Study selection

In total, 23,180 records were identified through database searches. Deduplication removed 4,853 records (4,850 automatically identified by Covidence and three manually identified), leaving 18,327 records for title and abstract screening. Of these, 18,250 were excluded as irrelevant. A total of 77 full text articles were assessed for eligibility, with 49 reports excluded for the following reasons:

• • No nutrition outcomes reported or tested pre/post (n = 21);

• • Study protocols (n = 7);

• • Duplicate feasibility studies where the main study was already included (n = 7);

• • Wrong population (n = 7);

• • Wrong outcomes (n = 5); and

• • No digital component (n = 2).

A total of 28 reports met the inclusion criteria, representing 26 unique studies. Multiple reports relating to the same underlying study (n = 2) were linked and treated as a single study in the review (Figure 1).

Study characteristics

The included unique studies (n = 26), reported across 28 publications, implemented a wide range of interventions, which can be broadly grouped into three categories.

1. 1) Lifestyle interventions with a digital component, where digital tools were incorporated to enhance or support in-person programmes (n = 5)^{(Reference Aschbrenner, Naslund and Bartels30–Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}.

2. 2) Fully digital lifestyle interventions, delivered entirely through digital formats (n = 17)^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35–Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)}.

3. 3) Digital nutrition-specific interventions, which focused specifically on dietary behaviour change via digital platforms (n = 3)^{(Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54–Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}.

One study^{(Reference Lee, Waite, Piernas and Aveyard57)} offered participants a choice between digital or face-to-face delivery modes.

Intervention duration

Intervention durations ranged from shorter durations of four weeks^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Fischer55)} , six weeks^{(Reference O’Keeffe and Clarke47)}, and eight weeks^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} . Other mid-length interventions lasted from ten weeks^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)} to three months^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37,Reference Hobson43,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52,Reference Lee, Waite, Piernas and Aveyard57)} . Longer interventions ranged from six months in duration^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} to 12 months^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)} . Some studies specified the number of sessions rather than a defined duration, including either eight^{(Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42)} or 13 telephone-based sessions^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}.

Study design

RCTs were the most commonly used study design (n = 12). These included RCTs^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Wong, Ho, Shi, Tong, Chung and Yeung45)} , parallel two-arm RCTs^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} , a cluster RCT^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}, feasibility RCT^{(Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42)}, pilot RCT^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)}, non-inferiority RCT^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)}, and a mixed-method RCT^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)}. Other study designs included pilot studies (n = 3)^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Fischer55)} , feasibility studies (n = 3)^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Cheung, Dutta, Kovic, Stojcevski and Fan56,Reference Lee, Waite, Piernas and Aveyard57)} , qualitative studies (n = 2)^{(Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37,Reference O’Keeffe and Clarke47)} , and cohort studies (n = 2)^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} . Additional designs comprised mixed-methods studies (n = 2)^{(Reference Aschbrenner, Naslund and Bartels30,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , an exploratory study (n = 1)^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31)}, and a quasi-experimental quantitative interventional project (n = 1)^{(Reference Hobson43)}.

Sample sizes

Sample sizes varied considerably across the included studies, ranging from five participants in a small pilot study^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51)} to large-scale interventions with up to 761 participants^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}.

Geographical location

The studies were conducted across a range of countries. The United States accounted for the largest number of studies (n = 11)^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36–Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Hobson43,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Fischer55,Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)} , followed by Australia (n = 8)^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} . Other countries included the United Kingdom (n = 2)^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Lee, Waite, Piernas and Aveyard57)} , with the remaining countries each contributing one study: Ireland^{(Reference O’Keeffe and Clarke47)}, Sweden^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)}, the Netherlands^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}, South Africa^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}, and Hong Kong^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}.

