GDM care re‐imagined: Maternal and neonatal outcomes following a major model of care change for gestational diabetes mellitus at a large metropolitan hospital

The rapidly rising prevalence of gestational diabetes mellitus (GDM) poses major challenges to the efficient, timely and sustainable provision of diabetes care.


INTRODUCTION
In company with most Australian and International maternity facilities, Mater Mothers' Hospital (MMH) Brisbane has seen a rapid increase in the prevalence of gestational diabetes mellitus (GDM) over the past decade.For MMH this has equated to an additional 200 women per annum requiring GDM care over the last triennium, with no substantial increase in staffing or resources.A lack of skilled credentialled diabetes educator (CDE) midwives in our geographical location and the impost of COVID-19 restrictions further exacerbated resource issues.
Traditional GDM diagnosis and care delivery varies widely across Australia and New Zealand. 1 Common features include diagnosis between 24 and 28 weeks gestation using a fasting 75-g oral glucose tolerance test, with subsequent care for women with GDM including group education with CDEs and/or dietitians, transition from standard antenatal care to a hospitalbased/high-risk model and then additional fortnightly faceto-face review of glycaemia and medication titration in many settings. 1With the high numbers at our facility and the complexity of scheduling, obstetric or midwifery appointments were conducted separately and often relied on the woman presenting to hospital twice in a week or occasionally on consecutive days.Pre-implementation service scoping review of our traditional GDM schedule revealed insufficient staffing to fulfil the best practice appointment schedules, further compounded by significant levels of staff stress and sick leave.
In addition, over 30% of our women have cultural and linguistically diverse (CALD) needs, some requiring interpreter services, rendering group sessions inappropriate and/or highly inefficient.
Across the whole cohort, a 30% 'failure to attend' rate further exacerbated the inefficiencies.At this time, we had no digital 'dashboard' overview of the GDM cohort to inform our care planning.
All women with GDM were seen at the same frequency regardless of their glycaemic control or medication requirement.Finally, for the woman, continuity of maternity care was lost at the time of transition to the GDM schedule, the appointment schedule was burdensome, and the clinic wait times frequently prolonged.
At the time of pre-implementation review, there were only small case series demonstrating the use of app-based care in GDM in the literature. 2 These had demonstrated non-inferior clinical outcomes, and similar strategies had been used in low resource settings. 35][6] In the context of multiple barriers to change and years of cumbersome 'work arounds', key stakeholder clinicians set out to re-imagine and redesign GDM care.
The objective of this study was to use an iterative approach to radically alter the GDM model of care at our facility and adapt to limited resources and increasing prevalence, without compromising clinical outcomes.The primary hypothesis was that a digital care package for GDM would produce equivalent maternal and neonatal clinical outcomes.

Setting and participants
Our quaternary maternity facility, located in metropolitan Brisbane, delivers approximately 12 000 babies per annum, 6500 babies in the described public sector.The GDM prevalence is currently 18%, amounting to 1200 women per annum.The GDM service is staffed by 2.0 full time equivalent (FTE) CDE midwives and 2.0 FTE dietitians.There are 0.

Procedure
The project consisted of three phases; pre-implementation (the period prior to any change in model), the interim (the period where a smartphone app -clinician portal was implemented but patients remained in pre-existing appointment schedules), and post-implementation.
The novel model of care included: All women were also booked for fetal growth and wellbeing scans between 28 and 30 weeks gestation and again between 34 and 36 weeks gestation.All other GDM appointments were removed and women remained in their previously assigned, often low-risk, maternity model of care (eg midwifery or general practitioner shared care) using established hospital protocols -adding virtual, specialist GDM supervision.The decision to commence pharmacological treatment for GDM (metformin and / or insulin) was made on an individual patient basis by the treating clinicians.
Women on insulin therapy had their results reviewed routinely every week, or more regularly if identified as significantly out of range.Women on metformin or diet therapy were reviewed weekly by the CDE.Guidance was provided directly as required and by an agreed protocol (Supporting Information) which included glycaemic targets and ultrasound fetal abdominal circumference centiles.
In parallel to steps 1 to 5, an iterative approach drawing on the framework of Kotter's eight steps of change and recurrent Plan-Do-Study-Act (PDSA) cycles occurred. 8Communication within the team was managed by the project lead, with regular local publication of patient data by a 'run chart'.The clinical lead also performed weekly safety audit of glycaemic management and rapid response to technical and workflow issues.A pragmatic decision to progress to full implementation, as opposed to a randomised or a cohort-led approach, was made due to the large number of patients, the very limited staffing, significant staff stress, COVID and the difficulties of synchronously running two models of care.

