Program methodology

The most common F&V prescription intervention model was F&V vouchers (n 19), which provide vouchers to participants to be redeemed for fresh fruit and vegetables.^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Briefel, Chojnacki and Gabor31–Reference Bryce, Wolfson and Cohen33,Reference Buscail, Margat and Petit35–Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference La Barba and Weber Cullen44,Reference Richie47,Reference Saxe-Custack, LaChance and Hanna-Attisha50–Reference Xie, Price, Curran and Østbye54)} The second most frequent model was delivery and/or collection (n 8), with fresh F&V delivered to participants or made available for collection.^{(Reference Aiyer, Raber and Bello26,Reference Black, Vally and Morris28,Reference Black, Vally and Morris29,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Kerr, Barua and Glantz43,Reference Orsega-Smith, Slesinger and Cotugna45,Reference Palar, Napoles and Hufstedler46,Reference York, Kujan, Conneely, Glantz and Kerr55)} Lastly, two studies used a cashback rebate on F&V purchases, via EBT card and associated with enhancement to SNAP benefits.^{(Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49)} In addition, one study employed a combination of these models.^{(Reference Bowling, Moretti, Ringelheim, Tran and Davison30)}

Another variable within program methodology is provision of education, ranging from simple distribution of information, cooking tips and recipes via handouts^{(Reference Aiyer, Raber and Bello26,Reference Briefel, Chojnacki and Gabor31,Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37)} (n 3), to cooking classes^{(Reference Bryce, Guajardo and Ilarraza32)} (n 1), to regularly scheduled education sessions led by nutrition professionals^{(Reference Cavanagh, Jurkowski and Bozlak38,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49)} (n 3). The most common approach (n 14) was to employ a combination of formats.^{(Reference Black, Vally and Morris28–Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Bryce, Wolfson and Cohen33–Reference Buscail, Gendreau and Daval36,Reference Fertig, Tang and Dahlen40,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference La Barba and Weber Cullen44,Reference Orsega-Smith, Slesinger and Cotugna45,Reference Slagel, Newman and Sanville51–Reference Trapl, Smith and Joshi53)}

Providers of both prescriptions and produce varied depending on the program. The most common provider of F&V prescriptions (n 14) was community health centres,^{(Reference Aiyer, Raber and Bello26,Reference Black, Vally and Morris28,Reference Black, Vally and Morris29,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Cavanagh, Jurkowski and Bozlak38,Reference Freedman, Choi, Hurley, Anadu and Hébert41–Reference Richie47,Reference Saxe-Custack, LaChance and Hanna-Attisha50,Reference Snailer, Painter and Duncan52)} followed by non-governmental organisations, other non-profit, and charitable foundations^{(Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Ferdinand, Torres, Scott, Saeed and Scribner39,Reference Fertig, Tang and Dahlen40,Reference Palar, Napoles and Hufstedler46,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49,Reference Slagel, Newman and Sanville51,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 9). Other providers included federal/state programs^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Briefel, Chojnacki and Gabor31,Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Trapl, Smith and Joshi53)} (n 4) and schools^{(Reference Burrington, Hohensee, Tallman and Gadomski34)} (n 1). The produce mainly came from one of three sources: Farmers’ markets^{(Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Ferdinand, Torres, Scott, Saeed and Scribner39,Reference Freedman, Choi, Hurley, Anadu and Hébert41,Reference Richie47–Reference Trapl, Smith and Joshi53)} (n 12), grocery stores^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27–Reference Black, Vally and Morris29,Reference Briefel, Chojnacki and Gabor31,Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Fertig, Tang and Dahlen40,Reference La Barba and Weber Cullen44,Reference Xie, Price, Curran and Østbye54)} (n 8), or food banks and pantries^{(Reference Aiyer, Raber and Bello26,Reference Orsega-Smith, Slesinger and Cotugna45,Reference Palar, Napoles and Hufstedler46)} (n 3), with several (n 7) utilising a combination.^{(Reference Burrington, Hohensee, Tallman and Gadomski34–Reference Buscail, Gendreau and Daval36,Reference Cavanagh, Jurkowski and Bozlak38,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Kerr, Barua and Glantz43,Reference York, Kujan, Conneely, Glantz and Kerr55)}

