The objectives of this review are:

a) to assess the impact of personal financial incentives, six or more months after recruitment into an incentive scheme, on the performance of habitual health-related behaviours:

1. regardless of whether the incentive is still being offered at that time-point, and

2. when the incentive has been discontinued for at least one month.

b) to assess the extent to which the impacts reported in (a.1) and (a.2) are modified by:

1. behaviour type (smoking related vs. eating-related vs. physical activity-related),

2. incentive scheme characteristics (value of the incentive and whether attainment is certain vs. uncertain),

3. participant characteristics (level of social and material deprivation),

4. study characteristics (level of risk of bias relating to the standardization of study procedures across groups and the reliability of the outcome measures).

c) to assess the impact of personal financial incentives on motivation (intrinsic vs. extrinsic) to sustain outcomes after the incentive has been discontinued.

Electronic searches

We will conduct computerised searches of the following databases:

• MEDLINE (Ovid SP) (1948 to present)

• EMBASE (Ovid SP) (1974 to present)

• PsycINFO (Ovid SP) (1806 to present)

• CINAHL (EBSCO Host) (1981 to present)

• Cochrane Database of Systematic Reviews (The Cochrane Library) (1991 to present)

• Cochrane Central Register of Controlled Trials (CENTRAL), The Cochrane Library (1991 to present)

• SCOPUS (Elsevier) (1996 to present)

• Database of Abstracts of Reviews of Effects (The Cochrane Library) (1994 to present)

We will limit searches to studies of adults (18+ years of age). We will not apply restrictions with regard to the language of publication.

Searching other resources

In order to identify relevant ongoing and unpublished studies (e.g. dissertations, conference proceedings, working papers etc) we will search the following resources:

• HMIC (Ovid) (1983-present)

• Online clinical trials registers

o www.controlled-trials.com/mrct/ for UK trials

o clinicaltrials.gov/ct2/search for US trials

• Google Scholar (using a basic keywords such as “financial incentives” “smoking cessation”, “physical activity”, “weight-loss”; the first 1000 references will be scanned)

• Websites of key organisations in the area of health and incentives for health promotion

o Center for Health Incentives and Behavioral Economics (http://chibe.upenn.edu/)

o Healthy Incentives (www.healthyincentives.org.uk/)

o Weight Wins (www.weightwins.co.uk/)

o Departments of Health for England, Scotland, Wales and Northern Ireland

o Australian Federal and States Departments of Health

o The World Health Organisation

o United States Department of Health

In addition, we will search reference lists of eligible articles and contact key researchers and authors to identify further potentially eligible published, unpublished or ongoing studies.

We will include randomised controlled trials and cluster randomised controlled trials, which assess the impact of personal financial incentives on habitual health-related behaviours (smoking cessation, physical activity and healthier eating, including alcohol consumption), and/or the proximal direct consequences of such behaviours. At least one comparison group in the trials must have been randomised to receive personal financial incentives and compared to either groups not receiving financial incentives and/or groups receiving financial incentives that differ in type and/or amount. Trials must have measured outcomes up to at least 6 months from the start of the intervention.

We will exclude all studies other than randomised controlled trials to minimise the risk of bias. We will only include studies with a minimum follow-up of 6 months because we are interested in the sustainability of habitual health-behaviours. Performance of the target behaviour at six months from the beginning of the intervention is the gold standard for smoking cessation (Hughes 2003). We are applying this criterion to the other target behaviours and outcomes for reasons of standardisation and comparability. We will include studies with multiple comparison groups in which participants are offered personal financial incentives that differ in specific characteristics, such as type and/or monetary value. We will include studies of the effects of multi-component interventions if two or more comparison groups are exposed to interventions that differ only in the offer of personal financial incentives (or in the offer of personal financial incentives that differ in specific characteristics). However, we will exclude studies of the effects of multi-component interventions in which personal financial incentives feature as one component, but the study design precludes collection of data relating to the independent effect(s) of incentives.

