Targeted Lung Health Check (TLHC) evaluation

The Targeted Lung Health Checks (TLHC) pilot programme is a flagship programme of work in England which contributes to the ambition of the NHS Long Term Plan to improve early diagnosis and survival for those diagnosed with cancer. The TLHC programme targets those most at risk of lung cancer and works with Integrated Care Systems who have some of the highest rates of mortality from lung cancer.

The TLHC pilot programme works with seventeen projects covering fourteen Cancer Alliances to deliver the programme to approximately 600,000 eligible participants. People most at risk of lung cancer are identified based on age (i.e. 55-76 years of age), smoking status and other lifestyle factors. They are invited for a lung health check and where required, a Low Dose CT Scan to check their lungs for cancer.

NHS England (NHSE) is conducting a national evaluation to understand the impacts and economics of the TLHC programme. The evaluation is expected to demonstrate impacts on patient health outcomes, patient experience and wider health inequalities. Findings from the evaluation will ensure an evidence-based approach to NHSE's longer-term strategy for wider roll out of a targeted screening programme of this kind.

NHS England currently holds a separate data sharing agreement covering access to the NHS England - Cancer TRE environment for a separate purpose - DARS-NIC-411785-Z6X7M-v1.2 (Rapid Diagnostic Centre – Cancer TRE). NHS England requests use of select data products within this environment for the purpose described within this agreement under the TLHC programme. Data processed for the purpose outlined within this agreement will only be from those datasets named within this agreement.

NHS England has specified specific data periods of interest to them within the purpose section of this request, however as the method of access to this data remains via the NHS England - Cancer TRE, this agreement covers access to all available fields and periods within the TRE for the datasets listed within this request. Current functionality of the TRE environment limits the ability to set data minimisation controls via system controls and data provisioning. Therefore only contractual controls are applied to the Data Controller's / Processors based on the data and processing described in this request. This is also supported by the functionality to monitor the processing of data performed within the TRE environment.

Ipsos, together with the Strategy Unit at the Midlands and Lancashire Commissioning Support Unit (CSU), have been appointed jointly as the national evaluation partner for the TLHC pilot programme. NHSE have commissioned Ipsos to undertake the evaluation work. Ipsos has, in turn, sub-contracted the Strategy Unit hosted by Midlands and Lancashire Commissioning Support Unit (CSU) to undertake separate elements of the programme. NHS Midlands and Lancashire CSU will not be undertaking the processing outlined within this request and will not have access to any of the data from NHS Digital.

Data Processors included within this agreement are able to access the following datasets only for the purposes set out within this request:

- National Cancer Registration Data Set

- Civil Registration - Deaths

- Cancer Waiting Times (CWT) Data Set

The evaluation will seek to understand:

1. Impact on participant outcomes (including cancers detected; stage shift in diagnosis; other diagnoses)

• How did the targeted lung health check programme deliver across ten discrete projects in terms of detection of lung cancer at an earlier stage than it may otherwise have been?

• From the participants offered lung health checks, in how many was lung cancer found (conversion rate)?

• What other health conditions were detected during a lung health check and how often did these occur?

• What was the entry and completion rate of smoking cessation courses and what were the outcomes?

2. Health inequalities

• How were participant cohorts determined?

• Did the programme attract those most at risk of developing lung cancer, including the most deprived, vulnerable populations, and minority groups?

• How were these groups targeted, how effective were different engagement strategies, and how did take up rates vary by different engagement approaches?

• What impact has the programme had on reducing the variation in lung cancer outcomes across each of the ten projects?

The full evaluation of this pilot involves an analysis of data collected by the programme, information recorded in interviews with patients and staff about their experiences, and data collected through national data sets.

Through the TRE, Ipsos will be able to access patient level data from national datasets to compare outcomes in areas where the TLHC pilot programme is being implemented (intervention) against areas without this (control).

