Scheduled Remote Review of Continuous Glucose Monitoring Data with Targeted Intervention versus Standard Care in Children with Type 1 Diabetes: A Parallel-Groups Study Authors: Frances Ellen Gehrmann, Lauren Haddow, Peter William James Delaney, Grant James Smith, Craig E. Taplin, Timothy W. Jones, Elizabeth A. Davis Published: 2026 ABSTRACT AbstractIntroductionThis study evaluated the impact of scheduled review of continuous glucose monitoring (CGM) data with targeted remote intervention on glycaemic and person-reported outcomes in children with type 1 diabetes (T1D).MethodsChildren aged 5–14 years with T1D using CGM were recruited from two metropolitan clinics at Perth Children’s Hospital in Western Australia; one clinic group was assigned to intervention and the other to control. The intervention group underwent CGM review by a diabetes educator every 14 days. Participants with >10% decrease in time in range (TIR: 3.9–10.0 mmol/L) from the previous review period, or who were not meeting consensus CGM targets, were prioritised and offered personalised remote advice via phone or email; the number receiving contact was constrained based on allocated educator time in each 14-day period. Those improving or meeting targets received positive feedback by SMS. Change in TIR from baseline to 6 months was compared between groups using difference-in-difference analysis. Person-reported outcomes were assessed using Type 1 Diabetes and Life (T1DAL) and Diabetes Treatment Satisfaction (DTSQ) questionnaires.ResultsA total of 113 children were recruited, 60 in the intervention group (53% male, age 10.2 [2.7] years, baseline TIR 66.9% [13.4], HbA1c 7.2% [1.3]) and 53 in the control group (51% male, age 9.6 [2.6] years, baseline TIR 69.4% [13.4], HbA1c 6.9% [0.8]). Most participants used automated insulin delivery systems; 64% in the control and 72% in the intervention. Across 14-day periods, mean (95% CI) 71% (67, 75) of participants in the intervention group were identified for targeted support; 39% (30, 49) received remote advice. Positive feedback was sent to 38% (32, 43) in each 14-day period. Compared with the control group, the intervention group showed a non-significant 3.1 percentage point increase in TIR (β = 3.1, p = 0.11). Over 6-months, mean TIR decreased in the control group (69.4%–66.6%) and was unchanged in the intervention group (66.9%–67.0%). No significant between-group differences were observed in T1DAL or DTSQ scores, limited by low end-of-intervention questionnaire response rates.ConclusionRegular CGM review identified many participants not meeting CGM targets, and over half received targeted intervention. However, the intervention did not significantly improve TIR or person-reported outcomes. While it did not bring participants into target range in this well-controlled cohort, the model may have greater effect in children with lower baseline TIR. Future trials should evaluate its impact in higher-risk youth. Introduction A key goal of the management of Type 1 diabetes is to prevent complications by maintaining blood glucose levels as close to normal as possible [1]. Until recently a target HbA1c of <7.0% was recommended to minimise long term complications of diabetes [2], however most recent international guidelines recommend a target HbA1c of ≤6.5% in those with access to advanced diabetes technologies [3]. With the advent of continuous glucose monitoring (CGM), glycaemic targets have expanded to include CGM metrics, with percentage of sensor glucose time in target range (TIR [3.9–10.0 mmol/L]), time below range (TBR [<3.9 mmol/L and <3.0 mmol/L]) and time above range (TAR [>10.0 mmol/L and >13.9 mmol/L]), along with mean sensor glucose and coefficient of variation (CV) used as measures of glycaemic control [4]. Population based registries have shown that most young people with T1D do not achieve these targets [5–7]. Within Australia, population-based registries report that only 18–27% of youth with T1D achieve a HbA1c <7.0% [8, 9]. The model of care recommendations for young people living with diabetes has changed little over the last 4 decades. In-person clinical visits four times a year remains standard care [1]. Whilst models of care remain unchanged, diabetes technology has advanced significantly over the past decade with continuous glucose monitoring (CGM), insulin pumps and integrated advanced hybrid closed-loop insulin delivery (AHCL). Clinical and real-world trials of CGM use in youth demonstrate reduction in HbA1c and increase in TIR with CGM use, however, with CGM alone most youth still do not meet glycaemic targets [10–12]. AHCL systems have further improved glycaemic outcomes but even with these systems many do not achieve optimal glycaemic goals [13]. Real-world glycaemic outcomes of Western Australian youth show AHCL systems deliver an average 8% increase in TIR, but only 39% of youth meet TIR targets of >70% [14] evidencing that technology use improves outcomes but on its own does not deliver optimal glycaemic outcomes. This suggests that models of care to maximise outcomes need