Description of ‘JolinchenKids – fit and healthy in daycare’
JolinchenKids - fit and healthy in daycare is comprised of five modules, three focusing on children, one on parental participation, and one on promoting health among DF staff. The three modules focusing on children’s health were designed to affect dietary and PA habits among 3- to 6-year-olds and to improve their mental well-being . Program Implementation was based on the Public Health Action Circle and included the four steps:  needs analysis,  module selection,  module implementation, and  evaluation. Starting point of the program was a two-day training session for teachers followed by the needs analysis and module selection in each individual DF. During the two-day training session, staff of the AOK health insurance familiarized teachers with content and objectives of each module, explained how to use the intervention materials and gave practical tips for the selection and pedagogical and didactic implementation of the individual modules of JolinchenKids into the day-to-day routines of the daycare setting. Generally, DFs were free to choose which modules they would like to implement and in which order. In addition, several modules could be implemented at the same time. More detailed information about the program and intervention components belonging to the five different intervention modules can be found elsewhere .
Study design and participants
The evaluation of JolinchenKids – fit and healthy in daycare was conducted with a nationwide sample of DFs. A cluster-controlled trial was conducted to assess relevant outcome measures at baseline and after 1 year. The AOK provided two lists, one containing addresses of DFs implementing the program in 2016 and one containing DFs that planned to implement the program 1 year later. We invited a random selection of DFs (n = 473) from both lists to participate in the study. To ensure that DFs from rural and urban parts of Germany would take place in the study, the number of inhabitants was determined for each DF on the basis of the postal code. The selection then followed a ranked list of DFs classified into quintiles based on the number of inhabitants (≤ 5.000; ≤ 20.000; ≤ 100.000; > 100.000). In the invitation letter, we provided a short questionnaire to determine eligibility to participate in the study. Further information on the short questionnaire and the inclusion criteria can be found elsewhere . The recruitment took place in summer and fall 2016. DFs were closed during school holidays depending on the federal state. We started recruiting directly after the closing times. The first surveys could therefore be carried out in September. To ensure that both groups were similar in structural characteristics and in parents’ socio-demographic characteristics, information from the short questionnaire was used to match intervention and control DFs.
Once a DF consented to participate in the study, was deemed eligible, and was matched to a DF from the control group, we asked the DF to choose two kindergarten groups that included at least some three-to-five-year-old children. Parents of three-to-five-year-old children of those two groups were invited to have their child to participate in the study. Prior to the tests, parents were asked for informed consent to participate in each survey. On the measurement day, children with written parental informed consent were informed appropriately and asked for verbal assent. Ethical approval for conducting the study was obtained from the Medical Association in Bremen (HR/ RE – 522, April 28th, 2016).
Between September and December 2016, baseline measurements were performed followed by a post-test assessment during the same time period 1 year later (Sept – Dec 2017). The same two study nurses who were trained by two researchers collected data at both time points. After the baseline assessment, program implementation started at intervention DFs with a two-day training session for teachers for the introduction of the programme and distribution of intervention materials provided by the AOK. Control DFs did not receive the intervention and continued with their usual routine. Upon completion of the final data collection, the control DFs were offered the same training and materials that had been delivered to the intervention DFs (i.e. wait listed control).
Parental education, monthly household income, and children’s sex, age, and migration background were reported by parents (or legal guardians) . If information on education was available from both parents/legal guardians, parental education was classified into low, medium, and high according to Lampert et al. , otherwise it was set to missing. Data on migration background was compiled based on information on the country of birth and the nationality of both parents. Children classified as having a two-sided migration background had parents who both had immigrated to Germany and/or parents who were not German citizens; children classified as having a one-sided migration background had one parent that had immigrated to Germany from another country and/or did not hold German citizenship . If information on the country of birth or the nationality of one parent was not available, migration background of the child was set to missing. To determine urbanity, DFs from municipalities with more than 20.000 inhabitants were classified as urban whereas those from municipalities with less than 20.000 inhabitants were classified as rural.
