How representative are responders to the english cancer patient experience survey of english breast cancer patients? A cohort study using linked national cancer patient experience survey and cancer registry data

Abstract

Objective

The annual National Cancer Patient Experience Survey (CPES) gives insights into the quality of care received by patients with cancer treated in the NHS. This study assesses how the sample of breast cancer (BC) patients who responded to CPES in England compares to the population diagnosed with breast cancer.

Design

Retrospective population-based cohort study, conducted as part of the National Audit of Primary Breast Cancer and the National Audit of Metastatic Breast Cancer.

Setting

National Cancer Patient Experience Survey (CPES) responses linked to National Cancer Registration data for BC patients (stage 0–4) diagnosed between 2015 and 2021.

Participants

315 676 people diagnosed with BC 2015–2021.

Primary and secondary outcome measures

Logistic regression was used to assess the association between personal, disease, and clinical characteristics and the odds of responding to CPES.

Results

CPES responses were provided by 44 809 individuals,14.2% of breast cancer registration records. Most survey responses (69%) were from patients in their first year since diagnosis. Response rates were significantly lower among men (adjusted odds ratio (aOR) 0.79; CI 0.68 to 0.92; p=0.002), younger (18–39 years, aOR 0.51; CI 0.48 to 0.54; 9<0.001) and older patients (≥90 years, aOR 0.22; CI 0.18 to 0.27; p<0.001), individuals from non-White ethnic backgrounds (e.g. Asian/Asian British, aOR 0.70; CI 0.66 to 0.75; p<0.001), and those living in the most socioeconomically deprived areas (aOR 0.70; CI 0.67 to 0.72; p<0.001), leading to underrepresentation of these groups. By contrast, higher response rates were observed among patients with stage 4 disease (aOR 1.69; CI 1.57 to 1.82; p<0.001) and those receiving multimodal anti-cancer treatment (aOR 3.01; CI 2.92 to 3.10; p<0.001). CPES responses were also less common among patients diagnosed during 2020–2021, during the COVID-19 pandemic.

Conclusion

People with breast cancer who return CPES questionnaires differ in their characteristics to the population of people diagnosed with breast cancer. CPES is a valuable tool for supporting quality improvement, but this study highlights the need for targeted strategies to increase participation among underrepresented groups.

Strengths and Limitations of this study.

• Large, national cohort with over 3 15 000 patients, enabling robust and comprehensive comparison between CPES responders and the cancer registry population.

• Novel linkage of individual-level cancer registry data with CPES responses across multiple years, allowing detailed examination of response patterns.

• Unable to describe characteristics of patients invited but who did not respond due to CPES sampling limitations.

• COVID-19 pandemic and changes in CPES methodology during 2020–2021 may affect comparability with earlier years.

Introduction

The patient experience of care is a fundamental dimension of healthcare quality, alongside patient safety and treatment effectiveness.¹ For people with cancer, the care pathway can be complex, often involving multiple treatments and interactions with different healthcare professionals. Understanding how patients experience care coordination, the provision of information, and involvement in decision-making can provide valuable insights and help drive improvements in the overall quality of cancer care (2).

The National Cancer Patient Experience Survey (CPES) in England has served as a key tool for capturing patient-reported experiences of care within the National Health Service (NHS) since 2010.² The survey evaluates multiple aspects of the care pathway, including diagnosis, treatment and aftercare. Adult patients (aged 16 and over) with a confirmed diagnosis of cancer are eligible for inclusion if they: (a) had an inpatient hospital admission or (b) were seen as a day case for cancer related treatment within the sampling period. A consistent 3 month sampling window from first April to 30^th June is used each year. The survey is sent to all patients undergoing cancer treatment and not just those newly diagnosed.

The results from the survey are published annually in a national report alongside reports for individual NHS trust, cancer alliances and integrated care boards.^{3 4} The data have also been made available for studies by other research groups.^5–8 These publications have highlighted various aspects of patient experience, and where the quality of care might be improved. However, the method of sampling used by CPES means that not all people with cancer are equally likely to be sampled. In an analysis of CPES responders between 2010 and 2014 with four common cancers (lung, breast, prostate and colorectal), Alessy et al⁹ reported that people from a White ethnic background, resident in less deprived areas, and diagnosed with earlier stage disease were over-represented. Nartey et al¹⁰ extended these results for people with lung cancer, reporting that older people, and those with worse performance status, and multiple comorbidities were under-represented in CPES responses between 2009 and 2015. This study undertook an in-depth evaluation of CPES responses from recent years (2015–2021) among people with breast cancer, comparing the personal, disease, and clinical characteristics of respondents with those of all individuals diagnosed with breast cancer in England each year.

