related behavior (Abraham & Michie, 2008; Bartholomew, Parcel, Kok, Gottlieb, & Fernández, 2011; Craig et al., 2008). This knowl- edge is essential for understanding how to design complex inter- ventions to induce change. In addition, there is a growing need for an evidence-based instrument that can be used to evaluate and qualify the potential of existing interventions (e.g., Brug et al., 2010; Craig et al., 2008).
Various meta-analyses have shown that interventions targeted on health behavior change may be effective (Kroeze, Werkman, & Brug, 2006; Shahab & McEwen, 2009). These studies have diffi- culties in identifying which particular BCTs are responsible for heterogeneity in effectiveness of interventions. An exception is the study by Albarracin et al. (2005), which made an attempt to identify theoretically derived BCTs within HIV prevention pro- grams and showed that some technique types were more likely to effectively change behavior than others, increasing our under- standing of why variation in intervention effectiveness exists. In general, however, comparisons between BCTs in intervention ef- fectiveness studies and meta-analyses have been hampered by the lack of a systematic framework for identifying BCTs within inter- ventions.
Recent developments of taxonomies of BCTs provide frame- works that can be used to classify interventions in a systematic way. As such, they provide the possibility to systematically eval- uate theory-based BCTs within complex interventions (Abraham & Michie, 2008). A BCT taxonomy clarifies differences and similarities in content of interventions targeting similar behaviors in similar settings. It provides a detailed definition of each BCT, including essential elements. In the taxonomy of Abraham and Michie (2008), for example, Prompt intention formation is defined as “encouraging the person to decide to act or set a general goal” (see Table 1 for an overview of the taxonomy).
Some meta-analyses that have used a BCT taxonomy have succeeded in identifying BCTs that influence the effectiveness of interventions. Webb, Joseph, Yardley, and Michie (2010) exam- ined the effectiveness of Internet-based interventions using a tax- onomy adapted from Hardeman, Griffin, Johnston, Kinmonth, and Wareham (2000). They found that the two BCTs that were asso- ciated with the greatest changes in behavior were Stress manage- ment and General communication skills training. Moreover, they found that intervention effectiveness was larger when more BCTs were included. Michie, Abraham, Whittington, McAteer, and Gupta (2009) examined the effectiveness of physical activity (PA) and healthy eating (HE) interventions using the taxonomy of 26 BCTs from Abraham and Michie (2008). They showed that inter- ventions were most likely to be effective when Self-monitoring was used as a technique, or when Self-monitoring plus an addi- tional self-regulation technique were used. Using the same taxon- omy, Dombrowski et al. (2012) identified several BCTs (including Self-monitoring) with greater probability of intervention success on weight and kilocalorie consumption. In addition, Dombrowski et al. (2012) identified several BCTs that hampered intervention success, including Provide general information and Provide infor- mation on consequences. De Bruin, Viechtbauer, Hospers, Schaalma, and Kok (2009) also used the same taxonomy to code standard care in control groups of studies evaluating the effective- ness of interventions for HIV. They showed that the control groups differed greatly in number and type of BCTs. Finally, refined taxonomies have been used to successfully identify BCTs that
increase effectiveness for reducing excessive alcohol consumption (e.g., Michie et al., 2012).
When interventions use multiple BCTs, several situations may occur: (a) the effects of the BCTs are additive, (b) the effects of the BCTs cancel out, or (c) the effects of the BCTs amplify. This latter effect is the focus of our study. The amplification of effects implies that a combination of BCTs has a synergistic effect, which is also called an interaction effect. A synergistic effect occurs if the combination of two or more BCTs has a more potent effect than would be expected by their additive effect. For instance, from literature on fear appeals, it has been suggested that Fear arousal as a strategy can only be effective when also Skill information is provided (Rogers, 1995; Ruiter, Abraham, & Kok, 2001). Thus, Fear arousal and Skill information do not enhance success when applied separately, but their combined use can be quite effective (Peters, Ruiter, & Kok, 2013). Similarly, Implementation inten- tions have been suggested to be effective only when people are sufficiently motivated to engage in specific behavior (Sheeran, Webb, & Gollwitzer, 2005). Thus, Implementation intentions have to be combined with a motivation-enhancing technique to achieve success. Generally, it is expected that BCTs have synergistic effects (Malotte et al., 2000; Michie et al., 2009; Rothman, Bald- win, & Hertel, 2004), and it is considered to be important to gain an understanding on which combinations of BCTs matter (Dixon & Johnston, 2010; Michie et al., 2009). In addition, insight into the combination of techniques is essential with regard to the develop- ment of interventions. The synergistic effects of BCTs, however, generally cannot be examined by means of meta-analysis due to the lack of power in meta-regression to identify interaction effects (e.g., Michie et al., 2009). As such, only univariate or, to a lesser extent, additive effects have been examined.
