This paper is in the following e-collection/theme issue:

Digital Health Reviews (1289) Prevention and Health Promotion (986) Innovations and Technology for Physical Activity Education (742) Instruments and Questionnaires for Physical Activity and Lifestyle (72)

Published on in Vol 25 (2023)

This is a member publication of University of Sheffield (Jisc)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/46622, first published .
Effectiveness of Computer-Tailored Health Communication in Increasing Physical Activity in People With or at Risk of Long-Term Conditions: Systematic Review and Meta-Analysis

Effectiveness of Computer-Tailored Health Communication in Increasing Physical Activity in People With or at Risk of Long-Term Conditions: Systematic Review and Meta-Analysis

Effectiveness of Computer-Tailored Health Communication in Increasing Physical Activity in People With or at Risk of Long-Term Conditions: Systematic Review and Meta-Analysis

Authors of this article:

Longdan Hao1 Author Orcid Image ;   Stefan Goetze2 Author Orcid Image ;   Tourkiah Alessa3 Author Orcid Image ;   Mark S Hawley1 Author Orcid Image
Article Authors Cited by (2) Tweetations (8) Metrics

    Review

    1Centre for Assistive Technology and Connected Healthcare, School of Health and Related Research, University of Sheffield, Sheffield, United Kingdom

    2Department of Computer Science, University of Sheffield, Sheffield, United Kingdom

    3Biomedical Technology Department, College of Applied Medical Sciences, King Saud University, Riyadh, Saudi Arabia

    Corresponding Author:

    Mark S Hawley, PhD

    Centre for Assistive Technology and Connected Healthcare

    School of Health and Related Research

    University of Sheffield

    The Innovation Centre

    217 Portobello

    Sheffield, S1 4DP

    United Kingdom

    Phone: 44 114 2222000

    Email: mark.hawley@sheffield.ac.uk


    Background: Regular physical activity (PA) is beneficial for enhancing and sustaining both physical and mental well-being as well as for the management of preexisting conditions. Computer-tailored health communication (CTHC) has been shown to be effective in increasing PA and many other health behavior changes in the general population. However, individuals with or at risk of long-term conditions face unique barriers that may limit the applicability of CTHC interventions to this population. Few studies have focused on this cohort, providing limited evidence for the effectiveness of CTHC in promoting PA.

    Objective: This systematic review and meta-analysis aims to assess the effectiveness of CTHC in increasing PA in individuals with or at risk of long-term conditions.

    Methods: A systematic review and meta-analysis were conducted to evaluate the effect of CTHC in increasing PA in people with or at risk of long-term conditions. Hedges g was used to calculate the mean effect size. The total effect size was pooled and weighted using inverse variance. When possible, potential moderator variables were synthesized, and their effectiveness was evaluated by subgroups analysis with Q test for between-group heterogeneity Qb. Potential moderator variables included behavior change theories and models providing the fundamental logic for CTHC design, behavior change techniques and tailoring strategies to compose messages, and computer algorithms to achieve tailoring. Several methods were used to examine potential publication bias in the results, including the funnel plot, Egger test, Begg test, fail-safe N test, and trim-and-fill method.

    Results: In total, 24 studies were included in the systematic review for qualitative analysis and 18 studies were included in the meta-analysis. Significant small to medium effect size values were found when comparing CTHC to general health information (Hedges g=0.16; P<.001) and to no information sent to participants (Hedges g=0.29; P<.001). Half of the included studies had a low to moderate risk of bias, and the remaining studies had a moderate to high risk of bias. Although the results of the meta-analysis indicated no evidence of publication bias, caution is required when drawing definitive conclusions due to the limited number of studies in each subgroup (N≤10). Message-tailoring strategies, implementation strategies, behavior change theories and models, and behavior change techniques were synthesized from the 24 studies. No strong evidence was found from subgroup analyses on the effectiveness of using particular behavior change theories and models or from using particular message-tailoring and implementation strategies.

    Conclusions: This study demonstrates that CTHC is effective in increasing PA for people with or at risk of long-term conditions, with significant small to medium effects compared with general health information or no information. Further studies are needed to guide design decisions for maximizing the effectiveness of CTHC.

    J Med Internet Res 2023;25:e46622

    doi:10.2196/46622

    Keywords

    computer-tailored health communication; meta-analysis; long-term conditions; physical activity promotion 


    Long-Term Health Conditions and Physical Activity

    A long-term health condition, also referred to as a chronic condition or chronic disease, “is a health problem that requires ongoing management over a period of years or decades” [1]. Unlike acute illnesses, long-term health conditions cannot be cured completely. However, they can be effectively controlled and managed through appropriate medical treatments, therapies, and lifestyle adjustments over the long term [1]. Long-term conditions caused more than 70% of global deaths in 2019. Seven out of the “top 10 causes of death” are long-term conditions: ischemic heart disease; stroke; chronic obstructive pulmonary disease; trachea, bronchus, and lung cancer; Alzheimer disease and other dementias; diabetes mellitus; and kidney disease [2].

    Physical activity (PA) has been shown to improve symptoms in people with long-term conditions [3-7]. However, extra barriers caused by the symptoms of long-term conditions can reduce people’s adherence to the recommended PA level [6,8]. Tailored information and personalized feedback have been identified as effective facilitators for increasing PA [9].

    Computer-Tailored Health Communication

    Computer technology has increasingly made tailoring feasible on a large scale, enabling access to a wide population [10-12]. Computer-tailored health communication (CTHC) uses computer-based platforms for individual information collection and processing, tailored information, and personalized feedback provision [10-12]. By embracing 3 different disciplines (computer science, health informatics, and behavior science) [10-12], CTHC has developed its own concepts: message-tailoring strategies and message implementation strategies [12]. Message-tailoring strategies cover two subcategories: (1) the tailoring criteria (based on different behavior change theories and models), such as the stage of change [13,14] or individual risk factors, and (2) tailoring mechanisms (what information is used and how it is used, such as demographic information for personalization and content adaptation, or performance records for feedback and training) [12]. Message implementation strategies include three subcategories: (1) general implementation strategies (how do messages depend on user assessment and tailoring iteration; will professionals be involved in message tailoring or technical support), (2) delivery modalities (eg, web technologies or traditional channels), and (3) tools for promoting behavior (eg, goal-setting tools, skill training, reminders, or monitoring tools) [12].

    The Effectiveness of CTHC in Promoting Health Behaviors

    Previous systematic reviews have suggested that CTHC can provide promising results in behavior change interventions [15,16]. For the general population, CTHC has been shown to be more effective than general (nontailored) information on dietary promotion (eg, fruit and vegetable consumption and fat intake decrease) [17-21]; mammography screening [20,22]; smoking cessation [20]; and PA [17,20]. However, Lustria et al [15] suggested that CTHC has a larger impact on behavior change in healthy populations than in people with long-term conditions. Only 1 systematic review and meta-analysis has assessed tailored messages for people with a long-term condition, type 2 diabetes, and this review found CTHC to be more effective for self-management promotion than general health information for this condition [23].

    These results indicate that CTHC may play a role in promoting greater PA in people with long-term conditions. However, the effectiveness of CTHC may vary across different populations, considering factors such as health status, age, and gender [15]. In addition, variations in effectiveness may be observed when considering different behaviors, such as dietary habits, smoking, PA, and screening tests for common diseases [15]. To the best of the authors’ knowledge, no reviews have been conducted to evaluate the effectiveness of CTHC systems on PA behavior change specifically for people with or at risk of long-term conditions; nor have reviews jointly analyzed psychological theories for behavior change and computer algorithms used in CTHC, and so there is limited information to understand how tailoring systems work and interact with individuals with long-term conditions as a whole. Given that increasing PA is a crucial aspect of self-management for individuals with long-term conditions [3], exploring the effectiveness of CTHC in this regard is valuable. Therefore, it is essential to comprehend not only the impact of CTHC on PA but also the influential factors contributing to its effectiveness.

    The primary aim of this systematic review was to synthesize evidence on the effectiveness of CTHC on PA behavior change in people with or at risk of different long-term conditions. The secondary aims were to synthesize and evaluate message-tailoring strategies, message implementation strategies, and computer algorithms used in the included studies to provide suggestions to guide future CTHC designs for people with long-term conditions.


    Inclusion and Exclusion Criteria

    The Participants, Intervention, Comparison, Outcome, Study framework [24] was used to structure the inclusion and exclusion criteria (for details, see Textbox 1).

    Textbox 1. Inclusion and exclusion criteria.

    Inclusion criteria

    1. Participants

    1.1. Adult population (aged ≥18 years).

    1.2. People with long-term conditions (in total, 103 long-term conditions [25] were identified for study selection) or people with a BMI of ≥25 kg/m2 (overweight). This review includes people with overweight and obesity since the risk of having long-term conditions increases with a BMI of ≥25 kg/m2 [26-28].

