Trying to make a difference, was it really planned?: the journey of a clinical researcher Trainees meeting 2015 Paul Little Professor of Primary Care Research University of Southampton
Jan 18, 2016
Trying to make a difference, was it really planned?: the
journey of a clinical researcher
Trainees meeting 2015
Paul Little Professor of Primary Care Research
University of Southampton
Really - me?
The story of research: the story of a researcher
Outline: Antibiotics for common infections– How come infections?….
– Asking a series of questions • From Nepal (descriptive epidemiology) • …through Ivan Illich (sociology) • ….trialling pragmatic strategies (trials methodology)• ….and 'Killer bugs’ (microbiology) • …to complex intervention development to change
behaviours (health psychology)
– Recurring theme: good consultation skills
The story of a researcher:
How come infections?
Not initially!
Lifestyle change in hypertension (Brighton; Lord Trafford )
Lord Trafford
Why infections?: serendipity
‘Behind every successful man is a surprised woman!’
What is the connection?
.…..Neil Weir (ENT) and the charity BRINOS
(British Nepal Otology Service;medical team of the year 2013)
BRINOS’s question?:
What is the prevalence, and main causes of ear disease and hearing impairment in Nepal?
Why Ear Disease in Nepal?• Deafness the biggest disability (WHO survey)
Early research: how difficult could research be?...
The inexperience of youth?
• BRINOS is/was a small charity – Offered £20,000 to do the study…..
• Research in developing countries is difficult • and I still knew nothing...(at least I knew I knew nothing!?)
– Liverpool School Tropical Medicine (LSTM)– Professor Newell
The inexperience of youth?
Professor Newell
Methods
• Screened all reporting ear or hearing problems: field audiometer, otoscopy– and sample of those with ‘no problem’
• Stratified random sample n=15,845– Eastern (wet) and MidWest (dry) regions– 3 areas in each: terai, hills, mountain
Map of Nepal
Nepal - findings
• 16.6% had hearing impairment• Mostly due to otitis media• 55% school age otitis media
• Traditional remedies prevalent• animal urine, leaves!!
• 61% with ear pathology had never been to health post – and when they did, often no antibiotics!
Back to the registrar year
Why infections in a developed
country?…
–The commonest symptoms:• impact NHS/society
sickness disability
–25-30% consult each year RTIs
–Very high expectations for antibiotics, most got antibiotics
‘I’ve got tonsillitis again doctor’
Did the evidence support antibiotics?
BUT..its Friday pm, you are running late…… would you say no antibiotics to these ladies?
Medicalising illness?
Sociology: Ivan Illich – Medical Nemesis and ‘medicalisation’
‘Modern medicine is a negation of health. It isn't organized to serve human health, but only itself, as an institution. It makes more people sick than it heals’
How important is medicalisation in
acute illness?
The potential problem with medicalisation: the iceberg
Self care
Pharmacy/NHS direct
General practice: 1:9
Secondary care: 1:3300
How to assess the importance of
medicalisation?
Sore throat trial
Open trial of prescribing strategies:–No offer of antibiotics– Immediate antibiotic prescription –Delayed prescription
The boss?...not again!?
A great mentor, a good environment!
Compete with training (LSHTM)
Develop (‘bad’) open trial methodology– ?Drug vs no drug: NO: prescribing strategies
– ?Placebo blinded: NO: Patients had to know;
• Structured support for placebo effect
– ?Outcome ‘objective’? (inspect;swabs;pills count etc):
NO: all medicalising, needed light follow-up
• validated diary informed by qualitative work
Main results sore throat trial (n=715)
010
2030
4050
6070
8090
100
% better satis belief Ab future
AntibioticNo antib.delayed
%
p<0.001p<0.001
Even one antibiotic prescription is strongly
medicalising
..fuelling reconsultations,antibiotic use…
Antibiotic prescriptions vs antibiotics usedFigure 1: Time trend in antibiotic prescribing to children in UK general practice 1993-2004 estimated from national prescribing data and the IMS GP prescribing database (1993=100)
0.0
20.0
40.0
60.0
80.0
100.0
120.0
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
IMS data
PPA data
Serendipity: patients’ perception of communication
• Satisfaction:– doctor dealing well
with concerns (chi square 362 kappa 0.79)
00.5
11.5
22.5
33.5
44.5
5
very not
z=3.3, p=0.001
satisfied
Durationdays
Back to antibiotics……
Do antibiotics work any better in
the other infections we see? (series of studies:OM/chest/conjunctivitis/sinus)
More studies…..:…….similar messages!
