Leave a comment

6.8 Community-based intervention trials in developed countries

6.8 Community-based intervention trials in developed countries
Oxford Textbook of Public Health

Community-based intervention trials in developed countries

H. Hoffmeister and G. B. M. Mensink

Health promotion of large populations

Strategies to enhance the population’s health

Determining the impact of community intervention
Rationale of community intervention trials
Objectives of community intervention trials

Improvement of health knowledge, attitudes, and behaviour

Societal conditions

Risk factors, early signs, and symptoms

Morbidity and mortality rates

Early estimate of mortality outcome: the multiple logistic function
Intervention in communities

Local communities as the field of intervention activities

Theories and frameworks focused on community intervention trials
Study design

The choice of intervention and reference population

Survey sampling

Problems with secular trends

Efforts to detect weak intervention effects
Evaluation of changes

Interview instruments

Analytical procedures

Morbidity and mortality estimates

Mortality evaluation

Statistical procedures
Chapter References

Health promotion of large populations
Community health in industrial societies is most affected by a few chronic diseases, in particular cardiovascular diseases, cancer, adult-onset diabetes, arthropathies, and chronic diseases of the respiratory tract, liver, and gastrointestinal tract. These diseases develop over a long period of time without severely influencing a person’s quality of life. In the early stages no symptoms appear and the subjective health status is not necessarily altered because the body can compensate for the effects of the disease with physiological and metabolic changes.
Despite the role of age, genetic predisposition, and possible influence of micro-organisms, the risk of developing one of these diseases early in life depends considerably on an individual’s health behaviour. Habits like smoking, alcohol abuse, high fat consumption, physical inactivity, incorrect posture, and lack of hygiene are likely to raise risk factor levels (for example, increased blood lipid levels, high blood pressure, overweight, atherosclerotic lesions of blood vessels, musculoskeletal disorders, chronic inflammations, and infections). These risks play an important role in the initiation and progression of severe chronic diseases.
Strategies to enhance the population’s health
Preventive medicine uses two approaches to reduce the population’s risk of contracting one of these diseases.
Medical or high-risk approach
In this approach members of the medical profession screen for elevated risk factors and primarily treat individuals with high risk. The identification of high-risk individuals, their consequent medical treatment, and repeated advice to change their lifestyle is one way to enhance the individual’s health and thereby the health of the population. Several controlled intervention studies and experimental studies have shown significant reductions in risk factors as described below. Successes have been particularly high in the reduction of high blood pressure and in the treatment of hypercholesterolaemia, disorders of carbohydrate metabolism, and subsequent cardiovascular diseases. Currently, several projects are being conducted in which people are screening for prevailing infections, for example Helicobacter pylori, hepatitis B and C viruses, and Chlamydia pneumoniae, and subsequently given treatment to validate the impact on long-term outcomes like cancer, liver cirrhosis, and cardiovascular disease. The high-risk approach has, however, limitations and disadvantages. The previously mentioned chronic diseases are also prevalent among many people without elevated risk factor levels and early medical indications. Screening, drug treatment, and individual health counselling are expensive and, in addition, serious undesired side-effects of drug treatment have been observed.
Public health approach
The public health approach is directed towards general populations in communities, regions, or whole countries rather than to individuals. It is concerned with creating a healthy lifestyle, convincing the population to avoid health risks, and teaching skills to lower or avoid drug abuse and medical consumption and to cope with crucial life events. This public approach tries to enhance the knowledge about risky and preferable health behaviour as well as to change unhealthy attitudes and beliefs. If necessary it may also aim to change the community environment. It is the public health answer in the fight against widespread diseases and the promotion of the population’s health. Various uncontrolled activities and programmes are carried out in communities, on a regional and state level, using different ways to increase health knowledge and to influence attitudes and health behaviour of community inhabitants. While it can be assumed that none of these activities have adverse effects, it is not clear whether they really work. Sufficient knowledge about the most effective intervention strategies is missing and often the cost-effectiveness has not been taken into account appropriately.
Determining the impact of community intervention
Scientific evidence is needed to demonstrate that the community approach to promote public health is broadly effective. The different aspects and elements of underlying theories (such as social cognition theory and persuasive communication theory) must be tested in appropriate studies and evaluated with respect to their health impact. The evaluation should also be used as a feedback for the development of lifestyle intervention methods. This is a difficult and expensive scientific task. These aspects of community health are so complex that it is difficult to find a general definition or an agreement about a set of indicators to measure them appropriately. Changes in knowledge—for instance about the unhealthy effects of smoking; good dental hygiene; or attitudes, beliefs, and health behaviour (such as jogging, safer sex, or increased vegetable consumption)—can be observed early by ‘process evaluation’. Such changes, however, do not provide certainty of health improvement. The process of determining the physically measurable outcomes of intervention is complicated by the fact that there will be a time delay in changes in risk factor levels and disease prevalence. Changes in morbidity and mortality rates are usually not detectable until several years later.
During the last 25 years a few large community-oriented intervention studies have been conducted in the United States and Europe (Table 1). The objective of these studies was to reduce the occurrence of cardiovascular diseases and other chronic diseases in large populations by means of primary prevention. These studies represent a logical consequence of the modern understanding of non-communicable disease patterns and their development in industrial societies. Nevertheless, these cardiovascular disease intervention studies report only minor changes in indicators of risk, morbidity, and mortality. In addition, the outcomes are inconsistent although intervention measures were more or less similar. Even the well-designed studies, performed with massive intervention effort, show only limited success. The somewhat disappointing results are not necessarily due to a general failure of community-based interventions. Methodological difficulties in design or analysis, of which the researchers were originally unaware, may account for the lack of success. This can be derived from a meta-analysis conducted with published cardiovascular intervention studies, fulfilling the criteria of comparability (Sellers et al. 1997). To a certain extent the variability in outcome can be subscribed to different evaluation characteristics like length of follow-up time, response rates, matching of intervention and reference population, and adjustment for covariates.

