Use of Technology in Older Adults with Type 1 Diabetes: Clinical Characteristics and Glycemic Metrics

Abstract

Background: The use of diabetes-related technology, both for insulin administration and glucose monitoring, has shown benefits in older adults with type 1 diabetes (T1D). However, the characteristics of older adults with T1D and their use of technology in real-world situations are not well documented.

Methods: Older adults (age ≥65 years) with T1D, using insulin pump or multiple daily injections (MDI) for insulin administration, and continuous glucose monitoring (CGM) or glucometer (blood glucose monitoring [BGM]) for glucose monitoring were evaluated. Participants wore CGM for 2 weeks, completed surveys, and underwent laboratory evaluation.

Results: We evaluated 165 older adults with T1D; mean age 70 ± 10 years, diabetes duration 40 ± 17 years, and A1C 7.4% ± 0.9% (57 ± 10 mmol/mol). For insulin administration, 63 (38%) were using MDI, while 102 (62%) were using pump. Compared to MDI, pump users were less likely to have cognitive dysfunction (49% vs. 65%, P = 0.04) and had lower scores on the hypoglycemia fear survey (P = 0.03). For glucose monitoring, 95 (58%) used CGM, while 70 (42%) used BGM. Compared to BGM, CGM users were more likely to report impaired awareness of hypoglycemia (IAH) (P = 0.01), and had lower A1C (P = 0.02). Participants who used any technology (pump or CGM) had lower A1C (P = 0.04, 0.006), less hypoglycemia ≤54 mg/dL (P = 0.0006, <0.0001) and <70 mg/dL (P = 0.0002, 0.0001), and fewer glycemic excursions (coefficient of variation %) (P = 0.0001, <0.0001), while reporting more IAH (P = 0.04, P = 0.006) and diabetes distress (P = 0.02, 0.004).

Conclusion: Older adults with T1D who use newer diabetes-related technology had better glycemic control, lower hypoglycemia risk, and fewer glycemic excursions. However, they were more likely to report IAH and diabetes-related distress.

Clinical trials.gov NCT03078491.

Introduction

With the aging of the population and successful management of Type 1 Diabetes (T1D), the number of older adults with T1D has been increasing over the past few decades. In the recent years, there has been unprecedented expansion in diabetes-related technologies, both in the areas of insulin administration and glucose monitoring. Within the past 30 years, people with diabetes have seen the development and implementation of glucometers (blood glucose monitoring [BGM]) and, more recently, continuous glucose monitoring (CGM) in their daily routine.^1–3 Insulin administration devices, such as insulin pens and, more recently, insulin pumps, became widely available in the 1990s and have grown in prevalence ever since. The use of diabetes-related technologies (CGM and pump) in several studies have shown improvement in glycemic control, as well as quality of life-related outcomes, in adults with T1D.^4–8

The majority of studies evaluating diabetes-related technology do not include the older population as their focus. Subpopulation analyses and small retrospective cohort studies in recent years have shown the benefits of using CGM in older adults with T1D.^9–11 One study prospectively focusing on the use of CGM in older adults with diabetes has shown a decrease in the risk of hypoglycemia and improvement in glycemic control.¹² Most studies evaluating the benefits of pump in older adults were small, retrospective chart reviews.^13,14 In addition, it is not clear if the results regarding the use of technology, derived from controlled study environments, are also seen in community-living patients who are not part of any studies.

Aging, in general, brings a set of unique challenges in persons with diabetes, including the presence of medical comorbidities and frequently progressive cognitive and functional decline.^15,16 As people with T1D age, their higher risk of additional comorbidities may impair their ability to keep up with daily diabetes self-management behaviors.^17,18 In addition, older individuals may find learning new information and new technologies harder and more stressful than younger adults.¹⁹ If older adults with T1D are not able to use the diabetes-related technologies appropriately, errors can contribute to grave damage in the context of hypoglycemia, disease-related distress, and in quality of life. Thus, although diabetes-related technologies can help people with diabetes improve their glycemic control, it is important to identify the characteristics of older individuals who will be able to benefit from the use of these technologies.

