Medical, behavioural and social preconception and interconception risk factors among pregnancy planning and recently pregnant Canadian women

Objectives The objective of this study is to describe the clustering of medical, behavioural and social preconception and interconception health risk factors and determine demographic factors associated with these risk clusters among Canadian women. Design Cross-sectional data were collected via an online questionnaire assessing a range of preconception risk factors. Prevalence of each risk factor and the total number of risk factors present was calculated. Multivariable logistic regression models determined which demographic factors were associated with having greater than the mean number of risk factors. Exploratory factor analysis determined how risk factors clustered, and Spearman’s r determined how demographic characteristics related to risk factors within each cluster. Setting Canada. Participants Participants were recruited via advertisements on public health websites, social media, parenting webpages and referrals from ongoing studies or existing research datasets. Women were eligible to participate if they could read and understand English, were able to access a telephone or the internet, and were either planning a first pregnancy (preconception) or had ≥1 child in the past 5 years and were thus in the interconception period. Results Most women (n=1080) were 34 or older, and were in the interconception period (98%). Most reported risks in only one of the 12 possible risk factor categories (55%), but women reported on average 4 risks each. Common risks were a history of caesarean section (33.1%), miscarriage (27.2%) and high birth weight (13.5%). Just over 40% had fair or poor eating habits, and nearly half were not getting enough physical activity. Three-quarters had a body mass index indicating overweight or obesity. Those without a postsecondary degree (OR 2.35; 95% CI 1.74 to 3.17) and single women (OR 2.22, 95% CI 1.25 to 3.96) had over twice the odds of having more risk factors. Those with two children or more had 60% lower odds of having more risk factors (OR 0.68, 95% CI 0.52 to 0.86). Low education and being born outside Canada were correlated with the greatest number of risk clusters. Conclusions Many of the common risk factors were behavioural and thus preventable. Understanding which groups of women are prone to certain risk behaviours provides opportunities for researchers and policy-makers to target interventions more efficiently and effectively. This study was part of a large cross-sectional survey of preconception care attitudes, beliefs and intervention preferences of women and men across Canada, undertaken in May to June 2019. Participants were recruited via advertisements on public health unit websites and social media accounts and parenting webpages, referrals from ongoing studies and identification of eligible individuals through existing research datasets. Women and men were eligible to participate if they could read and understand English, were able to access a telephone or the internet, were either planning a pregnancy (preconception) or had ≥1 child in the past 5 years and were thus in the interconception period. Individuals interested in participating in the study received an introductory email after contacting the research team. Those who were eligible and agreed to participate received a link to an online consent form and questionnaire using the Research Electronic Data Capture system. Research staff assisted individuals who had difficulty accessing the online questionnaire and sent reminder follow-up telephone calls. For this study, only women were included. This study was completed as formative work for a large randomised controlled trial evaluating a preconception–early childhood intervention on the prevention of child obesity among pregnancy-planning women and their partners (HeLTI Canada).15 The measures selected for this study were guided by the Centre for Effective Practice Preconception Health Care Tool, which is used in the province of Ontario to guide preconception and interconception healthcare during primary care visits.14 Twelve of the 15 risk categories of the tool were assessed: reproductive history, sexual health, chronic medical conditions, medications, mental health, tobacco use, alcohol and other substance use, infectious diseases, nutrition, weight status, physical activity and psychosocial stressors. Three categories were omitted due as they were not relevant for the future trial: vaccinations and immunity, family and genetic history, and environmental exposures. Women were defined as pregnancy planning if they indicated that they were currently trying to get pregnant, or considering a pregnancy in the next 5 years. Reproductive history was assessed using the question, ‘Have you ever experienced any of the following with a pregnancy?’ Response options were: miscarriage, stillbirth, use of artificial reproductive therapies (ART), uterine abnormalities, caesarean section (planned and unplanned), preterm birth, low or high birth weight, gestational diabetes, high blood pressure that developed during pregnancy and birth defects. Five sexually transmitted infections (STIs) were assessed to evaluate sexual health using the question, ‘Have you ever tested positive for any of the following in the past year?’ with response options: chlamydia, syphilis, trichomoniasis, gonorrhoea and genital herpes. Responses were combined to create an indicator for testing positive for any STI. Chronic medical conditions were assessed using the question, ‘Have you ever been diagnosed with any of the following conditions?’ Response options were: asthma, cancer, diabetes, hypertension, inflammatory bowel disease (IBD), phenylketonuria, renal disease, seizure disorder, systemic lupus erythematosus or rheumatoid arthritis or another autoimmune disease, thromboembolic disease and thyroid disease. Infectious diseases were evaluated using the same question as for chronic medical conditions but for the following response options: cytomegalovirus, hepatitis B, hepatitis C, HIV, parvovirus, toxoplasmosis and tuberculosis. An indicator was created for a diagnosis of any of the listed infectious disease. Medication use was evaluated using the question, ‘Do you currently use any of the following medications?’ with the response options: prescribed medications, over-the-counter medications and alternative/complimentary medications (including herbal, natural, and weight-loss medications, and athletic products or supplements). Mental health was ascertained using two screening tools. Depressive symptoms were measured using the Patient Health Questionnaire (PHQ-9),16 a 9-item scale assessing a range of potential symptoms experienced in the last 2 weeks. Response options range from ‘not at all’ (0) to ‘nearly every day’ (3). Items are summed to create a total score; those scoring >10 are considered to have moderate to severe depressive symptoms. This scale has been shown to be valid and reliable in similar populations17 and in our sample had a Cronbach’s alpha of 0.83. Anxiety symptoms were measured using the Generalised Anxiety Disorder (GAD-7),18 a 7-item scale assessing symptoms experienced in the last 2 weeks. Response options, scoring and threshold for identifying significant symptoms are the same as for the PHQ-9. The GAD-7 has been shown to be valid and reliable in similar populations19 and had a Cronbach’s alpha of 0.89 in our sample. Those who scored >10 on the PHQ-9 and the GAD-7 were considered to have comorbid symptoms of depression and anxiety. Tobacco use was assessed based on a question asking, ‘On a typical day, how many cigarettes do you smoke?’ Responses were collapsed into none vs any. Regular alcohol use was defined as drinking more than once per week, which was evaluated using the question, ‘How often do you drink a beverage containing any alcohol?’. Regular cannabis use was defined as at least monthly medicinal or recreational use based on the question, ‘In the past 12 months, have you used cannabis (marijuana) for non-medical/recreational reasons?’ A similar question was asked for medical drug use. For those with a positive response to either of those questions, another question was asked about frequency of use. The PRIMEScreen tool20 consists of 18 questions about the average frequency of consumption, over the previous year, of specified foods and food groups and another seven items about vitamin and supplement intake. It particularly targets intake of fruits, vegetables, whole and low-fat dairy products, whole grains, fish and red meat as well as other foods that are major contributors to the intake of saturated and trans fats. Those with a total score of <12 were categorised as having fair/poor eating habits. Weight status was assessed using self-reported weight and height, which were used to calculate body mass index (BMI) based on the kg/m2 formula. An indicator was created for those with calculated BMI ≥25 kg/m2 (ie, overweight or obese). Physical activity was evaluated using the WHO guidelines for healthy physical activity levels, which define low physical activity as getting less than 600 metabolic equivalent (METS) per week. The number of METS obtained weekly was estimated using the Global Physical Activity Questionnaire (GPAQ).21 Based on responses to the GPAQ, an indicator was created to identify those achieving less than 600 METS weekly. Psychosocial stressors included three measures: loneliness, low household income and unemployment. Loneliness was assessed using the three-item UCLA Loneliness Scale;20 22 individuals scoring >5 were defined as being lonely. Socioeconomic status was assessed using self-reported household income, which was defined as low if reported as <$C50 000 annually. Current unemployment was determined by self-report. The following maternal demographic variables were also assessed: age (years and categorised as <35 or ≥35), birth outside Canada (yes or no), marital status (married or common-law, and single, divorced, or widowed), number of children (<2 or ≥2) and education level (post-secondary degree, and less than postsecondary degree). A description of the sample was provided using summary statistics, including means and SDs for continuous variables, and frequencies and percentages for categorical variables. Analyses were undertaken in SPSS (V.25), unless otherwise indicated. Statistical significance was established at p<0.05. For each of the 12 risk categories investigated, the prevalence of having a risk factor was calculated by dividing positive cases by the sample of those with non-missing data on that risk factor. For multifactorial categories, prevalence was also calculated for having at least one risk factor from the category. Next, the total number of risk factors experienced by each woman was determined and the mean, median, and range were calculated, along with the SDs and IQR. To establish clusters of risk factors, exploratory factor analysis was undertaken using Mplus, V.7.23 Tetrachoric correlations were estimated to address the binary nature of the risk factors using the categorical option in Mplus. The analysis was undertaken using a robust mean and variance adjusted weighted least squares estimator. The Geomax rotation was used, which allows for correlation between the clusters. Model convergence was initially a problem due to the inclusion of risk factors that were present in a very small proportion of the sample. To obtain a reliable solution, therefore, we included only risk factors prevalent in >5% of the sample. To determine the optimal number of clusters to extract, parallel analysis was undertaken using a SAS V9.4 program.24 The goal of parallel analysis is to determine if the number of factors found in the solution accounts for more variation than the number of factors extracted using random data. Model fit was assessed using root mean square error of approximation (RMSEA<0.06, recommended), Comparative Fit Index (CFI>0.95), Tucker-Lewis Index (TLI<0.95, recommended) and standardised root mean square residual (SRMR <0.08, recommended).25 Once the optimal number of clusters was determined and adequate model fit was confirmed, the final solution was reviewed for clinical meaningfulness. Based on the number of risk categories established in objective 1, the sample was divided into two groups: those with the mean number of risk factors or fewer and those with greater than the mean number of risk factors. Using this measure as the outcome, a multivariable logistic regression model was specified to determine the demographic factors that were associated with having greater than the mean number of risk factors. Demographic variables were selected a priori including age, education level, marital status, parity and country of birth. These variables were selected because they were not directly included in the Centre for Effective Practice Preconception Health Care Tool, but were hypothesised to be related to many of the risk categories based on the literature. All independent variables were entered into the model simultaneously and left in regardless of p value. Next, for each of the risk clusters established in objective 2, a score was calculated based on the number of risk factors present for the individual. For example, for a cluster with three risk factors, women either received a score of 0, 1, 2 or 3, depending on if she had none of the risk factors, one, two or all the risk factors in the cluster. Spearman’s rho was calculated to determine how the demographic characteristics listed above related to the scores derived for each of the clusters. The sample included both pregnancy planning and recently pregnant individuals. To test if the findings would in those who were actively pregnancy planning vs those who were in the interconception period but not actively pregnancy planning, we ran a sensitivity analysis focusing on only women who reported that they were planning for a pregnancy (n=529).