    2. Intervention

    2.1. Participants receive tailored messages to promote physical activity (PA). Tailored messages are defined as individualized and customized to fit the participants’ personal needs, status, interests, and concerns. Tailored messages follow at least one type of tailoring strategy, such as personalization (eg, mentioning the participant’s name) or feedback on PA performance. Messages are tailored according to (1) the participant’s personal characteristics, such as demographics, preferences, and physical and psychological status, or (2) behavior performance, preferences, and barriers for PA. Tailored messages can be combined with standard PA interventions including rehabilitation programs, self-management programs, and eHealth interventions. Message channels include SMS text messages, email, websites, smartphone apps, printed materials, fliers and other channels.

    2.2. Tailored messages are generated or selected by computer algorithms.

    3. Comparison

    3.1. Participants in the control group receive general information (nontailored messages) or receive no messages.

    4. Outcome

    4.1. The outcome must include at least one measure reflecting PA change. PA could be measured as activity performance (eg, steps), activity length and frequency (eg, minutes, times/week), or energy consumption.

    5. Study

    5.1. Only randomized controlled trials (RCTs) are included.

    5.2. Journal papers

    5.3. Written in English

    Exclusion criteria

    1. Participants

    1.1. Children (aged <18 years, unless national law defines a person as being an adult at an earlier age).

    1.2. Healthy population with a BMI <25 kg/m2.

    2. Intervention

    2.1. Tailored messages are targeted to promote behaviors other than PA. Messages are population based or community based (unlike individualized messages, population-based or community-based messages provide the same content to the whole targeted population or a community);

    2.2. Tailored messages are generated or selected by humans, for example, health professionals.

    3. Comparison

    3.1. Absence of a control group

    4. Outcome

    4.1. Only qualitative outcomes or indirect outcomes of PA

    5. Study

    5.1. Non-RCTs (eg, reviews, pilot studies, feasibility studies, and single-arm studies)

    5.2. Protocols, reviews, unpublished studies, conference abstracts, posters, theses, and articles

    5.3. Written in languages other than English

    Search Method

    We used search keywords in initial searches in Embase (OVID). After developing the search strategy (for details, see Multimedia Appendix 1), an adaptation was made for other databases. Studies from the earliest date in the selected databases until the search date (February 18, 2023) were included.

    Four keywords were used to develop Medical Subject Headings terms in the search strategy: (1) PA and exercise, (2) chronic disease, (3) tailored message, and (4) randomized controlled trial in OVID. The electronic databases included MEDLINE (OVID), Embase (OVID), PsycINFO (OVID), CINAHL, the Cochrane Central Register of Controlled Trials (The Cochrane Library), and the Web of Science Core Collection.

    Study Selection, Data Extraction, and Management

    After duplicate studies were removed, three steps were followed for selecting and analyzing the remaining studies: (1) title screening, (2) abstract screening, and (3) full-text screening. At the beginning of title screening, 2 assessors (LH and TA) independently screened the first 20 studies ordered alphabetically and discussed the results for a deeper understanding of the inclusion and exclusion criteria. For all studies, each assessor graded the studies independently: 0 for exclusion, 1 for not sure, and 2 for inclusion. Studies were included in the next screening if the sum of the grades was ≥2. Independent screening of the title, abstract, and full content was conducted successively. After each screening, the assessors discussed the difference in the results. If any disagreement persisted after the discussion, the final decision was made by a third assessor (MH).

    The main assessor (LH) extracted study characteristics, intervention details, and outcomes. The risk of bias was assessed by using the Cochrane risk of bias tools [29] (Risk of bias section).

    Study characteristics included the number of randomized controlled trial (RCT) arms (study design), duration, country of intervention, diagnosis of the participants, age, gender, and sample size. Intervention details included information channel, baseline assessment for message tailoring, message delivery frequency and time (not applicable for post and website log-in), message type (text, video, hyperlink, and external resources), and direction (1-way messages and 2-way messages), behavior change theories applied in the study, behavior change techniques (BCTs; based on 93 BCTs taxonomy defined by Michie et al [30]), tailoring on aspects of individuals, computer algorithms for achieving message generation and selection, and professional involvement. Outcomes included the types of PA, follow-up assessment time, and measurement tools.

    Data Collection and Analysis

    Meta-Analytic Approach

    The goal of the meta-analysis was to examine the effect of tailored messages on the level of PA. Studies were categorized based on the types of comparison groups: a nontailored control group (a control group receiving general health messages) and a nontreatment control group (a control group receiving no messages).

    The calculations were supported by Review Manager 5.4 (RevMan, provided by Cochrane). The meta-analysis approach used standardized mean difference with a random effects model and inverse-variance weighted and pooled effect size (standard mean difference between 2 comparison groups) [31] across studies. If the studies reported multiple assessment points in time for PA outcomes, the final point in time was used. This procedure is consistent with that used in other studies [20]. If studies reported multiple assessment methods for PA, objective continuous outcomes (eg, daily steps measured by pedometers) reported by most studies were chosen first to minimize heterogeneity.

    Eligible studies were analyzed using the mean difference and SD to measure the change from baseline to the last end point in the intervention period. The effect size of continuous PA outcomes with different measures was calculated together with Hedges adjusted g. This step increases the heterogeneity of the result. However, owing to the limited number of tailoring studies and nonstandard methodologies, this calculation has been widely applied in previous meta-analyses on tailoring intervention studies. The effect size was interpreted as small (Hedges g=0.2), medium (Hedges g=0.5), or large (Hedges g=0.8) [32].

    Heterogeneity in this study is presented by I2 [29]. Compared with the chi-square test for heterogeneity, I2 measures the impact of heterogeneity on the meta-analysis (the percentage of the variability in effect size that is caused by heterogeneity) [29]. Heterogeneity I2 is classified as low (0%≤I2<25%), medium (25%≤I2<75%), or high (I2≥75%) [33].

    Between-Groups Analysis Approach

    Between-groups analysis considered the main behavioral change theories and models reported in tailoring designs of included studies, the number of reported behavioral change theories, and types of outcome measurement (objective or subjective). The potential moderators for between-groups analysis were first translated into dichotomous values (presence or absence), and the average effect size was calculated when a minimum of 3 studies could be included at both levels of the moderators. A Q test for between-groups heterogeneity (Qb) was used to measure the significance of the difference between subgroups. Qb was calculated by the chi-square test with df=n–1 (n being the number of subgroups) [34,35].

    Qualitative Analysis Approach

    The qualitative analysis aimed to understand the design and implementation of tailoring interventions and the effective elements of behavior change interventions. The qualitative approach focuses on three subcategories: (1) message-tailoring strategies (behavior change theories and models and BCTs); (2) message implementation strategies (delivery channel, iteration, message type, direction, format, frequency, timing, and content); and (3) computer algorithms. Applied strategies were summarized, cumulated, and categorized if possible.

    Risk of Bias

    Risk of bias was assessed based on Cochrane Collaboration’s Risk of Bias Tools [29]. These tools are used to assess the study biases caused by the study design and execution and their impact on the findings of the study [29]. For each study, the main reviewer assessed the risk of bias using the Cochrane Collaboration’s Risk of Bias tool and criteria. The criteria evaluated various aspects of the included studies, such asselection bias, performance bias, detection bias, attrition bias, and reporting bias. Each criterion was assessed as having a low, high, or unclear risk of bias based on the information provided in the study report or other sources. For example, the selection bias criterion for a RCT would assess the method of random sequence generation, allocation concealment, and baseline comparability of the groups. The overall risk of bias for each study was then classified as low, high, or unclear based on the judgments for each criterion. Studies with a low risk of bias were considered to have a low likelihood of bias affecting the study results, whereas studies with a high risk of bias were considered to have a high likelihood of bias affecting the study results. Studies with an unclear risk of bias had insufficient information to make a definitive judgment.

    Publication Bias

    Given the small sample size and high heterogeneity of the included studies, we assessed publication bias using a range of methods, including funnel plot testing, Egger test, Begg test, the trim-and-fill method, and the fail-safe N test, as appropriate, given the known limitations of these methods. These analyses were conducted using the meta package in Stata/MP (version 17; StataCorp) [29,36].


    Description of the Studies

    Overview

    A total of 1791 studies were identified from 5 electronic databases after duplicates (n=620) were removed. Figure 1 shows the selection process based on the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [29,37]. Following title screening, 464 studies were retained for abstract screening and 93 studies were subsequently taken forward for screening of the complete article. In total, 24 studies were identified to meet the eligibility criteria for inclusion in this review. Agreement between the assessors was very high (Cohen κ=0.811). Among the included studies, 2 articles [38,39] reported results based on the same RCT but with different assessment times. Five articles [38,40-43] reported the details of the study design in separate papers. Therefore, the data reported in these separate papers were extracted if necessary. The details of PRISMA checklist is reported in Multimedia Appendix 2.

    ‎
    Figure 1. Updated version of PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram of the literature search and selection process—computer-tailored health communication intervention on increasing physical activity (PA) in people with or at risk of long-term conditions. RCT: randomized controlled trial.
    Study Characteristics

    All 24 studies [17,39,41-61] included in this review were RCTs published in English between 1999 and 2023. Details of the selected studies are summarized in Table 1. In total, 19 studies were 2-arm RCTs, 3 studies were 3-arm RCTs, and 2 studies were 4-arm RCTs. The sample sizes ranged from 36 to 1366, resulting in a total sample size of 11,754. Studies had an average duration of 10 months. The study population was aged 55.7 (SD 8.67) years on average, and 57.6% (6771/11,754) were female. All participants in the included studies had or were at risk of long-term conditions, including cardiovascular diseases (8 studies), cancer and survivors of cancer (7 studies), diabetes (8 studies), chronic obstructive pulmonary disease (1 study), and overweight or obesity (1 study).