Table 1 Major community trials on cardiovascular disease prevention

In a minor way, the variability of the study results in this meta-analysis was explained by differences in intervention measures. A major critical point for these intervention studies may be that the effects on risk factors are still too weak, thus explaining the lack of impact of these studies concerning morbidity and mortality. Furthermore, even if results are promising after the first phase of intervention, long-term compliance is necessary to achieve substantial effects on disease endpoints. The efforts to achieve prolonged changes in health behaviour may have been insufficient. More understanding of which interventions work and research on the most effective and efficient ways to change the population’s health behaviour—especially in the fields of smoking cessation, promoting healthy eating habits (and provision of high-quality foods), and enhancing physical activity in the general population—will help to improve the outcome of future interventions.
In recent years some well-conducted community intervention studies have shown the advantages of this approach. For example, projects concerning HIV infection/AIDS prevention have been successful, at least in Western community intervention projects, in limiting the epidemic (CDC 1999).
The dietary supplementation of communities in general or of high-risk regions with deficiencies in trace elements, minerals, or vitamins is also a possibility. In the 1970s for example, the World Health Organization (WHO) and other national health boards recommended the fluoridation of drinking water. Studies conducted in several countries showed lower rates of dental caries after supplementation of fluoride in drinking water compared with reference areas. Partly because of perceived unwanted side-effects, enforced supplementation of this kind is unacceptable in many countries. However, fluoride supplements for treatment in childhood, nutrient-enriched products in regions suffering from micronutrient deficiencies (such as iodine-enriched salt in Western Europe, and selenium-enriched bread in Finland), and folate supplements to be taken before and during pregnancy are commonly available and can be recommended in community health promotion programmes.
Particularly in developing countries, community-based nutrition intervention projects can be successful in the fight against malnutrition, if the underlying problem is identified and understood, and goals and objectives are clearly set. The impact is enhanced if the project concentrates on specific target group like pregnant or breast-feeding mothers.
The final evaluation of an intervention trial should not only focus on health outcome variables but should also include issues such as social and structural characteristics of the intervention communities which strongly determine the outcome of intervention. The intensity and density of the intervention should be optimized to meet the specific goals. The study design should also guarantee enough statistical power to evaluate the achievement of these goals. It is important that the natural variation in health indicators within the observed communities should be estimated before initiation of the trial. The main concerns in the design and the evaluation of community intervention trials are discussed below.
Rationale of community intervention trials
Health promotion within the structure of local communities or regions seems to be a good strategy to reduce common diseases. It is hypothesized that this approach has advantages compared with both individual treatment and to large national programmes (which often lack components that fit the individual). In this context several statements have to be considered.

Most people in a community need greater and more specific knowledge about health issues before they are likely to change unhealthy behaviour. In addition, widespread attitudes and beliefs incongruent with evidence-based knowledge have to be corrected.

A change to a healthier lifestyle is beneficial, not only for people at high risk but also for all inhabitants of a community. It may also improve life quality and satisfaction.

In Western societies most adults have one or more elevated risk factors for cardiovascular (and other chronic) diseases. A reduction of their risk factor levels is likely to have a large impact on the health status of populations.

Programmes and activities to promote health in relation to particular disease outcomes will not only reduce the risk for the intended diseases but may also affect many other disease outcomes. For example, healthy lifestyle campaigns directed towards smoking cessation, increased vegetable consumption, and physical activity are useful in preventing many non-communicable diseases (such as cardiovascular diseases, major forms of cancer, adult-onset diabetes, and the chronic diseases of the respiratory and gastrointestinal tract).

An action centre implemented in the community which initiates, conducts, and co-ordinates health-oriented programmes is an important tool to improve the health status of a community.

Structures already existing within communities such as schools, sports grounds, clubs, and health facilities, and community leaders will facilitate health promotion and preventive measures and make them more cost-effective compared with health promotion solely focused on individuals.

An individual is usually influenced by the community in which he or she lives. Health-improving activities within the community will encourage personal involvement in health-related issues.

A wider acceptance of health programmes and projects supported by local opinion leaders will enhance confidence in the benefits of such activities and will make it easier for individuals to accept and use them.
Objectives of community intervention trials
The typical conditions of social life and the environment in a community often determine the objectives and the likelihood for success of community intervention. The overall health in such a community is influenced by many aspects such as knowledge, awareness, and behaviour of the individuals, health facilities and provisions within the communities (health-oriented activity groups, hospitals, sports grounds, local media, and so on), and the specific risk profile of the community (regional eating habits and leisure time behaviour, mean risk factor levels, specific morbidity and mortality). Since these aspects interact with each other, the community intervention measures should try to influence all relevant aspects to improve health.
Within the observed community a positive preventive atmosphere has to be created. This process of change in particular health aspects should be studied in intervention trials (in subsequent time periods). In community intervention trials, this ‘process evaluation’ was seldom used, although it could give a deeper insight into the influences of intervention on the process and determinants of health changes. This cannot be obtained sufficiently solely from measuring the outcome. The achievement of the objectives can be observed by specific outcome measurements, which reflect the changes of various health aspects (Fig. 1).

Fig. 1 Elements, pathways, and outcomes of community intervention.