Diagnosis

The studies included participants with a broad range of mental health diagnoses. Some focused on individuals with severe mental illness, including schizophrenia spectrum and other disorders such as bipolar^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37,Reference O’Keeffe and Clarke47,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , while other studies included broader mental health conditions such as anxiety, depression, personality disorders, and post-traumatic stress disorders^{(Reference Aschbrenner, Naslund and Bartels30,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40–Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Lee, Waite, Piernas and Aveyard57)} . Five studies focused solely on depression^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Hobson43–Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} , with only one study focusing specifically on psychosis^{(Reference Fischer55)}. One study did not specify the type of mental health diagnosis included^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}.

Age and gender

The included participants were adults across a broad age range, from 18^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Hobson43,Reference O’Keeffe and Clarke47)} to 70^{(Reference Hobson43)} years. Nearly half of all studies (n = 11) reported the mean participant age in the 40s^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Lee, Waite, Piernas and Aveyard57)} . Regarding gender distribution, most studies included more females than males^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33–Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference O’Neil, Perez, Young, John, Turner and Saunders39–Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Wong, Ho, Shi, Tong, Chung and Yeung45–Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Lee, Waite, Piernas and Aveyard57)} . Six studies had predominantly male samples^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Hobson43,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52,Reference Fischer55)} , including one study with an all-male sample^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}. Gender was not reported in one study^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51)}.

Table 1 presents a summary of the characteristics of the included studies, including country, design, population, intervention details, digital technology used, and nutrition-related outcomes measured. Full details of the technology components and nutrition outcomes are presented in Tables 2, 3, and 4, respectively. Feasibility and acceptability of included studies are presented in Table 5.

Table 1.

Characteristics of included studies

Table 1. Long description

A table with 4 columns and 100 rows. The columns are labeled Gene, Gene ID, Chromosome, and Location. The Gene column lists gene names, the Gene ID column lists corresponding gene identifiers, the Chromosome column lists chromosome numbers, and the Location column lists genomic locations. The table includes various genes, their IDs, associated chromosomes, and specific locations on those chromosomes. The data is organized in rows, with each row representing a different gene and its details.

BMI = body mass index; CVD = cardiovascular disease; F = female; GHS = Get Healthy Coaching and Information Service; M = male; MDD = major depressive disorder; N = sample size; NR = not reported; PA = physical activity; RCT = randomised controlled trial; SMI = severe mental illness; T2D = type 2 diabetes; WL = waitlist.

Table 2.

Types of digital technologies, their uses, and frequency of application

Table 2. Long description

A table with 19 rows and 4 columns. The columns are labeled Author, Type of technology, Technology use/purpose, and Frequency/duration. The table lists various studies and their use of digital technologies for health interventions, including mobile phones, social media platforms, wearable activity trackers, Zoom platforms, smartphone apps, web-based tools, and text messages. Each row details the specific technology used, its purpose, and the frequency or duration of its application in the study.

FB = Facebook, MFP = My Fitness Pal, msg = messaging, PA = physical activity.

Table 3.

Quantitative nutrition outcomes

Table 3. Long description

A table with 21 rows and 6 columns summarizing the characteristics of included studies. The columns are Country, Design, Population, Intervention details, Digital technology used, and Nutrition-related outcomes measured. Each row provides specific details for each study. For example, Row 1: Afghanistan, Cluster randomized controlled trial, Pregnant women, 1. Counseling on maternal nutrition and health, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 2: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 3: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 4: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 5: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 6: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 7: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 8: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 9: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 10: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 11: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 12: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 13: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 14: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 15: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 16: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 17: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 18: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 19: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 20: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain. Row 21: Bangladesh, Cluster randomized controlled trial, Pregnant women, 1. Nutrition education, 2. Mobile phone reminders, 1. Gestational weight gain, 2. Birth weight, 3. Infant weight gain.

CTF = CoachToFit; NR = not reported; NS = not statistically significant; I = intervention; C = control; BMI = body mass index; CI = confidence interval; SD = standard deviation; OR = odds ratio; PP = per-protocol; ITT = intention-to-treat; LMG = lifestyle modification group; WLG = waitlist group.

Table 4.

