Chen 2019.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel group How were missing data handled? data from non‐boarding students and from boarding students who did not have Hb concentration, height and weight measurements were excluded from the analysis. Randomisation ratio: 1:1:2 (nutrition subsidy intervention:nutrition subsidy + monetary incentive intervention:control) Recruitment method: in each 1 of 30 townships, with the necessary permission from the Chinese government, schools were chosen if they offered fourth and fifth grade classes and accommodated ≥ 400 students. Once schools were selected, written assent was requested from students and their parents. Written consent was also requested from the students' legal guardians at school (schoolmasters and head teachers). Sample size justification and outcome used: NR Sampling method: 30 townships were randomly selected from 2 provincial‐level administration units, Qinghai Province and Ningxia Autonomous Region. Schools were selected from these townships and all fourth and fifth graders from these schools with assent and consent were included. Study aim or objective: how does the provision of nutrition subsidies translate into observed nutritional and health outcomes? Will policy targets with different levels of specificity, e.g. general ones such as malnutrition prevention and specific ones such as anaemia reduction, lead to different behavioural responses and, thus, nutritional and health outcomes, partly through different incentives attached to these policy targets (since certain incentives are presumably needed to achieve any specific policy target)? Study period: October 2009–May 2010 Unit of allocation or exposure: cluster: elementary schools
Participants	Baseline characteristics Nutrition subsidy Age: NR Place of residence: NR Sex: entire group (analysed group): mean proportion of boys: 0.52 (0.49) Ethnicity and language: entire group (analysed group): mean proportion of ethnic‐minority students: 0.62 (0.64); mean proportion with migrant father: 0.71 (0.80); mean proportion with migrant mother: 0.11 (0.12) Occupation: NR Education: entire group (analysed group): mean proportion of fifth graders: 0.49 (0.56); mean education of father, years: 6.12 (6.11); mean education of mother, years: 4.32 (4.32); mean proportion of senior‐level teaching staff: 0.48 (0.48) SES: entire group (analysed group): mean number of siblings: 2.18 (2.19) Social capital: NR Nutritional status: entire group: mean BMI‐for‐age z‐score: –0.78; mean Hb concentration (g/L): 128.06; mean proportion of underweight students: 0.07; mean proportion of anaemic students: 0.22; DDS (adapted from FAO guidelines): 5.12. Analysed group: mean Hb concentration: 128.51 (SD 12.63); mean proportion anaemic: 0.18 (SD 0.38); mean BMI‐for‐age z‐scores: –0.70 (SD 0.91); mean proportion underweight: 0.07 (SD 0.25); mean DDS: 4.75 (SD 2.17). Morbidities: proportion anaemic 0.19 Concomitant or previous care: NR Nutrition subsidy + monetary incentive Age: NR Place of residence: NR Sex: entire group (analysed group): mean proportion of boys: 0.53 (0.53) Ethnicity and language: entire group (analysed group): mean proportion of ethnic‐minority students: 0.73 (0.74); mean proportion with migrant father: 0.64 (0.69); mean proportion with migrant mother: 0.12 (0.09) Occupation: NR Education: entire group (analysed group): mean proportion of fifth graders: 0.51 (0.59); mean education of father, years: 6.16 (5.19); mean education of mother, years: 3.94 (3.25); mean proportion of senior‐level teaching staff: 0.46 (0.46) SES: entire group (analysed group): mean number of siblings: 2.20 (2.19) Social capital: NR Nutritional status: entire group: mean BMI‐for‐age z‐score: –0.77; mean Hb concentration (g/L): 127.21; mean proportion of underweight students: 0.09; mean proportion of anaemic students: 0.25; DDS (adapted from FAO guidelines): 5.05. Analysed group: mean Hb concentration: 127.84 (SD 12.80); mean proportion anaemic: 0.23 (SD 0.42); mean BMI‐for‐age z‐scores: –0.63 (SD 0.91); mean proportion underweight: 0.06 (SD 0.24); mean DDS: 4.65 (SD 2.20). Morbidities: NR Concomitant or previous care: NR Control Age: NR Place of residence: NR Sex: entire group (analysed group): mean proportion of boys: 0.52 (0.54) Ethnicity and language: entire group (analysed group): mean proportion of ethnic‐minority students: 0.64 (0.62); mean proportion with migrant father: 0.63 (0.67); mean proportion with migrant mother: 0.12 (0.11) Occupation: NR Education: entire group (analysed group): mean proportion of fifth graders: 0.51 (0.51); mean education of father (years): 6.87 (6.72); mean education of mother (years): 4.10 (3.88); mean proportion of senior‐level teaching staff: 0.40 (0.40) SES: entire group (analysed group): mean number of siblings: 2.35 (2.29) Social capital: NR Nutritional status: entire group: mean BMI‐for‐age z‐score: –0.83; mean Hb concentration (g/L): 129.82; mean proportion of underweight students: 0.09; mean proportion of anaemic students: 0.19; DDS (adapted from FAO guidelines): 5.28. Analysed group: mean Hb concentration: 128.03 (SD 12.95); mean proportion anaemic: 0.22 (SD 0.42); mean BMI‐for‐age z‐scores: –0.68 (SD 0.94); mean proportion underweight: 0.08 (SD 0.26); mean DDS: 5.33 (SD 2.32). Morbidities: NR Concomitant or previous care: NR Overall: NR Inclusion criteria: townships: located in Qinghai Province or Ningxia Autonomous Region. Schools: offering fourth and fifth grade classes; accommodating ≥ 400 students. Children: fourth or fifth grade students; written assent provided; consent from guardians provided; boarding (to be included in