Demonstrated Cost-Effectiveness of a Telehomecare Program for Gestational Diabetes Mellitus Management

Introduction

The prevalence of gestational diabetes mellitus (GDM) has increased steadily in recent years. In 2013, using the World Health Organization (WHO) criteria, the global prevalence of hyperglycemia in pregnancy was estimated to be 16.9%, resulting in more than 20 million live births. ¹ Pregnant women with GDM are at risk for obstetrical and neonatal complications. ² In 2008, the HAPO study established a continuous association between maternal glycemia and maternal and fetal complications with a demonstrated increased risk of macrosomia, caesarean section, and neonatal hypoglycemia. ³ The risks of pre-eclampsia, shoulder dystocia, prematurity, neonatal respiratory distress, and stillbirth were also significantly increased in mothers with GDM. ³

Fortunately, large randomized controlled trials and meta-analyses demonstrated that adequate treatment of women with GDM significantly reduces the rates of major complications such as fetal overgrowth, pre-eclampsia, and shoulder dystocia. ^{4 –8}

These patients therefore require close multidisciplinary follow-up to ensure the optimal management of GDM to reduce the risk of perinatal morbidity and mortality. This involves many appointments for patients and a significant use of hospital resources such as nurses, physicians, and nutritionists to allow the transfer of a high volume of new teachings. In this context, the use of hospital resources for education and management of GDM has increased markedly. Patients also encounter frequent delays before the first appointment in the clinic because of frequently overloaded clinic schedules and wasted time in the waiting room at each visit before seeing the doctor or other professionals. The related consequences and indirect costs of each visit for those women are work absenteeism and costs for transportation and/or babysitting of others' children. The health care system is facing an essential need to optimize organization of care to maintain access and cost-effectiveness.

Thus, the implementation of a telehomecare (THCa) system in this population could be an interesting option. THCa is defined as a remote mode of intervention using electronic data transmission for follow-up, education, and therapeutic adjustment. ^{9 –12} There is currently limited literature regarding the use of THCa in GDM patients, with most studies conducted in patients with type 1 and type 2 diabetes outside pregnancy. ¹³

Previous studies comparing THCa intervention to usual care in pregnant women with GDM reported similar glycemic control and maternal-fetal complications rates ^{14 –16} in both groups, but in the THCa group, empowerment in terms of GDM management ¹⁴ as well as satisfaction with care were improved. ¹⁶ Furthermore, Pérez-Ferre et al. demonstrated that THCa reduced unplanned medical visits by 62%, a decrease that was more important in insulin-treated patients (83%). ¹⁶

However, the results of a recent review and meta-analysis showed that current evidence remains insufficient for superiority of telemedicine in women with diabetes during pregnancy even if it confirmed that there was no evidence of harm regarding perinatal outcome. ¹⁷ Moreover, the important aspect of THCa cost-effectiveness remains largely unexplored. To the best of our knowledge, there are no data in the extant literature regarding the cost-effectiveness of THCa in GDM women in a Canadian context.

Hence, the objectives of this study were to evaluate in women with GDM the impact of THCa on clinical cost-effectiveness, pregnancy outcomes, and patient satisfaction with care services.

Subjects and Methods

The THCa initiative is a clinical Remote Patient Management project led by the endocrinology division at the Centre hospitalier de l'Université de Montréal (CHUM) assisted by a group of information technology technicians for the implementation of a platform that supports clinical processes.

A prospective stratified, noninferiority, and controlled clinical trial was designed in close collaboration with clinicians (physician, nurses, nutritionist, etc.). Women were not formally randomized, but rather assigned to either an intervention group provided with a THCa system for transmission and online analysis of blood glucose data or to a control group receiving usual care in the clinic. The study groups were formed sequentially: the THCa group was formed first by asking eligible newly diagnosed GDM women if they wanted to be followed with the THCa system instead of usual care after their first visit in the clinic. The control group was formed after the THCa group, in a period in which the clinic was not recruiting new women on the platform and was following newly diagnosed GDM women with only usual care. Women of both groups were enrolled between February 2016 and May 2017 (THCa group in the first 8 months and the control group in the following 7 months). All subjects were recruited at the obstetric clinic at the CHUM.

