Target group 1: Overweight women.

Overweight women (n = 54) were recruited at their first antenatal visit using convenience sampling. The power analysis was calculated based on two scales, self-efficacy in eating (WEL) and physical activity (PASE). Calculations were done with one sided t-tests with an effect size of 0.50, a power level of 0.80 and a significance level of 0.05. A 20% loss was included in the calculation. Based on both scales, a sample of 54 overweight pregnant women was needed to detect the difference in changes in eating self-efficacy or self-efficacy in physical activity.

Women were eligible for inclusion if they were 1) pregnant ≤15 weeks of gestation, 2) ≥18 years of age and 3) overweight or obese (BMI ≥ 25kg/m2) and 4) had adequate language skills in Finnish. Women were excluded if they 1) did not have a mobile device to download applications and synchronize data, 2) had a diagnosis of severe mental illness, 3) were diagnosed diabetes type 1 before pregnancy and/or 4) had mobility limitations.

Target group 2: Public health nurses.

All PHNs (n = 9) working in selected maternity and child health clinics were eligible for the study. The researcher applied for approval from the Nursing Director of the health center and contacted the PHNs via e-mail and scheduled meetings. Public health nurses were already recruited in the intervention development phase.

Data collection.

In the initial antenatal appointment, the PHNs provide a letter of information and ask permission for the researchers to be in contact with the potential participants. The researchers then explain the study purpose and procedures to the women in a scheduled appointment. In addition, an Oura ring, and instructions on how to use it and other applications are given to each participant by the researchers after informed consent.

Outcome measures.

The primary outcomes for the effectiveness of the intervention are self-efficacy in eating and physical activity and reduced weight gain during pregnancy (secondary outcome). To track variations and trends, health parameters are collected, these include: weight retention postpartum, depressive symptoms, quality of life, pregnancy specific anxiety, perceived stress, sense of coherence, eating behavior and lifestyle patterns, physical activity levels, sleep quality, stress levels and dietary intake. (Fig 1).

Measurements.

Participants complete the questionnaires via the SLIM application at recruitment (<15 pregnancy week), at 34 gestational weeks and 12 weeks after delivery.

Primary outcomes.

Eating self-efficacy and physical activity self-efficacy are measured with following questionnaires:

Weight Efficacy Life-Style Questionnaire (WEL): is a standard, validated and clinically meaningful measure of eating self-efficacy. WEL consists of 20 items, which are scored on a Likert-type scale from 0 to 9 (0: Not confident, 9: Very confident). Answers are assembled into five factors (four questions for each factor): confidence in being able to resist the desire to eat in the following contexts: negative emotions, availability, social pressure, physical discomfort and positive activities. The eating self-efficacy score is calculated by adding together all 20 items. A level of eating self-efficacy is defined by a high score on the WEL questionnaire [34, 35]. The validation studies reported an internal consistency for Cronbach’s α of 0.70–0.90 in overweight Caucasian population.

Self-Efficacy for Physical Activity Scale (PASE) is a valid questionnaire to assess confidence in the ability to exercise when faced with specific barriers. It has been modified for pregnant women. Participants are asked to rate their confidence in engaging in physical activity across a variety of situations. Responses are measured on a Likert scale ranging from 1 to 5 (1 = not at all confident, 5 = extremely confident). The PASE score is calculated by taking the sum of the 5 items. The internal consistency for Cronbach’s α was between 0.70–0.78 in a Latino women sample [36, 37].

Secondary outcome.

Weight gain is measured by PHNs at every antenatal appointment, on average eight times. The women’s weight before delivery is then compared to the national recommendations based on their BMI before pregnancy. For those women whose pre-pregnancy BMI is 25–29.9, the recommended weight gain during pregnancy is 7–11 kg and for those women whose pre-pregnancy BMI is 30 or more, a gain 5–9 kg weight during pregnancy is recommended [38].

Health parameters.

Edinburgh Postnatal Depression Scale (EPDS) is a reliable and valid 10-item questionnaire that screens women for symptoms of emotional distress during pregnancy and the postnatal period. The total score is calculated by adding the numbers selected for each of the 10 items. Depression is defined by a high score on the EPDS. Cut-off scores range from 9–10 to 13–14 [39, 40]. In this study, a cut-off score ≥10 has been used to indicate possible depression. The reliability values of the Edinburgh Depression Scale indicated by Cronbach’s α coefficient per trimester were 0.82, 0.83, and 0.84, respectively among an unselected sample of 845 pregnant women [41, 42].