Selection Bias

For optimal analysis, the pilot sites should be compared to CCGs that are equivalent to the pilot sites in all influential respects apart from their participation in the programme. For this pilot, CCGs were selected on a non-random basis (those will highest lung cancer mortality rates). As a consequence, the selection of a comparison group should also consider the potential for systematic differences between the pilot CCGs and comparator areas to bias findings. A baseline analysis is required to identify suitable comparator areas for this evaluation.

Unit of analysis

Whilst it is possible to access patient level data through the TRE, there are a couple of reasons why patient level analysis is not appropriate for this evaluation, including construction of longitudinal measures at an individual level and access to a comparator population. These difficulties will be eased by focusing on outcomes at the population level by aggregating patient level outcomes over units of administrative geography (or other units of aggregation) and normalising these results to the size of the local population. As a result, the majority of processing conducted with the data will be in the form of aggregated data with small numbers unsuppressed. Aggregation of the data will be performed by the Ipsos MORI, therefore this agreement includes access to patient level, pseudonymised data via the TRE initially.

A strategy document outlining the analytical questions to be answered for each element of the programme evaluation is maintained by NHSE. NHS England is the sole data controller for this programme and all work undertaken as part of this programme. Ipsos and Midlands and Lancashire CSU are all data processors acting under the instruction of NHSE. Ipsos is unable to make decisions about how or why the data will be processed. Any such decisions, including decisions about the questions to be answered for the purpose of the evaluation and decisions about which data processors will undertake which work packages require approval from NHSE.

The GDPR legal basis for NHSE to process personal data is GDPR Article 6(1)(e) ‘task in the public interest’ and for processing special categories of personal data NHSE rely on GDPR Article 9(2)(h) ‘processing is necessary for the purposes of preventative or occupational medicine, for the assessment of the working capacity of the employee, medical diagnosis, the provision of health or social care or treatment or the management of health or social care…’.

Using data from the TRE, this analysis will take place in two stages:

Stage 1:

Baseline analysis (covering years 2016 to 2019 prior to the start of the TLHC programme) of the following patient level outcomes to identify suitable comparator areas for the subsequent impact evaluation:

• Cancer diagnosis (incidence),

• Cancer stage at diagnosis,

• Mortality,

• One-year survival rates.

Ipsos will use these outcomes aggregated on the Middle Layer Super Output Area (MSOA)-level alongside other characteristics aggregated on the MSOA-level (e.g. age-profile, deprivation quintiles, etc.) to apply propensity score matching (PSM). Using PSM a suitable set of comparator areas (i.e. counterfactuals) that ‘match’ TLHC treatment areas can be identified (i.e. MSOAs with similar outcomes and characteristics that are not delivering the TLHC intervention but are otherwise similar to TLHC areas).

Further baseline analysis will assess the plausibility of parallel trends in the aggregated patient outcomes on the MSOA-level in the pre-intervention period (both before matching and after matching), required for the planned difference-in-difference (DID) estimation in the impact evaluation of the TLHC.

Stage 2:

The impact evaluation will focus on establishing the causal impact of the TLHC on the above-mentioned patient level outcomes (i.e. cancer diagnosis, cancer stage at diagnosis, mortality, and one-year survival rates) using:

• The matched TLHC treated and untreated MSOAs in a DID analysis if parallel trends in pre-intervention outcomes are observed (see baseline analysis above) including covariates and time and unit fixed effects

• If parallel trends in pre-intervention outcomes are not observed, only matched analysis will be used including covariates and time and unit fixed effects.

• Regression discontinuity design (RDD) boundary analysis exploiting spatial proximity and thus likely similarity of treated and untreated MSOAs at boundaries of treated and untreated CCGs.

• DID and matching will provide the average treatment effect of the TLHC on the specified outcomes whereas RDD will provide the local average treatment effect of the TLHC on outcomes (i.e. treatment effect for MSOAs in proximity to boundaries).