Qualitative nutrition outcomes

Table 4. Long description

A table with six columns and eight rows summarizing studies on dietary behaviors and digital technology. Columns are labeled Author, Nutrition-related themes/key findings, Measurement method, Key quotes, Authors of this review’s interpretation, and EDH. Each row presents data from different studies, detailing the author, key nutrition-related themes, measurement methods, notable quotes, interpretations by the review authors, and EDH. The table captures various aspects of dietary behavior changes, digital interventions, and their outcomes.

Table 5.

Feasibility, acceptability, measurement methods, and challenges of included studies

Table 5. Long description

A table comparing the feasibility, acceptability, measurement methods, and challenges of studies on digital technology and nutrition outcomes. The table has 25 rows and 5 columns. Column headers are Feasibility, Acceptability, Measurement Methods, and Challenges. Row 1: Byaruhanga et al, 2017, 90 percent participants agreed that the intervention was feasible, 90 percent participants agreed that the intervention was acceptable, SMS messages, Prominent barriers included limited network coverage and high cost of airtime. Row 2: Chib, et al, 2014, 90 percent participants agreed that the intervention was feasible, 90 percent participants agreed that the intervention was acceptable, SMS messages, Prominent barriers included limited network coverage and high cost of airtime. Row 3: Chib, et al, 2014, 90 percent participants agreed that the intervention was feasible, 90 percent participants agreed that the intervention was acceptable, SMS messages, Prominent barriers included limited network coverage and high cost of airtime. Row 4: Chib, et al, 2014, 90 percent participants agreed that the intervention was feasible, 90 percent participants agreed that the intervention was acceptable, SMS messages, Prominent barriers included limited network coverage and high cost of airtime. Row 5: Chib, et al, 2014, 90 percent participants agreed that the intervention was feasible, 90 percent participants agreed that the intervention was acceptable, SMS messages, Prominent barriers included limited network coverage and high cost of airtime.

ADePT = Acceptability of Digital Health Interventions with People; AIM = Acceptability of Intervention Measure; CEQ = Credibility/Expectancy Questionnaire; CTAP = Contextualised Technology Adaptation Process; FIM = Feasibility of Intervention Measure; IAM = Intervention Appropriateness Measure; IQR = interquartile range; M = mean; MI = motivational interviewing; NR = not reported; SD = standard deviation; SMS = Short Message Service; SUS = System Usability Scale; WLG = waitlist group.

Risk of bias assessment

Risk of bias (see supplementary file 4) was assessed according to study design. RCTs (n = 12)^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44–Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} were assessed using the Cochrane Risk of Bias 2 tool^{(Reference Higgins, Altman, Gotzsche, Juni, Moher and Oxman58)}. Five studies were rated as low risk of bias^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)} , three were rated as having some concerns^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} , and four were judged to be at high risk of bias^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)} . Non-randomised studies (n = 2)^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} were assessed using the Newcastle–Ottawa Scale^{(Reference Wells, O Shea, Peterson, Welch, Losos and Tugwell59)} and rated as high quality. Five qualitative studies or qualitative components ^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37,Reference O’Keeffe and Clarke47,Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)} were assessed using the CASP checklist⁽⁶⁰⁾ and were all rated as high quality. Seven feasibility, pilot, or preliminary efficacy studies/components were not formally assessed, as their primary focus was feasibility, acceptability, or preliminary efficacy rather than intervention effectiveness. Risk of bias due to missing results (reporting bias) was not formally assessed, no statistical syntheses or meta-analyses were conducted.

1. i. What digital technologies are being utilised to support interventions aimed at improving nutrition-related outcomes in adults with mental health conditions?

