The Australian New South Wales COVID-19 pandemic response, beginning in March 2020, included strict lockdowns and restrictions on gatherings and movement throughout the state, resulting in changes to access for ambulatory care services, including a pivot toward telehealth. Eligible participants were nonpregnant adults aged 18 years and older with a diagnosis of diabetes mellitus who had attended the Diabetes Centre at Royal Prince Alfred Hospital (RPAH) in Sydney, Australia, in the immediate prepandemic period between January 2019 and March 2020. They were invited to participate in an NLTA as a health surveillance strategy with the specific aim of monitoring well-being over 12 months. The TSG intervention comprised 3-monthly telehealth contacts for 6 months, then again at 12 months post-enrollment. Contact was made via phone or video with a diabetes nurse educator. Baseline data included participant demographics, self-reported medication and medical history, and collection of comprehensive blood and urine tests, including hemoglobin A_1c (HbA_1c%), by standard pathology laboratory protocols. At each subsequent assessment, demographic data, self-reported anthropometric measurements, and medication regimen were gathered; results of recent (ie, within 3 months) blood and urine tests were collected and discussed. A main aim in the nurse consultations was to help avoid deterioration in glycemia through maintenance of self-care and optimizing adherence to prescribed lifestyle and medication regimens. Bespoke health and well-being questionnaires, as well as validated psychological screening tools, were administered, including the EuroQol Visual Analogue Scale (EQ-VAS), Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder 7-item (GAD-7) scale, and Problem Areas in Diabetes (PAID-5) scale [19-22]. Following best clinical practice and as recommended by the institutional Human Research Ethics Committee, if psychological inventories revealed concern for the presence of major depression including self-harm, the participant’s general practitioner (GP) and community health team were to be contacted. The participant was encouraged to remain in regular contact with their health care team, in addition to telehealth surveillance visits.

Periods for analysis were defined as follows: prepandemic (pre-enrollment) data were captured between January 2019 and February 2020 (T0). Enrollment data (T1) were obtained during the enrollment period beginning from March 2020 onward. Follow-up data were collected at approximately 3 months (T2), 6 months (T3), and 12 months post-enrollment (T4). Final follow-up visits were completed in January 2022.

A retrospective, non-intervention comparison cohort (non-TSG) was established using data collected in the RPAH Diabetes Centre electronic medical record. Criteria for inclusion in the comparison cohort required the participant to have received 4 or more clinical services delivered between 2019 and 2021, aligned with the TSG participants. Assessment time points were matched to those of the intervention group.

To assess the unadjusted effect of independent variables across metabolic outcomes between the intervention and comparison groups, we used nonparametric analysis of variance for continuous variables and chi-square (χ²) tests for categorical variables, along with descriptive statistics over time. Fisher exact tests were applied when expected cell frequencies were less than 5. Due to the non-normal distribution of the outcome variables, Wilcoxon rank sum tests were conducted at each time point (T0, T1, T2, T3, and T4) to compare metabolic indices.

A stepwise regression approach was used to identify potential covariates, excluding interaction terms and independent variables that lacked association, introduced model skewness, or were deemed clinically irrelevant. To assess differences in effects across time points, specifically between the prepandemic period (T0) and subsequent assessments, linear regression models were applied, adjusting for key factors such as time, age, sex, diabetes type and duration, ethnicity, perceived personal risk of COVID-19 infection, and food stockpiling. For the regression analyses, the very few individuals with maturity-onset diabetes of the young, pancreatogenic diabetes, or atypical diabetes were classified as either T1DM if on an insulin regimen or T2DM if on an oral hypoglycemic regimen.

For the psychological outcome measures (PHQ-9, GAD-7, and PAID-5), regression models were restricted to the TSG group, as these measures did not apply to the non-TSG group. Results were reported as odds ratios with 95% CIs and corresponding P values. Statistical significance was defined as P<.05. All analyses were conducted using SAS 9.4 (SAS Institute Inc) software.