To be included, women were required to be older than 18 years and have a singleton pregnancy. A diagnosis of GDM was based on International Association of Diabetes and Pregnancy Study Groups/WHO criteria following a one-step diagnostic strategy with a 75-g oral glucose tolerance test (OGTT) if one of the glycemic results were ≥5.1 mmol/L (fasting), ≥10.0 mmol/L (at 1 h), and/or ≥8.5 mmol/L (at 2 h). To standardize the cohort, women with early or late GDM diagnosis were excluded. Only women with GDM diagnosed with an OGTT performed between the 21 0/7th and 30 0/7th weeks of pregnancy were included. Finally, patients had to be willing to use and able to learn the THCa software application to be included. We excluded women with type 1 or type 2 diabetes and twin pregnancy. Patients not able to use the technology or those who missed their first appointment more than once in the clinic after GDM diagnosis were also excluded.

The primary outcome was to evaluate the cost-effectiveness of THCa by the assessment of directs costs, including the number of medical/paramedical visits to the outpatient clinic as well as all interventions by the health care professionals done specifically for GDM management. Secondary outcomes were the impacts of THCa on (1) diabetes control, (2) GDM-related complications, and (3) patient satisfaction with care compared with usual care.

All patients gave informed written consent, and the study was approved by the CHUM Research Ethics Board Committee. Canada Health Infoway, Orion Health, Bell, and the Quebec Health care authorities provided funding and support for the study. No employee of those organizations was involved in the study conception, data collection, analysis, or writing of this article.

For each subject, capillary blood glucose was determined and self-monitored using Contour^® Next One (Ascensia) glucose meter provided for the study. All women were asked to record their fasting capillary blood glucose values as well as before and 1 h after meals alternately (e.g., minimum of 3 values a day, 1 day before meal, and the next day after meal). Maternal hypoglycemia was defined as a capillary blood glucose <3.8 mmol/L and hyperglycemia as a capillary blood glucose >5.0 mmol/L fasting, >5.3 mmol/L before lunch/dinner, >7.8 mmol/L 1 h after a meal, or >6.7 mmol/L 2 h after a meal. Insulin was introduced after 1 or 2 weeks of lifestyle counseling if more than two capillary blood glucose values per week were above target at the same time of the day without documented failure to follow dietary recommendations. Insulin therapy was monitored.

Furthermore, weight gain during pregnancy was documented. If the weight before pregnancy was not available, the first registered weight at the beginning of the pregnancy before the 12th week was used. Excessive gestational weight gain was computed based on clinic weight values, prepregnancy body mass index (BMI), and the 2009 Institute of Medicine guidelines. ¹⁸ In addition, gestational age at delivery, mode of delivery, and indications for caesarean delivery were registered and analyzed. Maternal complications were also collected and included hemorrhage, hypertension induced by pregnancy (recorded blood pressure ≥140/90 mmHg), pre-eclampsia, eclampsia, and HELLP syndrome.

The following neonatal issues were also fully documented to ensure the safety of THCa monitoring: birth weight (BW), preterm delivery (i.e., less than 37 weeks), birth defects, shoulder dystocia, neonatal respiratory distress syndrome, transfer to the neonatal intensive care unit, prenatal and neonatal deaths, jaundice/hyperbilirubinemia treated by phototherapy, and neonatal hypoglycemia. A blood glucose level of less than 2.5 mmol/L was used to define neonatal hypoglycemia. It should be noted that we reported all glucose levels less than 2.5 mmol/L, even if that event did not require any additional intervention on the newborn.

BW >90th percentile was used to define large for gestational age (LGA) and <10th percentile corresponds to small for gestational age (SGA) using the Canadian Perinatal Surveillance System BW chart. Macrosomia was analyzed using two definitions (BW ≥4 kg or LGA) as well as intrauterine growth restriction (low BW <2.5 kg or SGA).

Patients in the THCa group were given access to a web patient portal through their personal computer/tablet/phone upon participant's own preference. Each THCa patient was instructed to enter on the platform all the blood glucose measurements performed as outlined by their care plan (Fig. A1 in Supplementary Appendix A) and had to answer a set of five questions each week related to their general well-being and their feelings about GDM management. The patient's health status and blood glucose values were captured in the system and were accessible by clinicians at all times.