A wide variety of digital technologies was employed across the included studies to support nutrition-related outcomes for individuals with mental health conditions, with many interventions incorporating multiple modes of delivery. Common technologies included mobile phones^{(Reference Aschbrenner, Naslund and Bartels30,Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Hobson43,Reference O’Keeffe and Clarke47,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , telephones^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Lee, Waite, Piernas and Aveyard57)} , smartphone applications^{(Reference Hobson43,Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Fischer55)} , web-based platforms^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Lee, Waite, Piernas and Aveyard57)} , social media platforms^{(Reference Aschbrenner, Naslund and Bartels30,Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)} , wearables^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} , and video conferencing tools^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference O’Keeffe and Clarke47,Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)} . Mobile phones were frequently used for text messaging to provide daily support, tailored feedback^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)} , deliver personalised messages supporting nutrition and health behaviours^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , as well as sending reminders^{(Reference Fischer55)} to encourage adherence and support^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}, or supplementing monthly in-person coaching^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}. Smartphone applications were used for the delivery of interventions^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} , for self-monitoring of dietary intake via MyFitnessPal^{(Reference Hobson43,Reference Fischer55)} , and to support health education and self-monitoring ^{(Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)}. Telephone calls were commonly integrated to support behaviour change^{(Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Hobson43,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)} , deliver coaching^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} , or facilitate assessment check-ins^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)}.

Social media platforms, particularly Facebook, were used alongside other components to facilitate peer support^{(Reference Aschbrenner, Naslund and Bartels30,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)} . Wearable devices such as Fitbits^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}, pedometers^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)}, and activity trackers^{(Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} supported self-monitoring of physical activity. Video conferencing tools such as Zoom facilitated the delivery of remote interventions including Food4Thought^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}, Choices4Health^{(Reference O’Keeffe and Clarke47)}, and CALM^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)}. Web-based tools and platforms were widely used; some studies used structured web-based tools or applications to deliver their interventions^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Lee, Waite, Piernas and Aveyard57)} , while others used modular web-based programmes such as WebMOVE^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)} or self-guided platforms^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)} . The frequency of these digital components varied widely across studies. Several interventions involved daily engagement, such as daily text messages^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)} , daily Fitbit tracking^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}, and daily self-monitoring via smartphone apps^{(Reference Hobson43,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Fischer55)} ; while a substantial amount of studies followed a weekly format^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38–Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Wong, Ho, Shi, Tong, Chung and Yeung45–Reference O’Keeffe and Clarke47,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Cheung, Dutta, Kovic, Stojcevski and Fan56,Reference Lee, Waite, Piernas and Aveyard57)} . Two studies offered fortnightly or monthly telephone calls during maintenance phases^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36)} , while others scheduled coaching calls or check-ins over several months^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)} . Table 2 displays the types of technology used, their intended purposes, and the frequency of their use across the included studies.

1. ii. What is the effectiveness of digital interventions in enhancing nutritional outcomes for adults with mental health conditions?

The included studies assessed a range of primary and secondary nutrition outcomes, using both self-report and objective measures. Primary outcomes most commonly focused on dietary intake and diet quality^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Fischer55)} ; however, other studies also assessed nutrition knowledge^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}, lifestyle behaviours (that included unhealthy diet)^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)}, a nutrition domain (addressing questions on awareness and benefits of a nutritious diet)^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}, nutrition self-efficacy and stage of behaviour change ^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , and health risk behaviours (including dietary behaviours associated with increased health risk)^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40)}.

Findings on dietary intake and diet quality were mixed across the included studies. A few interventions reported no significant changes, including those that reported no significant change in dietary intake^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32)}, fruit and vegetable intake^{(Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)} , dietary risk categories^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)}, or food frequency measures^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}. Schizophrenia and diabetes Mobile-Assisted Remote Trainer (SMART) interventions also showed no significant improvements in the measured dietary outcomes, with no significant changes in fruit and vegetable intake in the pilot study^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51)} or Mediterranean Diet Adherence Screener (MEDAS) scores in the subsequent study^{(Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)}. Only two interventions reported positive effects. Fruit and vegetable intake improved^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40)}, and significant improvements in dietary change scores (P < 0.001)^{(Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} from baseline to follow-up/across timepoints, were found.

