Ethnic variation in cancer patients’ ratings of information provision, communication and overall care

ABSTRACT Objective. Ethnic inequalities in cancer patient experience exist but variation within broad ethnic categories is under-explored. This study aimed to describe variation by ethnic sub-category in experiences of information provision and communication (key domains of patient experience) using National Cancer Patient Experience Survey (NCPES) data. Design. The NCPES 2012–2013 contained responses from 68,737 cancer patients treated at 155 NHS Trusts in England. Multivariate logistic regression was used to investigate associations between ethnicity and patients’ ratings of overall care, information provision and communication. Results. Variation by and within broad ethnic categories was evident. Non-White patients (particularly Asian patients (ORadj:0.78; 95%CI:0.67-0.90, p=0.001)) were less likely than White patients to receive an understandable explanation of treatment side effects. Among Asian patients, those of Bangladeshi ethnicity were least likely to receive an understandable explanation. Conclusions. Effective communication and information provision are important to ensure patients are well informed, receive the best possible care and have a positive patient experience. However, ethnic inequalities exist in cancer patients’ experiences of information provision and communication with variation evident both between and within broad ethnic categories. Further work to understand the causes of this variation is required to address ethnic inequalities at practice and policy level.


Introduction
Cancer care and treatment is a resource-intensive area of healthcare which is set to expand, due in part to an ageing population. It is also becoming more complex as treatment options increase and more patients have co-morbidities. As a result, many countries, including the UK (where healthcare is available free of charge to ordinary residents through the National Health Service (NHS)), have developed and implemented strategies designed to ensure the best possible cancer care and treatment is available to all patients (New Zealand Ministry of Health 2002;Department of Health 2010;National Cancer Equality Initiative 2010;Koh, Graham, and Glied 2011;Department of Health 2011a). Despite these strategies, patients' experience of cancer care is known to vary by sociodemographic factors such as ethnicity (Ayanian et al. 2005;Shadmi et al. 2010;Penner et al. 2013). Ethnic disparities in patient experience are well documented in the NHS, with many studies finding that non-White patients (especially Chinese and Asian groups) report less positive experiences than White British groups across a range of healthcare services (Raleigh et al. 2007;Sizmur 2011;Lyratzopoulos et al. 2012;Henderson, Gao, and Redshaw 2013), including cancer care (Quality Health 2013; Bone et al. 2014).
In order to facilitate health service improvements, the Department of Health has identified a number of areas that are important to patient experience, a key domain of highquality healthcare (Department of Health 2011b). Information provision has been identified as being critical to good patient experience (NHS National Quality Board 2012) leading to a focus on high-quality written and verbal communication with the aim of effectively informing and empowering patients and their families. However, different ethnic groups are known to have different preferences for information provision and communication; for example, there is variation between ethnic groups in the preferred amount of information and mode of delivery (e.g. printed, online and person-to-person) Kakai et al. 2003;Mitchison et al. 2012). Ethnic minority patients who do not speak the native language of the healthcare setting also report communication issues (Ayanian et al. 2005;Shadmi et al. 2010). We therefore hypothesised that there would be wide variation in reported patient experience related to communication and information provision among cancer patients in England.
The National Cancer Patient Experience Survey (NCPES) is undertaken in England annually on behalf of the Department of Health to evaluate NHS cancer services and contains several questions related to patients' experiences of communication and information provision. While published NCPES reports have identified inequalities in reported patient experience by ethnicity in areas such as overall ratings of and access to care (Department of Health 2009), there have not been any detailed analyses of ethnic variation in communication. Furthermore, each broad ethnic category used for analysis in these reports represents a patient population containing several diverse groups (i.e. the Asian ethnic group contains Indian, Pakistani and Bangladeshi patients), yet inequalities are not explored by small ethnic sub-category. The aim of this study was to gain further insight into variation in cancer patients' reported experiences of information provision and communication, through analysis of relevant NCPES data by ethnic categories and sub-categories.

Sources of data
Secondary analysis was performed on data from the 2012-2013 NCPES, conducted by Quality Health on behalf of the Department of Health. The 2012-2013 NCPES was sent to every adult (aged over 16 years) with a primary diagnosis of cancer, admitted as an inpatient or day case patient for treatment in an NHS hospital in England and discharged between 1 September and 30 November 2012. Questionnaires were not sent to patients who were known to have died since discharge. Non-responders were followed up with two postal reminders (Quality Health 2012). The 2012-2013 NCPES data set contained responses to 70 multiple-choice questions and associated demographic information for 68,737 cancer patients from 155 English NHS Trusts, the administrative bodies responsible for commissioning primary, community and secondary health services from providers.