In the present study, the use of classification and regression trees (CART; Breiman, Friedman, Olshen, & Stone, 1984) has been proposed to identify synergistic effects. CART is especially suit- able for data with many predictor variables that could interact. CART has been used in the field of health psychology and medical sciences, for example, to examine which combination of factors can predict cancer (van Dijk, Steyerberg, Stenning, & Habbema,
2004), or to stratify patients on disease severity (Trujillano, Badia, Serviá, March, & Rodriguez-Pozo, 2009). As far as known, CART has never been used in the field of meta-analysis, except for a small study by Dusseldorp (2001). The aim of the present study was to gain a further understanding of synergistic effects of BCTs by applying CART in a special way to meta-analytic data. This novel approach will be referred to as Meta-CART. The main objective was to examine which combinations of BCTs explain intervention success.
Method
Data from the 101 studies that were included in the meta- analysis of Michie et al. (2009) were used. The studies were published between 1990 and 2008 in peer-reviewed journals writ- ten in English. The study effect size data and the scores on the taxonomy of 26 BCTs were obtained from the authors of the meta-analysis.
พฤติกรรมที่เกี่ยวข้อง (อับราฮัมและ Michie, 2008 Bartholomew หีบห่อ กก Gottlieb และ Fernández, 2011 Craig et al., 2008) ขอบ knowl นี้เป็นสิ่งจำเป็นสำหรับการทำความเข้าใจวิธีการออกแบบที่ซับซ้อนอินเตอร์-ventions เพื่อก่อให้เกิดการเปลี่ยนแปลง นอกจากนี้ มีเครื่องมือที่ใช้หลักฐานที่สามารถใช้ในการประเมิน และรับรองศักยภาพของงานวิจัยที่มีอยู่ (เช่น Brug et al., 2010 ต้องเติบโต Craig et al., 2008)วิเคราะห์ meta ต่าง ๆ ได้แสดงให้เห็นว่า เป้าหมายของการเปลี่ยนแปลงพฤติกรรมสุขภาพการรักษาอาจมีผล (Kroeze, Werkman, & Brug, 2006 อื่นเป้าหมายหลักของ & McEwen, 2009) Diffi-culties ในระบุซึ่ง BCTs เฉพาะ heterogeneity ในประสิทธิภาพของงาน ศึกษาเหล่านี้ได้ ข้อยกเว้นเป็นการศึกษาโดย Albarracin et al. (2005), ซึ่งทำให้ความพยายามในการระบุครั้งแรกราคามา BCTs ภายในเชื้อเอชไอวีป้องกัน pro-กรัม และพบว่า บางชนิดเทคนิคมีแนวโน้มที่จะเปลี่ยนลักษณะการทำงานมากกว่าผู้อื่น มีประสิทธิภาพเพิ่มของเราภายใต้มีของสาเหตุความผันแปรในการแทรกแซงอยู่ ทั่วไป อย่างไรก็ตาม เปรียบเทียบระหว่าง BCTs ในการแทรกแซงของ ef fectiveness ศึกษาและการวิเคราะห์เมตาได้รับการขัดขวาง โดยไม่มีกรอบการทำงานระบบสำหรับการระบุ BCTs ภายใน ventions อินเตอร์พัฒนาการล่าสุดของระบบของ BCTs ให้กรอบทำงานที่สามารถใช้เพื่อจัดประเภทงานวิจัยในอย่างเป็นระบบ เช่น พวกเขาให้สามารถระบบ eval-uate BCTs ตามทฤษฎีในงานวิจัยซับซ้อน (อับราฮัมและ Michie, 2008) ระบบภาษี BCT ชี้แจงความแตกต่างและความเหมือนกันในเนื้อหาของมาตราที่กำหนดเป้าหมายลักษณะคล้ายกันในการตั้งค่าที่คล้ายกัน ให้คำนิยามรายละเอียดของแต่ละ BCT รวมทั้งองค์ประกอบที่สำคัญ ในการจำแนกประเภทของอับราฮัมและ Michie (2008), ตัวอย่าง ผู้แต่งเจตนาให้กำหนดเป็น "ส่งเสริมให้บุคคลตัดสินใจกระทำ หรือตั้งเป้าหมายทั่วไป" (ดูตารางที่ 1 ภาพรวมของระบบ)Some meta-analyses that have used a BCT taxonomy have succeeded in identifying BCTs that influence the effectiveness of interventions. Webb, Joseph, Yardley, and Michie (2010) exam- ined the effectiveness of Internet-based interventions using a tax- onomy adapted from Hardeman, Griffin, Johnston, Kinmonth, and Wareham (2000). They found that the two BCTs that were asso- ciated with the greatest changes in behavior were Stress manage- ment and General communication skills training. Moreover, they found that intervention effectiveness was larger when more BCTs were included. Michie, Abraham, Whittington, McAteer, and Gupta (2009) examined the effectiveness of physical activity (PA) and healthy eating (HE) interventions using the taxonomy of 26 BCTs from Abraham and Michie (2008). They showed that inter- ventions were most likely to be effective when Self-monitoring was used as a technique, or when Self-monitoring plus an addi- tional self-regulation technique were used. Using the same taxon- omy, Dombrowski et al. (2012) identified