The role of intervention mapping in designing disease prevention interventions: A systematic review of the literature

Objective To assess the role of Intervention Mapping (IM) in designing disease prevention interventions worldwide. Methods Systematic search and review of the relevant literature—peer-reviewed and grey—was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines. Findings Only five of the twenty two included studies reviewed were RCTs that compared intervention using IM protocol with placebo intervention, and provided the outcomes in terms of percentage increase in the uptake of disease-prevention programmes, and only one of the five studies provided an effect measure in the form of relative risk (RR = 1.59, 95% CI = 1.08–2.34, p = 0.02). Of the five RCTs, three were rated as strong evidences, one as a medium evidence and one as a weak evidence, and they all reported statistically significant difference between the two study groups, with disease prevention interventions that have used the intervention mapping approach generally reported significant increases in the uptake of disease-prevention interventions, ranging from 9% to 28.5% (0.0001 ≤ p ≤ 0.02), On the other hand, all the 22 studies have successfully identified the determinants of the uptake of disease prevention interventions that is essential to the success of disease prevention programmes. Conclusion Intervention Mapping has been successfully used to plan, implement and evaluate interventions that showed significant increase in uptake of disease prevention programmes. This study has provided a good understanding of the role of intervention mapping in designing disease prevention interventions, and a good foundation upon which subsequent reviews can be guided.


Conclusion
Intervention Mapping has been successfully used to plan, implement and evaluate interventions that showed significant increase in uptake of disease prevention programmes. This study has provided a good understanding of the role of intervention mapping in designing PLOS

Introduction
Background Health promotion entails the use of both educational and environmental interventions to improve conditions of living favourable to health. [1] Different health promotion intervention models have been used in the past, such as the logic model described by Kirby and associates [2] and the PRECEDE/PROCEED model described by Green and Kreuter. [1] The logic model provides a background of risk behaviours and their determinants in the at-risk groups. It also depicts the environmental factors and their determinants that impact directly or indirectly on the risk behaviours, as well as identifying characteristics that distinguish between effective and ineffective programmes. [2] The PRECEDE component involves an analysis of the behavioural and environmental determinants of health and their correlates. While the PROCEED component involves the development, implementation and evaluation of a health promotion programme. [1] Intervention mapping (IM) is a health promotion protocol for selecting and applying social and behavioural science theories, such as theories of health psychology, to the planning, implementation and evaluation of health promotion programmes. [3] Intervention mapping is not a new health promotion theory, it is a framework that tries to bridge the gap between theories and practice, because despite the wide range of social and behavioural science theories, their use in the development and implementation of health promotion programmes has been a constraint to planners. [4] According to Kok and Mesters, the concept of intervention mapping, came into practice because of the problems encountered from using existing models such as: [5] "trying to change behaviour that was not related to the problem, trying to change determinants for behaviours that were not relevant to the behaviour, trying to change individual behaviour while environmental factors were responsible, trying to apply change methods that were never shown to be effective, trying to implement programmes by health professionals that were inadequately trained to do so, and so forth".
These led to a retrospective review of prototype health promotion programmes to come up with the intervention mapping protocol which describes a health promotion intervention in six steps as shown in Table 1. Although the process is defined by a series of steps, the planning process is iterative and cumulative rather than linear, with programmers moving in both directions as new themes and concepts are evolving, and each step depends on findings of the preceding step. [3] Step one above is based on the PRECEDE model described by Green and Kreuter [1] and is done before starting the actual intervention plan. It entails an assessment of the at risk population by reviewing all relevant literature related to the problem area, as well as collecting new information from the community though interviews and focus groups to have a better understanding of attitudes, beliefs and behaviours. [3] Step two forms the basis for the intervention by stating who and what is expected to change at both individual and ecological levels by stating the performance objectives and their determinants to produce the matrices of proximal performance objectives. [3] Each of the performance objectives has determinants, for example, self-efficacy is a determinant for choosing not to have sex and confidence is a determinant for negotiating condom use. [6] Step three entails the selection of theory informed methods and practical strategies. An intervention method is a process of using theory-based approach to change behaviour or environmental conditions, [3] for example, social learning theory (Bandura) has been used as the basis for methods improving self-efficacy and skills, theory of reasoned action (Fishbein & Ajzen) for changing subjective normative belief and behavioural beliefs, health belief model (Rosenstock) for changing knowledge, and risk perception. [6] Step four basically involves description of the programme protocols and contexts by participants, as well as pilot-testing the strategies and materials. It provides the vehicles for conveying the program design to producers. [3] In step five, programme sustainability is also considered in addition to adoption and implementation. In step six, the evaluation plan (which starts in step one, and developed together with the intervention map) is finalized. [3] This involves both process and effect evaluation. Process evaluation basically assess the fidelity of implementation, while effect evaluation assesses the impact of the intervention in the target population. [3] This can be done by randomizing intervention population with non-intervention populations to assess the effect of the intervention. [3,7] The intervention mapping framework has been used to successfully adapt health promotion programmes. [6,8] When adapting a programme from one population to another using the intervention mapping protocol, each of the tasks of all the steps described in Table 1 must be considered in terms of the following: what is still relevant to and can be maintained in the new population, what needs to be added for the new population, what needs to be deleted as inappropriate for the new population and what needs to be deleted or adapted as impractical in the new population. [6] This is also achieved by reviewing all the relevant literature, collecting new data from the new population through individual interviews and focus groups, as well as key informant interviews with major stakeholders. [6] Despite its widespread use in designing health promotion and disease prevention programmes, little evidence exists on the magnitude of the role IM plays in promoting uptake of disease prevention interventions. Therefore, this review aims at critically appraising the literature to assess the role of intervention mapping in designing disease prevention interventions worldwide.

