Melanie Morris

Ethnicity, deprivation and screening: survival from breast cancer among screening-eligible women in the West Midlands diagnosed from 1989 to 2011

Background:Social inequalities in breast cancer survival are smaller when the cancer is screen-detected. We examined survival from screen-detected and non screen-detected breast cancer by ethnicity and deprivation.Methods:Cancer registry data for 20 283 women aged 50–70 years, diagnosed between 1989–2011 and invited for screening, were linked with screening and ethnicity data. We examined Asian, Black and White groups, less deprived and middle/more deprived women. Net survival was estimated using ethnic- and deprivation-specific life tables. Estimates were corrected for lead-time bias and over-diagnosis.Results:Net survival varied by screening history. No significant differences in survival were found by ethnicity. Five-year net survival was 90.0% (95% CI, 89.3–90.8%) in less deprived groups and 86.7% (85.9–87.4%) among middle/more deprived women. Screening benefitted all ethnic and both deprivation groups. Whether screen-detected or not, more deprived women had significantly poorer outcomes: 5-year net survival was 78.0% (76.7–79.2%) for deprived women who were not screen-detected compared with 94.0% (93.1–95.1%) for less deprived women who were screen-detected.Conclusions:The three ethnic groups differed little in their breast cancer survival. Although screening confers a survival benefit to all, there are still wide disparities in survival by deprivation. More needs to be done to determine what underlies these differences and tackle them.

An investigation of social inequalities in help-seeking and use of health services for fertility problems in a population-based sample of UK women.

Although infertility is an important public health problem, treatment can be expensive and resources are increasingly scarce. This study investigates possible inequalities in the use of medical services for fertility problems. We analysed data from a population-based survey for associations between socio-economic characteristics and help-seeking or use of services, to establish whether inequalities existed. More women of higher social status and education reported fertility problems, but there was no clear trend in help-seeking, investigations or treatments for infertility by social status and education level. New work is planned to investigate these issues more fully, particularly the role of family income.

A novel ecological methodology for constructing ethnic-majority life tables in the absence of individual ethnicity information

Background Deprivation-specific life tables have been in use for some time, but health outcomes are also known to vary by ethnicity over and above deprivation. The mortality experiences of ethnic groups are little studied in the UK, however, because ethnicity is not captured on death certificates. Methods Population data for all Output Areas (OAs) in England and Wales were stratified by age-group, sex and ethnic proportion, and matched to the deaths counts in that OA from 2000 to 2002. We modelled the relationship between mortality, age, deprivation and ethnic proportion. We predicted mortality rates for an area that contained the maximum proportion of each ethnic group reported in any area in England and Wales, using a generalised linear model with a Poisson distribution adjusted for deprivation. Results After adjustment, Asian and White life expectancies between 1 and 80 years were very similar. Black men and women had lower life expectancies: men by 4 years and women by around 1.5 years. The Asian population had the lowest mortality of all groups over age 45 in women and over 50 in men, whereas the Black population had the highest rates throughout, except in girls under 15. Conclusions We adopted a novel ecological method of constructing ethnic-majority life tables, adjusted for deprivation. There is still diversity within these three broad ethnic groups, but our data show important residual differences in mortality for Black men and women. These ethnic life tables can be used to inform public health planning and correctly account for background mortality in ethnic subgroups of the population.