Methods

Study population, data sources & linkage

This study was undertaken as part of the National Audits of Primary Breast Cancer (NAoPri) and Metastatic Breast Cancer (NAoMe) and used pseudonymised patient and tumour-level data supplied by the National Cancer Registration and Analysis Service (NCRAS).^11–13 Details of NAoPri and NAoMe can be found in their State of the Nation Reports and associated Methodology documents.^{14 15} NCRAS curate the National Cancer Registry Dataset that contains information on all patients diagnosed with breast cancer and registered in England and includes information from hospital pathology systems. There is a delay of several years before the information on individuals diagnosed in a specific year are made available to ensure the dataset is complete.

The study population comprised adults (≥18 years) diagnosed with breast cancer (ICD-10 codes: C50, D05; TNM stages 0–4) in England between January 2015 and December 2021. Cancer registration records were linked at an individual patient level to CPES responses from the surveys conducted between 2017 and 2021. Not all CPES records were linked (see online supplemental figure 1 for details): 4315 records had patient identifiers that did not match any in registration records, 4480 were excluded because they were from individuals with multiple responses and only the response closest to diagnosis was retained. A further 93 were excluded as respondents were diagnosed after recruitment closed.

Registry records were also linked to other national healthcare datasets to derive treatment details and capture events related to cancer care, including the Cancer Outcomes and Services Dataset (COSD), Hospital Episode Statistics – Admitted Patient Care (HES APC), the Radiotherapy Dataset (RTDS), and the Systemic Anti-Cancer Therapy (SACT) dataset.

National cancer patient experience survey

CPES captures patient perspectives on cancer care quality and identifies areas for improvement. All eligible patients — those ≥16 years, admitted as inpatients or day cases for cancer-related care between 1st April and 30^th June — are invited to complete the survey following discharge. As the survey is not limited to newly diagnosed patients, individuals may be invited in multiple years. The questionnaire contains approximately 70 multiple choice questions that cover the care pathway from diagnosis to aftercare,¹⁶ with responses categorised as positive, negative or neutral. It is distributed by post, with an online completion option. Individuals receive up to two reminders, with helplines and translation services are available. The survey uses a consistent annual sampling method and questionnaire to allow trend analysis. In 2020, only patients from 55 English trusts were invited to participate due to COVID-19-related service prioritisation. In other years, around 130 NHS trusts participated.

Study variables

The Cancer Registration dataset was the primary source for patient demographics (age, gender, ethnicity) and clinical characteristics (route, date, and stage of diagnosis).¹⁷ Socioeconomic status was measured using area-based Index of Multiple Deprivation (IMD) quintiles. Performance status, derived from COSD was used to assess fitness for treatment. Comorbidity burden was assessed with Charlson Comorbidity Index (CCI) scores derived from ICD-10 codes in HES APC records that related to any admissions during the 24 months before diagnosis.¹⁸ Treatment details were drawn from the HES (surgery), RTDS (radiotherapy), and SACT (chemotherapy) datasets. Six treatment categories were analysed: (1) no treatment, (2) mastectomy, (3) mastectomy and radiotherapy, (4) breast conserving surgery (BCS), (5) BCS and radiotherapy, and (6) multimodal treatment, incorporating surgery and chemotherapy (with or without radiotherapy). No endocrine treatment details were included in this study. Surgical variables were defined as a mastectomy (with or without immediate reconstruction) or BCS within 12 months of their diagnosis date.¹⁹ Receipt of adjuvant radiotherapy was defined as the first date of radiotherapy being after the date of surgery. Receipt of neoadjuvant or adjuvant chemotherapy was combined to report on the effect of any chemotherapy. Patients recorded as receiving radiotherapy only (n=552), likely reflecting missing surgical data from independent hospitals, were categorised as having received BCS plus radiotherapy. As this group was much larger than the mastectomy plus radiotherapy group, any misclassification is expected to have minimal impact on the results.