Search strategy and selection criteria
The selection criteria was jointly decided and agreed by the two reviewers based on the study objective. This research is limited to studies that used intervention mapping in disease-specific prevention interventions, as shown in Table 2.
The PRISMA guidelines for reporting systematic reviews was followed to allow for systematic reporting. The PRISMA checklist is shown as File A in S1 Supporting Information. [9] Search for published literature was conducted in three main electronic databases: MEDLINE, EMBASE and Web of Science. It was initially designed using MEDLINE on 8th August 2014 and then adapted to the other databases on the same date using their specific terms. This allowed for testing of precision and sensitivity. The Campbell Collaboration and the Cochrane Library were also searched for any available and/or on-going systematic reviews on the topic. Search for grey literature (mainly unpublished research) was carried out on the following databases: OpenGrey and NYAM Grey Literature Report. A very broad search was initially conducted, which aimed at identifying all intervention mapping studies on health promotion and disease prevention programmes. The titles and abstracts of all the identified studies were read to select those that used intervention mapping in disease-specific prevention interventions. After going through series of modifications, two main search categories were identified; "intervention mapping" related and "disease prevention" related. This gave the broadest search that captured all the relevant studies. The choice of disease prevention related terms was guided by a listing of areas of health promotion that intervention mapping has been applied, as stated on the intervention mapping website.
[10] However, this underwent a series of iterative process leading to modifications as recommended in the PRISMA guidelines. [11] The Boolean operators AND/OR and truncation were used to link words and to identify all the possible endings of the search terms respectively as shown below: Intervention Both indexed and free-text terms for intervention mapping were searched for. Details of the search strategy are attached as File B in S1 Supporting Information.

Data extraction and quality of studies
All the identified studies were transferred to Endnote X7 and duplicates were removed. Ineligible studies were removed after reading the abstracts, while the full texts of the eligible studies were obtained and read for further ascertainment of eligibility or otherwise. Data extraction was carried out independently by the two authors, and compared for consistencies.
Critical appraisal. This involved an assessment of methodological quality as well as the risk of bias in the eligible studies. The appraisal status was not meant to ascertain eligibility status of the studies, but to identify strengths and weaknesses. Since the intervention mapping protocol designs and assesses the effect of an intervention, a combination of (mainly) the quality assessment tool for quantitative studies developed by the Effective Public Health Practice Project (EPHPP) and (to a lesser extent) the Centre for Evidence-Based management (CEBMa) checklist for surveys were considered appropriate, because the last step of the intervention mapping framework involves an evaluation of the intervention effect in quantitative terms. The EPHPP tool uses eight items to assess the quality of quantitative studies. [12] Each of these eight items is rated as strong, moderate or weak to collectively give rise to an overall methodological ratings of strong (four strong ratings with no weak ratings), moderate (less than four strong ratings and one weak rating) or weak (two or more weak ratings) evidence.
[13] Details of the quality appraisal are attached as File C in S1 Supporting Information. Results of the critical appraisal were used to guide construction of the tool for data extraction and synthesisAuthors were contacted where information was not clear or when documents were not available.