to be developed and tested. In Western Australia, CGM is fully subsidised and is used in over 95% of young people with T1D. The use of CGM data to detect changes in glycaemic outcomes provides an opportunity to adapt our model of care to deliver more personalised and targeted care for young people with diabetes. We therefore designed a study to test a novel model of care aiming to investigate if frequent review of CGM data by health care providers to detect adverse changes in glycaemic control that is followed by targeted intervention to those requiring additional support can improve glycaemic and patient reported outcomes compared to those receiving the current traditional care model. Methods Study Design This was a quasi-experimental, parallel-groups study comparing outcomes between an intervention clinic and a comparison clinic before and after implementation of the intervention using a difference-in-differences analysis. Children with T1D were recruited from two metropolitan clinics with comparable mean age and HbA1c at Perth Children’s Hospital, a tertiary paediatric diabetes centre in Western Australia, between November 2023 and May 2024. Children recruited from one clinic were allocated to the control group and the other clinic allocated to the intervention group, with intervention delivered by a diabetes educator already working within that clinic. Each clinic was led by different clinicians, but had an identical model of care, procedures and access to multi-disciplinary team specialties. This method of allocation was chosen to mimic a real-life clinical model, where children and families receive care from and develop relationships with a consistent team. A rolling recruitment was conducted, meaning that some participants had started the intervention prior to final participants being recruited. All children willing to use CGM and data-share with the clinic were invited to participate; non-English speaking families were excluded. All children attending the clinic had access to CGM funding through the NDSS and those not using CGM were doing so by choice. All eligible children were invited to participate with information sent via email and families invited in person when they attended their routine clinic visits. Intervention The intervention model of care was compared with a control group receiving conventional outpatient care. All participants had access to conventional care which consisted of 3-monthly multi-disciplinary clinic visits and 24-h phone support with a Diabetes Nurse Educator (DNE)-run triage line during the day and an emergency medical (Endocrinology trainee) line after hours. The intervention consisted of remote review of CGM data by a diabetes educator every 14-days. Glycaemic metrics were ranked based on ascending TIR, international CGM metrics targets and change in TIR as pre-defined major and minor flags. Flags of concern were determined using international consensus CGM metrics targets [3] and baseline analysis of a sample of the hospital cohort’s CGM data. Criteria were further separated into major flags which identified approximately 50% of the local population of youth on CGM and combined major and minor flags identifying approximately 70%. Major flags were defined as TIR <50%, >10% reduction in TIR over 2 weeks, >4% TBR and <75% sensor wear. TIR between 51 and 69% was assigned a minor flag. Positive flags included meeting all CGM metric targets or metrics demonstrated significant improvement in metrics from the last 14-day. The participants’ number of concerning flags and then ascending TIR were used to prioritise the 24 individuals each 14-day that were offered remote clinical review; this was capped to allow time for meaningful clinical consultation within the diabetes nurse’s allocated time of two 4-h slots per 14-day. Participants were offered the option of clinical review with a text message via WhatsApp (Meta Inc.), a secure messaging platform that allows confirmation of messages being received and read. If participants opted for clinical review, they chose to receive either a phone call with the research diabetes nurse educator (DNE) or an email with clinical support based on CGM and pump data review. Participants who met positive flags received a text message via WhatsApp with positive feedback regarding their CGM report. The intervention period ran for 26 weeks. To best reflect clinical practice, the research DNE was recruited from the intervention group’s routine clinical team to utilize the clinical relationship that develops in real-world clinical care. Data Collection Demographic and clinical data were collected from the Western Australia Children’s Diabetes Database (WACDD). Variables collected included age, sex, diabetes duration, postcode, insulin delivery regimen and HbA1c. Postcode data were used to classify the participants’ area of residence and calculate relative socio-economic disadvantage using the Australian Bureau of Statistics Socio-Economic Indexes for Areas (SEIFA) Index of Relative Socioeconomic Disadvantage (IRSD), with quintiles 1 and 2 