6  Health economic evaluation

The health economic evaluation (03_econ_engine.R, called from 04_master_loop.R) costs the service year by year and values what it achieves from two perspectives:

The two tiers are two valuations of the same quits. They are never added together.

6.1 The forward programme

Every headline figure is for the forward programme: the service years from the policy effect year (2026) to the time horizon (2045), and the admissions avoided by the patients treated in those years over each cohort’s five follow-up years (so savings run to 2049). Costs and savings are discounted to 2026 at discount_rate (3.5%). The years 2019 to 2025 are history: they are modelled, so that coverage and the carry-over of earlier quitters build up as they did, and their costs appear year by year in Tab_3b_Costs_By_Year, but they are not part of any headline, cumulative position, return on investment or break-even year.

6.2 Incremental volumes

Every cost and benefit is the difference between the treatment arm (with the service) and the control arm (without it) in the same year: assessments, nicotine replacement therapy (NRT), identified smokers, screening, very brief advice, quits with specialist support after discharge, and 4-week, 12-month and lifetime quits.

6.3 Costing year by year

Each service year is costed from its own activity (cost_service_years()).

Cost How it is worked out
Tobacco dependence advisors whole-time equivalents (WTE) needed for the year’s assessments: in-person assessments plus phone assessments at their share of an in-person assessment’s time (cost_phone_assessment_mins / cost_specialist_assessment_mins), divided by one advisor’s annual capacity (advisor_patients_per_day x working_days_per_year)
Band 8a leads band_8a_fte_per_trust for each live trust
Consultant time consultant_pa_per_trust programmed activities a week for each live trust, a full-time consultant being 10
Band 6 staff band_6_per_advisor x advisor WTE, with at least band_6_minimum_fte per live trust
Admin staff admin_per_advisor x advisor WTE (0.33, one administrator to three practitioners as in the CURE pilot (Evison et al. 2020)), with at least admin_minimum_fte per live trust
NRT net NRT supplies x cost_pharmacotherapy
Overhead staff costs x overhead_share_of_staff_cost (20%)
Specialist support after discharge (Tier 2 only) net 4-week quits with specialist support after discharge x cost_support_after_discharge_per_quit
Ward staff time (Tier 2, shown separately in Tier 1) net screening x cost_screen_mins plus net very brief advice x cost_vba_mins, at ward_nurse_hourly_rate

Live trusts in a year are the number of trusts in the area (counted from trusts_in_scope.csv: all 129 for England, the ICB’s trusts for an ICB, one for a trust) times Trust_Coverage_Pct for the year. Salaries are applied to each grade’s WTE. Staff are costed as exact fractions of a post, with no rounding up to whole posts.

6.3.1 Unit costs

Staff costs are annual salary plus employer’s on-costs from the PSSRU Unit Costs of Health and Social Care 2025 (Jones et al. 2025), and the other unit costs follow NHS England’s costing of the service. Overheads are added as a share of staff costs rather than taken from PSSRU, whose overheads include estates and capital that a ward-based service shares with the rest of the hospital.

Parameter Value (GBP) Source
Advisor, band 4 (salary_advisor) 34,849 PSSRU 2025 Table 11.2.2, hospital nurses: salary 26,606 + on-costs 8,243. NHS England: advisors are band 3 or 4
Band 8a lead 76,444 Table 11.2.2: 56,857 + 19,587
Band 6 55,087 Table 11.2.2: 41,325 + 13,762
Admin, band 3 32,768 Table 12.2.4, band 3 administration pay 25,017, with on-costs at the band 4 rate (31.0%)
Consultant, per year full time (salary_consultant) 170,707 Table 11.3.2, consultant (medical): 125,412 + 45,295. One session a week per trust is an assumption
NRT per patient prescribed 36 NHS England: a two-week course of dual NRT (PSSRU 2025 section 7.3 gives 19.95 a week for combination NRT)
Overhead 20% of staff costs NHS England
Ward nurse, per hour 47 Table 11.2.2, band 5, cost per working hour excluding qualifications

Advisor productivity (advisor_patients_per_day) is the largest single influence on the cost to Trusts, because it sets the advisor WTE and, through the ratios above, the band 6 and admin WTE. The base case is 5 assessments a working day, an assumption. CURE’s specialist consultation takes 45 minutes (Evison et al. 2020), so five take half of a 450-minute working day, leaving time for follow-up calls (chapter 4), finding patients on the wards and records. Published estimates are lower: 3.72 in the CURE pilot (1,450 patients assessed by 3 WTE nurses over 130 working days), whose nurses also made 2.1 follow-up contacts after discharge for each patient assessed (Evison et al. 2020); about 3 in NHS England’s original model of the service (90 minutes a patient at 61% patient-facing time); and 1.51 in the QUIT service in South Yorkshire. Sens_Productivity_Low tests 3 a day. It is an input in scenarios_control_panel.csv and can be changed there for any scenario.

