PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • This is a longitudinal study of newly recruited participants in traditional and alternative interventions. Participants were recruited in a two-stage cluster sampling frame. Clustering units were community organizations working on food insecurity in the Montreal Metropolitan Region (MMR). The MMR includes 82 municipalities and a population of nearly four millions people [27]. The list of organizations involved in food insecurity interventions was validated through the confirmation of experts with in-depth knowledge of the food insecurity intervention network in the Montreal area. Organizations exclusively targeting children such as school lunch and breakfast programs were excluded. A total of 451 organizations were selected. We administered a phone survey to the directors of 30.1% (136 / 451) of these organizations, randomly selected, to identify the interventions each organization implemented and the number of new and overall participants. Of the 136 surveyed organizations, 61 uniquely offered traditional interventions and 75 offered a form of alternative intervention. Organizations offering both traditional and alternative interventions were classified as alternative. Participants selected from these organizations were uniquely participating in alternative interventions Organizations were invited to participate in our study based upon their number of new participants each year. A minimum number of 30 and 50 participants having begun a food insecurity intervention in the past 6 months was set respectively for alternative and traditional interventions. This criteria for participation was determined first, by considering what could be expected in terms of participant recruitment and second, what was needed in terms of statistical analysis. Community partners’ expertise first informed our criteria for the minimum number of new participants. Namely, new participants in alternative interventions are less numerous than in traditional interventions whereas participation is stable and constant for longer periods of times. While the number of new participants in traditional interventions is higher, participation is more sporadic and related to temporary situations. Second, power calculations for the minimum number of participants required for statistical analysis, as further explained below, considered the hierarchical nature of the data. Individuals between 18 and 65 years of age registered for the first time, and for less than 6 months in selected MMR food insecurity community organizations were invited to participate in our study. People older than 65 years of age were excluded from the study because, in Québec, they can benefit from income supplement and have preferential paths to fight food insecurity. Homeless people were also excluded for two reasons. First, homeless people represent a sub-population extremely vulnerable and their strategies to cope with food are different from the rest of the population [28]. Second, their inclusion in the study would have biased results because of the lack of long term strategies for food insecurity directed to this vulnerable population. Services and resources provided by organizations, although classified in the same category (i.e. food bank) could differ among organizations according to policies, quality of food or frequency of access. Participants in our study were nested within organizations. With nested data observations may not be independent and hierarchical multilevel modeling is recommended [29]. To account for these differences along with the longitudinal nature of the data, analysis accounted for data structured into three levels. The first level was the change overtime in dependent variables, the second level were individuals, the third corresponded to the organization where participants were recruited. With this hierarchical structure of data, and since the intraclass (among individuals) and interclass (among different organizations) coefficients for our outcome variables were unknown, we needed a large number of participants in each organization to detect a 10% difference in the variables measuring food security with an acceptable degree of precision (0.9) and at a statistically significant threshold (0.05). Among the 136 organizations, 16 organizations carrying out traditional interventions and 6 implementing alternative interventions met the criteria for a minimum number of new registered participants. From October 2011 to May 2012 a questionnaire to investigate health, food insecurity and vulnerability was completed by participants with the support of research assistants specifically-trained to accompany participants through the completion of the questionnaire. The questionnaire took approximately 30–45 minutes and was administered face to face in French or in English, according to the preference of participants. Interviews took place in the organizations providing food insecurity interventions or in the nearby area. Participants were informed that they would be called back and invited to participate in the second part of the study and were asked to inform the research team in case of changes of address. Six months after the first interview, a postcard inviting participants to communicate possible changes of address was sent to each participant. Nine months after the first interview, participants were invited by phone to complete the follow up interview. A nine month follow-up was considered adequate to detect intervention effects on the level of food security and health status because according to community partners, sufficient time has passed such that participation in alternative food security interventions has become regular. Non respondents were contacted by mail and invited to contact project managers. The second interview took place in the same location as the first. In case of changes of address or impracticability of the first location, another place was chosen by mutual agreement. The same questionnaire was used for the first and the second survey. The categorizing variable corresponded to participants’ enrollment in one of the two intervention strategies, traditional or alternative. Dependent variables were food security status and perceived physical and mental health. Control variables considered in the study were gender, age, country of birth, marital status and income. Food security status was measured using the food security module included in the Canadian Community Health Survey(CCHS)[30]. The food security