To identify the health inequality impacts of the TLHC pilot programme, further patient level information is required (from the TRE), which will be used in corresponding sub-group analyses using the above discussed methodologies:

• Sex – by male/female

• Age – by age-groups 55-64 and 65-76

• Ethnicity – by ethnic groups numbers permitting

As mentioned above, to identify a suitable comparison group, Ipsos will use baseline data from the Cancer Registry to ‘match’ comparator areas to treatment areas. Although TLHC sites were selected at CCG level, the impact evaluation requires a smaller level of aggregation (i.e. MSOA-level) for two reasons:

1. To maximise the number of units in the sample to provide sufficient statistical power for analysis.

2. To scope the feasibility of using a boundary regression discontinuity design, which compares areas just within the CCG boundary to those just outside the boundary, requiring smaller area data, i.e. MSOA-level aggregate patient data. Specific steps for the aggregation are described in the following section “5b. Processing Activities”.

The following linked national datasets will be required (through the TRE) for the purposes of this evaluating the TLHC pilot programme:

i) National Cancer Registration Dataset

This data source contains details of individuals’ diagnoses including details of cancer staging, first treatments, how advanced cancer was when diagnosed, etc.

ii) Civil Registration Mortality

This data source will be used to identify one-year survival and mortality impacts of the TLHC using cause of death information provided in the dataset.

iv) Cancer Waiting Times (CWT)

This data source will be used for the baseline assessments of existing service delivery differences between TLHC and control areas, and will further be used as a matching variable to identify the counterfactuals for TLHC areas for the impact evaluation.

The specific data items required from the datasets listed above include the following:

• TUMOUR_SITE - Site of neoplasm (topography)

• BIRTHDATEBEST - Patient's date of birth

• SEX - Patient's Sex

• CCG_CODE / CCG_NAME - Clinical Commissioning Group code and name (at diagnosis)

• MSOA11_CODE - 011 Middle Super Output Area (at diagnosis)

• QUINTILE - Measure of deprivation at small area level made up from the income domain, quintiles are calculated from populations

• DIAGNOSISDATE1 - Indicative date of diagnosis of the cancer

• STAGE_BEST - Best ‘registry’ stage at diagnosis of the tumour

• TREATMENT_START_DATE - Start date of cancer treatment given to a patient who is receiving care for a cancer condition

• DEATHDATEBEST - Date of Death

• DAYS_TOFROM_INDEXDATE_DIAGNOSIS (Derived per applicant's requirements) - Option to an interval between two dates (e.g. Days from date of diagnosis to death)

• ETHNICITY - Ethnicity code

• ETHNICTY_GROUP_BROAD - Option to group ethnicities (e.g. white/ non-white/ unknown)

Individually authorised analysts, substantively employed by Ipsos, will be granted remote secure access to the Cancer Trusted Research Environment (TRE) within NHS Digital’s data platform, the Data Processing Service (DPS).

This application requests the processing of National Cancer Registration Data Set, Civil Registration - Deaths & Cancer Waiting Times (CWT) Data Set available within the NHS England Cancer TRE associated with the Data Sharing Agreement - DARS-NIC-411785-Z6X7M-v1.2 (Rapid Diagnostic Centre – Cancer TRE) for the purpose outlined in this agreement. Processors will utilise all available data for the products indicated above within the TRE. Where available, historic data will be accessible from the 2014/15 year. Data processed for the purpose outlined within this agreement will only be from those datasets named within this agreement.

Within the NHS England - Cancer TRE, the analysts will be able to access pseudonymised linked data from the datasets as outlined in the objective for processing section. No details which directly identify data subjects, such as names, NHS Numbers, etc., will be accessible within the TRE.

Ipsos will access patient level data through the TRE and aggregate these to conduct a population level impact analysis.