In this study, we evaluated the characteristics of older adults (≥65 years) with T1D seen in a tertiary care diabetes clinic, who were using various diabetes-related technologies for insulin administration (multiple daily injections [MDI] vs. insulin pump) and glucose monitoring (BGM vs. CGM). We assessed the association of the use of these technologies on glycemic control and risk of hypoglycemia.

Methods

In this cross-sectional analysis, we examined baseline data from two ongoing studies that included older adults with T1D: “Technological Advances in Glucose Management in Older Adults,” a study assessing the use of CGM in older adults with T1D (Clinical trials.gov), and “Assessing the Impact of Aging on Adults with T1D and T2D.” Data were collected between April 2017 and March 2021. Eligibility criteria for this analysis included a diagnosis of T1D, age 65 years or older, and stable use of diabetes technology for glucose monitoring (BGM or CGM) and insulin administration (pump or MDI) for at least 6 months before assessment. If using BGM, participants needed to be willing to wear masked CGM for at least 14 days. No participants in either study were using any hybrid closed loop systems or smart Bluetooth insulin pen technology. All participants provided written informed consent. The study protocols were approved by the Institutional Review Board at the Joslin Diabetes Center.

At enrollment in each study, demographic, medical information, and clinical data on diabetes management were collected.

CGM parameters

A masked CGM (Dexcom^® G4) was worn for 14 consecutive days (2 sensor sessions; participants changed sensor at home) by the participants who were CGM-naive. Participants who were already using their own personal CGM (Dexcom G5 or G6) consented to device download by the study staff. CGM data were downloaded and the data from the 2-week period before enrollment were analyzed from all participants using personal CGM; a minimum of 192 h of CGM data were required for inclusion in the study analyses. Coefficient of variation (CV%) was calculated as standard deviation of glucose divided by mean glucose level × 100. CGM data from both CGM-naive participants and participants using personal CGM were analyzed to determine duration of hypoglycemia and hyperglycemia by the following metrics in min/day: time spent ≤54 mg/dL, time spent <70 mg/dL, time in range 70–180 mg/dL, and time spent >180 mg/dL.²⁰

Clinical assessments

Cognitive function (by Montreal Cognitive Assessment [MoCA])²¹ and depression (by Geriatric Depression Scale [GDS]), medication use, or reported diagnosis)²² were assessed, and laboratory A1C test was performed. Survey measures for impaired awareness of hypoglycemia (IAH) (by Clarkes' method),²³ hypoglycemia fear (by HFS-98),²⁴ diabetes distress (by Diabetes Distress Scale—Type 1 [DDS-Type 1]),²⁵ and overall perceived well-being (Short Form-36 [SF-36])²⁶ were also performed.

Clinical parameters

Cognitive dysfunction was defined as MoCA score <26. Depression was defined as follows: GDS score >5, the use of antidepressant medications, or a reported diagnosis of depression on demographic and medical history surveys. Hypoglycemia unawareness was defined as ≥4 responses as reduced awareness on the Clarke survey. The HFS-98 was scored by dichotomizing all responses into <1 and >1, adding responses to questions 16–33 to calculate the Worry subscore, adding responses to questions 1–15 to calculate the Behavior subscore, and adding both subscores together to calculate the Total Score. The DDS was scored by taking the average of all seven subscale scores as well as the Total Score. The SF-36 was scored by calculating Physical Component Scores (PCS) and Mental Component Scores (MCS). The PCS and MCS are summary based t-scores derived from taking the average of all physical components (general health, physical function, physical role limitations, and pain), and the average of all mental components (emotional role limitations, energy fatigue, emotional well-being, and social functioning).²⁷

Statistical analysis

Descriptive statistics for demographic and clinical data are reported as percentages for categorical variables. For continuous variables, data were reported as mean ± standard deviation for data with a normal distribution and median and first and third quartile “median (Q1, Q3)” for data with nonnormal distribution. SAS 9.4 software version was used for all analyses and included Pearson's correlations, Student's t-tests, Fisher's exact tests, and logistic regression analysis.