Do cancer survival statistics for every hospital make sense

1192 www.thelancet.com/oncology Vol 17 September 2016 In recent years, there has been increasing pressure from politicians for researchers to produce cancer survival estimates for every hospital and health-care provider in England, to inform national cancer strategy and patient choice. Here we set out why cancer survival statistics for each hospital would not be useful for either of those purposes. Population-based cancer survival fi gures, derived from cancer registry data for all patients with cancer living in a defi ned region, have been used to highlight geographic, socioeconomic, and international inequalities in survival for many years. Such survival fi gures have underpinned every national cancer strategy in the UK since 1995, and trends in survival have been used to evaluate the eff ect of those strategies. Unlike hospital-specifi c survival fi gures, populationbased fi gures include all patients with cancer. They are unbiased, and they take precise account of the risk of death from causes other than cancer, allowing survival from the cancer to be compared between populations. Therefore, population-based fi gures are ideal for studying trends in survival, and for making regional and international comparisons. Persistent inequalities in survival have led politicians to demand survival estimates for ever-smaller geographies, such as for the 209 Clinical Commissioning Groups in England (population 65 000–880 000 per group). However, producing survival estimates for every Clinical Commissioning Group that are suffi ciently robust to be interpreted for managerial purposes, year on year, is extremely diffi cult, even for the most common cancers. This approach can be taken for 1-year but not 5-year survival (as there are more deaths in the fi rst year after diagnosis), and even that requires modelling the entire national dataset of several million patients with cancer diagnosed over more than 10 years. Failure to understand the diff erence between the robustness of population-based survival fi gures, and the unreliability of hospital-specifi c survival fi gures has led the Department of Health and NHS England to demand routine production of hospital survival fi gures to judge the performance of every health service provider (ie, a single hospital, or a group of hospitals that provide cancer services). It has even been suggested that hospital-specifi c cancer survival statistics should be produced every year, for each type of cancer, and for each stage at diagnosis, as a way of informing “patient choice”. A proposal to that eff ect was removed from the 2015 cancer strategy only at a very late stage of drafting. We argue that provider-level survival statistics cannot be usefully interpreted for surveillance of hospital performance: they are prone to referral bias, and to random fl uctuation because of small numbers; they can lead to spurious comparisons and inappropriate management decisions; and they do not support patient choice. The fact that more than 1 million people in England have opted out from the use of their health data other than for their own clinical management will introduce bias and further undermine the utility of such statistics. Annual snapshots of outcome data cannot be safely interpreted in isolation. For example, 90-day mortality rates are published each year for individual colorectal surgeons, but the warnings for patients are striking: “the results this year will not tell us the real situation and will not be completely accurate for some surgeons”; “very few [surgeons] do a suffi cient number [of procedures], even over the four-year period, for these data to allow reliable comparison between surgeons”; “[it is] hard to identify one surgeon as being in charge in one operation” [because surgeons work in teams]”. Readmission rates, and 30-day or 90-day mortality rates, are short-term outcome measures that might relate to a single surgeon or hospital, but fair comparisons require adjustment for patient case-mix, and suffi cient numbers to produce statistically robust results. One consequence could be misplaced confi dence when evidence of poor performance is not strong, which can be misinterpreted as refl ecting adequate performance. By contrast, population-based outcome measures, such as 1-year, 5-year, or 10-year survival, refl ect the overall eff ectiveness of the health service in managing all patients with cancer in the longer term. Variation in such metrics cannot generally be ascribed to a single surgeon or hospital, or a single component of the quality of care. In addition, it is not straightforward to assign a patient with cancer to a particular hospital for survival For information on NHS clinical care choices see http://digital. nhs.uk/catalogue/PUB20527 Do cancer survival statistics for every hospital make sense?

No ‘cure’ within 12 years of diagnosis among breast cancer patients who are diagnosed via mammographic screening: women diagnosed in the West Midlands region of England 1989–2011

Despite dramatic improvements in survival over past decades, diagnosis with breast cancer leads to a small but persistent, long-term increased risk of death for all groups of women. This is also true for those whose cancer is detected asymptomatically via screening mammography.

What might explain deprivation-specific differences in the excess hazard of breast cancer death amongst screen-detected women? Analysis of patients diagnosed in the West Midlands region of England from 1989 to 2011

Background Breast cancer survival is higher in less deprived women, even amongst women whose tumor was screen-detected, but reasons behind this have not been comprehensively investigated. Methods The excess hazard of breast cancer death in 20,265 women diagnosed with breast cancer, followed up to 2012, was estimated for screen-detected and non-screen-detected women, comparing more deprived to less deprived women using flexible parametric models. Models were adjusted for individual and tumor factors, treatment received and comorbidity. For screen-detected women, estimates were also corrected for lead-time and overdiagnosis. Results The excess hazard ratio (EHR) of breast cancer death in the most deprived group, adjusted only for age and year of diagnosis, was twice that of the least deprived among screen-detected women (EHR=2.12, 95%CI 1.48-2.76) and 64% higher among non-screen-detected women (EHR=1.64, 95%CI 1.41-1.87). Adjustment for stage at diagnosis lowered these estimates by 25%. Further adjustment had little extra impact. In the final models, the excess hazard for the most deprived women was 54% higher (EHR=1.54, 95%CI 1.10-1.98) among screen-detected women and 39% higher (EHR=1.39, 95%CI 1.20-1.59) among non-screen-detected women. Conclusion A persistent socio-economic gradient in breast cancer-related death exists in this cohort, even for screen-detected women. The impact of differential lifestyles, management and treatment warrant further investigation.