module presents the same questions used in the United States Household Food Security Survey Module (HFSSM). The HFSSM was validated to measure change in food security status overtime [31]. The CCHS calculates three scores of food insecurity for the previous year: for the respondent, for dependent children (when applicable) and for the respondent’s household. The CCHS food security module is composed of 10 adult-related and 8 children-related questions investigating whether the respondent or other household members experienced indicators of food insecurity. Questions query the severity of the experiences associated with food, such as an anxiety that food will run out, a need to modify the amount of food consumed, experiencing hunger, and in the extreme, going a whole day without eating. Each multiple choice answer is recoded scoring 0 or 1 point, where 0 corresponds to food security and 1 to food insecurity status. For example, if the question “You and other household members worried that food would run out before you got money to buy more. Was that often true, sometimes true, or never true in the past 12 months?” was answered “never true” the question was coded as 0 while responses of “often true” or “sometimes true” were coded 1. The answers “often” and “sometimes” are both considered affirmative responses because they indicate that the condition occurred at some time during the year [31]. The final score ranges between 0 to 10 for adults and 0 to 8 for children. The food security module defines three levels of food security: food security, with scores of 0 or 1, moderate insecurity with a score between 2 and 5 for adults and 2 and 4 for children, and severe insecurity with score respectively above 5 for adults and above 4 for children. Household food security status is dependent on both adult and child scores. In families with children, the household is food secure if both adults and children are food secure; the household is moderately food insecure if either adults or children are moderately food insecure but neither is severely food insecure; the household is severely food insecure if either adults or children are severely food insecure. In childless households, adults’ food security status corresponds to household food security status. Health related quality of life: Generic health-related quality of life was measured using the SF-12-v2 questionnaire [32]. The SF-12-v2 is a shorter and validated version of SF-36, regularly used to assess perceived health [33]. The questionnaire tests physical and mental health in the last four weeks generating 8 subscales: 1- physical functioning (composed of 2 items: health limitations in accomplishing moderate activities such as moving a table, or pushing a vacuum cleaner; health limitations in climbing several flights of stairs); 2- role limitations due to physical problems (composed of 2 items: limitations accomplishing what one desires to accomplish; limitations in the kind of work or activities); 3- bodily pain (composed of 1 item: pain interference in the accomplishment of normal work); 4- general health perceptions (composed of 1 item: own health perception); 5- vitality (composed of 1 item: perceived energy); 6- social functioning (composed of 1 item: interference of physical health or emotional problems); 7- role limitations due to emotional problems (composed of 2 items: limitations in accomplishing what wanted as a result of feeling depressed or anxious; less attention in doing work or other activities); and 8- mental health (composed of 2 items: perception of calm and peace; perception of downhearted and depression). Based on these subscales, two summary scores were calculated: the physical (PCS) and the mental (MCS) component summary scores. PCS was built with the subscales 1 through 4 and MCS was built with subscales 5 through 8.PCS and MCS scores were transformed in a 0–100 score according to published regression weights and scoring rules (a higher score indicating better health-related quality of life) as suggested in the user manual. [34] Household income and other control variables: Respondent’s gender, age, country of birth, being part of a visible minority, marital status and household income were self-reported. According to the Employment Equity Act of Canada, we defined visible minorities "persons, other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour". Household income was grouped into 7 categories ranging from “no income” to “income superior to $40000”. In our analysis, missing values were not imputed but excluded pair-wise. Since our data were repeated measures from individuals nested in organizations and categorized in two different interventions, the influence of dependent variables on the outcomes (food insecurity and health) was investigated with multilevel regression analyses. We used multilevel models to account for the hierarchical structure of data and likewise, to avoid an underestimation of the group effect and incorrectly rejecting the null hypothesis of no difference (i.e. type I error)[29]. The analyses were executed separately for participants in traditional and alternative interventions. Generalized linear latent regression models (GLLAMM) were used. GLLAMM performs maximum likelihood estimation by using adaptive quadrature. Three-level random intercept regression models were constructed for food security for each intervention strategy. A first model was constructed using food insecurity as the dependent variable. Subsequently, a sequence of controlling variables (respondent’s gender, age, country of birth, marital status and income) were entered as covariates at the individual level of the model. No organization level factors were added to the models. Six linear random intercept regression GLLAMM models were used to assess perceived health (three for physical health and three for mental health) in each intervention group. The first two models considered physical and mental health unadjusted, the following two models were adjusted for respondents’ gender, age, country of birth, marital status and income, while the last two models were also adjusted for adults’ food security status. Differences across time in the two intervention groups were also tested through GLAMM models. The STATA v11.2 software was used to perform statistical analysis. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the health research ethics committee (CERES) of the University of Montreal. Written informed consent was obtained from all subjects/patients.
rdf:type