Literature search and study selection
Outcome of the literature search and selection of studies is summarized in Fig 1. The Cochrane library and the Campbell collaborations were searched for existing systematic reviews but none was found. Since most of the studies did not provide information on the effects of the interventions, authors were contacted to obtain that information. Out of the 16 authors contacted, only 6 responded at the time of compiling his review. Two authors provided the documents, [14,15] while the other four said the intervention effect trials were not yet published. [16][17][18][19] Study characteristics A total of 22 studies were included in this review. Three studies were cluster RCTs, [20][21][22] and two are individual RCT. [23,24] Twelve studies were conducted in the Netherlands, [14,17,21,22,[25][26][27][28][29][30][31] five were conducted in the U.S.A, [15,16,19,32,33] two were conducted in Taiwan, [23,34] one was conducted in Tanzania, [35] and one was conducted in both Tanzania and South Africa. [18]  Seven studies were on HIV/STI prevention, [18,19,[25][26][27][28]35] one was on chlamydia prevention, [30] one was on Hepatitis B virus (HBV) prevention, [29] four were on influenza prevention, [17,21,22,31] four were on cervical cancer prevention, [15,16,23,34] one was on breast & cervical cancer prevention, [33] one was on secondary stroke prevention, [32] and one was on physical activity related injury prevention. [14] Byrd

Assessment of risk of bias
Since only five studies were RCTs out of which only one provided an effect estimate in form of relative risk, and also considering the high possibility of heterogeneity, the risk of publication bias could not be assessed using funnel plot. However, the possibility of selection bias, the study design, possibility of confounders, blinding, data collection methods, withdrawal rate, intervention integrity and robustness of analysis, as contained in the EPHPP tool, were assessed to rate the methodological qualities of the studies as described in Appendix 3.

Heterogeneity
Even though all the studies used intervention mapping to design the respective disease prevention interventions, there is still a high possibility of heterogeneity (both clinical, methodological and statistical) because most of them differ in the following ways; epidemiologic design, sampling technique, study population, disease categories, methods of data collections and methods/robustness of data analysis. However, heterogeneity was not formally assessed because meta-regression could not be performed due to incomplete data, and sub-group analysis stands the danger of over interpretation.

Results of individual studies and synthesis of results
A summary of study characteristics is presented in Table 3, and the summary of findings is provided afterwards, while the adapted appraisal tool is summarized in Table 4.

Summary of findings
Only five of the twenty two included studies provided the effects in terms of percentage increase in the uptake of prevention programmes, and only one of the five studies provided an effect measure (between interventions using IM protocol and interventions using placebo with standard care) in the form of relative risk (RR = 1.59, 95%CI = 1.08-2.34, p = 0.02). All the five studies were RCTs, three of which were rated as strong evidences, one as a medium evidence and one as a weak evidence. These studies are: 9% increase in the uptake of influenza vaccine (RR = 1.59, 95%CI = 1.08-2.34, p = 0.02), [22] 10.9% increase in the uptake of mammography & a 15.9% increase in the uptake of pap-smear test, [33] 19.7% increase in the uptake of pap-smear test (p = 0.002), [23] 28.5% increase in the uptake of pap-smear screening test (p<0.0001), [24] and a 50% reduction in physical activity-related injury among low active children (HR, 0.47; 95% CI, 0.21-1.06). [20] However, all the studies have identified the participants-tailored and theory-driven determinants of uptake of the respective disease prevention interventions, which is essential (or even a pre-requisite) to the success of any disease prevention programme. Results are presented by intervention/disease type: Needs assessment data was obtained from a questionnairebased study from 11 determinants associated with influenza vaccine compliance were obtained using a multivariable analysis, with odds ratios ranging from 1.7 to 28.9. Both qualitative and quantitative programme evaluation data were obtained using a web-based questionnaire in the following season, but detailed analysis not described The effect in the intervention (IM) clusters relative to the control clusters was not given.  In-depth interviews, focus groups and quantitative surveys were conducted to obtain data on needs assessment. A cluster randomized controlled trial was conducted to assess the effect of the intervention, but described in a separate paper. Need assessment data was obtained from focus groups and in-depth interviews. However, details of intervention data collection and analysis were not provided in this paper.
The behavioural outcomes identified were: abstinence, condom use among sexually active, and healthy dating/ relationship. Effect evaluation of the programme outcomes after implementation was not provided in this paper.
(Continued) Both qualitative and quantitative methods were used for needs assessments. HIV specialist nurses were used for programme implementation and data collection, but details and analysis were not provided.
Programme effect evaluation was planned to be presented in a separate paper.