considered socioeconomic disadvantage [15]. CGM data were retrieved from device-specific platforms (CareLink™, Glooko® and LibreView) with CGM reports reviewed online by the study team every 14-day to identify major, minor and positive flags and CGM data exported at the end of study period for analysis of outcomes. Clinical service delivery data were collected on a 14-day basis including number of participants enrolled, number meeting major and minor concerning flags and positive flags, number offered contact, number opting for contact and number receiving positive feedback. Clinician time taken reviewing the data of and contacting those that were offered intervention was recorded each 14-day cycle. Study data and digital consent were collected and managed using REDCap, electronic data capture tools [16, 17]. Outcomes The primary outcome was the change in TIR from baseline to end-of-intervention. Secondary glycaemic outcomes included change in HbA1c and glycaemic metrics across the study period, including TBR (<3.9 mmol/L and <3.0 mmol/L), TAR (>10.0 mmol/L and >13.9 mmol/L) and mean sensor glucose. Person-reported outcomes collected included quality of life (QoL), assessed using the Type 1 Diabetes and Life (T1DAL) Questionnaire for children and parents, and treatment satisfaction was assessed using the Diabetes Treatment Satisfaction Questionnaire (DTSQs) for parents and teens. Questionnaires were completed at baseline and 6 months later in both the intervention and control groups. The DTSQ change (DTSQc) was administered at the end of intervention in parents and teens; it assesses change in satisfaction with a treatment regimen over time, with a more positive score indicating increased satisfaction with treatment regimen and a more negative score reflecting decreased satisfaction. Children ≥8 years are validated to complete age specific versions of the TIDAL and teens ≥12 years can complete the DTSQs and DTSQc. Statistics Mean (SD) is presented for continuous measures and frequencies (%) for categorical measures. A difference-in-differences (DID) analysis under a linear mixed model framework was used to determine the association between the intervention and changes in the primary outcome over time. The model included a random intercept for participant, effects for group (control, intervention), phase (start, end) and an arm by phase interaction (the parameter of interest) and key demographic/clinical characteristics (regimen, age, diabetes duration, sex). Additional sensitivity analyses were also conducted: standard ANCOVA, a linear mixed effects beta regression, a non-covariate adjusted DID model and a robust regression using iteratively reweighted least squares to reweight influential cases. Intervention effects for secondary glycaemic outcomes were estimated using regression models of change scores adjusting for key sociodemographic characteristics as were person reported outcomes (change in T1DAL and DTSQs scores). Where residual outliers were noted, sensitivity analyses using robust regression modelling were also conducted. Further sensitivity tests of secondary outcomes were also conducted using standard ANCOVA and generalized linear mixed models with distributional family appropriate for the data (e.g., negative binomial, beta, Gaussian). As an exploratory, hypothesis-generating exercise, interaction models were conducted to explore differences in intervention effects by three sociodemographic factors: age group (<12 years, ≥ 12 years), baseline TIR (<50%, 50–69%, ≥70%), and area of socioeconomic disadvantage (lower IRSD = quintiles 1–2, higher IRSD = quintiles 3–5). For all tests, a p value of <0.05 was considered statistically significant. The number of participants who met CGM metric flags in each 14-day period is presented with mean (SD) percentage of participants, and clinician time is presented as mean (SD) minutes per 14 days and per participant. No correction for multiple testing was applied. Ethics Approval This study was performed in accordance with the Declaration of Helsinki and approved by the Child and Adolescent Health Service Human Research Ethics Committee in Western Australia (RGS0000005749). It is registered with the Australian New Zealand Clinical Trials Registry (ACTRN12623000955617) https://www.anzctr.org.au/. Recruitment took place from November 2023 to May 2024. All parents, guardians or next of kin provided written informed consent for their child to participate in this study. Results Over the recruitment period, 113 of a potential 268 children (42%) joined the study, 60 (55% male) in the intervention group and 53 (51% male) in the control group. Online supplementary Figure 1 (for all online suppl. material, see https://doi.org/10.1159/hrp/adaag028) provides a CONSORT-like diagram of the recruitment process. Table 1 provides baseline demographic and clinical characteristics. Two participants (one control and one intervention) withdrew from the study after commencement due to CGM connectivity difficulties (intervention participant) and a desire not to complete further