Background
Gestational diabetes mellitus (GDM) is one of the most common disorders in pregnancy and now affects around 12% of pregnancies in Australia [
1,
2]. The short-term risks of GDM which includes large-for-gestational age babies, interventional delivery and birth trauma are reduced when GDM is well managed [
3,
4]. In the long-term, women with GDM and their infants are at an increased risk of cardio-metabolic disorders and type 2 diabetes mellitus [
5]. The primary intervention for GDM is lifestyle changes, including medical nutrition therapy provided by a qualified dietitian [
6].
There is strong evidence that women with GDM who are provided with individualised medical nutrition therapy from a qualified dietitian over a minimum of three appointments are less likely to require medication [
7‐
9]. This minimum schedule of dietetic appointments was first recommended in the Academy of Nutrition and Dietetics Nutrition Practice Guidelines [
10] and has since been incorporated into the 2015 Queensland Clinical Guideline for GDM [
11]. Specifically, the minimum schedule of appointments consists of a 1 h initial counselling session, a minimum of two review appointments and one postnatal follow up visit [
11]. Reviews should be scheduled on a two to four weekly basis according to clinical need and further reviews are recommended if pharmacological treatment is initiated [
11]. A recent study has demonstrated that the Queensland Clinical Guideline for GDM has been poorly implemented across Queensland Health with fewer than one-third of the hospitals achieving the minimum schedule of dietetic appointments [
12].
The translation of clinical guidelines into practice rarely occurs in the short-term without a systematic approach to implementation [
13]. Implementation science attempts to address the gap between best-available evidence and clinical practice, and successful implementation involves a multi-faceted approach that considers individual, local and system-based influences on change [
14,
15]. As implementation science has grown in popularity, so too have the theories, models and frameworks used to describe, understand or evaluate an implementation process [
16].
One well known framework for implementation is the Promoting Action on Research Implementation in Health Services (PARIHS) [
17]. The PARIHS framework has undergone considerable content and construct validity, refinement and testing since its original inception in 1998, and has recently been updated to increase its usability as the
integrated or
i-PARIHS framework [
18‐
21]. The
i-PARIHS framework proposes that successful implementation consists of the following three core constructs: the innovation; the context in which the change is to be implemented; and the intended recipients [
19]. Each construct is considered at multiple levels (local, organisational and outer layers) and the final element, facilitation, is the ‘active ingredient’ that combines all constructs to enable implementation [
19]. The
i- PARIHS framework is described as a determinant framework used to understand which constructs or domains acts as barriers and enablers to influence implementation outcomes [
16]. However, despite the authors’ intention for
i-PARIHS to be used as a prospective framework to guide implementation processes, it has been predominately used as an evaluation tool [
19]. The
i-PARIHS framework was selected as a development, implementation and evaluation tool for this project in conjunction with the
i-PARIHS Facilitation Guide due to its alignment with health service research and the specific training undertaken by two of the authors [
18].
An opportunity existed to change service delivery to improve women’s access to dietetic input for GDM through the implementation of the minimum schedule of dietetic appointments as recommended by the Queensland Clinical Guideline at a regional hospital in Queensland, Australia [
11]. The recommended schedule of care had been successfully implemented using a theory-driven implementation science approach at three other hospitals in Queensland, Australia [
8,
9]. The research team for this study saw an opportunity to prospectively use the
i-PARIHS framework to guide the implementation and so increase the chances of a successful change while uniquely contributing to the existing
i-PARIHS implementation science literature. The aim of this study was two-fold: to evaluate the impact of the dietitian-led model of care for women with GDM on clinical outcomes; and to understand the utility of the
i-PARIHS framework to prospectively guide an implementation process.
Results
There were 169 women who were referred for GDM management in the pre-intervention group and 141 women in the post-intervention group. After excluding women who did not birth at the public hospital, 125 women were included in the pre-intervention group and 119 in the post-intervention group. Prior to the model of care changing, dietetic staff resourcing dedicated to GDM was 0.25 full time equivalents (FTE), which was increased to 0.4 FTE after the model of care was implemented. The increase was possible due to increased dietetic staffing allocations which had occurred just prior to the study commencing. GDM staff resourcing for diabetes educators and obstetric physicians remained unchanged. However, in practice, the requirement for diabetes educator resourcing reduced by approximately 0.2 FTE.
