Sturm 2013.

Study characteristics
Methods	Study design: PCS How were missing data handled? NR. It is unclear how many HHs had incomplete data at follow‐up. Data were analysed as monthly observations for each HH. Randomisation ratio: N/A Recruitment method: N/A Sample size justification and outcome used: NR Sampling method: HHs enrolled in Discovery Vitality used the Discovery visa credit card for purchases at pick n pay supermarkets. Scanner data from Pick n Pay for credit card purchases from November 2009 to March 2012 were linked to 169,485 HHs. Study aim or objective: to examines the effect of a price reduction for healthy food items on HH grocery shopping behaviour among members of South Africa's largest health plan. Study period: 28 months: November 2009 to March 2012 Unit of allocation or exposure: HHs
Participants	Baseline characteristics Intervention or exposure Age: NR Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: NR Education: NR SES: NR Social capital: NR Nutritional status: ratio of healthy to total food expenditure, mean: 10% rebate: 0.21 (SD 0.11); 25% rebate: 0.21 (SD 0.12). Ratio of fruit and vegetable to total food expenditure, mean: 10% rebate: 0.10 (SD 0.08), 25% rebate: 0.10 (SD 0.08). Morbidities: NR Concomitant or previous care: NR Distance from home to nearest Pick and Pay store: mean distance, km: 10% rebate: 1.96 (SD 1.88); 25% rebate: 1.96 (SD 1.91) Control Age: NR Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: NR Education: NR SES: NR Social capital: NR Nutritional status: ratio of healthy to total food expenditure, mean: 0.17 (SD 0.13). Ratio of fruit and vegetable to total food expenditure, mean: 0.09 (SD 0.09) Morbidities: NR Concomitant or previous care: NR Distance from home to nearest Pick and Pay store: mean distance, km: 2.11 (SD 1.99) Overall: NR Inclusion criteria: all Vitality members were eligible for the healthy food benefit at no additional cost to them, but they had to activate the benefit online or by telephone. Only purchases made with a Visa credit card issued by Discovery were analysed, as this was the only identifying information for purchases for which there was no rebate. Exclusion criteria: none reported Pretreatment: HHs who became eligible for a rebate during the study period already had a larger proportion of overall food expenditure going towards healthy foods and a smaller proportion towards less desirable foods at baseline (when nobody was eligible for the rebate) than HHs that never participated in the programme. Participants also lived closer to a Pick n Pay supermarket than to a competing supermarket; the opposite was true for non‐participants. The difference between non‐participants and participants was significant for all variables. Attrition per relevant group: NR Description of subgroups measured and reported: effects for all shoppers vs nearby shoppers (those living ≥ 1 km closer to nearest Pick n Pay supermarket relative to nearest Shoprite or Woolworths supermarket) Total number completed and analysed per relevant group: total: 169,485 Discovery visa holders; intervention (HealthyFood benefit): 100,344 (10% rebate); 67,343 (25% rebate); control (non‐participants): 69,141 Total number enrolled per relevant group: total: 169,485. Discovery visa holders; intervention (HealthyFood benefit): 100,344 (10% rebate); 67,343 (25% rebate); control (non‐participants): 69,141 Total number randomised per relevant group: N/A
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: food prices Intervention type: rebate for healthy food purchases (cash‐back payments) Description: large subsidy programme that operated nationwide, started in 2009 by South Africa's largest health insurer: the HealthyFood programme. Provided a rebate of up to 25% on healthy food purchases in > 400 designated supermarkets across South Africa. All Vitality members were eligible for the healthy food benefit at no additional cost to them, but they had to activate the benefit online or by telephone. Following activation, members immediately received a 10% rebate for healthy foods, which increased to a 25% rebate on completion of an online health risk assessment questionnaire. The rebate was capped at a ZAR 4000 maximum monthly purchase (about USD 480) per HH and a limit related to participation in health promotion activities. Food items eligible for the rebate programme were selected by a panel consisting of nutritionists, physicians and behavioural scientists based on international guidelines on healthy nutrition, including those from South Africa and the US. Complete list of eligible items (> 6000) was on Discovery's website (www.discovery.co.za) and distributed as brochures to programme participants. Participating supermarkets had in‐store signs identifying eligible foods; they are also marked on the store receipt. The labelling was implemented prior to the study period and was not changed during the study Duration of intervention period: 28 months; scanner data from Pick n Pay for credit card purchases from November 2009 to March 2012. Frequency: monthly. Purchases were collapsed into monthly observations, resulting in 1,909,740 observations (HH months). Number of study contacts: NR. Scanner data available every time the card was used to purchase items at Pick n Pay. Providers: Discovery health insurance company in collaboration with Pick n Pay supermarkets. Delivery: activate HealthyFood benefit online or by telephone Co‐interventions: none reported Resource requirements: NR Economic indicators: NR Control: no intervention
Outcomes	Proportion of HH expenditure on food: ratio of healthy to total food expenditure for 10% rebate; ratio of healthy to total food expenditure for 25% rebate; ratio of fruit and vegetables to total food expenditure for 10% rebate; ratio of fruit and vegetables to total food expenditure for 25% rebate; ratio of less desirable food to total food expenditure in a HH for 10% rebate; ratio of less desirable food to total food expenditure in a HH for 25% rebate.
Identification	Sponsorship source: National Cancer Institute (grant R21CA161287); the National Institute of Child Health Human Development (grant R21HD071568); and the Anne and James Rothenberg Dissertation Award, 2011–2012. Country: South Africa Setting: supermarkets Author's name: Roland Sturm Email: sturm@rand.org Declarations of interest: yes; no financial disclosures. Study or programme name and acronym: HealthyFood programme Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	High risk	CBA, where participants self‐selected into intervention.
Allocation concealment (Selection bias)	High risk	CBA.
Baseline characteristics similar (Selection bias)	Unclear risk	Intervention HHs lived closer to a Pick and Pay supermarket than controls. Other sociodemographic characteristics NR. The authors used a DID analysis with matched HHs but, since no characteristics were reported, it was not possible to assess baseline imbalance.
Baseline outcome measurements similar (Selection bias)	High risk	Study authors reported that the expenditure on healthy food as a ratio of total expenditure was higher in the intervention HHs than in the controls.
Blinding of participants and personnel (Performance bias)	Low risk	Intervention was assessed based on sales data as part of participants' usual grocery shopping behaviour.
Blinding of outcome assessment (Detection bias)	Low risk	Outcome was measured using scanner data from supermarkets, therefore, blinding unlikely to influence these outcomes.
Protection against contamination (Performance bias)	Low risk	Non‐participants could not have benefited from the intervention (rebates) as they were not registered to receive benefits.
Incomplete outcome data (Attrition bias)	Unclear risk	Missing or excluded data among scanner data collected NR.
Selective outcome reporting (Reporting bias)	Unclear risk	No protocol available.
Other bias	Unclear risk	Misclassification bias: unlikely. Linked with use of Vitality credit card. Measurement bias: unclear. Potential bias due to use of scanner data which may not be representative of total purchasing behaviour. Seasonality bias: unlikely. Seasonality in grocery shopping patterns were controlled for with a set of dichotomous variables for each specific month in a year.