surveys (control participant). At baseline, the intervention group mean (SD) age was 10.2 (2.7) years with the diabetes duration of 4.5 [3] years, TIR 64.8 (13.9)%, and HbA1c 7.2 (1.3)%. Insulin regimen included 72% using hybrid closed loop, 20% using multiple daily injections (MDI), and 8% using other modes of pump therapy. The control group had a mean age of 9.6 (2.6) years at baseline, with 3.2 (2.5) years diabetes duration, TIR 69.4 (13.4)%, and HbA1c 6.9 (0.8)%. Hybrid closed loop systems were used by 64%, with 28% MDI and 8% other pump modes. All CGM devices and insulin delivery systems approved for use in Australia were represented within the study cohort, with type of device chosen by the child and family based on their individual needs. Table 1.Baseline demographic and clinical characteristics of control and intervention groups​ControlIntervention p value N 5360​Male sex, n (%)27 (51)33 (55)0.81aAge, yearsb9.6 (2.6)10.2 (2.7)0.12cDiabetes duration, yearsb3.3 (2.5)4.4 (3)0.06cAge category at baseline, n (%) <8 years15 (28.3)10 (16.7)0.33a 8–12 years25 (47.2)32 (53.3) ≥12 years13 (24.5)18 (30)Socioeconomic profile Low SES (IRSD 1 and 2)5 (9)12 (20)0.19d Non low SES (IRSD 3–5)48 (91)48 (80)Insulin delivery, n (%) HCL34 (64)43 (72)0.58d Other pump4 (8)5 (8) MDI15 (28)12 (20)TIR, %b69.4 (13.4)66.9 (13.4)0.35eBaseline TIR, n (%) <50%6 (11.3)7 (11.7)0.82d 50–69%16 (30.2)20 (33.3) ≥70%31 (58.5)29 (48.3) N/A (<70% CGM use)0 (0)4 (6.7)HbA1c, %b6.9 (0.8)7.2 (1.3)0.23eQuintile 1 represents greatest socioeconomic disadvantage; 5 represents least socioeconomic disadvantage; IRSD, Index of Relative Socioeconomic Disadvantage; HCL, hybrid closed loop; other pump, pump not using advanced insulin delivery technology: either manual mode or predictive low glucose suspend; MDI, multiple daily injections; TIR, time in range (3.9–10 mmol/L). a Chi-squared test for categorical variables with large count frequencies. b Values reported as mean (SD). c Wilcoxon test for non-normally distributed continuous variable. d Fisher’s exact test for categorical variables with small count frequencies. e Unpaired t test for normally distributed continuous variable. Baseline demographic and clinical characteristics of control and intervention groups Quintile 1 represents greatest socioeconomic disadvantage; 5 represents least socioeconomic disadvantage; IRSD, Index of Relative Socioeconomic Disadvantage; HCL, hybrid closed loop; other pump, pump not using advanced insulin delivery technology: either manual mode or predictive low glucose suspend; MDI, multiple daily injections; TIR, time in range (3.9–10 mmol/L). Chi-squared test for categorical variables with large count frequencies. Values reported as mean (SD). Wilcoxon test for non-normally distributed continuous variable. Fisher’s exact test for categorical variables with small count frequencies. Unpaired t test for normally distributed continuous variable. In the adjusted linear mixed model, there was no significant difference in the change in TIR between intervention and control groups. As shown in Table 2, the control group went from 69.4 (14.7)% to 66.6 (14.3)% (−2.8% change, p = 0.02), and the intervention group went from 66.9 (13.4)% to 67 (10.8)% (0.11% change, p = 0.31); the difference in difference was not statistically significant (β = 2.8; 95% CI: −5.8, 3.4; p = 0.13). Table 2.CGM metrics and HbA1c at baseline and end of study with difference in change adjusted for treatment arm and demographic and clinical factors% Time, mmol/LControl, n = 53Intervention, n = 60Difference in changebaselineend of studychange p valuebaselineend of studychange p valueΒ coefficient (95% CI) p value3.9–10.069.4 (14.7)66.6 (14.3)−2.8 (8.3) 0.02** 66.9 (13.4)67 (10.8)0.11 (10.6)0.942.8 (−5.8, 3.4)0.133.9–7.848.8 (14.8)46.5 (13.8)−3.0 (9.6) 0.03** 43.9 (13.3)43.3 (11.5)−0.5 (10.1)0.692.4 (−6.4, 2.6)0.21<3.0a0.4 (0.6)0.3 (0.8)0 (0.6)0.530.3 (0.6)0.4 (0.7)0.1 (0.5)0.38−0.05 (−0.3, 0.4)0.73% time <3.9a2.1 (3.3)2.3 (3.9)0.1 (2.8)0.581.8 (2.5)1.8 (3)0.1 (1.7)0.440.1 (−1.0, 1.2)0.75% time >10.0a23.4 (18.8)27.3 (20.3)3.2 (9.5) 0.01** 27.7 (17.2)29.2 (16.4)0 (10.2)0.58−3.0 (−3.8, 6.0)0.11% time >13.9a5.9 (9.6)7.7 (12.9)1.2 (5.2) 0.02** 7.5 (9.2)6.8 (7.4)0.7 (5.3)0.94−2.4 (−6.4, 2.6)0.22Mean SGL8.5 (1.5)8.8 (1.4)0.2 (0.8) 0.03** 8.9 (1.4)8.8 (1.1)−0.1 (1.1)0.58−0.3 (−0.4, 0.6)0.10SD3.2 (0.8)3.4 (0.9)0.1 (0.5) 0.04** 3.3 (0.8)3.3 (0.7)−0.1 (0.6)0.48−0.2 (−0.2, 0.4)0.06CGM metric data reported as mean (SD) with paired t test where normally distributed.Β coefficient (95% CI) = percentage points and 95% CI difference-in-difference change in CGM metrics after adjustment for treatment arm, age, diabetes duration, sex, and treatment regimen.