Maternal characteristics
A comparison of the maternal characteristics for the pre- and post-intervention group are reported in Table
2. Of the characteristics which are known predictors for pharmacotherapy use in GDM [
31], age and pre-pregnancy body mass index (BMI) were similar between the two groups (Table
2). However, nearly three times more women were diagnosed early (before 24 weeks gestation) in the post-intervention group (21% vs 7.3%,
p < 0.01), whereas women in the pre-intervention group were more likely to be diagnosed on a fasting BGL (
P = 0.09), and were also somewhat more likely to have reported a family history of diabetes (
P = 0.17) (Table
2).
Table 2
Maternal characteristics for the pre-and-post intervention groups, before and after the GDM model of care
Total participants | 125 | 119 | |
Age (SD), years | 32.2 (5.7) | 32.7 (5.9) | 0.81 |
Gestational age at diagnosis (SD), weeks | 26.9 (3.6) | 25.6 (5.6) | < 0.01 |
Early diagnosis (under 24 weeks), n (%) | 9 (7.3%) | 25 (21.0%) | < 0.01 |
Diagnosis based on fasting result, n (%) | 53 (43%) | 39 (33%) | 0.09 |
Diagnosis based on 2 or more results, n (%) | 38 (31%) | 27 (24%) | 0.20 |
Parity (SD) | 1.1 (1.2) | 0.8 (1.1) | 0.78 |
Nulliparous, n (%) | 50 (40%) | 56 (47%) | 0.29 |
PP BMI (SD), kg/m2 | 27.1 (6.4) | 27.9 (7.7) | 0.19 |
PP BMI underweight, n (%) | 6 (5.0%) | 3 (2.9%) | |
PP BMI normal weight, n (%) | 47 (39%) | 40 (38%) | |
PP Overweight, n (%) | 29 (24%) | 29 (28%) | |
PP Obese (BMI ≥30 kg/m2), n (%) | 39 (32%) | 32 (31%) | 0.81 |
Indigenous Status, n (%) | 2 (1.6%) | 5 (4.2%) | 0.23 |
Previous GDM, n (%) | 19 (16%) | 24 (20%) | 0.37 |
Family History Diabetes, n (%) | 58 (49%) | 48 (40%) | 0.17 |
Smoking, n (%) | 9 (8.9%) | 15 (13%) | 0.37 |
Pre-pregnancy hypertension, n (%) | 9 (7.5%) | 4 (3.3%) | 0.16 |
Polycystic Ovarian Syndrome, n (%) | 8 (6.7%) | 9 (7.5%) | 0.77 |
Adherence to dietetic schedule of appointments
There was a large increase in dietetic appointments in the post-intervention group and the number of women achieving the minimum number of dietetic appointments greatly increased (82% vs 29%,
p < 0.001). There was also an increase in the mean number of appointments with the diabetes team obstetric physician in the post-intervention group, but a decrease in the number of diabetes educator appointments (Table
3). This trend was similar when only women diagnosed from 24 weeks was analysed (Table
3).
Table 3
Dietetic, diabetes educator and obstetric physician appointments, before and after changing a GDM model of care
Dietitian, number of appointments, mean (SD) | 2.4 (0.8) | 3.8 (1.7) | < 0.001 | 2.4 (0.8) | 3.7 (1.4) | < 0.001 |
Adherence to dietetic schedule of appointments, n (%) | 36 (29%) | 98 (82%) | < 0.001 | 33 (29%) | 79 (84%) | < 0.001 |
Obstetric Physician, number of appointments, mean (SD) | 1.9 (2.5) | 2.5 (2.8) | 0.08 | 1.8 (2.3) | 2.3 (2.4) | 0.11 |
Diabetes Educator, number of appointments, mean (SD) | 3.1 (2.0) | 2.6 (2.6) | 0.10 | 3.0 (1.4) | 2.3 (2.0) | 0.007 |
Total appointments, mean (SD) | 7.4 (4.3) | 8.9 (5.1) | 0.01 | 7.1 (3.6) | 8.3 (3.9) | 0.02 |
Pharmacotherapy use
Pharmacotherapy use increased in the post-intervention group by 10% (47% v 37%,
P = 0.10), a clinically significant increase (Table
4). Insulin use more than doubled in the post-intervention group and metformin use was halved (Table
4). The gestational age for commencing pharmacotherapy was also earlier in the post-intervention group (27.8 weeks vs 30.1 weeks,
p = 0.05) (Table
4). Total pharmacotherapy use in women diagnosed from 24 weeks was 36% vs 44% (
p = 0.18), an increase of 8% in the post-intervention group (not shown).