For the intervention group, coaching was delivered directly through the platform, repetitively and interactively when required, by the clinical team or automatically from a set of preprogrammed algorithms activated if the patient entered blood glucose values out of the normal range. Automated and adjusted feedback provided directly to the patient by those preprogrammed clinical algorithms promote better health behaviors and help those women to adjust adequately their diet, lifestyle, and/or, if applicable, their insulin therapy in case of hypoglycemia or hyperglycemia.

Alerts for hypoglycemia and hyperglycemia were divided into two levels: orange for results modestly out of range (e.g., between 2.6 and 3.7 mmol/L for hypoglycemia and between 7.9 and 9.0 mmol/L for hyperglycemia noted 1 h after meals) and red for more pronounced values of hypoglycemia or hyperglycemia (e.g., less than 2.6 or greater than 9.0 mmol/L after meals). Such clinical algorithms were programmed in the platform for hypoglycemia and for hyperglycemia noted fasting and 1 h or 2 h after meals, as well as randomly throughout the day. Each time a result classified as hypoglycemia or hyperglycemia was detected by the platform, a set of different questions and instructions for the patient regarding their symptoms, potential reasons for out-of-range values, diet, physical activity, and, if applicable, insulin treatment was matched for each case, asking afterward in most situations for a timely control of blood glucose.

At any time, the patient in the intervention group could access a health library to review specific teachings around GDM management (Fig. A2 in Supplementary Appendix A), view a weekly summary of their blood glucose values (Fig. A3 in Supplementary Appendix A), or contact the health team through the messaging system of the platform.

Clinicians received alerts based on those predetermined algorithms and could adjust therapy based on the symptoms and results documented by the patient. The remote care of patients was assigned to a small group of nurses working on the unit (Figs. A4 and A5 in Supplementary Appendix A). The glycemic control assessment was done daily on weekdays by the registered nurse in charge if there were any alerts (such as hypoglycemia or hyperglycemia), and there was a review of all data and charts with the medical team at least every 2 weeks for all patients.

Overall, the registered nurse was in charge to answer alerts, first view and analyze data entered by the participant, contact them through the platform or by phone if details were needed, present the report to the physician for evaluation, recommendations, and modifications of treatment if needed, and finally, transmit those medical conclusions and/or modifications in the plan of care to the participant. When required, patients came to the hospital to receive specific interventions by their clinical team (e.g., insulin use education with the registered nurse, management review for suboptimal GDM control, and fetal ultrasound for weight estimation).

Patients in the control group were followed according to usual care. Neither the THCa nor the control group had a predetermined number of visits in clinic; the frequency was based on clinical judgment as in a real-life clinical setting. Women in the THCa group were informed at inclusion that part of their follow-up for GDM would be through THCa, but they would need to come back to the outpatient clinic if we had any concern about their GDM control, for introduction of insulin therapy or review of different teachings, and so on. Both groups had the same obstetrical follow-up according to usual care, and no change was made during the study.

Patient satisfaction with care was measured in both groups using Likert scales adapted from a previous publication. ¹⁹ For each question, all participants had to evaluate, on a scale from 1 to 10, their capacity to achieve diabetes goals and to act adequately with the information received, their satisfaction with the services received and with the educational tools offered, as well as their overall diabetes-related psychosocial self-efficacy.

All interventions of both groups for GDM care, including time duration for each intervention (visit, e-mail writing, phone calls, GDM-related ultrasound, nonstress test (NST), meeting with the nurse or nutritionist, and so on), were recorded during the study to perform the cost-effectiveness analysis. All medical visits in the outpatient clinic with physicians for GDM, including ultrasound made specifically for GDM concerns as well as nurse, nutritionist, and social worker interventions, were registered. All unplanned visits at the emergency room or the obstetrical ward and inpatient hospitalizations were also registered.

In our institution, all paramedical interventions (nurse, nutritionist, and social worker) as well as administrative interventions (secretary and so on) have a code equivalent and a corresponding time duration filled accordingly by each member of the health care team, which allowed us to obtain precisely the time and the number of interventions per patient. In addition, the administrative costs for each visit, the salary scale of all employees, as well as the cost of each medical visit (sum of the costs of each medical act made by different physicians) were obtained, which enabled us to establish precisely each cost per patient for the management of their GDM.

