Association of travel time, patient characteristics, and hospital quality with patient mobility for breast cancer surgery: A national population‐based study

This national study investigated hospital quality and patient factors associated with treatment location for breast cancer surgery.


INTRODUCTION
5][6] Patients who bypass their nearest hospital are typically younger, fitter, and more affluent.Patients are more likely to travel to hospitals offering advanced technologies, such as robotic surgery, 7 but not necessarily hospitals with the best disease-specific outcomes. 8derstanding patient mobility for cancer services is important.
First, new policies are being developed that aim to use patient choice to better match demand (patients experiencing long waiting times) to supply (hospitals with shorter waiting times), especially in the context of the postpandemic cancer backlogs. 9Second, there is emerging evidence that hospitals located in competitive areas where patients have the opportunity to select from a wider range of providers offer better quality care, which suggests that there is a trade-off between centralizing health services and using competition to incentivize quality improvement. 10,11Third, when more affluent and younger patients are able to travel to hospitals that provide better quality, this may widen inequalities in access and outcomes for marginalized groups. 12east cancer is an important cancer type in which to investigate patterns of patient mobility because the care for patients with breast cancer is rapidly evolving; for example, the diffusion of oncoplastic techniques for breast-conserving surgery (BCS) and breast reconstruction. 13In many countries, there is also a wide range of patient forums and cancer charities helping them to make decisions about the hospitals in which they receive treatment. 14,15e NHS in the United Kingdom is an ideal health system in which to understand the impact of patient choice policies in cancer care. 16It is a publicly funded, single-payer system in which, in principle, patients have the opportunity to select any hospital providing cancer treatment. 17Over 95% of all cancer care is delivered in the NHS, and national administrative data sets are available that capture the care they receive. 18In addition, it is one of the few countries to publicly report information on the quality of cancer care. 19 this national population-based study, we investigate how many patients with breast cancer who had a major primary surgical resection in the English NHS bypassed their nearest surgical center for treatment.We then present these mobility patterns according to patient characteristics as well as the extent to which certain characteristics make hospitals more attractive to patients.Finally, we discuss the implications of our results for designing national health policies and provider incentives to ensure effective, efficient, and fair functioning of health care systems.

Data sources
We used data from the National Cancer Registration and Analysis Service 20 linked to the NHS Hospital Episode Statistics (HES) database. 18HES data provided information on patient-level characteristics, including the patient's residence, age, sex, the number of comorbidities according to the Royal College of Surgeons' Charlson comorbidity index, 21 ethnicity, and socioeconomic deprivation expressed in terms of quintiles of the national distribution of the Index of Multiple Deprivation (IMD) in 2015. 22HES data also provided information regarding the treating hospital, the date of surgery, and the type of breast cancer surgery; for example, mastectomy, autologous reconstruction, and the occurrence of breast re-excisions after BCS.Breast cancer surgery procedure information was coded according to the Office of Population Censuses and Surveys Classification of Surgical Operations and Procedures, 4th Revision. 23The rurality of the area of residence was captured as rural, urban (non-London), or London. 24The National Cancer Registry data provided information on cancer stage.

Population
We obtained individual patient-level data for all patients who had been diagnosed with breast cancer between January 1, 2016, and December 31, 2018, and who subsequently underwent either BCS or a mastectomy with or without reconstruction in the NHS in England.
Patients with breast cancer were identified in the National Cancer Registry data using the International Classification of Disease, 10th Revision 25 code C50.Patients with these breast cancer codes were included if their sex was recorded as female and if there was no other cancer diagnosis 1 month before and 1 month after the breast cancer diagnosis.For patients who had multiple diagnoses of breast cancer in the National Cancer Registry data, we used information on the earliest diagnosis record.Patients were included in our analysis if they had undergone elective resection and if they were treated in one of the 166 English NHS hospitals routinely performing breast cancer surgery (hospitals that perform at least 10 procedures per year).Patients with metastatic disease as well as patients who underwent surgery in a private sector hospital were excluded.

Patient characteristics
Six patient characteristics were included in our analysis: age, socioeconomic deprivation, 22 the number of comorbidities, 21 ethnicity, cancer stage, and residential area classified as rural', urban (outside London), or London. 26

