Executive summary

Hospital-TTD-Mod is a model of tobacco dependence treatment for adult inpatients in acute hospitals in England. It estimates what the inpatient tobacco dependence services set up under the NHS Long Term Plan cost, how many more patients they help to stop smoking, and what that does to the patients’ hospital use, deaths, quality of life and lifetime NHS costs. It was built at the University of Sheffield with NHS England, and runs in R for England, any integrated care board (ICB) or any trust.

The model covers adult inpatients in the 129 trusts with inpatient tobacco dependence services: 126 acute trusts and 3 community trusts with inpatient beds. The maternity and mental health pathways, which are separate in the NHS Standard Treatment Plan (Papadakis et al. 2023), are not included.

Each year from 2019, the index year, to 2045, the time horizon, the model takes the year’s patients, from Hospital Episode Statistics (HES) for 2023/24, by the condition they were admitted with, age group, sex and deprivation, and puts them through the care pathway with and without the service: screening and very brief advice at admission, assessment by a specialist advisor in person or by phone, NRT, follow-up calls, and referral for specialist support after discharge. Smoking among patients comes from a synthetic population of England (Head et al. 2026), calibrated to national audits of smoking among inpatients. The years before 2026, the policy effect year, are calibrated to NHS England’s data on referrals to the services, so the scenarios start from the coverage the services have actually reached.

Each year’s patients are then followed for five years after discharge. Patients who quit have a lower risk of admission for each of 42 smoking-related conditions, falling with time since quitting, and a lower risk of dying from them. For the conditions with studies of quitting after diagnosis (ischaemic heart disease, ischaemic stroke, COPD and lung cancer), the effects come from those studies; for the rest, from studies of new disease. The benefits come only through quitting: patients seen by the service who carry on smoking are no better off.

The results are given from two sides, which value the same quits and are never added together. Tier 1 is the trust: the cost of the service against the cost of the admissions it avoids. Tier 2 is the health system: the cost of the service, specialist support after discharge and ward staff time, against the NHS treatment costs avoided and the quality-adjusted life years (QALYs) gained over the lifetimes of the extra people who quit for life, valued as NICE values stop smoking interventions and judged against NICE’s threshold. Costs are in 2024/25 prices. Results are given for one year of the service, for the roll-out to 2030 and for the programme to 2045, by deprivation, age and sex, and with sensitivity analyses of the main assumptions.

Earlier models gave every patient who took up support the readmission reduction seen in the Ottawa evaluation (Mullen et al. 2017), counted admissions as if they were patients, and used one effect for every condition. This model gives the benefit only to the extra quitters, counts patients and admissions separately, and works condition by condition. The Ottawa and London evaluations (Mullen et al. 2017; Robins et al. 2024) are used as checks, not as targets.

This document describes the model as it is in version 3: its data, how they are prepared, the clinical pathway, the epidemiology, the economics, the checks, the analysis of inequalities, and how to run it.

Use of AI tools

AI assistants (Claude, Anthropic; Gemini, Google) were used to help write and review parts of the model code and this documentation. They had no access to any sensitive data. Everything they contributed has been thoroughly reviewed by the author: the logic has been checked against the evidence and the model design, the code has been tested, and the text has been checked against the code it describes. Duncan Gillespie takes full responsibility for the model, its code, documentation and results.