**p value <0.05 considered statistically significant. a Median (IQR) reported with paired Wilcoxon-rank test for non-normally distributed outcomes. CGM metrics and HbA1c at baseline and end of study with difference in change adjusted for treatment arm and demographic and clinical factors CGM metric data reported as mean (SD) with paired t test where normally distributed. Β coefficient (95% CI) = percentage points and 95% CI difference-in-difference change in CGM metrics after adjustment for treatment arm, age, diabetes duration, sex, and treatment regimen. **p value <0.05 considered statistically significant. Median (IQR) reported with paired Wilcoxon-rank test for non-normally distributed outcomes. There were no statistically significant differences between groups in the change for any secondary glycaemic outcomes including TBR (<3.0 mmol/L [p = 0.73], <3.9 mmol/L [p = 0.75]), TAR (>10.0 mmol/L [p = 0.11], >13.9 mmol/L [p = 0.22]), time in tight range (3.9–7.8 mmol/L [p = 0.21]) mean sensor glucose (p = 0.10), and sensor glucose standard deviation (p = 0.06), with Table 2 documenting the glycaemic outcomes at baseline and end of intervention for each treatment arm with difference-in-change coefficients and 95% CI calculated for each metric. In all sensitivity models, the effects remained similar. The within-group change showed a statistically significant deterioration in glycaemic metrics of TIR (p = 0.02), time in tight range (p = 0.03), TAR (>10.0 mmol/L [p = 0.01] and >13.9 mmol/L [p = 0.02]), mean sensor glucose (p = 0.03), and sensor glucose standard deviation (p = 0.04) in the control group, whilst there was no statistically significant change in any glycaemic metrics for the intervention group (Table 2). Figure 1 shows the estimated marginal means (EMMs) of change in TIR for control and intervention groups, stratified by baseline TIR (<50%, 50–69% and ≥70%), age at study commencement (<12 years, ≥12 years), and area of socioeconomic disadvantage (lower IRSD, higher IRSD). Results did not show a significant interaction effect between arm and baseline TIR (p = 0.11), age at commencement (p = 0.43), or lower IRSD (p = 0.71). EMMs of change in TIR across the study period in those starting with a baseline TIR <50% were 7.1 (95% CI: −1.1, 15.3)% in the control group and 16 (95% CI: 8.1, 23.9)% in the intervention, compared to −2.2 (95% CI: −7.1, 2.7)% in the control and −1.9 (95% CI: −7.1, 3.4)% in the intervention group for those starting with a baseline TIR ≥70%. In children aged ≥12 years, EMM of change in %TIR was −1.25 (95% CI: −8.8, 6.3)% in the control group versus 4.1 (95% CI: −2.7, 10.9)% in the intervention group, compared to children aged <12 years with −2.2 (95% CI: −7.3, 2.9)% change in TIR in the control group and −0.2 (95% CI: −5.9, 5.5)% in the intervention group. For children living in areas of greater socioeconomic disadvantage (lower IRSD), EMM of change in TIR was −1.92 (95% CI: −12.8, 9.0)% in the control group and 3.0 (95% CI: −4.8, 10.8)% in the intervention group, compared to −2.2 (95% CI: −7.3, 2.9)% in the control group and 0.4 (95% CI: −5.2, 6.1)% in the intervention group for those living in higher IRSD areas. EMMs of change in TIR from baseline stratified by baseline TIR, age at study commencement, and area of relative socioeconomic disadvantage. EMMs (±95% confidence intervals) of change in percentage TIR from baseline by treatment arm, stratified by baseline TIR (<50%, 50–69%, ≥70%) (a), age at study commencement (<12 years, ≥12 years) (b), and relative socioeconomic disadvantage (lower IRSD, higher IRSD) (c) as calculated by postcode and index of relative socioeconomic disadvantage (IRSD). Points represent model-based EMMs; error bars indicate 95% confidence intervals; number above the plots represents n in each group. Figure 2 demonstrates the percentage of the cohort meeting flag criteria for intervention or positive feedback every 14 days and the percentage of the intervention group who received contact. Each 14-day period, 68 (6)% (mean [SD]) of participants in the intervention group and 67 (6)% in the control group met at least 1 major or minor concerning flag based on their CGM report. In the intervention group, 38 (4)% met a major flag of concern every 14 days and 41 (4)% of the control group met major flags of concern. When the top 24 concerning flags were prioritized each 14-day period, 61 (15)% of all participants who flagged were offered clinical review. Of those offered, 59 (14)% received nurse review each 14-day period, with 67% being via phonecall and 33% email. A mean of 28 (3.6) minutes was spent on each contact with an average of 7.6 (4.0) minutes spent per total participant enrolled. The majority of those who did not receive contact did not reply to the initial message, with only 12 participants replying to decline contact across the study period. In the intervention, 37 (4)% met criteria for and received positive feedback via WhatsApp each 14-day period, and 34 (5)% of the control group met positive criteria. Percentage of the control or intervention group flagging for intervention or positive feedback and the percentage of the intervention group contacted each 14-day period. Box and whisker plot of % of participants from the control and intervention groups meeting flag criteria for intervention or positive feedback each fortnight (noting the control group did not receive contact) and percentage of participants