Population characteristics

The majority of F&V prescription interventions exclusively included adults^{(Reference Aiyer, Raber and Bello26,Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Cavanagh, Jurkowski and Bozlak38–Reference Freedman, Choi, Hurley, Anadu and Hébert41,Reference Kerr, Barua and Glantz43,Reference La Barba and Weber Cullen44,Reference Palar, Napoles and Hufstedler46,Reference Richie47,Reference Slagel, Newman and Sanville51–Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 17), 30 % targeted families^{(Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Briefel, Chojnacki and Gabor31,Reference Burrington, Hohensee, Tallman and Gadomski34–Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Orsega-Smith, Slesinger and Cotugna45,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49)} (n 9), and children made up the target population for the smallest proportion of the studies^{(Reference Black, Vally and Morris28,Reference Black, Vally and Morris29,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Saxe-Custack, LaChance and Hanna-Attisha50)} (n 4). Exactly half of the studies targeted populations with a specific health status, most commonly type 2 diabetes^{(Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Freedman, Choi, Hurley, Anadu and Hébert41,Reference Kerr, Barua and Glantz43,Reference Snailer, Painter and Duncan52,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 7), followed by overweight and obesity^{(Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49)} (n 3), others included hypertension^{(Reference Trapl, Smith and Joshi53)} and HIV-positive status.^{(Reference Palar, Napoles and Hufstedler46)} Participation in a government assistance program was required in 27 % (n 8) of studies,^{(Reference Black, Vally and Morris28–Reference Briefel, Chojnacki and Gabor31,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Ferdinand, Torres, Scott, Saeed and Scribner39,Reference Xie, Price, Curran and Østbye54)} although all of the studies described one or more indicators of low income or food insecurity measurement.

Other trends were seen within the target populations; however, these could not be collated as frequencies due to the differences in reporting between studies. Target populations were majority female,^{(Reference Aiyer, Raber and Bello26,Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Buscail, Margat and Petit35,Reference Buscail, Gendreau and Daval36,Reference Ferdinand, Torres, Scott, Saeed and Scribner39,Reference Fertig, Tang and Dahlen40–Reference La Barba and Weber Cullen44,Reference Richie47–Reference York, Kujan, Conneely, Glantz and Kerr55)} and majority non-white,^{(Reference Aiyer, Raber and Bello26–Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Cavanagh, Jurkowski and Bozlak38,Reference Ferdinand, Torres, Scott, Saeed and Scribner39,Reference Freedman, Choi, Hurley, Anadu and Hébert41–Reference La Barba and Weber Cullen44,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49,Reference Saxe-Custack, LaChance and Hanna-Attisha50,Reference Trapl, Smith and Joshi53–Reference York, Kujan, Conneely, Glantz and Kerr55)} where specified.

Study designs

There was considerable heterogeneity in the design of the different studies, with three main study designs used. The first being non-randomised observational study designs (n 24);^{(Reference Aiyer, Raber and Bello26–Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Bryce, Guajardo and Ilarraza32,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Cavanagh, Jurkowski and Bozlak38,Reference Ferdinand, Torres, Scott, Saeed and Scribner39–Reference Saxe-Custack, LaChance and Hanna-Attisha50,Reference Snailer, Painter and Duncan52–Reference York, Kujan, Conneely, Glantz and Kerr55)} the second, randomised controlled trials (n 5);^{(Reference Briefel, Chojnacki and Gabor31,Reference Bryce, Wolfson and Cohen33,Reference Buscail, Margat and Petit35–Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37)} and the third, non-randomised clinical trial (n 1).^{(Reference Slagel, Newman and Sanville51)}

Outcomes measured

Change in dietary quality was measured in the majority of studies^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Black, Vally and Morris29,Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Buscail, Margat and Petit35,Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Fertig, Tang and Dahlen40–Reference Palar, Napoles and Hufstedler46,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49,Reference Saxe-Custack, LaChance and Hanna-Attisha50,Reference Slagel, Newman and Sanville51,Reference Trapl, Smith and Joshi53)} (n 17), only one of which reported no statistically significant improvement in diet quality.^{(Reference Burrington, Hohensee, Tallman and Gadomski34)} The number and type of instruments used to measure diet quality varied widely, as did their levels of reliability, these included food frequency questionnaire (FFQ)^{(Reference Buscail, Margat and Petit35)} (n 1), validated 24-hour recall methods^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Black, Vally and Morris29,Reference Fertig, Tang and Dahlen40,Reference Slagel, Newman and Sanville51)} (n 4), HEI scores^{(Reference Basu, Akers, Berkowitz, Josey, Schillinger and Seligman27,Reference Slagel, Newman and Sanville51)} (n 2), and nutritional biomarkers^{(Reference Black, Vally and Morris29)} (n 1). Several of the studies used externally validated screeners to measure F&V intake, including NCI dietary screener^{(Reference Cabili, Briefel, Forrestal, Gabor and Chojnacki37,Reference Freedman, Choi, Hurley, Anadu and Hébert41,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi49)} (n 3), the Block screener^{(Reference Saxe-Custack, LaChance and Hanna-Attisha50)} (n 1), and the Behavioural Risk Factor Surveillance System (BRFSS)^{(Reference Jones, VanWassenhove-Paetzold and Thomas42)} (n 1), in addition to other instruments not often cited in the literature^{(Reference La Barba and Weber Cullen44,Reference Palar, Napoles and Hufstedler46,Reference Trapl, Smith and Joshi53)} (n 3). Internally designed, non-validated surveys were used to assess F&V intake in the remainder^{(Reference Bowling, Moretti, Ringelheim, Tran and Davison30,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Kerr, Barua and Glantz43,Reference Orsega-Smith, Slesinger and Cotugna45)} (n 4).