The aggregation within the TRE will be applied as follows:

1. Ipsos colleagues access the pseudonymised data within the TRE. Analysts will be able to access only the data they are permitted to see and can utilise a variety of analytical tools available within the TRE platform

2. Ipsos colleagues will aggregate patient level cancer data in the TRE (i.e. from the National Cancer Registration Dataset, Cancer Waiting Times, and Civil Registration Mortality) to MSOA-level geographies for patients aged 55-76 years of age at the time of diagnosis with lung cancer in half-yearly periods (i.e. first half of 2016, second half of 2016, etc.) to provide aggregated information (as base numbers and proportions of populations) on the aforementioned outcomes cancer incidence, stage at diagnosis, one-year survival, and mortality.

This information will be calculated for all persons aged 55-76 per MSOA, and to understand health inequality effects of the TLHC, broken down per MSOA by the following sub-groups:

• Male/female

• Age brackets (e.g. 55-64, 65-76)

• Ethnicity groups (where numbers permit)

3. Subsequent baseline and impact analysis will be performed within the TRE by Ipsos. Ipsos’ analysts will process this data in aggregated form with small numbers unsuppressed, any outputs produced will be aggregated with small numbers suppressed.

4. Only summary, aggregate results data (data will be aggregated with small numbers suppressed in line with the HES analysis guide) will be exported from the TRE and this will be subject to review and approval by the NHS Digital team providing the TRE. The objective of this will be to ensure that no output contains information which could be used either on its own or in conjunction with other data to breach an individual's privacy.

Analysts will be able to access only the data they are permitted to see and can utilise a variety of analytical tools available within the TRE platform.

Only summary, aggregate results data (data will be aggregated with small numbers suppressed in line with the HES analysis guide) will be exported from the TRE and this will be subject to review and approval by the NHS Digital team providing the TRE. The objective of this will be to ensure that no output contains information which could be used either on its own or in conjunction with other data to breach an individual's privacy.

Following similar principles to those adopted by the SAIL Databank for Wales (https://saildatabank.com/saildata/data-privacy-security/#secure-access) and the Scottish National Data Safe Haven (https://www.isdscotland.org/Products-and-Services/EDRIS/Use-of-the-National-Safe-Haven/), only summary, aggregate results data are exported from the TRE by Ipsos MORI, subject to NHS Digital audits of outputs from the system by trained output checkers. This ensures that no output contains information which could be used either on its own or in conjunction with other data to breach an individual's privacy.

All possible ways of data minimisation have been considered and undertaken where possible, therefore, the Data Controller has considered their legal obligations under UK General Data Protection Regulation (UK GDPR).

Output 1: Baseline analysis – to identify suitable counterfactual areas

Anticipated timing of analysis (i.e., cleaning the data, setting up the data set and conducting the baseline analysis): Late 2022, dependant on setup time to access the TRE

• The selected methodological approaches and the findings of the baseline analysis will be written up by Ipsos in form of a methodological protocol and shared with NHSE. The protocol will also outline the econometric approach for the impact evaluation and a discussion of the implications of the findings from the baseline analysis.

• Ipsos hopes that this output will be delivered in early 2023.

• The protocol will include aggregated data for the treatment and selected comparison areas in tabular format and graphical illustrations of trends in pre-intervention outcomes between treated and comparison units. It will further include the standard output required to assess the quality of a matching analysis (e.g. graphical illustration of common support areas, t-test of means, and reported reduced standardised percentage bias between treated and matched untreated units after matching). Variable means and standard deviations will be presented for descriptive statistics and mean-estimates with standard errors and p-values (0.1, 0.05, and 0.01 level) indicating statistical significance will be presented for estimations. Small numbers will be suppressed in line with HES analysis guide.

Output 2: Impact evaluation analysis

Anticipated timing of impact analysis (i.e. cleaning new data, updating the existing dataset with the new data, and conducting the impact analysis: mid-2024

• Results from the impact evaluation will be written up by Ipsos and shared with NHSE as part of the final TLHC programme evaluation report.

• It is anticipated that this report will be delivered in September 2024.