A lifestyle therapy programme demonstrated significant improvements in diet quality compared to psychotherapy, in both per-protocol (P = 0.001) and intention-to-treat analyses (P < 0.001)^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)}. Additional measures from individual studies showed mixed results. One study reported improvements in diet assessment, dietary motivation, and diet quality, though changes did not reach statistical significance^{(Reference Fischer55)}. Nutrition knowledge was also examined in a separate study, which found no significant improvement following intervention^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}. A third study, however, found that the nutrition domain of a broader health-promoting behaviour measure showed significant improvement from baseline to week nine in the intervention group compared to control (P < 0.05)^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}.

Secondary outcomes mainly focused on weight^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31–Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52,Reference Lee, Waite, Piernas and Aveyard57)} and BMI^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Hobson43,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} . Other outcomes included waist circumference^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)} , attempts to change behaviour^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}, readiness to change behaviour^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}, and loss of control eating^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}. Several interventions reported clinically significant weight loss from baseline to six months and 12 months (P = 0.010)^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}, as well as short-term reductions (−4.1 kg) at 12 weeks^{(Reference Lee, Waite, Piernas and Aveyard57)}. In the CoachToFit intervention, mean weight and BMI differences at 12 months were not significant; however, participants had greater odds of achieving ≥5% and ≥7% weight loss compared to usual-care participants^{(Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)}. One study demonstrated significant weight loss from baseline to six months to 12 months, but the maintenance phase was not significant^{(Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36)}. Significant weight loss was also observed in participants with low symptom severity and high SMS (short message system) engagement (P < 0.001)^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}. Participants achieving ≥5% weight loss or improved fitness contributed more interactions in the Facebook group versus participants who did not, with this difference approaching statistical significance (P = 0.060)^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31)}. Medium-sized treatment effects were reported in another intervention^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}, and descriptive data from one study^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)} showed weight declining from 87.4 kg (standard deviation (SD) 19.8) at baseline, to 82.6 kg (SD 17.1) at follow-up; although statistical significance was not reported for this study.

Conversely, some interventions reported minimal or non-statistically significant changes in weight, including one where intervention group weight changed minimally from 91.55 kg (SD 24.7) at baseline to 91.16 kg (SD 23.61) at six months^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}. Similarly, the SMART studies reported minimal or no changes in weight and BMI, with no significant anthropometric or metabolic changes reported in the 12-week intervention^{(Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)}. In another study^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)} comparing lifestyle therapy versus psychotherapy, weight changes at eight weeks were not significant in either per-protocol (P = 0.099) or intention-to-treat (P = 0.114), and one study reported no difference in weight between groups (P = 0.960)^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32)}.

BMI outcomes were likewise mixed. Significant improvements were observed in one study (P < 0.001)^{(Reference Hobson43)}, while another reported no statistically significant change at six months (P = 0.380) or at 12 months (P = 0.600)^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}. The WebMOVE intervention demonstrated a significant BMI change from baseline to six months (P = 0.001) compared to MoveSMI and usual care, which both showed no significant change^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)}. One study reported changes across BMI categories (18–24: healthy weight; 25–29: overweight; and ≥30: obese), although statistical significance was not reported^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}. In other studies, small changes were observed over time in both intervention and control groups^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)} , with one showing slight reductions^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32)} and another showing slight increases^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}. Medium-to-large effects were observed in a separate study^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}, while minimal changes with no statistical significance (P = 0.300) were reported in another intervention^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}.

