NHS Digital Data Release Register - reformatted
University of Warwick
Opt outs honoured: Y
Basis: Section 251 approval is in place for the flow of identifiable data
Format: Identifiable Non Sensitive
How often: Ongoing
When: unknown — 11/2016
HSCIC Id: DARS-NIC-351810-N3G6N-v0.3
Data: MRIS - List Cleaning Report
Data: Hospital Episode Statistics Admitted Patient Care
Data: Hospital Episode Statistics Outpatients
Data: Office for National Statistics Mortality Data (linkable to HES)
Output: Specific outputs expected, including target date:
The main product of this study will be the analysis carried out to answer the following important main questions:
1. What explains the observed variations in suspected bowel cancer referral rates? Is it possible to identify any particular groups of patients and particular areas that need to be targeted for efficient and effective referrals?
2. How have these observed variations changed over time and how this has impacted on mortality?
3. What has been the effect of the bowel screening program on the diagnosis of bowel cancer and the subsequent mortality?
As part of the analysis there will be various tables with the results from the estimation methodology employed to conduct the analysis. The results will be grouped to the lowest Primary Care Trust (PCT) level, such that no patient nor doctor information will be identifiable. In cases where the number of practices/GPs in a given PCT is small, PCT sharing boundaries will be grouped together so that anonymity of the doctors and patients will be preserved.
Expected outputs & dissemination activities:
The journal articles will be authored by the two users of the data.
The applicant will aim to publish the papers in general interest journal such as the British Medical Journal and specialist journal such as the Clinical Colorectal Cancer journal. These are peer-reviewed prestigious journals that reach a wide audience including health care professionals and individuals involved in policy making in the NHS. The statistical analysis to be carried out will be easily generalisable to other cancer incidences and hence the importance of also targeting of the general interest journals. The applicant will also ensure that the papers are made widely accessible through open-access policies of these journals to ensure that the results reach a wide audience including NHS staff who work with bowel cancer patients, policy makers and General Practitioners. Cancer Research UK will also be made aware of the research findings.
The findings from the project will also be presented at relevant national and international conferences to disseminate the results. Some of the conferences that are of particular relevance to this study include the Annual conference and exhibition by the National Association of Primary Care (NAPC) and the Annual Education, Research and Innovation Symposium by the Royal College of General Practitioners (RCGP). Both NAPC and RCGP aim to improve the quality of the primary care services as well as to bridge the gap between the primary and the secondary care. RCGP Symposium is considered as a ‘must attend’ event by General Practitioners, GP registrars and academics who want to learn more and potentially contribute in primary care service improvement. Additionally, the findings will also disseminated at The King's Fund events and the Cancer Research UK Cambridge Institute (CRUK CI) Seminar events. The participants at these conferences will in turn help to disseminate the results to wider audiences.
Non-technical executive summaries of the papers will be circulated to all the relevant organisations including those mentioned above. Warwick University has a very good media department which will ensure that all relevant results are publicised in an appropriate manner to reach the intended groups to achieve maximum benefits. If papers are accepted by the aforementioned journals, this it self will ensure that the results from this study reaches the right health professionals. Press-releases at the same time as publications will also ensure that the results get the maximum publicity.
The tentative target dates for conducting the study are as follows:
Months 0-3: The initial data cleaning is expected to take about 3 months.
Months 4-12: Cross-sectional analysis of variations in observed patterns of referral rates and writing of the paper. A paper for a general interest medical journal such as the British Medical Journal, or a specialist interest journal such as the Clinical Colorectal Cancer journal.
Months 9-36: Longitudinal analysis of variations, modelling using survival analysis, writing of the paper. A paper for a general interest medical journal such as the British Medical Journal, or a specialist interest journal such as the Clinical Colorectal Cancer journal.
All outputs will have small numbers suppressed in line with the HES analysis guide.
Activities: Data creation for analyses
i) The HES data extract will provide the Lower super Output Area (LSOA) identifiers which will be used to merge the local area population characteristics from the Census data with the HES information. This is crucial for the statistical analysis as this will enable the analysis to account for confounders and eliminate the bias in the estimation of the statistical models.
ii) “The Quality Outcome Framework (QOF) is a voluntary annual reward and incentive programme for all GP surgeries in England, detailing practice achievement results. It is not about performance management but resourcing and then rewarding good practice.” See http://www.hscic.gov.uk/qof. QOF data provides some information on GP practice characteristics such as, the overall demographics of the pool of registered patients like the size of registered patients, their age group, etc. It also provides some information about the prevalence of different type of diseases like Cancer, Diabetes, Depression, Heart Disease, etc. But none of the information in QOF data are sensitive and these are freely available to retrieve from the HSCIC’s website.
Matching of the QOF data to HES will enable the analysis to account for confounders such as the health of the patients attached to the GP practice. This will enable the applicant to model the probability of the referral leading to a confirmed diagnosis of bowel cancer. In summary, the HES data will be merged with the local patient characteristics that can be extracted from the QOF data and also the Census data at the LSOA level. This is crucial for valid statistical analysis.