Patient, clinical and Trust-level factors
Patient, clinical and Trust-level characteristics are known to impact on ratings of patient experience (Bone et al. 2014) and were adjusted for in our analyses. Self-reported data was used for these factors where possible as it is considered to be 'gold-standard', particularly for ethnicity (Saunders et al. 2013). From NCPES responses, self-reported patient (i.e. gender, age, ethnicity, presence of a long-standing condition, sexual orientation and employment status) and clinical characteristics (i.e. time since diagnosis and response to treatment) were extracted. Hospital-reported data was used for NHS Trust, tumour group, patient classification (day case or inpatient) and index of multiple deprivation (IMD), a composite measure of area-level deprivation associated with a patient's postcode (Department for Communities and Local Government 2011). Analysis of the data by ethnicity was performed using six broad categories and fourteen sub-categories (Table 1). Chinese and Other were included as separate broad categories in line with other studies showing different rates of patient satisfaction in these groups (Department of Health 2009;Sizmur 2011;El Turabi et al. 2013). White and White British were used as reference categories for the broad and sub-categories, respectively, as they contained the largest number of participants.

Data analysis
The 2012-2013 NCPES was screened for questions related to overall care, information provision and communication. For the 23 questions selected (Appendix), patients' responses were coded into binary outcomes ('positive' or 'not positive') as per the survey guidance (Quality Health 2013). Firstly, univariate logistic regression was used to explore associations between a positive rating of overall care (as assessed by Q70 'Overall how would you rate your care?') and patient and clinical factors, including ethnicity. A mixed effects multivariate model, adjusting for confounders, was then developed by including fixed effects for patient and clinical factors associated with a positive overall care rating at a significance level of p < .05 and a random effect for NHS Trust of treatment. This multi-level regression framework accounts for differences that may arise because ethnic minority patients are clustered in poorly performing Trusts. Finally, a likelihood ratio test was used to compare models using the six broad ethnic categories and fourteen sub-categories and to determine if variation within broad ethnic groups was statistically significant, taking into account the number of respondents. The same method was used to explore the associations between patients' ratings of overall care and being given the right amount of information (as measured by Q67 'How much information were you given about your condition and treatment?').
The remaining 21 questions were then grouped by the stage of the pathway to which they referred (i.e. diagnosis, diagnostic tests, treatment side effects, operations, support or leaving hospital) as indicated by the section of NCPES questionnaire in which they appeared (Appendix). Composite scores by broad and sub-categories of ethnicity were determined for each stage of the pathway by calculating an unweighted average of the proportion of positive responses to the constituent questions. While this method does not address variation in the meaning of a positive response to a question or the relative contribution of a 13.4 a Ethnic sub-categories are italicised. b Patients may have multiple long-standing conditions (therefore the column total may exceed 100%). c Such as (but not limited to) HIV, diabetes or chronic heart disease. Gender was unknown for 3.4% of respondents (n = 2319), age group for 4.4% (n = 2992), ethnicity for 3.8% (n = 2590), long-standing conditions for 9.0% (n = 6175), time since first treatment for 4.2% (n = 2905), employment status for 4.4% (n = 3023), response to treatment for 9.6% (n = 6574) and IMD quintile for 0.5% (n = 359).
question to overall patient experience, it is consistent with the NCPES survey guidance and provides an overview of variation in reported patient experience. Based on this exploration of the data, information related to treatment side effects was selected for further analysis as it was the lowest rated part of the cancer care pathway. Univariate and multivariate multi-level logistic regression was performed as previously described to examine associations between ethnicity and the three NCPES questions related to information about treatment side effects (Q17, 18 and 19) adjusting for potential confounding factors. Communication with staff is an important domain of patient experience and ratings of communication vary by staff group (Quality Health 2013); however, ethnic differences have not been described. Therefore, three comparable questions which asked how often patients received understandable answers to important questions from different staff members (Q24, 37 and 41) were also selected for further analysis. Associations with ethnicity were explored as previously described using univariate and multivariate multilevel logistic regression. All statistical analyses were performed using Stata v12 and complete case analysis was undertaken (i.e. respondents with missing demographic, clinical or Trust-level data, or those who did not answer the question of interest, were excluded).

Respondent characteristics and rating of care
In total, 68,737 patients completed 2012-2013 NCPES representing a response rate of 64% (Quality Health 2013). Most respondents were female, aged 66-75 years and described themselves as White British (Table 1).