Waist circumference scores varied across studies. One study reported gradual reductions from baseline (101.5 cm) to follow-up (94.4 cm), although statistical significance was not reported^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}, while another reported small reductions in both intervention and control group^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32)}. The SMART pilot also showed a small reduction in waist circumference from 116.40 cm to 116.00 cm; however, statistical significance was not tested due to the small sample size^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51)}. A further study reported no statistically significant change^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}. Behavioural outcomes showed positive effects. Attempts to change behaviours were greater in intervention participants compared to controls (odds ratio (OR) = 3.7, 95% confidence interval (CI) 1.03–13.23); however, there were no significant differences at six months (P = 0.440)^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}. Similarly, readiness to change dietary behaviour increased in the intervention group at six months (40% versus 23%, P = 0.049) and 12 months (40% versus 20%, P = 0.023) compared to the controls^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}. Loss of control eating did not show significant interactions with treatment settings (P = 0.150) or Loss of Control Eating Score (LOCES) (P = 0.900)^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}. Across the studies, primary and secondary nutrition outcomes were assessed using a variety of methods, which are reported in Table 3, alongside the quantitative nutrition outcomes reported across the included studies.

Qualitative nutrition outcomes

A small number of studies explored nutrition-related outcomes qualitatively (n = 8)^{(Reference Aschbrenner, Naslund and Bartels30,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference O’Keeffe and Clarke47,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53,Reference Fischer55,Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)} . Key findings highlighted the value of sharing health goals and making healthy choices at the same time (‘being in the same boat’)^{(Reference Aschbrenner, Naslund and Bartels30)}. The benefits of tracking and viewing their health^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40)} and lifestyle^{(Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)} behaviours (diet, physical activity, and sleep) were the most popular topics in message exchanges in two studies. Participants also reported increased awareness of current dietary habits, improved dietary choices, and were motivated to improve the quality of their diet^{(Reference Fischer55)}. Similarly, participants receiving the SMART intervention described increased awareness and knowledge of healthy eating, greater motivation to make healthier choices, improved portion control, increased fruit and vegetable intake, and reductions in unhealthy food or sugar consumption ^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)} . Digital technologies were reported to facilitate engagement by helping to maintain social connections and reducing the anxiety around meeting in person^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}.

Additional themes included being ready and not overburdened, gaining knowledge and implementing it, and perceiving the intervention as appropriate and effective^{(Reference O’Keeffe and Clarke47)}. However, one study^{(Reference Fischer55)} identified some key barriers to engagement, including lack of family support, unhealthy options being easier than healthier options, living situation, lack of availability/motivation to cook healthy meals, other obligations, and side-effects of medications. Additionally, in one study participants noted a preference for more personalised text messages^{(Reference Fischer55)}.

Qualitative measurement methods varied across studies and included focus group interviews^{(Reference Aschbrenner, Naslund and Bartels30)}, thematic analysis of text messages^{(Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)}, thematic analysis of interviews^{(Reference O’Keeffe and Clarke47,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)} , interviews analysed inductively^{(Reference Fischer55)}, open-ended questions^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40)}, and listening and information sessions^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}. Table 4 summarises the qualitative nutrition outcomes, key themes/findings, measurement methods, key quotes, and these authors’ interpretation of the eight included studies.

1. iii. Are nutrition interventions supported by digital technology feasible and acceptable for individuals with mental health conditions?

Most included studies reported on some aspects of feasibility and/or acceptability of digital technology for nutrition-related outcomes for individuals with mental health conditions. Table 5 reports the feasibility, acceptability, measurement methods, and challenges of the included studies.

Measurements

The included studies used a wide range of methods to assess feasibility and acceptability of the interventions, with findings largely focused on recruitment, retention, session attendance, and engagement with digital tools. Quantitative measures included a variety of questionnaires. These included the group climate questionnaire form^{(Reference Aschbrenner, Naslund and Bartels30)}, pre/post surveys^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}, the Acceptability of Intervention Measure (AIM) and Intervention Appropriateness Measures (IAM)^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , Feasibility of Intervention Measure (FIM)^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , five-question satisfaction survey after intervention based on the Contextual Technology Adaption Process (CTAP^{)(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}, System Usability Scale (SUS)^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} , Credibility Expectancy Questionnaire (CEQ)^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}, process evaluation questionnaire^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}, and the Usability, Satisfaction, and Ease of use questionnaire (USE)^{(Reference Fischer55)}. Other quantitative measures included session attendance^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31–Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference O’Neil, Perez, Young, John, Turner and Saunders39,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Lee, Waite, Piernas and Aveyard57)} , level of adherence to the intervention^{(Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41)}, text message responses^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Aschbrenner, Naslund, Gill, Bartels and Ben-Zeev37)} , follow-up rates^{(Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Looijmans, Jörg, Bruggeman, Schoevers and Corpeleijn41,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen50)} , dropout rates^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35,Reference Wong, Ho, Shi, Tong, Chung and Yeung45)} , use of platforms^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31,Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54,Reference Fischer55)} , modules completed^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Wong, Ho, Shi, Tong, Chung and Yeung45,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)} , and completion of programme logbooks^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}. Some studies also used qualitative aspects, such as thematic analysis on interviews^{(Reference O’Keeffe and Clarke47,Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53,Reference Lee, Waite, Piernas and Aveyard57)} and open-ended feedback^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}.