Aaro et al, 2006.[18]
Prevention of HIV: Promotion of condom use and delaying onset sexual debut, the SATZ intervention.
Students aged 12-14 years in Dar es Salaam (Tanzania), Cape Town and Polokwane (South Africa). 24-30 schools (3000-5600 students) were selected from each study site and randomly allocated to intervention and control groups (cluster randomization).
Authors demonstrated good understanding and application of IM in developing a theory based promotion of condom use and delaying sexual debut, intervention, following the six steps Data were collected using questionnaires at baseline, immediately after the intervention and after one year. Using a cluster effect 5.5% gave a power 80%, acceptable loss to follow up of 20% and required at least 11 pairs of schools. Therefore, 12, 13 and 15 pairs were respectively used for the three study sites.
Results of the evaluation were not provided (Continued) Intervention mapping in designing disease prevention interventions HIV/STI prevention. Uptake of screening tests was found to be associated with attitude, self-efficacy, perceived norms of partners/friends/parents, perceived susceptibility, shame, pros, and characteristic test site accessibility. While the performance objectives for condom use were found to be; decide to use condom, obtain/buy condom, always carry condom along, be confident to negotiate using condom with your partner by communicating and persuading, agree to use condom or not to have sex, use condom correctly and persist on using condom for every act of sexual intercourse. [18,19,[25][26][27][28][29][30]35]  Individually randomized controlled trial of papsmear screening for cervical cancer.
Chinese women aged 30 years and above (or younger if married), in Taiwan. Study population (424) was obtained from relatives of inpatient and randomly allocated to intervention and control groups.
Details of the IM framework application was described in different paper (Hou et al, 2004) The primary outcome is screening behaviour (uptake) and intention in the following year assessed in a survey using pretested and evaluated instruments. Data was collected in both arms at baseline and after three months. Chi squared test was used to compare groups, while t-test and linear regression were used to analysed the mean scores of the secondary outcomes obtained on 5 point Likert scale.
51.2% of women in the intervention group and 31.5% in the control group reported having a pap-smear test within 3 months post intervention (p = 0.002). However, no significant difference in intention to take a pap-smear test between the two groups (p = 0.31). https://doi.org/10.1371/journal.pone.0174438.t003 Intervention mapping in designing disease prevention interventions Intervention mapping in designing disease prevention interventions Influenza prevention. The most important determinants of influenza vaccine uptake among HCWs were found to be: longer opening hours, more test locations, use of mobile carts, written policy, active request, working for more than 15 years, perceived personal risks, perceived reduction of risk to patients, awareness of the existence of guidelines and the influence of media attention on avian influenza. [17,21,31] Cervical and breast cancer prevention. The determinants of screening uptake were found to be; knowledge, perception of susceptibility, perceived pros and cons, cultural norms, physician referral, insurance coverage, access & regularity of care, cost, flexibility of place-of work policy, embarrassment & discomfort, fatalism, language barrier and fear of outcome & confidentiality. [15,16,33,34] Secondary stroke prevention. Determinants of prevention were found to be; the need for provider (discharge) check-off list, clinical reminders, training and education on risk factors & local resources, stroke support groups, IEC materials and administration support. [32] Physical activity-related injury prevention. Determinants were found to be; type of sport played (contact/no contact), weather, time of season/time of day, playing surface, equipment (protective/footwear), rules, previous injury, age, sex, fitness level and psycho-social factors. [14] Discussion Intervention mapping is now widely being used to design disease prevention interventions worldwide. It has been applied in a wide range of health promotion programmes including communicable and non-communicable disease preventions, as well as general health promotion. Most studies on intervention mapping reported that it has been found to be useful in designing disease prevention programmes, with varying degrees across different disease categories and populations. While some of the variations could be explained by differences in study designs and study populations, they also portray the need for methodological improvements in the use of intervention mapping to design prevention programmes.
Even though a meta-analysis could not be performed because only one study reported an effect estimate with confidence interval, an attempt is made to summarise the findings of the reviewed literature. All the five randomised controlled studies reported statistically significant difference between the IM intervention and placebo control groups, with the IM group Intervention mapping in designing disease prevention interventions associated with an increase in the uptake of disease-prevention intervention ranging from 9% to 28.5% (0.0001 p 0.02), and one study reported a 50% decrease in the incidence of physical activity-related injury among low active children (HR, 0.47; 95% CI, 0.21-1.06) in the IM group. On the other hand, all the 22 studies have successfully identified the determinants of the uptake of disease prevention interventions, which is essential to the success of disease prevention programmes.
The findings of this review show that uptakes of influenza vaccine among healthcare workers, mammography for breast cancer screening and pap-smear test for cervical cancer screening among sexually active women, as well as reduction in physical activity related injury among low active school children can be improved by designing disease prevention programmes using the intervention mapping protocol. [20,[22][23][24]33] Since most of the identified determinants of the uptake of these prevention programmes are potentially modifiable, health planners can target to encourage (or discourage) their uptake through all the known effective ways, such as education, training and even incentives. This can be said with some level of certainty for pap-smear screening for breast cancer because three of the reviewed studies conducted in different populations are significantly associated with increased uptake. In the case influenza vaccine uptake, mammography for breast cancer and prevention of physical activity related injuries that were reported by only one study each, more work needs to be done to increase reliability of the findings. In the study on the prevention of physical activity related injury, the overall finding shows no statistically significant difference between the intervention (IM) and control groups, but a sub group analysis showed a 50% reduction in the incidence of injuries among the low active students in the intervention group. This shows that the outcome of interventions can be influenced by some specific characteristics of the study population, hence the need to design participants' tailored interventions with detailed sub-group analyses.
Restricting the literature search to only English language published literature was a limitation, as this may limit the inclusion of useful evidence, thus possibly introducing some form of selection bias which makes generalization difficult. The study was also limited by the fact that 15 of the 22 studies reviewed were rated as weak evidences and 4 as medium evidences, thus making the evidence less reliable. The possibility of publication bias resulting in over-representation of the positive effects of interventions could not be ruled out, because studies with positive effects are more likely to be published and vice versa. However, this could not be assessed because only one study provided an effect estimate.

