COMMUNITY INTERVENTION TRIAL

1
COMMUNITY INTERVENTION TRIAL

• Evaluating interventions to promote population
  health

2
QUESTIONS

• How can we evaluate the effects of interventions
  and policies which target whole populations? Can
  we use the tool-box of epidemiology for this
  purpose?
• What are the main threats to validity of such
  evaluation studies? How can we make these studies
  as powerful as possible?
• What is the best strategy to build an
  evidence-base to underpin public health
  interventions and policies?

3
CONTENTS

• General typology of epidemiological research
  designs
• Community intervention trial definition,
  illustrative examples, relationship with other
  study designs
• Indications for community intervention trials
• Methodological aspects
• - the classical design example North Karelia
  project- main threats to validity- possible
  solutions

4
GENERAL TYPOLOGY OF EPIDEMIOLOGICAL STUDY DESIGNS
(1)

• Unit of observation individual
• - Experimental study
• - Observational study
• . follow-up study
• . case-control study
• . cross-sectional study

5
GENERAL TYPOLOGY OF EPIDEMIOLOGICAL STUDY DESIGNS
(2)

• Unit of observation group
• - Experimental study (community intervention
  trial)
• - Observational study
• . time series study
• . multiple group comparison study

6
COMMUNITY INTERVENTION TRIAL

• Experimental evaluation of an intervention with
  groups as units of observation
• Examples
• - effect of water fluoridation on dental
  caries
• - effect of fly control on diarrheal diseases
• - effect of vaccination on poliomyelitis
• - effect of health education on
  cardiovascular diseases

7
Number and per cent of children age 6-9 caries
free, Kingston and Newburgh, N.Y., 1954-1955
8
THREE-WEEK MOVING AVERAGE OF HIGH GRILL INDEX OF
TOTAL FLIES
9
Reported death rates per 1000 per annum under 2
years of age in Latin-American children of group
A and group B towns
10
RELATIONSHIP WITH OTHER STUDY-DESIGNS

• Within epidemiology Randomized Controlled Trial
  --gtCluster-Randomized Trial --gtCommunity
  Intervention Trial
• Within the social sciences Posttest-Only Control
  Group Design ?Pretest-Posttest Control Group
  Design

11
INDICATIONS FOR COMMUNITY INTERVENTION TRIALS

• Intervention is aimed at group (community)
  instead of at individuals
• Possible reasons (often overlapping)
• - it can only be done that way
• e.g. fly control
• - it is more (cost)-effective that way
• e.g. fluoridation of drinking water supply
  mass media health education
• - it simply works out that way
• e.g. vaccination against poliomyelitis using
  Sabin vaccine
• No control persons within same community left!

12
METHODOLOGICAL ASPECTS (1)

• Classical design
• - one or more experimental communities
• one or more control communities
• usually small number
• - usually without randomization
• experimental community given
• control community selected to be comparable

13
METHODOLOGICAL ASPECTS (2)

• Classical design
• - before-and-after comparisons
• outcome measures
• process measures
• - using registries
• and/or separate cross-sectional samples

14
METHODOLOGICAL ASPECTS (3)

• Example North Karelia project
• - objective to assess the effectiveness of a
  health education campaign using (a.o.) mass
  media
• - North Karelia local people asked for
  governmental action to counter high
  prevalence of heart disease
• - control community neighbouring county with
  many characteristics in common, Kuopio
• - more than 10 years observation
• . repeated surveys of risk factor levels
• . registries (CHD mortality)

15
METHODOLOGICAL ASPECTS (4)

• Results risk factors, mortality
• Interpretation
• - why did Kuopio county improve so much?
• - catching up of North Karelia?

16
Main target health behaviours and risk factors in
North Karelia (NK) and the reference area Kuopio
(REF) according to cross-sectional population
surveys in 1972, 1977, and 1982
17
Percentage decline in CHD mortality of 35-64
year-old men in North Karelia and the rest of
Finland
18
MAIN THREATS TO VALIDITY (1)

• Comparability of experimental and control groups
  is frequently suboptimal
• - no randomization
• self-selection of experimental community is
  common
• matching of control community at best only
  on base-line measurements- small number of
  observation units
• statistical control for incomparability not
  feasible

19
MAIN THREATS TO VALIDITY (2)

• Comparability of measurements is frequently
  suboptimal (information bias)
• - not double-/triple/blind
• the population knows (survey data!)
• the interveners know (registry data!)
• the evaluators know

20
MAIN THREATS TO VALIDITY (3)

• Link between exposure and effect not obvious
  (causal inference problematic)
• - cross-sectional surveys do not permit
  assessment of change in response to exposure
  to the intervention
• Contamination of control communities with
  spill-over of intervention programme is frequent
• - the community is not a laboratory, or a
  health care institution, in which one can
  effectively control the application of an
  intervention

21
RECENT DESIGN INNOVATIONS (1)

• Increase comparability of experimental and
  control groups
• - create possibility of randomization
  simplify intervention and measurements, so
  that number of communities can substantially be
  increased
• - if randomization is impossible, give highest
  priority to matching on relevant base-line
  measurements
• - prevent self-selection of experimental
  communities

22
RECENT DESIGN INNOVATIONS (2)

• Increase opportunities for statistical control of
  imcomparability between experimental and control
  groups
• - larger number of (control) communities
• - assess trends before intervention

23
RECENT DESIGN INNOVATIONS (3)

• Increase comparability of measurements
• - let surveys be held by independent
  contractors, and do not disclose purpose
  either explicitly or implicitly (e.g. ask for
  intervention exposure after assessment of
  endpoints)
• - monitor changes in registration
• - do blinded analysis

24
RECENT DESIGN INNOVATIONS (4)

• Strengthen causal inference through supplementary
  data and analyses
• - add longitudinal (cohort) measurements
  measure degree of exposure to intervention,
  relate this to outcomes- trend analysis
  establish correct timing of presumed effect
• Prevent contamination
• - choose communities far part

25
CONCLUSIONS

• The Community Intervention Trial is a powerful
  design for evaluating the effect of
  population-wide interventions
• The population is not a clinic, let alone a
  laboratory, and control over the intervention and
  its evaluation is often suboptimal
• In view of the large resources needed to
  implement a Community Intervention Trial, this
  should be reserved for situations in which the
  intervention can reasonably be expected to work
  (i.e., separate components have already been
  evaluated properly)