traditionally underrepresented populations in our sample. We fielded surveys from May 2016 through June 2018 via email link using Research Electronic Data Capture (REDCap) data collection software (26). This research was determined to be exempt by the Memorial Sloan-Kettering Cancer Center Institutional Review Board (X15-021).

Participants.

At each healthcare organization, we identified the person “best able to describe the language services offered,” a strategy successfully used in a prior study (20). We identified this person via web search or telephone calls, then sent the REDCap survey to their e-mail address.

Data sources.

For data on interpreter policies, healthcare organizations submitted answers to the REDCap survey. Respondents were offered options to provide the facility’s written policy on interpreter use or to complete a survey on the policy’s content. For hospitals that sent their written policies, two coders separately abstracted and coded data from the policies using the survey questions as a guide. For data on hospital characteristics, we matched the survey data with 2016 American Hospital Association (AHA) Annual Survey data using the hospital name or address. Characteristics of AHA hospitals included hospital teaching status, government/non-government status, bed size (<100 beds, 100-299 beds, ≥300 beds), and Joint Commission accreditation. Hospitals were additionally classified into Census Regions (Midwest, Northeast, South, West) by state. Healthcare organizations that are not hospitals, such as some of the NCORP sites, were matched to their admitting hospital, which was confirmed by calling the individual NCORP site.

We also used publicly available data from the US Census Bureau’s American Community Survey (2012-2016) to characterize the nativity and language diversity of the counties where each healthcare organization was located. These variables included (i) LEP households, (ii) US nativity, (iii) language spoken at home by ability to speak English for population aged ≥5 years, and (iv) language spoken at home for population aged ≥5 years (27). For each county, we then calculated the percentage of LEP households, whether the foreign-born population was ≥20%, and whether the non–English-speaking population increased ≥10% from 2005 to 2009 through 2012 to 2016.

Statistical analyses.

We compared the characteristics of hospitals that did and did not respond to the survey by geographic, hospital, patient-related, and workforce characteristics using a Chi-square test for categorical variables and the Wilcoxon rank sum test for continuous variables.

For our primary aim, we report the proportion of hospitals in our sample that have a written policy on clinicians’ use of non-English language skills and whether they conduct language proficiency assessments for bilingual clinicians. We also report the most commonly requested interpreter languages.

For our secondary aim, we assessed whether geographic, hospital, patient, and workforce characteristics were associated with hospitals having bilingual clinician policies. For bivariate associations, we used Chi-square tests for categorical variables and Wilcoxon rank sum tests for continuous variables. We then fit a multivariable logistic regression model to evaluate characteristics associated with hospitals having bilingual clinician policies, including characteristics that were significant in bivariate analyses or of clinical interest. As a sensitivity analysis, only characteristics that were significant in bivariate analyses were included in a multivariable model.

Additionally, we used content analysis to qualitatively assess responses to two free-text questions: one on difficulties providing LEP patient care and another inviting any remaining comments on hospitals and language services.

All analyses were conducted using STATA version 15 (College Station, TX) (28) and R software version 3.5.1 (Vienna, Austria) (29).

RESULTS

Most hospitals (71%, 127/178) responded to our survey. NCORP sites had a higher response rate compared to non-NCORP sites. There were no significant differences between responders and non-responders by geographic or patient-related characteristics (Table 1). Responding hospitals were in the South (32%), Midwest (29%), Northeast (20%), and West (18%). Most hospitals were non-government (72%), had ≥300 beds (71%), were teaching hospitals (54%), and were Joint Commission accredited (93%). A sizable proportion (43%) were in counties where the non–English-speaking population increased by ≥10% from 2005 to 2009 through 2012 to 2016.

Table 1.

Characteristics of Responder and Non-Responder Hospitals

Characteristics		Overall, no. (%)	Non-responders, no. (%)	Responders, no. (%)	P value
Hospitals invited		178 (100)	51 (29)	127 (71)
Geographic characteristics
	Census region (n=177)				0.06
	Midwest	57 (32)	20 (40)	37 (29)
	Northeast	35 (20)	9 (18)	26 (20)
	South	48 (27)	7 (14)	41 (32)
	West	37 (21)	14 (28)	23 (18)
Hospital characteristics
	Hospital classification				0.31
	Government	46 (26)	10 (20)	36 (28)
	Non-government	132 (74)	41 (80)	91 (72)
	Size by no. of beds				0.51
	<100	15 (8.4)	5 (9.8)	10 (7.9)
	100-299	41 (23)	14 (27)	27 (21)
	≥300	122 (69)	32 (63)	90 (71)
	Teaching hospital	90 (51)	21 (41)	69 (54)	0.16
	Accredited by Joint Commission	163 (92)	45 (88)	118 (93)	0.37
Patient-related characteristics
	LEP households in county, median (IQR)	3.6 (2.0, 7.4)	3.5 (2.1, 7.0)	3.6 (2.0, 7.8)	0.48
	County foreign-born population ≥20% (n = 177)	51 (29)	13 (26)	38 (30)	0.74
	Non–English-speaking population increase ≥10%*	73 (41)	19 (37)	54 (43)	0.63
NCORP Status					0.03
	NCORP	103 (58%)	36 (71%)	67 (53%)
	Non-NCORP	75 (42%)	15 (29%)	60 (47%)

Abbreviations: IQR, interquartile range; LEP, limited English proficient.