in the intervention group contacted each fortnight. Plots with median (horizontal line), mean (dot), and IQR (box). Flagged = met at least one criteria for contact by the diabetes educator. Flagged positive = met at least one criteria for positive feedback. Contacted = received contact as part of the intervention group. I, intervention group; C, control group. The percentage of participants in the intervention group who met each different flag of concern criterion is depicted in Figure 3a. The minor flag of TIR <70% was most frequent, with 43 (8)% (mean [SD]) flagging each 14-day period. Within the total cohort, 21.6 (4.5)% flagged for TBR >4% and 12 (3.2)% flagged for TIR <50%. Less frequent flags included 8.2 (4.8)% for a >10% drop in TIR and 4.8 (2.2)% for percentage sensor wear <75%. In those participants who were prioritized for contact, mean(SD) number of flags of concern was 1.4 (0.6) with 76 (8)% meeting major flags of concern each review period, 49 (10)% meeting minor flags of concern, and 24 (7)% meeting minor flags only. Mean percentage of participants who were offered contact meeting each flag of concern is depicted in Figure 3b. Percentage of participants in the intervention group meeting each flag criteria each 14-day period. Box and whisker plot of percentage of participants from the intervention group flagging for individual flag criteria each fortnight. a Percentage of the total intervention cohort meeting flag criteria. b Percentage of the cohort prioritized and offered contact. Plots with median (horizontal line), mean (dot), and IQR (box). TIR, time in range; TBR, time below range; SW, sensor wear. Collection of person-reported outcomes at baseline and end of intervention was incomplete, with 102 (90%) parents and 71 (63%) children completing baseline questionnaires versus 56 (50%) parents and 47 (42%) children completing questionnaires after 6 months. The number of children completing DTSQ was limited at both baseline and 6 months due to it being designed for children ≥12 years. Table 3 presents T1DAL and DTSQs questionnaire scores and completion rates for parents and children/teens at baseline and end of intervention. There was no significant difference in change of quality-of-life scores (T1DAL) between groups for parents (B = −0.28, 95% CI: −5.1, 4.5, p = 0.91) or children (B = 1.3, 95% CI: −6.0, 8.6, p = 0.72) when adjusted for clinical and demographic factors. Similarly, there was no significant difference in change in treatment satisfaction (DTSQs) between groups for parents (B = −2.1, 95% CI: −5.7, 1.5, p = 0.25) or teens (B = 0.1, 95% CI: −12.4, 12.5, p = 0.98). The DTSQc was completed at the end of intervention by 32 parents (mean [SD] score: 11.3 [8.75]) and 13 teens (mean [SD] score: 10.5 [7.5]) in the intervention group and 23 parents (mean [SD] score: 10.0 [9.6]) and 6 teens (mean [SD] score: 16.2 [14.3]) in the control group. Scores were positive, indicating increased treatment satisfaction in both the control and intervention groups after 6 months with no significant difference between groups (parent: p = 0.58, teen: p = 0.39). Table 3.Change in quality of life and treatment satisfaction scores from baseline to end of study​Control (total n = 53)Mean (SD)Intervention (total n = 60)Mean (SD) p value n n T1DALParent (aggregated <8, child and teen scores) Baseline4762.8 (15.5)5563.3 (13.3)0.91 End of study2261.6 (11.8)3465 (11.1) Change22−0.6 (9.2)342.1 (8.8) Difference in change−0.28 (−5.1, 4.5)Child (aggregated child and teen scores) Baseline2972.9 (13.4)4268.3 (15.4)0.72 End of study1772.7 (18.5)3075.2 (13.6) Change171.7 (9.1)304.4 (11.9) Difference in change1.3 (−6.0, 8.6)DTSQsParent Baseline4742 (7.6)5442.7 (7.6)0.25 End of study2544.1 (6.9)3644.7 (7.1) Change252.0 (7.8)361.4 (6.1) Difference in change−2.1 (−5.7, 1.5)Teen Baseline835 (8.5)1735.9 (6.7)0.98 End of study633 (7)1235.8 (7.5) Change62.2 (8.0)120.7 (6.4) Difference in change0.1 (−12.4, 12.5)Survey scores reported as mean (SD).Difference in change reported as B coefficient (95% CI) using linear regression model adjusting for age, diabetes duration, sex, and insulin regimen.T1DAL, Type 1 Diabetes and Life Questionnaire (measures diabetes related quality of life); DTSQs, Diabetes Treatment Satisfaction Questionnaire. Change in quality of life and treatment satisfaction scores from baseline to end of study Survey scores reported as mean (SD). Difference in change reported as B coefficient (95% CI) using linear regression model adjusting for age, diabetes duration, sex, and insulin regimen. T1DAL, Type 1 Diabetes and Life Questionnaire (measures diabetes related quality of life); DTSQs, Diabetes Treatment Satisfaction Questionnaire. Discussion This study in children with T1D tested whether frequent remote CGM monitoring and review followed by targeted intervention resulted in improved glycaemic outcomes. Although the study identified that a high proportion of participants (67–71%) did not meet glycaemic targets, the intervention was not associated with significant improvements. Although the primary