Overall rating of care
A large proportion of respondents (88%, n = 58,525) rated their overall care positively (i.e. as 'very good' or 'excellent'). However, there was variation between ethnic categories with non-White patients (particularly Chinese and Asian patients) significantly less likely to rate their care positively compared to White patients (Table 2). There was also variation in overall care rating within broad ethnic categories by sub-category. For example, among Asian patients Bangladeshis had the lowest rating of overall care (54.2% vs.72.4% of Indian patients, p = .01) and among White patients those of 'Any other background' rated their care less positively than British patients (82.4% vs. 89.6%, p = .001). These differences by ethnicity in reported patient experience persisted after adjusting for potential confounders such as age, gender and NHS Trust of treatment.

Communication and information provision
Patients who felt they received the right amount of information (i.e. responded positively to Q67) were far more likely to rate their overall care as 'very good' or 'excellent' (OR adj :7.74, 95%CI:7.25-8.25, p < .001), adjusting for age, gender, ethnicity, long-standing condition, time since first treatment, tumour group, response to treatment, patient classification (day case or inpatient) and Trust. However, patients' ratings of communication and information provision varied across the stages of the cancer care pathway, as well as by ethnicity (Table 3). Overall, patients reported being most satisfied with information provision when leaving hospital and least satisfied with information about the side effects of treatments. Chinese and Asian patients (particularly Bangladeshi patients) were less satisfied with communication and information provision across the entire pathway.

Information about side effects of treatment
Three NCPES questions related to information about the side effects of treatment asked patients if they had been (a) given an understandable explanation of potential side effects (Q17), (b) given written information that was easy to understand about the side effects of their treatment (Q18) and (c) told about potential side effects that may develop in the future (Q19). Variation in responses to these questions by ethnicity was evident (Table 4). Black and Asian (especially Bangladeshi) patients were less likely than White patients to report that they had received an explanation of side effects that they could understand from staff or been told about side effects that might develop in the future. A similar pattern was seen with regards to receiving understandable written information about the side effects of treatment. Among patients who reported that they According to the likelihood ratio test, the model using ethnic sub-categories was significantly better than the model using broad categories (χ 2 (10) = 52.00, p < .001). a Adjusted for age, gender, long-standing condition, time since first treatment, tumour group, patient classification, response to treatment, IMD and Trust. b For comparisons by broad ethnic categories, the reference group was White patients. c For comparisons by ethnic sub-categories, the reference group was White British patients.
were not given written information about the side effects of treatment that was 'easy to understand', 27.1% (n = 2436) reported that they had been given written information but found it difficult to understand. Adjusting for age, gender, long-standing condition, tumour group, time since first treatment, response to treatment, patient status, IMD and Trust, non-White participants were less likely to report that they found the written information provided easy to understand (e.g. Black patients OR adj :0.36, 95%CI:0.28-0.47, p < .001).

Understandable answers to important questions
Three NCPES questions asked patients how often they received understandable answers to important questions from different staff members, namely (a) Clinical Nurse Specialists (CNSs) (Q24), (b) doctors (Q37) and (c) ward nurses (Q41). Overall patients reported that CNSs were most likely to give responses to questions that were easy to understand and, for all three staff groups, worse communication was reported by non-White patients (Table 5). These differences by ethnicity remained after adjusting for potential