To measure height, weight, and body composition, children had to be barefoot. Height was measured to the nearest 1 cm using a stadiometer (Seca® type 213 stadiometer, Invicta Plastics Ltd., Leicester, UK). Body weight was measured to the nearest 0.1 kg. Bioelectrical impedance and body mass were assessed once in each child using a prototype leg-to-leg device that is based on the TANITA® BC 420 SMA digital scale. The prototype was developed by TANITA Europe (TANITA Europe GmbH, Sindelfingen, Germany) specifically for this study to assess leg-to-leg bioelectrical impedance in children whose feet are too small for the currently produced devices. The BIA measurements took place according to the instructions of the manufacturer . Body type was entered as ‘standard’ for all children. For ethical reasons, children were measured without fasting. Before each individual measurement, the child’s shoes and socks and, if worn, tights were removed. The body mass index (BMI) was calculated as weight in kilos measured by the TANITA scale divided by body height in meters squared. We did not weigh children’s clothes; therefore, we did not correct BMI for the weight of the clothes. Children were classified as underweight/normal or overweight/obese according to age- and sex-specific cut-offs derived from percentile curves by Cole and Lobstein . These (International Obesity Task Force – IOTF) childhood BMI cut-offs for overweight, obesity, and thinness are widely used and are based on nationally representative survey data from six countries covering the age range of 2 to 18 years . Data were fitted using the LMS method to standardize the distribution of BMI using age- and sex-specific parameters on skewness (L), median (M), and coefficient of variation (S), respectively in the age range from 2 to 18 years to eventually fit corresponding BMI categories for adults at age 18. If information on height and weight was not available, BMI and BMI category were set to missing. Percent body fat was derived from bioelectrical impedance assessment (BIA) . Due to the lack of a fasting state at the time of measurement, implausible BIA values could not be avoided. According to Goran et al.  values under 500 and above 900 Ω were excluded from the data analyses later on. If information on BIA was not available or in case of an implausible BIA value, percent body fat was set to missing.
Motor skills, screen time and physical activity
Children participated in the five test items of the KindergartenMobil-Test (KiMo). A detailed description of the testing procedure is given by Klein and colleagues . Further information of the five test items can be found elsewhere . Exercises were explained and demonstrated according to the KiMo manual. We analyzed items separately for each motor skill.
To assess screen time, parents were asked about information on hours of television/digital video disk/video viewing (television time) and computer/smartphone/tablet/games-console use (computer use) for both a typical weekday and weekend days. Response categories were 0 = ‘not at all’, 1=’ < 1 h/day’, 2 = ‘between 1 and < 2 h/day’, 3 = ‘between 2 and < 3 h/day’, and 4=’ > 3 h/day’. Children’s television time and computer use were summed up to total screen time per week as follows: (television time on weekdays*5) + (television time on weekend days*2) + (computer use on weekdays*5) + (computer use on weekend days*2). Daily PA levels were assessed via parental reports based on a question in the parental questionnaire: “Think for a moment about a typical weekday (weekend day) for your child in the last month. How much time would you say your child spends playing outdoors on a typical weekday (weekend day)?”. In addition, parents answered the following questions:” Is your child member of a sports club?” and “How much time does he/she spend doing sports in a sports club per week?”. Hours that children spent playing outdoors on a typical weekday and on weekend days and weekly participation in sports club activities were assessed to calculate total daily time spent on PA as follows: [(PA playing outdoors on weekdays*5) + (PA playing outdoors on weekend days*2) + weekly sports participation]/7. These parental-report measures of outdoor playtime and sports participation were found to be positively correlated with accelerometry-derived moderate to vigorous PA in a previous study .
We assessed consumption of unsweetened beverages, fruits and vegetables, snacks, and the number of meals per day during the last week in a self-developed food frequency questionnaire (FFQ) which has not been validated so far. However, items were based on a validated FFQ previously used in wave 2 of the longitudinal cohort study “German Health Interview and Examination Survey for Children and Adolescents” (KiGGS) [36, 37] and food categories were based on the food pyramid of the German Nutrition Society (DGE, Deutsche Gesellschaft für Ernährung). In the self-administered FFQ, respondents are asked questions about the frequency and about the portion size of a limited number of usually consumed foods. It is relatively inexpensive, easy and quick to administrate . However, only a limited number of foods can be included in a FFQ for feasibility reasons and to limit the burden for participants. Intervention goals in the nutrition module of JolinchenKids – fit and healthy in daycare are based on the nutrition pyramid of the AID Infoservice of the Federal Office for Agriculture and Food  which is based on the recommendations of the DGE. When developing the FFQ used in our study, we therefore adapted a predefined list of foods to the nutrition pyramid of the AID. Items were developed and used to measure the consumption of water and unsweetened drinks, fruits, vegetables, bread, cereals and side dishes, milk (products), meat, fish and eggs, fats and oils, and sweets, sweet spreads, pastries or salty snacks. To assess consumption of unsweetened beverages parents were asked the following “How often did your child drink water (mineral water, tap water, homemade soda water) and unsweetened drinks (fruit tea, herbal tea) in recent weeks?” Response categories were “never/ less than one glass per week”, “one to three glasses per week”, “four to six glasses per week”, “one glass a day”, “two to three glasses a day”, “four to five glasses a day”, and “more than five glasses a day” wherein one glass was defined as 135 ml. To assess number of snacks per day we asked parents the following question: “How often did your child eat sweets, sweet spreads, pastries or salty snacks such as chips and french fries in recent weeks?” Response categories were “never/ less than one serving per week”, “one to three servings per week”, “four to six servings per week”, “one serving a day”, “two servings a day”, “three servings per day”, “four servings per day”, and “more than four servings a day”. One portion size was quantified as one child’s hand full of a snack. In addition, the number of meals on weekdays and weekends was assessed. Based on the recommendations of the nutrition pyramid , nutrition was considered healthy, if a child consumed i) at least four glasses of unsweetened beverages, ii) at least five portions of fruits and vegetables, iii) not more than one snack per day, and iv) at least three portions of milk and dairy products per day.
