2 Data sources
This chapter describes the data the model uses and what each is used for. The last section says briefly why each main source was chosen.
2.1 The care pathway
The steps of the pathway follow the NHS Inpatient Tobacco Dependence Treatment Care Bundle as set out in the NHS Standard Treatment Plan (Papadakis et al. 2023): at admission, screening, very brief advice, NRT and an opt-out referral to the trust’s tobacco dependence team; in hospital, an assessment by a specialist advisor, in person or by phone; at discharge, stop smoking aids and a referral for specialist support after discharge; and after discharge, follow-up calls from the trust team. The rates at each step, with and without the service, come from national audits and evaluations of services in England (chapter 4).
2.2 Routine data
2.2.1 Hospital Episode Statistics (HES) Admitted Patient Care
HES records every admission to an NHS hospital in England. The model uses the financial year 2023/24, the HES year, for the number of patients and admissions, the condition each admission was for (ICD-10), each patient’s other smoking-related conditions, and the cost and length of each admission.
The model uses ordinary admissions of adults (16 and over), not maternity and not ending in death (chapter 3, “Admissions that end in death”). The main condition of an admission is found by scanning the first five diagnosis positions of the first episode of a continuous inpatient spell: if a smoking-related diagnosis is found, the one in the highest position is the main condition; otherwise the admission is “Other”. Each patient is placed in the group of the main condition of their first admission of the year, and all their admissions in the year count in that group. A patient’s other conditions come from every diagnosis position of every episode in the year.
2.2.2 NHS England patient-level data on referrals
NHS England’s patient-level data for the tobacco dependence services, extract 67_017 (June 2026), give the admissions of acute inpatients recorded as smoking with an in-house referral to the service, by month of admission from April 2022 to February 2026, by age group, sex, ethnic group and IMD quintile. NHS England rounds counts to the nearest 5 and suppresses small counts. The extract stays on the virtual machine with the other restricted data. It is used to calibrate the service’s coverage in 2023 to 2025 (chapter 7). Not every trust submitted every month, so it undercounts referrals.
NHS England also supplied the share of acute trusts with a live service by financial year (5% in 2020/21, 16% in 2021/22, 51% in 2022/23, 86% in 2023/24, 93% in 2024/25, 91% in 2025/26 and 92% in 2026, with at most 95% expected).
2.2.3 Acute trust catchments
The Office for Health Improvement and Disparities (OHID) publishes the population each NHS acute trust serves, built from HES. The model uses Table 2 of the April 2026 release: patients from each Middle layer Super Output Area (MSOA, 2021 boundaries) admitted to each acute trust, all admissions, 2021/22 to 2023/24. These give the population each trust and ICB serves when estimating its smoking prevalence. They don’t cover community trusts, so the three community trusts with inpatient beds take the catchment of the acute trusts serving the same county.
2.2.4 Office for National Statistics (ONS) deaths
Patients who die after discharge can’t be readmitted, so the model includes deaths after discharge. ONS deaths from all causes in England in 2023, by age, sex and IMD, give the all-cause death rate. ONS deaths in England in 2018 by underlying cause, by single year of age, sex and IMD quintile, with each of the model’s conditions as a cause, give each condition’s share of those deaths. With HES counts of the patients who have each condition, these give each condition’s deaths a year per patient who has it, and the death rate from other causes (chapter 5). The rates are for England and are applied to every ICB and trust, each with its own mix of patients. The deaths by cause file holds small counts and stays on the virtual machine.
2.3 Smoking
HES doesn’t record smoking reliably, so smoking among patients is estimated from survey data on the population each hospital serves, and calibrated to national data on smoking among inpatients.
2.3.1 Health Survey for England (HSE)
The HSE 2017, 2018 and 2019 waves are pooled. Each adult respondent’s smoking status (current, former or never cigarette smoker), age group, sex and IMD quintile come from the UK Data Service end user licence files. Missing smoking status (189 of 24,379 adult respondents) is imputed from age, sex, IMD quintile, survey year, ethnicity, partnership status, qualifications, socioeconomic classification and general health.
