Determinants of use of care provided by complementary and alternative health care practitioners to pregnant women in primary midwifery care: a prospective cohort study

Data for this analysis was obtained from the DELIVER study (Dutch acronym for ‘data primary care delivery’) conducted by the Department of Midwifery Science of VU University Medical Center, Amsterdam. The DELIVER study is a descriptive study that aimed to provide information about the organization of midwifery care, the accessibility of midwifery care and the quality of primary midwifery care in the Netherlands [18].

In the DELIVER study, a two-stage sampling procedure was used. Firstly, midwifery practices were recruited by using purposive sampling. Three stratification criteria were used: region (north, east, south, west), level of urbanisation (urban or rural area), and practice type (dual or group practice) to ensure that different types of practices in different regions were represented. Subsequently, all clients receiving care in the participating primary midwifery practices at any moment in a 12 month study period in 2009–2010 were eligible to participate if they were able to understand Dutch, English, Turkish or Arabic. The participating practices (20 of the 519 midwifery practices in the Netherlands) comprised 110 midwives and a caseload of 8200 clients per year, with all regions of the Netherlands being represented [18].

Clients participating in the DELIVER study completed up to three questionnaires. The first questionnaire was administered before 34 weeks of gestation, the second between 34 weeks of gestation and birth, and the third in the postpartum period. In addition, information was collected about the care provided by midwives by extracting data from electronic client records of participating clients and from the Netherlands Perinatal Registry. The latter consists of information provided by midwives, GPs and obstetricians. Reporting to this Registry is obligatory. The three data sources were linked using unique, anonymous client and midwifery practice identifiers [18]. The Medical Ethics Committee of VU University Medical Center, Amsterdam approved the study protocol of the DELIVER study. Participants provided written informed consent.

Our study comprised those pregnant women who filled in the first and third questionnaires (up to 13 weeks postpartum) and whose questionnaire data could be linked to the electronic client record data and the Netherlands Perinatal Registry data. To maximize the homogeneity of the low-risk population, we excluded women who were referred to secondary care during pregnancy. Women who were referred during labour were classified as non-referred because the pregnancies of these women were low-risk (Fig. 1). All women filled in the questionnaires at home without interference from a professional. We used data from electronic client records with regard to two independent variables; 1. Health care utilization in midwifery practices (Fig. 2, health behaviours: ‘health care utilization of pregnant women in primary midwifery care), and 2. Parity (Fig. 2, need variables: ‘parity’). These variables were shown to be invalidly measured in the client questionnaires. We assume that midwives recorded visits to their practice and parities of women more validly.

CAM practitioner use was measured by two items in the third questionnaire of the DELIVER study: ‘Please indicate whether you have seen any of the following practitioners of complementary or alternative medicine since the beginning of your pregnancy’ and ‘What other practitioner(s) of complementary or alternative medicine did you see?’ For each practitioner, women had to specify contact rates in predefined categories (0, 1–3, 4–6, 7–9, 10–12, 13–15, >15 meetings). Various types of CAM practitioners were stated explicitly in the questionnaire: acupuncturist, anthroposophical practitioner, homeopath, manual therapist (chiropractor, osteopath, manual therapist), naturopath (diet therapy, neural therapy, herbal therapy) or paranormal practitioner (psychic, faith healer, magnetic therapist), and respondents also had the option to choose other alternative practitioner. Women who reported at least one consultation with a CAM practitioner were defined as CAM users.

Potential determinants of CAM practitioner use concerned predisposing, enabling, need and health behaviour variables. Data on possible determinants were obtained from the first questionnaire and the electronic client records. Several variables, based on Andersen’s model [17], were considered to be potential determinants of CAM practitioner use. In the Andersen model use of health services depends on individual and contextual characteristics, and on health behaviour. The following components were measured: predisposing, enabling, need, and health behaviour characteristics. Predisposing characteristics are existing conditions that predispose people to use (yes/no) healthcare services. Enabling/disabling characteristics facilitate or impede use. Need characteristics are conditions that patients or health providers recognize as requiring medical treatment. Health behaviour characteristics are behaviours on the part of the individual that influence health status [17]. Potential determinants were categorized into one of these components by using existing literature of the Andersen’s model and by discussion of the authors.

Operationalizations of the independent variables are shown in Fig. 2.