    Table 1. Details of the 24 selected studies.
    Study and yearSample size, nCountry of settingLength (months)RCTa armsDiagnosis of participantsAge (years), mean (SD)Gender (female), %
    Agboola et al [44], 2016126United States62Type 2 diabetes51.5 (11.6)51.6
    Almeida et al [45], 2015452United States1 out of 6b4Sedentary cardiac patients58.6 (9.7)59.1
    Antypas and Wangberg [46], 201469Norway222Cardiac diseases59.1 (8.7)21.7
    Beleigoli et al [47], 20201298Brazil63BMI ≥25 kg/m233.6 (11.0)76.7
    Block et al [48], 2016339United States62Prediabetes55 (8.9)31.0
    Boudreau et al [49], 2011325Canada82Type 2 diabetes49.1 (6.6)44.3
    Broekhuizen et al [17], 2012340The Netherlands122Familial hypercholesterolemia45.9 (12.9)56.8
    Bull et al [50], 1999272United States34Overweight sedentary patients with risk of cardiovascular diseases39.6 (13.3)63.2
    Demark-Wahnefried et al [38] and Demark-Wahnefried [51], 2007543United States102Survivors of breast and prostate cancer57 (10.8)56.0
    Dobson et al [40], 2018366New Zealand92Poorly controlled diabetes47 (15.0)48.4
    Frederix et al [52], 2015140Belgium62Cardiac patients61 (8.5)18.6
    Golsteijn et al [41], 2018478The Netherlands42Patients with and survivors of prostate and colorectal cancer66.5 (7.6)13.0
    Gomersall et al [53], 201936Australia32People with cancer and survivors64.8 (9.6)36.1
    Kanera et al [42], 2017462The Netherlands122Survivors of cancer55.9 (11.4)79.9
    Khunti et al [54], 20211366England483People at risk of diabetes61 (9.1)48.8
    Migneault et al [55], 2012337United States82African American individuals with hypertension56.6 (11.0)70.3
    Ottenbacher et al [39], 2012c400United States122Survivors of breast and prostate cancer57 (11.1)56.0
    Peacock et al [62], 2020204United Kingdom62People at risk of cardiovascular disease and type 2 diabetes64 (6.0)35.8
    Rock et al [43], 2015693United States24 out of 482Any early-stage breast cancer56.3 (9)100.0
    Short et al [57], 2015330Australia33Survivors of breast cancer55 (—)d100.0
    Storm et al [58], 2016790Germany and the Netherlands82Cardiac patients50.8 (12.2)62.9
    Vluggen et al [59], 2021478The Netherlands62Patients with T2DMe60 (6.76)32.4
    Volders et al [60], 2020585The Netherlands122Older adults with chronic illnesses74.5 (6.4)48.4
    Voncken-Brewster et al [61], 20151325The Netherlands62People at risk for or diagnosed with COPDf57.6 (7.2)52.7

    aRCT: randomized controlled trial.

    bThe length of study was 6 months in total. However the reported data collected from follow-up assessment was at the end of the first month.

    c2-year follow-up study of Demark-Wahnefried [47].

    dSD is not available in the paper.

    eT2DM: type 2 diabetes mellitus.

    fCOPD: chronic obstructive pulmonary disease.

    Risk of Bias

    All included studies presented some degree of potential bias when assessed using the Cochrane Collaboration Risk of Bias Tool. Figure 2 shows the degree of risk of bias, and Figure 3 [17,38-55,57-62] shows the identified risks in each study.

    ‎
    Figure 2. Summary of risk of bias.
    ‎
    Figure 3. Risk of bias assessments for studies in a systematic review of computer-tailored health communication on increasing physical activity for people with chronic obstructive pulmonary disease. Key: +=low risk of bias; ?=unclear risk of bias; −=high risk of bias.

    In total, 12 studies were considered to have a moderate to high risk of bias [46-48,50,52-54,58-61,63], and the remaining 12 studies were considered to have a low to moderate risk of bias. All the studies reported random sequence generation, except for Volders et al [60], which did not report details on randomization. In total, 13 studies concealed the allocation until the trials started. Five studies [17,43,46,58,60] blinded the participants, and 10 studies [38,40,44,47,48,52,54,56,58,59] blinded the outcome assessors. Fourteen studies had a low attrition rate or found no difference in intention-to-treat assessment, while 7 studies had a very high attrition rate [46,47,50,54,58-60]. In total, 21 studies reported the study registration and study design. Two studies have a high risk of bias in other areas owing to the small sample size.

    Meta-Analysis on PA Outcomes

    After the study characteristics were summarized from 24 studies, 6 studies were excluded from the meta-analysis for the following reasons: Bull et al [50] and Agboola et al [44] provided different types of outcomes (dichotomous outcomes and log of risk ratio) to most studies. It was not possible to combine these 3 outcomes with other outcomes. Limited information was provided by Boudreau et al [49], Migneault et al [55], and Voncken-Brewster et al [61] to calculate the standardized mean difference between the groups. The effect size reported by Broekhuizen et al [17] was much larger than that reported in other studies and the SE was very small. This resulted in a significant increase in overall heterogeneity. Therefore, the study by Broekhuizen et al [17] was considered as an outlier and was excluded from the meta-analysis.

    A total of 18 studies were included in the meta-analysis and were categorized into two subgroups to be analyzed using forest plots as follows:

    1. Subgroup 1: intervention group with tailored messages versus control group with general messages (tailored vs general): 9 studies were included in this subgroup; the results and forest plot are shown in Table 2 and Figure 4.
    2. Subgroup 2: intervention group with tailored messages versus control group with no message (tailored vs none): 10 studies were included in this subgroup; the results and forest plot are shown in Table 3 and Figure 5.
    Table 2. Subgroup 1: forest plot of tailored versus general messagesa.
    Study of subgroup, yearExperimentalControlStandard mean difference

    		Values, mean (SD)Totalb, nValues, mean (SD)Total, nWeight (%)IVcRandom, 95% CI
    Almeida et al [45], 2015109.71 (89.33)12574.04 (89.76)1158.50.400.14 to 0.65
    Beleigoli et al [47], 20202.7 (3.13)4202.4 (2.21)47016.20.11−0.02 to 0.24
    Demark-Wahnefried et al [38] Demark-Wahnefried [51], 2007112.7 (126.6)25383.8 (119.1)26613.10.230.06 to 0.41
    Frederix et al [52], 20152077 (2357)691075 (1203)705.70.530.20 to 0.87
    Khunti et al [54], 2021−296 (2969)317−385 (2217)37314.80.03−0.12 to 0.18
    Ottenbacher et al [39], 201260 (180)17130 (150)22911.50.18−0.02 to 0.38
    Peacock et al [62], 2020161.9 (58.12)134163 (58.12)707.2−0.02−0.31 to 0.27
    Rock et al [43], 2015165 (185.2)343157 (205.2)34814.90.04−0.11 to 0.19
    Short et al [57], 20159644.8 (7745.6)1098301.2 (3373.4)1108.10.22−0.04 to 0.49
    TotalN/Ad1941N/A20511000.16e0.07 to 0.25

    aHeterogeneity: τ2=0.01 (an estimate of the between-study variance in a random effects meta-analysis [τ2]. The square root of this number [ie, τ] is the estimated SD of the underlying effects across studies); χ28=15.8; P=.05; I2=49%. Test for overall effect (Z test and P value of standardized mean difference): Z=3.43 (P<.001).

    bNumber of participants in the treatment and control group of each study.

    cIV: inverse variance.

    dN/A: not applicable.

    eOverall effect.

    ‎
    Figure 4. Subgroup 3 Forest Plot tailored message or General message.
    Table 3. Subgroup 2: forest plot of tailored versus no messagesa.
    Study of subgroup, yearExperimentalControlStandard mean difference

    		Values, mean (SD)Totalb, nValues, mean (SD)Total, nWeight (%)IVcRandom effect, 95% CI
    Almeida et al [45], 2015108.46 (89.046)10577.97 (89.0058)10310.60.340.07 to 0.62
    Antypas and Wangberg [46], 20145613 (2828)71356 (2937)112.21.40.32 to 2.48
    Block et al [48], 20161.21 (1.3607)1000.42 (1.3496)14710.90.580.32 to 0.84
    Dobson et al [40], 20183.45 (2.03)1693.48 (2.19)17211.8−0.01−0.23 to 0.2
    Golsteijn et al [41], 20181145 (883)222943 (769)21312.30.240.05 to 0.43
    Gomersall et al [53], 201917.1 (54.0477)16−50.1 (52.3672)153.71.230.45 to 2.01
    Kanera et al [42], 2017688.1 (570.6)162512.2 (452.1)20611.90.350.14 to 0.55
    Storm et al [58], 20161 (1.66)4030.34 (1.53)38713.20.410.27 to 0.55
    Vluggen et al [59], 2021833 (741)111884 (777)17711.3−0.07−0.30 to 0.17
    Volders et al [60], 2020200 (206)164206 (197)24612.1−0.03−0.23 to 0.17
    TotalN/Ad1459N/A16771000.29e0.12 to 0.46

    aHeterogeneity: τ2=0.05 (an estimate of the between-study variance in a random effects meta-analysis [τ2]. The square root of this number [ie, τ] is the estimated SD of the underlying effects across studies); χ29=43.6; P<.001; I2=79%. Test for overall effect (Z test and P value of standardized mean difference): Z=3.43 (P<.001).

    bNumber of participants in treatment and control group of each study.

    cIV: inverse variance.

    dN/A: not applicable.

    eOverall effect.