Poor habitual health-related behaviours, including tobacco smoking (Batty 2008; Teo 2006), poor diet-related behaviours (including the harmful use of alcohol) (Cox 2000; He 2007; Heidemann 2008) and lack of physical activity (Andersen 2000; Batty 2001), contribute greatly to the development of major risk factors for non-communicable diseases (NCDs). These disease, which include cardiovascular diseases, type 2 diabetes, certain types of cancers and chronic respiratory diseases, together account for more than 50% of preventable deaths worldwide (3four50.com; WHO 2011). The morbidity and mortality burden of NCDs affects people in all age groups, imposing large, increasing and avoidable costs in human, social and economic terms (Beaglehole 2011; WHO 2011).

The prevalence of NCD-related risk factors, such as obesity, hypertension, raised blood glucose and cholesterol, as well as the physiological or metabolic consequences of tobacco smoking, can be reduced by changing individuals' and populations’ health-related habits, so as to promote certain healthy behaviours, including smoking cessation, physical activity and healthier eating (including the responsible consumption of alcohol). Achieving this could in turn reduce the prevalence and burden of NCDs (Katz 2009).

Modifying habitual health-related behaviours however, is difficult. Although many people report that they want to change their behaviour to improve their health, most find it difficult to implement and maintain the necessary changes (Ogden 2007; Sutton 1998).

One possible way to improve individuals’ health-related behaviours is through the use of personal financial incentives. Personal financial incentives are increasingly being considered and applied in health policies around the world in an attempt to promote health-enhancing behaviours (Le Grand 2007; Largarde 2007). Several aspects of the effectiveness of personal financial incentives to promote health-related behaviours however, remain unclear (Marteau 2009). For example, although there is evidence that they can be effective in promoting non-habitual health-related behaviours, such as attendance at clinic appointments, uptake of immunisations, mammography screening and tuberculosis screening, and adherence to healthcare treatments (Sutherland 2008), the currently limited evidence base indicates that the impact of such incentives on more habitual health-related behaviours, such as smoking-, diet- and physical activity-related behaviours, is less straightforward (Sutherland 2008). Furthermore, evidence for the sustained effectiveness of personal financial incentives beyond the period of intervention remains to be established (Marteau 2009).

Research is needed which will synthesise the available evidence across various habitual health-related behaviours, in order to establish the exact conditions under which incentives are effective in changing such behaviours, i.e. to determine which types of personal financial incentives, for which participants and which behaviours (or related outcomes) result in greatest changes. Furthermore there is a need to determine whether these behaviour changes are:

(a) sustained after the incentive is discontinued;

(b) maintained for the duration incentives are offered but “crowd-out” intrinsic motivation, making it less likely than before incentivisation that people engage in the healthy behaviour after the incentive is discontinued; or

(c) maintained only for the duration incentives are offered with the behaviour returning to baseline levels after they are discontinued.

There is also a need to elucidate the circumstances under which each possibility might occur.

This review will address these gaps in the literature by focusing on the use of incentives for changing poor habitual health-related behaviours, i.e. tobacco smoking, physical inactivity and unhealthy eating, including the harmful drinking of alcohol, and for promoting healthier habitual behaviours, i.e. smoking cessation, physical activity and healthier eating, including the responsible drinking of alcohol. In addition, it will consider the impact of personal financial incentives on the proximal indicators of eating behaviour and performance of physical activity (body weight, body mass, blood glucose, blood cholesterol, blood lipids) (i.e. major, modifiable physiological or metabolic risk factors for NCDs). In will also attempt to determine the variables that modify the effect of financial incentives on habitual health-related behaviours.

Although various existing reviews have examined the use of incentives for changing health-related behaviour, no single review has focused explicitly on habitual health-related behaviours and has asked the same questions as those proposed in this review. Although the proposed review has some overlap with existing reviews in terms of the included studies, it will differ through the inclusion of further trials and the analysis of variables which hitherto have remained unexamined. By building upon existing reviews it will endeavour to produce a more complete and comprehensive picture of the impact of personal financial incentives allowing generalizations across habitual health-behaviours, both about the impact and the modifiers of such impact.