2.3.2 Synthetic population of England
The respondents are placed in neighbourhoods through a synthetic population of the adults of England (Head et al. 2026), built by iterative proportional fitting: HSE 2017 to 2019 adults were reweighted so that each of the 33,755 Lower layer Super Output Areas (LSOAs, 2021 boundaries) matches the 2021 Census for age and sex, ethnicity, highest qualification, partnership status, general health and socioeconomic classification (NS-SEC), and its IMD 2019 quintile. It was scaled to mid-2022 populations and integerised, giving 46.2 million synthetic adults, each drawn from one HSE respondent. It covers ages 16 to 89.
2.3.3 Index of Multiple Deprivation 2019
Each LSOA has its quintile of the English Index of Multiple Deprivation 2019, from the official File 1 (2011 LSOAs), linked to 2021 LSOAs with the ONS exact fit lookup. This is the same measure as the IMD quintile recorded in HES.
2.3.4 Smoking among hospital inpatients
The calibration targets are in user_inputs/smoking_calibration_panel.csv:
- Overall prevalence: 16.3% of eligible admissions with a smoking status recorded in NHS England’s patient-level data for 2024 (Agrawal et al. 2026), or 21% of adult inpatients in the British Thoracic Society national audit of 2021 (Devani et al. 2023), chosen with
SMOKING_HEADLINE_SOURCEin00_config.R. - Prevalence by condition group, relative to the overall prevalence, from national clinical audits (lung cancer, COPD, stroke), published studies and assumptions for the remaining groups. For ischaemic heart disease the target is weighted by how patients were first admitted: 26.3% among patients admitted as an emergency with ST-elevation myocardial infarction (Khattak et al. 2024) and 8.6% among patients having planned coronary artery bypass grafting (Sewart et al. 2025), weighted by the shares of IHD patients whose first admission in HES was non-elective or elective.
- The spread of patients who smoke by IMD quintile, sex and age group, from the 185,195 patients recorded as smoking at their first eligible admission in 2024 (Agrawal et al. 2026).
2.4 Epidemiology
2.4.1 Relative risks
The relative risks of each smoking-related condition for current smokers against never smokers come from a review of the risks of smoking for adult diseases (Webster et al. 2018).
2.4.2 Recovery curves
The excess risk of a former smoker falls with time since quitting, quickly for cardiovascular disease, more slowly for respiratory disease and most slowly for cancer. The model uses the three curves of Kontis et al. (Kontis et al. 2014; Oza et al. 2011). The curve each condition follows is set in the table of effects of quitting (chapter 5).
2.4.3 Effects of quitting after a diagnosis
For conditions with studies of people who quit after their diagnosis, the effect of quitting on admissions and deaths comes from one study for each: ischaemic heart disease (Wu et al. 2022), ischaemic stroke (Chen et al. 2019), COPD (Josephs et al. 2017; Doo et al. 2023) and lung cancer (Caini et al. 2022). Chapter 5 gives the reasons for each choice and the other studies considered.
2.5 Costs and outcomes
2.5.1 NHS Payment Scheme prices
The cost of an admission avoided is its price in the 2023/25 NHS Payment Scheme (2023/24 prices), by Healthcare Resource Group, including payments for long stays, as in chapter 3, converted to 2024/25 prices with the NHS Cost Inflation Index (chapter 6, “Prices”).
2.5.2 Staff and other unit costs
Staff costs are from the PSSRU Unit Costs of Health and Social Care 2025 (Jones et al. 2025): salary and on-costs for hospital nurses by Agenda for Change band (Table 11.2.2), administrative staff by band (Table 12.2.4) and consultants (Table 11.3.2). The cost of NRT per patient, the overhead and consultant time follow NHS England’s costing of the service (chapter 6).