    ‎
    Figure 5. Subgroup 3 Forest Plot tailored message or No message.

    Subgroup 1 (tailored vs general messages; Table 2) showed a small statistically significant weighted mean effect size between the treatment and control groups (Hedges g=0.16, P<.001, 95% CI 0.07-0.25). In total, 9 studies with 1941 participants in the intervention groups and 2051 participants in the control groups were included. The heterogeneity [64] among these studies was medium (I2=49%).

    In subgroup 2 (tailored vs no messages; Table 3), a small to medium and statistically significant weighted mean effect size was observed between the treatment and control groups (Hedges g=0.29, P=.001, 95% CI 0.12-0.46). In total, 10 studies with 1459 participants in the intervention groups (“Total” in column 4) and 1677 participants in the control groups were included. The heterogeneity [64] in this subgroup was high (I2=79%).

    In both between-groups analyses, heterogeneity [29] was caused by the diversity of participants, different types of long-term conditions, the different tailoring strategies and implementation strategies, and the different methods of PA measurement.

    Publication Bias

    The funnel plot analysis for both meta-analyses (tailored vs general messages and tailored vs no messages interventions) did not reveal any significant asymmetry. The effect size estimate obtained from the fill-and-trim method (observed+imputed: effect size=0.139, 95% CI 0.036-0.242 for tailored vs no message interventions; effect size=0.225, 95% CI 0.003-0.447 for tailored vs general messages) did not differ significantly from the effect size estimate obtained from the meta-analysis, and the Begg test supported the null hypothesis H0 of no small study effect for both analyses (P=.59). In contrast, Egger test rejected the none hypothesis of no small study effect for both meta-analysis: for tailored versus no messages (P=.05) and for tailored versus general messages (P=.02). It is important to note that the power of all methods used to test for the risk of publication bias was low because of the limited number of studies (n=10); thus, caution is needed when drawing conclusions from the results [36] (Multimedia Appendix 3: output data for publication bias analysis).

    Message-Tailoring Strategies

    The message-tailoring strategies applied in the 24 studies were analyzed to provide suggestions for designing future tailoring studies. Behavior change theories and behavior change techniques were extracted from the 24 included studies.

    Behavior Change Theories and Models

    In total, 22 behavior change theories and models were identified in the studies. The most frequently reported theories were social cognitive theory (SCT; 9 studies) [13], the I-change model (6 studies) [65], the stage of change in the transtheoretical model (5 studies) [13,14], the health action process approach (HAPA; 4 studies) [66], and the theory of planned behavior (3 studies) [13,67]. The impact of each behavior change theory or model mentioned earlier on effect sizes was tested using Qb (Table 4). In subgroup 1, the effect size for studies using SCT was higher than that of studies without SCT. However, the opposite result was obtained in subgroup 2, and neither difference reached significance (Qb=0.42, P=.52 in subgroup 1 and Qb=0.08, P=.77 in subgroup 2). Few studies used the I-change model or HAPA in subgroup 1 for a between-groups analysis. In subgroup 2, the effect size for studies without the I-change model was higher than that for studies with the I-change model, and the difference was significant (Qb=3.75; P=.05). The effect size for studies using HAPA in subgroup 2 was higher than that of studies without HAPA. However, no difference reached significance (Qb=0.66; P=.42). Few studies have used the stage of change model and the theory of planned behavior for a between-groups analysis.

    Comparing studies with different numbers of behavior change theories or models (none or one vs multiple), no differences reached significance in between-groups analyses (Table 5).

    Table 4. Between-groups analysis on behavior change theories and models.
    Tailoring strategySubgroup 1: tailored or generalaSubgroup 2: tailored or nonea

    		Studies, nSample size, nStandard mean difference (95% CI)P valueStudies, nSample size, nStandard mean difference (95% CI)P value
    Test for subgroup differences between interventions with and without applying SCTb

    		SCT520690.19 (0.08 to 0.31).001515990.29 (0.11 to 0.47).002

    		No SCT419230.13 (−0.04 to 0.29).13515370.35 (−0.01 to 0.70).05
    Test for subgroup differences between interventions with and without applying I-CHANGE

    		I-CHANGEN/AcN/AN/AN/A415010.13 (−0.07 to 0.32).20

    		No I-CHANGEN/AN/AN/AN/A616350.45 (0.19 to 0.71)<.001
    Test for subgroup differences between interventions with and without applying HAPAd

    		HAPAN/AN/AN/AN/A312430.38 (0.14 to 0.62).002

    		No HAPAN/AN/AN/AN/A718930.25 (0.02 to 0.47).03

    aTest for subgroup differences between interventions with and without applying social cognitive theory: χ21=0.4, P=.52, I2=0% for subgroup 1; χ21=0.1, P=.77, I2=0% for subgroup 2. Test for subgroup differences between interventions with and without applying I-CHANGE: χ21=3.8, P=.05, I2=73.4% for subgroup 2. Test for subgroup differences between interventions with and without applying health action process approach: χ21=0.66, P=.42, I2=0% for subgroup 2.

    bSCT: social cognitive theory.

    cN/A: not applicable.

    dHAPA: health action process approach.

    Table 5. Between-groups analysis—using multiple behavior change theories to single and none.
    Subgroup 1: tailored or generalaSubgroup 2: tailored or nonea
    Studies, nSample size, nStandardized mean difference (95% CI)P valueStudies, nSample size, nStandardized mean difference (95% CI)P value
    Test for subgroup differences
    Multiple311590.25 (0.13 to 0.37)<.001720270.27 (0.08 to 0.46).006
    Single or none628330.11 (0.00 to 0.22).04311090.39 (−0.08 to 0.86).11

    aTest for subgroup differences: χ21=2.9, P=.09, I2=65.7% for subgroup 1; χ21=0.2, P=.64, I2=0% for subgroup 2.

    Behavior Change Techniques

    Thirty-one BCTs were identified based on the 93 BCTs taxonomy [30]. The number of identified BCTs in studies was between 3 and 12 BCTs. Most studies conducted before 2020 did not clarify the BCTs used in tailoring design; therefore, a high probability remains that the list of identified BCTs is incomplete. Hence, the applied BCTs are presented as a list in Textbox 2, instead of the accumulated numbers of use.

    Textbox 2. Summary of behavior change techniques applied in selected studies based on 93 hierarchically clustered techniques (note that the numbering follows that from the study by Michie et al [30]).

    1. Goal setting and planning

    1.1. Goal setting (behavior)

    1.2. Problem solving

    1.4. Action planning

    1.5. Review behavior goals

    1.6. Discrepancy between current behavior and goal

    1.7. Review outcome goals

    2. Feedback and monitoring

    2.2. Feedback on behavior

    2.3. Self-monitoring of behavior

    2.4. Self-monitoring of outcomes of behavior

    3. Social support

    3.1. Social support (unspecified)

    3.2. Social support (practical)

    3.3. Social support (emotional)

    4. Shaping knowledge

    4.1. Instruction on how to perform a behavior

    4.2. Information about antecedents

    5. Natural consequences

    5.1. Information about health consequences

    5.3. Information about social and environmental consequences

    5.6. Information about emotional consequences

    6. Comparison of behavior

    6.1. Demonstration of the behavior

    6.2. Social comparison

    7. Associations

    7.1. Prompts or cues

    9. Comparison of outcomes

    9.2. Pros and cons

    9.3. Comparative imagining of future outcomes

    10. Reward and threat

    10.1. Material incentive (behavior)

    11. Regulation

    11.2. Reduce negative emotions

    12. Antecedents

    12.1. Restructuring the physical environment

    12.2. Restructuring the social environment

    12.3. Avoidance or reducing exposure to cues for the behavior

    13. Identity

    13.2. Framing or reframing

    15. Self-belief

    15.1. Verbal persuasion about capability

    15.3. Building self-belief

    16. Antecedents

    16.3. Imaginary reward

    Tailoring on Different Characteristics of Individuals

    On the basis of different behavior change theories and models, studies chose to tailor messages on different characteristics of individuals. This study synthesized the individuals’ characteristics used in tailoring to provide suggestions for future CTHC studies.

    In total, 6 different categories were identified from the included studies (Textbox 3): barriers, personal characteristics, planning, disease, psychological factors, and performance. Tailoring on barriers to PA (1.1), personal goals (3.4), motives for increasing PA (5.6), stage of behavior change (5.12), and performance regarding the goal (6.7) were seen in most of the included studies.

    Textbox 3. Categorizing tailoring information on different aspects of individuals.

    Category, subgroup first layer, and subgroup second layer

    1. Barriers

    1.1. Barriers for physical activity (PA)

    2. Personal characteristics

    2.1. Demographic information

    2.1.1 Ethnic mapping

    2.2. Cultural characteristics

    2.2.1. African American voice

    2.2.2. Tone and language

    2.2.3. Content and wording

    2.3. Personal determinant

    2.4. Signed off using the first name of the coach

    2.5. Select the timing, duration and frequency of messages and reminders

    2.6. Weight

    3. Planning

    3.1. Action planning

    3.2. Coping planning

    3.3. Goal setting

    3.4. Personal goals

    4. Disease

    4.1. Cancer-coping style

    4.2. Comorbid conditions

    5. Psychological factors

    5.1. Risk

    5.1.1. Awareness of the disease risk

    5.1.2. Change in risk perception

    5.1.3. Risk factor profiling

    5.2. Cues to action

    5.3. Exercise preferences

    5.4. Motivational modifications

    5.5. Moral obligation

    5.6. Motives for increasing PA

    5.7. Motivational determinants or constructs

    5.7.1. Attitude

    5.7.2. Intrinsic motivation

    5.7.3. Self-efficacy

    5.7.4. Social influences

    5.7.5. Intention

    5.8. Self-regulation

    5.9. Pros and cons of being more physically active.

    5.10. Response efficacy

    5.11. Social support

    5.12. Stage of behavior change

    5.13. Outcome expectations

    5.14. User’s values

    6. Performance

    6.1. Awareness of one’s own performance

    6.2. Ability to prepare and execute plans to achieve goals and to overcome potential barriers

    6.3. Baseline behavioral information

    6.4. Behavioral modifications

    6.5. Daily performance steps

    6.6. PA level

    6.7. Performance regarding the goal

    Message Implementation Strategies

    This review summarizes two categories of applied message implementation strategies: (1) assessment on individuals and message iteration and (2) message delivery channel, direction, and frequency.

    Assessment and Iteration

    All the studies conducted baseline assessments for message tailoring. Baseline assessments included sociodemographic information, lifestyle (PA level, smoking status, and diet), psychological characteristics (stage of change, anxiety, PA motives, and attitude toward PA), health status (BMI and long-term conditions), and user habits regarding the use of websites. Ten studies [41,43,44,52-55,57,60,62] used objective assessment of PA (pedometer or accelerometry) and 14 studies [17,39,40,42,45-51,58,59,61] collected PA data through subjective questionnaires. In 18 studies included in the meta-analysis, using subjective PA-related data for tailoring iteration showed a larger effect size than using objective PA data for both subgroups, but neither of the between-group differences were significant (Table 6; Qb=0.69, P=.40 and Qb=0, P=.97).

    Table 6. Subgroup analysis: objective and subjective physical activity measures.
    Subgroup 1: tailored or generalaSubgroup 2: tailored or nonea
    Studies, nSample size, nStandardized mean difference (95% CI)P valueStudies, nSample size, nStandardized mean difference (95% CI)P value
    Test for subgroup differences
    Objective519430.12 (−0.02 to 0.27).1038760.29 (−0.10 to 0.68).14
    Subjective420490.20 (0.10 to 0.31)<.001722600.30 (0.10 to 0.50).004

    aTest for subgroup differences: χ21=0.7, P=.40, I2=0% for subgroup 1; χ21=0.0, P=.97, I2=0%.

    In total, 20 out of the 24 studies used iterations for message tailoring. In the studies included in the meta-analysis, only 1 study did not use iteration, so it was not possible to calculate Qb, only qualitative synthesis on the iteration is presented in this paper. In total, 50% (10/20) of the studies included in the systematic review with iteration showed a positive significant effect size and 25% (1/4) of the studies without iteration showed a positive significant effect size. One study [61] iterated tailored messages based on the determinants of PA (eg, barriers to performing PA and motivational beliefs) obtained from predesigned questions, whereas 19 studies iterated tailored messages based on individuals’ PA-related behavior data (eg, previous steps).

    Delivery Channels and Message Features

    The identified message channels include (1) mobile apps, (2) the web, (3) computer programs, (4) interactive voice response system phone calls, (5) postal printed material, (6) email, and (7) SMS text messages. Nine studies used multiple channels for information dissemination, where postal printed materials and the web were the most used channels among the studies. The forms of message content included text, audio, video, photos, graphs, and hyperlinks, where 14 studies used only text and the remaining 10 studies used multiple forms of messages. For message direction, 1-way messages (unidirectional: participants only receive messages) were applied in 16 studies, and 2-way messages (bidirectional: participants can have interactive communication with senders) were used in the remaining 8 studies. Four different message doses (sending frequency) were found in the studies: high (≥3 times a week), moderate (weekly), low (monthly or longer), and variable (participants decide when to log in for a message). Nine studies used low doses, 4 studies used moderate doses, and 4 studies used high doses. The remaining 7 studies used variable message doses.

    Message Selection Algorithms

    The information on message selection algorithms used for generating tailored content provided in the studies was limited. In total, 17 studies did not report information related to computer algorithms, and 5 studies reported that a third-party platform was involved: Overnite (tailored builder) for 4 studies [41,42,58,61] and PropeloTM platform for 1 study [53]. In 6 studies [41,42,44,46,58,61], the stage of change was reported as the decision root (the very first node in a decision tree structure) of the algorithms.


    Principal Findings

    This systematic review and meta-analysis aimed to synthesize evidence regarding the effectiveness of CTHC to promote PA for people with or at risk of long-term conditions. Overall, CTHC can be considered as effective, with a small to medium significant effect size for increasing PA among people with or at risk of different long-term conditions. By tailoring the intervention to the individual’s unique needs, preferences, and characteristics, CTHC may be able to address the barriers to, facilitators of, and motivations for PA in a more personalized manner than generic interventions. In contrast to previous suggestions [15], the effect size of CTHC for people with or at risk of long-term conditions was found to be larger than that for tailored messages (including computerized and professional-led tailoring) in promoting PA in the general population [15,16]. However, both the study by Lustria et al [15] (n=6) and our study have a limited number of included papers for individuals with or at risk of long-term conditions. Therefore, further studies on CTHC are required to strengthen our findings. CTHC’s effect size in increasing PA is comparable with that of physiotherapist-led PA interventions for adults with musculoskeletal injury, having or at risk of noncommunicable diseases [68] (standardized mean difference 0.15, 95% CI 0.03-0.27). Thus, CTHC may serve as an effective alternative or complement to traditional physiotherapist-led interventions, particularly for individuals with long-term conditions who face barriers to accessing or adhering to in-person tailored interventions. By tailoring the intervention to the individual’s unique needs and preferences, CTHC may also be able to provide a more personalized and engaging experience that encourages sustained behavior change.

    Approximately half of the included studies combined ≥2 behavior change theories and models to construct the tailoring rules. No significant difference was found in the between-groups analysis compared with studies with single or no behavior change theory or models, in contrast to the results obtained by previous studies [10,16].

    SCT and stage of change in transtheoretical model are the most commonly used theories and models in the included studies. However, in the studies included in the meta-analysis, no behavioral change theories or models were shown to play a decisive role on treatment effect.

    Goal setting and feedback and monitoring were the most commonly used tailoring strategies (22 of 24 studies). Because of the limited information available, it was not possible to determine the effectiveness of these strategies.

    For the implementation strategies, no significant difference was found between studies with subjective measurements and studies with objective measurements in either subgroup. Another implementation strategy is iteration, which was used in most of the included studies. In the qualitative analysis, the limited evidence tends to support the use of iteration in CTHC for superior outcomes.

    It is important to note that the current analysis was limited by the small number of studies included in the between-groups analysis, which may have reduced the statistical power to detect significant differences. In addition, the specific theories and models used in each study may have varied in their relevance and applicability to the target behavior and population with long-term conditions, which may have influenced the results. Further research is needed to explore the potential benefits and limitations of using different combinations of behavior change theories and models in CTHC interventions to promote PA in individuals with long-term conditions. As more studies become available, it may be possible to identify which specific tailoring components, such as the use of specific BCTs, delivery frequency, and channels, are most effective in promoting PA. This information could help optimize the design and delivery of CTHC interventions and enhance their impact on behavior change in individuals with long-term conditions.

    As noted by Mummah et al [69], it is crucial to recognize the key factors for the targeted population when designing CTHC interventions, particularly when targeting individuals with different long-term conditions. Because different behavior change theories emphasize different factors related to behavior change, it is possible that the same theory may have different effects on different populations. However, given the limited number of studies included in this meta-analysis, it is premature to draw conclusions regarding the effective factors for each specific long-term condition. Nonetheless, the factors applied in the studies included in this analysis can serve as a useful reference when designing CTHC interventions to promote PA in individuals with long-term conditions. Future research should continue to explore the underlying mechanisms and moderators of behavior change in CTHC interventions to better tailor these interventions to the needs of specific populations and optimize their effectiveness.