Outcomes related to health status were also measured in eleven of the studies,^{(Reference Black, Vally and Morris28,Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Cavanagh, Jurkowski and Bozlak38,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Kerr, Barua and Glantz43,Reference Palar, Napoles and Hufstedler46,Reference Richie47,Reference Snailer, Painter and Duncan52,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} with all but one^{(Reference Xie, Price, Curran and Østbye54)} reporting a statistically significant improvement in at least one health measure. All outcome measures were conducted by a trained medical or research staff, or retrieved from medical records, none relied on self-report. The different health outcome measures included frequency of sick visits to healthcare provider and/or hospital attendances^{(Reference Black, Vally and Morris28)} (n 1), number of oral antibiotic prescriptions^{(Reference Black, Vally and Morris28)} (n 1), iron status^{(Reference Black, Vally and Morris28)} (n 1), weight^{(Reference Bryce, Guajardo and Ilarraza32,Reference Kerr, Barua and Glantz43,Reference Richie47,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 4), BMI^{(Reference Black, Vally and Morris28,Reference Cavanagh, Jurkowski and Bozlak38,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Palar, Napoles and Hufstedler46,Reference Richie47,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 7), blood pressure^{(Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Kerr, Barua and Glantz43,Reference Richie47,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 6), glycated haemoglobin (HbA1c) level^{(Reference Bryce, Guajardo and Ilarraza32,Reference Bryce, Wolfson and Cohen33,Reference Kerr, Barua and Glantz43,Reference Palar, Napoles and Hufstedler46,Reference Snailer, Painter and Duncan52,Reference Xie, Price, Curran and Østbye54,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 7), waist circumference^{(Reference Kerr, Barua and Glantz43,Reference Richie47,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 3), glycaemic control (continuous glucose monitor and fasting blood sugar measurements)^{(Reference Kerr, Barua and Glantz43,Reference Palar, Napoles and Hufstedler46,Reference Richie47)} (n 3), and antiretroviral therapy adherence^{(Reference Palar, Napoles and Hufstedler46)} (n 1).

Food security status was also an outcome of interest in eleven of the studies,^{(Reference Aiyer, Raber and Bello26,Reference Briefel, Chojnacki and Gabor31,Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Buscail, Gendreau and Daval36,Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Kerr, Barua and Glantz43,Reference Orsega-Smith, Slesinger and Cotugna45,Reference Palar, Napoles and Hufstedler46,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference Saxe-Custack, LaChance and Hanna-Attisha50,Reference York, Kujan, Conneely, Glantz and Kerr55)} with only one study not reporting a statistically significant improvement in food security status.^{(Reference Saxe-Custack, LaChance and Hanna-Attisha50)} There was less variation in the measurement instruments compared to diet quality measurements, with the majority using the widely used USDA 18-item Household Food Security Survey^{(Reference Briefel, Chojnacki and Gabor31,Reference Buscail, Gendreau and Daval36,Reference Palar, Napoles and Hufstedler46,Reference Ridberg, Bell, Merritt, Harris, Young and Tancredi48,Reference York, Kujan, Conneely, Glantz and Kerr55)} (n 5). Others used the 2-item^{(Reference Aiyer, Raber and Bello26)} (n 1) and 6-item^{(Reference Jones, VanWassenhove-Paetzold and Thomas42,Reference Kerr, Barua and Glantz43,Reference Saxe-Custack, LaChance and Hanna-Attisha50)} (n 3) versions which have been shown to have high sensitivity and specificity when compared to the original Household Food Security Survey tool.^{(Reference Hager, Quigg and Black56,Reference Blumberg, Bialostosky and Hamilton57)} Two studies used internally designed, non-validated surveys.^{(Reference Burrington, Hohensee, Tallman and Gadomski34,Reference Orsega-Smith, Slesinger and Cotugna45)}