several BCTs (including Self-monitoring) with greater probability of intervention success on weight and kilocalorie consumption. In addition, Dombrowski et al. (2012) identified several BCTs that hampered intervention success, including Provide general information and Provide infor- mation on consequences. De Bruin, Viechtbauer, Hospers, Schaalma, and Kok (2009) also used the same taxonomy to code standard care in control groups of studies evaluating the effective- ness of interventions for HIV. They showed that the control groups differed greatly in number and type of BCTs. Finally, refined taxonomies have been used to successfully identify BCTs thatincrease effectiveness for reducing excessive alcohol consumption (e.g., Michie et al., 2012).When interventions use multiple BCTs, several situations may occur: (a) the effects of the BCTs are additive, (b) the effects of the BCTs cancel out, or (c) the effects of the BCTs amplify. This latter effect is the focus of our study. The amplification of effects implies that a combination of BCTs has a synergistic effect, which is also called an interaction effect. A synergistic effect occurs if the combination of two or more BCTs has a more potent effect than would be expected by their additive effect. For instance, from literature on fear appeals, it has been suggested that Fear arousal as a strategy can only be effective when also Skill information is provided (Rogers, 1995; Ruiter, Abraham, & Kok, 2001). Thus, Fear arousal and Skill information do not enhance success when applied separately, but their combined use can be quite effective (Peters, Ruiter, & Kok, 2013). Similarly, Implementation inten- tions have been suggested to be effective only when people are sufficiently motivated to engage in specific behavior (Sheeran, Webb, & Gollwitzer, 2005). Thus, Implementation intentions have to be combined with a motivation-enhancing technique to achieve success. Generally, it is expected that BCTs have synergistic effects (Malotte et al., 2000; Michie et al., 2009; Rothman, Bald- win, & Hertel, 2004), and it is considered to be important to gain an understanding on which combinations of BCTs matter (Dixon & Johnston, 2010; Michie et al., 2009). In addition, insight into the combination of techniques is essential with regard to the develop- ment of interventions. The synergistic effects of BCTs, however, generally cannot be examined by means of meta-analysis due to the lack of power in meta-regression to identify interaction effects (e.g., Michie et al., 2009). As such, only univariate or, to a lesser extent, additive effects have been examined.In the present study, the use of classification and regression trees (CART; Breiman, Friedman, Olshen, & Stone, 1984) has been proposed to identify synergistic effects. CART is especially suit- able for data with many predictor variables that could interact. CART has been used in the field of health psychology and medical sciences, for example, to examine which combination of factors can predict cancer (van Dijk, Steyerberg, Stenning, & Habbema,2004), or to stratify patients on disease severity (Trujillano, Badia, Serviá, March, & Rodriguez-Pozo, 2009). As far as known, CART has never been used in the field of meta-analysis, except for a small study by Dusseldorp (2001). The aim of the present study was to gain a further understanding of synergistic effects of BCTs by applying CART in a special way to meta-analytic data. This novel approach will be referred to as Meta-CART. The main objective was to examine which combinations of BCTs explain intervention success.MethodData from the 101 studies that were included in the meta- analysis of Michie et al. (2009) were used. The studies were published between 1990 and 2008 in peer-reviewed journals writ- ten in English. The study effect size data and the scores on the taxonomy of 26 BCTs were obtained from the authors of the meta-analysis.
การแปล กรุณารอสักครู่..