^*Increase from 2005 to 2009 through 2012 to 2016, American Community Survey data

Primary aim results

All responding hospitals reported seeing LEP patients, and 98% of hospitals had a written policy on interpreter use (Table 2). Over three-quarters (81%) reported having bilingual clinicians (n=103). However, only 59% (n=74) had written policies on bilingual clinicians who provide their own professional services in both English and another language, and only 62% (n=64) of hospitals with bilingual clinicians reported a way of assessing the language proficiency of these providers. Of the 64 hospitals that reported assessing proficiency, about two-thirds (n=42, 66%) assessed medical terminology knowledge.

Table 2.

Responses to select survey questions

Survey Question (n=127)	Responders, no. (%)
Does your hospital have a written policy on interpreter use for patients with LEP?
Yes	125 (98)
No	2 (2)
Does your hospital’s policy on interpreter use for patients with LEP include a policy on bilingual clinicians*? (n=125)†
Yes	74 (59)
No	51 (41)
Are bilingual clinicians available at your hospital?
Yes	103 (81)
No	24 (19)
Is any assessment of language proficiency done for bilingual clinicians? (n=103)
Yes	64 (62)
No	24 (23)
Don’t know	15 (15)
Hospitals reporting assessing knowledge of medical terminology (n=64)
Yes	42 (66)
No	22 (78)
Please rank the most common languages interpreters are requested for at your hospital, with #1 being the most common. (n=124)
Spanish	113 (91)
American Sign Language	7 (6)
Arabic	2 (2)

Abbreviation: LEP, limited English proficient.

^*Bilingual clinicians provide their services in more than one language

^†Denominator is 125 responding hospitals with a written policy on interpreter use for patients with LEP.

Spanish was the most commonly requested language by LEP patients in nearly all hospitals (n=113, 91%). The second and third most-requested languages were American Sign Language and Arabic.

Secondary aim results

In bivariate analyses, there were statistically significant differences in the likelihood of having written policies on bilingual clinicians based on census region (p=0.03), hospital classification (government/non-government) (p=0.03), NCORP status (p<0.01), and bilingual clinician availability (p<0.01) (Table 3). In the multivariable analysis, which included those factors in addition to teaching hospital status and the proportion of LEP households in the county, the only characteristics significantly associated with having a bilingual clinician policy were having bilingual clinicians at the hospital (OR 4.58, 95% CI 1.48, 16.16; p<0.01) or being a non-NCORP site (OR 0.14, 95% CI 0.04, 0.38; p<0.01). Of note, we did not find evidence of an increase in the odds of a hospital having bilingual clinician policies per 5% increase in proportion of LEP households in the hospital’s county (OR 81 [0.41, 1.55]) (Figure 1).

Table 3.

Characteristics of hospitals with or without written policies on bilingual clinicians’ use of non-English language skills

Characteristics		Hospitals, no.	Without Policy, no. (%)	With Policy, no. (%)	P value
			Overall			127	53 (42)	74 (58)
Geographic characteristics
	Census region				0.03
	Midwest	37	19 (51)	18 (49)
	Northeast	26	9 (35)	17 (65)
	South	41	11 (27)	30 (73)
	West	23	14 (61)	9 (39)
Hospital characteristics
	Hospital classification				0.03
	Government	36	9 (25)	27 (75)
	Non-government	91	44 (48)	47 (52)
	Bed size				0.42
	<100 beds	10	3 (30)	7 (70)
	100-299 beds	27	9 (33)	18 (67)
	≥300 beds	90	41 (46)	49 (54)
	Teaching hospital	69	27 (39)	42 (61)	0.64
	Accredited by Joint Commission	118	47 (40)	71 (60)	0.16
Patient-related characteristics
	% total LEP households in county, median (IQR)		3.0 (2.0, 6.7)	4.0 (2.2, 8.9)	0.30
	% foreign-born in county ≥20%	38	16 (42)	22 (58)	>0.95
	% change in non–English-speaking population ≥10%	54	24 (44)	30 (56)	0.73
Workforce characteristics
	Availability of bilingual clinicians	103	35 (34)	68 (66)	<0.01
NCORP site					<0.01
	NCORP	67	39 (74%)	28 (38%)
	Non-NCORP	60	14 (26%)	46 (62%)

Abbreviations: IQR, interquartile range; LEP, limited English proficient.