Hospital characteristics
We used seven hospital characteristics that that may make a hospital more attractive to patients and their primary or secondary care physicians when considering where to have surgical treatment.These variables were informed by the peer-reviewed literature, 7 the national breast cancer organizational survey undertaken by the UK National Breast Cancer Audit, 27 and the study's patient and public involvement group and Steering Committee.
� Treatment availability: We identified 49 multidisciplinary cancer centers as those hospitals that offer both breast cancer surgery and radiotherapy on the same site and all provide neoadjuvant chemotherapy and adjuvant radiotherapy for breast cancer.
� Specialist breast reconstruction center: We identified 45 hospitals that performed at least 20 breast reconstructions after mastectomy using autologous non-implant or expander-based techniques per year.
� Media reputation: We identified 11 hospitals with a strong media reputation based on employing breast cancer surgeons who were listed in 2018 as the best breast cancer surgeons in the United Kingdom according to The Daily Mail, a leading national paper in the United Kindom. 28Overall hospital performance rating: We identified 12 hospitals as providing inadequate care according to the performance rating system of the UK Care Quality Commission, which provides a composite metric for hospital quality and is published online. 29Research activity: We defined 31 high-research activity hospitals using an established method based on trial recruitment 30 that considered research activity at a hospital according to the number of participants recruited at each hospital per year to studies funded by the National Institute for Health Research in 2018-2019. 6Cancer waiting times: We identified hospitals that met cancer waiting time targets (i.e., to start treatment within 31 days from the decision to treat date) between January 2016 and during the 36-month period from 2016 to 2018.
� Re-excision rates: We identified 33 centers with the highest reexcision rates (greater than 20% re-excision rates) after elective BCS.

Travel time
Patients' residential locations were represented by the populationweighted centroids of their Lower-Layer Super Output Areas (LSOAs).There are 32,844 LSOAs in England, defined as small areas that typically include 1500 residents or 650 households. 24

Bypassing hospitals
Patients who were not treated in the hospital nearest to them, were classified as bypassers.

The association of travel time, hospital, and patient characteristics with treatment location
We applied conditional logistic regression models to estimate the association between travel time and hospital and patient characteristics and where patients received surgery. 5,6For each of the 166 hospitals providing breast cancer surgery, a binary dependent variable was created to indicate the hospital that provided the surgery.
Travel time was included in the model as the additional travel time relative to the nearest hospital and was grouped into four categories: ≤10, 11-30, 31-60, and >60 minutes.backgrounds [IMD, 3-5] vs, patients from less deprived backgrounds [IMD, 1-2]), rural-urban classification (patient living in non-London urban areas or in London vs. patients living in rural areas), and cancer stage (patients with stage II or III cancers vs. patients with stage I cancer).We obtained robust standard errors to take into account potential clustering around the 42 regional Integrated Care Systems, which are responsible for the coordination of services provided by the English NHS. 31 Multiple imputations using chained equations were applied to create 10 imputations for the missing values in stage (2.4% for patients who underwent BCS, 5.0% for patients who underwent mastectomy) and ethnicity (4.3% for patients who underwent BSC, 3.2% for patients who underwent mastectomy).Of note, the proportion of ethnic groups identified in the NHS HES data has been shown to correlate with National Census results, which are considered a gold standard. 32Multiply imputed data sets were used for the regressions, including patient characteristics, as well as case-mix adjustment in hospital-level reoperation rates.Regression results were combined using Rubin rules.All analyses were conducted using Stata 17 (Stata Corporation).

Ethics
Ethics approval for use of secondary, anonymized, patient-level data sets for these analyses was received from the NHS Research Ethics Committee on January 6, 2020 (reference: 20/WA/0161).Informed consent was not required for use of this information.
Patient partners have co-designed this research study, including defining the primary research objectives and design of the study as part of the patient and public involvement committee.Three of the authors have lived experience of breast cancer, and they have actively contributed to the writing of this article.

RESULTS
We identified 101,750 patients who were diagnosed between January 2, 2016, and December 31, 2018, and who underwent breast cancer surgery with curative intent.Of these patients, 69,153 received underwent an elective BCS, and 33,686 underwent an elective mastectomy (see Figure S1).Of the 23

Hospital bypassing
Of the 69,153 patients undergoing BCS, 22,622 (32.7%) bypassed their nearest hospital providing breast cancer surgery; and, of the 23,536 patients undergoing a mastectomy without reconstruction, 7179 (30.5%) bypassed their nearest hospital providing breast cancer surgery (see Table S1).The proportion was higher among patients living in rural areas (36.3% for BCS and 34.1% for mastectomy without reconstruction) compared with patients living in urban non-London areas (29.1% for BCS and 27.3% for mastectomy without reconstruction; see Table S2).Figure 1 illustrates the area of residence for patients who had their BCS at a surgical center in North West England.This included patients who lived within the local area of the hospital as well the those who traveled from outside of the local area to receive care there (bypassers).Table S1 highlights the median travel time for nonbypassers and bypassers according to the number of hospitals bypassed.