Adults aged 18 years or over (no restrictions for socio-economic or clinical characteristics or prognostic factors).

Given the prediction that the impact of personal financial incentives is moderated by recipients’ level of social and material deprivation, (e.g. Sutherland 2008), we will classify participants at the study level as either highly social and materially deprived (“High”) or not highly socially and materially deprived (“Other”).

Interventions will consist of the offer of personal financial incentives, provided directly to patients or consumers (as opposed to health-care providers), contingent upon: smoking cessation; performance of a pre-specified level of physical or sedentary activity; achievement of a pre-specified target relating to the eating of healthier or less healthy foods and drinking of alcoholic beverages; achievement of a pre-specified calorific or nutritional target related to nutrient intake; achievement of a pre-specified level of energy expenditure; and/or achievement of a pre-specified level of weight loss.

We will exclude incentives of little or no monetary value and those of symbolic value(e.g. certificates, stickers, badges, key-rings, t-shirts, caps, hats or mugs) and incentives that are not contingent on individual performance of the target behaviour(s) or achievement of the target outcome(s) (e.g. consumer sales promotions, direct pricing, income transfer programs, tax credits).

For the purposes of this review, we will classify personal financial incentives according to two dimensions, presented in order of expected importance:

1. the monetary value of the financial incentive (whether high or low; see 'Data Extraction'). This variable has been frequently proposed as an important modifier of the effect of financial incentives on health-related behaviour (e.g. Sutherland 2008; Lussier 2006; Paul-Ebhohimhen 2008)

2. whether attainment of the financial incentive is certain (i.e. the possibility of obtaining the incentive depends only on performance of the pre-specified target behaviour or achievement of the pre-specified target outcome) vs. uncertain (i.e. the possibility of obtaining the incentive depends both on performance of the pre-specified target behaviour or achievement of the pre-specified outcome and chance. Performance of the pre-specified target behaviour or achievement of the pre-specified target outcome entitles participants’ to the possibility of winning the incentive by being entered into a draw/lottery/sweepstake/competition/contest. Actually attaining the incentive however, depends on chance). Assessing this distinction is important, as research in related areas suggests that participants might respond differentially to a certain vs. an uncertain incentive (e.g. Leung 2002).

Eligible comparison groups will be those in which participants are exposed to:

a) no treatment;

b) the same treatment as the incentivised group(s), but without the offer of a personal financial incentive; or

c) a personal financial incentive that differs from that offered to the treatment group in type (i.e. certain vs. uncertain), and/or monetary value.

There will be no restrictions relating to the geographical or organisational setting(s) or context(s) in which the intervention(s) are provided.

Primary outcomes

Achievement of the desired habitual health-related behaviour or related outcome – i.e. performance of the target health behaviour or achievement of the target outcome, at least 6 months after recruitment into the personal financial incentives scheme and one month after the personal financial incentive has been discontinued, where the target behaviour or related outcome refers to that for which the incentive has been offered.

For each of the habitual health-behaviours we are considering, we are interested in the following outcomes:

Smoking cessation:

• cessation (dichotomous - measured by carbon monoxide reading or cotinine test of urine, saliva or blood).