Nine sensitivity analyses test the costing, one input at a time, on one rule: each cost input 25% below and above its base case value, except advisor productivity, the most uncertain, which is tested at 3 and 10 assessments a day (Sens_Productivity_Low and Sens_Productivity_High). Sens_NRT_Low and Sens_NRT_High set NRT at 27 and 45 a course. Sens_Overhead_Low and Sens_Overhead_High set overheads at 15% and 25% of staff costs. Sens_Consultant_Low and Sens_Consultant_High set consultant time at 0.75 and 1.25 sessions a week per trust. Sens_Advisor_Band3 costs advisors at band 3 (32,768: there is no band 3 in the nursing tables, so administration pay with on-costs at the band 4 rate).

Stop smoking medicine is costed as combination NRT (cost_pharmacotherapy) for every patient given it. Varenicline and cytisine are not modelled separately: the quit rates come from studies of quit attempts with and without support, not of particular medicines, so the choice of medicine would change only the cost, and a two-week course of varenicline costs less than the NRT modelled, within the range of Sens_NRT_Low and Sens_NRT_High.

Costs are at national rates. For ICB and trust runs, staff pay, ward staff time and the unit costs of admissions avoided can be adjusted for local costs by the market forces factor (MFF_MODE, chapter 8); NRT, specialist support after discharge, lifetime savings and QALYs are not adjusted.

Of the cost to Trusts, the fixed costs of each live trust (the band 8a lead and the consultant session, with their overheads) are a larger share while services reach only part of their patients, and a smaller share at full coverage.

Ward staff time is an opportunity cost, not cash: the ward nurses who screen patients and give very brief advice are employed anyway, and the time comes out of other ward work. It is therefore included in the Tier 2 system cost and shown as a separate line in Tier 1, outside the trust’s cash position.

Tab_3b_Costs_By_Year shows, for every service year, the activity, the WTE and cost of each grade, every cost line, the discount factors and the discounted values. The report’s workforce table (tbl_workforce) shows the WTE and cost by grade, with overheads and NRT, in 2026 and the first year at full coverage.

6.3.2 Prices

Every cost and saving is in 2024/25 prices (MODEL_PRICE_YEAR in 00_config.R). Inputs published in other years are converted with the NHS Cost Inflation Index (data/base/raw/nhs_cost_inflation_index.csv, applied by R/prices.R): the HCHS pay and prices index to 2014/15, then the NHSCII (overall) from the PSSRU manual (Jones et al. 2025). The 2024/25 figure (4.02%) is provisional.

Input Published in Converted by
Staff pay, ward nurse time, NRT 2024/25 (PSSRU 2025) none
Cost of an admission avoided (NHS Payment Scheme) 2023/24 1.0402, when the model reads it (main.R)
Specialist support after discharge (cost_support_after_discharge_per_quit) 2023/24 (642 GBP) 1.0402, entered as 668 GBP in the control panel
Lifetime NHS savings (Godfrey et al.) 2008/09 1.4217, in 05_build_lifetime_values.R

The money values in the control panel are entered in 2024/25 prices. To move the model to a later price year, add the year’s index to the index file, set MODEL_PRICE_YEAR, and re-enter the control panel’s money values in that year’s prices.

6.3.3 Specialist support after discharge

Specialist support after discharge is costed at £668 for each extra four-week quit made with it: the cost per quitter of local authority stop smoking services in England, the median across regions, £642 in 2023/24 (NHS Digital 2024), in 2024/25 prices. It is a cost to the local health system, in Tier 2. It is not in the trust’s costs in Tier 1, though it may be a cost to a trust whose own team gives the support.