WHOQOL-BREF is a well-accepted, valid generic 26-item instrument and the mostly commonly used instrument to assess quality of life. Responses to questions are measures on a 1–5 Likert scale where 1 represents "disagree" or "not at all" and 5 represents "completely agree" or "extremely”. WHOQOL-BREF has four domains: Physical health, psychological health, social relationships and environment. The score is calculated by adding together the point values for the questions corresponding to each domain and then transforming the scores to a 0-100-point interval. The first two questions on the WHOQOL-BREF do not correspond to a domain, but are meant to provide a global assessment of quality of life. Higher scores in each of the domains correspond to a greater perceived quality of life. The internal consistency measured by Cronbach’s alpha ranged from 0.71 to 0.77 in the sample of 190 postpartum women [43, 44].

PRAQ-R2 is a valid pregnancy-specific anxiety questionnaire. Responses are measured with a Likert type scale ranging from 1 (definitely not true) to 5 (definitely true). The items in the PRAQ-R2 can be ordered into three subscales: Fear of giving birth, worries about bearing a physically or mentally handicapped child and concern about one’s own appearance. Cronbach’s alphas across the total scale were good (above 0.80) in a sample of Finnish pregnant women (n = 1152) [45].

Perceived Stress Scale (PSS-4) is a validated measure which rates on a 5-point scale the extent to which participants have experience stress-related symptoms in the past month. Responses are measured with a Likert type scale ranging from 0 (never) to 4 (very often), the Cronbach’s α = 0.77 (English sample). Higher values indicate more stress [46, 47].

Sense of Coherence (SOC-13): is measured with a 13-item SOC questionnaire, which has three subscales: Meaningfulness (4 items), Comprehensibility (5 items), and Manageability (4 items). Responses are scored on a 7-category semantic differential scale with two anchoring responses tailored to the content of each item. Five items (1, 2, 3, 5, and 7) are reversed before adding together the scores to obtain the total score. The total score can range from 13 to 91 and a higher score indicates a higher SOC. The internal consistency measured by Cronbach’s alpha ranged from 0.70 to 0.92 [48]. The sense of coherence scale shows high internal consistency [49].

The three-factor eating questionnaire (TFEQ-R18) is used to measure eating behavior tendencies [50], which has three domains: cognitive restraint (6 items), uncontrolled eating (9 items) and emotional eating (3 items). Responses are measured on a 4-point Likert scale, except for one item using an 8-point Likert scale. The range of total scores are from 0–100. To calculate the scores of the subscales, items 1–13 are reverse-coded, so that higher scores indicate a higher level of eating behavior tendencies. Cronbach’s alphas for the three domains were 0.71, 0.88 and 0.89 respectively (Swedish sample).

6-FQ is a 27-item self-administered instrument with psychometric properties that measures lifestyle pattern factors: Convenient Diner, Fast Pacer, Easily Enticed Eater, Exercise Struggler, Self-Critic and All-or-Nothing-Doer. Responses are measured on a 4-point Likert scale: “don’t agree at all” (0), “agree a little” (1), “agree” (2) and “strongly agree” (3). The score range is from 0–81, and a higher score indicates unhealthy lifestyle pattern factors. The Cronbach’s reliability estimates for internal consistency ranged from 0.76 to 0.85 (sample of overweight or obese adults) [51].

Background information questionnaire: The maternal background characteristics including age, BMI, marital status, education and employment status are collected with a self-administered background information questionnaire.

Patient record data.

Data on pregnancy and delivery is comprised of gestational weight gain, postpartum weight retention, blood pressure, pregnancy complications, birth outcomes, newborn data such as birthweight and LGA. In addition, data is collected about several other maternal items: chronic diseases, psychological history, use of medications, diagnoses and complications in previous pregnancies (such as, GDM, preeclampsia, preterm delivery (<37 weeks of gestation) and miscarriage).

Well-being data.

Data on dietary intake that is collected with the mobile app (FatSecret), and the nutritional intake (Calories, Protein, Fat and Carbohydrate) are both included in analysis. The app also displays the data by meal and snack eaten to observe meal patterns. FatSecret has been shown to be accurate and practical [52]. The outcome variable of interest in this study is the total kilocalories per day.

Data on activity, sleep and stress is collected with the Oura-ring. Studies in the literature have indicated the reliability of the ring in terms of sleep, physical activity, heart rate, and heart rate activity [53–55].

Data analysis.

The data will be analyzed statistically. Appropriate descriptive statistics will be calculated for the background variables, WEL, PASE, weight change, EPDS, WHOQOL-BREF, PRAQ-R2, PSS-4, SOC-13, TFEQ-R18, 6-FQ, calories consumed, and the Ouras parameters. Variables like the EPDS, baseline BMI and Ouras wear time will be tested for their association with changes in self-efficacy in eating and physical activity. Mixed effect models will be used to evaluate change over time in self-efficacy in eating and physical activity and weight change. Simple mediation models will be used to assess calories consumed and MVPA as mediators between self-efficacy and weight change.