difference-in-differences analysis did not provide evidence that the TIR change differed between the intervention and control groups, descriptively, the control group showed a small statistically significant decline in TIR and the intervention group showed no evidence for significant change. Whilst these within-group patterns may warrant further investigation in larger studies in the future, they should be interpreted as descriptive and hypothesis generating rather than evidence of an intervention effect. When stratified by baseline TIR, age, and IRSD, there was no evidence for a differential intervention effect based on patient characteristics; however, numbers in the groups were small and the study was not powered to detect these differences. Visual inspection of the effect sizes and confidence intervals suggests future adequately powered research should explore whether these types of interventions differ in their effectiveness for those with a baseline TIR <50% and those aged 12 years and older. No significant change in QoL or treatment satisfaction was identified in questionnaires. There are limited studies using remote monitoring of glucose levels between clinics in youth with T1D. A randomized control trial by Ganrud et al. [18] of 117 youth aged 8–17 with mean HbA1c 8.9% (0.67) compared weekly review of pump uploads (using self-measured or CGM glycaemic data) with standard outpatient care. Whilst results were non-significant, a trend of lower HbA1c at 6 months was seen in their intervention group (−0.35%) compared to the control (−0.05%). The magnitude of glycaemic change was greater compared to our study; however, older mean age and higher HbA1c at baseline make direct comparison difficult. Teenagers (aged 13–17 years) demonstrated greater glycaemic gains compared to younger age groups; this supports the need to further explore effects of intervention in a teenage age group. Crossen et al. performed a feasibility study in youth aged 1–20 (mean 9 years) comparing standard care (mean Glycaemic Management Index [GMI] 7.5%) to monthly CGM review (mean GMI 7.6%) and contact by a clinician if CGM metrics demonstrated excessive hyperglycaemia (>15% time above 13.9 mmol/L) or hypoglycaemia (>5% of the time below 3.9 mmol/L) [19]. This model showed a treatment effect (−0.25%) on GMI, with greater reduction seen in those with a higher baseline GMI [19], also suggesting potential for greater gain in those with poorer baseline glycaemia. A series of studies conducted by a single institution in the newly diagnosed T1D population, the Teamwork, Targets and Tight Control (4T) Study utilized an automated dashboard that extracted data from Dexcom Clarity to remotely monitor and prioritize those requiring intervention [20–22]. Alerts for intervention were comparable to our study; data were reviewed on a weekly basis; and participants were prioritized based on number of alerts and ascending TIR and contact was provided to 22 top-priority participants each week by electronic message through an electronic health record. Whilst there was no significant difference in HbA1c at 12 months between those receiving remote monitoring compared to standard care in the 4T cohort, TIR was 8.8% higher in those receiving weekly remote monitoring [20–22]. It may be speculated that newly diagnosed families may be more receptive to advice and behaviour change before individual diabetes management behaviours have established. A strength of our study was the inclusion of children using any available CGM technology despite this requiring review across multiple platforms. It should be noted that the study DNE was involved in the clinical care of participants in the intervention group, allowing for the real-world clinical relationship that develops between a child/family to be incorporated into this trial model of care. A limitation of this study was that participants had relatively optimized glycaemic outcomes at baseline, although most still did not meet consensus targets. Given mean TIR was approaching targets at baseline (66% intervention and 69% control), there was a smaller margin for improvement following intervention, making detection of significant effect on glycaemic outcomes less likely. Only 11% of participants started with TIR <50% compared to 32% of eligible children who did not enrol in the study, limiting the power to detect significant change in this group despite visual appearance of effect in this subgroup. There was a self-selection bias with the cohort that consented to participate in the trial, mostly already very motivated with a low mean HbA1c (6.9% control and 7.2% intervention). This makes results of this study different to generalize across the wider population of children with T1D and future work will need to address barriers to adopting this model for those with poorer glycaemic outcomes. Another limitation is the lack of randomization, with control and interventions groups allocated to two existing clinic groups, with matched models of service delivery but run by different clinicians. This makes it impossible to