Table 4
Pharmacotherapy use and maternal and infant outcomes for the pre-and-post intervention groups
Binary outcomes |
Requiring any pharmacotherapy, n (%) | 46 (37%) | 56 (47%) | 0.10 | 0.15 | 1.53 (0.86–2.81) |
Metformin | 20 (16%) | 10 (8.3%) | | | |
Insulin | 14 (11%) | 33 (28%) | | | |
Metformin + insulin | 12 (10%) | 14 (12%) | | | |
Large-for-gestational age, n (%) | 10 (8.0%) | 12 (10.1%) | 0.57 | 0.56 | 1.30 (0.54–3.15) |
Small-for-gestational age, n (%) | 10 (8.0%) | 12 (10.1%) | 0.57 | 0.81 | 1.12 (0.45–2.79) |
Continuous outcomes |
Gestational age for pharmacotherapy, mean (SD), weeks | 30.1 (4.7) | 27.8 (6.8) | 0.05 | | |
Infant birthweight, mean (SD), grams | 3352 (499) | 3290 (470) | 0.32 | | |
Infant birthweight
The mean infant birthweight decreased by 62 g in the post-intervention group (Table
4). Large-for-gestational age and SGA infants increased by 2% in the post-intervention group (Table
4). Interestingly, of the ten LGA infants in the pre-intervention group, 40% (
n = 4) of the mothers were treated with pharmacotherapy whereas in the post-intervention group, 67% (
n = 8) of the mothers were treated with pharmacotherapy.
NoMAD survey instrument
Five of the 11 staff involved in the GDM team completed the NoMAD survey both before and after the change, a response rate of 45%. Due to the anonymity of the survey, it is not known whether they were the same five staff each time. There was no change in the responses from before and after the model of care was implemented for 59% (n = 13) of the questions. Specifically, all responses were positive before and after the model was changed, indicating that the staff felt the model of care was familiar and they had an understanding of the purpose of the model of care, how it affected the nature of their work and the potential value of the model. There was also agreement that the staff felt there were key people driving the model of care forward, that participating in the model was a legitimate part of their work and they would continue to support the model.
Prior to the model of care changing, one of the five staff did not agree with the following statements: they could easily integrate the model of care into their existing workload; sufficient training was provided to enable staff to implement the model; and that staff agree the model of care is worthwhile. Two staff did not agree that sufficient resources were available to support the new model of care. After the model had been implemented one respondent did not agree that sufficient resources were available, and a separate respondent did not agree that they could modify how they worked with the model of care.
The i-PARIHS framework
The study processes as they occurred within the
i-PARIHS framework is summarised in Table
5, including a description of the pre-post models of care, literature review findings and facilitation activities. It was discovered that the
i-PARIHS framework was most useful during the development phase to understand specific barriers and enablers within each construct and as a diagnostic tool and checklist when developing the innovation and preparing for implementation. While the
i-PARIHS framework was also useful as a reflection tool in the evaluation phase, particularly within each construct, the authors did not feel that the
i-PARIHS framework uniquely contributed to evaluation as it did with the development phase.
Table 5
The development and implementation of a GDM dietitian-led model of care using the i-PARIHS framework
Development Phase |
Overview | Starting point: • Minimum schedule of dietetic appointments (Queensland Clinical Guideline for GDM) • Goal to increase women’s access to dietetic support and reduce pharmacotherapy requirements. Organisational fit: • Task duplication identified • Low and high-risk models of care (diet-controlled vs pharmacotherapy + diet) • Models of care: Low risk as dietitian-led, high risk as diabetes educator and physician led • Increased surveillance for low-risk GDM patients (due to third dietetic appointment) • Timing of appointments and changes to ongoing monitoring of all women with GDM. Supporting material: • Escalation of care flow chart for dietitians • Low and High-risk model of care summary flowcharts • Updated patient information • Pre-implementation checklists | Recipients (Staff): • Diabetes team members: Dietitians, Diabetes Educators, Nursing Unit Manager, Clinical Nurse Consultant, Director of Endocrinology, Obstetric Physicians, Administration Officers. • Working party: Clinical Nurse Consultant (opinion leader/ authority), Dietitians (champions/ opinion leader), Nursing Unit Manager (authority), Diabetes Educators (champions) | Local: • Increasing GDM diagnosis requiring efficient model of care • Task duplication within the team • Leadership change Organisational: • Change to organisational structure. • Period of transition (opening of new hospital). External Health Systems: • State-wide publication of Clinical Guideline for GDM (2015) | Problem identification: • Clinical guideline recommendation for MNT not met Acquiring/appraising evidence: • Prior research (Surveys) [ 12, 23] • Service mapping Consensus building: • Stakeholder mapping and engagement • Team meetings • Goal setting • Local context assessment: • Diagnosis using i-PARIHS guidance • Model of care development meetings • Working party contributions |