All TSG participants consented to de-identified data being used for research purposes. This study was approved by the Human Research Ethics Committee (HREC/X20/RPAH/0206) of the Sydney Local Health District. All participant data were handled with strict confidentiality following ethical guidelines, and personal information was de-identified for analysis. Participation was voluntary, with the opportunity to withdraw at any time, and no financial compensation was provided.

One hundred TSG participants were enrolled, with 91 completing 4 tele-surveillance visits over a 12-month follow-up, comprising the final TSG analysis set (Table 1). Participants were predominantly male (57%, 52/91), Caucasian (63%, 57/91), and had a mean age of 57 years. Overall, 48% (44/91) and 51% (46/91) of the participants had a diagnosis of T1DM and T2DM, respectively. One percent (1/91) were post-pancreatectomy. The median diabetes duration and age of diabetes onset were 14.0 years and 37.1 years, respectively, with 12% (11/91) of participants aged <30 years. Demographic data for the comparison cohort (n=115) are also shown in Table 1. The comparison cohort was also predominantly male, with a similar percentage of T1DM and T2DM, but was slightly older, had a longer diabetes duration, and was more likely to be non-Caucasian.

Over the pandemic period, different trajectories of glycemia were observed in the TSG and non-TSG cohorts; more favorable blood glucose levels were maintained in the TSG cohort throughout the study period (Figure 1). We observed an absolute reduction in HbA_1c% from pre-enrollment (T0) to 12 months of 0.3% in the TSG and 0.1% in the non-TSG; the percentage reduction in HbA_1c% over 12 months was significantly greater in the TSG compared to the non-TSG (4% vs 1%, respectively; P=.04). It is notable, however, that the prepandemic HbA_1c% for the non-TSG group was higher, with a mean of 8.2% (SD 1.2%) compared to a mean of 7.8% (SD 1.5%) in the TSG group (P=.04; Table 2). After adjusting for prepandemic HbA_1c% for each cohort separately across all the time points, we observed a significant overall improvement in HbA_1c% for the TSG, particularly at 3 months (T2; P=.02) and 6 months (T3; P=.047), with significance sustained across all time points (P=.048). In contrast, there was no significant change in HbA_1c% across any of the time points for the non-TSG (P=.61).

Between-group differences were significant between the TSG and non-TSG at each time point (Table 2 and Figure 1). Notably, the final regression model showed that TSG participants were 40% less likely (adjusted odds ratio 0.6, 95% CI 0.5‐0.7; P<.001) to experience an unfavorable rise in HbA_1c% compared to the non-TSG participants, adjusting for sex, age, prepandemic HbA_1c%, ethnicity, diabetes type, and duration (Table 3).

For the data points available, we observed stability in BMI of ~29.4 kg/m² over the 12 months for both the TSG and non-TSG. Mean blood pressure at enrollment compared to the final visit was similar for the TSG and non-TSG. Lipid levels showed similar stability over the pandemic period for both groups. Albuminuria status was progressively more unfavorable in the non-TSG; however, there was more non-TSG data missing for this variable, including cardiac, renal, and liver disease status (Table 2 and Multimedia Appendix 1).

In the TSG alone, we observed a decline in participants reporting having seen their GP or endocrinologist over the observation period (Figure 2). Specifically, we observed a significant decline in diabetes service attendance over time (P<.001; Multimedia Appendix 2). This was true for both face-to-face and telehealth visits over the 12 months. In contrast, we observed an increase in email contact with the Diabetes Centre over 12 months. Visits to allied health care practitioners remained low throughout the study period. At enrollment (T1), 13% (12/91) of TSG participants had been to the hospital in the prior 3 months, 5.5% (5/91) had visited the emergency department, and 5.5% (5/91) had been admitted as inpatients. We did not observe any changes in the number of hospital visits, including emergency department, hospital admissions, or day procedures, over the observation period. Eighteen percent (16/91) of participants reported that they had specifically avoided or delayed a diabetes-related health care service, and this remained steady throughout the study period (Multimedia Appendix 2).