The averages for the PCS and MCS scores were used to calculate individual z-scores, which were then transformed to t-scores using the national average mean (50) and standard deviation (10) to calculate individual scores.²⁷

The cohort was divided into multiple different groups for analysis. After data cleaning, including calculating proper survey response scores described above, preliminary analysis included running Pearson's correlation tests for all data by all categories of technology usage (MDI, pump, BGM, and CGM). The first method of secondary analysis was to compare individuals based on insulin administration (pump vs. MDI) and glucose monitoring (CGM vs. BGM). Student's t-tests were used to compare demographics, survey responses, and CGM metrics (where applicable) among groups. Next, we used the same methods of analysis to compare individuals based on nonexclusive groups of pump users and CGM users versus those people using no new technology (MDI and BGM). We then created exclusive technology usage categories: Pump and CGM users, Pump-only users, CGM-only users, and MDI and BGM users. Logistic regression models using technology usage categories as the dependent variable to generate all regression statistics, including odds ratios and 95% confidence intervals were performed. Variables, including survey scores, laboratory (A1C) tests, and CGM metrics that were statistically significant in the model, were visualized in a forest plot.

Results

We analyzed 165 older adults with T1D (mean age 70 ± 10 years, 52% female, 97% white). In this community-living cohort, 22% of the participants were living alone, 55% had cognitive dysfunction, and 31% had depression. The number of average medical comorbidities was 10 per patient and they were taking an average 10 medications per day. A majority of the participants had a college diploma or higher education (88%). Mean A1C in this cohort was 7.4% ± 0.9% (57 ± 10.3 mmol/mol). For insulin administration, 38% of the population was using MDI, while 62% were using an insulin pump. For glucose monitoring, 58% were using personal CGM, while 42% were using glucometers. In addition, 41% of participants were using both insulin pump and CGM, while 21% were using MDI and BGM. The characteristics of the overall cohort, stratified by insulin administration and by glucose monitoring method, are shown in Table 1.

Table 1.

Characteristics of Study Population Stratified by Methods of Insulin Administration and Glucose Monitoring, n = 165