Improvement of health knowledge, attitudes, and behaviour
Instead of concentrating on an individual’s behaviour, community intervention is concerned with the manipulation of health knowledge, health attitudes, and health behaviour in whole communities. Thus, an important goal of community intervention is to achieve a favourable change in these health aspects in the whole population or large groups within the population.
A primary target of community intervention studies is a verifiable improvement of health knowledge in the community which should be maintained for a long period of time. This does not guarantee improved health, but it supports all the other efforts to achieve better health behaviour. A change in health knowledge within the community is one of the earliest detectable effects of community intervention. Health-oriented lessons in schools are very effective in this respect. They may not only influence the children but also their families.
Changes in attitudes and beliefs are further aims of community intervention. Attitudes and beliefs are not always in accordance with the prevailing knowledge. Although many attitudes may be based on particular knowledge, this is not a necessary condition for health attitudes and beliefs. A community may have developed certain health attitudes (like ‘too much coffee is bad for your heart’) without knowing why the behaviour is unhealthy (it may increase serum cholesterol levels). On the contrary, an increase in knowledge does not always change a person’s attitude. Although a certain behaviour (like smoking) may be generally accepted as harmful, it is not necessarily regarded as a risk for oneself (‘My grandfather smoked his whole life and lived until 90’). Changes in attitudes and beliefs can also be observed as early effects in intervention populations.
The objective which is most difficult to achieve via intervention actions and programmes is to gain significant influence on the health behaviour of a large part of the community. Although a change in knowledge and attitude is possible, past experience of community intervention trials shows that changes in behaviour with high impact on health (such as smoking, nutritional habits, and physical activity) are difficult to achieve and are even more difficult to maintain over a long period of time.
Changes in behaviour are difficult to achieve even in communities with a well-established infrastructure to enhance and maintain health. This becomes obvious by looking at smoking behaviour. The majority of people are aware that smoking is the most dangerous single health risk of industrialized societies. Even so, a large number of people in these societies smoke. This discrepancy between health knowledge and attitudes was also observed in a study among German schoolchildren. All of the children who smoked on a regular basis knew about the main diseases caused by smoking. But despite this knowledge, 20 per cent of them wanted to continue smoking. The 80 per cent saying that they would like to stop or reduce smoking are more susceptible to programmes, courses, and other community activities concerning smoking cessation.
Societal conditions
An improvement of health knowledge, attitudes, and beliefs does not inevitably lead to changes in behaviour. These changes are likely to occur if the community’s health infrastructure is appropriate (that is, enough health-related information sources, courses to stop smoking or other drug consumption, availability of fresh vegetables, low-fat foods, sports grounds, and so on). An improvement in an insufficient infrastructure should be an additional target of intervention. Societal characteristics of a community should therefore be analysed and taken into account before starting a community intervention project. In Germany, there has been a strong growth in self-help groups during the last 30 years. This was initiated by so-called coronary heart groups for patients after myocardial infarction. They started lifelong endurance training and organized healthy lifestyle programmes. Local groups are now concerned with cardiovascular prevention, primary prevention therapies against rheumatic disease, therapies for diabetes, and rehabilitation of specific cancers—these groups meet regularly in many communities. A validation of the effectiveness and efficiency of these forms of community intervention has not been performed.
Risk factors, early signs, and symptoms
Altered behaviour should lead to a measurable change in physical risk factors. Changes in risk factors like obesity, high blood pressure, hypercholesterolaemia, specific liver enzyme levels (as indicators of alcohol abuse), and high resting heart rate (as an indicator of poor physical fitness) can give important indications of the success of intervention. Achieving favourable changes in risk factors is the most common and verifiable goal of community intervention. It is assumed that a favourable change of risk levels or of early signs and symptoms of diseases will improve the health of a population. A community intervention trial is needed to prove this assumption.
The interdependency of risk factor changes, and sometimes contrasting impacts on different disease outcomes, complicates the determination of overall success. An improvement of certain risk factors may induce an undesirable trend in others. Several studies observed an increase in body mass index among people who stopped smoking. Measures to reduce a risk factor may not be beneficial for the prevention of all kinds of disease. Individuals with high levels of cardiovascular risk factors have an increased risk of developing coronary heart disease and premature death. Replacing a high amount of saturated fatty acids by polyunsaturated fatty acids in the diet leads to lowered serum cholesterol levels. This is a favourable change concerning the development of atherosclerosis and cardiovascular diseases, but some studies suggest that there may be an increased risk of specific cancers among individuals who have consumed a high amount of polyunsaturated fatty acids (Hursting et al. 1990; Staessen et al. 1997; Veierod et al. 1997).
It is observed in many studies that people consuming moderate amounts of alcohol have a lower risk for cardiovascular disease than teetotallers and heavy drinkers. This observation is biologically plausible because alcohol intake enhances serum high-density lipoprotein cholesterol and lowers the risk for thrombosis. Even total mortality rates are lower in moderate drinkers compared with teetotallers or heavy drinkers (Hoffmeister et al. 1999; Liao et al. 2000), although alcohol consumption increases the risk for certain cancers, liver diseases, some other diseases, and traffic accidents. An intervention which recommends total abstinence may therefore not be appropriate. It is possible that societies with very restricted alcohol rules suffer more from alcoholism than those with a liberal attitude.
Several hypotheses about causal links between certain risk factors and the development of disease have been proposed but many could not be confirmed. Therefore, the objective should not be to achieve solely a reduction of risk factor levels. For example, at present, there is a controversy among health scientists whether a high intake of antioxidants (such as tocopherol, vitamin C, provitamin A, and selenium) has a preventive effect on cancer and cardiovascular diseases. As long as convincing evidence for a causal link is missing, it is not regarded as necessary to adapt current recommendations.
Morbidity and mortality rates
The promotion of a healthy lifestyles and improvement of risk factor levels is just one necessary step on the way to increasing health. The final success of a community intervention programme must be analysed on the basis of changes in disability, morbidity, and mortality rates. At the community level, these outcomes are usually measured by prevalence rates, incidence rates, lost years of life, mean age at onset of disease, or mean age at death from a specific disease. Improvement of the quality of life may be measured as well as the average years of disease-free life of a population. Not only are changes in specific mortality rates important but also changes in morbidity rates and premature disabilities. These outcomes are often overlooked in the evaluation of community trials, although they can largely contribute to the healthy life expectancy.
In addition, the intervention should have a positive impact on total mortality. A successful reduction of coronary heart disease mortality through intervention should, for example, not be accompanied by higher cancer mortality rates. A real improvement of health can only be ascertained if total mortality rates are decreased and life expectancy is improved.
Most chronic diseases occur at older ages. To reach older people through community intervention, and to have an essential influence on the morbidity and mortality of very old people, seems impossible. For example, high mortality rates of cardiovascular diseases occurring at old age in a society are not alarming but, on the contrary, provide an indication of superior health status and medical care in this society. Health promotion cannot prevent all diseases and will not lead to immortality. The purpose of the intervention measures is to reduce the occurrence of diseases and deaths early in life. In the Western world this means a reduction of mortality and morbidity before reaching the age of 70 to 80 years. Thus this should be a criterion to measure the success of intervention, which is possible in a well-designed trial.
Early estimate of mortality outcome: the multiple logistic function
As mentioned above, the final evaluation of success of community trials should be based on the changes in premature morbidity and mortality of the disease(s) of interest as well as total mortality. In practice, this can only be evaluated after a long time period in which the detectable changes in disease rates will occur. Changes in risk factor levels will occur much earlier. For example, the elevated risk of smoking on lung cancer will remain for about 10 to 20 years after smoking cessation (for cardiovascular disease, the risk will drop faster). Consequently, most researchers want to summarize risk factor changes to predict changes in morbidity or mortality in the long term. Project financers and politicians in public health are also eager for an early estimate of the success of the intervention programme.
A classical tool to summarize the changes in risk factor levels is the ‘multiple logistic function’. Such a function weighs the changes in major risk factors according to their importance as contributors to mortality risk. The weighting factors are derived from multiple logistic regressions of longitudinal mortality data. Multiple logistic functions are used by many epidemiologists to estimate mortality risk from early measured risk factors.
Although these functions are widely used for summary evaluation of risk factor changes, it is important to be careful when applying them. The estimated weights (from a different population) may not reflect the impact of risk factors on mortality in the observed population. The logistic model does not consider follow-up time and censoring (owing to unfinished observation times for drop-outs and individuals still alive). Basically, a multiplicative effect of risk factors is assumed which is not appropriate, for instance, for all coronary heart disease risk factors. Inclusion of new risk factors will therefore change the estimates of the others. The estimated function is likely to be more a reflection of the model than of the data. More recent multiple logistic functions are based on more sophisticated proportional hazard models (Cox and Oakes 1984).
Intervention in communities
The development of chronic diseases early in life may result from unhealthy behaviour patterns and unfavourable life conditions. A major fraction of the population shows such behaviours. Therefore, health promotion and disease prevention programmes try to manipulate this behaviour and try to establish structures which support a healthy lifestyle.
Local communities as the field of intervention activities
Local communities play a key role in the realization of intervention measures. They form the field in which most events of social life take place. They provide the structures and institutions for daily life. For example, the local media are effective instruments for spreading health-related information. Churches, schools, sports clubs, self-help groups, and other organizations with regular meetings are important institutions for creating a positive intervention climate and for establishing norms and reinforcing them frequently. The health system and the food distribution system can also be used for the purposes of health promotion.
Available health structures, information resources, and people and groups with influence and credibility within the community are important tools for reaching the objectives in community intervention trials. They will provide the conditions by which a larger population can be exposed to messages about healthy lifestyle, become involved in health topics, and experience the advantages of improved health behaviour. Different methods of communication, education, and advertising concerning health issues can be used to achieve improvements in health knowledge, attitudes, and behaviour.
Theories and frameworks focused on community intervention trials
Intervention methods and activities in community studies should have a theoretical foundation. Social science research has provided profound experience about the psychological and sociological mechanisms by which new norms in daily life can be achieved and propagated, new fashions can be created, and more involvement in health issues can be generated. Different social theories and frameworks attempt to explain individual behaviour as well as trends and fashions within populations. All of these general behaviour theories can be applied to issues relevant for health. They should be used to define a set of measurements in a community intervention study. Important points and elements applicable to community intervention are listed in Fig. 2.