It is possible to get bored with the same message….
More studies OM/sinusitis/chest infections:
Do Abs help symptoms? : not much!
Evidence from RCTs, systematic reviews
prior duration
duration after seeing doctor
total duration untreated
Benefit from antibiotics
NNT
otitis media
1-2 days 3-5 days 4 days 8-12 hours 18
sore throat
3 days 5 days 8 days 12-18 hours 10-20
sinusitis 5 days 7-10 days 12-15 days 24 hours 13
bronchitis 10 days 10-12 days 20-22 days 24 hours 10-20
Increasing concern?:Headline News March 2012
• Resistance to antibiotics could bring "the end of modern medicine as we know it", WHO claim
………but is it our problem?
Is it our problem?: yes!• Medicine > Agriculture• 80% of (medical) antibiotics are prescribed in
primary care– most patients still get antibiotics!
• But does our prescribing really impact resistance?
Trends in Prescribing of Antibacterials in General Practice in England
© Copyright NHSBSA 2012
0
100
200
300
400
500
600
700
800
Apr.97-Mar.98
Apr.98-Mar.99
Apr.99-Mar.00
Apr.00-Mar.01
Apr.01-Mar.02
Apr.02-Mar.03
Apr.03-Mar.04
Apr.04-Mar.05
Apr.05-Mar.06
Apr.06-Mar07
Apr.07-Mar08
Apr.08-Mar09
Apr.09-Mar.10
Apr.10-Mar.11
Apr.11-Mar.12
Item
s pe
r 100
0 Pa
tient
s
Penicillins Tetracyclines MacrolidesCephalosporins Sulphonamides & trimethoprim QuinolonesMetronidazole & tinidazole All other antibacterial drugs
What is happening to primary care prescribing in England?: progress reversed 2004 onwards
7% increase
Trends in prescribing of antibacterials in General Practice in England
What is going on?...time for qualitative work
What do GPs think about resistance?
What are their key concerns?
GPs views of resistance? (Wood et al)
• recognise the importance of resistance
• BUT: ..not a problem in their practice!
• Blame hospitals/other prescribers!
GPs’ key concerns? (Kumar et al )
• Main concerns: severe symptoms, complications– Matches patient concern
• ad hoc targeting • pus, temperature,demographic, diet etc
Can we improve the evidence for better targeting of antibiotics
for bacterial infections?
…..and will it make any difference!
Flesh eating killer bugs! (=streptococci)
June 2014:
‘A long-serving and well-loved pastor has been killed by a flesh-eating bacteria’
(i.e. necrotising fasciitis)
Can we target strep. in sore throat?
• Options for targeting– Clinical score? e.g. Centor (pus, nodes, fever, no cough):
– Developed for Lancefield Group A
– RADTs? (rapid streptococcal antigen tests)– Group A Strep only
• What about non Group A? (C,G)• Major virulence factors/rates of septicaemia similar
Are C+G strep relevant?: probably!– Strep. in 34-40% (n=517;n=606)
• 25% C or G– similar clinical presentation to group A
Predicting A/C/G streptococci?
FeverPAIN (AUC 0.70):•*Fever last 24h• Pus•*Attend rapidly (<=3 days)•*severely Inflamed tonsils• No cough or coryza (i.e. pharyngeal
illness)
*=univariate and multivariate in both cohorts
Does better diagnosis/targetting (using a clinical score or RADTs)
lead to better outcome?