The intervention -development and content
The development and implementation of the novel GDM care package was generated from the best available evidence and clinical expertise from a multidisciplinary team, with a strong woman-centred focus.The process acknowledged and attempted to account for the complexities of our diverse patient cohort, insufficient staffing and scope of practice limitations, as well as a finite resource pool for program development.
The GDM education video was developed inhouse with multidisciplinary input from clinician stakeholders including dietitians, diabetes educator midwives, obstetric physician/endocrinologists, clinical educators and interpreter services.The script for the video described multiple core knowledge components related to GDM including the definition of GDM, basic pathophysiology, food and exercise requirements, use of a home blood glucose meter and blood glucose targets.The video was translated by the Special Broadcasting Service Australia (SBS) translation services into the five main languages for CALD women with GDM at our facility, namely, Cantonese, Mandarin, Somali, Vietnamese and Hindi.Culturally appropriate foods and actors were included in those versions, informed through close work with women from each cultural and/or language group.These videos are available free to view on You Tube, with links provided directly to women.

Outcome measures and data management
Outcomes included maternal (gestational age at birth, mode of delivery) and neonatal (birthweight, LGA (large for gestational age), neonatal intensive care unit (NICU) admission/length of stay and hypoglycaemia) clinical outcomes.Patient data were prospectively collected using the electronic maternity record.Following birth, data regarding maternal and neonatal clinical outcomes for delivered women with liveborn, term babies were extracted from the hospital maternity database.Cut-offs from New Australian birthweight centiles 9 were used to determine which neonates were LGA (>90th centile).Diabetes-related NICU admission was defined as admission due to hypoglycaemia, respiratory distress syndrome, transient tachypnoea of the newborn or cardiac dysfunction attributable to LGA.Data on home blood glucose levels (BGLs) were obtained from the CSIRO portal and classified according to frequency of readings within predefined BGL targets.

Data analysis
Maternal and neonatal characteristics and birth outcomes were described using number (percent) for categorical variables, mean (standard deviation (SD)) for normally distributed continuous variables, and median (interquartile range (IQR)) for non-normally distributed continuous variables.Evidence for a subgroup effect related to treatment received (diet / metformin / insulin) was examined using interaction terms between model of care and treatment type in linear and logistic regression models.An interaction term with P < 0.05 was considered indicative of a subgroup effect.

Multivariable robust linear regression was used to estimate
the adjusted association between model of care and birthweight.Statistical analysis was completed using Stata v15.1 and a P-value <0.05 was considered statistically significant throughout inferential analysis.

RESULTS
During the pre-and post-implementation periods, 941 term, liveborn delivery outcomes in 935 women were examined (six sets of twins).This number included 469 women on diet therapy alone, 142 treated with metformin and 324 treated with insulin ± metformin.

Pre-vs post-implementation, all treatment types combined
There was no evidence of a difference between pre-and postimplementation groups regarding treatment type, maternal age, maternal body mass index (BMI) at booking, gestational age or onset of labour (Table 1).There was weak evidence of a difference for mode of birth (P = 0.049), with fewer than expected forceps and vacuum deliveries post-implementation (Table 1).There was no evidence of a difference between groups in neonatal gender distribution, birthweight, proportion of neonates who were LGA, whether neonates were admitted to NICU (all or maternal diabetes-related admissions) or length of stay in NICU (Table 1).

By treatment type
Women treated with metformin and women treated with insulin were older and had a higher BMI at booking than those treated with diet alone (Table 2).Gestational age at delivery, onset of labour and mode of birth also differed by treatment type (Table 2).There were no differences by treatment type in baby gender, multiple gestation, birthweight, the proportion of neonates who were LGA, NICU admission (any cause or diabetes-related), or NICU length of stay (Table 2).The median number of BGL results entered by women treated with metformin and insulin was higher than for those treated with diet.The overall frequency of home BGL readings in the target range was similar across treatment groups (67.8%) (Table 2).

Pre-vs post-implementation, separately by treatment type (diet, metformin, insulin): birthweight
The unadjusted association between pre/post-implementation and birthweight differed by treatment type (P = 0.005).On average birthweight decreased by 100 g pre-to post-implementation  in the diet group, increased by 100 g pre-to post-implementation in the insulin group and was not statistically significantly different pre-to post-implementation in the metformin treated group (Table 3).When adjusted for gestational age, baby gender, maternal age and maternal BMI, birthweight tended to increase pre-to post-implementation in the insulin (45 g, 95% CI 0 to 90 g) and metformin (43 g, 95% CI −13 to 99 g) groups and tended to decrease in the diet group (−32 g, 95% CI −73 to 9 g) pre-to postimplementation.Pre/post-implementation differences were not statistically significant after adjustment.

Other characteristics
There was no evidence that the association between pre/postimplementation and each of LGA (P = 0.06), admission to NICU (for all P = 0.47; for maternal diabetes-related P = 0.14), length of stay in NICU (for all, P = 0.82; for maternal diabetes-related P = 0.32), or incidence of hypoglycaemia (P = 0.09), differed by treatment type (Table 3), meaning that any pre/post-implementation differences were similar for all treatment types.