Feasibility

Recruitment feasibility was mixed across interventions. One study^{(Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42)} reported 43% recruitment, with 77% of participants completing ≥4/8 sessions, while another^{(Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46)} highlighted the role of clinician involvement in enhancing recruitment, although missing data at follow-up presented challenges. The 12-week SMART intervention reported a recruitment rate of 67.4%, with adherence of 93% and engagement remaining ≥90% across most of the intervention^{(Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)}. The CoachToFit intervention also reported substantial engagement, with 58% completing all modules, 84% completing ≥50% of modules, and participants completing a median of 25 peer-specialist calls^{(Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49)}. One study demonstrated high initial uptake but declining longer-term engagement^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}, whereby 75% of participants attended at least one lifestyle session and 70% attended an exercise class in the first six months, which declined to only 42% attending an exercise session between months 7 and 12^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}. Despite the decline, digital engagement remained high in this study: 66% of participants used Facebook, while text messaging was utilised by 78% of PeerFIT participants (group lifestyle + mHealth) and 91% of those in the BEAT (one-on-one mHealth coaching)^{(Reference Aschbrenner, Naslund, Gorin, Mueser, Browne and Wolfe33)}. Similarly, another study found that 76% of participants joined their secret Facebook group, with higher weekly session attendance among Facebook users compared to non-users^{(Reference Naslund, Aschbrenner, Marsch, McHugo and Bartels31)}.

Retention rates varied across studies. The SMART intervention reported 92.9% retention and 6.9% attrition ^{(Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)}. In one study^{(Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36)}, retention ranged from 91% at six months to 85% at 12 months, with higher session attendance significantly associated with weight loss. Another study^{(Reference Lee, Waite, Piernas and Aveyard57)} reported 94% retention (16 out of 17) of participants over the study period (12 weeks). Retention was found to be higher in a lifestyle intervention (78%) compared to psychotherapy (70%), with nearly 80% of participants completing three or more sessions^{(Reference O’Neil, Perez, Young, John, Turner and Saunders39)}. In another study^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)} low dropout rates were observed, although intervention usage varied widely. Across interventions, SMS response rates exceeded 80% in one programme^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}, and app uptake was as high as 83.6% in another^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}. Despite strong initial engagement, some challenges were noted, such as only 58% of participants submitting logbooks^{(Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)}. In another study, although app uptake was high, 38.6% of participants did not complete the study, with most dropouts (83%) occurring within the first three months^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}.

Some interventions used specific benchmarks to evaluate feasibility. The NEEP intervention reported that it met targets for uptake (62.5%) and engagement (53.3%) but not for food diary completion^{(Reference Fischer55)}. In the four-week SMART pilot, module-level response rates ranged from 75% to 95%, exceeding the predefined ≥50% threshold for good adherence^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51)}. Another intervention reported better feasibility for its online WebMOVE intervention, with 31% completing all modules compared to none in the MoveSMI group (in-person)^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)}. Feasibility was also supported by the intervention design process, with collaboration between developers and the Genesis Club (non-profit clubhouse) being attributed to programme feasibility^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}. Feasibility was evaluated using the ADePT framework in one study, which concluded that short, digitally supported interventions may be both feasible and effective^{(Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)}.