Statistical analysis

All statistical analyses were conducted using Stata version 17. The dataset variables were summarised using descriptive statistics. For each variable group, the number of patients represented in CPES were expressed as a percentage of the study population within that group, and the significance of the difference in proportions of CPES responders between variable groups was assessed with Chi-square tests. Smoothed plots were generated to visualise patterns in response rates by age at diagnosis, stratified by outcome variables such as year of diagnosis, disease stage, and treatment group. The underlying trends were estimated using a non-parametric nearest-neighbour smoother.

Univariable and multivariable logistic regression models were used to estimate ORs for the likelihood of responding to CPES, according to personal, disease and clinical characteristics. The multivariable model included year of diagnosis, gender, age at diagnosis, ethnicity, deprivation quintile, disease stage, CCI, screen-detected status, and treatment received. Multicollinearity between explanatory variables was assessed during the data exploration stage using cross-tabulations, with no evidence of strong correlations identified. Although treatment modality reflects underlying clinical characteristics, including disease stage, age, and comorbidity, it was retained because the CPES sampling frame was structured such that treatment patterns directly influenced the likelihood of patient inclusion. Consequently, the estimated associations should be interpreted as measures of association rather than causal effects. All statistical tests were two sided and p-values<0.001 were considered to demonstrate statistical significance given the large sample size.

We performed two sensitivity analyses (see online supplemental figure 2). Sensitivity analysis 1 included only patients diagnosed before the sampling window closed in their diagnosis year and who responded in that same year, thereby making the care pathways of eligible individuals more similar. Sensitivity analysis 2 further restricted this group to diagnoses from 2017 to 2019 to reduce the impact of the COVID-19 pandemic.

Several variables had low levels of missing data: stage of disease (9.47%), ethnicity (5.604.84%), and Charlson Comorbidity Index (CCI) (3.47%). For each of these variables, patients with missing values were assigned to an “unknown” category, allowing them to be retained in the regression analysis. Performance status (PS) was frequently missing (40.24%) and was reported only in the descriptive analysis.

Results

The cancer registration dataset contained 315 676 people diagnosed with breast cancer (stage 0–4) in England between January 2015 and December 2021. There were between 44 840 and 48 690 patients diagnosed each year, with the exception of 2020, when the first wave of the COVID-19 pandemic affected the delivery of cancer services. The number of people diagnosed during 2020 was 38 291.

A total of 44,809 CPES questionnaire responses were linked to patients in the cancer registration dataset, representing 14.2% of cancer registration records 2015–2021 (see online supplemental figure 1) for a flowchart of study population generation). The highest linkage proportions were observed for patients diagnosed in 2017 (21%) and 2018 (23%), reflecting the availability of survey data for the year of diagnosis as well as subsequent CPES surveys. The proportion of linked responses declined in later years, with only 12.7% of patients diagnosed in 2020–2021 having a linked CPES response.

Table 1 shows how patients diagnosed in a given year could contribute responses from different CPES surveys. With the exception of 2020, over 60% of CPES responses linked to patients diagnosed in a particular year came from the survey conducted in that year and would reflect the initial diagnostic / treatment planning process and primary treatment. This proportion increased as the availability of subsequent CPES years decreased, although few CPES responses were from surveys conducted more than 2 years after diagnosis. Few patients diagnosed in 2015–2016 had linked survey responses, as the first CPES data were from 2017. Overall, 97% of CPES responses were received within the year of diagnosis or the following year.

Table 1. Number of CPES responses linked to the people diagnosed with breast cancer, by year of diagnosis.

Year	No. of patients diagnosed in cancer registration records	Responses from each CPES year linked to cancer registration records					Total linked	(%)
		2017	2018	2019	2020	2021
2015	44 964	465	198	124	28	64	879	2%
2016	44 840	3258	375	169	40	70	3912	9%
2017	44 906	5796	3474	222	30	79	9601	21%
2018	46 920		7158	3396	67	152	10 773	23%
2019	47 065			6791	1262	546	8599	18%
2020	38 291				1721	3154	4875	13%
2021	48 690					6170	6170	13%
Total	315 676	9519	11 205	10 702	3148	10 235	44 809	14%

CPES, Cancer Patient Experience Survey.