Predisposing variables encompassed socio-demographic and belief factors, consisting of age, ethnicity, marital status, occupation, educational level, intended place of delivery, and religion. Enabling variables included finance (health care insurance) and organization (accessibility of care) variables. Regarding health insurance, we distinguished between basic and supplementary health care insurance.

Need variables comprised the health status of the client. The descriptive component of EuroQol (EQ) was used to measure self-reported health status [19]. This component asked the respondent to consider and rate her actual health on five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Responses to questions on each of these dimensions can take one of five values, which concern five levels of severity (no problems/slight problems/moderate problems/severe problems/extreme problems). Health status values ranged from extreme problems on all five dimensions (value =-0.109) to no problems on any dimension (value = 1.0). A single health status value was calculated by applying scores from a UK valuation set [19]. We then dichotomized the scores as ‘poor’ (lowest quartile) and ‘the remainder’. Feelings towards pregnancy were measured by using the Pregnancy Related Anxiety Questionnaire (PRAQ) [20]. The scales used were ‘fear of giving birth’ (two items), ‘fear of bearing a handicapped child’ (four items) and ‘concern about one’s appearance’ (three items). Items were scored on a four-point scale (4 = very true, 3 = true, 2 = not true, 1 = certainly not true). Every item score was dichotomized based on the median score. BMI was calculated using the weight and height before pregnancy reported by the respondent. We classified BMI according to the World Health Organization classification of adult underweight, normal weight, overweight and obesity [21]. Finally, we computed a variable ‘gravidity/parity difference’ which measured the difference between the number of pregnancies and the number of deliveries. We hypothesised that there could be a difference in prenatal health care use between women with miscarriage(s) and/or abortion(s) in their obstetric history.

Health behaviour variables consisted of questions related to smoking, soft and hard drug use, alcohol use, adequate folic acid use, locus of control and adequacy of prenatal health care utilization of pregnant women in primary midwifery care. We did not include drug use because none of the pregnant women reported drug use, which concurs with our sampling of low-risk pregnancies [18]. The locus of control was measured by a single question about the extent of the perceived possibility of influencing lifestyle and/or health behaviour (‘To what extent do you feel that you can influence your health by changing your lifestyle and/or behaviour?’). Folic acid use was labelled as adequate when started at least four weeks before pregnancy [22]. Adequacy of prenatal health care utilization of women in primary midwifery care was measured using the Kotelchuck Index, which is widely used in the US [23]. We constructed a revised assessment index of the adequacy of prenatal care use in Dutch primary midwifery care (Table 1), modified according to the guidelines of the Royal Dutch Organization of Midwives, concerning the number of prenatal visits during pregnancy. This index combines the timing of initial prenatal health care and the number of prenatal health care visits. Prenatal care entry regarded on the gestational age at the first prenatal visit and classified into ‘timely’ (gestational age at onset < 12 weeks) and ‘late’ (gestational age at onset ≥ 12 weeks). The number of prenatal visits was derived from the electronic client record, and compared to the “expected” number of visits as described by the Dutch prenatal guideline for primary midwifery care taking the gestational age at which women gave birth into account. Adequacy of prenatal health care utilization was trichotomized into ‘adequate plus’, ‘adequate’ and ‘inadequate’ (inadequate and intermediate) care.

First, we described the background characteristics of the study population, and second the prevalence of CAM practitioner use. Third, we performed univariable logistic regression analyses for all determinants. Next, we performed multivariable logistic regression with a backward selection procedure, i.e., stepwise deletion of the variables that contributed least to the model that predicts use of CAM practitioners until all remaining variables contributed significantly at p < 0.05 level. The results are presented as odds ratios (ORs) and 95 % confidence intervals (CI). Women reporting no use of a CAM practitioner were our reference group. The structure of the data was hierarchical, i.e., respondents were clustered by midwifery practice. Characteristics of practices may affect all women who received care in that practice, which might lead to dependency of data regarding women coming from the same practice [24]. To adjust for this potential clustering, multilevel analytical methods were used. A two-tailed p-value of 0.05 or lower was considered statistically significant. Missing data accounted for less than 1.5 % of all variables, with the exception of 6.5 % for BMI. SPSS 21.0 (SPSS Inc., Chicago, IL) was used for all analyses.