PRISM Trial
– Empirical delayed prescribing (control )– 5 item clinical score (FeverPAIN)
• 0-1 <20% strep (none), • 2-3 39% strep (delayed), • 4+ 63% strep (immediate)
– RADT • Similar but test for higher scores (3+)
Results: Delayed(control)
FeverPAIN RADT
Duration (moderately bad or worse Sx)
Median 5 days
HR 1.30* (1.03 to 1.63 )
HR 1.11(0.88 to 1.40)
Antibiotic use 75/164 (46%)
RR 0.71*(0.05 to 0.95)
RR 0.73*(0.52 to 0.98)
All models controlled for fever and symptom severity at baselineNo difference in returns within one month or following
So better diagnosis (targeting using FeverPAIN) improves symptom control and lowers
antibiotic use
• RADTs similar but no clear advantages to a clinical score alone.
Targeting:How common are complications
and can we predict them?
Does delayed prescribing prevent complications?
DESCARTE sore throat cohort
• n>13,000!
• Multi-centre collaboration: Friendship groups SW SAPC– trusted colleagues to build major multi-centre
collaborations – The social nature of research
DESCARTE
Results: Complications are uncommon
No antibiotics Antibiotics DelayedAntibiotics
Complications (total) 73/4536 (1.6%) 75/5750(1.3%) 16/1664 (1.0%)
Quinsy 11/4,536 (0.2%) 30/5750 (0.5%) 4/1,664 (0.2%)
Sinusitis 23/4,536(0.5%)
10/5750(0.2%) 2/1,664 (0.1%)
Otitis media
30/4,536(0.7%)
26/5750 (0.5%)
10/1,664 (0.6%)
Celluliltis/impetigo 10/4,536(0.2%)
9/5750 (0.2%) 0/1,664 (0.00%)
Can we predict complications in sore throat?: Not very well!
• Only two variables:– severe tonsillar inflammation (OR 1.92) – severe earache (OR 3.02)
• modest utility AUROC 0.61 (chance=0.5!)– 70% complications when neither variable present!
Delayed prescribing prevents complications as effectively as
immediate antibiotics.
No antibiotics Antibiotics DelayedAntibiotics
Adjusted RRs
StratifiedPropensity score(Multiple Imputation)
1.00 0.61(0.40;0.94)
0.55(0.31,0.98)
..and lowers reconsultations more effectively than immediate
antibiotics
Adjustment No antibiotics Antibiotics DelayedAntibiotics
Adjusted RRsAll control for clustering
StratifiedPropensity score(Multiple Imputation)
1.00 0.76(0.68;0.86)
0.58(0.49,0.67)
• So complications are uncommon, and we cannot very effectively predict them
• But if considering an antibiotic, consider delayed prescribing?– prevents complications, reduces reconsultations – at least as effective as immediate antibiotics.
DESCARTE sore throat cohort
Targeting in chest infections?:
The patient: ‘I’ve got green sputum doctor’
The doctor: ‘how do I know antibiotics won’t work for my particular patient?’
(green sputum, smoker etc)?
…….the overall data is modest (few RCTs) and not helpful for subgroups?
• 3012 adult patients with LRTI in 12 countries– acute cough (<28 d) main symptom
• or GP suspects acute bronchitis or pneumonia
• 2061 randomised: amoxycillin 1 gr TID or placebo
GRACE trial :
GRACE Network: Again friendships to build major
collaborations!
1. General practice Respiratory Infections Network (GRIN)
2. Flexibility; trust;
.....?requiring less dosh for the University?
Resolution of bad symptoms: overall data
Day 7-8 ‘survivor’
Log rank Hazard ratio P/NNT
Whole data set(n=1799)
0.465 vs 0.395
P=0.172 1.06
(0.96 to 1.18)
NNT 15
P=0.229
Resolution of moderately bad symptoms ( whole data set)
0.0
00.2
50.5
00.7
51.0
0
0 10 20 30analysis time
groupnumber = 0 groupnumber = 1
Kaplan-Meier survival estimates
time to resolution of moderately bad symptoms
What about my particular patient?Hazard ratio P
Interaction term?(p)
Smokersn=486
1.20(p=0.121)
1.23(1.01 to 1.50)
0.044
NNT 9
Age 60+n=550
0.86(p=0.166)
0.95(0.79 to 1.14)
0.555
NNT 143
Green Sputumn=346
1.28(p=0.059)
1.31(1.05 to 1.65)
0.019
NNT 8
Comorbid n=438
0.99(p=0.914)
1.06(0.86 to 1.31)
0.581
NNT 14
Green sputum subgroup0.0
00.2
50.5
00.7
51.0
0
0 10 20 30analysis time
groupnumber = 0 groupnumber = 1
Kaplan-Meier survival estimates
time to symptom resolution - green phlegm subgroup
Smokers0.0
00.2
50.5
00.7
51.0
0
0 10 20 30analysis time
groupnumber = 0 groupnumber = 1
Kaplan-Meier survival estimates
time to symptom resolution - current smoker subgroup
Benefits vs harms: side effects!
• Nausea, rash, or diarrhoea
Antibiotic 29%
Placebo 24%• NNH 21
Targetting?: 6 symptoms and signs predict consolidation
History (day 1)
Severe cough present
931 (33)
56 (40)
1.4 (1.0-2.0)
. 1.1 (0.7-1.6)
Phlegm present 2239 (79) 120 (86) 1.6 (1.0-2.6) N.A. Breathlessness present Severe breathlessness present
1594 (57) 197 (7)
96 (69) 17 (12)
1.7 (1.2-2.5) 1.9 (1.1-3.4)
1.4 (1.0-2.1) 1.3 (0.7-2.4)
0.025 0.419
Runny nose absent 807 (29) 61 (44) 2.0 (1.4-2.8) 1.9 (1.3-2.7) <0.001 Fever present 989 (35) 82 (59) 2.8 (2.0-3.9) N.A. Chest pain present Severe chest pain present
1304 (46) 141 (5)
80 (57) 13 (9)
1.6 (1.1-2.2) 2.1 (1.2-4.0)
1.2 (0.8-1.7) 1.5 (0.8-3.1)
0.402 0.224
Diarrhoea present 199 (7) 15 (11) 1.6 (0.9-1.8) 1.5 (0.8-1.8) 0.165 Physical examination (day 1) General toxicity 739 (26) 43 (31) 1.3 (0.9-1.8) 1.1 (0.7-1.6) 0.728 Diminished vesicular breathing 362 (13) 31 (22) 2.0 (1.3-3.1) 1.7 (1.1-2.6) 0.013 Crackles 264 (9) 44 (31) 5.3 (3.6-7.7) 3.5 (2.3-5.2) <0.001 Tachycardia (pulse >100 beats/min) 111 (4) 17 (12) 3.7 (2.2-6.5) 2.3 (1.3-4.3) 0.003 Tachypnoea (>24 breaths/min) 55 (2) 6 (4) 2.4 (1.0-5.7) 1.4 (0.9-2.0) 0.421 Blood pressure <90/60 mmHg 71 (3) 9 (6) 2.9 (1.4-5.9) N.A. Temperature >37.8°C 156 (6) 22 (16) 3.5 (2.1-5.7) 2.5 (1.4-4.4) <0.001
Simple risk stratification
• 2 history: Breathless, no coryza• 2 chest signs: bronchial, crackles• 2 vital signs: pulse >100, temp. >37.8
• 0 = 1% have consolidation• 1-2= 5% (most here…)• 3 = 20%
Will antibiotics work for my particular patient with a chest infection?:
• Modest benefits even in key clinical subgroups, modest disbenefits…
• Don’t prescribe for the vast majority! – Consider antibiotics/delayed antibiotics for
3+ key symptoms/signs?
