Impact smoking questions (#388)

Co-authored-by: Ralph <ralph.hawkins1@nhs.net>

Author: selina-dxw

app/lung-health-check/2026/02/impact-smoking-questions/Cigarettes-highest.png

@@ -0,0 +1,64 @@
+---
+title: Balancing clinical accuracy with keeping our users engaged  
+description: Understanding the impact of our questions on users and on the lung cancer screening programme.
+date: 2026-02-19
+---
+This is the second post in a 3-part series about our approach to designing questions asking participants about their smoking history. 
+
+We are designing a new digital service for the lung cancer screening programme. The service will ask eligible participants questions about their medical history and lifestyle to work out their risk of getting lung cancer in the next 5 years. If a participant is considered at higher risk of lung cancer they will be offered a CT scan. This service is currently delivered by phone and is called the [lung health check](https://www.nhs.uk/tests-and-treatments/lung-cancer-screening/). We are exploring whether a digital offering would be desirable to users and clinically safe.  
+
+## Understanding the impact of our questions on users 
+
+For the calculation, we need to know the total number of years someone has smoked tobacco. And the average number of cigarettes they smoke or used to smoke a day. The information we need may seem simple but we know these questions can have an emotional impact on our users. Someone’s smoking history can be very personal and mixed with feelings of shame and anxiety.  
+
+One participant told us during a research interview: “I’m always fearful of it, and I know I smoked for almost 20 years or over, the damage was done, I put junk into my body, the damage was done.” 
+
+We mapped the possible impact of different ways of asking users questions about their smoking history. We need to make sure that the questions we ask users don’t make someone feel judged or anxious. Not only for the obvious reason of not causing our users harm. But also because we want to keep our users engaged with our service, so they complete their lung cancer screening and can be offered a CT scan if they need one.  
+
+We considered asking users: 
+
+- the highest amount of tobacco they smoked 
+- minimal details about their smoking history 
+- for an average amount of tobacco smoked 
+
+We first tested designs that only asked users for the highest amount of tobacco they smoked, because this is what was included in the phone scripts for the current service. These designs didn’t work for users. Some users shared that they would reduce the number because they felt it didn’t accurately reflect their smoking history.  
+
+Someone might have smoked 20 cigarettes a day for several years, tried to quit a few times, and then reduced the amount they smoke. The question could feel judgemental to someone who’s struggled to quit and worked hard to reduce the amount they smoke. It didn’t acknowledge the effort they made trying to stop smoking.
+
+![Design that asks users 'what is the highest number of cigarettes you have smoked per day on a regular basis?'](Cigarettes-highest.png 'Our first design asking users for the highest number of cigarettes they smoked on a regular basis') 
+
+If users felt like they couldn’t share an accurate history it brought into question how accurate their result would be. This risked undermining a participant’s trust in our service, and possibly whether they would complete it. As a result, one of our user needs for the service is:  
+As a user of the service, I need to be able to explain my unique health context, so that I feel accurately understood. 
+
+Additionally, small inaccuracies in a participant’s history might not pose a significant clinical risk to that individual. But when we scale the service nationally, they could have a significant impact at a population level.  
+
+We decided not to test questions that asked users for an average amount of tobacco they smoked because we don’t think this would be suitable for all users. Some users might find it difficult to work out the average amount of tobacco they smoked. NHS services should be designed for everyone, including people with [low health literacy](https://service-manual.nhs.uk/content/health-literacy). We can’t easily explain to users how they can work out an average amount of tobacco they smoked in their lifetime. This was particularly relevant to users with the first mindset: I want to do the right thing, just don’t make me feel judged or confused. 
+
+## Quantifying smoking is an imperfect science 
+
+We learnt from our subject matter experts that trying to quantify the amount of tobacco someone has smoked is an imperfect science. The information we need for the calculation needs to be good enough to give an indication of someone’s level of risk. But it will never be perfect.  
+
+We also know that people cannot always recall their smoking history accurately. During our pop-up research with shisha smokers, we spoke to people who were smoking shisha alone, but in our questionnaire they would say they only ever smoke in a group.  
+
+How a participant interprets questions is tied to their mental model of smoking. A user’s mental model is based on their individual background and past experiences. It can change how someone identifies. For example, as someone who ‘used to smoke’, or ‘quit smoking but still has the occasional cigarette’. A participant could answer questions honestly from memory, and still not be giving an accurate answer.  
+
+We need to get the balance right between: 
+
+- asking questions that give us enough information to calculate an accurate result 
+- allowing users to feel like they’ve told us enough information so they trust the result we give them 
+- not overwhelming users by asking too many questions 
+- not asking users to recall too many specific details about their smoking history
+
+During our pilot we are asking participants to complete their phone appointment after testing our digital service. We will then compare their responses to help us understand if a digital service is viable. We don’t expect their responses to match. And we know that a participant's responses to the smoking questions are likely to be different in our service from their phone call. What we’ll need to consider is our level of tolerance for differences between the 2 services.  
+
+Although the phone service has the benefit of call handlers clarifying questions with participants, this doesn’t necessarily mean that it's more accurate. One thing we’re anticipating with a digital service is that participants might feel able to be more honest when they are answering a questionnaire in private, rather than telling another person. 
+
+## The impact of our questions on the programme 
+
+If the way we ask someone about their smoking history consistently over-estimates the amount of tobacco they smoked in their lifetime, we could end up over screening the population. That is, giving a higher number of participants a CT scan when it might not have been necessary. The risk from the low-dose radiation exposure is low to an individual participant. But the risk of exposing participants at a population level means it needs to be considered by the programme. 
+
+The opposite could also be true. If we consistently under-estimate the amount of tobacco someone smoked in their lifetime, we could under screen the population. We could end up not offering someone a CT scan when they would have benefitted from it. For example, if someone is diagnosed with lung cancer, when it could’ve been diagnosed by a radiologist during lung cancer screening. This could undermine trust in the screening programme.  
+
+Finally, someone could stop engaging with the lung cancer screening programme altogether if they feel like our service isn’t worth their time. Or if they don’t understand the benefit of completing the questionnaire. Someone might also withdraw from the service if it makes them feel judged, or if the questions are confusing.  
+
+Our next design history shares how we iterated our designs based on these insights.