Participant characteristics	Total N = 165	MDI N = 63 (38.2)	Pump N = 102 (61.8)	P	BGM N = 70 (42.4)	CGM N = 95 (57.6)	P
Demographics
Age (years)	69.8 ± 9.7	69.9 ± 10.5	69.8 ± 9.3	NS*	70.2 ± 9.8	69.6 ± 9.7	NS
Female	86 (52)	34 (54)	52 (51)	NS	44 (63)	42 (44)	0.01
Body mass index	25.5 ± 4.1	25.3 ± 4.0	25.6 ± 4.2	NS	25.0 ± 4.5	25.9 ± 3.7	NS
Race (non-Hispanic white)	160 (97)	60 (95)	100 (98)	NS	67 (96)	93 (98)	NS
Education (college diploma or higher)	146 (88)	52 (83)	94 (92)	NS	60 (86)	86 (91)	NS
Living alone	32 (22)	15 (26)	17 (19)	NS	15 (23)	17 (21)	NS
Total medication/day	9.8 ± 4.4	9.9 ± 4.5	9.7 ± 4.3	NS	8.9 ± 4.0	10.4 ± 4.5	0.03
Comorbidities/medical conditions	10.3 ± 3.7	10.2 ± 3.8	10.4 ± 3.5	NS	10.7 ± 3.4	10.1 ± 3.8	NS
Cognitive dysfunction	92 (55)	41 (65)	50 (49)	0.04	40 (57)	51 (54)	NS
Depression	51 (31)	16 (25)	35 (35)	NS	20 (29)	31 (33)	NS
Diabetes characteristics
Age at diagnosis (years)	30.2 ± 18.8	33.2 ± 21.0	28.3 ± 17.3	NS	32.0 ± 18.0	28.7 ± 19.4	NS
Duration of diabetes (years)	39.7 ± 16.7	37.3 ± 16.9	41.2 ± 16.5	NS	38.1 ± 15.4	40.8 ± 17.6	NS
Personal CGM use	95 (58)	28 (44)	67 (66)	0.007	—	—	—
Insulin doses/day on MDI (unit/kg)	0.49 ± 0.25	0.49 ± 0.25	—	—	0.51 ± 0.28	0.43 ± 0.18	NS
Personal pump use	102 (62)	—	—	—	35 (50)	67 (71)	0.007
Insulin doses/day on pump (unit/kg)	0.48 ± 0.17	—	0.48 ± 0.17	—	0.49 ± 0.17	0.47 ± 0.17	NS
A1c (%)	7.4 ± 0.93	7.5 ± 1.1	7.3 ± 0.85	NS	7.6 ± 1.0	7.2 ± 0.87	0.02
A1c (mmol/mol)	57.1 ± 10.3	58.1 ± 11.8	56.6 ± 9.3	NS	59.2 ± 11.1	55.6 ± 9.5	0.02
Diabetes distress survey
Powerlessness	2.2 ± 0.86	2.2 ± 0.89	2.2 ± 0.84	NS	2.0 ± 0.69	2.3 ± 0.96	NS
Management distress	1.5 ± 0.68	1.5 ± 0.64	1.5 ± 0.71	NS	1.5 ± 0.57	1.6 ± 0.76	NS
Hypoglycemia distress	2.2 ± 0.86	2.2 ± 1.1	1.8 ± 0.90	0.03	1.8 ± 0.76	2.1 ± 1.2	NS
Negative social perception	1.5 ± 0.68	1.3 ± 0.49	1.4 ± 0.70	NS	1.2 ± 0.43	1.4 ± 0.73	0.01
Eating distress	1.6 ± 0.80	1.6 ± 0.65	1.7 ± 0.88	NS	1.6 ± 0.72	1.7 ± 0.86	NS
Physician distress	1.2 ± 0.52	1.2 ± 0.32	1.2 ± 0.60	NS	1.2 ± 0.47	1.1 ± 0.55	NS
Friend and family distress	1.5 ± 0.66	1.4 ± 0.55	1.5 ± 0.73	NS	1.4 ± 0.55	1.6 ± 0.73	0.02
Total score	1.6 ± 0.54	1.7 ± 0.55	1.6 ± 0.54	NS	1.5 ± 0.42	1.7 ± 0.61	0.04
Short Form-36
Physical component scores	50.0 ± 10.0	50.4 ± 9.7	49.8 ± 10.3	NS	51.0 ± 9.1	49.2 ± 10.7	NS
Mental component scores	50.0 ± 10.0	50.0 ± 10.7	50.0 ± 9.6	NS	51.0 ± 9.7	49.0 ± 10.2	NS
Hypoglycemia fear survey
Worry	14.2 ± 7.9	15.3 ± 7.3	13.6 ± 8.1	NS	13.3 ± 6.9	15.0 ± 8.6	NS
Behavior	15.0 ± 12.5	18.1 ± 14.1	13.1 ± 11.1	0.03	12.9 ± 13.3	16.9 ± 13.6	NS
Total score	29.2 ± 17.8	33.4 ± 19.1	26.8 ± 16.7	0.03	26.2 ± 14.4	31.9 ± 20.1	NS
IAH
Participants with IAH	60 (46)	22 (44)	38 (48)	NS	21 (34)	39 (57)	0.01
CGM metrics
Time ≤54 mg/dL (min/day)	11.8 (3.5, 32.4)	16.0 (5.5, 71.5)	10.3 (1.7, 25.3)	0.009	31.5 (15.4, 61.3)	5.5 (1.0, 13.3)	<0.0001
Time <7 0 mg/dL (min/day)	48.0 (22.7, 99.5)	60.3 (32.5, 146.8)	44.5 (19.1, 77.0)	0.01	83.7 (48.3, 137.6)	29.9 (10.9, 59.9)	<0.0001
Time in range  (70–180 mg/dL) (min/day)	879.0 ± 218.4	834.8 ± 197.1	901.2 ± 225.0	NS	832.4 ± 163.7	911.1 ± 246.5	0.02
Time >180 mg/dL (min/day)	489.6 ± 235.5	513.5 ± 221.8	479.1 ± 242.4	NS	501.5 ± 188.1	484.5 ± 267.0	NS
CV (%)	37.7 ± 7.0	39.4 ± 7.1	36.7 ± 6.6	0.03	41.8 ± 6.1	34.3 ± 5.7	<0.0001

