CGM Benefits and Burdens: Two Brief Measures of Continuous Glucose Monitoring

Abstract

Background:

Continuous glucose monitors (CGM) are underutilized by individuals with type 1 diabetes (T1D), particularly during the adolescent years. Little is known about perceptions of CGM benefit and burdens, and few tools exist to quantify this information.

Methods:

Two questionnaires were developed and validated—Benefit of CGM (BenCGM) and Burdens of CGM (BurCGM)—in a sample of adolescents ages 12-19 years involved in the T1D Exchange Registry. We chose to start the validation process with adolescents given their low CGM uptake and high risk for suboptimal glycemic outcomes. Exploratory and confirmatory factor analyses were conducted to confirm factor structure and select items. The resultant scales were tested for internal reliability and convergent/divergent validity with critical diabetes and quality of life outcomes: age, depression, diabetes distress, self-efficacy, technology attitudes, and diabetes technology attitudes.

Results:

A total of 431 adolescents with T1D completed the questionnaires (51% female, mean age 16.3 ± 2.26, 83% white non-Hispanic, 70% having used CGM). Two single factor scales emerged, and scales were reduced to 8 items each. Those who perceived higher benefit of CGM exhibited lower diabetes distress, higher self-efficacy, and more positive attitudes toward technology. Those who perceived higher burden of CGM exhibited higher diabetes distress, lower self-efficacy, and less positive technology attitudes.

Conclusion:

The BenCGM and BurCGM questionnaires each comprise 8-items that demonstrate robust psychometric properties for use in adolescents with T1D, and can be used to develop targeted interventions to increase CGM wear to improve diabetes management.

Type 1 diabetes (T1D) management has evolved in the past decade due to a rapid advancement in diabetes technologies. Continuous glucose monitoring (CGM) has contributed to dramatic changes in diabetes care due to its near-continuous surveillance of glucose levels, ability to integrate with insulin pump technology, and most recently replace necessary blood glucose checks for insulin dosing.¹ CGM devices have demonstrated glycemic benefit for individuals who use them consistently by reducing glycemic variability, reducing hypoglycemia, and lowering HbA1c levels.^2-6

Despite the glycemic benefits of using CGM, uptake is suboptimal, with adolescents using CGM least of all age groups.^7,8 Data from the 2015 T1D Exchange Registry indicated that <10% of adolescents were using a CGM as part of their diabetes self-management,⁷ with only a slight increase reported in 2017 of 14%.⁸ Reasons for low uptake are not well understood and research is limited. Known barriers to routine CGM use in children and adults include perceived inaccuracy of glucose readings,^9,10 high cost reported by parents,^9-12 frequent annoyances from alarms, insertion pain, and body image issues related to wearing CGM,^12,13 and interruptions to daily life.^14,15 Barriers in children and adolescents are less studied than in adults and are important to identify given the unique emotional, physical, and developmental challenges in this age¹⁶ that ultimately culminate in the poorest glycemic control of any age group with T1D.⁷

This study sought to determine the perceived benefits and burdens of CGM in an adolescent population by developing and validating two brief measures of perceptions. These measures could then be used to better understand the primary reasons adolescents with T1D choose to wear or not wear CGM as part of their diabetes self-management, and may inform behavioral intervention targets about diabetes technology use. The aims of this study were to (1) develop scales with relevant content related to perceptions of benefit and burden of CGM use and (2) assess their psychometric properties in a sample of adolescents with T1D.

Methods

Item Development

Scale content was intended to represent a range of practical and quality-of-life issues related to CGM use. Initial items were derived from a checklist of barriers to diabetes technology¹² including cost of supplies, trust in the device, wearing devices on the body, understanding the data, and alerts. We added additional burden content based on our clinical experience working with adolescents with T1D (LHM is an RN and certified diabetes educator, MLT, KAD, and KKH are clinical psychologists with diabetes expertise, and all work with adolescents with T1D). Benefit content was developed by the authors to address specific benefits of CGM including detection of hypoglycemia and hyperglycemia, conducting fewer blood glucose checks, sleeping or exercising better, and feeling more secure.^17,18 Content validity was ascertained based on extensive review and revision from the author group as well as two additional certified diabetes educators/RNs who provided expert consensus for the preliminary scale items. Next, the preliminary scales were field tested with 2 adolescents (mean age 16.5 years) and 4 young adults (mean age 32.25) with T1D who used CGM. Scales were reviewed for thematic content and wording clarity, and slight modifications were made.

The initial item pool included 14 Likert-type scale items to assess the benefits of CGM, with positively worded statements beginning with the stem “I think. . .,” as well as 15 items with negatively worded statements that were intended to assess burdens of CGM use. The total item pool had a Flesch-Kincaid reading level of grade 3.5.

Recruitment and Study Procedures

This study was part of a web-based online survey administered to people with T1D ages 12-19 years old who are part of the T1D Exchange Clinic Registry. An announcement about the study was emailed to 13,152 participants who had email addresses on file in the spring of 2018. Adults ages 18-19 years old provided informed consent online and completed the questionnaires. Parents of 12- to 17-year-olds provided consent for their child’s participation, and then the adolescent provided study assent and completed the questionnaires online. Data were collected via REDCap,¹⁹ and the surveys took 20-35 minutes to complete. The Stanford University IRB approved all study procedures and a $20 gift card was provided to participants as compensation.

Measures

In addition to the items on CGM benefits and burdens, participants completed questionnaires and demographic information to demonstrate construct validity, specifically convergent, divergent, and discriminant validity.²⁰ Demographic and diabetes information included age, gender, ethnicity, self-reported HbA1c, insulin pump and CGM use. Although there is high correlation between self-reported HbA1c and lab values,²¹ a significant portion of individuals in this sample did not report their HbA1c, and therefore we did not use this metric to establish validity. We hypothesized that those who choose not to use CGM perceiving less benefit to the technology or higher burden of use than those who have been exposed to CGM.

The Patient Health Questionnaire–8 (PHQ-8) was used to measure symptoms of depression,²² which has been validated in samples as young as 9 years old.²³ The PHQ-8 is identical to the PHQ-9 except for the omission of the suicidality item. Higher scores are indicative of more depressive symptoms.

Diabetes Distress was assessed by the Problem Areas in Diabetes–Pediatric version (PAID-Peds),²⁴ a 20-item questionnaire designed to assess burden and distress related to diabetes in the past month, and is scored so that higher scores indicate more distress. It has been validated in samples of individuals ages 8-17.²⁴

The Self-Efficacy for Diabetes Management (SEDM) questionnaire was used to assess personal perceptions of self-efficacy to care for their diabetes.²⁵ The SEDM is a 10-item scale developed for and validated in adolescents, and has been positively correlated with self-management in youth with T1D.²⁶

Two additional questionnaires were included that had previously been used with adults with T1D, the General Technology Attitudes Survey and the Diabetes Technology Attitudes Survey.¹² These surveys contain 6 and 5 questions, respectively, and were scored on a 5-point scale. The General Technology Attitudes Survey included 6 statements rated on a 5-point agree/disagree scale (eg, “Technology has made my life better”). The Diabetes Technology Attitudes survey included 5 questions rated on the same scale that were specific to diabetes (eg, “Diabetes technology has made my life better”).