analysis). Exclusion criteria: NR Pretreatment: baseline characteristics and outcome variables presented in Table A1 and 2. Table A1: most variables were quite balanced across groups, with minor differences due to sampling errors, suggesting that the random group assignments were done reasonably well. However, due to the modest number (i.e. 59) of project schools, some student or school characteristics may have been balanced across the 3 groups (Appendix A Table A1), even under randomised group assignments. To address this issue, the authors modified the estimating equation. Quote: "Unless otherwise stated, all estimates presented below are obtained after controlling for the full set of covariates reported in Appendix A Table A1." Attrition per relevant group: 61% of boarding students (who were considered in the analysis) were not analysed because they did not have health information collected due to budgetary reasons. Considering the full sample of 6994, attrition was 88%. Description of subgroups measured and reported: NR Total number completed and analysed per relevant group: to fully capture the impacts of the treatments, the authors' analysis focused only on the 2199 boarding students in the sample (of 6994 students). Due to budgetary reasons, while dietary information was collected from all (boarding) students, health information was collected only from about half of them: 1020 boarding students had information on Hb concentration level, 952 had height and weight information, and 866 had both sets of information. Thus, the final analytical sample comprised 866 boarding students with information available on all 3 dimensions. Students: 219/582 (38%) for nutrition subsidy; 210/563 (36%) for nutrition subsidy + monetary incentive; 437/1550 (28%) for control group. Total number enrolled per relevant group: total sample 6994. Not clear per group, as only reported on those with Hb and height/weight information. Total number randomised per relevant group: total of 15 schools for nutrition subsidy, 15 for nutrition subsidy + monetary incentive and 29 for control group.
Interventions	Intervention characteristics Nutrition subsidy (treatment group 1) Food access intervention category: food prices Intervention type: food subsidy Description: nutrition subsidy with a general policy target of 'malnutrition reduction'. The monetary equivalent of the total amount of nutrition subsidy provided to each school in this group was CNY 225 (USD 33) per enrolled student (which cost CNY 1.5/day, enough for purchasing 60 g of red meat, for 150 days). The subsidy money was transferred to the bank account of each school and the schoolmasters were able to use this money for nutrition‐related expenses (in any way they deemed reasonable). In addition, each schoolmaster in this treatment group was informed about the main aim of the intervention project (i.e. to reduce child malnutrition) and was given 3 pieces of additional information: 1. the proportion of enrolled students who were anaemic (not the specific individuals but the mean rate of the whole school), 2. descriptions of effective methods for reducing iron‐deficient anaemia and 3. details about anaemia's relation with school attendance, educational performance and cognitive development. However, the project team did not provide any specific instructions or stipulate specific requirements on what foods the schools should purchase; schoolmasters were allowed to make their own decisions on how the subsidy money was spent to achieve the goal of malnutrition reduction. Duration of intervention period: 6 months: November 2009–May 2010, with a 1‐month pause during the winter break in February 2010. Frequency: 1 Number of study contacts: 2 contacts: baseline (October 2009) and follow‐up (May 2010). Providers: project team and local government. Delivery: CNY 225 (equivalent to USD 33) per student, enough to purchase 60 g of red meat per day for 150 days, was transferred into the school's bank account. Co‐interventions: NR Resource requirements: CNY 225 per student. Human resource requirements were minimal, as the money was transferred to the school bank account and administered by the schoolmasters. Economic indicators: NR Nutrition subsidy + monetary incentive (treatment group 2) Food access intervention category: food prices Intervention type: food subsidy with additional incentive Description: nutrition subsidy identical to treatment group 1, + a specific policy target of 'anaemia reduction'. Involved an incentive in the form of a potential monetary bonus provided to schoolmasters; amount was tied to actual reductions in anaemia prevalence among students in their schools – more specifically, a schoolmaster would receive a CNY 150 (or USD 22) bonus for each of his other students whose status changed from being anaemic to being non‐anaemic over the course of the intervention. As with schoolmasters in treatment group 1, those in treatment group 2 were informed about the main aim of the intervention project (i.e. to reduce child malnutrition) and were provided with the same 3 pieces of anaemia‐related information. Similarly, they were allowed to make their own decisions on how the subsidy money was spent; no requirements were imposed by the project team or the local government. Duration of intervention period: November 2009–May 2010, with a pause of 1 month (February 2010) during the winter break. Frequency: 1 Number of study contacts: 2 contacts: baseline (October 2009) and follow‐up (May 2010). Providers: project team and local government. Delivery: CNY 225 (equivalent to USD 33) per student, enough to purchase 60 g of red meat