Values are mean ± standard deviation, median (q1, q2), or n (%). Values in boldface type are statistically significant for α of 0.05.

BGM; CGM, continuous glucose monitoring; CV, coefficient of variation; IAH, impaired awareness of hypoglycemia; MCS, Mental Component Scores; MDI, multiple daily injections; NS, not significant.

First, we compared characteristics of people using pump versus MDI technology for insulin administration (Table 1). Pump users had a lower prevalence of cognitive impairment (49% vs. 65%; P = 0.04) and lower scores on hypoglycemia distress category measured on DDS (1.8 vs. 2.2; P = 0.03) and hypoglycemia fear survey (27 vs. 33; P = 0.03), compared to MDI users. Pump users were also more likely to use personal CGM for glucose monitoring compared to MDI users (66% vs. 44%; P = 0.007). Although laboratory A1C levels were not different between the two groups (7.3% vs. 7.5%; p = ns), the pump users had lower duration of hypoglycemia, both for glucose ≤54 mg/dL (10 min/day vs. 16 min/day; P = 0.009) and <70 mg/dL (45 min/day vs. 60 min/day; P = 0.01). Pump users also had fewer glycemic excursions as shown by lower CV% (37% vs. 39%; P = 0.03).

Next, we compared the characteristics of people using CGM versus BGM technology for glucose monitoring (Table 1). CGM users were majority male (56% vs. 37%, P = 0.01), on more daily medications (10 vs. 9; P = 0.03), were more likely to report IAH (57% vs. 34%; P = 0.01), and were more likely to also use insulin pump (71% vs. 50%; P = 0.007), compared to the BGM users. CGM users had higher scores on DDS (1.7 vs. 1.5; P = 0.04), especially in the categories of negative social perception and friend and family distress. CGM users had better glycemic profiles, including lower A1C (7.2% ± 0.9% vs. 7.6% ± 1.0%; P = 0.02), lower duration of hypoglycemia for both glucose ≤54 mg/dL (6 min/day vs. 32 min/day; P < 0.0001) and <70 mg/dL (30 min/day vs. 84 min/day; P < 0.0001), and fewer glycemic excursions as measured by CV (34% vs. 42%; P < 0.0001).

Next, we compared people using either pump or CGM, with those using neither of these technologies. As shown in Table 2, pump or CGM users had better glycemic control as measured by A1C (7.3% ± 0.85% and 7.2% ± 0.87% vs. 7.8% ± 1.1%; P = 0.04, 0.006) and better CGM metrics, including time of glucose ≤54 mg/dL (10 and 6 min/day vs. 53 min/day; P = 0.0006, < 0.0001), time of glucose <70 mg/dL (45 and 30 min/day vs. 121 min/day; P = 0.0002, <0.0001), time in range (70–180 mg/dL) (901 and 911 min/day vs. 790 min/day; P = 0.005, 0.004), and CV (37% and 34% vs. 43%; P = 0.0001, <0.0001). Pump users in particular were younger at the age of diagnosis (28 years vs. 36 years; P = 0.02). In addition, compared to those using MDI or BGM, people using either pump or CGM were more likely to have IAH (48% and 57% vs. 27%; P = 0.04, 0.006), and higher score for negative social perception (1.4 and 1.4 vs. 1.2; P = 0.02, 0.004).