Fig. 2 Focus points for influencing behaviour and elements of intervention.

One of the earliest and most often applied theories is the ‘social cognition theory’ formerly known as ‘social learning theory’ (Bandura 1986). The focal point of this theory is that changes of behaviour can be achieved through intensive exposure to important models (ideals or archetypes, such as pop stars or sport stars). Self-efficacy (including self-esteem, self-regard, self-respect, self-confidence, competence, and effective functioning) and group efficacy play an important role in changes of behaviour. It is influenced by personal, observed, or otherwise transferred experiences. Furthermore, according to this theory, for maintenance of a newly adapted behaviour, a supportive social setting and the development of skills are needed.
To adopt this framework for health purposes, opinion leaders in a community (for example, the mayor, the medical professionals, a local sport star) should be involved in intervention management. They should convincingly and repeatedly appeal to the public to stop smoking, eat less fat, be more physically active, and so on. In addition, skills training should be organized within communities.
The ‘theory on reasoned action’, which became the ‘theory of planned behaviour’, analyses and predicts behaviour, and was mainly developed by Ajzen and Fishbein and extended by Ajzen and Madden (1986). This framework applied to health promotion, concentrates on establishing the credibility of people distributing information about health issues, favourable lifestyle, and disease prevention measures. The theory suggests that individuals pass through a series of steps from awareness, attitudes, and knowledge acquisition through to motivation and skill development, and finally take action to change behaviour. The prevailing subjective norms in the community have a high impact on the health behaviour of its members. Perceived behaviour control is the ability to cope with difficulties by adapting to positive behaviour. To sustain adapted behaviour, skills of self-management have to be learned. The theory can be used in community intervention trials by regularly distributing information about healthy lifestyle in the media (such as local press and television) whilst ensuring high credibility.
‘Persuasive communication’ campaigns (McGuire 1984) try to convince individuals to take more responsibility for their own health maintenance. Based on psychological theories concerning communication, attitudes, and behaviour, a seven-step procedure is proposed.

Reviewing the realities.

Axiological analysis.

Surveying the sociocultural situation.

Mapping the mental matrix.

Teasing out the target themes.

Construction of the communication.