Areas give this support through different routes (Papadakis et al. 2023): the local authority stop smoking service, the NHS Community Pharmacy Smoking Cessation Service, the trust’s own team for 3 to 12 months after discharge, or a digital service. The model uses one cost per quit for all of them, the local authority figure, for three reasons. It is the only national cost per quit, from the national returns of local authority services. Local authority services are the route the Standard Treatment Plan names first and the one the registration and quit rates in the model come from (Cherodian et al. 2025). And no cost per quit is published for the other routes: a pharmacy is paid £30 for the first consultation, £10 for each interim one and £40 for the last (Community Pharmacy Surrey and Sussex 2025), plus the stop smoking aids, but how many patients finish the course and quit is not published; the cost of the trust’s extended support depends on how many contacts each patient has; and digital services cost little per user but have no national quit rates. A local system that knows its own mix can set cost_support_after_discharge_per_quit to its own cost per quit. The cost of support after discharge is shown on its own in the results, so it can be read against local figures.

The advisors’ follow-up calls in the first four weeks after discharge, which the Standard Treatment Plan asks the trust team to make, are not part of this cost: they are within the advisors’ time costed in Tier 1 (chapter 4).

6.4 Bed days freed

Each avoided admission frees the average length of stay for its condition and patient group, from the HES inputs. Bed days freed build up as successive cohorts are treated and each cohort’s effect runs through its five follow-up years.

6.5 Tier 1: the trust

Tier 1 sets the trust’s discounted cost of the forward programme (staff, NRT and overhead) against the discounted cost of the admissions the forward cohorts avoid, valued at the unit cost of each condition. Only proportion_cashable_savings (50%) of the avoided cost is counted as cashable, because a freed bed is not all cash: much of the saving is capacity that gets used for other patients.

Tab_3_Cash_Flow_Profile shows the forward programme by calendar year from 2026: the trust’s cost in each service year, the cashable savings falling in that year, the net position in the year, and the cumulative position and return on investment from 2026. The break-even year is the first year the cumulative position turns positive. The cash flow and Tab 1 are built from the same per-year costs and savings, and the run stops if their totals differ.

6.6 Tier 2: the health system

Tier 2 sets the cost to the health system (the trust’s cost, specialist support after discharge and ward staff time) against what the extra people who quit for life gain over their lifetimes: the NHS treatment costs they avoid and the quality-adjusted life years (QALYs) they gain, with the life years they gain alongside. It values quitting as NICE’s economic models value stop smoking interventions, so the service can be compared with other ways of helping people to stop.

6.6.1 Who is valued

The people valued are the extra lifetime quitters: the extra people still quit at 12 months (treatment arm minus control arm), times pr_smokefree_lifetime_given_12m (0.65), the share who stay quit for life. The 0.65 is Stapleton and West’s allowance for relapse after 12 months (Stapleton and West 2012). They are valued whatever their admission was for.

6.6.2 The lifetime value of each quitter

05_build_lifetime_values.R works out, for each age band and sex, the life years, QALYs and NHS savings of one person who quits for life, as present values at the age they quit, and writes them to data/base/derived/lifetime_values_per_quitter.csv. The assumptions are in one editable file, user_inputs/lifetime_benefits_panel.csv, one row per age band. The model reads only the derived file, so values from another source can replace it (see “Changing the lifetime assumptions” below).

With the panel as supplied, a person who quits for life gains:

Age band Quit age used Life years QALYs Life years and QALYs, lower NHS savings, men NHS savings, women
16-24 20 0.77 1.00 0.62 £1,943 £1,534
25-34 30 1.08 1.41 0.87 £2,214 £1,793
35-49 42 1.60 2.08 1.20 £2,707 £2,433
50-64 57 1.34 1.74 1.05 £2,112 £2,466
65-74 70 0.65 0.84 0.56 £1,316 £1,642
75-89 82 0.32 0.42 0.28 £423 £504
90+ 92 0.17 0.23 0.15 £242 £163

All are discounted at 3.5% from the age of quitting, and the money is in 2024/25 prices. The NHS savings are the same in the central and lower values.