The remaining analysis was restricted to patients diagnosed with breast cancer between 2017 and 2021, ensuring both datasets covered the same time period. Following this exclusion, the study population comprised 225 872 individuals, of whom 40 018 (17.7%) had a CPES response.

Tables 2 and 3 describe the proportion of people who returned a CPES response across a variety of personal, disease and clinical characteristics, respectively. The tables give the raw figures and unadjusted ORs for the various subgroups as well as the adjusted ORs from the multivariable regression model and accounts for the effect of the other variables presented.

Table 2. Comparison of patient demographics of CPES responders to full study population.

Characteristic	People diagnosed n=2 25 872	Total CPES responders n=40 018	% of people diagnosed	Unadjusted OR of CPES response	95% CI		Adjusted OR of CPES response	95% CI		P value
Year of diagnosis										<0.001
2017	44 906	9601	21.4%	(Ref)1.00			(Ref)1.00
2018	46 920	10 773	23.0%	1.10	1.06	1.13	1.13	1.10	1.17
2019	47 065	8599	18.3%	0.82	0.80	0.85	0.85	0.83	0.88
2020	38 291	4875	12.7%	0.54	0.52	0.56	0.53	0.51	0.55
2021	48 690	6170	12.7%	0.53	0.52	0.55	0.53	0.51	0.55
Gender										0.002
Female	224 222	39 798	17.7%	(Ref)1.00			(Ref)1.00
Male	1650	220	13.3%	0.71	0.62	0.82	0.79	0.68	0.92
Age (years)										<0.001
18-39	10 092	1824	18.1%	0.83	0.78	0.88	0.51	0.48	0.54
40-49	29 487	5740	19.5%	0.91	0.88	0.94	0.67	0.65	0.70
50-59	54 293	10 752	19.8%	0.93	0.90	0.96	0.82	0.80	0.85
60-69	55 528	11 671	21.0%	(Ref)1.00			(Ref)1.00
70-79	43 972	7796	17.7%	0.81	0.78	0.84	0.94	0.91	0.98
80-89	26 129	2127	8.1%	0.33	0.32	0.35	0.65	0.61	0.68
90-99	6371	108	1.7%	0.06	0.05	0.08	0.22	0.18	0.27
Ethnicity										<0.001
White	193 402	35 478	18.3%	(Ref)1.00			(Ref)1.00
Asian or Asian Bri	9163	1279	14.0%	0.72	0.68	0.77	0.70	0.66	0.75
Black or Black Bri	5123	746	14.6%	0.76	0.70	0.82	0.75	0.69	0.82
Mixed	1543	231	15.0%	0.78	0.68	0.90	0.75	0.65	0.87
Other Ethnic Group	3990	648	16.2%	0.86	0.79	0.94	0.83	0.76	0.90
Unknown	10 942	1603	14.6%	0.76	0.72	0.81	0.81	0.77	0.86
IMD Quintile										<0.001
Least deprived - 1	53 311	10 106	19.0%	(Ref)1.00			(Ref)1.00
2	50 936	9586	18.8%	0.99	0.96	1.02	0.97	0.94	1.00
3	47 339	8546	18.1%	0.94	0.91	0.97	0.92	0.89	0.95
4	40 512	6862	16.9%	0.87	0.84	0.90	0.85	0.82	0.88
Most deprived - 5	33 774	4918	14.6%	0.73	0.70	0.76	0.70	0.67	0.72

Adjusted ORs are from a multivariable model including year of diagnosis, gender, age, ethnicity, deprivation, disease stage, Charlson Comorbidity Index, screen-detected status, and treatment received.

CPES, Cancer Patient Experience Survey; IMD, Index of Multiple Deprivation; OR, Odds Ratio.

Table 3. Comparison of clinical characteristics of CPES responders to full study population.