Table 2.

Clinical and Continuous Glucose Monitoring Characteristics by Pump and Continuous Glucose Monitoring Usage Versus Multiple Daily Injections+Blood Glucose Monitoring, n = 165

	Pump N = 102	CGM N = 95	MDI+BGM N = 35	Pump vs. MDI+BGM P	CGM vs. MDI+BGM P
Clinical characteristics
Age (years)	69.8 ± ± 9.3	69.6 ± 9.7	71.4 ± 5.3	NS	NS
Body mass index	25.6 ± 4.2	25.9 ± 3.7	24.9 ± 3.8	NS	NS
Female	52 (51)	42 (44)	22 (63)	NS	NS
Race (non-Hispanic white)	100 (98)	93 (98)	33 (94)	NS	NS
Education (college diploma or higher)	94 (92)	86 (91)	29 (83)	NS	NS
Living alone	17 (19)	17 (21)	7 (21)	NS	NS
Total medication/day	9.7 ± 4.3	10.4 ± 4.5	9.3 ± 4.6	NS	NS
Comorbidities/medical conditions	10.4 ± 3.5	10.1 ± 3.8	10.1 ± 3.7	NS	NS
Cognitive dysfunction	50 (49)	51 (54)	22 (63)	NS	NS
Depression	35 (34)	31 (33)	7 (20)	NS	NS
Diabetes characteristics
Age at diagnosis (years)	28.3 ± 17.3	28.7 ± 19.4	35.7 ± 16.2	0.02	NS
Duration of diabetes (years)	41.2 ± 16.5	40.8 ± 17.6	35.7 ± 15.6	NS	NS
Insulin doses/day on pump (unit/kg)	0.48 ± 0.17	0.47 ± 0.17	—	—	—
Insulin doses/day on MDI (unit/kg)	—	0.43 ± 0.18	0.51 ± 0.28	—	NS
A1c (%)	7.3 ± 0.85	7.2 ± 0.87	7.8 ± 1.1	0.04	0.006
A1c (mmol/mol)	56.6 ± 9.3	55.6 ± 9.5	61.2 ± 12.2	0.04	0.006
Diabetes distress survey
Powerlessness	2.2 ± 0.84	2.3 ± 0.96	2.0 ± 0.63	NS	NS
Management distress	1.5 ± 0.71	1.6 ± 0.76	1.5 ± 0.71	NS	NS
Hypoglycemia distress	1.8 ± 0.90	2.1 ± 1.2	1.9 ± 0.81	NS	NS
Negative social perception	1.4 ± 0.70	1.4 ± 0.73	1.1 ± 0.33	0.02	0.004
Eating distress	1.7 ± 0.88	1.7 ± 0.86	1.5 ± 0.60	NS	NS
Physician distress	1.2 ± 0.60	1.1 ± 0.55	1.1 ± 0.29	NS	NS
Friend and family distress	1.5 ± 0.73	1.6 ± 0.73	1.3 ± 0.45	NS	0.01
Total score	1.6 ± 0.54	1.7 ± 0.61	1.5 ± 0.39	NS	NS
Short Form-36
Physical component scores	49.8 ± 10.2	49.2 ± 10.7	52.3 ± 8.1	NS	NS
Mental component scores	49.9 ± 9.5	48.9 ± 10.2	51.6 ± 10.4	NS	NS
Hypoglycemia fear survey
Worry	13.6 ± 8.1	15.0 ± 8.6	13.5 ± 7.3	NS	NS
Behavior	13.1 ± 11.1	16.9 ± 14.0	13.8 ± 10.0	NS	NS
Total Score	26.8 ± 16.6	31.9 ± 20.1	27.3 ± 16.6	NS	NS
Impaired awareness of hypoglycemia
Participants with IAH	38 (48)	39 (57)	8 (27)	0.04	0.006
CGM metrics
Time ≤54 (min/day)	10.3 (1.7, 25.3)	5.5 (1.0, 13.3)	53.4 (13.4, 85.1)	0.0006	<0.0001
Time <70 (min/day)	44.5 (19.1, 77.0)	29.9 (10.9, 59.9)	120.7 (40.2, 190.9)	0.0002	0.0001
Time in range  (70–180 mg/dL) (min/day)	901.2 ± 225.0	911.1 ± 246.5	790.4 ± 154.8	0.005	0.004
Time >180 (min/day)	479.1 ± 242.4	484.5 ± 267.0	530.6 ± 174.8	NS	NS
CV (%)	36.7 ± 6.6	34.3 ± 5.7	42.6 ± 7.3	0.0001	<0.0001