So we need to be able to communicate effectively……
Can we improve communication, and will it help reduce antibiotic use?
Lack of time
Acute infection: a quick consultation?
How to change clinician prescribing behaviour?
GRACEINTRO (INternet TRaining for
antibiOtic use) Trial
Web based training: four groups
• No training
• Communication– enhanced communication training + booklet
• CRP training – cut points; kit demonstration
• Both: Communication and CRP
n=6771 baseline Post-intervention n=4264
Enhanced Communication/Information sharing
• Addressing the patients world– Concerns, – Expectations, – Attitudes
• Information exchange: booklet– Natural history; – Risks/benefits of antibiotics– Self-help – Safety netting
• Wrap up– Summarise– Check understanding, other concerns
We can communicate effectively and it makes an important difference:
RR(adjusted for patient variables)
p
Control 1.0
CRP 0.47 (0.35 to 0.64) <0.001
Communic’n 0.66 (0.50 to 0.85) <0.001
Both 0.39 (0.28 to 0.54) <0.001
Can we help symptoms with simple advice (PIPS study)?
PIPS study • Randomised strategies
– analgesic strategies
• Paracetamol vs ibuprofen vs combination;
– steam inhalation (factorial design)
Patients complied … BUT trivial differences in symptom severity overall
Pmol(control)
Ibuprofen Both
Whole cohort(743/889;84%)
1.67 +0.04(-0.11 to 0.19)
+0.11(-0.04 to 0.26)
• 10 symptoms: 0=no problem……6 as bad as it could be
Ibuprofen better in chest infections and for children
Ibuprofen interfering with the immune response?:
More reconsultations:same/new/worse symptoms
Pmol(control)
Ibuprofen Both
Reconsultation(same Sx, new Sx, or worse Sx)
35/300(12%)
58/295(20%)**
48/295(17%) 1
Adjusted RR
1 1.67 (1.12 to 2.38)
1.49(0.98 to 2.18)
** p=<=0.01 1 p=0.06
Complications higher
Pmol(control)
Ibuprofen Both
Complication 2/300 (0.6%)1 cellulitis1OM
11/295 (3.7%):1 Quinsy3 sinusitis1 meningitis1pneumonia5 OM(2 not new)
4/295 (1.4%):1 Quinsy2 sinusitis(1 not new)1 cerv. adenitis
Steam?– No benefit– Mild thermal injury in 4 patients (2%) who
returned full diaries• No reconsultations with scalding
We are probably doing more harm than good with widely given self help advice!
• ibuprofen little help overall– ? for chest infections and children. – BUT progression of symptoms/complications
• Advice to use steam does not help– and occasional harm
Can we prevent infections?
The PRIMIT trial of a web based behavioural intervention to reduce infection transmission
(in press)
Handwashing?
• Hand-washing widely advocated– e.g. H1N1 pandemic– but role of handwashing debated!
• No good randomised evidence among adults in our (resource rich) setting.
Developing a complex intervention
• Initial qualitative/questionnaire studies: – Confirm useful target behaviour (handwashing)
• determinants; barries/facilitators
• Further qualitative/questionnaire studies: – (‘think aloud’) for draft materials– tested key assumptions
• Randomised pilot of prototype website: – Changed behaviour (increased handwashing) .
Results: infections prevented 20,066 randomised
16,908 (84%) followed-up
Intervention Control p
Any RTI at 4 months 51% 59% <0.001
Any RTI (in household) 44% 49% <0.001
Transmission reduced to and from household members
Intervention Control p
Transmission to household
7.8% 9% <0.001
Transmission from household
6.8% 8.8% <0.001
Infections slightly less severe
Intervention Control p
Days more severe symptoms if RTI
4.1 days 4.3 days 0.008
Reduced consultations, reduced GI infections
Intervention Control p
Consultations for RTIs (notes)
18.9% 20% <0.001
GI infections
21.5% 25% <0.001
We can prevent infections!