At enrollment (T1), the median EQ-VAS to assess participants’ health on a scale from 0 to 100 was relatively high at 80. Nevertheless, at T1, 20% (18/91) of participants reported scores indicative of moderate to moderately severe depression (PHQ-9), while 3.3% (3/91) reported scores consistent with severe depression. Twelve percent (11/91) reported moderate to severe anxiety (GAD-7), and 21% (19/91) reported scores suggesting significant diabetes-related emotional distress at T1. All mental health indices generally remained consistent over the 12 months. An exception was the observation of anxiety scores worsening over the 12-month study only in those with an enrollment GAD-7 score ≥10 (P<.001; Multimedia Appendix 3).

In the TSG, we observed significant differences over the study period in pandemic-related beliefs and behaviors, including perceived personal risk of clinically contracting COVID-19 and stockpiling additional food (Multimedia Appendix 4). These were adjusted for in the final mental health regression models using backward stepwise elimination (Table 4).

In the final model for PHQ-9 outcomes, adjusting for time, age, sex, type of diabetes, and significant covariates from the univariate analyses (Table 4), females were 9 times (P<.001) and those younger than 30 years of age were 42 times (P<.001) more likely, respectively, to experience an unfavorable rise in depression score. Similarly, in the GAD-7 outcome model, females and those younger than 30 years of age were 10.3 times (P<.001) and 93 times (P<.001) more likely, respectively, to experience an unfavorable rise in anxiety score. Finally, in the PAID-5 model, females and those younger than 30 years of age were 4.1 times (P<.001) and 12.8 times (P<.001) more likely, respectively, to experience an unfavorable rise in diabetes-related distress score. An adverse increase in PHQ-9 and GAD-7 scores was significantly greater for those with T2DM.

Although the COVID-19 pandemic resulted in overall reduced access to usual health services [23], our study observed a more favorable HbA_1c% trajectory over 12 months in the NLTA group, which differed from the comparison, usual care group. This was achieved in the context of fewer medical physician encounters. The NLTA was effective in monitoring the psychological health of people with diabetes and allowed the identification of patients who were psychologically vulnerable. We observed that the most vulnerable demographic in this cohort comprised women, young people, and those with high baseline GAD-7 scores, with potential implications for future pandemic planning. Perhaps more significantly, this study highlights the feasibility and positive potential for a nurse-led telehealth surveillance model to support traditional doctor-led models of care in diabetes management. Our results suggest that a nurse-led telemonitoring service is also acceptable to patients with diabetes, as evidenced by high participation rates at 12 months.

Our observation that the NLTA achieved stability in glycemic control during the challenging early stages of the COVID-19 pandemic is similar to other studies, including those in Australia, that involved telehealth use [24-28]. These findings are also in contrast to data highlighting deterioration of glycemic control in diabetes cohorts during the COVID-19 pandemic [29,30]. In addition, the intervention cohort achieved glycemic stability despite a significant decline in contact with their usual endocrinologist and usual diabetes educator via face-to-face or telehealth consultations. Contact with primary care providers within the TSG remained consistently unchanged over this period. One might speculate that nurse-led telemonitoring facilitated less frequent endocrinology and diabetes educator contact in the context of relatively stable glycemic control in the TSG, reserving targeted contact for those at higher risk and with complications [31-33]. We also observed stability of cardiovascular risk factors over the 12 months in the nurse-led telemonitoring group, again in contrast to other global reports of deterioration in control and despite reduced contact with auxiliary services [34,35]. Similarly, there was no increase in unplanned hospital attendance throughout the intervention period. Regular nurse-led telehealth surveillance was associated with stability in the overall rates of medication-taking and glucose self-monitoring throughout the pandemic period.