Implications for practice, future research and policy
Even though the review process has some limitations, and the methodological qualities of most of the reviewed literature was also low, recommendations can be made to improve the design and implementation of intervention mapping on disease prevention programmes: Most of the studies only described the development of disease prevention interventions using intervention mapping, but did not provide details of the epidemiological processes such as sampling techniques, methods of data collection and analysis, study design etc. Therefore, future studies on intervention mapping should take these into account in order to improve the methodological quality and validity of the studies. As much as possible, quantitative outcomes and effect estimates should be provided and where they are published in different articles, titles should refer to the previous articles. This is because the two articles obtained by contacting authors did not contain the key search term 'intervention mapping' in their titles and abstracts, hence, not captured by the search. Subsequent reviews on the effects of intervention mapping on disease prevention should, at the onset, focus on evaluation trials of interventions developed using intervention mapping (because they provide the effects needed) not on studies that just described how intervention mapping is used to develop a disease prevention intervention. However, there may be need to contact authors of the latter to get information on the former. There is need for larger reviews that would include all the relevant literature, which should be conducted in a more ideal setting and following all the principles and guidelines for conducting a systematic review. This will be needed to make some policy recommendations. There is a need to create global awareness and training on the use of intervention mapping in disease prevention, so that more research would be conducted in different parts of the world, which would add to the existing database. This is because most of the published studies were conducted in Europe and USA.

Conclusion
Despite the widespread use of intervention mapping in designing disease prevention interventions, little evidence exist on magnitude of the role IM plays in promoting uptake of disease prevention interventions. IM has been successfully used to plan, implement and evaluate interventions that showed significant increase in uptake of disease prevention programmes. This study has found that disease prevention interventions that have used the intervention mapping approach have generally reported significant increases in the uptake of disease prevention programs.
This implies that it can be recommended for designing such interventions with some level of certainty. However, these findings should be interpreted with caution in making generalization because of the limitations of this review. Nonetheless, this has provided an insight on the role of intervention mapping in designing disease prevention interventions, and a good foundation upon which subsequent reviews can be planned and conducted. It is also recommended that the use of IM to promote primordial and secondary prevention should be reviewed.