Physical activity:

• achievement of target level or frequency of physical activity (dichotomous - measured objectively, e.g. by pedometer, activity record, diary, questionnaire or scale)

Eating healthier foods:

• achievement of target amount or frequency of specified healthier food(s)/drink(s) (including alcoholic beverages) consumed (dichotomous - measured objectively, e.g. by diet record or diary, food frequency questionnaire)

Eating unhealthier foods:

• achievement of target amount or frequency of specified unhealthier food(s)/dink(s) (including alcoholic beverages) consumed (dichotomous - measured objectively, e.g. by diet record or diary, food frequency questionnaire)

Proximal direct consequences of eating behaviour and/or performance of physical or sedentary activity:

• achievement of target calorific or nutritional profile of food(s)/drink(s) consumed (dichotomous - measured objectively, e.g. based on diet record or diary, food frequency questionnaire)

• achievement of target level of energy expenditure (dichotomous - measured objectively, e.g. based on activity record, diary, questionnaire or scale)

• achievement of target level of cardio-respiratory fitness (dichotomous - measured by maximal oxygen intake VO2 max)

Risk factors for NCDs:

• achievement of target body weight/body fat distribution/body mass/related proxies (e.g. leptin, adipocytokines and other obesity or inflammatory markers), given target weight loss/fat loss/body mass/related proxies if applicable (dichotomous - measured objectively)

• achievement of target blood cholesterol level/blood lipid profile/blood glucose level (dichotomous - measured objectively)

Motivation (intrinsic vs. extrinsic) to engage in target health-related behaviour (dichotomous- measured using self-report questionnaires)

Where these outcomes are not available, we will record and dichotomise the following continuous outcomes:

Physical activity:

• amount of physical activity (continuous - measured objectively, e.g. number of minutes or steps performed)

Eating healthier/unhealthier foods:

• amount of healthy/unhealthy food(s)/drink(s) (including alcoholic beverages) consumed (continuous - measured objectively, e.g. by diet record or diary, food frequency questionnaire)

Proximal direct consequences of eating behaviour and/or performance of physical or sedentary activity:

• amount of calories from food(s)/drinks(s) consumed (continuous -measured objectively, e.g. based on diet records or diary, food frequency questionnaire)

• level of energy expenditure (continuous - measured objectively, e.g. based on activity records, diary, questionnaire or scale)

• level of cardio-respiratory fitness (continuous - measured by maximal oxygen intake VO2 max)

Risk factors for NCDs:

• level of weight loss/fat loss/body mass improvement (continuous - measured objectively)

• level of blood cholesterol/blood glucose/blood lipid (continuous - measured objectively)

Psychological variables:

• Motivation (intrinsic vs. extrinsic) to engage in target health-related behaviour (continuous- measured using self-report questionnaires)

We will extract only dichotomous outcome data and present it in tables describing and summarising the results of each study. Where dichotomous data are not available, we will extract continuous outcome data and dichotomise it, by converting SMDs directly to odds ratios.

We will deal with varying time-points of assessment of the outcome by creating time-assessment categories. These will begin at six months after recruitment into an incentive scheme and will consist of six month intervals (i.e. 6 months, 6-12 months, 12-18 months, 18-24 months etc since recruitment). We will also create time-assessment categories for after removal of the incentive. These will consist of one month intervals between 1 and 3 months, a three month interval between 3 and 6 months and six month intervals thereafter (i.e. 1-2 months 2-3 months, 3-6 months, 6-12 months, 12-18 months, etc. after discontinuation of incentives). We will calculate odds ratios for outcomes.

The extraction and (where necessary and possible) conversion of outcome data into dichotomous measures is intended to allow an overall estimate of behaviour change across the three sets of target behaviours.

Secondary outcomes

None

Two authors (EM and FV) will independently extract all data. If outcome data are unavailable or are not presented in the published full-text reports of individual studies in the forms pre-specified in 'Types of Outcome Measures' (i.e. dichotomous data), or we cannot converted them to the necessary format, we will contact study authors with a request to provide these data. The first author (EM) will reconcile the two sets of independently completed data extraction forms. If there are inconsistencies between the two sets, we will re-check extracted data and verify them against the corresponding full-text study report. If uncertainly remains, the two data extractors will meet to discuss and reach a consensus. If consensus cannot be reached a final decision will be made following discussion with a third author (IS).