Characteristic	People diagnosed n=2 25 872	Total CPES responders n=40 018	% of people diagnosed	Unadjusted OR of CPES response	95% CI		Adjusted OR of CPES response	95% CI		P value
Stage										<0.001
Stage 0	22 737	2228	9.8%	0.44	0.42	0.46	0.67	0.64	0.70
Stage 1–3 a	163 601	32 625	19.9%	(Ref)1.00			(Ref)1.00
Stage 3b,3c	8199	1624	19.8%	0.99	0.94	1.05	0.91	0.86	0.97
Stage 4	9947	1327	13.3%	0.62	0.58	0.66	1.69	1.57	1.82
Unknown	21 388	2214	10.4%	0.46	0.44	0.49	0.74	0.70	0.77
Performance Status
0	107 739	21 659	20.1%	(Ref)1.00
1	16 495	2521	15.3%	0.72	0.69	0.75
2	6089	471	7.7%	0.33	0.30	0.37
3	3808	110	2.9%	0.12	0.10	0.14
4	855	23	2.7%	0.11	0.07	0.17
Unknown	90 886	15 234	16.8%	0.80	0.78	0.82
Co-morbidity (Charlson Index)										<0.001
0	182 683	36 131	19.8%	(Ref)1.00			(Ref)1.00
1	21 276	2844	13.4%	0.63	0.60	0.65	0.82	0.78	0.86
2	8571	771	9.0%	0.40	0.37	0.43	0.70	0.64	0.75
3+	5511	243	4.4%	0.19	0.16	0.21	0.46	0.40	0.52
Unknown	7831	29	0.4%	0.02	0.01	0.02	0.05	0.03	0.07
Screen Detected										0.382
No	165 041	27 941	16.9%	(Ref)1.00			(Ref)1.00
Yes	60 831	12 077	19.9%	1.22	1.19	1.24	0.99	0.96	1.02
Treatment Received										<0.001
No-Treatment	38 591	1747	4.5%	0.25	0.24	0.26	0.34	0.32	0.37
Mastectomy	24 897	2966	11.9%	0.71	0.68	0.74	0.82	0.78	0.86
Mastectomy+RT	9048	1535	17.0%	1.07	1.01	1.14	1.18	1.11	1.25
BCS	21 753	2412	11.1%	0.66	0.63	0.69	0.79	0.75	0.83
BCS+RT	75 123	11 999	16.0%	1.00			1.00
Multimodal	56 460	19 359	34.3%	2.75	2.67	2.82	3.01	2.92	3.10

Adjusted ORs are from a multivariable model including year of diagnosis, gender, age, ethnicity, deprivation, disease stage, Charlson Comorbidity Index, screen-detected status, and treatment received.

BCS, Breast Conserving Surgery; CCI, Charlson Comorbidity Index; CPES, Cancer Patient Experience Survey; OR, Odds Ratio; RT, Radiotherapy.

As noted above, the odds of having a linked CPES response declined in later diagnosis years, reflecting the reduced follow-up within the study. Response rates for those diagnosed in 2020 and to a lesser extent 2021, were further impacted by the COVID-19 pandemic. There were 1650 men in the study cohort (0.73%). The proportion of responses received from men was lower than women, but this effect was reduced slightly after adjustment (adjusted odds ratio (aOR) 0.79; CI 0.68 to 0.92; p=0.002).

There was a strong association between age at diagnosis and the proportion of CPES responses, with a small increase up to 60–69 years of age, after which there was a steady decline. This pattern was evident for each year of diagnosis (figure 1). Following adjustment, this pattern became more evident, suggesting other factors were masking the influence of age on response rates.

Figure 1. Relationship between age at diagnosis and percentage representation in CPES, stratified by survey year.

All ethnic groups had lower odds of returning a CPES response compared those from a White ethnic background. People from Asian, black and mixed ethnic groups were the least likely to return a CPES questionnaire (aOR 0.70; CI 0.66 to 0.75 Asian or Asian British, aOR 0.75; CI 0.69 to 0.82 black or black British, aOR 0.75; CI 0.65 to 0.87 mixed; p<0.001). The proportion of people with a CPES response also reduced steadily with increasing deprivation, falling from 19.0% in the least deprived to 14.6% in most deprived quintile. This was not ameliorated after accounting for other factors (Quintile 5 - aOR 0.70; CI 0.67 to 0.72; p<0.001).