Figure 1.

Factors associated with hospitals having a policy for LEP patients that included bilingual clinicians (showing odds ratios and confidence intervals)

In the sensitivity analysis, restricting the multivariable model to only include factors that were significant in the bivariate analysis, we again found that the only factors significantly associated with having a bilingual clinician policy were having bilingual clinicians at the hospital (OR 4.74, 95% CI 1.63, 15.58; p<0.01) and being a non-NCORP site (OR 0.19, 95% CI 0.07, 0.47; p=<0.01).

There were two free-response questions on (i) LEP patient care provision difficulties and (ii) any other comments, with 60 and 43 responses, respectively. Prominent themes included (a) inadequate financial and workforce resources to provide language services, particularly for languages of lesser diffusion for which the population of speakers in a geographic region is relatively small; (b) inadequate compliance to existing policies by providers; and (c) challenges including logistics, last-moment changes, and inadequate in-person interpreters to meet need.

DISCUSSION

Nearly all the NCI-designated comprehensive cancer centers, NCORP-participating healthcare organizations, and NYC safety-net hospitals surveyed had policies for interpreter use, but, as hypothesized, a large proportion (41%, n=51) did not have written policies on bilingual clinicians. This is consistent with prior work showing that many healthcare organizations do not fully incorporate CLAS Standard concepts (30). More specifically, a large proportion of hospitals treating cancer patients do not have written policies on bilingual clinicians to help enforce adherence to CLAS Standard 7, “Ensure the competence of individuals providing language assistance, recognizing that the use of untrained individuals and/or minors as interpreters should be avoided” (17). Our findings suggest that CLAS Standards implementation remains variable and incomplete in cancer hospitals, similar to non-cancer hospital findings (31).

While 81% of cancer hospitals reported having bilingual clinicians, only 59% had a written bilingual clinician policy, and only 62% of hospitals with bilingual clinicians assessed their non-English language proficiency. This gap suggests that clinicians are using their language skills in clinical settings without explicit guidance or quality assurance. While it is well-established that clinicians of varying skill levels use their language skills in clinical settings (32,33), only now do we know that a significant proportion do so without written organizational policies. The absence of such policies makes it more likely that clinicians provide language services variably and inconsistently and that patients and staff have less clarity on the language access standards to which they may hold their providers and hospitals accountable. The absence of clear policies providing oversight also opens hospitals to increased risk and liability, which historically has led to clinical, legal, and financial consequences (16,34–36). To avoid these consequences, compliance with the CLAS Standards might be facilitated by adding specific, evidence-based metrics, such as having written bilingual clinician policies, to existing CLAS implementation tools (37,38). Policymakers and health system leaders can also consider developing payment models that reward quality care to LEP patients, systemically collecting language information, and making the business case for excellent LEP care provision as they promote efforts to encourage compliance to CLAS Standards (39).

The two characteristics associated with cancer hospitals having bilingual clinician policies in the multivariable analysis were having bilingual clinicians available and being a non-NCORP site. The first finding was not unexpected. We speculate that bilingual clinicians may have expressed greater demand for policies, that the hospital may have initiated policies for liability reasons, or both. Little is known of the drivers of new policy adoption at cancer hospitals, and this topic deserves further study. The second finding was not expected, as NCORP sites include the Minority/Underserved Community Sites, which focus on underrepresented communities, including LEP patients.(40) Our study suggests efforts to clarify and adopt written policies on bilingual clinicians can be targeted towards NCORP sites.

Notably, we did not find an association between patient demographics-related characteristics and having written bilingual clinician policies. Our finding may appear contrary to data showing that the proportion of the population that speaks a language other than English predicts the likelihood of having language services at hospitals (22,41). However, written bilingual clinician policies may represent a more specific commitment to ensuring language access, and for unclear reasons, cancer hospitals may be less responsive to local language needs. We hypothesize that this may reflect a combination of lag time between cancer hospitals being aware of the magnitude of language services need based on the local population and the adoption of formal policies, lack of awareness of written bilingual clinician policy importance, and/or absence of mechanisms for reflecting local population demographics when creating cancer hospital policies. For example, it was unexpected that 61% of Western cancer hospitals did not have a written bilingual clinician policy (Table 3), despite having over 7.5 million LEP people (42). Similarly, of the 54 hospitals in counties where the increase in the non-English speaking population was at least 10%, 44% did not have bilingual clinician policies, highlighting an important gap (Table 3). Patient-centered and locally responsive mitigation strategies could include consulting patient and family advisory councils, which have been successfully implemented in both cancer (43,44) and non-cancer settings (45,46), and incorporating local data on language needs (41) when creating hospital policies.