• The impact evaluation section will include a light touch presentation of the methodologies and main findings from the selected econometric approaches (e.g. the average treatment effects of the TLHC on the specified outcomes on the full sample and by sub-samples).

• The report will be accompanied by a technical appendix explaining the used econometric methodologies in more detail and presenting various iterations of findings tabular and graphical form. As for the presentation of findings form the baseline analysis, only aggregated data for the treatment and selected comparison areas will be presented with small numbers suppressed. Likewise. Variable means and standard deviations will be presented for descriptive statistics and mean-estimates with standard errors and p-values (0.1, 0.05, and 0.01 level) will be shown for any output derived from an estimation.

NHSE may produce information to be made publicly available, where this information is not deemed sensitive or inappropriate. Any published reports will contain only information that is aggregated with small numbers suppressed in lie with the HES Analysis Guide.

The TLHC programme is a flagship programme of work for NHSE which will contribute to the ambition of the NHS Long-term Plan to improve early diagnosis and survival for those diagnosed with cancer. The principle aims of the programme are: to increase the number of lung cancers diagnosed at stage one or two by identifying those at risk of developing lung cancer and inviting them for a lung health check.

Without an intervention of this kind, approximately only 30% of lung cancers are found at an early stage. Findings from lung cancer research trials and initial results from the TLHC programme, show lung cancer screening can increase early stage detection to as high as 75% - (Reference for UKLS trial: Field, J.K., Duffy, S.W., Baldwin, D.R., Brain, K.E., Devaraj, A., Eisen, T., Green, B.A., Holemans, J.A., Kavanagh, T., Kerr, K.M. and Ledson, M., 2016. The UK Lung Cancer Screening Trial: a pilot randomised controlled trial of low-dose computed tomography screening for the early detection of lung cancer. Health technology assessment (Winchester, England), 20(40), p.1.)

An evaluation of this programme is necessary to determine the real-world impacts of this pilot. The findings from this analysis (using data from the TRE) hope to quantify how the TLHC pilot has contributed to the NHS Long Term Plan ambition of diagnosing 75% of cancers at an early stage by 2028; and be used to predict the ongoing impact of a national rollout of this programme.

Evaluation Questions:

This analysis is hoped to answer some of the key question from the TLHC evaluation including how effective are TLHCs in relation to cancer outcomes (incidence, stage at diagnosis, mortality) in areas where the TLHC programme is not yet established.

Accessing anonymised patient level data through the TRE will allow Ipsos colleagues to conduct sub-group analyses (i.e. by gender, age and ethnicity) of the TLHC pilot impacts on outcomes at an MSOA-level, which would not be possible through accessing aggregated data due to data not meeting the anonymisation threshold. This means Ipsos would be able to explore whether there are a stronger effects of the TLHC pilot in specific sub groups.

How the findings will be used:

Findings from this evaluation are hoped to help identify the need for a lung cancer screening programme of this kind across England and further afield. It will also inform how the programme’s roll out should take place, for instance if people with certain characteristics, demographics or living in particular geographical areas, should be targeted first over others. Findings from the sub group analysis will inform whether the programme should adapt its target population or change its promotional materials so that specific people are targeted to drive uptake in areas where it is especially low.

Having access to the combined data supports the wider aim of evaluating the impact and effectiveness of the TLHC in achieving the intended goals. If the evaluation demonstrates that TLHCs are effective and are having a positive impact in terms of improving the patient experience and outcomes for patients with suspected cancer, the findings of this evaluation would enable NHSE to work with Cancer Alliances and local authorities to expand TLHC programmes and improve the effectiveness of TLHCs. If the evaluation demonstrates the intended goals for TLHCs are being achieved, the information derived from the data in the NHS England - Cancer TRE will support conversations with commissioning groups about developing sustainable funding models and evidencing the wider benefits of the TLHCs.

Unable to evidence as data has not yet been used for this purpose.