Determinants of treatment location
For both BCS and mastectomy, the univariable and multivariable analyses demonstrated that travel time was strongly associated with the hospital where the patients underwent their surgery (see Table S3 and Table 2).The odds of a patient traveling to another hospital than the nearest rapidly decreased with the additional travel time.For example, the odds of patients undergoing mastectomy (without reconstruction) traveling to a hospital that was up to 10 minutes farther away than their nearest hospital was considerably lower (adjusted odds ratio [OR], 0.27; 95% confidence interval [CI], 0.22-0.33),which is in keeping with most patients receiving care at their nearest hospital.
For women receiving BCS, we found that patients were more likely to undergo surgery at a specialist breast reconstruction center (adjusted OR, 1.85; 95% CI, 1.36-2.50)and less likely to undergo treatment at hospitals with the shortest waiting times (adjusted OR, 0.65; 95% CI, 0.46-0.92).We did not identify a statistically significant association between hospital-level re-excision rates and the hospital where patients received their surgical treatment.For women undergoing a mastectomy alone, we observed that patients were more likely to travel to a specialist breast reconstruction center (adjusted OR, 1.52; 95% CI, 1.14-2.02).Women who underwent a mastectomy and IBR were five times more likely to receive this at one of the 45 specialist breast reconstruction centers (adjusted OR, 5.53; 95% CI, 3.65-8.37)although all 166 centers offered breast reconstruction.
Women were also more likely to have this procedure at hospitals that employed surgeons with a strong media reputation for breast cancer -1225 surgery (adjusted OR, 2.41; 95% CI, 1.28-4.52).In addition, patients undergoing IBR were less likely to travel to hospitals with the shortest waiting times (adjusted OR, 0.60; 95% CI, 0.40-0.90).For all three types of procedures, we did not find any association between the odds of patients traveling to a particular hospital and the overall UK Care Quality Commission hospital performance ratings, the research activity of the hospital, or whether the hospital was a multidisciplinary cancer center.
The interaction terms presented in Table 3 and Table S4) establish the variation in the association between travel time and treatment location according to six patient characteristics.We However, patients who were younger, those who had fewer comorbidities, those who were of a White ethnic background, and those who lived in rural areas were more likely to travel to alternative hospitals farther away for treatment.Patients undergoing any type of breast cancer surgery were more likely to be treated at hospitals classified as specialist breast reconstruction centers irrespective of travel time or whether or not they were undergoing a mastectomy.
Patients receiving a mastectomy with IBR were more likely to travel to hospitals that employed surgeons who had a strong media reputation for breast cancer surgery.For both BCS and mastectomy with IBR, we observed that women were less likely to travel to hospitals with the shortest surgical waiting times for treatment.
The findings of this work have several policy implications.First, the finding that elderly patients, those with comorbidities, as well as those from ethnic minority backgrounds were more likely to receive care at their local hospital suggests that the increasing centralization of services (e.g., for complex breast reconstruction) could result in inequalities in access to recommended treatments or hospitals that provide better quality care. 13,33,346][37] Therefore, it is important that policies that allow patients to choose where they receive their care include measures that mitigate against the risks that they increase inequalities in access and outcomes.This can include provision of free transport, accommodation, or even protection against loss of salary or income.In addition, given the association of travel time on the likelihood of bypassing, this is more likely to occur in larger urban conurbations, such as London, where there is a higher density of provision from which patients can select compared with rural areas.
Second, patients undergoing BCS or mastectomy were more likely to travel to hospitals that were known specialist breast reconstruction centers irrespective of whether they required a reconstruction. 38Although not standardized, there is evidence that oncoplastic techniques, such as therapeutic mammoplasty and autologous reconstructions, are associated with better patient outcomes.Therefore, it is likely that this knowledge is being informally disseminated through primary care networks and patient groups, and the availability of these oncoplastic techniques in hospitals could confer a competitive advantage increasing market share. 10,35,39,40tients undergoing BCS were not more likely to receive care at multidisciplinary cancer centers, where they would be able to have all their care (surgery and radiotherapy) at one location.
Third, we observed that the reputation of individual surgeons was associated with the hospitals where patients who received immediate reconstruction were treated.The impact of The Daily Mail on health-seeking behaviors-the newspaper with the largest circulation in the United Kingdom-has also been similarly demonstrated in prostate cancer for men receiving radical prostatectomies. 4itically, for the above two points, we do not know whether the clinical outcomes for patients treated at these hospitals are better.It is important to improve this disconnect between perceived and actual cancer treatment quality because, otherwise, it has the potential create perverse incentives (e.g., technology adoption) without necessarily affecting patient outcomes. 41Conversely, with adequate incentives, patient choice and hospital competition could support improved outcomes of care. 42[48] Three co-authors, including those with lived experience of breast cancer, have provided recommendations based on their own experiences and the findings from this study (Table 4).These recommendations demonstrate that initiatives that aim to improve information about where patients can have their treatment should not only focus on making this information more readily available but they should also ensure that this information is easy to understand and presented in a format that can support the trade-offs that patients have to make.
Our fourth major finding is that patients undergoing BCS were less likely to travel to hospitals with shorter waiting times because patients have a preference for other aspects of health care quality, which means the patient mobility that we observed in the study may lead to a lengthening rather than a shortening of waiting times.These findings go against policy initiatives that enable patients to select health care providers with shorter waiting times to manage treatment backlogs after the coronavirus disease 2019 pandemic. 49,50r modelling of patient mobility does highlight several conceptual and methodological challenges.In this report, we have studied where patients receive their treatment in relation to where they live.Decisions are made by patients together with primary or secondary care physicians (at the diagnosing hospital) and are influenced by pre-existing referral patterns.However, distinguishing between the preferences of the patient and their primary and secondary care physicians is beyond the scope of this analysis and requires further qualitative investigation. 45,51,52The data presented are from 2016-2018, which was the time frame available within our data set.Although activity levels have now returned to prepandemic levels, we do acknowledge that patterns of referral may have changed.
The study used centroids of small geographical areas, typically representing 650 households, to represent the location of patients' residence, and this could have masked variations in travel times, thus attenuating rather than enhancing the observed associations between travel time and patient mobility. 53We also acknowledge A geographic information system (ESRI ArcGIS) was used to determine average daytime travel times by car by inputting the populationweighted centroids of the patients' LSOAs and full postal codes of the 166 hospitals performing breast cancer surgery by using the Ordnance Survey Master Map Highways Network.Travel time was included in the model as the additional travel time patients had to travel beyond their nearest hospital to reach an alternative hospital providing breast cancer surgery.
Univariable and multivariable conditional logistic regression models were estimated to assess the impact of travel time and hospital characteristics with and without adjustment of confounders.Patient characteristics were then included in the adjusted model as interactions with travel time to investigate the extent to which the associations with travel time were modified by the six patient variables: age (patients older than 70 years vs. patients aged 70 years or younger), ethnicity (patient from ethnic minority groups vs. White ethnic groups), comorbidity (patients with one or more comorbidity as defined by the Charlson comorbidity index vs.patients with no comorbidity), socioeconomic status (patients from more deprived AGGARWAL ET AL. observed that older patients (p < .01 for BCS and mastectomy alone), ethnic minority patients (p < .01 for BCS), patients with comorbidity (p = .01for mastectomy alone), and patients with more advanced disease (stage II vs. I; p < .01 for BCS and mastectomy with or without IBR; p = .01)were less likely to travel to a hospital other than their nearest for treatment.Conversely, rural patients were more likely to travel to a hospital other than their nearest for treatment (p < .001for BCS and mastectomy with and without reconstruction).For example, patients aged 70 years or older were less likely to travel to a hospital up to an additional 30 minutes away from nearest hospital for either BCS (adjusted OR, 0.76; 95% CI, 0.69-0.84)or mastectomy alone (adjusted OR, 0.64; 95% CI, 0.56-0.73).Conversely, additional travel time was less strongly associated with the odds of traveling to a particular hospital for patients who lived in rural areas (OR for interaction term always > 1.00) compared with patients living in urban areas, which demonstrates that patients living in rural areas had a greater willingness to travel.DISCUSSIONThis national, population-based study demonstrated that up to one in three patients who have breast cancer are bypassing their nearest hospital offering cancer surgery.Travel time is the most important determinant of where patients receive their breast cancer treatment.

F I G U R E 1
Map of the Northwest region of England (UK) illustrating the mobility patterns of patients who received breast conserving surgery (BCS) at a selected NHS hospital (indicated with a star symbol).The crosses represent other hospitals providing BCS in the region, and the colored dots represent individual patients who underwent BCS.Patients treated at the hospital (star symbol) who traveled from outside the local area (arrivers) are represented as blue dots.Patients from the hospital's local area who received treatment there are represented as green dots, and patients from the hospital's local area who traveled to other hospitals for surgery are represented as red dots (leavers).The map includes a scaled magnification of the region (inset) and a national overview.Contains National Statistics and National Records of Scotland data (source: Northern Ireland Statistics and Research Agency) as well as Ordnance Survey data.©Crown copyright and database right 2022.
Characteristics of patients diagnosed with breast cancer between 2016 and 2018 who underwent breast-conserving surgery or mastectomy in English National Health Service hospitals.
Includes patients recorded as mixed White and Asian, mixed White and Black African, mixed White and Black Caribbean, and any other mixed background according to linked National Health Service Hospital Episode Statistics data.Chinese, other, and any other ethnic group according to linked National Health Service Hospital Episode Statistics data.
a b Includes patients recorded as Adjusted impact of travel time and hospital characteristics for 69,153 patients who underwent breast-conserving surgery and 32,591 patients who underwent mastectomy in English National Health Service hospitals.
T A B L E 2