Data Analyses

Descriptive statistics were used to analyze the individual items. Missing data were assessed for pattern and maximum likelihood estimation was used to mitigate a small amount of missing data.²⁷

Structural validity was assessed by splitting the sample randomly into two samples and undergoing two types of factor analysis. First, half the sample was used for principal component analysis (PCA). The PCA was computed for each subset of items separately (benefits and burdens) using eigenvalues >1 and Varimax rotation (assuming scales were orthogonal in nature). The second half of the data were used for confirmatory factor analysis (CFA) in order to test model fit. Adequate model fit was assumed if the χ²/degrees of freedom was <5, the root mean square error of approximation (RMSEA) was <.8, the comparative fit index (CFI) was >.9, and the standardized root mean square residual (SRMR) was <.08.^28,29 Mplus version 8 (Los Angeles, CA) was used for CFA calculations.³⁰ Modification indices were examined to reduce the item pool and achieve optimum model fit.

After the BenCGM and BurCGM scales were refined via factor analysis, they were analyzed for interitem and item-total correlations. Reliability was tested using Cronbach’s α test of internal consistency of the two scales, with a score >.8 being acceptable internal consistency.

Convergent and divergent validity were tested based on correlations with age, PHQ-8, PAID-Peds, SEDM, and general and diabetes technology attitude questionnaires. Discriminant validity was tested based on a t-test between means of individuals who have used CGM and individuals who have not used CGM. With the exception of the CFA, all other statistics were computed using SPSS v. 24 (Armonk, NY).

Results

Descriptive Statistics

In total, 431 adolescents with T1D completed the online survey battery related to their diabetes management. The participant characteristics are shown in Table 1. Of those who currently use CGM or have used CGM in the past, 99% indicated using either a Medtronic CGM or a DexCom CGM, as flash glucose monitors and implantable CGMs were not available in the United States at the time of this survey. Less than 1% of the data were missing for the BenCGM and BurCGM items. Of the data, 29% were missing for self-reported HbA1c and therefore only reported in Table 1. Data were imputed using maximum likelihood estimation.²⁷

Table 1.

Sample Characteristics (n = 431).

Characteristic	N (%)	Mean ± SD (min-max)
Age (years)		16.3 ± 2.26 (8.9-19.9)
Gender
Female	210 (51)
Male	218 (49)
Other/missing	3 (<1)
Race/ethnicitya
White, non-Hispanic	361 (83.8)
Hispanic	47 (10.9)
Black	18 (4.2)
Native American/Alaskan Native	11 (2.6)
Asian/Pacific Islander	9 (2.1)
HbA1c (%)		8.14 ± 1.17 (5.6-13)
Prior/current CGM use
Yes	302 (70)
No	129 (30)

^aRace/ethnicity equals more than 100% because categories were not mutually exclusive.

The initial items for each scale are found in Table 2. The Benefits of CGM items mean score was 3.91 (SD ± 0.64) and was not normally distributed, thus, nonparametric correlations were used for analysis. The Burden of CGM items mean score was 2.27 (SD ± 0.67). The Benefit and Burden items were moderately correlated to each other, ρ = –.64 (P < .001), indicating they measured related but distinct concepts.

Table 2.

Item Descriptions and Principal Component Analysis Results.

Mean (SD)	Item-total correlation (initial), ρ	Item-total correlation (final), ρ	Item #, final scale	Item	PCA factor loadings (1 factor)
BenCGM scale
3.89 (0.73)	.64	−	−	. . . CGM reads correctly most of the time	.67
4.22 (0.79)	.78	.82	1	. . . CGM makes taking care of diabetes easier	.83
4.09 (0.87)	.76	.76	2	. . . CGM helps take care of low blood sugars	.78
4.03 (0.82)	.69	−	−	. . . CGM helps take care of high blood sugars	.74
3.93 (0.88)	.61	.66	3	. . . CGM alarms are helpful	.63
4.03 (0.89)	.78	.80	4	. . . CGM makes me/would make me feel more secure	.84
4.02 (0.95)	.70	.74	5	. . . CGM lets me/would let me do less fingersticks	.67
4.12 (0.99)	.63	.63	6	. . . my family wants me to wear a CGM	.62
3.21 (0.99)	.54	−	−	. . . my friends want me to wear a CGM	.50
4.04 (0.93)	.58	−	−	. . . my diabetes care team wants me to wear a CGM	.53
3.95 (0.94)	.80	.81	7	. . . I take/would take better care of my diabetes with a CGM	.82
3.40 (1.12)	.68	−	−	. . . CGM helps with getting better sleep	.66
3.96 (0.91)	.76	.77	8	. . . CGM helps with managing blood sugar during exercise	.75
3.92 (1.00)	.83	−	−	. . . I feel/would feel better wearing a CGM	.83
BurCGM scale
2.31 (0.88)	.63	.72	1	. . . CGM sensor readings cannot be trusted	.70
2.25 (0.99)	.81	−	−	. . . CGM takes too much effort to use	.81
2.17 (0.93)	.80	.81	2	. . . CGM takes too much time to use	.84
2.73 (1.09)	.52	−	−	. . . CGM alarms too much	.58
1.89 (0.82)	.74	.80	3	. . . CGM is not helpful	.81
2.42 (1.05)	.71	.74	4	. . . CGM is painful to wear	.71
2.67 (1.07)	.48	.56	5	. . . the CGM is too expensive to wear regularly	.44
2.84 (1.09)	.65	−	−	. . . CGM causes too many skin issues (won’t stay on, rashes, etc)	.68
2.16 (0.94)	.757	.81	6	. . . CGM causes too much worry about blood sugars	.79
2.24 (1.15)	.761	.71	7	. . . I feel/would feel embarrassed about wearing CGM	.73
1.65 (0.77)	.594	−	−	. . . my family does not want me to wear a CGM	.64
1.68 (0.79)	.554	−	−	. . . my diabetes care team does not want me to wear a CGM	.58
1.79 (0.79)	.705	.72	8	. . . it is too hard to understand CGM information	.74
2.65 (1.29)	.676	−	−	. . . I do not/would not like how a CGM sensor looks on my body	.59
2.65 (1.28)	.622	−	−	. . . I do not/would not like when people notice me wearing a CGM	.55

Exploratory Factor Analysis

The PCA was performed in half of the total sample (n = 212), reserving the remaining data for CFA. For Benefits of CGM, a single factor PCA explained 50.8% of the variance, and all factor loadings were 0.50-0.84 indicating adequate to strong loadings onto the single factor (Table 2). For the Burdens of CGM scale, a single factor solution explained 47% of the variance and factor loadings ranging from moderate to strong at 0.44-0.84 (Table 2). Based on the theoretical consistency of one factor with the benefits construct and a separate single factor with the burdens construct, this model was chosen for further analysis.

Confirmatory Factor Analysis

CFA was run on the remaining sample (n = 219) that was not used for PCA. The 29 items from the BenCGM and BurCGM scales were tested together as correlated scales, independent scales, and as a bifactor model, but the initial models did not produce acceptable fit (RMSEAs > .10). Modification indices were used to determine which variables were more correlated with each other. Items with high cross-correlations between scales were removed from one of the scale to prevent contamination between scales. Some reciprocal items seemed to make model fit worse on both scales, and therefore were removed entirely (eg, “My diabetes care team wants me to wear CGM”/“My diabetes care team does not want me to wear CGM”). The final model was reduced to 8 items on the BenCGM scale and 8 items on the BurCGM scale (Figure 1). The final model fit included RMSEA = .078 (CI = .065-.09), CFI = .917, SRMR = .049, and χ²/df = 2.32 (Figure 1). This satisfied all model fit parameters indicating an acceptable fit for the data.

Figure 1.

Confirmatory factor analysis results for BenCGM and BurCGM scales.

Final Model Reliability

After reduction based on CFA, the final BenCGM and BurCGM scales each consist of 8 items (Table 2, Supplemental Table 1). The BenCGM mean score was 4.04 (SD ± 0.67), and individual items and overall mean were skewed moderately to the left (negative skew), meaning the individuals in this sample perceived relatively high benefits of CGM (Table 2). The highest item mean for the final scale was that “CGM makes taking care of diabetes easier” at 4.22 (±0.79). The final BenCGM scale items had a moderate mean correlation of 0.50, and the item-total correlations were all moderate to high, ranging from ρ = .63 to ρ = .82, all P < .001 (Table 2).