separate intervention effects from unique clinic features. The incomplete questionnaire completion at end of intervention is another limitation. Small numbers of completed questionnaires compromise power to detect a significant difference in treatment satisfaction or quality of life between the intervention and control groups and there may be differences in the populations who completed or did not complete questionnaires making any effects potentially subject to bias. Ongoing optimization of target population, triggers for contact and modes of communication will be important to create a successful model of care. Previous similar studies have shown that children with lower TIR or HbA1c and older, teenage participants made greater glycaemic gains with remote CGM review and intervention [18, 19] and subgroup analysis of this cohort showed a non-significant suggestion of greater gains in these populations. Thus, it is plausible that targeting children within our population with lower TIR and the adolescent subgroup could result in more significant improvement in glycaemic outcomes. Further exploration of this model should include adolescents, where contact could be targeted at the young person with diabetes rather than their parent/guardian, as well as prioritizing individuals with lower TIR who stand to make greater gains with this form of intervention. Furthermore, when considering implementation of a new model of care, feasibility is dependent on resource availability and allocation. Data have been collected to complete a full economic evaluation of this model of care. In this trial, CGM reports were manually reviewed from different device platforms by members of the study team and flags manually allocated to participants. This is a labour-intensive process and would not be feasible across a large clinical cohort. To offer this model of care to all, optimization of processes and contact methods is needed. Adoption of an automated algorithm to identify and prioritize patients who met criteria for contact would be necessary. In this study, an average of 7.6 min per participant per 14 days was spent on clinical review and contact, which may not be feasible across a large cohort. Reduction in time spent per individual could be achieved by using predominantly messaging services to streamline contact; this has been used successfully in similar models [20, 21], but such an intervention requires further stakeholder engagement and co-design as it is likely a range of preferred methods of receiving information across patients and families exists. Further refinement of clinical flags of concern for intervention could also be considered to prioritize clinician time and target those at greatest risk. Conclusion In a group of children with generally optimized glycaemia and high technology use, frequent review of CGM with targeted remote intervention over 6 months identified a large proportion of participants for remote intervention. In its current form, in this population, it is unlikely that the intervention will result in clinical improvements. However, targeted delivery of the intervention to those with suboptimal glycaemic outcomes, or adolescent groups, warrants further exploration. Automating key processes such as CGM review and prioritization/ranking processes to streamline patient identification and contact may improve clinical feasibility and wider adoption. Acknowledgments The authors are grateful to the children and young people with diabetes and their families who participated and provided the data. Statement of Ethics This study was performed in accordance with the Declaration of Helsinki and the study protocol was reviewed and approved by the Child and Adolescent Health Service Human Research Ethics Committee in Western Australia (RGS0000005749). All parents, guardians, or next of kin provided written informed consent for their child to participate in this study. Conflict of Interest Statement Timothy W. Jones was a member of the journal’s Editorial Board at the time of submission. Funding Sources This work is/has been supported by the Western Australian Future Health Research and Innovation Fund, which is an initiative of the WA State Government. The study team was supported by Breakthrough T1D Australia (grant No. 5-SRA-2021-1088-M-X). Frances Ellen Gehrmann acknowledges the Western Australian Child and Adolescent Health Service, The Kids Research Institute, and University of Western Australia combined clinician researcher PhD scholarship to support her work in this project. Author Contributions F.E.G., G.J.S., C.E.T., T.W.J., and E.A.D. contributed to the conception and design of the study and results interpretation. P.W.J.D., L.H., and F.E.G. contributed to data collection and cleaning. F.E.G. and G.J.S. contributed to statistical analysis. F.E.G. wrote the first draft of the manuscript. All authors edited, reviewed, and approved the final version of the manuscript. E.A.D. acts as guarantor of this work and, as such, had full access to all the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis.