Increase response rate to a mobile phone text message sent to discharged patients from Chelsea & Westminster NHS Hospital Trust


Across all NHS Hospital Trusts in the UK, patients are given a survey to assess performance. This is commonly at the ward level, e.g., A&E, sexual health, etc. This survey is one question with a standardised scale. Each trust is evaluated on a, its score, and B, the percentage of people who actually respond to the survey (at the time hovering around 10-15%). It is clear that increasing the response rate is a task fit for her behavioural intervention, it is not appropriate to try to manipulate the score of the hospital – as this should be as neutral and as objective as possible.

There were two ways in which discharge patients were asked to respond: either a text message a day or two after they had been discharged, or a paper survey whilst they were in the process of being discharged. As a group we decided to focus on the text message, seeing as this was both at a scalable, and be the most common use of the survey aid trust.

So it became clear that we should focus on the txt messages, and increase the response rate from the existing 10-15%.

Research & insight

There are a range of techniques possible from our experience and the literature, such as social validation – as an injunctive rather than a descriptive as the majority of people weren’t doing the survey, reciprocity, requests for help, perceptual completeness (meaning explicitly messaging ‘to complete your stay’ to invoke desire to complete) and a pre-message to evoke consistency.


With the text message as was written it became clear only one sentence could be changed, as much of text was instructional and/or legal.

So we rewrote the key opening sentence in five different ways using different behavioural techniques, and wrote our test procedure to the highest controlled trial standard.

Participants were all patients discharged from 6 hospital wards in Chelsea & Westminster and West Middlesex Hospitals between December 2017 to February 2018 (N= 47,623 patients). The wards comprised of two emergency and outpatient departments, a sexual health clinic and a therapies department.

The design was between-subjects, in which each patient received one SMS version only. The independent variable was message type (with 5 levels), and the dependent variable was the response rate (i.e. ratio of responses to sends). We also look at patient’s actual FFT ratings.

All patients who were discharged from the specified wards received one version of SMS ­­– either the control message or one of the 4 new messages. The different message types were evenly distributed across patients in each ward.

We did not change any other aspect of the process typically undertaken when gathering data ­– only the content of the message varied. This meant that if changes to the response rate occurred, the type of message patients received could uniquely explain it.

Data for each ward were analysed separately. We first looked at overall trends in response rates, and then, for wards with adequate sample size (and statistical power), we conducted tests to determine the p-values, i.e. whether the trends were statistically meaningful.

  • Reciprocity & Liking SMS (‘The staff at Chelsea and Westminster Hospital who looked after you recently would benefit from your feedback’) consistently gave the highest response rates across all wards/hospitals compared to any other SMS tested

Emphasising the care and treatment that patients have recently received from medical staff appears to be most persuasive and effective strategy to motivate patients to respond. This trend was consistently established (across multiple ward specialties and in two different UK hospitals in one Trust), suggesting the Reciprocity & Liking text does have some promise in increasing responses rates in other Trusts.


32% increase in response

  • Tested on 47,623 patients
  • Strength of result high (p<0.001 compared to original)
  • Margin of error below 5% on larger wards, trend continued across smaller wards