Evaluating the effectiveness.
The sixth step is the crucial one, because it is the practical application. It uses a communication set containing aspects such as credibility, attractiveness, and power. Elements mediating the communication, such as exposure, skill acquisition, and motivation, are also implemented in this framework.
The ‘precede–proceed model’ (Green and Kreuter 1991) for educational intervention is a framework to plan and administer health education programmes. It begins with five phases: social, epidemiological, behavioural/environmental, educational/organization, and administration/policy diagnosis. This is followed by implementation, process, impact, and outcome evaluation. This plan of action covers the multidimensionality of health and its large number of collaborators.
‘Social market theories’ analyse the needs of a target population (Kotler and Clarke 1987). Based on these needs adequate products can be offered and costs and benefits for the provider and consumer can be estimated. Preventive health services are products for which the audience has to be defined, messages have to be developed, and the most effective channels for acceptance selected. These theories combine and apply elements of the theories and frameworks discussed above.
Study design
The community intervention study has by its very nature a quasi-experimental design. It is experimental in the sense that the observer manipulates the intervention community with public health programmes and observes the changes in the population against a reference (without such manipulation). It is quasi-experimental in so far as the observer cannot control for every exposure or health-related change in the intervention or in the reference community.
A well-chosen study design can partly prevent unwanted confounding exposures and will control more efficiently for unknown or unexpected influences on the population’s health. Observing several intervention versus reference communities which are spread over the region of interest is one way of ruling out initial inequalities and lowering the chance of unexpected confounding trends between intervention and reference areas. This can also be achieved by taking ’embedded’ intervention communities out of a larger region which may serve as a reference.
Most of the intervention studies listed in Table 1 follow a design comparing intervention with reference communities. A similar design was also used in the successful Centers for Disease Control and AIDS Community Demonstration Projects in the United States. To each of five intervention communities spread over the United States, a reference community was matched by density of community members at risk, availability and use of illegal drugs, and prostitution rate. During 4 years of intervention, around 1000 people in the intervention regions were recruited and trained to distribute and verbally reinforce prevention messages and materials. Intervention was based on behaviour change theories (DiClemente and Wingood 1995) and provided basic AIDS-related information, instruction on the use of condoms, ‘small-media’ materials (booklets, leaflets, flysheets, and so on) about the role model’s progress, and increased availability of condoms and bleach kits (for needle-sharing people). Evaluation was conducted by anonymous interviews with 15 000 target people in 10 cross-sectional waves. Two waves were completed before the implementation of intervention activities. At the end of the intervention, more than half of the target population was reached at least once. A statistically significant increase by 74 per cent in people carrying condoms and safer sexual behaviour (40 to 50 per cent higher), as well as increased bleach use in those sharing needles, was observed in the intervention communities as compared with the reference communities.
The choice of intervention and reference population
Theoretically, the design for a community intervention trial should include a group of communities randomly chosen from a country or region of interest. The chosen units should be assigned by chance as intervention or reference units.
If the findings of an intervention programme are intended to be applied to a larger population in the future, they should reflect this population. The units must represent the main health-relevant differences within the country or region. Depending on the infrastructure of the country it should include rural regions, medium-sized towns, and parts of large towns. Inhabitants within those units should reflect the variety of socio-economic groups and variation in lifestyle. The samples of these intervention and reference regions would then allow generalization of the outcomes of intervention measures to the total population. From a health policy perspective, this is an important aspect of a community intervention study. This type of study would have the additional advantage of avoiding misinterpretation of results due to unknown secular trends and variance of outcome variables.
Such a design, although desirable from a scientific perspective, is often not feasible for many practical reasons. It requires complex organization and logistical support, and would be very expensive. After a long period of intervention and evaluation, no major health impact might be visible and cost-effectiveness might be unfavourable.
Therefore, in most conducted trials, one or only a few regions were taken as intervention and reference areas. Table 1 shows the main characteristics of the major cardiovascular disease community intervention studies. The number of selected communities, size of intervention populations, intervention and observation period, and other typical issues are described. In most of the community intervention trials conducted, between one and three cities or parts of cities were chosen as intervention units and one or two as reference cities.
In such a situation, it is even more important to ensure that the reference community resembles the intervention community as closely as possible. Both regions should be identical with respect to sociodemographic structure, initial risk factor levels, and mortality risks. In practice, however, the researcher is not free in choosing appropriate communities. Political, practical, or financial reasons, such as the need for intervention due to high cardiovascular mortality risk, proximity to the intervention centre or research group, or the pressure to conduct intervention from the public or community leaders of a certain region, may play a role in the final design.
The appropriate reference should also be chosen carefully. The need for similarity with the intervention community often makes it necessary to find a region nearby, although this may raise specific problems. Unexpected changes in health status or health perspectives can occur due to other interventive measures in the reference community. Spillover of information and cognition from the intervention population, changes in socio-economic conditions like increasing unemployment, changes in the health-care system or improvement of medical diagnoses, and spontaneous lifestyle change due to health-relevant events can occur in the reference community. For example, when a local individual or team becomes a sport champion in a certain sport and gains popularity, many inhabitants may start to participate in this sport. Such changes can also occur in the intervention population, independently of the intervention efforts.
Secular trends in the reference population could differ from the intervention population. These differences cannot be detected from initial levels of risk factors. Several of these unwanted coincident changes are unlikely to occur when a nested design is chosen. This can be realized if the pre- and post-surveys are representative independent samples of the whole nation, and the intervention population is an independent representative (clustered) sample of the total population of this nation. This was the case in the German Cardiovascular Prevention Study. In this study, six intervention communities and rural regions scattered over the country were selected (Table 1) (Hoffmeister et al. 1996). The pooled intervention units closely represent the German population by age, gender, and socio-economic status. The risk factor means of the pooled intervention regions were nearly identical with those of the national reference at the beginning of the study. Although there were differences in initial levels between the six single regions, the close resemblance of the risk factor means between the pooled intervention regions and the reference regions shows convincingly that the chosen regions together represent the German health conditions. Figure 3 shows the magnitude of differences occurring between communities or regions due mainly to socioeconomic and cultural settings. The mean levels of total serum cholesterol in the six intervention regions of the German Cardiovascular Prevention Study are presented as an example (corrected for differences in age and gender distribution). The figure also shows that the intervention had a delayed impact for this risk factor. Often the financial aspects of such a widespread reference will force projects to use a different reference design. National representative surveys, however, can also be used as national health surveys.

Fig. 3 Changes in mean total serum cholesterol values (German Cardiovascular Prevention Study study, 1984 to 1992).

Survey sampling
Population surveys should be conducted before, during (often at midterm), and after the intervention programme. An additional survey, some years after the end of intervention measures, could provide insight into the endurance and implementation of improved health behaviour. The surveys should preferably be independent samples with the same age and socio-economic distributions. Ideally, the samples for both intervention and reference communities should be random, representative, or stratified samples of the community. In a situation where intervention programmes focus on certain risk groups (age, ethnic, or socio-economic factors), it might be more effective to restrict the sample to this target population. Usually community intervention is directed towards the whole population but the disease of interest is only restricted to a certain group (for example, 40- to 60-year age groups for coronary heart disease). Even so, it would be appropriate to sample just this group.
The researcher should make a great effort to achieve maximal response rates in the survey samples. To avoid a substantial bias, the response rate should be at least 70 per cent. Non-respondents can differ considerably from (primary) respondents in health behaviour and sensitivity to health intervention programmes. For a valid evaluation of the intervention-related changes it is important to achieve a high response rate. Non-respondents should be contacted repeatedly to persuade them to participate in the surveys. At the very least, minimal information about non-respondents should be gathered, for instance through a short questionnaire. This should provide some information about the difference between non-respondents and participants.
The attempts to enlarge the response rate should be equal in both the intervention and the reference communities. If response rates between the intervention and reference communities differ considerably, this response bias could have major consequences for evaluation of the intervention trial. In the German national health surveys, for example, the smoking prevalence among spontaneous respondents (first 50 per cent of the sample) was 1 to 2 per cent lower than in the next 20 per cent of respondents. Fortunately, the non-respondents did report no higher smoking prevalence in a short questionnaire.
The use of a cohort sample may have a serious impact on observed risk changes, which possibly restricts the intervention evaluation. Repeated screenings are powerful intervention measures themselves. They induce a strong version of the ‘Hawthorne effect’. People, being aware that they are under study, are likely to change their behaviour in a positive way. Thus the screening effect of the cohort survey can affect health behaviour of the participants independent of the intervention programme or could make them react differently to the intervention programme. Therefore, an unbiased estimate of the impact of intervention on the total community is not possible.
In contrast with cross-sectional independent samples, cohort samples may provide information about the people who are susceptible to intervention. In this way, a better evaluation of the intervention programme is possible because the subpopulations which have responded to intervention can be identified. It may also give information about the parts of the intervention programme which were most effective. Ideally, a cohort sample is drawn in combination with independent cross-sectional samples, and so the advantages of both designs can be used.
An attempt to minimize the effect of repeated screenings among cohorts was made in the Minnesota Heart Health Study (Luepker et al. 1994) by repeating the surveys for half of the cohort after 2 years and for the other half after 4 years (midterm samples). Finally, the complete cohort sample was remeasured after 7 years of intervention. Whether this really eliminates a cohort screening effect, however, is not clear. This study additionally included independent cross-sectional samples.
Problems with secular trends
For several risk factors as well as for specific morbidity and mortality rates, a gradual upward or downward trend over long periods of time (decades) occurs in various countries and regions. This phenomenon is referred to as the secular trend.
Health behaviours, for example smoking prevalences, will also follow trends, although they have rarely been documented. The outcome variables in community intervention trials therefore have to be controlled for such trends. An intervention programme has to be successful in modifying these secular trends in a favourable direction. An intervention effect will be validated by comparing it with the secular trend estimated from the reference communities. In the German Cardiovascular Prevention Study Study, risk factor changes of the six pooled intervention regions were corrected for the national secular trends by using the entire nation as the reference population. The magnitude of a secular trend in a large population can be seen from the reference trend in Fig. 3.
In the case of small communities a good estimate of secular trends may be a problem. An observer should have carefully gathered information about secular trends in lifestyle and risk factors that he or she wants to modify before the initiation of the intervention programme. If there is already a very sharp decline in their levels, it will be difficult to modify this change additionally by intervention.
In the North Karelia Project (Table 1), the Finnish North Karelia region was chosen as an intervention area because it had the highest cardiovascular mortality rates worldwide at the beginning of the study. Public awareness and concern about this fact might have had an influence on lifestyle and health behaviour changes of the North Karelian inhabitants. This already initiated favourable secular trend cannot be separated from the intervention effects. In North Karelia the favourable trend in risk factors as well as in cardiovascular morbidity and mortality rates continued several years after the intervention measures had stopped. The initial occurrence of cardiovascular diseases was lower in the reference region. It is assumed that the decreasing secular trend in the reference region (Kuopio) was less pronounced. In addition to an intervention effect, differences in secular trends in the outcome variables (risk factors, cardiovascular morbidity, and mortality rates) observed between North Karelia and Kuopio may therefore partly explain the final study results.
Nevertheless, a population being aware of its high cardiovascular disease risk might be highly susceptible to health issues and intervention. Thus, it is possible that the intervention mainly induced the downward trends. The design of this project does not allow for relying on a single explanation, although from a public health view the study was very successful.
In the Kilkenny Heart Project (Shelley et al. 1995), the strong secular changes and probably a certain ‘contamination’ of the reference community by the intervention project resulted in similar risk factor reductions in both the intervention and reference counties.
In contrast with a multi-community intervention trial or a nested intervention trial, a study based on only one or two matched intervention–reference pairs should be matched for more criteria than described above. The paired communities should have the same starting levels of risk factors, morbidity, and mortality rates. If these parameters differ initially, it is an indication for inequalities concerning lifestyle and living conditions between those communities. These differences also could be caused by varying secular trends. Trends (for example, in mortality rates) should ideally be observed in both intervention and reference regions over several years before starting an intervention to achieve an impression of the grade of congruence of the matched communities.
Efforts to detect weak intervention effects
Unexpected or unknown inequalities between the communities can be responsible for different variances and trends in the outcome parameters. These circumstances, together with the limited population size of the units, make it difficult to draw conclusions and to detect the real intervention effects.
The three major community intervention trials in the United States—the Stanford Five City Project (Farquhar et al. 1990), the Minnesota Heart Health Program (Luepker et al. 1994), and the Pawtucket Heart Health Study (Carleton et al. 1995)—did not observe homogeneous and substantial net reductions of risk factors (Table 1), although significant changes occurred for single risk factors in selected sex or age groups. The change in total cardiovascular risk score did not differ significantly between the intervention and reference populations in these trials except for the cohort sample in the Stanford Five City Project. However, even small risk factor reductions conducted in large populations are likely to contribute substantially to the community health.
The researchers in these three American intervention trials concluded that the influence of intervention might have been too weak to change behaviour. This conclusion might be true. Contrary to the findings mentioned above, however, the Stanford trial shows favourable net reductions for most of the risk factors, although significant changes could not be ascertained. It may be that the sample sizes were too small to ascertain the weak net changes that could be achieved in the Stanford Study and in several other studies (Table 1).
Evaluation of changes
The complexity of community intervention trials requires evaluation instruments which measure the various dimensions of possible influences. Changes in health-related knowledge, attitudes, and behaviour, means and prevalences of risk factors, and physical symptoms and signs should be followed in the intervention versus reference populations. Specific morbidity as well as the prevalence of pain and life quality are outcomes of interest which should be evaluated. Finally, the overall impact of intervention must be evaluated by following specific and total mortality rates. For assessment and explanation of these changes, data on the social environmental and economic conditions (societal evaluation) in communities must be gathered as well. The evaluation instruments and methods used in community intervention trials are discussed in the following subsection.
Interview instruments
The main instrument to measure health-related knowledge, attitudes, and behaviour, as well as subjective levels of symptoms, signs, diseases, and quality of life in population samples, is the questionnaire. Widely used and validated scales and question blocks from epidemiological and social science research are available for intervention trials. Self- or interviewer-instructed questionnaires are useful to measure the following issues and many others in a standardized, reliable, and valid way. Interactive computer questionnaires with integrated quality, plausibility, and validity checks are often used.
The following parameters are assessable through questionnaire:


drinking (including alcohol consumption)

food consumption and eating behaviour (with food frequency lists, food protocol, 24-hour recall, and diet history questionnaires)

sport or physical activity

prevalence of chronic diseases, symptoms, signs, and pain

medication use

subjective health status and life quality

frequency of doctor visits or visits to health institutions

occupation and social status.
Analytical procedures
Physical and biochemical parameters can be measured with high precision and accuracy. Following the principles of good laboratory practice, analytical procedures or measurements will normally result in methodological errors in the range of 2 to 5 per cent. Changes achieved through intervention programmes thus can be determined in representative population samples. The expected changes should be pinpointed at the beginning of the trial. These expected differences should be used for calculating the sample size sufficient to confirm statistical significance of the observed differences.
The physical measurements should be performed under the highest quality standards for preanalytical and analytical procedures. Specially trained people should perform the sampling, storage, and transport of blood and other materials in order to minimize differences in method. Altering procedures of blood sampling (such as using the sitting versus the lying position), for example, would lead to unacceptable variation. Since considerable seasonal differences of risk factor levels and behaviour (for example, cholesterol level, nutrition, and physical activity) can occur, the intervention and reference should be sampled during the same time of the year. Clinical analyses should be performed in a central laboratory with internal and external quality controls. In this way methodological differences between laboratories are reduced. A field investigator could systematically measure too high blood pressure values due to his personality (‘Rosenthal effect’), so this should be checked and interviewers should be rotated regularly from the intervention to reference samples.
Morbidity and mortality estimates
Despite the generally large populations in intervention studies, objective and complete assessment of disease events by standardized medical examinations is difficult to achieve. Only frequently occurring and strictly defined diseases like ischaemic heart disease can be assessed sufficiently well. A continuous registration might be the ideal way to measure morbidity rates. This has, however, not been done in community trials. Morbidity data can also be gathered from practising doctors or from hospitals in the communities. This has to be done with careful monitoring of completeness and comparability of the data.
During morbidity and mortality evaluation, it is also important to ensure that availability of medical services and medical treatment do not differ between intervention and reference populations. Although the impact of curative measures like bypass surgery, programmes for early detection of certain cancers (breast, cervix, colon, and so on) on morbidity and mortality rates is controversial, it seems likely that improved medical treatment has a positive influence on morbidity and mortality.
Mortality evaluation
Every intervention trial should evaluate disease-specific mortality and total mortality rates. Since deaths from certain diseases and even from all causes are rare events, long observation times and large communities are necessary. As with risk factor changes, the expected reduction in mortality by intervention has to be pinpointed at the start of the study to estimate the necessary community size. In most intervention trials, the total number of mortality cases in the intervention and reference populations are counted and cumulative mortality rates during the study time (extended with a certain period after intervention) are compared. Causes of death can be ascertained from death certificates, hospital records, and interviews with relatives. In the German Cardiovascular Prevention Study the official national and regional (age-adjusted) mortality rates from the Federal Statistical Office have been used.
As can be seen from Table 1, no convincing intervention effect could be observed on mortality for most trials. This may be due to insufficient influence on risk factors and health behaviour by intervention activities as well as uncontrolled secular trends. Problems in evaluating mortality rates can also arise because of different developments in the population structure between intervention and reference regions. Thus, not only the mortality cases but also the drop-out and drop-in rates have to be considered. The denominator, which means the living part of population in communities and their movements, has to be estimated carefully.
Statistical procedures
The mean levels (or prevalence rates) of outcome variables of the independently drawn samples represent the initial and final health status in the intervention and reference population. In intervention trials with more than two surveys, the additional measurements during intervention time can give insight in the process of intervention-induced changes.
The statistical analysis procedures normally derive their statistical power from the number of individuals within the survey samples. The community is the unit of intervention, but changes in response to the intervention programme can only be measured by looking at individuals. This apparently paradoxical situation for the evaluation process limits the use of usual statistical procedures to measure the changes. Furthermore, the evaluator has to correct for the initial differences between the intervention and reference groups.
The classical way to measure intervention effects is by the use of simple formulas for net changes. Figure 4 shows the imaginary risk factor level in the intervention community (I0) and in the reference (R0). At the end of the intervention period, the level was changed in the intervention community (I1) and reference (R1). In this example a reduction in the intervention community (I1 – I0) was achieved whereas the level increased in the reference (R1 – R0). The net change is often expressed as the percentage change in the intervention community minus the percentage change in the reference (formula (a)) which can be rewritten as a subtraction of ratios (I1/I0 – R1/R0). Variations of this formula have been used in literature ((b) and (c)) to measure net changes. The second formula has the advantage that it allows a direct estimation of confidence intervals because it is defined as a ratio of ratios. The third formula gives the relative change divided by the baseline level in the reference which is assumed to be the baseline for the total population.

Fig. 4 Formulas to calculate net changes on the basis of community levels.

The general procedures used to calculate the net changes in intervention communities assume that baseline differences are still the same at the final measurement in the absence of an intervention programme (Fig. 4). This assumption, however, is not very realistic and will be the exception when health changes within communities are considered. Initial differences between community levels can cause complications for the evaluation. If initially measured levels differ between the observed communities, this can partly be due to chance and it is likely that effects similar to a ‘regression to the mean’ will occur: strong differences due to chance at a first measurement are likely to become smaller at the second measurement. Moreover, an initial high risk-factor level could motivate individuals within the community to change their lifestyle independently of the intervention programme.
In addition, the possible occurrence of a ‘ceiling effect’ has to be considered: individuals (or parts of communities) with high risk-factor levels may have reached their upper limits. A rising secular trend for this group will stagnate. The evaluation analyses would then over- or underestimate the potential of an intervention measure depending on the occurrence of such an effect in the intervention or reference community. In Fig. 5 a possible ceiling effect is graphically presented. The level of a certain risk factor at baseline in the reference (R0) rises further during the intervention period and reaches its biological upper limit (R1). The observed net changes (I1 – I0 versus R1 – R0) give an underestimation of the ‘true’ intervention effect which would be observed when no ceiling effect occurred. A larger effect would be measured if, for instance, the reference were to start at a lower baseline level. It is possible to argue that such an effect is a part of natural circumstances and an intervention should be successful in altering such ‘natural’ trends. However, this example illustrates how difficult it may be to detect intervention-based changes in reality. On the contrary, a very healthy population can hardly be improved in their risk factor levels due to a similar effect. Multiple samples between the start and the endpoint of the intervention, and the use of multivariate statistical models, can be used to correct for such effects.