2.5.3 Specialist support after discharge
The cost of specialist support after discharge is the cost per quitter of local authority stop smoking services in 2023/24, £642 (NHS Digital 2024) (£668 in 2024/25 prices), taken for every route (chapter 6).
2.5.4 Lifetime NHS savings, life years and QALYs
The lifetime value of each person who stops smoking for good is worked out by Stapleton and West’s method (Stapleton and West 2012): life years gained by age at quitting from Doll et al. (Doll et al. 2004) below 65 and Le et al. (Le et al. 2024) at 65 and over (Le et al. at every age for the lower values), discounted at 3.5% and net of the years that would have been gained by stopping later anyway, with QALYs at 1.3 per life year (1.0 for the lower values). The NHS treatment costs avoided are from Godfrey et al. (Godfrey et al. 2011), put on the same basis. Remaining life expectancy, used for the timing of both, is from the ONS single-year life tables for England, 2024 (Office for National Statistics 2025). Chapter 6 sets out the method and how to change each assumption.
2.6 Why these sources
| Source | Used for | Why this one |
|---|---|---|
| HES 2023/24 | Patients, admissions, conditions, costs, lengths of stay | Every NHS admission in England, with diagnoses, so patients, admissions and conditions can be counted for every trust and ICB |
| NHS Standard Treatment Plan (Papadakis et al. 2023) | The steps of the pathway | NHS England’s specification of the care bundle the services deliver |
| NHS England patient-level data | Coverage in 2023 to 2025; service results in 2024 (Agrawal et al. 2026) | The only national data on the services’ activity |
| CURE pilot (Evison et al. 2020; Evison et al. 2021), London evaluation (Robins et al. 2024), BTS audits | Pathway rates with and without the service; consultation times, staffing ratios, follow-up after discharge | Evaluations of services in England with the stage-by-stage counts the pathway needs |
| Meta-analysis of very brief advice (Cheng et al. 2024) | A check on the effect of very brief advice alone | Pooled trials of advice lasting three minutes or less against no advice |
| Synthetic population (Head et al. 2026), HSE 2017-2019 | Smoking in each area’s population by age, sex and IMD | Smoking for every neighbourhood, consistent with national survey data, so each trust’s population can be described |
| BTS audit 2021 (Devani et al. 2023), patient-level data (Agrawal et al. 2026), condition audits and studies | Smoking among inpatients | National measures of smoking among inpatients, by condition where they exist |
| Relative risks (Webster et al. 2018), recovery curves (Kontis et al. 2014) | Risk of each condition by time since quitting | Condition-specific risks for a wide list of conditions, with recovery after quitting for three disease groups |
| Studies of quitting after diagnosis | Effects of quitting for IHD, stroke, COPD and lung cancer | The populations most like the patients: people who already have the disease (chapter 5) |
| ONS deaths | Deaths after discharge | National deaths by cause, age, sex and IMD |
| NHS Payment Scheme | Cost of admissions avoided | What trusts are paid for an admission |
| PSSRU 2025 (Jones et al. 2025) | Staff costs | The standard UK source for staff unit costs |
| NHS Digital stop smoking services statistics (NHS Digital 2024) | Cost of support after discharge | The only national cost per quitter |
| Stapleton and West (Stapleton and West 2012), with Doll et al. (Doll et al. 2004) and Le et al. (Le et al. 2024) | Life years and QALYs gained over a lifetime | A simple, published UK method for valuing a quit, from the longest follow-up of people who stopped (Doll et al.), with the only estimates that cover quitting at 65 and 75 (Le et al.) |
| Godfrey et al. (Godfrey et al. 2011) | Lifetime NHS savings | Lifetime NHS costs for England by single year of age at quitting and sex; NICE’s later models give larger savings (chapter 6), so this is the cautious choice |
| ONS life tables (Office for National Statistics 2025) | Timing of the lifetime benefits | Current life expectancy for England by single year of age and sex |