Discussion
This study investigated the relationship between ethnicity and patients' reported experience of cancer care with a focus on communication and information provision. Patients' overall rating of care varied both between and within broad ethnic groups and was correlated with ratings of communication and information provision. There was considerable variation in patients' reported experiences of communication and information provision across the cancer care pathway and by staff member, as well as by ethnicity. Non-white patients, especially Chinese and Asian (and in particular Bangladeshi) individuals, rated their experiences of communication and information provision less positively than White patients. Studies have demonstrated that non-White cancer patients are less likely than White patients to rate their overall care and treatment positively (Quality Health 2013; Bone et al. 2014) and the results of our analysis provide additional evidence of such ethnic inequalities. They also go further by highlighting the varied experiences of patients within broad ethnic categories. For example, among Asian patients, those of Bangladeshi ethnicity reported the worst experiences, while, among White patients, those of British or Note: Significant associations (p < .05) are highlighted in bold. a Adjusted for age, gender, long-standing condition, time since first treatment, tumour group, patient classification, response to treatment, IMD quintile, Trust location (in/outside Greater London) and Trust. b Analysis was restricted to 52,029 respondents and excluded patients who did not respond to Q17 (n = 5031) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 11,677). According to the likelihood ratio test, the model using ethnic sub-categories was significantly better (at 95% confidence level) than the model using broad categories (χ 2 (10) = 28.77, p = .001). c For comparisons by broad ethnic categories, the reference group was White patients. d For comparisons by ethnic sub-categories, the reference group was White British patients. e Analysis was restricted to 51,115 respondents and excluded patients who did not respond to Q18 (n = 6413) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 11,209). According to the likelihood ratio test, the model using ethnic sub-categories was not significantly better (at 95% confidence level) than the model using broad categories (χ 2 (10) = 6.75, p = .75). f Analysis was restricted to 48,025 respondents and excluded patients who did not respond to Q19 (n = 10,230) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 10,482). According to the likelihood ratio test, the model using ethnic sub-categories was significantly better (at 95% confidence level) than the model using broad categories (χ 2 (10) = 28.51, p = .002). Irish ethnicity were more likely to rate their care positively. There are many hypotheses as to why patients' ratings of care experiences vary by ethnicity; for example, it has been suggested that ethnic minority patients may receive the same care but rate it more poorly due to different expectations (Lyratzopoulos et al. 2012). However, other studies have presented evidence that expectations of care are in fact similar between ethnic groups, particularly in the area of communication (Weinick et al. 2011). An alternative hypothesis asks if ethnic minority patients have more challenging clinical cases. As this was a secondary analysis of patient experience survey data limited clinical information was available and it was not possible to determine the complexity of patients' clinical cases, but we were able to control for the effects of tumour group and self-reported response to treatment in our analyses and so it is unlikely that clinical differences could account for the results observed in this study. Other studies have shown that a large proportion of observed variation in care ratings may be due to ethnic minorities being concentrated in poorly performing services (Lyratzopoulos et al. 2012); however, the inclusion of a random effect for NHS Trust of treatment in our multi-level model accounts and adjusts for this potential confounding due to clustering. Note: Significant associations (p < .05) are highlighted in bold. a Adjusted for age, gender, long-standing condition, time since first treatment, tumour group, patient classification, response to treatment, IMD quintile, Trust location (in/outside Greater London) and Trust. b Analysis was restricted to 40,880 respondents and excluded patients who did not respond to Q24 (n = 18,931) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 8926). According to the likelihood ratio test, the model using ethnic sub-categories was not significantly better (at 95% confidence level) than the model using broad categories according to the likelihood ratio test (χ 2 (10) = 12.80, p = .24). c For comparisons by broad ethnic categories, the reference group was White patients. d For comparisons by ethnic sub-categories, the reference group was White British patients. e Analysis was restricted to 34,877 respondents and excluded patients who did not respond to Q37 (n = 26,507) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 7353). According to the likelihood ratio test, the model using ethnic sub-categories was significantly better (at 95% confidence level) than the model using broad categories according to the likelihood ratio test (χ 2 (10) = 18.06, p = .05).
f Analysis was restricted to 33,138 respondents and excluded patients who did not respond to Q41 (n = 28,685) or had missing data for age, gender, ethnicity, time since first treatment, long-standing condition status, response to treatment or IMD quintile (n = 6914). According to the likelihood ratio test, the model using ethnic sub-categories was not significantly better (at 95% confidence level) than the model using broad categories according to the likelihood ratio test (χ 2 (10) = 9.74, p = 0.47).
Patients from different ethnic groups may also have different preferences for information related to their condition, care and treatment. For example, in a study examining the use of health information among cancer patients in Hawaii, Japanese patients preferred printed materials from healthcare providers, non-Japanese Asian patients preferred person-to-person communication and Caucasian patients preferred online sources (Kakai et al. 2003). There may also be disparities in information preferences between ethnic minority patients and their families. For example, discordance with regards preferences about prognostic information was found among migrant cancer patients and their families in Australia with families preferring non-disclosure (Mitchison et al. 2012). Although these studies were conducted in non-UK settings and involved different ethnic groups, it is likely that similar issues exist amongst minority ethnic groups in England. For example, ethnic minority patients may not be able to access information in their preferred format or receive the information they want, especially if there are language barriers, thereby leading to a less positive experience of care.
