1. The use of regular text messaging over one year to collect primary outcome data in an RCT
Reuben Ogollah1,2, Martyn Lewis1,2, Kika Konstantinou1 Sarah Lawton2, Jamie Garner2, Nadine Foster1,2
1Arthritis Research UK Primary Care Centre, Research Institute for Primary Care and Health Sciences, David Weatherall Building, Keele University, Staffordshire, ST5 5BG
2Keele Clinical Trials Unit, David Weatherall Building, Keele University, Staffordshire, ST5 5BG

2. RCTs with time-to-event outcome data
The primary endpoint of interest is time-to-event or failure time
Ideally individuals are observed until event has occurred
Frequent repeated data collection needed to accurately capture the event of interest with reduced risk of recall bias
A ubiquitous and inevitable problem of missing data

3. Time-to-event outcome data
Loss to follow-up
Study end
Censored
Lost to follow-up: event time censored
Time (months)

4. Minimising dropout at design stage
Postal or electronically-sent questionnaire surveys are not ideal for collecting frequent data
Often yield poor response rates even after several reminder mailings
Alternative: short message service (SMS) to obtain prospective, real-time data

5. Objectives
Describe the use of SMS to obtain weekly primary outcome data
Describe the response patterns to SMS
Implications for data analysis

6. Study design (the SCOPiC trial)
Usual, non-stratified care
No systematic subgrouping or targeting treatment
Stratified care
Group 1: Advice and support to self-manage
Group 2: Physiotherapist-led treatment
Group 3: Fast-tracked, with MRI, to spinal specialists
470 adults with sciatica/suspected sciatica from ~30 GP practices

7. Primary outcome Total followed up to date, n= 422 (target, n= 470)
“Compared to how you were at the SCOPiC clinic X weeks/months ago, how are your back and leg symptoms today?”
Completely recovered
Much better
Better
Same/ no change
Worse
Much worse
Total followed up to date, n= 422 (target, n= 470)
90% (n=380)
10% (n=42)

8. Participant receive postcard on Thursday
Primary outcome
Compared to how you were at the SCOPiC clinic X weeks/months ago, how are your back and leg symptoms today?”
Completely recovered
Much better
Better
Same/ no change
Worse
Much worse
First weekly SMS sent on Sunday following the SCOPIC clinic
Non-responders after 48 hours sent SMS reminders (Tuesday)
Non-responders receive post-card reminders after 48 hours (mailed Wednesday 1st class)
Participant responds
7 days
Second weekly SMS to collect outcome data (2nd Sunday)
Participant receive postcard on Thursday
Weekly for first 4 months
Monthly until recovered, or up to 12 months

9. Weekly/ monthly response rates
7,125 out of 7,975 (89%) valid responses received
78% Phone call
91% SMS
[P<0.001]
Pattern of non-response: both intermittent and dropout

10. Weekly/ monthly response rates
All participants (n=422; 380 SMS, 42 phone calls)
Percentage of complete responses

11. Weekly response rates
Median (IQR) weekly response rates for the first 16 weeks
SMS
Phone call

12. Implications for data analysis
High response rate: reduced amount of missing outcome data
Increased power
Minimise bias
Improve generalizability of the results
Strengthen intention to treat principle - data from all randomised participants utilised
Plausible non-informative censoring (ignorable missing) and reduced problems around interval censoring
Non-informative censoring: the possibly unknown true time to the event for a patient is the same regardless of whether or not it is actually observed (or whether censoring occurs or not prior to it).

13. Conclusions
Collecting frequent follow-up outcome data with SMS is feasible in a RCT
High response rate (>90%) for weekly text data collection very promising
This could be an additional and/or alternative strategy for collecting regular primary outcome data (especially time-to-event outcome) in large pragmatic trials
Future work: build a dynamic workflow of questions, presenting the next item based on previous responses
Response modifies the next question
Alternative when you ask short key outcomes

14. Acknowledgments
This study is funded by the National Institute for Health Research Health Technology Assessment Programme (NIHR HTA project number 12/201/09) and will be published in full in Health Technology Assessment. Nadine Foster is supported through an NIHR Research Professorship (NIHR-RP ) and NIHR Senior Investigator award. Kika Konstantinou is supported through a HEFCE Senior Clinical Lecturer award.  The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the HTA programme, NIHR, NHS or the Department of Health.

15. Keele Clinical Trials Unit David Wetherall Building Keele University Newcaslte-under-Lyme ST5 5BG
Tel: Fax: ​

16. Weekly/ monthly response rates
Those who have been followed up for at least 16 weeks (n=359; 321 SMS, 38 phone calls)

17. Weekly/ monthly response rates
···
W16 M5
M10
M11
M12 X
100% • Y% Z%
7,125 out of 7,975 (89%) valid responses received
78% Phone call, 91% SMS
Predominantly intermittent pattern of non-response

18. Weekly/ monthly response rates
Median (IQR) monthly response rates from months 5 to 12