6.6.2.1 Life years

Life years follow Stapleton and West’s method (Stapleton and West 2012). A person who stops smoking for good gains life_years_gained years of life (undiscounted). The years come at the end of life, so they are discounted from the age of quitting to the age they arrive (gain_arrives_at_age). Some of the gain would have come anyway, because many smokers stop later in life without help; that share (quit_later_share) is taken off:

\[\text{life years} = \frac{\text{life years gained}}{1.035^{\,(\text{arrives} - \text{quit age})}} \times (1 - \text{share who would quit later})\]

With quit ages of 30, 40, 50 and 60, 10, 9, 6 and 3 years gained (Doll et al. 2004) and the gain arriving at 82, this gives Stapleton and West’s Table 1 rows B and F exactly for three of their four age bands; their 35 to 44 row is 1.5% above what their stated method gives. 05_build_lifetime_values.R checks this each time it runs. The share who would quit later is their row E: 0.351 under 35, 0.263 at 35 to 44, 0.210 at 45 to 54 and 0.175 at 55 to 64 (2.5% of smokers stopping each year), weighted by years of age where a model band spans two of theirs.

Age band Life years gained, central Lower Gain arrives at Share who would quit later
16-24, 25-34 10 8.0 82 0.351
35-49 8.4 6.32 82 0.245
50-64 3.9 3.06 82 0.187
65-74 1.38 1.20 86.4 0.175
75-89 0.52 0.45 90.2 0.175
90+ 0.24 0.21 95.8 0.175

Below 65 the central values are Doll et al.’s gains from 50 years’ follow-up of British doctors (Doll et al. 2004), as Stapleton and West used them, at the middle of each model band. Doll et al. stop at 60, and many hospital patients who smoke are older, so the older bands use Le et al. (Le et al. 2024), the only estimates that cover 65 and 75: quitting at 55, 65 and 75 gains 3.4, 1.7 and 0.7 years. Their 65 to 74 value is the average of 65 and 75 (1.2 years; the same averaging of 55 and 65 gives their own 55 to 64 value of 2.6), scaled by 3/2.6 so it joins Doll et al.’s 3 years at 55 to 64. Above 75 the gain is taken to fall in proportion to remaining life expectancy (Office for National Statistics 2025). For people who quit at 65 or over the gain arrives at the end of their remaining life expectancy, not at 82. Taylor et al. found similar gains at 65 (1.4 to 2.0 years for men, 2.7 to 3.7 for women) (Taylor et al. 2002).

The lower values take Le et al.’s gains at every age. Le et al. used 2018 United States mortality and relative risks by years since quitting, and their gains below 65 are 13% to 25% smaller than Doll et al.’s.

Men and women gain the same. Stapleton and West note that women gain slightly more, but after discounting the difference is small enough to ignore.

6.6.2.2 QALYs

QALYs are life years times qalys_per_life_year: 1.3 in the central values, 1.0 in the lower. Stapleton and West counted life years only (one QALY per life year), as earlier UK guidance for commissioners had (Parrott et al. 1998), so that results stay comparable over time, and noted that smoking cessation models have used 0.9 to 1.4 QALYs per life year, because quitting also reduces years lived with disease. NICE’s own models count both: per extra person quit at 12 months they give about 1.0 QALY for the general population (Filby and Taylor 2018) and 1.1 for general hospital inpatients (Mallender et al. 2013). With 1.3, this model gives a similar value per extra 12-month quitter (the report compares them); with 1.0 it gives about a quarter less.

6.6.2.3 NHS savings

The NHS savings are from Godfrey et al. (Godfrey et al. 2011), who modelled the lifetime NHS costs of heart attack, stroke, lung cancer and COPD, and of deaths from other causes, for 1,000 people who stop smoking at each age from 35, against 1,000 who carry on, all free of the four diseases at the start. Their Appendix 9 gives the saving by single year of age and sex, undiscounted, in 2009 prices. Three steps put it on the same basis as the life years:

  1. Discounting. Godfrey et al. give the discounted saving only at 35 (Table 6): 52% of the undiscounted saving for men and 44% for women at 3.5%. That is what the whole saving would be worth if it came 18.8 years (men) or 23.9 years (women) after quitting. At other ages the delay is in proportion to remaining life expectancy (Office for National Statistics 2025), so savings for older quitters come sooner and are discounted less.
  2. People who would have quit later. The saving is reduced by the same share as the life years, because Godfrey et al.’s comparison group never quits.
  3. Prices. From 2008/09 (the HCHS index Godfrey et al. used) to 2024/25 with the NHS Cost Inflation Index, a factor of 1.4217 (see “Prices” above).

Under 35 takes the saving at 35, discounted from the person’s own age. 90 and over takes the saving at 89, because Godfrey et al.’s 90-and-over row is larger than at 89 and not on the same basis. Each band is the average of its single years of age.