    The limited reporting of computer algorithms in the included studies is an important observation as it highlights the need for greater transparency and detail in reporting the technical aspects of CTHC interventions. Only 7 of the studies [41,42,44,46,53,58,61] provided information on the computer algorithms used in their CTHC interventions, and these studies primarily used rule-based systems with stage of change as the decision root for categorizing participants and selecting appropriate messages [44,46]. Although the use of the stage of change as a preferred element for message selection is consistent with the recommendations of Noar et al [70], it is important to note that there may be other factors that could also contribute to the effectiveness of CTHC interventions, such as the use of specific BCTs or the personalization of messages based on individual characteristics. Furthermore, the limited reporting of computer algorithms in the included studies may have also limited the reproducibility and scalability of CTHC interventions in other settings or populations. Without clear and detailed information on the technical aspects of CTHC interventions, it may be difficult for researchers and practitioners to adapt or replicate these interventions in other contexts.

    Strengths and Limitations

    The possible limitations of this review should be considered when interpreting the results. First, this review only included studies published in English, which increases the likelihood of omitting relevant research published in other languages. Second, the overall heterogeneity of this meta-analysis was medium to high. Studies included in this review covered different long-term conditions, which may have increased the heterogeneity of the pooled population. Third, the limited number of studies included in the meta-analysis presents a challenge for assessing publication bias. With fewer studies available, it becomes more difficult to identify possible sources of bias and draw reliable conclusions from the analysis. Therefore, caution is needed when interpreting the results, and further research with a larger number of studies is needed to confirm these findings and provide more robust evidence. In addition, nonstandard study designs for behavior change interventions, such as different measures of PA and tailoring strategies, may also increase the heterogeneity of the intervention effects among the included studies. However, as Michie et al [71] suggested, high heterogeneity is common in the meta-analysis of behavior change interventions.

    Most of the studies provided only limited information related to behavior change theories and computer algorithms. Specifically, when the studies reported that a combination of behavior change theories was used, details about how theories were combined were often not provided. These absences made it difficult to analyze the effect of tailoring features in detail.

    Despite these limitations, all the included studies were RCTs. This review assessed computer-tailored interventions on 1 behavior (PA) among different long-term conditions, which may offer a new perspective for future intervention designers when selecting effective tailoring features.

    Recommendations for Further Study

    CTHC studies on increasing PA in people with or at risk of long-term conditions are still very limited. It is still difficult to make suggestions related to CTHC design for future studies. On the basis of the limited information obtained from this review, the stage of change could be considered as the root of the decision rules for participant and message categorization. An individual’s stage of change in increasing PA can be used to match the tailored messages for the same stage initially. Within each stage, different determinants of performing PA can be promoted by emphasizing different types of information (eg, benefits of increasing PA for individuals in stage 1; behavior data compared with goals for individuals in stages 3 and 4).

    Iteration is also suggested as an advantageous implementation strategy. With iteration, future message content can be adjusted based on individuals’ responses to previous messages. In addition to the participants’ PA performance data, iterations can be combined with other tailoring strategies, such as the participants’ preferences regarding the messages’ content, participants’ preferences regarding the message sources, participants’ objective status, and inner values. Iterating only on behavior performance might still lead to a low use rate or high attrition rate if the message content is not sufficiently attractive to the recipients. Future studies could include recommendation systems supported by machine learning to match tailored messages to participants’ preferences regarding the message itself. These new approaches have drawn more attention in recent behavior change studies [68-72].

    Most studies have assessed short-term effectiveness. There were 7 studies that assessed outcomes over 12 months, and most of them obtained nonsignificant results. Future studies should assess the results over a longer period to understand the long-term effects of tailored messages.

    The lack of sufficient information related to the CTHC design was a major challenge in evaluating the effectiveness of the tailoring elements in the included studies. More detailed reporting of behavior change theories and models, BCTs, and computer algorithms for message selection in future CTHC studies would be invaluable for understanding how CTHC works in behavior change interventions.

    This analysis underscores the importance of transparent and detailed reporting of the technical aspects of CTHC interventions, particularly the computer algorithms used for message selection. Providing clear information on the decision rules, message selection criteria, and personalization methods can improve the reproducibility, scalability, and effectiveness of CTHC interventions for promoting PA in individuals with long-term conditions.

    Therefore, it is critical for future studies to prioritize the transparent reporting of CTHC design elements, as this can enable researchers and practitioners in diverse settings to adapt or replicate these interventions with greater confidence. By improving the clarity and detail of reporting, we can enhance our understanding of the underlying mechanisms and moderators of behavior change in CTHC interventions and ultimately optimize their impact on PA behaviors in individuals with long-term conditions.

    Conclusions

    Overall, the CTHC intervention was shown to be effective in increasing PA levels in people with or at risk of different long-term conditions. Most CTHC studies aimed to provide tailored PA goal setting, feedback, and suggestions based on individuals’ PA-related data. With the support of behavior change theories and models, CTHC studies also targeted individuals’ intentions, motivations, barriers, and preferences. Although no behavior change theory or model was found to be superior to others, using the stage of change as the root of decision rules for message selection could be beneficial. In terms of the message implementation strategy, iteration appears to be an effective element of CTHC. Owing to the limited information available related to tailored content design and computer algorithms, it is still challenging to synthesize or conclude which tailoring elements are effective and why. More CTHC studies are needed for a comprehensive evaluation in the future.

    Acknowledgments

    The authors would like to thank Angie Rees who provided great support on searching methodologies. The authors would also like to thank Marrissa Martyn-St James who provided valuable suggestions on the meta-analysis methodologies. This is self-funded research.

    Data Availability

    The data sets generated and analyzed during this study are available from the corresponding author upon reasonable request. However, all data were generated based on the original data from the included studies, making it easy to extract data accordingly.

    Authors' Contributions

    LH was responsible for the data collection, study design, data analysis and interpretation, and manuscript drafting. SG and MH were responsible for verifying the concept, structure, and methodology as well as proofreading this meta-analytic review. LH and TA were responsible for study inclusion and exclusion, and MH gave final suggestions when LH and TA were in disagreement. All the authors made substantial contributions to the writing of the manuscript.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    Search strategy.

    DOCX File , 15 KB
    Multimedia Appendix 2

    PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist.

    DOCX File , 29 KB
    Multimedia Appendix 3

    Publication bias analysis.