Barriers | • Staff resourcing • Education/knowledge • Managing schedule of appointments | • Some resistance to change (minor) • Competing interdisciplinary priorities • Differences of opinion • Perceived workload pressures • Motivation and engagement | Local: • Historical resistance to change • Team culture Organisational: • Period of high organisational change and transition | Project management: • Increase to dietitian FTE/ clinic days • Appointment template changes • Working party meetings • Newsletters/ email updates Improvement methods: • Professional development sessions • Team meetings Conflict management and resolution: • Leadership involvement • One-on-one meetings Team building • Team meetings • Acknowledging key contributions |
Enablers | • Strong evidence-base • State-wide guidelines • Well-established team • Dedicated researcher | • Leadership support • Local opinion leaders/ champions • Minimal disruption to usual workflow • Individuals and team able to implement change • Low staff turnover | Local: • Team autonomy • Leadership support Organisational: • Executive support • Alignment with organisational and research priorities External Health System: • State-wide mandate | Team building: • Acknowledging enablers • Feedback |
Implementation Phase |
Intervention/ change in practice | • New schedule of dietetic appointments and reduction of diabetes educator appointments • Dissemination of supporting materials | • Increase to dietetic staffing time for GDM • Procedures and policies to inform local system changes | • Procedures and policies to inform local system changes • Informed stakeholders and executive of change to model of care | Communication and feedback: • Fortnightly meetings • Newsletters/ email updates Conflict management and resolution: • One-on-one meetings • Leadership involvement |
Evaluation Phase |
Successes | • Adherence to schedule of dietetic appointments (29% vs 88%) | • NoMAD survey: familiar, understanding of purpose, support for the model of care, change in negative perceptions | Local: • Dietitian-led model of care adopted as standard practice | |
Confounders | • Appointment timing deviated from original Academy of Nutrition and Dietetics Nutrition Practice Guidelines • Initial education as group rather than individual • Fidelity: patient satisfaction survey not implemented • Sustainability: FFQ data collection not completed at second review | • Lack of perceived value for understanding patient satisfaction and FFQ • Significant differences in baseline characteristics between pre-and-post intervention groups (early diagnosis, family history of diabetes mellitus, previous diagnosis of GDM) | Local: • Increased surveillance of women with GDM to the end of their pregnancy | Communication and feedback: • Newsletters/ email updates • Post-implementation presentation to team members |
Discussion
The aims of this study were to evaluate the impact of a dietitian-led model of care for women with GDM on clinical outcomes and to understand the utility of the
i-PARIHS framework as a prospective tool in an implementation process. We used a theoretical-approach [
19] to develop, implement and evaluate the changes to the GDM model of care in order to create real change in the care of women with GDM. As a result, adherence to the minimum schedule of dietetic appointments (one initial education and at least two review appointments) was greatly increased after implementing the model of care changes.
Despite improving dietetic input, requirements for pharmacotherapy increased in the post-intervention group by 10%, a clinically significant increase. It is thought this result was due to important differences between the two groups rather than the model of care itself. A strong independent predictor of requiring pharmacotherapy to treat GDM is an early diagnosis (before 24 weeks gestation) [
31,
34,
35]. The number of women diagnosed early in the post-intervention group was almost three times that in the pre-intervention group. There were also differences in the percentage of women who had previously been diagnosed with GDM, with 4% more women in the post-intervention group reporting a previous GDM diagnosis [
31]. However, despite adjusting for these confounders in logistic regression modelling and performing a post-hoc analysis on women diagnosed from 24 weeks, there was still a clinically significant increase in women requiring pharmacotherapy after increasing dietetic input.
One of the biggest factors that likely influenced pharmacotherapy use was increased surveillance by the diabetes team. Prior to the model changing, women who achieved their BGL targets at the first review appointment were often discharged from the diabetes team and it was expected their BGLs would continue to be monitored by their usual antenatal care providers. Once the model of care changed, women in the post-intervention group were followed by the diabetes team to the end of their pregnancy and women with elevated BGLs beyond the first review appointment were more likely to be picked up and referred to the obstetric physician for pharmacotherapy.