Language differences between staff and patients have been suggested as another possible cause of variation in ratings by ethnicity. As the NCPES does not gather information on the languages spoken by respondents, we were not able to investigate or control for the effects of language. According to a study estimating access to NHS translation services (Gill et al. 2009), Bangladeshis have the highest proportion of non-English speakers which suggests language may play a role in cancer patient satisfaction. Yet, this hypothesis does not explain all the patterns observed in our analysis; for example, White Irish patients rated communication with CNSs more positively than White British patients though both groups are native English speakers. Although language can be integral to ethnicity, there are many other social parameters that contribute to ethnic groupings including external aspects, such as communities and networks, and internal aspects, such as a sense of identity and values, group obligations and feelings of security and comfort (Isajiw 1993). Thus, ethnic variation in ratings may be attributable to several factors. Evidence from studies related to maternity care indicate that differences in language may in fact have less influence on care ratings than cultural barriers and staff stereotyping (Puthussery et al. 2010). Different ethnic group practices can be difficult for outsiders to understand; for example, in a study exploring communication with ethnic minorities in primary care, staff found it challenging to understand and accommodate Bangladeshi patients' needs to fit appointments around particular work requirements including 'restaurant hours' (Hawthorne, Rahman, and Pill 2003). However, there is evidence to suggest that healthcare professionals who build up long-term relationships with Bangladeshi patients begin to understand and be more tolerant of their particular needs than their colleagues (Hawthorne, Rahman, and Pill 2003). The finding from our study, that Bangladeshi participantsrated communication with a CNS more positively than with other staff groups, may indicate that a long-term, personal relationship with a named staff member improves communication among ethnic minority individuals. The mode of communication may also be an important factor in explaining the poor communication ratings from Bangladeshi patients observed in this study as Sylheti, spoken by many Bangladeshis, has no agreed written form (Duff, Lamping, and Ahmed 2001) and so individuals may not have access to understandable written information.
The data used in this analysis was taken from a national survey (2012-2013 NCPES) with a large sample size (N = 68,737) and a relatively high response rate (64%). Patients' responses were dichotomised as per the survey guidance thereby facilitating comparison of results with other studies based on this data set, and the data set was large enough to allow analysis by ethnic sub-category to reveal patterns previously masked by the broad categories more commonly used in research. Despite these strengths, there are several limitations to this study. For example, non-White patients are known to be less likely to respond to NCPES, potentially introducing selection bias and, although language support leaflets were available for survey participants, the NCPES was written in English and therefore patients with low English skills are less likely to respond and would be excluded from the study. Also, the phrasing of some questions in the NCPES makes it impossible to unpick the mode of communication or information provision the patient is rating. For example, when responding to Q25 'Did hospital staff give you information about support or self-help groups for people with cancer?' patients may be rating verbal and/or written communication. Finally, as the dataset is cross-sectional there is a possibility of recall bias.
This analysis offers insight into successes and failings in different areas of communication along the cancer care pathway; however further work is needed to understand the causes of these inequalities. Interviews or focus groups with minority ethnic groups living in the UK may be useful to understand further their experiences of care. In particular, to explore the inequalities in patient experience reported by Bangladeshi patients observed in this study, focus groups or interviews in Sylheti may be useful (Duff, Lamping, and Ahmed 2001). Additionally, surveys which record both ethnicity and language of preference may provide insight into the role of language in patient experience.
Regardless of the causes, the inequalities identified in our analysis highlight challenges in communicating with and providing care to a diverse patient population and illustrate the need for services to adapt in order to provide the best possible care to all patients, regardless of ethnicity. These inequalities may also indicate that previously implemented policies, such as national protocols for delivery of information related to treatment and operations, have been insufficient. It is important to note that effective communication is not simply about giving all the relevant information to a patient, but also about tailoring information delivery to the needs, beliefs and values of the individual (Watts et al. 2004). Services should therefore be adapted to improve communication in the presence of language and/or cultural variation and policy-makers need to be aware of the inequalities that exist and ensure that new policies address these issues and support individuals in most need (El Ansari et al. 2009). Alternative methods of information provision, tailored to specific needs and preferences of minority ethnic groups may prove effective, such as audio or video information in Sylheti for Bangladeshi patients who find understanding written information challenging . The education of healthcare professionals, from students to consultants, should address ethnic bias (Dedier et al. 1999) and the need to understand patients as individuals (Kai et al. 1999) so as to ensure the best possible care and patient experience.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
The research was supported by the Imperial College Healthcare Charity (7006/P31U) and the National Institute for Health Research (NIHR) Biomedical Research Centre based at Imperial College Healthcare NHS Trust and Imperial College London. The views expressed are those of the authors and not necessarily those of the NHS, NIHR or Department of Health.

Key messages
(1) Cancer patient experience is known to vary by ethnicity, but variation within the broad ethnic categories typically used in research is poorly-described.
(2) This secondary analysis of data from the 2012-2013 National Cancer Patient Experience Survey provides evidence of variation in patients' ratings of communication and information provisionkey domains of patient experienceboth between and within ethnic categories.
(3) Compared to White patients, Non-White patients (particularly Asian patients) were less likely to report positive experiences, and among Asian patients those of Bangladeshi ethnicity reported the poorest experiences.
(4) Further work to understand the causes of variation in cancer patients' experiences of information provision and communication is required to address ethnic inequalities at practice and policy level.