Prior work suggests that LEP clusters with other traits that contribute to suboptimal cancer care. Low health literacy and LEP combined predict even lower cancer screening rates than LEP alone (47), and associations have been found between neighborhood LEP rates, proportion of recent immigrants, and late-stage colorectal cancer diagnosis (48). These studies emphasize the importance and utility of anticipating language access needs for cancer hospitals, as neighborhoods are located in a landscape of risk informed by social and structural determinants of health (49). Efforts such as linguistically responsive cancer patient navigation programs (50) appear promising in mitigating some of these inequities (51). Upstream and large-scale interventions are also necessary. Cancer hospitals can consider incorporating equity measures – such as written language access policies – into quality assessment, an approach that has been used successfully in cardiovascular disease (52).

Clinical practice and public health implications

Having bilingual clinicians practicing without written policies has the potential to exacerbate existing cancer care inequities. Language is the most important barrier for LEP patients with cancer in understanding information and treatment decision making (53–55). For example, bone marrow transplant patients with LEP report difficulty accessing information in their requested language (56). Surveyed LEP patients with cancer have notable knowledge gaps, including their cancer diagnosis (57), stage, and metastatic status (58). Inadequate language service access hampers goals-of-care discussions (59) and advance care planning (60), which are important in aligning patient and provider goals in cancer treatment and end-of-life care. Information, when conveyed, may also carry more weight for LEP patients; one study showed that Spanish-speaking Latina breast cancer patients were more likely than English-speaking Latina and Black patients to rely on physicians’ treatment recommendations than to participate in shared decision making (61). Unsurprisingly, LEP has been associated with higher symptom distress, lower quality of life (62), and worse treatment outcomes (13,63) among cancer patients. These gaps could be narrowed by ensuring the quality of language services through written language policies, including language assessment, for bilingual clinicians.

Efforts to increase language concordance, which is associated with positive patient outcomes (15), should also be sustained and expanded. After the 2010 census, it was estimated that at least 13% of the US population would be Spanish-speaking by 2020 (64). While national data on physician non-English language skills are not readily available, the percentage of Spanish-speaking physicians is unlikely to approach this number. A web-based survey found that almost all surveyed metropolitan areas had weak matches between the languages that physicians and patients speak (65). In the short term, hospital language services must meet immediate demand with high-quality interpreting services and appropriate use of clinician non-English language skills, particularly to ensure access to patients’ most-requested languages. In the long term, clinicians with non-English language skills matching those of their LEP patients must be recruited, retained, and empowered.

Limitations

This study was the first nationwide survey to assess whether cancer hospitals have written bilingual clinician policies. We relied on health care organizations to accurately report information about bilingual physicians’ language proficiency and medical terminology knowledge rather than asking bilingual clinicians to corroborate this information. Our response rate was 71%; while this is high compared to usual e-mail survey response rates (66,67), our study excludes data from non-responders. Although health care organizations within NCORP sites were directly surveyed, the surveyed health care organization may not have been representative of the larger NCORP site. NCORP sites may have variation amongst its health care organizations; to decrease this bias, health care organizations at NCORP sites were chosen randomly. Additionally, variables that we did not examine may better explain variability in the presence/absence of bilingual clinician written policies. Our ability to assess reasons for not having bilingual clinician policies was limited and could have benefited from including other clinicians such as nurses or non-clinical staff, or from further qualitative methods such as interviews and focus groups. Finally, our US data may not be generalizable to other countries.

Conclusions and future research areas

Future research should investigate drivers of and barriers to having written bilingual clinician policies, which could inform interventions to introduce such policies more consistently and broadly across US hospitals. Data collection should be expanded to non-cancer hospitals to better characterize the scope of this issue, and could include clinical and non-clinical staff besides physicians. The implementation of written policies should also be examined to ensure compliance and assessment of bilingual physicians’ language skills, including their medical terminology knowledge. Such implementation might also benefit from having champions within nursing units or clinic teams who can help promote adherence to language policies and quality of language concordant care; these efforts should be studied for best practices for success. This research is vital to ensuring the safety of LEP cancer patients and to promoting patient-centered outcomes, such as being empowered, informed, and able to participate in shared decision making. Hospitals have an opportunity to proactively intervene on language disparities and comply with CLAS standards by writing, disseminating, and enforcing language policies that enable bilingual clinicians to use their non-English language skills responsibly.

Availability of data and materials:

The datasets during and/or analyzed during the current study available from the corresponding author on reasonable request.