The likelihood of a person returning a CPES survey was also associated with various clinical factors (table 3). A greater proportion of responses was received from people with invasive compared with those with non-invasive disease. Among invasive cancers, fewest responses were received from people with stage 4 disease. However, after adjustment for other factors, patients with stage 4 disease were more likely to return a survey response than the reference (aOR 1.69; CI 1.57 to 1.82; p<0.001). This finding was consistent across age deciles 30–90 years (figure 2).

Figure 2. (a) Stage of disease by age in the breast cancer population. (b) The relationship between response rates, stage of disease and age for CPES years 2017–2021.

The proportion of responses decreased with increasing performance status and greater numbers of comorbidities. There was little difference in the proportion of responses depending on whether or not a person’s cancer was screen detected.

Treatment patterns had a strong association with the likelihood of patients returning a CPES questionnaire. The proportion of responses increased with greater treatment intensity, reaching its highest level among patients in the multimodal group, who received chemotherapy alongside surgery with or without radiotherapy (multimodal aOR 3.01; CI 2.92 to 3.10; p<0.001). The observed proportions within the treatment groups varied slightly among people of different ages, with all older age groups less likely to receive multimodal treatment (figure 3). However, the addition of an interaction term for treatment category and age did not improve the model fit substantially and was therefore omitted from the table to aid interpretation.

Figure 3. (a) Treatment received by age in the BC population. (b) The relationship between response rates, treatment received and age for CPES years 2017–2021.

Sensitivity analysis 1 included 102 297 breast cancer patients diagnosed before survey closure (2017–2021), of whom 27 554 (26.9%) responded in their diagnosis year. Annual response rates ranged from 27.8–33.6%, except in 2020, when participation fell to 10.6% due to modified survey methods and restricted NHS trust recruitment. Patterns of representation mirrored the main cohort: the youngest and oldest patients, non-white ethnicities, the most deprived quintile, and those with greater frailty or comorbidity remained underrepresented. Representation by disease stage was similar, and treatment modality continued to influence participation, with multimodal therapy associated with higher response, though effects were attenuated. Additional restrictions in sensitivity analysis 2 did not reduce this variation in representation (see online supplemental table 1 for aORs).

Discussion

Summary of findings

This study analysed data on 315 676 breast cancer patients (stages 0–4) linked to 44,809 CPES responses to explore how the characteristics of people who return a CPES questionnaire corresponded to the characteristics of people diagnosed each year. It also used data from multiple registration and survey years to investigate whether patients were responding in their year of diagnosis or in subsequent years. Overall, there were various distinct patterns in the response rates. A lower proportion of responses were received from men compared with women, and responses fell among patients of older age, with less than 2% of people aged 90+returning a questionnaire. The proportion of responses was lower among patients of non-White ethnicity, or from greater socioeconomic deprivation areas. Non-invasive disease, and higher levels of frailty were also associated with reduced response rates.

Some of the patterns of response were not linked to patient characteristics. Treatment intensity influenced the likelihood of an individual being eligible for the survey and it was not surprising to find a greater proportion of responses from people receiving multimodal therapy (surgery+chemotherapy ± radiotherapy). In addition, the latter 2 years in the study were affected by reduced follow-up, whereby those responding to CPES surveys one (2021) or two (2020) years after diagnosis were not captured. Patients diagnosed in 2020 were also underrepresented compared with those diagnosed 2017–2019 because of changes in sampling methods and treatment pathways resulting from the COVID-19 pandemic. Unlike other years, patients diagnosed in 2020 were mostly likely to respond to CPES in the year following their diagnosis (64.7%). Two sensitivity analyses, designed to reduce the impact of (1) heterogeneity due to the sampling design and (2) the COVID-19 pandemic, did not change the extent of variation observed.

Study strengths and limitations

The large sample size, national scope, and linkage of cancer registration data and CPES responses at individual patient level across multiple years are significant strengths of this study, and enable a comprehensive analysis of how CPES response rates vary among different subgroups of people with breast cancer.

The study has various limitations. First, the study was unable to describe the characteristics of (1) the patients who were invited but did not respond to the CPES, or (2) the 4,315 CPES respondents who could not be linked to a cancer registration record and were therefore excluded. Consequently, potential systematic differences between included and excluded respondents could not be assessed. The CPES sampling process involves inviting all patients undergoing treatment within a defined 3 month sampling window, and the resulting population will differ from the cancer registration population as its process focuses on patients after their initial diagnosis. In addition, residual confounding from unmeasured factors cannot be excluded.