To allow for assessment of the role of the pre-specified moderating variables (i.e. incentive scheme characteristics (incentive value and certainty) and participants’ level of social and material deprivation)), during the data extraction process we will categorise incentives and their recipients at the study level. Specifically, we will classify incentives according to:

a) their value i.e. low (>$400) vs. high (=< $400).We will make judgements of “High value” if the total value of incentives is larger than the minimum weekly income required to be earned per household for individuals to be above the USA poverty threshold. We have chosen to follow USA guidelines because currently the majority of research in this field has been conducted in this country. The average number of family members per household in the USA is three (rounded off to the nearest figure) (United States Census Bureau, 2011) with the equivalent poverty threshold set at approximately $18530 annually ($386 weekly) per household (US Department of Health & Human Services, 2011).Based on this, we will classify the value of incentives worth $400 (total value) and above as “high” and those worth $400 (total value) and below as “low”.

b) their type, i.e. certain (all incentives, such as cash, deposits, gifts, vouchers etc., excluding lotteries) vs. uncertain (i.e. lotteries)

We will collect information on participants’ level of social and material deprivation and make judgements based on any relevant information that is available in the included studies (e.g. income, employment, education, ethnicity, SES scores). We will aggregate this information to allow studies to be categorised as either highly social and materially deprived (“High”) or not highly socially and materially deprived (“Other”). We have chosen this categorisation because our primary interest is to determine whether incentives are more effective for the most deprived, rather than to assess the level of effectiveness associated with each level of deprivation. We will make categorisations at the study level to allow for between-studies comparisons. We will make judgements of “High deprivation” when any or all of the following conditions are met:

1. Majority of study participants have not completed high school or the mean number of years in education is less than 12 years

2. Majority of study participants earn less than $ 20,000/year ($1,666/month), or the mean reported income is less than $20,000/year or the majority of participants are allocated to the lowest income category,

3. Majority of study participants are unemployed or in unskilled, semi-skilled, skilled, or blue collar jobs

4. Majority of participants have a low SES score or the mean SES score is low. Decisions about whether SES scores are indicative of high deprivation will be made by referring to the scoring of the scale used and any related instructions for interpreting these scores.

5. Majority of study participants are non-White. This information will be used when income, education, occupation and SES have not been measured, or when the information provided by these variables does not allow for definite categorisations (e.g. income is low but education is borderline, such as just above 12 years) Judgments of high deprivation based on these variables will not be affected if the sample is predominantly white.

6. Majority of study participants are underinsured or lacking insurance, receiving Medicaid, or attending public clinics or Women Infant and Children (WIC) programmes.

7. Majority of study participants are living in an area of deprivation, or receiving welfare benefits (including, in the UK, free school meals).

If the information provided by two variables is contradictory, e.g. income is low but education is high, then we will take into account the information provided by a third variable, such as occupation or ethnicity, to make a judgment.

If no relevant information is reported in the paper, then we will contact authors and enquire about the availability of relevant data.

We will assess risk of bias of included studies at the outcome level. For both randomised controlled trials and cluster randomised trials, we will assess risk of bias by applying of the Cochrane Collaboration risk of bias tool (Higgins 2011). We will assess the risk of bias for the following domains

Selection bias:

1. Random sequence generation

2. Allocation concealment

Performance bias:

1. Blinding of participants and personnel

We do not expect knowledge of intervention allocation by participants to lead to performance bias. In fact, blinding of participants is usually not relevant in studies assessing the impact of financial incentives on health-related behaviours. For the intervention to work participants need to be aware of their entitlement to incentives, so that they can perform the necessary behaviour/achieve the outcome necessary for their attainment. Consequently, we will not consider studies in which participants were not blinded to be at high risk of bias. We will make risk of bias judgements regarding blinding of personnel (and whether their knowledge of the intervention may have altered the way they interacted with participants, and has thus influenced outcomes)

2. Standardization of study procedures

A related potential source of performance bias specific to trials assessing the impact of financial incentives on health-enhancing behaviours that we will assess, is whether studies have controlled for the additional processes inheret in the delivery of the incentive, compared to regular treatment: Attainment of incentives often requires additional involvement, on behalf of both participants and personnel, in the form of frequent clinical appointment attendance, monitoring of the formers’ performance etc, which may confound the impact of financial incentives, leading to an overestimation of their effectiveness.

We will make judgements of low risk of bias when study procedures have ensured that all processes are standardised between groups (i.e. all participants attend an equal number of clinical appointments and their performance is monitored a comparable number of times) apart from the provision of financial incentives contingent on performance of a target behaviour/achievement of a target outcome. A lack of such standardisation will result in judgments of high risk of bias, whereas we will make a judgement of unclear risk of bias when there is insufficient information regarding the procedures relating to the non-intervention groups. We will incorporate this risk of performance bias assessments into the analysis to determine whether the impact of financial incentives co-varies with such between-study differences.

Detection bias

1. Blinding of outcome assessment

In trials assessing the impact of financial incentives on health-related behaviours outcome assessors are often responsible for disseminating the incentives. We expect it to be often the case therefore that assessors are aware of which group a participant has been allocated to. Whether or not a lack of blinding of outcome assessment leads to bias will largely depend on the robustness/reliability of the outcome measure used in each study, and the extent to which it requires judgements on behalf of the outcome assessors.

2. Reliability of outcome measure

A related source of detection bias, the risk of which we will assess in studies included in this review, concerns the method of outcome assessment employed and the extent to which it is reliable or can be deceived. We expect easily falsifiable measurements to be deceived more by participants in conditions where delivery of the financial incentive is contingent on the outcome of the assessment, thus leading to bias. We will consider studies in which the outcome assessment relies purely on self-report measures at high risk of bias, compared to those which include an objective outcome measure, such as a biochemical indicator. For example, in the case of physical activity and healthier eating, we will consider studies at low risk of bias if they rely on biochemical indicators such as weight-loss, maximal oxygen intake, blood lipid/glucose profiles, as opposed to diaries or questionnaires. With regards to smoking cessation, we will consider studies at low risk if smoking status is measured using the Russell standard (West 2005), as opposed to relying on self-report or monitoring of carbon monoxide level. We will incorporate these risk of detection bias assessments into the analysis to determine whether the impact of financial incentives co-varies with the type of method used to assess outcomes.

Attrition bias

Incomplete outcome data

We expect that in studies assessing the impact of financial incentives on health-related behaviour, greater levels of attrition will be observed in non-incentivised groups compared to the incentivised groups. We will analyse originally dichotomous and/or dichotomised outcome data missing due to participant drop-out via intention-to treat analysis, with a conservative assumption being made that participants dropping-out have not sustained (or achieved) the target behaviour or related outcome.

Reporting bias

1. Selective outcome reporting

2. Other sources of potential bias

For cluster randomised trials we will also consider the following:

Recruitment bias

For this domain, we will make high risk of bias judgements for studies where participants were recruited into clusters after randomisation was completed. We will make low risk of bias judgements for studies where recruitment was completed before randomisation. We will make unclear risk of bias judgments for studies where there is a lack of information regarding the order of recruitment and randomisation.

Two authors will independently apply the risk of bias tool. Additionally, each author will collect and record the source of information for each risk of bias judgement (e.g. quotation or summary of information from trial report). Where judgements are based on assumptions made on the basis of information provided outside publicly available documents, this should be stated. Any inconsistencies between the two authors with respect to coding judgements or information in support of judgements will be resolved by consensus. If consensus cannot be reached a final decision will be made following discussion with a third author

If possible, we will combine data from cluster-randomised controlled trials and individually randomised controlled trials for the analysis. We will consider cluster-randomised controlled trials that have not taken their design into account (i.e. have performed statistical methods that allow analysis at the level of the individual while accounting for the clustering in the data) at high risk of bias, and will perform corrected analyses where possible, if the following information can be extracted:

• the number of clusters (or groups) randomised to each intervention group; or the average (mean) size of each cluster;

• the outcome data ignoring the cluster design for the total number of individuals (for example, number or proportion of individuals with events, or means and standard deviations); and

• an estimate of the intracluster (or intraclass) correlation coefficient (ICC).