The costs related to obstetrical matters (e.g., regular obstetrical follow-up, C Section, and extra visit for obstetrical concerns) were not factor since not directly related to GDM follow-up. However, NST, extra obstetrical follow-up added for GDM matters or ultrasounds done because the women had GDM were analyzed and priced. The costs related to the development, launch, and maintenance of the e-platform, as well as the indirect costs for patients (parking or other travel expenses, work absenteeism, babysitting of other children, and so on) were not evaluated.

Results

A total of 161 pregnant women with GDM were recruited, with 80 and 81 women included in THCa and control groups, respectively. The main clinical characteristics are shown in Table 1. Women were 32 years old on average, and the mean BMI, the rates of excessive gestational weight gain, and insulin therapy were similar. Level of education as well as global family income were also similar in both groups.

Overall, both groups had similar baseline characteristics, except for previous caesarean delivery, which was more prevalent in the control group. GDM control was also similar between the two groups in terms of the hypoglycemia rates (mean per group of 1.4 in the THCa group and 1.5 in the control group and median of 1 and 0 per subject, respectively) and the hyperglycemia rates (mean per group of 17.7 and 22.8 and median of 13 and 17 per subject, respectively). Also, no difference between the two groups was noted in the mean plasma glucose level (5.4 and 5.5 mmol/L, respectively) (see Table 1).

In terms of the primary outcome, patients in the THCa group had an average of 1.5 compared with 3.3 medical visits in the control group, representing a decrease of 56% (P < 0.001). However, an increase in nursing interventions by about 10-fold was observed (10.3 compared with 0.9 interventions, P < 0.001) in the THCa group (mainly by e-mails or phone calls; see Table 2).

The cost-effectiveness analysis demonstrated a significant cost saving of 167.75 CAN$ or 16.1% per patient followed by THCa. This saving for direct cost in global health care for the management of GDM was noted, despite the increased nursing time needed for follow-up (Table 3). Indeed, a specific increase of 113.40 CAN$ or 44.2% in nursing care costs has been noted for the THCa group, which is an expected result in the context of THCa implementation.

Maternal and fetal outcomes were similar between the groups (see Table 4). Furthermore, there was no case of pre-eclampsia in either group. With respect to neonatal outcomes, it is important to note that complications typically associated with GDM, such as macrosomia (1.3% in the THCa group and 2.5% in the control group, P = 1.000), neonatal hypoglycemia (20% and 16%, respectively, P = 0.545), and shoulder dystocia (2.5% and 0%, respectively, P = 0.245) were similar in both groups. Only one newborn in the control group required a dextrose infusion for neonatal hypoglycemia and none in the THCa group.

Global satisfaction with care was similar in both groups (8.9/10 in the THCa group and 8.5/10 in the control group, P = 0.128), but satisfaction with educational support was significantly increased in the THCa group (9.0/10 for the THCa group and 8.5/10 in controls, P = 0.028).

Discussion

Results of this study show a significant decrease in medical visits (56%) and in total health care costs (16%) in women with GDM when follow-up is ensured by THCa, without compromising pregnancy outcomes, quality of care, safety, or patient satisfaction.

Indeed, THCa monitoring allows strict glycemic control similar to usual care used in our study. Moreover, maternal and fetal outcomes, including all rates of obstetrical complications evaluated in this study, were similar between both groups, demonstrating that, by using THCa monitoring, there is no loss in the quality of care and THCa is as safe as usual care. The total caesarean rate tended to be increased in the control group without reaching clinical significance, but this probably reflected the fact that a higher rate of previous caesarean section was noted in the control group at baseline since the caesarean rate in labor (not planned) was similar between both groups.

These results are consistent with most of the current literature, which reported a 38% to 83% decrease in medical visits for THCa patients with GDM without increasing maternal or perinatal complications. ^2,16,22 However, some data of other studies remain conflicting with regard to the reduction in medical visits with THCa, a result that could be partly explained by the heterogeneity of the THCa system and protocols used in the literature. ²³ The literature also reports comparable glycemic control and rates of maternal-fetal complications in patients followed by THCa, which is similar to the findings in this study. ^{14 –16,24}

An innovative aspect of this project is that THCa was integrated within traditional care, as the responsibility for the follow-up of the cohort of patients on THCa was part of the tasks of the nurses working directly at the clinic, allowing for a real-life experience in adding THCa to current services. Another innovative aspect of the study is the delivery of coaching directly through technology by using preprogrammed algorithms that are based on patients' blood glucose results. We believe that this innovation favors self-motivation and promotes a better health condition.