While we did not aim to, nor did we undertake, formal cost-effectiveness comparisons of health care delivery models, based on the NLTA model and reduction in GP and medical specialist consultations, we speculate that the NLTA model would not be more financially costly than the traditional non-TSG model comparator and yet delivered arguably better metabolic outcomes for the study cohort. The potential cost-effectiveness of the NLTA is supported by prior studies. A review paper examining the cost-effectiveness of telehealth for diabetes management found that, despite using different methodologies in varied health care settings, the majority of studies found telehealth to be cost-effective, citing direct costs per patient. The authors concluded that telehealth solutions have vast potential for cost-effective diabetes care [36]. Separately, there is a small body of literature to support the cost-effectiveness of nurse-led services in diabetes. In a study from the Netherlands, diabetes nurse specialists provided care comparable to physicians, with similar health-related quality-of-life measures and favorable costs per QALY gained [37]. Few contemporary studies exist on the cost-effectiveness of the combined approach of nurse-led telemedicine in diabetes. However, a modeling study on patients with heart failure found that nurse-led telephone support improved survival and was cost-effective in comparison with usual care [38]. A 2016 Belgian study reported that a nurse-led telecoaching program for T2DM, compared to usual GP care, was highly cost-effective with a reported incremental cost-effectiveness ratio well below suggested thresholds for the country [39]. Taken together, in the context of our findings, these data support the likely cost benefits of the NLTA but underscore the need for additional cost-effectiveness analysis of this model of care within a robust study design.

Through regular contact and the administration of validated psychological inventories, we were able to capture and track changes in COVID-related beliefs and deterioration in psychological health over the study period. We observed high levels of vaccine acceptance, low COVID infection rates, and postulate that provision of education through regular and reassuring health check-ins may have contributed to pandemic-safe, appropriate behaviors. In addition, the usage of validated standardized screening instruments as a component of the surveillance model enabled us to identify individuals with high levels of anxiety, depression, and diabetes distress—particularly prevalent among women and those aged <30 years. Our findings align with the increasing pandemic-related distress observed in the general UK population, which appears to be driven primarily by changes in younger people and women [40,41]. The Progression of Diabetic Complications (PREDICT) cohort study and other studies also observed a similar increased risk of adverse psychological health in younger people with diabetes during the same period [42,43]. The reasons behind the increased psychological vulnerability for these groups are not clear, but it would be reasonable to speculate that high baseline mental health issues in youth and adverse social determinants of health may play a role, as observed in previous studies [44]. Regardless of etiology, this model of care may help to identify at-risk young people with diabetes who may benefit from targeted, evidence-based intervention, such as an enhanced SMS text messaging support program [45]. The alignment of our findings with larger, robust studies suggests that a nurse-led telehealth surveillance model of care was able to accurately identify those most vulnerable and at risk. Beyond psychological health, physical activity and sedentary behavior showed modest changes, and self-care behaviors remained unchanged.

Optimal chronic management of diabetes involves monitoring and ongoing support to facilitate self-management and often behavior change, with a significant time and health resource burden. The COVID-19 pandemic led to innovations in patient care and widespread application of health care delivery via telemonitoring means. A global study found that 80% of providers from many countries adapted to new methods of telemedicine via telephonic or video technology [46]. Nurses in particular have embraced such technological advances, often with higher levels of acceptance than primary care providers [46]. A Canadian national survey found a ninefold increase in nurse involvement in telemonitoring care from prepandemic levels [47]. In the context of the rapid expansion of technologies to deliver health care, there is potential for a new nurse-led diabetes surveillance model of care to be scalable, accessible, and economically beneficial in the postpandemic era.

The limitations of this study include its observational and nonrandomized design, with the inherent potential bias. It is important to recognize that baseline differences in age, HbA_1c%, and diabetes duration between the TSG and non-TSG cohorts could have influenced our results, despite appropriate statistical adjustments, potentially rendering the glycemic benefits observed less significant. We also note that data beyond metabolic indices were not available for the comparator group. In addition, this was a single-center study of predominantly White, English-speaking participants within the context of the Australian Medicare-funded health care system. Thus, our results may not be entirely generalizable to the wider patient population or other health care systems. However, our results support the need for further study in a larger cohort using a robust, randomized, and controlled framework within the prevailing health infrastructure. Furthermore, successful implementation of the NLTA would depend on adequate technological infrastructure, funding, and reimbursement for nurse-led encounters, as well as workforce training.

Our data suggest that an NLTA in diabetes care is feasible, acceptable, and could have utility in the monitoring of metabolic indices and psychological well-being in people living with diabetes, allowing for appropriate triage and intensification of management—especially in the context of limited access to primary or specialist care. With this in mind, we believe further studies of a diabetes nurse-led telemonitoring model of care are warranted.