Randomising every patient you see:Software for clinic seat research
Bruce Arroll
Douglas Kingsford
Antonio Fernando III
2010
Department of General Practice & Primary Health Care
Faculty of Medical & Health Science
University of Auckland, Auckland, New Zealand
The problem with RCTs
Expensive Exclude many of the patients we see
May not be the patients we usually see-restrictive Expensive Time consuming Huge amount of paperwork If paper based need data entry
How generalisable are they How to get a cheap computer generated randomisation
code to ensure concealed allocation.
McVickers paper
– Vickers AJ, Scardino PT. The clinically integrated randomised trial proposed novel method for conducting large trials at low cost. Trials (Biomed Central 5/3/2009)
Integrate RCT in to clinical practice Randomise every eligible patient Eligible if clinician uncertain-broad-crucial point Adverse effects, me too drugs; lifestyle Patients could enter own data eg adverse effects
Randomisation “daily”
20 starter packs of Varenicline (Champix)
Give them out haphazardly or systematically
“Pressured” my colleagues to be systematic Mailed a letter to 40 Maori smokers offering 4 free visits
with or without a starter pack of Champix 4/20 with starter pack replied and 0/20 in control Conclusion “don’t send out letters to Maori patients”
Without a starter pack
Better way would be phone call or face to face
Pediatric Oncology-the challenge
Every child with a malignancy in the (developed
world) is in a randomised controlled trial
Contributed to increase in survival for leukemia from about 30% to 90%
Generalisability RCT
300 patients from 250 GPs
300 consecutive patients from 3 GPs
Which has the most generalisability?
Office RCT
Internet based software
Make RCTs from office chair simple
Answer simple research questions as part of every day clinical work
Gratitude diaries 3 things grateful for and what did to cause them daily for
one week. Control group writes down early life memories. ? Then once per week
RCT done with internet sample had an effect up to 6 months later. CES-D 13 vs 10.5 Seligman ME, American psychologist 2005;60;5:410-21
Once per week enough Lyubomirsky S.The How of Happiness: Penguin
No trials in primary care Anecdote 90%+ patients happy to do them.
Research Question Does giving depressed patients a gratitude diary vs
control writing improve their PHQ outcomes at 4 (?) months
Inclusion criteria broad Any one with PHQ ≥ 10 (600 pts in 4000 patients) Excl Bipolar; severe drug/alcohol;dementia; personality
disorder; eating disorder; persisting psychotic illness; life expectancy < 2 years; patient unreliable at attending appointments
Research Issues ? Follow up as they come in versus attempt a formal
follow up
A sample size of 98 patients in each arm will be required to demonstrate a 2 point reduction on the PHQ from 13 to 11 with a standard deviation of 5
Issues Follow up and analysis
Stop trial at 4 months and do last value carried forward
Or
Stop trial at 4 months and take data as found (akin to Kaplan Meir)
– Analyse using random coefficients which would compare the gradient of the intervention group and the control group
Develop office based software Able to get randomisation code from clinic computer
Data gets stored on web No extra data entry required
Multiple doctors/nurses can submit data from different clinics-international collaboration possible
Privacy assured as each clinic uses own patient file identifier Simple entry criteria and simple outcome criteria Eg 600 patients with PHQ ≥ 10
Validated in primary care In NZ Arroll et al 2010 (not published)
Short Gives a “diagnosis” categorical Also continuous score to monitor improvement Free from June 2010 Currently been used in decision support in our district
health board therefore familiar and used
Why PHQ-9 and GAD-7
Data collected after randomisation-next slide
Previous psych contact and admissions asked after
randomisation
Done to simplify recruiting
Not part of primary outcome
Used for exploring for future studies
? Cause any bias
Spectrum
No gold standard ie post mortem won’t tell you if the
person is depressed.
Sadness →→→→→→→→→→Depression
↓
Consider Treatment if severe or
persistent
What is in paper form?
Consent form –need to be kept/scanned
Information sheet
Intervention instructions
PHQ
9 questions over past month
Score 10-14 mild major depression
Score 15-19 moderate major depression
Score 20+ severe major depression
Advantage a dichotomous score and a continuous scale
to monitor improvement
Analysis
Both ways i.e. all at 4 months –send out/phone for follow up At 4 months according how they have come in
– Using random coefficients analysis of slopes
Analyse all Analyse by PHQ 10-14 and >14
>14 more difficulty with functioning and may be less able to do Gratitude diaries
–