One of the major limitations of the study is the absence of formal randomization and blinding. Blinding was not possible, and since the providers of the THCa system imposed maximum use of the platform in the first months after it launched, instead of formal randomization with its potential to slow down the use of the platform, we decided to move forward and build our cohort sequentially in an attempt to limit recruitment bias.

We first asked all eligible newly diagnosed GDM women in our center to enter the THCa group and in this way formed our intervention group. The control group was formed after the intervention group by approaching women during the period in which it was not possible to recruit new women into the THCa system (i.e., during this time, all women newly diagnosed with GDM were followed in the clinic by usual care). This way, we ensured that the control group was not limited to only those women who were not interested in the THCa system.

In addition, for the cost-effectiveness evaluation focused on GDM care, we chose to record time and price specifically the factors we considered to be directly related to GDM diagnosis and management. This evaluation accounted only the direct costs for our hospital but did not account indirect costs such as parking, babysitting, work absenteeism, and so forth. Also, the pricing process was specific to our health care team and institution. However, even if the pricing process cannot be generalizable and this cost-effectiveness evaluation was performed during implementation of a new system and not after full deployment of the solution, both groups were evaluated the same way and in the same period, allowing the proportion of the cost saving to be reliable and probably more conservative.

Global evaluation of direct and indirect costs after a complete running phase of a new platform with a randomized controlled trial would certainly provide more information on the total cost-effectiveness of the THCa platform in patients with GDM.

To our knowledge, this is the first study to analyze closely, in a Canadian setting, the cost-effectiveness of THCa follow-up compared with usual monitoring in GDM women. Also, the reduction of 56% in medical visits could even be considered conservative since women were included in both groups at their first visit at the outpatient clinic where, as protocolized in our center, they only meet with the nurse and the nutritionist to receive general lessons on GDM and to start register their blood glucose. Indeed, women newly diagnosed with GDM will meet first with the doctor 1 or 2 weeks after this first visit with the nurse and the nutritionist.

So, since all visits after the inclusion visit in our study were counted in for the economic analysis, all women in the intervention and control groups had a necessary first visit with the medical team. However, since one of our main goals was to try to help the clinicians predict, in a real-life setting, the impact of THCa implementation for its organization and for the patient, we decided to count the first medical visit in and price it since the patient had to come to the hospital for her GDM management. We felt it was closer to reality to keep this visit in our economic analysis. Therefore, we have demonstrated that THCa monitoring is cost-effective, despite increased nursing time, which will be an additional argument for health care centers wishing to implement such a system.

For our cohort, if we extrapolated that 80% of our 400 new cases of GDM per year were being followed by THCa, a conservative cost saving of 53,680 CAN$ can be estimated per year in our center.

Few studies have focused on the cost-effectiveness analysis of THCa. A recent systematic review targeted all studies that examined the economic analysis of telemedicine in diabetes. None of the studies included in this review directly concerned GDM, but it was found that use of telemonitoring and telephone reminders was cost-effective in diabetes management. ²⁵ However, it is important to note that initial deployment of THCa involves an additional burden in nursing time and related cost. Our experience suggests that for a successful deployment of this technology, a dedicated nurse must be present at the clinic to follow the cohort.

One of the strengths of this study was the large number of patients included in each arm. Another key aspect of our study was the analysis of patient satisfaction, which demonstrated that THCa is convenient, well accepted, and easy for patients to use. This suggests that THCa can play a role in improving quality of life and facilitating better acceptance of the GDM condition during pregnancy, a result similarly reported by Dalfrà et al. ²⁶

Conclusion

Among women with GDM, THCa monitoring, compared with usual care, significantly decreased medical visits and direct costs for GDM management in our institution without compromising pregnancy outcomes, quality of care, or patient satisfaction. This study demonstrated that an additional burden in nursing time should be considered when implementing a THCa program, although this measure is still cost-effective.

The THCa system seems to represent an interesting new avenue for delivering proper care for women with GDM and minimizing related health costs. Further analyses by formal multicentric randomized and controlled trials are needed to better characterize the impact of a THCa system on direct and indirect costs in women with GDM.