The final BurCGM scale mean score was 2.21 (SD = 0.69), and individual items and total scale were skewed to the right (positive skew), indicating relatively low perceptions of CGM burden (Table 2). The highest item mean score in final scale was “CGM is too expensive to wear regularly” at 2.67 (±1.07) The items correlated moderately with each other with a mean of 0.46, and all correlated moderately to strongly with the total scale, from ρ = .56 to ρ = .81, all correlations P < .001. The final BenCGM and BurCGM scales negatively correlated with each other at ρ = –.68, P < .001.

The final BenCGM had a Cronbach’s α of .89, indicating high internal consistency. Similarly, the final BurCGM scale had a Cronbach’s α of .87, also indicating strong internal consistency, Neither scale’s α improved with the deletion of any of items.

Convergent and Divergent Validity

Hypothesized relationships and validity testing of the scales are shown in Table 3. The final BenCGM scale did not correlate significantly with age or general depression, indicating divergent validity with these characteristics (Table 3). The BenCGM had a weak negative correlation with the PAID-Peds (ρ = –.10), representing a negative correlation to diabetes distress. It also showed a weak positive correlation with self-efficacy for diabetes management (ρ = .17). There were moderate positive correlations with general technology attitudes and diabetes technology attitudes (ρ = .40 and .41, respectively), indicating higher technology attitudes correlate with higher perceived benefits of CGM, and indicating convergent validity with these measures.

Table 3.

Convergent and Divergent Validity With Demographic Variables and Other Psychological Instruments.

	Hypothesized relationship	Type of validity	BenCGM, ρ (P)	BurCGM, ρ (P)
Age	No relationship	Divergent	.01 (P = .8)	−.02 (P = .77)
PHQ-8 (Depression)	No correlation to benefit or burden of CGM	Divergent	.06 (P = .26)	.04 (P = .44)
PAID-Peds (diabetes distress)	Negative correlation to benefits and positive correlation to burdens of CGM	Convergent	−.10 (P = .05*)	.35 (P < .001*)
SEDM (Self-efficacy)	Positive correlation to benefits and negative correlation to burdens of CGM	Convergent	.17 (P = .001*)	−.23 (P < .001*)
General Technology Attitudes	Positive correlation to benefits and negative correlation to burdens of CGM	Convergent	.40 (P < .001*)	−.35 (P < .001*)
Diabetes Technology Attitudes	Positive correlation to benefits and negative correlation to burdens of CGM	Convergent	.41 (P < .001*)	−.41 (P < .001*)

^*Statistical significance at alpha = .05.

The BurCGM scale did not correlate with age or general depression. It demonstrated convergent validity with the PAID-Peds (ρ = .35), indicating that higher perceived burden of CGM was correlated with higher levels of diabetes distress. There was a small negative correlation with self-efficacy (ρ = –.23), and small to moderate correlations with general and diabetes-specific diabetes technology attitudes (ρ = .35 and .46, respectively), indicating higher perception of CGM burden correlated with lower attitudes toward technology.

Discriminant Validity

Both the BenCGM and BurCGM scales detected a difference in mean scores between adolescents who had never worn CGM compared to those who had worn CGM. The BenCGM scale mean score was 3.61 (SE = 0.671) for those who had never worn CGM and 4.36 for those who had worn CGM (P < .001). The BurCGM mean score was 2.60 for those who had never worn CGM and 1.91 for those who had (P < .001). The mean BenCGM score was positively correlated with CGM exposure (ρ = .39, P < .001), and mean BurCGM score was negatively correlated to CGM exposure (ρ = –.34, P < .001). As hypothesized, those who have experience with CGM perceive higher benefits and lower burden to CGM than those who have not worn it before.