Family health climate
The family health climate was assessed with the Family Health Climate-Scales for PA (FHC-PA) and nutrition (FHC-NU), using a validated questionnaire . The FHC-PA Scale consists of 14 items and three subscales: value (e.g.,” In our family, it is normal to be physically active in our leisure time”), cohesion (e.g.,” ...we have fun doing physical activities together (e.g., bike tours, hikes)”), information (e.g.,” ... we collect information (e.g. on the internet) on physical activity and exercise”). The FHC-NU Scale is comprised of 17 items pertaining to four subscales: value (e.g.,” ... it is normal to choose healthful foods”), cohesion (e.g., “...we appreciate spending time together during meals”), communication (e.g., “...we talk about which foods are healthful”), consensus (e.g.,” ...we rarely argue about food- or diet-related matters”). The items were rated on a 4-point Likert-type scale ranging from 0 = ‘not true’ to 3 = ‘true’. Scores representing the mean of all items were calculated for the FHC-PA and FHC-NU, respectively.
Process evaluation data
To assess intervention dose and fidelity, intervention DFs were provided with a paper-and-pencil calendar to track implementation progress at individual DF groups from baseline to follow-up . In these calendars, DF staff documented module choices, as well as module specific activities for each week covering components of intervention modules for diet (healthy breakfast buffet, “drinking oasis”, dish of fruits & vegetables, short games), PA (PA games), wellbeing (time of card game “feel good island”, short time-outs), and parental participation (newsletter, “message in a bottle”, parent-staff evening) that were marked with checkboxes, as well as a documentation of the weekly amount of time (in minutes) spent on working with the respective intervention materials.
To quantify the intervention dose for each module, we considered a time frame of 40 weeks (i.e. 1 year excluding holidays) during which modules could be conducted within the DF groups. For each module, essential components were distinguished from additional components. For example, conducting 1 hour of PA games was considered as essential and counted as one point, while any additional 10 minutes of PA games was counted as 1/6 additional points for each week. The sum of points in the PA module of all weeks was then divided by 40 to assess the percentage of adherence. Likewise, for each week, a quarter point was given for each of the four components of the diet module, half a point was given for any activity in the parental participation module, and one point was given for 1 hour of games or timeouts for the mental wellbeing module, to assess the percentage of adherence. Eventually, the percentage of adherence for all DF groups of the intervention DFs was categorized into 0%, i.e. no adherence to the respective module, 1–50% of adherence, and more than 50% of adherence. Due to the points depending on the reported duration of components, for some modules an adherence above 100% was calculated.
Descriptive statistics, i.e. mean and standard deviation (SD) or percentage of categories were calculated for the baseline and follow-up survey. We investigated the differences in outcome variables between T0 and T1 between the intervention and control group by using linear mixed models. In case of binary outcome variables, logistic mixed models were used. We modelled fixed effects for intervention group and survey to investigate overall group and time effects, as well as an interaction of group and survey to identify the intervention effect across all intervention DFs, regardless of module choices. Due to the flexibility of mixed models, we were able to use data on participants at baseline without observations in follow-up and accounted for repeated measurements by means of a random effect on the residual side. The effect estimates for quantitative outcomes are expressed as the difference between the mean individual changes in the intervention and the mean individual changes in the control groups. The effect estimates for binary outcomes were obtained from logistic regression models and presented as odds ratios with 95% confidence intervals. All models were adjusted for sex, age, and migration background, and BMI of the children, as well as for household income, highest educational level of parents and urbanity of the DF. Distribution of residuals and model fit were assessed with regard to Q-Q plots, residual plots and the Bayesian information criterion (BIC). In a further step, all models were also stratified by migration background, BMI category, and urbanity of the DF to investigate intervention effects in subsamples, e.g. overweight/obese children, children with a one- or two-sided migration background or children from urban vs. rural areas.
In a second step, we investigated intervention effects taking module choices and module-specific intervention dose at the DF group level into account. Based on the process evaluation data, differences in outcome variables were therefore estimated depending on adherence to the respective intervention modules. The adherence categories were used considering control and intervention group at baseline as reference while investigating changes in the control group at T1, and the three categories of module adherence. Modules were chosen with regard to specific outcomes, i.e. adherence to the PA intervention module was considered for the outcome variables time spent in PA, screen time, and motor skills, adherence to the nutrition module was considered for the outcome variables on fruits and vegetables consumption and consumption of unsweetened beverages and snacks.
Significance level was set to α = 0.05. Statistical analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, North Carolina, USA)  and particularly the glimmix procedure to estimate linear and logistic mixed models. We did not adjust for multiple testing in the overall analyses.