6.6.3 A population prevention view

Both sources describe people who quit without the smoking-related diseases: Godfrey et al.’s cohort starts free of the four diseases, and Doll et al.’s and Le et al.’s gains are for general populations. This is how NICE valued quitting in its guidance, including for general hospital inpatients (Mallender et al. 2013). It gives a valuation that can be compared with community stop smoking services, but it is not a model of what quitting does for these patients:

  • Tier 2 counts new disease. Tier 1 counts mostly readmissions for conditions patients already have, which Godfrey et al. leave out. In the first five years both count some of the same benefits (new disease avoided), so the two tiers are never added together; each also counts benefits the other leaves out.
  • Hospital patients who smoke are older and sicker than the general population. They have less life left, which lowers the life years they can gain, but a higher absolute risk, which raises the gain from quitting. NICE’s model for patients with heart disease gave about twice the QALYs per quitter of its model for general inpatients (Mallender et al. 2013). So the population values may understate or overstate the lifetime benefit for these patients. The report shows how much of the lifetime benefit would need to be realised for the service to stay cost-effective.

The NHS savings are conservative against NICE’s own models. Godfrey et al. model four diseases from a disease-free start; NICE’s models also count the yearly costs of people living with heart disease and stroke, and give £3,300 to £3,900 per extra 12-month quitter in 2024/25 prices (Filby and Taylor 2018; Mallender et al. 2013), considerably more than this model’s. The report sets the model’s values per extra 12-month quitter beside NICE’s.

6.6.4 Timing and discounting

The lifetime values are present values at the age of quitting, so they are counted in the year the patient quits and discounted from there to 2026 like the costs. 05_build_lifetime_values.R builds them at the control panel’s discount_rate, which must be the same in every scenario, and records the rate and price year in the file; a run stops if they don’t match its own.

6.6.5 Life years and deaths within five years

Life years over a lifetime are not the same as deaths avoided in a period. The deaths avoided and life years gained within the five years of follow-up come from the effect of quitting on deaths after discharge (chapter 5, Tab_23), for these patients and their conditions. They are reported beside the lifetime results, never added.

6.6.6 The range in the results

Tier 2 is worked out twice in every run, with the central and the lower lifetime values (Tab_2, columns Value and Lower; Tab_3b, columns ending _Lower). The report and briefings show the lifetime results as the central value with the range from the lower to the central values in brackets. The lower and higher estimates used elsewhere change the effects of quitting on admissions, not who quits, so they leave the lifetime results unchanged.

6.6.7 ICER and net monetary benefit

The incremental cost-effectiveness ratio is the net system cost (system cost minus lifetime NHS savings) per QALY gained. The net monetary benefit values the QALYs at wtp_threshold (30,000 GBP) and takes off the net system cost. The service is judged against the range NICE uses for its guidelines, 20,000 to 30,000 GBP per QALY. (From April 2026 NICE’s technology appraisals use 25,000 to 35,000 GBP; that change applies to technology appraisals, not to guidelines (DIA Global Forum 2026).)

The model reports NICE’s valuation because these services are recommended in NICE guidance. HM Treasury’s Green Book values a QALY at about 70,000 GBP and discounts health at 1.5% a year rather than 3.5%; both would make the service look more cost-effective, so NICE’s valuation is the more demanding test.

6.6.8 Changing the lifetime assumptions

To change Edit Then
Life years gained by quitting at an age life_years_gained (and life_years_gained_lower) in user_inputs/lifetime_benefits_panel.csv Run 05_build_lifetime_values.R, then the model. 00_setup_and_run.R re-runs it when the panel is newer than its output
The age the quit is taken at for a band quit_age As above
When the years gained arrive gain_arrives_at_age As above
The share who would have quit later anyway quit_later_share (also applied to the NHS savings) As above
QALYs per life year qalys_per_life_year (and _lower) As above
The discount rate discount_rate in the control panel, the same for every scenario As above; the run stops until the lifetime values are rebuilt
The source altogether, for example values from a population model such as STAPM Write data/base/derived/lifetime_values_per_quitter.csv in the same layout: value_set (central, lower), age_group, sex, optionally imd_quintile, life_years, qalys, nhs_savings, discount_rate, price_year, source. Values per person who quits for life, as present values at the age of quitting The model reads it as it is; R/check_inputs.R checks it has every age band and sex (and IMD quintile, if given) for both value sets