    DOCX File , 425 KB
    1. Long term physical health condition. National Health Service. URL: https:/​/www.​datadictionary.nhs.uk/​nhs_business_definitions/​long_term_physical_health_condition.​html?hl=long%2Cterm%2Ccondition [accessed 2020-03-05]
    2. The top 10 causes of death. World Health Organization. Dec 9, 2020. URL: https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death [accessed 2021-06-18]
    3. Durstine JL, Gordon B, Wang Z, Luo X. Chronic disease and the link to physical activity. J Sport Health Sci. Mar 2013;2(1):3-11. [FREE Full text] [CrossRef] [Medline]
    4. Colberg SR, Sigal RJ, Fernhall B, Regensteiner JG, Blissmer BJ, Rubin RR, American College of Sports Medicine; et al. American Diabetes Association. Exercise and type 2 diabetes: the American College of Sports Medicine and the American Diabetes Association: joint position statement. Diabetes Care. Dec 2010;33(12):e147-e167. [FREE Full text] [CrossRef] [Medline]
    5. Hayes C, Kriska A. Role of physical activity in diabetes management and prevention. J Am Diet Assoc. Apr 01, 2008;108(4 Suppl 1):S19-S23. [FREE Full text] [CrossRef] [Medline]
    6. Schwartz J, Rhodes R, Bredin SS, Oh P, Warburton DE. Effectiveness of approaches to increase physical activity behavior to prevent chronic disease in adults: a brief commentary. J Clin Med. Mar 01, 2019;8(3):124-125. [FREE Full text] [CrossRef] [Medline]
    7. Williamson J, Pahor M. Evidence regarding the benefits of physical exercise. Arch Intern Med. Jan 25, 2010;170(2):124-125. [FREE Full text] [CrossRef] [Medline]
    8. Bullard T, Ji M, An R, Trinh L, Mackenzie M, Mullen SP. A systematic review and meta-analysis of adherence to physical activity interventions among three chronic conditions: cancer, cardiovascular disease, and diabetes. BMC Public Health. May 24, 2019;19(1):636-669. [FREE Full text] [CrossRef] [Medline]
    9. Okwose NC, O'Brien N, Charman S, Cassidy S, Brodie D, Bailey K, et al. Overcoming barriers to engagement and adherence to a home-based physical activity intervention for patients with heart failure: a qualitative focus group study. BMJ Open. Sep 21, 2020;10(9):e036382-e036393. [FREE Full text] [CrossRef] [Medline]
    10. Noar SM, Harrington NG. Tailored communications for health-related decision-making and behavior change. In: Diefenbach MA, Miller-Halegoua S, Bowen DJ, editors. Handbook of Health Decision Science. New York, NY. Springer; 2016.
    11. Harrington NG, Noar SM. Reporting standards for studies of tailored interventions. Health Educ Res. Apr 2012;27(2):331-342. [CrossRef] [Medline]
    12. Lustria ML, Cortese J, Noar SM, Glueckauf RL. Computer-tailored health interventions delivered over the web: review and analysis of key components. Patient Educ Couns. Feb 27, 2009;74(2):156-173. [CrossRef] [Medline]
    13. Glanz K, Rimer BK, Viswanath K. Health Behavior and Health Education Theory, Research, and Practice. Hoboken, NJ. Wiley; 2008.
    14. Prochaska JO, Velicer WF, Rossi JS, Goldstein MG, Marcus BH, Rakowski W, et al. Stages of change and decisional balance for 12 problem behaviors. Health Psychol. 1994;13(1):39-46. [CrossRef]
    15. Lustria ML, Noar SM, Cortese J, Van Stee SK, Glueckauf RL, Lee J. A meta-analysis of web-delivered tailored health behavior change interventions. J Health Commun. Dec 12, 2013;18(9):1039-1069. [FREE Full text] [CrossRef] [Medline]
    16. Noar SM, Benac CN, Harris MS. Does tailoring matter? Meta-analytic review of tailored print health behavior change interventions. Psychol Bull. Jul 15, 2007;133(4):673-693. [CrossRef] [Medline]
    17. Broekhuizen K, van Poppel MN, Koppes LL, Kindt I, Brug J, van Mechelen W. Can multiple lifestyle behaviours be improved in people with familial hypercholesterolemia? Results of a parallel randomised controlled trial. PLoS One. Aug 2012;7(12):e50032-e50080. [FREE Full text] [CrossRef] [Medline]
    18. Celis-Morales C, Abraham S, Keenan P, Ashor AW, Livingstone K, Lara J, et al. Effect of web-based tailored lifestyle interventions on fruit and vegetable consumption in adults: a systematic review and meta-analysis of randomised controlled trials. Proc Nutr Soc. Apr 15, 2015;74(OCE1):214-221. [FREE Full text] [CrossRef] [Medline]
    19. Eyles HC, Mhurchu CN. Does tailoring make a difference? A systematic review of the long-term effectiveness of tailored nutrition education for adults. Nutr Rev. Aug 01, 2009;67(8):464-480. [FREE Full text] [CrossRef] [Medline]
    20. Krebs P, Prochaska JO, Rossi JS. A meta-analysis of computer-tailored interventions for health behavior change. Prev Med. Oct 2010;51(3-4):214-221. [FREE Full text] [CrossRef] [Medline]
    21. Teasdale N, Elhussein A, Butcher F, Piernas C, Cowburn G, Hartmann-Boyce J, et al. Systematic review and meta-analysis of remotely delivered interventions using self-monitoring or tailored feedback to change dietary behavior. Am J Clin Nutr. Feb 01, 2018;107(2):247-256. [FREE Full text] [CrossRef] [Medline]
    22. Sohl SJ, Moyer A. Tailored interventions to promote mammography screening: a meta-analytic review. Prev Med. Oct 05, 2007;45(4):252-261. [FREE Full text] [CrossRef] [Medline]
    23. Sahin C, Courtney KL, Naylor PJ, E Rhodes R. Tailored mobile text messaging interventions targeting type 2 diabetes self-management: a systematic review and a meta-analysis. Digit Health. Dec 2019;5:2055207619845279. [FREE Full text] [CrossRef] [Medline]
    24. Amir-Behghadami M, Janati A. Population, Intervention, Comparison, Outcomes and Study (PICOS) design as a framework to formulate eligibility criteria in systematic reviews. Emerg Med J. Jun 20, 2020;37(6):387-386. [FREE Full text] [CrossRef] [Medline]
    25. Taylor SJC, Pinnock H, Epiphaniou E, Pearce G, Parke HL, Purushotham N, et al. A rapid synthesis of the evidence on interventions supporting self-management for people with long-term conditions: PRISMS – Practical systematic Review of Self-Management Support for long-term conditions. Southampton (UK): NIHR Journals Library; Health Services and Delivery Research, No. 2.53. Dec 2014 Appendix 9, List of long-term conditions [FREE Full text]
    26. Global BMI Mortality Collaboration; Di Angelantonio E, Bhupathiraju S, Wormser D, Gao P, Kaptoge S, et al. Body-mass index and all-cause mortality: individual-participant-data meta-analysis of 239 prospective studies in four continents. Lancet. Aug 20, 2016;388(10046):776-786. [FREE Full text] [CrossRef] [Medline]
    27. Peeters A, Barendregt JJ, Willekens F, Mackenbach JP, Al Mamun A, Bonneux L, et al. NEDCOM‚ the Netherlands EpidemiologyDemography Compression of Morbidity Research Group. Obesity in adulthood and its consequences for life expectancy: a life-table analysis. Ann Intern Med. Jan 07, 2003;138(1):24-32. [FREE Full text] [CrossRef] [Medline]
    28. Obesity and overweight. World Health Organization. Jun 9, 2021. URL: https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight [accessed 2023-07-14]
    29. Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al. Cochrane Handbook for Systematic Reviews of Interventions. 2nd Edition. Chichester, UK. Wiley; 2019.
    30. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, et al. The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions. Ann Behav Med. Aug 06, 2013;46(1):81-95. [FREE Full text] [CrossRef] [Medline]
    31. Deeks JJ, Higgins JP, Altman DG. Chapter 10: Analysing data and undertaking meta-analyses. In: Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.3. London, UK. The Cochrane Collaboration; 2022.
    32. Cohen J. Statistical Power Analysis for the Behavioral Sciences Second Edition. New York, NY. Routledge; 1988.
    33. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. Sep 06, 2003;327(7414):557-560. [FREE Full text] [CrossRef] [Medline]
    34. Cooper H, Hedges LV, Valentine JC. The Handbook of Research Synthesis and Meta-Analysis. New York, NY. Russell Sage Foundation; Jun 14, 2019.
    35. Borenstein M. Introduction to Meta-analysis. Chichester, UK. Wiley; 2009.
    36. Dalton JE, Bolen SD, Mascha EJ. Publication bias: the elephant in the review. Anesth Analg. Oct 17, 2016;123(4):812-813. [FREE Full text] [CrossRef] [Medline]
    37. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. Mar 29, 2021;372(1):n71. [FREE Full text] [CrossRef] [Medline]
    38. Demark-Wahnefried W, Clipp EC, Lipkus IM, Lobach D, Snyder DC, Sloane R, et al. Main outcomes of the FRESH START trial: a sequentially tailored, diet and exercise mailed print intervention among breast and prostate cancer survivors. J Clin Oncol. Jul 01, 2007;25(19):2709-2718. [FREE Full text] [CrossRef] [Medline]
    39. Ottenbacher AJ, Day RS, Taylor WC, Sharma SV, Sloane R, Snyder DC, et al. Long-term physical activity outcomes of home-based lifestyle interventions among breast and prostate cancer survivors. Support Care Cancer. Oct 01, 2012;20(10):2483-2489. [FREE Full text] [CrossRef] [Medline]
    40. Dobson R, Whittaker R, Jiang Y, Maddison R, Shepherd M, McNamara C, et al. Effectiveness of text message based, diabetes self management support programme (SMS4BG): two arm, parallel randomised controlled trial. BMJ. May 17, 2018;361(11):k1959. [FREE Full text] [CrossRef] [Medline]
    41. Golsteijn BH, Bolman C, Volders E, Peels DA, de Vries H, Lechner L. Short-term efficacy of a computer-tailored physical activity intervention for prostate and colorectal cancer patients and survivors: a randomized controlled trial. Int J Behav Nutr Phys Act. Oct 30, 2018;15(1):106. [FREE Full text] [CrossRef] [Medline]