Second, the COVID-19 pandemic limits the comparability of 2020–2021 data with earlier years because of the disruption to healthcare services, and the revised CPES methodology. Restricting analyses to those patients diagnosed in 2017–2019 did not, however, greatly change the size and type of variation in the patterns of response observed.

Finally, the supplied cancer registration data include only patients aged ≥18, whereas the CPES survey includes people aged ≥16 years; given the rarity of breast cancer in 16–17 year-olds, this mismatch is unlikely to affect the results.

Comparison with other studies

The results for this cohort of breast cancer patients are generally consistent with previous research. Studies of CPES response rates from people diagnosed with various types of cancer have also reported responders were on average younger, and from less deprived areas, and were less likely to be from ethnic minority backgrounds compared with the national cancer registry population.^{9 10 20} Efforts to improve representation from different ethnic minorities and more deprived backgrounds should be made by survey organisers to ensure cancer services understand the needs of these individuals, not least because people from ethnic minorities often have aggressive tumours, present with late stage disease and experience poorer survival.²¹ Individuals from lower socioeconomic backgrounds face greater barriers to healthcare engagement, which may reflect broader issues of access, healthcare literacy, or perceived value of participation.²²

The overrepresentation of those receiving multimodal treatment aligns with the findings of Nartey et al,¹⁰ who reported similar patterns when evaluating CPES representation in patients with lung cancer. As CPES sampling is based on inpatient or day-case admissions within the sampling window, the duration of anti-cancer treatments will directly influence inclusion and introduces length-biased sampling. Patients with early-stage disease are less likely to be sampled, while those undergoing multiple admissions for treatments like chemotherapy and radiotherapy are more likely to be included.

Accurate representation of the breast cancer population by CPES responders requires both appropriate sampling processes and adequate response rates. Sampling linked to treatment episodes introduces bias against less intensively treated patients, and post-sampling mortality can further limit representativeness, particularly for some people with advanced cancer.²³ Although CPES response rates are relatively high compared with other national surveys, they have been declining: from 63% in 2017 to 61% in 2019, 55% in 2021, and 52% in 2023.²⁴ This trend increases the risk of bias, particularly as younger patients remain underrepresented; data from the 2023 CPES survey across all tumour types showed that patients aged 55–64 were twice as likely to respond as those aged 25–34.²⁴

Questions have been raised about whether the current CPES sampling process provides a suitable basis for hospital-level performance comparisons.²⁵ An independent analysis of 2016 CPES data reported that hospitals with higher response rates tended to achieve better patient experience scores but concluded that non-response did not meaningfully bias overall scores.²⁶

Implications

Despite its limitations, CPES remains a uniquely valuable resource for understanding patient experience at a national level. The survey is the most comprehensive source of information on the experience of patients using English NHS cancer services, and its findings are helpful for informing policy and improving outcomes.²⁷ Our study highlights the benefit of linking individual CPES responses to cancer registration, and greater use of this might be made in the reporting of CPES results. The CPES survey respondents clearly differ from the population of people diagnosed with breast cancer each year in England, and approaches that remove some of this variation to aid the interpretation of results should be considered. Initiatives might involve weighting the results to remove the effects of the length-biased sampling, although whether it is possible to develop a robust weighting process would need to be evaluated. Study findings also suggest that institutional comparisons based on CPES-derived quality metrics should be interpreted with caution, particularly where differences in patient case mix or patterns of response rates may influence results. Future work should explore strategies to improve participation among underrepresented groups to enhance the representativeness and equity of CPES findings. An alternative is to use the linked data to report on the experiences of patients within specific subgroups, and at specific points in their care pathway. This would help to identify areas for targeted improvement more readily than overall summary statistics.

However, these developments still require better representation of specific patient groups, including older patients, and those from diverse backgrounds. This will require a variety of engagement strategies and care will be needed to consider how these can be done without introducing an excessive burden to hospital staff. Nonetheless, ensuring the CPES survey is more accessible and inclusive is required if the delivery of breast cancer care is to be shaped by a diversity of patient voices.

Data availability statement

Data may be obtained from a third party and are not publicly available.