Discussion

This study is the first to present brief and valid measures for perceptions of CGM in light of current CGM technology. In this sample of adolescents with T1D, the BenCGM and BurCGM scales are novel in their ability to briefly assess perceptions of current CGMs, and may contribute to greater understanding of CGM use patterns and behavioral intervention targets.

Previous measures of individual perception of CGM include the CGM-Satisfaction Scale, developed in 2005 based a first generation CGM devices.³¹ The scale is lengthy (37 items), and omits major themes relevant to modern CGM use including reduction in number of blood glucose checks needed, improved accuracy and trust, and utility of alarms. Another instrument used to assess perception of CGM is the Glucose Monitoring Satisfaction Scale that relates to general glucose checking practices.³² This scale is intentionally agnostic to device or technology, so can be used to assess satisfaction with blood glucose checking or CGM technologies, but this also makes the scale less strategic for assessing perceptions specific to CGM devices. The BenCGM and BurCGM scales are therefore important in their specificity and ability to reflect the current state of CGM technology.

The overall psychometric analyses of the BenCGM and BurCGM scales were robust, and demonstrated acceptable convergent and divergent validity with existing measures. The directions of the associations that we hypothesized were confirmed in this sample. The strongest associations with the BenCGM and BurCGM scales were found with technology and diabetes technology attitudes surveys, which was expected. This indicates that more general perceptions of benefits and burdens with technologies also have relevance to CGM as a specific tool for diabetes self-management.

Understanding individuals’ perceptions of CGM is imperative for both clinical and research settings. Both the BenCGM and BurCGM scales are able to discriminate between individuals who have worn CGM and those who have not worn CGM. There are potentially misconceptions about the technology among CGM naïve individuals that should be ascertained. The BenCGM scale may provide explicit cues for individuals and providers to consider as they weight how CGM might enhance self-management practices. Likewise, the BurCGM scale could assist clinicians in identifying primary barriers to CGM use, and create appropriate interventions to overcome them. Overall, these scales provide important patient-reported outcome metrics to further understand complex behaviors related to adherence and diabetes self-management.

As CGM technology rapidly evolves, perceptions of CGM will evolve as well. Although the content developed for these scales was based on currently existing technology and not future technologies, the fundamental issues related to CGM will likely remain the same. These scales may serve as a yardstick to measure the differences in perception of usability between iterations of future CGM technology and to inform future device design efforts.

The strengths of this analysis include the extensive content development related to perceptions of CGM in adolescents with T1D, high internal consistency for both the BenCGM scale and BurCGM scales, structural validity based on independent (split-sample) factor analyses, and structurally and theoretically sound final scales. In addition, construct validity testing demonstrated that the scales were significantly correlated with measures of general and diabetes technology attitudes, diabetes distress, and self-efficacy.

There are limitations to this instrument and study. The BenCGM mean score is high, meaning that due to the ceiling effect, the scale may have reduced ability to detect change toward more positive perceptions of CGM. Likewise, the BurCGM mean is low, reducing the ability to detect further worsening of CGM burden perception. The study data were gathered only in an adolescent population from the T1D Exchange registry, and may not be representative of the larger adolescent population or other age groups. HbA1c measures were self-reported, which is why a significant portion of the data was missing for this variable and why it was not used for validity analysis. Research will be needed to validate the BenCGM and BurCGM scales in adults and children as well as other populations such as individuals with type 2 diabetes. Further work should entail testing whether the scales can predict successful CGM use for individuals, either measured by frequency of use or duration of use over time.

Conclusion

Overall, the 8-item BenCGM and Bur CGM scales are brief instruments that demonstrate high internal consistency and reliability in a large sample of adolescents with T1D. Further testing will elucidate new properties of these scales and determine how to best use them in clinical practice and research with the goal of improving diabetes care for youth living with T1D.