Machine learning and event mining approaches are being exploited in the intervention to build personal models–such as sleep and physical activity–for each individual. The relationships between the collected parameters, including weight gain, physical activity, sleep, and stress will also be investigated. Signal processing and (for example, unsupervised and semi-supervised) machine learning techniques are being exploited to extract events–i.e., high abstraction level information–from the data collected via the Oura ring and the questionnaires. These events characterize pregnant women’s daily life or lifestyle: e.g., their physical activity during the day or stress levels during sleep and these events are defined based on different attributes in certain epochs varying from minutes to a day. The events will be divided into three categories: activity, health, and context. Subsequently, the relationships between the events and their attitudes will be investigated via statistical analysis techniques. For example, when an undesired event (e.g., stress) is detected during pregnancy, statistical analysis techniques (e.g., Chi-Square test) will be used to evaluate the relationships between the event and the mother’s activity, health, and context. Moreover, higher data abstraction will be performed to record events in a chronological order. Data analytic techniques ranging from simple rule-based algorithms to machine learning classifications will be developed to extract patterns from the collected data and events. To this end, machine-learning-based methods (e.g., autoencoders) will be developed to build personalized models, by which the pregnant woman’s real state is compared with the desired state.

Process evaluation of the implementation of the SLIM intervention.

A process evaluation is being conducted during the intervention period using a mixed-methods approach to examine the feasibility and acceptability of the intervention, the implementation strategy and the study processes including the data collection [23]. To enable this evaluation, the components of the primary intervention (i.e., its intervention and delivery strategies), and the implementation outcomes (in terms of acceptability, fidelity, mechanism of impact and contextual influences) are being defined with the help of a logic model (Fig 3) [56, 57]. To evaluate the reach of the implementation, both the number of acceptances to participate and the number of refusals have been documented. In addition, the level of missing data in returned questionnaires is being assessed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Logic model of implementation of SLIM intervention.

https://doi.org/10.1371/journal.pone.0279696.g003

Data collection for process evaluation.

Public health nurses. The PHNs perception of the appropriateness, acceptability and feasibility of the intervention is being evaluated with validated AIM, IAM & FIM questionnaires [58] every six months. In addition, all PHNs will be invited to take part in focus group exit interviews to assess the fidelity and adoption of the intervention at the end of the study. A semi-structured interview guide will be used and questions will aim to explore if the intervention was delivered as planned and whether adaptations to the intervention were made during delivery. Beliefs about the quality of the intervention delivery, the capacity and resources needed to deliver the intervention and the contextual influences on intervention implementation will also be explored. In addition, in order to assess the fidelity of the intervention, PHNs will complete a logbook of those contents of the participant’s maternity care appointments that are related to the SLIM intervention.

Overweight women. Women’s satisfaction with and usage of the SLIM intervention will be assessed using a validated questionnaire measuring appropriateness, acceptability and feasibility (AIM, IAM & FIM) of the intervention [58]. In addition, women will be asked to participate in a semi-structured phone interview at the end of the follow-up. A semi-structured interview guide will be used that will include questions aiming to capture the women’s perspectives on using the intervention. The daily wear time of the Oura-ring will be used to assess the fidelity of the intervention.

The Acceptability of Intervention Measure (AIM), Intervention Appropriateness Measure (IAM), and Feasibility of Intervention Measure (FIM) are four-item validated measures of implementation outcomes. Responses are measured on a scale from 1 to 5: 1 = Completely disagree, 2 = Disagree, 3 = Neither agree nor disagree, 4 = Agree, 5 = Completely agree. Scales can be created for each measure by averaging responses. The score is calculated by taking the sum of the four items. Higher scores indicate greater acceptability, appropriateness, or feasibility. Cronbach’s alphas should be 0.85 for acceptability, 0.91 for appropriateness, and 0.89 for feasibility using the 4-item scales from the structural validity survey, and 0.83 for acceptability, 0.87 for appropriateness, and 0.88 for feasibility using the test-retest reliability survey [58].

For those measures used in this study that did not have a validated translation, we undertook a translation from English to Finnish. Following the guidelines for translation and cultural adaptation of research questionnaires [59], the material was first translated by a professional translator to the target language. Next, the researchers (JS and HNV) made minor modifications to reflect the local cultural context, and this version was subsequently translated back to English by another independent translator. After back-translation, the accuracy of the translation was verified by the researchers (JS and HNV) and any discrepancies discussed until a consensus was reached. The AIM, IAM & FIM was translated in collaboration with its original developer, Prof. Bryan Weiner.

Data analysis of the process evaluation.

Qualitative data, recorded interviews, will be transcribed verbatim and transferred to a NVivo Version 12.1.0 template. Data will be analyzed using thematic analysis [60]. Quantitative data will be analyzed statistically. Descriptive statistics will be calculated to AIM, IAM and FIM questionnaires and logbooks. Qualitative and quantitative data will be integrated as a part of the process evaluation. The change in AIM, IAM and FIM will be analyzed between pregnancy week 34 and 12 weeks after birth, using paired t-tests or Wilcoxon signed rank tests.