There were other factors that may have impacted pharmacotherapy requirements for women in the post-intervention group. The timing of the dietetic appointments was based on consultation with the implementation recipients (diabetes team members) and most women were seen at fixed intervals: initial education within a week of diagnosis; first review appointment a week later; and second review appointment between 34 and 36 weeks gestation. On reflection, appointments with a dietitian are likely to be more effective in the early stages of diagnosis where women are most able to make positive dietary and lifestyle changes with intensive support [
6]. While close attention was paid to the adherence to the minimum schedule of dietetic appointments, we were unsuccessful at implementing the timing recommended in the Academy of Nutrition and Dietetics Nutrition Practice Guidelines [
10]. By scheduling the second review appointment at 34 weeks gestation, there was effectively little difference between the amount of dietetic support in the pre-and post-intervention groups in the weeks following diagnosis.
Prior research on dietetic input for women with GDM [
7], including studies undertaken at three Queensland hospitals [
8,
9], has shown that women who achieve a minimum of three dietetic appointments are less likely to require pharmacotherapy. In the Queensland studies, only one of the three hospitals achieved a similar level of adherence to the minimum schedule (one initial and two reviews) as our study, yet all hospitals achieved clinically relevant reductions in pharmacotherapy [
8,
9]. Interestingly, one of the hospitals only marginally improved their adherence to the minimum schedule of appointments (4.8% of women received at least 3 appointments vs 3.4%), yet still achieved a 9.1% decrease in pharmacotherapy requirements [
9]. However, the number of women receiving individual dietetic appointments as the initial consult increased from 2 to 43%, indicating that individual education may be a more important than the number of appointments women receive [
9]. In this study, the initial consult for women was always via group education, therefore it is likely our results may have improved had we been able to change this to an individual format. Furthermore, previous research [
7‐
9] did not evaluate outcomes for women diagnosed before 24 weeks, a predictor for pharmacotherapy requirements, as previously described. Finally, in the study by Reader, Splett and Gunderson assessing the impact of the Academy of Nutrition and Dietetics Nutrition Practice Guidelines, the overall results demonstrated a decrease in insulin use at sites adhering to the Nutrition Practice Guidelines but this difference was not detected at sites where women with GDM were managed by a specialist diabetes team [
7]. In the present study, both the pre- and post-intervention groups were managed by a specialist diabetes team. It is likely that the interplay between the local context (specialist diabetes centre), the characteristics of the women, and the initial education session and timing of appointments (innovation) meant we were unable to achieve a reduction in pharmacotherapy as predicted.
Despite the increased pharmacotherapy use following the model of care changes, the overall development and implementation of the model of care was considered successful as defined by achieving adherence to the minimum schedule of dietetic appointments for more than 75% of women. According to the NOMAD survey, the changes were easily understood and valued by the staff who responded and since the completion of this study, the model of care has been accepted as standard practice within this team.
The prospective use of the i-PARIHS framework provided structure and guidance during the development and implementation phases of this study while highlighting the important role of the facilitator. During the evaluation phase, it also provided the researchers with key constructs to consider as influencing factors to the clinical outcomes. In contrast to much of the previous research which has focused on the retrospective application of the i-PARIHS framework, we found i-PARIHS was most useful during the development phase as a diagnostic tool and checklist. However, during the evaluation phase we were able to reflect on the influencing factors to the outcomes within each i-PARIHS construct.
Despite using a theoretical approach to implementation, not all aspects of changing the model of care were adhered to, demonstrating some of the difficulties with real world health services research. For example, collection of the food frequency questionnaire was incomplete and the patient satisfaction survey was never implemented, highlighting issues with fidelity, acceptability, and sustainability. It is possible the recipients of the implementation did not understand the value of this data, an aspect overlooked by the researchers. Furthermore, collaboration and negotiation with recipients was an important change management strategy but this negotiation meant specifying the follow-up schedule of appointments, resulting in a deviation from the Academy of Nutrition and Dietetics’ recommendations. Attempting to balance best available evidence with organisational ‘fit’ may have impacted the success of the outcome of pharmacotherapy use.
The main limitation of the study was the small sample size and differences in baseline characteristics between the two groups, thus we were unable to determine the true change in the outcome of pharmacotherapy. Due to its low response rate, the results of the NoMAD survey cannot be considered representative of all staff recipients. We were also limited in our evaluation with the exclusion of the pre-specified outcomes of the patient satisfaction survey and the food frequency questionnaire. It is possible that bias was introduced due to the main facilitator also being responsible for the evaluation and interpretation of data, although most data collection was performed by a research assistant who was not involved in data analysis. Most data were collected from chart entries, which has the potential to introduce inaccuracies due to incorrect data entry. Our study reported on a small group of pregnant women, residing in South Eastern Queensland and cannot be considered generalisable to the wider Queensland or Australian population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.