Statistical analyses
This has three parts:
(1) Starting with a linear model and cross-sectional data, first analysis with look at the effect of observed patients and GP Practice characteristics on suspected bowel cancer referral rates. All referrals will be grouped into three categories; immediate, urgent and non-urgent, and then standardized and adjusted for demographic and population characteristics obtained from the LOSAs. The results from this estimation will be used to test the hypothesis of whether the observed patients and GP Practice characteristics have a significant effect on explaining the variations in referral rates for suspected bowel cancer.
(2) The second stage of the analysis will use the entire longitudinal dataset from 2003 to 2013 and panel data analysis to model the changes in behaviour through time. In addition, the analysis will be used to estimate the effect of the bowel screening program (introduced in 2006) on the observed patterns of referrals.
(3) The final part of the project will use survival analysis to model the probability of death from bowel cancer conditioning on the type of referral and the characteristics of the patient and the GP practice.
None of the processed data will be shared with any third party nor will they be used for commercial purposes or for marketing purposes. All individuals with access to the data are employees of the University of Warwick.
Objective: Background:
There has been increasing concerns in recent years about differences in patient referral rates for suspected cancer across GPs. Each year there are nearly one million urgent GP referrals for suspected cancer (National Cancer Intelligence Network, 2014). In 2011, the rate of urgent referrals for suspected cancer in England ranged from 919.8 to 2957.4 per 100,000 populations. This 3.2 fold difference is evidence of a wide variation in cancer referral rates (Cancer Research UK, 2012). Socioeconomic factors and doctors characteristic have been shown to play a part in explaining variations in overall referral rates (O'Donnell, 2000). Overall GP referral rates for medical and surgical outpatient referrals are shown to be higher in high deprived areas, (Hippisley-Cox, Hardy, Pringle, Fielding, & Carlisle, 1997). This has not been tested for suspected bowel cancer referral rates. The characteristics that are expected to be associated with the referral rates are observed patients, local area characteristics, GP Practice characteristics and the incidence of bowel cancer in the particular GP area covered by the GP practice.
Objective for processing:
The main objective in processing the data is to carry out statistical analyses to explain observable variations in GP-practice referral rates for suspected bowel cancer and how this leads to positive diagnosis of cancer using patient, GP practice level characteristics and the local population characteristics. The research will try to explain how these observed variations has changed over time and how this has impacted on mortality. Another objective is to look at the effect of the Bowel Screening Program on the diagnosis of bowel cancer and the subsequent mortality.
A comprehensive analysis using mortality data will enable the modelling of survival durations after referrals.
The main product of this study will be the analysis carried out to answer three main questions:
1. What explains the observed variations in suspected bowel cancer referral rates? Is it possible to identify any particular groups of patients and particular areas that need to be targeted for efficient and effective referrals?
2. How have these observed variations changed over time and how this has impacted on mortality?
3. What has been the effect of the bowel screening program on the diagnosis of bowel cancer and the subsequent mortality?
The research papers written as part of this project will be included as evidence as part of a PhD thesis if they fall within the timeframe for submission.
Benefits: This study aims to statistically model (i) variations in bowel cancer referral rates across GPs, and (ii) the survival rates, controlling for confounders.
There has been an increasing concern about the differences in the referral rates for patients suspected of cancer (http://www.rightcare.nhs.uk/downloads/Right_Care_Diagnostics_Atlas_hi-res.pdf.). Variations in referral rates are considered as a source of inefficiency in the primary care services. This project will aim to identify various sources (for example specific socio-economic factors) that contribute to observed variations in referral rates. This will help policies to be targeted to ensure that the relevant population receives the appropriate care and hence reduce inefficiencies in the delivery of care. Given the very wide dissemination strategy, this will help policies to be targeted at the right population to ensure that they receive timely and appropriate care, and hence increase their health and wellbeing.
Any beneficial change as a result of the outputs will be dependent on when the outcome of our research will be published. It will be ensured that all relevant bodies/organisations such as the NHS England, Public Health England, the Department of Health, Cancer Research UK, GPs and patients, involved in decision making with regard to bowel cancer diagnosis and treatment are informed about the research. For example, in 2010, Right Care working closely with NHS England, the Department of Health and the Public Health England, published for the first time the NHS Atlas of Variation in Health Care with the main aim to increase awareness regarding the variation existing in some clinical domains across different regions in England (http://www.rightcare.nhs.uk/atlas/qipp_nhsAtlas-LOW_261110c.pdf). The variation in cancer referral rates was considered very concerning and as they pointed out “Awareness is the first important step in identifying and addressing unwarranted variation” (Right Care 2010). By increasing awareness, they aimed to trigger the research for unwarranted variation and assess the value of the healthcare provided both to populations and to individuals. This research will contribute towards understanding part of the variation in cancer referral rates.
The methodology used will be very easily generalizable to incidences of other types of cancers too. The outputs will therefore be beneficial to other researchers who wish to look at similar issues in other diseases which show very wide variations in detection and treatment across different geographical areas.
Source: NHS Digital.