• A free standing web intervention increases hand-washing – reduces infections, their severity, and transmission,
• ?Pandemic: will access internet for advice.
Had enough?
Have we learned anything useful?
• Antibiotic resistance: a public health problem we generate!• Antibiotic prescribing: medicalises, fuels demand/resistance• For symptoms: antibiotics overall/subgroups mostly not helpful
• A ‘bacterial’ score in sore throat helps symptoms, reduces antibiotics • Commonly given advice (steam/ibuprofen) is probably harmful!
• For complications:• Sore throat: Uncommon/difficult to predict:
– Good safety netting skills– if antibiotics are considered, consider delayed prescribing?
• Chest infections: basic clinical history/exam. help identify consolidation
• For prevention: a behavioural web handwashing intervention helps• prevents infections, reduces severity, and reduces transmission
Reflections on infection research• Acute infections are relatively quick/easy consultations
in a world of increasing demands…BUT:– a central public health role– a central role for better communication
• a little more time, BUT saves time in future!• brief training for experienced GPs helps
• Practice changing research:– Good mentorship and training – Large collaborations (friendship!)
• Powered for subgroups/adverse outcome
– Mixed methods, carefully developing complex interventions • really understand and change behaviours!
• …Is such research valued?
Having ideas and getting grants…..rejection?
Rejection
Don’t let it get you down ?Recycle or resubmit:
• If the referees points are answerable
Picking the ones to resubmit (otherwise they might get tired of
you!)
• BUT……..Keep going if it’s a good idea!
Why we don’t write enough grants? Perfectionism?
• The best is the enemy of the good…… Insufficient protected time
• Structural (other responsibilities; competing demands)
• distraction;faffing……. Failure?......expectations?..assume
2:3 or more to go down• Aim to be working on 2-3 ideas at any
one time……• Try and recycle……
Other excuses…..
• SO PUT ASIDE TIME IN THE DIARY• DON’T ANSWER EMAILS………….(AND DON’T HAVE A BEER BEFORE!)
Success? (5 S’s)
Space/time, Story (idea), Sharp (methods),Support (peers/team/social)Stamina
James McKenzie
‘For some years I went blundering on, gradually falling into a routine, i.e. giving some drug that seemed to act favourably on the patient, till I became dissatisfied with my work and resolved to try and improve my knowledge by more careful observation.’
Lack of timeWhat aspects of communication
are important to patients?:
Patient-centredness cohort
• Assessed patients’ perceptions of patient centredness:– develop questionnaire – assess domains empirically (factor analysis)
• Communication and Partnership• A personal Relationship• Health Promotion• A positive and Clear approach to the problem• Interest (of the doctor) in the effect of the illness on life
• Determine relationship of 5 domains to outcome (n=865; all conditions)
Stem (unless specified): The doctor ..... % agree Factorloading
Factor 1 communication and partnership
Was interested in my worries about the problem 80 0.68Was interested when I talked about my symptoms 93 0.80Was interested in what I wanted to know 86 0.67(Full question:) I felt encouraged to ask questions 80 0.54Was careful to explain the plan of treatment 80 0.58Was sympathetic 85 0.59
Was interested in what I thought the problem was 80 0.80Discussed and agreed together what the problem was 75 0.62Was interested in what I wanted done 75 0.67Was interested in what treatment I wanted 62 0.52Discussed and reached agreement with me on the plan of treatment 76 0.56
Results: Factor analysis of patients’ perceptions
Cronbach’s alpha 0.96
Relationship of domains to outcome: Satisfaction (MISS) : communication, positive
beta
Enablement: interest in life, health promotion, positive
beta
Symptom burden (MYMOP): positive, health promotion
beta
Results III• Referrals less if ‘personal’ relationship
– odds ratio 0.70 (0.54 to 0.90)
So are domains of communication important?
• There probably are distinct domains of patients perceptions of communication– Probably reliable– support the patient centred model
• Different domains => different outcomes (satisfaction, enablement,symptoms,referrals)– important for both patients and health service