We will deal with data from studies with multiple treatment arms (i.e. in which participants have been randomised to different types of incentives) by conducting multivariate analyses, whereby we will model direct comparisons between each treatment arm and the control. We will combine data from multiple control groups (i.e. groups not offered treatment and groups offered the same treatment as the incentivised groups but without the offer of financial incentives).

We will conduct a narrative review, describing the interventions, review/study populations, review/study characteristics and the impact of financial incentives for changing the three habitual health-related behaviours of interest, namely smoking cessation, healthier eating, including alcohol consumption and physical inactivity.

Our statistical analysis will consist of a meta-regression, which will incorporate multivariate analyses for multiple treatment studies (in which participants are allocated to incentivised groups differing with respect to the type and/or size of the incentive offered), using metareg (White 2011). The analysis will involve the following stages:

Stage 1: The effect of incentives (all combined vs. control) on health-related behaviour (all combined) will be estimated through a standard meta-analysis

Stage 2: A meta-regression will be performed with behaviour type (i.e. smoking cessation, physical activity, healthier eating, weight-loss) as a covariate.

Stage 3: A meta-regression will be performed with incentive-scheme characteristics as covariates (certain vs. uncertain and value of incentive). A multivariate framework will be used for studies with multiple treatment arms in order for direct comparisons between each treatment arm and the control to be modelled (i.e. for studies with groups A’ vs. A” vs. C the multivariate framework will be used to estimate the effects of A’ vs. C and A” vs. C). Interaction terms will be included to investigate the joint effects of the incentive scheme characteristics (certain vs. uncertain and value of incentive).

Stage 4: A meta-regression will be performed with participant characteristics (i.e. level of material deprivation) and risk of bias (i.e. risk of performance and detection bias) as covariates. Behaviour type and incentive scheme characteristics will be re-entered into the model if they are found to be important predictors at stages 2 and 3 respectively.

We will calculate pooled effect sizes with 95% confidence intervals using random effects models. Given that we expect effect sizes to vary between studies according to the characteristics of the studied populations and target behaviours or related outcomes, random- as opposed to fixed-effect models are, ex ante, considered likely to be more appropriate for the purposes of this review

See Strategy for data synthesis

We will write up and submit our results for publication in a peer-review journal. We will also present our results at relevant conferences.

Eleni Mantzari

CSI Health

Health Psychology Section

Department of Psychology (at Guy's)

King's College London

5th floor Bermondsey Wing

Guy's Campus

London

SE1 9RT

eleni.mantzari@kcl.ac.uk

Centre for the Study of Incentives in Health, Health Psychology Section, King's College London

http://www.kcl.ac.uk/iop/research/centres/csihealth/index.aspx

Mrs Eleni Mantzari, Health Psychology Section, King's College London, London, UK
Dr Florian Vogt, Institute of Pharmaceutical Science, King's College London, London, UK
Mr Ian Shemilt, Behaviour and Health Research Unit, University of Cambridge, Cambridge, UK
Dr Yinghui Wei, MRC Clinical Trials Unit, London, UK
Professor Julian Higgins, MRC Biostatistics Unit, Cambridge, UK
Professor Theresa Marteau, Health Psychology Section, King's College London, London, UK

01 June 2011

20 December 2012

This research is funded by a Strategic Award in Biomedical Ethics from the Wellcome Trust; programme title: “The Centre for the Study of Incentives in Health” Grant number: 086031/Z/08/Z; PI Prof. TM Marteau

None known

English

England

Subject indexing assigned by CRD

Health Behavior; Humans; Motivation; Reward

19 July 2012

15 April 2013