DISCUSSION
The aim of this intervention was to radically alter the GDM model of care at our facility, adapting to limited resources and increasing prevalence, without compromising clinical outcomes.The data analysis pre-to post-implementation supports the hypothesis that the new digital care package for GDM provides clinically equivalent outcomes when compared with the previous traditional GDM model of care at our facility.
Gestational age at birth differed with differing treatment types, reflecting the protocol used at the time of the study.This recommended induction of labour in the insulin therapy group from 38 weeks gestation, the metformin only group from 39 weeks and the diet group from 40 weeks gestation.The gestational age considered for induction has since extended by seven days based on the 'every week matters' concept. 10e reason for fewer than expected forceps and vacuum deliveries post-implementation is unclear but may reflect more appropriate birth planning.A type 1 error (evidence for a difference when no real difference exists) is another possible explanation.
The neonatal outcome data (birthweight, LGA, nursery admission) are reassuring but a larger study would be required to detect a difference in uncommon outcomes such as maternal diabetesrelated NICU admission or neonatal hypoglycaemia.
Women who were older and had higher BMI at booking were more likely to require pharmacotherapy.This would be expected due to increased insulin resistance with both increasing age and BMI. 11 The strength of this novel model of care is in maintaining a woman-centric approach, focusing on language-specific care, in the woman's place, in a time convenient to her.Wide stakeholder consultation and co-design underlined the importance of multidisciplinary care and the need for sustainability across all parts of the service in the new model.While a randomised controlled trial design would have been optimal, with the pre/ post-design potentially introducing biases, significant staffing and resource limitations mandated a pragmatic approach in our facility.The consistency of maternal characteristics, even in the pandemic setting, is reassuring; however, persistent and

1 .
initial midwife phone call informing patient of the diagnosis of GDM at 24-28 weeks gestation and on the same day 2. delivery via email of the link to a GDM education video and registration for use of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) M♡THer app 7 3. a Bluetooth capable blood glucose meter was couriered to the patient's home address (at the hospital's expense) along with National Diabetes Support Scheme (NDSS) registration forms, with planned delivery within 48 h of initial phone contact 4. a face-to-face dual appointment with the CDE midwife and dietitian was scheduled for day seven and day 14 post-diagnosis (attendance rate was 95%), and 5. if required, a face-to-face insulin education appointment was arranged.
Univariable associations with treatment type (diet, metformin, insulin) and with model of care (pre (3 June 2019-30 June 2020) vs post (1 Nov 2020-2 June 2021) implementation) were examined using Pearson's χ 2 test or Fisher's exact test for categorical variables and analysis of variance (ANOVA)/two-sample t-test or Kruskal-Wallis ANOVA /Wilcoxon rank-sum test for normally and non-normally distributed continuous variables respectively.
Further investigation in the change in birthweight outcomes for women treated with insulin and those with diet is required.It is likely that the language and culturally specific videos have allowed for targeted and early dietary intervention and may account for the small decrease in birthweight seen in the diet group.Consumer feedback will be required to further elucidate this outcome.Of note, dietitians reviewing patients specifically counsel against restrictive eating practices.It is less apparent why the insulin-treated group's birthweights increased, as insulin titration was more intensive (weekly vs fortnightly) than in the traditional model.The unadjusted birthweight differences remain numerically small and are further reduced after adjustment for maternal factors.The frequency of home BGL readings in the target range was 67.8%.Further studies are required to determine the association between percentage in target range and clinical outcomes in GDM.
3 FTE obstetric physicians allocated for the GDM service.Most women are diagnosed with GDM between 24 and 28 weeks gestation.The mean maternal age of GDM women is 32.8 years and almost 60% of the women are overweight or obese.Approximately 30% of women are from cul-

TABLE 1 Maternal
and neonatal characteristics and birth outcomes pre-and post-implementation -for all treatment types Data are presented as mean (SD) or median (IQR) for continuous measures, and n (%) for categorical measures.†P-values obtained using Pearson's χ 2 test for categorical data and two-sample t-test for continuous data except where indicated.‡LGAnot calculated for n = 1 missing birthweight and n = 12 twins.§ Wilcoxon rank-sum test.¶ Includes hypoglycaemia, respiratory distress syndrome, transient tachypnoea of the newborn and cardiac dysfunction attributable to LGA.

TABLE 2 Maternal
, neonatal, birth and app use characteristics, overall and by treatment type P-values obtained using Pearson's χ 2 test for categorical data and analysis of variance (ANOVA) for continuous data except where indicated.‡LGAnot calculated for n = 1 missing birthweight and n = 12 twins.§ Kruskal-Wallis ANOVA.¶ Includes hypoglycaemia, respiratory distress syndrome, transient tachypnoea of the newborn and cardiac dysfunction attributable to LGA.

TABLE 2 (
Continued) a language-specific education video, streamlined schedule of visits and in partnership with the CSIRO, the M♡THer smartphone application.The maternal and neonatal clinical outcomes are reassuring and the potential benefit of improved visibility, timely and frequent review and a woman-centric approach are considered vital in re-imagining GDM care.

TABLE 3
Maternal and neonatal characteristics and birth outcomes pre-and post-implementation -separately by treatment type †LGA not calculated for n = 1 missing birthweight and n = 12 twins.‡ Includes hypoglycaemia, respiratory distress syndrome, transient tachypnoea of the newborn and cardiac dysfunction attributable to LGA.