Can we assess Cancer Waiting Time targets with cancer survival? A population-based study of individually linked data from the National Cancer Waiting Times monitoring dataset in England, 2009-2013

Background Cancer Waiting Time targets have been integrated into successive cancer strategies as indicators of cancer care quality in England. These targets are reported in national statistics for all cancers combined, but there is mixed evidence of their benefits and it is unclear if meeting Cancer Waiting Time targets, as currently defined and published, is associated with improved survival for individual patients, and thus if survival is a good metric for judging the utility of the targets. Methods and findings We used individually-linked data from the National Cancer Waiting Times Monitoring Dataset (CWT), the cancer registry and other routinely collected datasets. The study population consisted of all adult patients diagnosed in England (2009–2013) with colorectal (164,890), lung (171,208) or ovarian (24,545) cancer, of whom 82%, 76%, and 77%, respectively, had a CWT matching record. The main outcome was one-year net survival for all matched patients by target attainment (‘met/not met’). The time to each type of treatment for the 31-day and 62-day targets was estimated using multivariable analyses, adjusting for age, sex, tumour stage and deprivation. The two-week wait (TWW) from GP referral to specialist consultation and 31-day target from decision to treat to start of treatment were met for more than 95% of patients, but the 62-day target from GP referral to start of treatment was missed more often. There was little evidence of an association between meeting the TWW target and one-year net survival, but for the 31-day and 62-day targets, survival was worse for those for whom the targets were met (e.g. colorectal cancer: survival 89.1% (95%CI 88.9–89.4) for patients with 31-day target met, 96.9% (95%CI 96.1–91.7) for patients for whom it was not met). Time-to-treatment analyses showed that treatments recorded as palliative were given earlier in time, than treatments with potentially curative intent. There are possible limitations in the accuracy of the categorisation of treatment variables which do not allow for fully distinguishing, for example, between curative and palliative intent; and it is difficult in these data to assess the appropriateness of treatment by stage. These limitations in the nature of the data do not affect the survival estimates found, but do mean that it is not possible to separate those patients for whom the times between referral, decision to treat and start of treatment could actually have an impact on the clinical outcomes. This means that the use of these survival measures to evaluate the targets would be misleading. Conclusions Based on these individually-linked data, and for the cancers we looked at, we did not find that Cancer Waiting Time targets being met translates into improved one-year survival. Patients may benefit psychologically from limited waits which encourage timely treatment, but one-year survival is not a useful measure for evaluating Trust performance with regards to Cancer Waiting Time targets, which are not currently stratified by stage or treatment type. As such, the current composition of the data means target compliance needs further evaluation before being used for the assessment of clinical outcomes.

Which patients are not included in the English Cancer Waiting Times monitoring dataset, 2009–2013? Implications for use of the data in research

Background:Cancer waiting time targets are routinely monitored in England, but the Cancer Waiting Times monitoring dataset (CWT) does not include all eligible patients, introducing scope for bias.Methods:Data from adults diagnosed in England (2009–2013) with colorectal, lung, or ovarian cancer were linked from CWT to cancer registry, mortality, and Hospital Episode Statistics data. We present demographic characteristics and net survival for patients who were and were not included in CWT.Results:A CWT record was found for 82% of colorectal, 76% of lung, and 77% of ovarian cancer patients. Patients not recorded in CWT were more likely to be in the youngest or oldest age groups, have more comorbidities, have been diagnosed through emergency presentation, have late or missing stage, and have much poorer survival.Conclusions:Researchers and policy-makers should be aware of the limitations in the completeness and representativeness of CWT, and draw conclusions with appropriate caution.

Characteristics of patients with missing information on stage: a population-based study of patients diagnosed with colon, lung or breast cancer in England in 2013.

BackgroundStage is a key predictor of cancer survival. Complete cancer staging is vital for understanding outcomes at population level and monitoring the efficacy of early diagnosis initiatives. Cancer registries usually collect details of the disease extent but staging information may be missing because a stage was never assigned to a patient or because it was not included in cancer registration records. Missing stage information introduce methodological difficulties for analysis and interpretation of results. We describe the associations between missing stage and socio-demographic and clinical characteristics of patients diagnosed with colon, lung or breast cancer in England in 2013. We assess how these associations change when completeness is high, and administrative issues are assumed to be minimal. We estimate the amount of avoidable missing stage data if high levels of completeness reached by some Clinical Commissioning Groups (CCGs), were achieved nationally.MethodsIndividual cancer records were retrieved from the National Cancer Registration and linked to the Routes to Diagnosis and Hospital Episode Statistics datasets to obtain additional clinical information. We used multivariable beta binomial regression models to estimate the strength of the association between socio-demographic and clinical characteristics of patients and missing stage and to derive the amount of avoidable missing stage.ResultsMultivariable modelling showed that old age was associated with missing stage irrespective of the cancer site and independent of comorbidity score, short-term mortality and patient characteristics. This remained true for patients in the CCGs with high completeness. Applying the results from these CCGs to the whole cohort showed that approximately 70% of missing stage information was potentially avoidable.ConclusionsMissing stage was more frequent in older patients, including those residing in CCGs with high completeness. This disadvantage for older patients was not explained fully by the presence of comorbidity. A substantial gain in completeness could have been achieved if administrative practices were improved to the level of the highest performing areas. Reasons for missing stage information should be carefully assessed before any study, and potential distortions introduced by how missing stage is handled should be considered in order to draw the most correct inference from available statistics.