Risk of bias is non-randomised studies
| Selection bias | Performance bias | Detection bias | Attrition bias | Selective outcome reporting | ||
| Study | Selection of participants | Confounding variables | Measurement of exposure | Blinding of outcome assessment (subjective/objective outcomes) | Incomplete outcome data (subjective/objective outcomes) | |
| Dorresteijn | Data were collected prospectively. | Confounding variables (weather) were considered. | All were exposed, performance bias is not affected by exposure. | Blinding not possible and unlikely to affect objective outcomes. | Characteristics of study before and after the intervention are available. | All of the expected outcomes were included in the study descriptions. |
| Geaney |
‘A random sample of 100 individuals took part in the study (fifty staff from each hospital).’ ‘n the intervention hospital, all individuals who were asked to participate agreed. Less than five in the non-intervention hospital refused to participate in the study.’ | Confounding variables (age and gender) were considered. | All were exposed, performance bias is not affected by exposure. | No details. | Cross-sectional there was no attrition. | All of the expected outcomes were included in the study descriptions. |
| Holdsworth 43 | ‘Two premises applying for the HBA were unsuccessful in receiving it, and were used as a comparison group for the study.’ | Confounding variables (adjusted for age, gender, ethnicity, social class and body mass index) were considered. | All were exposed, performance bias is not affected by exposure. | Lack of blinding may affect subjective outcomes. | Numbers provided (low response rate). | All of the expected outcomes were included in the study descriptions. |
| Kwak et al 48 |
‘A sample of 128 worksites meeting the inclusion criteria were selected through the Chamber of Commerce and invited by letter and telephone to take part in this study …. 12 worksites agreed to participate.’ | Confounding variables (matched based on the social economic status of their employees) were considered. | All were exposed, performance bias is not affected by exposure. | Lack of blinding may affect subjective outcomes. |
‘At baseline, the average response rate to the questionnaire was 88.1% (487 of 553). At the first follow-up, the response rate average was 82.5% (n=376) and at the second follow- up 75.8% (n=303). During the course of the project, one of the worksites in the intervention group underwent reorganization; half of the employees were made redundant, thereby causing most of the dropout in our study. There was no selective dropout at the first follow-up. At the second follow-up, smokers were more likely to discontinue the study than non-smokers.’ | More items collected than reported. |
| Lassen et al | ‘The employees were randomly approached after having bought their meal and asked to participate in the study focusing on canteen food intake.’ | Confounding variables (demographics, seasonal change) were not considered. | All were exposed, performance bias is not affected by exposure. | Lack of blinding unlikely to affect objective outcomes. | Evidence significant differential in number of healthcare personnel between baseline and follow-up | ‘For the outcomes of refined sugar, wholegrain and plate waste a large number of participants had zero intake. These were analysed in two steps. First the probability of any intake yes/no was analysed …. then a regression model was fitted as described above with regard to the participants with a positive intake.’ |
| Lowe et al 34 | ‘Participants were recruited through letters distributed to cafeteria patrons that provided an overview of the study along with inclusion and exclusion criteria.’ | No details. | All were exposed, performance bias is not affected by exposure. | Lack of blinding is unlikely to affect objective outcomes. | Numbers provided (high attrition rate).”Data from ‘Participants who did not scan their cards an average of at least four times per month were excluded from analyses.’ | Reporting of data was not always per Intervention group with no justification. |
| For subjective outcomes ‘Dietary recalls were conducted by the Diet Assessment Center at The Pennsylvania State University.’ | ||||||
| Macdonald et al 38 | Five centres were chosen to participate including a range of contexts (metropolitan, rural, regional). | Confounding variables (changes at the sites) were not considered. | All were exposed, performance bias is not affected by exposure. |
‘Catering receipts received included generic food types such as sandwiches. Categorisation into food types in these instances were based on inferences of the project dietitian.’ ‘The number of serves provided where invoices listed general categories; for example, “large fruit platter” were made from inferences based on cost by the project dietitian.’ | ‘Two ACCHOs were able to provide catering receipts at baseline and follow-up.’ | All of the expected outcomes were included in the study descriptions. |
| Meyers 27 | Data were collected prospectively. | No details. | All were exposed, performance bias is not affected by exposure. | ‘Observations were carried out by undergraduate research assistants.’ | N/A. | All of the expected outcomes were included in the study descriptions. |
| Milich 28 | ‘From original total pool of sampled employees 50 subjects were randomly sampled for each body size in each experimental condition.’ | Confounding variables (weather, obesity and price change) were considered. | All were exposed, performance bias is not affected by exposure. | Outcome assessed by assessing food items on plate then calculating caloric value. | N/A. | All of the expected outcomes were included in the study descriptions. |
| Sato et al 29 | Data were prospectively collected. | Confounding variables (price change) were not considered. | All were exposed, performance bias is not affected by exposure. | Lack of blinding is unlikely to affect objective outcomes. | N/A. | All of the expected outcomes were included in the study descriptions. |
| Thorndike et al 30 | ‘Employee platinum plate users who made a purchase in the cafeteria at least three times during each 3 month period from baseline to the end of follow-up were included in the study cohort.’ | Confounding variables (age, gender, race, job type, and work status) were considered. | All were exposed, performance bias is not affected by exposure. | Lack of blinding is unlikely to affect objective outcomes. |
‘1.7% of employees missing any sociodemographic data were excluded from the analysis.’ Information about the number of participants before and after the study exists, however at each time point the cohort was not identical. | All of the expected outcomes were included in the study descriptions. |
| Van Kleef 31 | ‘Carried out in hospital staff restaurant and visitors are allowed to purchase displayed snacks at a self-service checkout counter throughout the day over a four week period. Data were collected prospectively.’ | No details. | All were exposed, performance bias is not affected by exposure. | Lack of blinding is unlikely to affect objective outcomes. | N/A. | All of the expected outcomes were included in the study descriptions. |
Green, low risk of bias; grey, not applicable; N/A: not applicable; orange, unclear risk of bias; red, high risk of bias; ‘quote from publication’.