Acceptability

Acceptability was generally high across interventions, with participants reporting favourable perceptions of digital technology on nutrition-related outcomes. Several studies^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34,Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42,Reference Fischer55)} demonstrated strong participant satisfaction, with one reporting that all participants rated the service as good or excellent^{(Reference Kelly, Baker, Fagan, Turner, Deane and McKetin42)}, or reported general satisfaction (92%)^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)}. The NEEP intervention met benchmarks for acceptability, with over 83% finding it useful, 75% believing it met their needs, and 83.3% agreeing or strongly agreeing that they were satisfied by the intervention^{(Reference Fischer55)}. Similarly, the SMART interventions demonstrated high acceptability and usability, with high scores across the AIM, IAM, and SUS scales^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51,Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)} . Participants particularly valued the personalisation and practical lifestyle information provided, and described the intervention as accessible, relevant, supportive, and encouraging ^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51–Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)}.

Interventions delivered in group formats were mixed in terms of acceptability. In one study^{(Reference Aschbrenner, Naslund and Bartels30)} participants reported a high level of shared purpose and active participation in the in-person group sessions (92%). Conversely, in another study participants disliked the in-person group format^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38)}. Remotely delivered programmes were well received^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)}, highly acceptable, broadly liked, not perceived as burdensome^{(Reference O’Keeffe and Clarke47)}, and were easy to use and consistent^{(Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40)}. One study highlighted the value of tailored support^{(Reference Lee, Waite, Piernas and Aveyard57)}. Acceptability was also assessed using validated measures, with high scores reported on the AIM and the IAM scales^{(Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48)}. In contrast, one study found no significant changes in intervention credibility from baseline to week 9^{(Reference Wong, Ho, Shi, Tong, Chung and Yeung45)}.

Challenges

Several studies reported challenges that may limit engagement. Common barriers included technology access^{(Reference Young, Cohen, Goldberg, Hellemann, Kreyenbuhl and Niv38,Reference Thornton, Kay-Lambkin, Tebbutt, Hanstock and Baker40,Reference O’Keeffe and Clarke47,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} , engagement and adherence issues^{(Reference Holt, Gossage-Worrall, Hind, Bradburn, McCrone and Morris32,Reference Green, Yarborough, Leo, Yarborough, Stumbo and Janoff36,Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44,Reference Kolaas, Axelsson, Hedman-Lagerlöf and Berman46,Reference Hanly, Campbell, Bartlem, Dray, Fehily and Colyvas48,Reference Chinman, Wang, Dodge, Frank, McCoy and Cohen49,Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)} , including high attrition^{(Reference Temmingh, Claassen, van Zyl, Carrara, Dayakalashe and Myer35)}, non-usage attrition^{(Reference Young, Mohebbi, Staudacher, Berk, Jacka and O’Neil54)}, and low use of food/exercise diaries^{(Reference Young, Drew, Kay-Lambkin, Collins, Callister and Kelly44)}. Other challenges were social or contextual, such as preference for more in-person interaction, or how emphasis on specific dietary requirements can hinder engagement^{(Reference O’Keeffe and Clarke47)}. Lack of access to kitchen facilities^{(Reference Cheung, Dutta, Kovic, Stojcevski and Fan56)} and technical issues including errors with SMS responses^{(Reference Nicol, Jansen, Haddad, Ricchio, Yingling and Schweiger34)} were two other reported challenges. Mobile phone credit/data barriers were also reported in the SMART studies, alongside preferences for greater interactivity and additional resources ^{(Reference Arnautovska, Ritchie, Soole, Menon, Korman and Milton51–Reference Arnautovska, Ritchie, Soole, Baker, Korman and Bialkowski53)}. Limited engagement with hyperlinks was also identified^{(Reference Arnautovska, Ritchie, Korman, Menon, Milton and Varnfield52)}.