per day for 150 days, was transferred into the school's bank account. Since the bonus provided to schoolmasters in treatment group 2 would not be realised until the end of the intervention period when the actual reductions in anaemia prevalence were revealed (note that it may not even be realised if there are no reductions in anaemia prevalence), the actual amount of subsidy per student received by the 2 treatment groups (i.e. the amount that could be used for food purchase per student terms) was identical during the intervention. Co‐interventions: NR Resource requirements: CNY 225 per student, + CNY 150 for each anaemic student converted to non‐anaemic. Human resource requirements were minimal, as the money was transferred to the school bank account and administered by the schoolmasters. Economic indicators: NR Control: no intervention
Outcomes	DDS (0–10) Anthropometry: BMZ; proportion underweight (BMZ < –2SD) Biochemical: Hb concentration Morbidity: proportion anaemic
Identification	Sponsorship source: National Natural Science Foundation of China [grant number 71603261]; Humanities and Social Science Fund of the Ministry of Education of China [grant numbers 16YJC880107, 18YJC790010]; The Fundamental Research Funds for the Central Universities [grant number 2019TC110]. Country: China Setting: fourth and fifth graders enrolled in elementary schools in rural Qinghai and Ningxia Authors' names: Qihui Chen; corresponding: Qiran Zhao Email: zhaoqiran@cau.edu.cn Declarations of interest: yes; no conflicts of interest. Study or programme name and acronym: NR Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	No information provided how the random sequence generation is done. Only (quote): "schools were randomly assigned into three groups."
Allocation concealment (Selection bias)	Low risk	Unit of allocation was by school; all assignments were done at start of study.
Baseline characteristics similar (Selection bias)	Unclear risk	Low for the analysed group, but no data available for the entire group. Quote: "Due to both randomisation plus taking this into account during the analysis. Baseline characteristics and outcome variables presented in Table A1 and 2. Table A1: Most of these variables are quite balanced across groups, with minor differences due to sampling errors, suggesting that the random group assignments were done reasonably well." "However, due to the modest number (i.e., 59) of project schools, some student or school characteristics may not be perfectly balanced across the three groups (Appendix A Table A1), even under randomized group assignments. To address this issue, we modify the estimating equation (1) in two ways." "Unless otherwise stated, all estimates presented below are obtained after controlling for the full set of covariates reported in Appendix A Table A1."
Baseline outcome measurements similar (Selection bias)	Unclear risk	Low for the analysed group, but no data available for the entire group. Quote: "Due to both randomisation plus taking this into account during the analysis. Baseline characteristics and outcome variables presented in Table A1 and 2. Table A1: Most of these variables are quite balanced across groups, with minor differences due to sampling errors, suggesting that the random group assignments were done reasonably well." "However, due to the modest number (i.e., 59) of project schools, some student or school characteristics may not be perfectly balanced across the three groups (Appendix A Table A1), even under randomized group assignments. To address this issue, we modify the estimating equation (1) in two ways." "Unless otherwise stated, all estimates presented below are obtained after controlling for the full set of covariates reported in Appendix A Table A1."
Blinding of participants and personnel (Performance bias)	High risk	Not explicitly reported whether students and school staff were blinded to assignment. Given the integral role of the schoolmaster in administering the intervention and their access to information on anaemia; however, it is not possible that these people could be blinded. The lack of blinding may have resulted in considerable performance bias, specifically in the group of schoolmasters not receiving incentives.
Blinding of outcome assessment (Detection bias)	High risk	NR whether the medical professionals who assessed some outcomes were blinded. Most outcomes were objective and not prone to detection bias, but outcomes such as dietary diversity may have been affected by unblinded self‐report from participants.
Protection against contamination (Performance bias)	Low risk	Since payment was made into the school account and information was provided to the headmasters, contamination at school level was not expected. Furthermore, only boarders were analysed: boarding arrangement ensures that almost all the food consumed by boarding students came from the intervention.
Incomplete outcome data (Attrition bias)	High risk	Very high levels of attrition among students: 363/582 (62%) for nutrition subsidy; 353/563 (64%) for nutrition subsidy + monetary incentive; 1113/1550 (72%) for control group. This is predominantly due to only including boarding students with Hb and anthropometric data, as well as the loss of 1 control cluster. Attrition appears to be differential for the control group versus the intervention groups.
Selective outcome reporting (Reporting bias)	Unclear risk	No study protocol available.
Other bias	Unclear risk	Recruitment bias: randomisation of schools occurred after students had been recruited; low risk. Incorrect analysis: adjustment for clustering NR, but robust SEs used; unclear risk.