    42. Kanera IM, Willems RA, Bolman CA, Mesters I, Verboon P, Lechner L. Long-term effects of a web-based cancer aftercare intervention on moderate physical activity and vegetable consumption among early cancer survivors: a randomized controlled trial. Int J Behav Nutr Phys Act. Feb 10, 2017;14(1):19. [FREE Full text] [CrossRef] [Medline]
    43. Rock CL, Flatt SW, Byers TE, Colditz GA, Demark-Wahnefried W, Ganz PA, et al. Results of the exercise and nutrition to enhance recovery and good health for you (ENERGY) trial: a behavioral weight loss intervention in overweight or obese breast cancer survivors. J Clin Oncol. Oct 01, 2015;33(28):3169-3176. [FREE Full text] [CrossRef] [Medline]
    44. Agboola S, Jethwani K, Lopez L, Searl M, O'Keefe S, Kvedar J. Text to move: a randomized controlled trial of a text-messaging program to improve physical activity behaviors in patients with type 2 diabetes mellitus. J Med Internet Res. Nov 18, 2016;18(11):e307. [FREE Full text] [CrossRef] [Medline]
    45. Almeida FA, Smith-Ray RL, Dzewaltowski DA, Glasgow RE, Lee RE, Thomas DS, et al. An interactive computer session to initiate physical activity in sedentary cardiac patients: randomized controlled trial. J Med Internet Res. Aug 24, 2015;17(8):e206-e285. [FREE Full text] [CrossRef] [Medline]
    46. Antypas K, Wangberg SC. An Internet- and mobile-based tailored intervention to enhance maintenance of physical activity after cardiac rehabilitation: short-term results of a randomized controlled trial. J Med Internet Res. Mar 11, 2014;16(3):e77-e92. [FREE Full text] [CrossRef] [Medline]
    47. Beleigoli A, Andrade AQ, Diniz MD, Ribeiro AL. Personalized web-based weight loss behavior change program with and without dietitian online coaching for adults with overweight and obesity: randomized controlled trial. J Med Internet Res. Nov 05, 2020;22(11):e17494-e17498. [FREE Full text] [CrossRef] [Medline]
    48. Block G, Azar KM, Romanelli RJ, Block TJ, Palaniappan LP, Dolginsky M, et al. Improving diet, activity and wellness in adults at risk of diabetes: randomized controlled trial. Nutr Diabetes. Sep 19, 2016;6(9):e231. [FREE Full text] [CrossRef] [Medline]
    49. Boudreau F, Godin G, Poirier P. Effectiveness of a computer-tailored print-based physical activity intervention among French Canadians with type 2 diabetes in a real-life setting. Health Educ Res. Aug 2011;26(4):573-585. [CrossRef] [Medline]
    50. Bull FC, Kreuter MW, Scharff DP. Effects of tailored, personalized and general health messages on physical activity. Patient Educ Counsel. Feb 03, 1999;36(2):181-192. [FREE Full text] [CrossRef] [Medline]
    51. Demark-Wahnefried W. Print-to-practice: designing tailored print materials to improve cancer survivors' dietary and exercise practices in the FRESH START trial. Nutr Today. May 2007;42(3):131-138. [FREE Full text] [CrossRef] [Medline]
    52. Frederix I, Hansen D, Coninx K, Vandervoort P, Vandijck D, Hens N, et al. Medium-term effectiveness of a comprehensive internet-based and patient-specific telerehabilitation program with text messaging support for cardiac patients: randomized controlled trial. J Med Internet Res. Jul 23, 2015;17(7):e185. [FREE Full text] [CrossRef] [Medline]
    53. Gomersall SR, Skinner TL, Winkler E, Healy GN, Eakin E, Fjeldsoe B. Feasibility, acceptability and efficacy of a text message-enhanced clinical exercise rehabilitation intervention for increasing 'whole-of-day' activity in people living with and beyond cancer. BMC Public Health. Jun 03, 2019;19(Suppl 2):542-548. [FREE Full text] [CrossRef] [Medline]
    54. Khunti K, Griffin S, Brennan A, Dallosso H, Davies MJ, Eborall HC, et al. Promoting physical activity in a multi-ethnic population at high risk of diabetes: the 48-month PROPELS randomised controlled trial. BMC Med. Jun 03, 2021;19(1):130. [FREE Full text] [CrossRef] [Medline]
    55. Migneault JP, Dedier JJ, Wright JA, Heeren T, Campbell MK, Morisky DE, et al. A culturally adapted telecommunication system to improve physical activity, diet quality, and medication adherence among hypertensive African-Americans: a randomized controlled trial. Ann Behav Med. Feb 23, 2012;43(1):62-73. [FREE Full text] [CrossRef] [Medline]
    56. Peacock OJ, Western MJ, Batterham AM, Stathi A, Standage M, Tapp A, et al. Multidimensional individualised Physical ACTivity (Mi-PACT)--a technology-enabled intervention to promote physical activity in primary care: study protocol for a randomised controlled trial. Trials. Aug 28, 2015;16(7):381. [FREE Full text] [CrossRef] [Medline]
    57. Short CE, James EL, Girgis A, D'Souza MI, Plotnikoff RC. Main outcomes of the Move More for Life Trial: a randomised controlled trial examining the effects of tailored-print and targeted-print materials for promoting physical activity among post-treatment breast cancer survivors. Psychooncology. Jul 2015;24(7):771-778. [FREE Full text] [CrossRef] [Medline]
    58. Storm V, Dörenkämper J, Reinwand DA, Wienert J, De Vries H, Lippke S. Effectiveness of a web-based computer-tailored multiple-lifestyle intervention for people interested in reducing their cardiovascular risk: a randomized controlled trial. J Med Internet Res. Apr 11, 2016;18(4):e78. [FREE Full text] [CrossRef] [Medline]
    59. Vluggen S, Candel M, Hoving C, Schaper NC, de Vries H. A web-based computer-tailored program to improve treatment adherence in patients with type 2 diabetes: randomized controlled trial. J Med Internet Res. Feb 23, 2021;23(2):e18524. [FREE Full text] [CrossRef] [Medline]
    60. Volders E, Bolman CA, de Groot RH, Verboon P, Lechner L. The effect of active plus, a computer-tailored physical activity intervention, on the physical activity of older adults with chronic illness(es)-a cluster randomized controlled trial. Int J Environ Res Public Health. Apr 10, 2020;17(7):209-213. [FREE Full text] [CrossRef] [Medline]
    61. Voncken-Brewster V, Tange H, de Vries H, Nagykaldi Z, Winkens B, van der Weijden T. A randomized controlled trial evaluating the effectiveness of a web-based, computer-tailored self-management intervention for people with or at risk for COPD. Int J Chron Obstruct Pulmon Dis. Mar 9, 2015;10(2):1061-1073. [FREE Full text] [CrossRef] [Medline]
    62. Peacock OJ, Western MJ, Batterham AM, Chowdhury EA, Stathi A, Standage M, et al. Effect of novel technology-enabled multidimensional physical activity feedback in primary care patients at risk of chronic disease - the MIPACT study: a randomised controlled trial. Int J Behav Nutr Phys Act. Aug 08, 2020;17(1):99. [FREE Full text] [CrossRef] [Medline]
    63. Boudreau F, Moreau M, Côté J. Effectiveness of computer tailoring versus peer support web-based interventions in promoting physical activity among insufficiently active canadian adults with type 2 diabetes: protocol for a randomized controlled trial. JMIR Res Protoc. Feb 11, 2016;5(1):e20-211. [FREE Full text] [CrossRef] [Medline]
    64. Higgins ET. Value from regulatory fit. Curr Dir Psychol Sci. Jun 24, 2016;14(4):209-213. [CrossRef] [Medline]
    65. Mummah SA, Robinson TN, King AC, Gardner CD, Sutton S. IDEAS (Integrate, Design, Assess, and Share): a framework and toolkit of strategies for the development of more effective digital interventions to change health behavior. J Med Internet Res. Dec 16, 2016;18(12):e317. [FREE Full text] [CrossRef] [Medline]
    66. Schwarzer R. Modeling health behavior change: how to predict and modify the adoption and maintenance of health behaviors. Appl Psychol. Jan 21, 2008;57(1):1-29. [FREE Full text] [CrossRef] [Medline]
    67. Ajzen I. The theory of planned behavior. Org Behav Human Decis Processes. Dec 1991;50(2):179-211. [CrossRef]
    68. Kunstler BE, Cook JL, Freene N, Finch CF, Kemp JL, OʼHalloran PD, et al. Physiotherapist-led physical activity interventions are efficacious at increasing physical activity levels: a systematic review and meta-analysis. Clin J Sport Med. May 24, 2018;28(3):304-315. [FREE Full text] [CrossRef] [Medline]
    69. Mummah SA, Robinson TN, King AC, Gardner CD, Sutton S. IDEAS (Integrate, Design, Assess, and Share): a framework and toolkit of strategies for the development of more effective digital interventions to change health behavior. J Med Internet Res. Dec 16, 2016;18(12):e317. [FREE Full text] [CrossRef] [Medline]
    70. Noar SM, Webb EM, Van Stee SK, Redding CA, Feist-Price S, Crosby R, et al. Using computer technology for HIV prevention among African-Americans: development of a tailored information program for safer sex (TIPSS). Health Educ Res. Jun 09, 2011;26(3):393-406. [FREE Full text] [CrossRef] [Medline]
    71. Michie S, Abraham C, Whittington C, McAteer J, Gupta S. Effective techniques in healthy eating and physical activity interventions: a meta-regression. Health Psychol. Nov 24, 2009;28(6):690-701. [FREE Full text] [CrossRef] [Medline]
    72. Cheung KL, Durusu D, Sui X, de Vries H. How recommender systems could support and enhance computer-tailored digital health programs: a scoping review. Digit Health. 2019;5:2055207618824727. [FREE Full text] [CrossRef] [Medline]

    ‎
    BCT: behavior change technique
    CTHC: computer-tailored health communication
    HAPA: health action process approach
    PA: physical activity
    PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses
    RCT: randomized controlled trial
    SCT: social cognitive theory

    Edited by A Mavragani; submitted 24.02.23; peer-reviewed by N Mungoli, MMAZ Asyura; comments to author 21.06.23; revised version received 06.08.23; accepted 25.08.23; published 04.10.23.

    Copyright

    ©Longdan Hao, Stefan Goetze, Tourkiah Alessa, Mark S Hawley. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 04.10.2023.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.