Values are mean ± standard deviation, median (q1, q2), or n (%). Values in boldface type are statistically significant for α of 0.05.

Finally, we compared technology usage in a logistic regression model using no technology use as the reference versus pump+CGM (Fig. 1A), pump+BGM (Fig. 1B), and MDI+CGM (Fig. 1C). Variables, including survey scores, laboratory (A1C) tests, and CGM metrics that were statistically significant in the model, were graphed in the forest plot. CGM users, independent of insulin administration methods, had better laboratory A1C and had fewer glycemic excursions as CV (%). CGM-only users also had comparable time spent in hypoglycemia to those people using pump and CGM. Overall, CGM users with or without pump had higher scores in the DDS subcategory “hypoglycemia distress,” Hypoglycemia Fear Survey total score and subcategory “behavior” score, and higher reporting of IAH than all other technology use categories.

FIG. 1.

Odds ratios for significant variables for all subgroups in the four categories of technology usage: (A) “Pump+CGM” (n = 67), (B) “Pump+BGM” (n = 35), and (C) “MDI+CGM” (n = 28), using “MDI+BGM” (n = 35) as the reference category. Odds ratios are plotted with their 95% confidence intervals and listed P-values. Boldface P-values were significant at α of 0.05. The x-axis is plotted on a logarithmic scale, with 1.0 and corresponding vertical line representing the null value of the odds ratio scale. BGM, blood glucose monitoring; CGM, continuous glucose monitoring; MDI, multiple daily injections.

Discussion

In this study, we examined the characteristics of older adults with T1D using diabetes-related technologies and the association of this technology use with glycemic metrics. In this cohort of community-living older adults with T1D, the use of diabetes-related technology for either insulin administration (pump) or glucose monitoring (CGM) was associated with a better glycemic profile, including better A1C, higher time in range, lower risk of hypoglycemia, and fewer glucose excursions (CV%). In addition, we identified several clinical characteristics that were associated with the use of insulin pump and CGM.

Older adults with T1D are a unique population, who have learned to manage a complex disease over many decades. They tend to have a higher need for insulin pump and/or CGM, compared to older adults with type 2 diabetes. However, unlike younger adults with T1D, older adults with T1D tend to have age-related challenges, including other comorbid medical conditions, and have variable decline in physical and cognitive fitness.²⁸ In our previous studies, we have found that older adults with T1D had five times the number of comorbidities than younger adults with T1D, including not only micro- and macrovascular complications of diabetes but also conditions such as cognitive dysfunction and polypharmacy.^28,29 Recognizing these comorbidities is important, as the use of technology by people with diabetes requires the ability to acquire new information, as well as the need to avoid errors. For insulin administration technology, pump use can be challenging or frustrating in older adults who have cognitive or physical decline and are unable to perform complex problem-solving tasks. The use of real-time CGM for glucose monitoring can be challenging if the overwhelming amount of data is not used appropriately, leading to overdoses or under dosing with insulin.

In our cohort, pump users had a lower likelihood of having cognitive dysfunction, which could suggest selection bias by either patients or clinicians to ensure the patient's ability to perform complex problem solving. On the contrary, lower risk of hypoglycemia and better glycemic control may benefit pump users in the context of preventing or worsening cognitive dysfunction. Further studies are needed to establish the directionality of this association. Of note, pump users were diagnosed at an earlier age and had longer duration of diabetes, compared to those people on MDI. These findings are consistent with previously published findings.³⁰ The cause for this association is not clear but may reflect clinician's preference to use pump from early age.

Our results for CGM users are consistent with a recent study evaluating the impact of personal CGM use in older adults with T1D, showing improved glycemic control and reduction in frequency of hypoglycemia episodes, along with their duration and severity.¹² Our data provide an important confirmation of these results in a real-world scenario. In our study, CGM users had a higher likelihood of having IAH. This finding is consistent with the commonly prescribed use of CGM in those people who are at a high risk for hypoglycemia. It is likely that both clinicians, as well as patients, were more inclined to start using CGM because of the high risk of hypoglycemia. It is interesting to note that CGM users had higher scores on the diabetes distress survey. They were particularly bothered by negative social perception and friends and family distress. It is possible that because CGM use includes alarms, and also offers the ability to share data with family, it may lead to feelings of interference and distress for the patient. These findings are contradictory to a previous online survey study that reported less hypoglycemia fear and less diabetes distress in older adults with T1D and T2D who were CGM users.¹⁰ This discrepancy might relate to a selection bias when online surveys are conducted, particularly in older adults, many of whom do not use internet.

Although the use of both pump and CGM was associated with improved glycemic metrics in our cohort, regression analysis showed that the lower A1C and shorter duration of hypoglycemia seen in pump users were largely driven by their use of CGM. These findings are consistent with a previous study showing no difference in glucose variability between groups of older adults with T2D using pump and MDI.³¹ Thus, CGM use might be encouraged in older adults with T1D, regardless of method of insulin administration.

The limitations of our study are the cross-sectional nature and multiple confounding factors that lead to the inability to establish directionality for the associations we found. In addition, this population of community-living older adults with T1D, cared for at a tertiary care diabetes center, may not include those people who are medically frail, or need assistance with their care.

In conclusion, our study identifies important characteristics of older adults with T1D using various diabetes-related technologies. Our results suggest the need for larger, prospective studies to develop criteria that can help clinicians choose the right technology for individual patients.

Authors' Contributions

Dr. Munshi had full access to all the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis. Study concept and design: Munshi, Slyne, and Toschi. Data acquisition, analysis, and interpretation: Munshi, Slyne, Michals, Davis, Dewar, Atakov-Castillo, and Toschi. Drafting of the article: Toschi, Slyne, Davis, Michals, Sifre, Dewar, Atakov-Castillo, and Munshi. Critical revision of the article of important intellectual content: Munshi, Slyne, and Toschi.

Author Disclosure Statement

M.M.: Consultant for Sanofi and Lilly. C.S., D.D., A.M., K.S., R.D., and A.A.-C.: No conflicts of interest relevant to this article were reported. E.T.: Consultant for Medtronic.

Funding Information

This study was supported by an NIH DP3 Grant (1DP3DK112214-01) and an NIH P30 Grant (P30DK036836). Continuous glucose monitoring materials were partially supplied by Dexcom. Role of the funder/sponsor: The funding source had no role in the design or conduct of the study; collection, management, analyses, interpretation of the data, or preparation and decision to submit the article for publication.