Fig. 5 Biased estimate of intervention net change due to ‘ceiling effect’.

An example of a multivariate model to estimate the effect of intervention on various risk factor changes is given in Fig. 6. Such a model allows correction for different time trends and omits variance problems. Additional corrections for within-community correlations have been proposed by Murray et al. (1994). Significant risk factor changes in the German Cardiovascular Prevention Study (using the model in Fig. 5) were so evident that no further refinement was considered necessary. Further community intervention research should, however, concentrate on stronger methods to improve health behaviour and to reduce risk factor levels in large populations rather than to refine the statistical handling to detect weak effects.

Fig. 6 Multivariate regression model to estimate intervention effects (Hoffmeister et al. 1996).

Various concepts to influence a population’s health behaviour and risk have been developed. The effectiveness of such strategies for primary prevention in large populations can be tested with local community intervention trials. The outcome of important community intervention trials has been described and discussed in this chapter with respect to their design. Such trials raise specific design problems (such as choice of intervention and reference, sampling, and secular trends), because the intervention is conducted and changes are measured at a community level and not at an individual level.
The success of prevention programmes can be measured from changes in health behaviour, risk factor levels, morbidity and mortality of specific diseases (diseases of interest), and total mortality. Process evaluation can give information about how such changes were achieved. For the final evaluation of intervention effects specific statistical methods are available.
Chapter References
Ajzen, I. and Madden, T.J. (1986). Prediction of goal directed behaviour, attitudes, intentions and perceived behavioural control. Journal of Experimental Social Psychology, 22, 414–53.
P>Bandura, A. (1986). Social foundation of thought and action. Prentice-Hall, Englewood Cliffs, NJ.
Carleton, R.A., Lasater, T.M., Assaf, A.R., Feldman, H.A., and McKinlay, S. (1995). The Pawtucket Heart Health Program: community changes in cardiovascular risk factors and projected disease risk. American Journal of Public Health, 85, 777–85.
CDC (Centers for Disease Control and Prevention) AIDS Community Demonstration Projects Research Group (1999). Community-level HIV intervention in 5 cities: final outcome data from the CDC AIDS Community Demonstration Projects. American Journal of Public Health, 89, 336–45.
Cox, D.R. and Oakes, D. (1984). Analysis of survival data. Chapman & Hall, New York.
DiClemente, R.J. and Wingood, G.M. (1995). A randomized trial of an HIV sexual risk-reduction intervention for young African-American women. Journal of the American Medical Association, 274, 1271–6.
Farquhar, J.W., Fortmann, S.P., Flora, J.A., et al. (1990). Effects of communitywide education on cardiovascular disease risk factors, the Stanford Five-City Project. Journal of the American Medical Association, 264, 359–65.
Green, L. and Kreuter, M. (1991). Health promotion planning: an educational and environmental approach (2nd edn). Mayfield Publishing, Palo Alto, CA.
Hoffmeister, H., Mensink, G.B.M., Stolzenberg, H., et al. (1996). Reduction of coronary heart disease risk factors in the German Cardiovascular Prevention Study. Preventive Medicine, 25, 135–45.
Hoffmeister, H., Schelp, F.-P., Mensink, G.B.M., Dietz, E., and Bohning, D. (1999). The relationship between alcohol consumption, health indicators and mortality in the German population. International Journal of Epidemiology, 28, 1066–72.
Hursting, S.D., Thornquist, M., and Henderson, M.M. (1990). Types of dietary fat and the incidence of cancer at five sites. Preventive Medicine, 19, 242–53.
Kotler, P. and Clarke, R.N. (1987). Marketing for health care organizations. Prentice-Hall, Englewood Cliffs, NJ.
Liao, Y., McGee, D.L., Cao, G., and Cooper, R.S. (2000). Alcohol intake and mortality: findings from the National Health Interview Surveys (1988 and 1990). American Journal of Epidemiology, 151, 651–9.
Luepker, R.V., Murray, D.M., Jacobs, D.R., et al. (1994). Community education for cardiovascular disease prevention: risk factor changes in the Minnesota Heart Health Program. American Journal of Public Health, 84, 1383–93.
Luepker, R.V., Rastam, L., Hannan, P.J., et al. (1996). Community education for cardiovascular disease prevention: morbidity and mortality results from the Minnesota Heart Health Program. American Journal of Epidemiology, 144, 351–62.
McGuire, W.J. (1984). Public communication as a strategy for inducing health promotion behavioural change. Preventive Medicine, 13, 299–319.
Murray, D.M., Hannan, P.J., Jacobs, D.R., et al. (1994). Assessing intervention effects in the Minnesota Heart Health Program. American Journal of Epidemiology, 139, 91–103.
Rossouw, J.E., Jooste, P.L., Chalton, D.O., et al. (1993). Community-based intervention: the Coronary Risk Factor Study (CORIS). International Journal of Epidemiology, 22, 428–38.
Sellers, D.E., Crawford, S.L., Bullock, K., and McKinlay, J.B. (1997). Understanding the variability in the effectiveness of community heart health programs: a meta-analysis. Social Science and Medicine, 44, 1325–39.
Shelley, E., Daly, L., Collins, C., et al. (1995). Cardiovascular risk factor changes in the Kilkenny Health Project. A community health promotion programme. European Heart Journal, 16, 752–60.
Staessen, L., De Bacquer, D., De Henauw, S., De Backer, G., and Van Peteghem, C. (1997). Relation between fat intake and mortality: an ecological analysis in Belgium. European Journal of Cancer Prevention, 6, 374–81.
Steyn, K., Steyn, M., Swanepoel, A.S.P., et al. (1997). Twelve-year results of the Coronary Risk Factor Study (CORIS). International Journal of Epidemiology, 25, 964–71.
Vartiainen, E., Puska, P., Jousilahti, P., Korhonen, H.J., Tuomilehto, J., and Nissinen, A. (1994). Twenty-year trends in coronary risk factors in North Karelia and in other areas of Finland. International Journal of Epidemiology, 23, 495–504.
Veierod, M.B., Thelle, D.S., and Laake, P. (1997). Diet and risk of cutaneous malignant melanoma: a prospective study of 50,757 Norwegian men and women. International Journal of Cancer, 71, 600–4.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: