Assessing Regimen Adherence of Adolescents With Type 1 Diabetes

Participants were recruited for two studies. One study was a randomized clinical trial testing the efficacy of a behavioral intervention designed to improve management of type 1 diabetes, and the other study was a longitudinal study of developmental influences on management of type 1 diabetes. Both study designs included multiple assessments over time and used the DSMP as the primary measure of adherence. To evaluate and compare the modified DBRS, the 1-year assessment in both studies included both measures of adherence.

Youth-parent dyads were recruited from three urban medical center pediatric diabetes clinics to participate in two studies: the randomized clinical trial (n = 81) and the observational study (n = 87). At baseline, the sample included 168 dyads; 146 parents (89%) and 144 youth (86%) completed the 1-year assessment reported in this study. Eligibility criteria at baseline included 1) youth diagnosed for a minimum of 1 year with type 1 diabetes requiring insulin treatment and 2) youth aged 10–16 years. Exclusion criteria were 1) major chronic illness (except for well-controlled asthma or thyroid problems) and 2) non–English speaking/writing youth or parent.

Parents provided written informed consent, and youths provided written informed assent. All assessments were completed in the home or at another convenient location selected by the parent. Trained interviewers administered the adherence measures, separated by other measures, to the parent and child independently. At two of the sites, these measures were administered in person; the interviewers administering the DSMP offered to read the self-administered DBRS survey items. At one of the sites, these measures were administered over the telephone and the interviewer read all assessment items. Each participant (parent and youth) was given an incentive of $25 for completing the assessment. To evaluate test-retest reliability, the DBRS was administered twice, within 1 week, in a random sample of 32 youth and 31 parents. Institutional review board approval was obtained from all institutions involved.

Both parent and youth completed a modified version of the DSMP (17) in structured interviews. A potential problem with the DSMP, particularly when administered to younger adolescents, is that responses to the interview questions may not fit the scoring system; standardized probes necessary to code the response are not provided. Furthermore, the coding scales for individual items vary from two-point scales to five-point scales, and the meaning and metric for these different scales is not uniform even when the scale is the same. In response, the DSMP was modified to standardize wording across items, permitting administration by nonmedical interviewers, and to make the scoring less subjective. For example, the original question on hypoglycemia asks the single question “What do you do to treat your low blood glucose reactions?” Responses are assessed for whether the child takes carbohydrates, whether the child tests again after taking the carbohydrates, and the time frame for these responses. Most respondents will not spontaneously give all of these details and, though the original DSMP does not specify it, further prompts from the interviewer may be necessary. The modified DSMP elicits each of these elements with a series of simple questions contingent on the respondent’s responses: “What do you usually do when you have a low blood glucose?” “Do you eat or drink something right away?” “What do you eat or drink?” “How much do you eat or drink?” “How many grams of carbs is that?” “Do you check your blood glucose afterward?” “How soon after you eat or drink do you check?” Unfortunately, the DSMP for youth on a flexible regimen (21) was not available for evaluation when these data were gathered; however, the modified DSMP included adjustments to make it more appropriate for intensive treatment regimens. Final revisions were made by a panel including a pediatric endocrinologist, a certified diabetes educator, two developmental psychologists, and a clinical psychologist.

The modified DSMP included 29 items assessing the following elements of diabetes management: insulin administration routine, adjustment of the routine in response to changes in the routine, dietary routine, blood glucose testing frequency, and exercise frequency. Responses to individual items were scored by the interviewer. An overall DSMP score was derived by averaging scores from the five elements and represents the proportion of adherence to an optimal diabetes regimen (possible scores of 0.00–1.00), with high scores reflecting greater adherence.

The original DBRS was a 35-item, self-administered, fixed-choice survey (18,19). Limitations of the original scale included lack of items addressing insulin administration with the pump, the need for updating to current intensive regimens, and the need for adequate respondent reading skills. To address the first two limitations, items were revised to reflect current diabetes management practices and a pump version of the DBRS was developed, replacing insulin injection–specific items with items relevant to administration of insulin with an insulin pump. To address the reading skills requirement for self-administrated in-person administration, interviewers offered to read all items; for approximately half of the sample, interviews were conducted by telephone and all items were read to both parents and youth. The new instrument was reviewed and modified by a pediatric endocrinologist, a certified diabetes educator, and a developmental psychologist. The revised measure was then administered to an independent sample of five parents and youth in the same geographic area. After a second revision, the measure was then administered by an independent investigator to five parents and youth in a different geographic area. Feedback from this administration produced the final instrument. The insulin injection version of the DBRS has 36 items, and the pump version has 37 items. To make the results comparable with the DSMP, scoring was calculated as a proportion of the maximum possible score (possible scores of 0.06–1.00), with high scores reflecting greater adherence.

Glycemic control was assessed with A1C assays obtained at the clinics. To obtain a more stable estimate A1C, the average of the assays conducted in the preceding year was used as an index of glycemic control. Demographic data were gathered during the parent interviews.

Internal consistency of the scales was examined with Cronbach’s α. The 1-week test-retest reliability of the DBRS and 6-month test-retest consistency of the DSMP were examined with intraclass correlations. Criterion validity was evaluated with correlations between parent and child report, correlations between the two measures by the same respondent, and zero-order and partial correlations with glycemic control as assessed with A1C.

Of 222 families who met the inclusion/exclusion criteria, 76% (n = 168) consented to participate and completed the baseline assessments. The average time from baseline to the 1-year assessment was 14 months, with 87% (n = 146) of the baseline sample remaining (81 girls and 65 boys). Data from the 1-year sample are reported here. The mean age of the sample was 14.8 years (range 11–18, SD 1.98), with an average of 8.3 years since diagnosis. The sample was 81.5% white, which is representative of the clinic populations from which it was drawn. Approximately 45% of the sample used insulin pumps, and 15% were on other basal-bolus regimens. At baseline, compared with 22 subjects who could not be followed to 1 year, the families that completed the 1-year assessment had more education (t = 3.73, P < 0.001), were less likely to be Hispanic [χ² (161) = 10.9, P = 0.001], and the youth, but not the parent, reported better adherence to the diabetes regimen (t = 2.28, P < 0.05). They did not differ with respect to age, sex, other aspects of race/ethnicity, adiposity, time since diagnosis, diabetes regimen, or A1C. Mean age, adherence scores, and A1C are presented in Table 1.

The internal consistency (Cronbach’s α) for the DSMP was acceptable for parents (0.75) and youth (0.70). Internal consistency of the DBRS was good in both parent (0.84) and youth (0.84).

T tests revealed no significant differences between the levels of adherence reported by parent and youth on either of these measures (PS > 0.30). Parent and youth reports were significantly related, but these correlations were moderate (for the DSMP, r = 0.51, P < 0.0001; for the DBRS, r = 0.48, P < 0.0001).

The 1-week test-retest intraclass correlations for parent and youth responses on the DBRS reflected good reliability: 0.71 (P < 0.0001) for parents and 0.71 (P < 0.0001) for youth. The intraclass correlation for youth-reported DSMP at 6-month and 1-year assessments was 0.69 (P < 0.0001); the intraclass correlation for parent-reported DSMP at 6 months and 1 year was 0.73 (P < 0.0001). The method of administration varied by clinic. This prevented a direct comparison of method of administration across clinics for both the DSMP and the DBRS without having potential confounding of differences between clinic populations. However, in all clinics the method of administration of the DSMP at 12 months was different from the method of administration at 6 months. A within-subject comparison of scores across methods indicated that the DSMP scores of youth, but not of parents, were significantly higher when the DSMP was administered by telephone than in person [for youth-reported DSMP, t (144) = 4.91, P < 0.0001; for parent-reported DSMP, t (144) = 0.55, P = NS]. This effect remained significant when age, sex, ethnicity, and intervention group were controlled. Therefore, the method of administration was controlled in subsequent analyses.

The DSMP and DBRS were significantly related to each other among parent (r = 0.72, P < 0.0001) and youth (0.74, P < 0.0001). Both measures were negatively related to age, which is consistent with studies showing a decline in diabetes management during the adolescent years (parent-reported DSMP, r = −0.26, P < 0.01; parent-reported DBRS, r = −0.20, P < 0.05; youth-reported DSMP, r = −0.29, P < 0.001; youth-reported DBRS, r = −0.18, P < 0.05). White youths reported significantly higher adherence on the DSMP (t = 3.27, P < 0.01) and the DBRS (t = 2.64, P < 0.01) than youths from minority backgrounds. Parent responses were not significantly related to minority status. There were no sex differences on either measure. Parent and youth responses to the DSMP were significantly related to the 1-year cumulative index of A1C (r = −0.35, P < 0.0001 and −0.36, P < 0.0001, respectively). Parent and youth responses to the DBRS were significantly related to A1C as well (r = −0.35, P < 0.0001 and −0.34, P < 0.001, respectively). These relationships did not change substantially for partial correlations controlling for age, clinic site, method of administration, and assignment to intervention or control groups (Table 2).

The study provides good evidence for criterion validity for the modified versions of the DSMP and the DBRS. Parent and youth responses to the two measures were correlated, and responses to both measures were significantly related to the index of glycemic control, i.e., average A1C values over the previous year. These findings were robust. The significant relationships to glycemic control were maintained when age, sex, clinic site, method of administration, and intervention condition were statistically controlled.

The structure, wording, and scoring of the original DSMP make it more suitable for use as a diagnostic tool in a clinical setting than as a standardized research tool. Administration of the original instrument requires well-trained interviewers (16). The original DSMP interview has open-ended questions, but necessary prompts for each question are not standardized. The interviewer is required to transform the response for most of the questions to fit a scoring system that may not always correspond to the response. Standardized probes necessary to code the response are not provided. Furthermore, the coding scales for individual items vary, and the metric for these different scales is not uniform even when the scale is the same. Thus, while this may be an effective means for a well-trained clinician to identify problem areas for adherence, it is less appropriate for empirical analyses of individual differences in overall adherence. In addition, the instrument takes ∼20 min to complete; thus, costs attributed to interviewer training, interviewer time, and subject burden make it less suitable for large research studies. In the original versions of the DSMP, simple sums are used to calculate total scores resulting in differential weight given to different areas. For example, in the flexible regimen DSMP, blood glucose monitoring had a maximum score of 30 versus hypoglycemia, exercise, insulin administration, and eating (maximum scores of 11, 12, 16, and 17, respectively) (21). Weighting of scores or standardization across areas of diabetes management should be explored. If used as a research instrument, standardized training of all interviewers across sites and standardized probes are essential.

The psychometric properties of the revised DSMP and its relationship with glycemic control are comparable or better than those previously reported for the original DSMP (17,21). The validation study for the original conventional-regimen DSMP reported a Cronbach’s α of 0.76 (17). In the evaluation study of the original flexible-regimen DSMP (21), baseline Cronbach’s α for the adolescents were poor (0.45), and those for parents were marginal (0.69); however, these improved slightly at a repeated assessment 6 months later (adolescents 0.65 and parents 0.70). Six-month test-retest correlations of the original flexible-regimen DSMP were poorer for adolescents than for parents (usual-care group test-retest: parents 0.73 and adolescents 0.42; intervention group test-retest: parents 0.73 and adolescents 0.51). Consistent with the current study, there was no significant difference between parents’ and adolescents’ scores and their scores were correlated. Criterion validity of the original DSMP was good (correlation with A1C ranged from −0.20 to −0.28) (17,21). Overall, the revised DSMP presented in this study appears to improve on the psychometric properties of the original and provides standardized wording for interviewers without expertise in diabetes care.

The findings for the revised DBRS are consistent with previous studies. The original DBRS had good psychometric properties, with a Cronbach’s α of 0.86 (18), comparable with the values for the revised DBRS reported in the current study. These values are superior to the current and previously reported internal consistency for the DSMP. However, unlike the revised DBRS, the original DBRS was not validated against the current standard for glycemic control, A1C values. The revised DBRS updates the original DBRS to include insulin administered with a pump and has good evidence for criterion validity; it was significantly related to the DSMP, parent and youth scores were correlated, and it was significantly related to A1C.

The revised DBRS may be less useful as a clinical tool, as the structure does not provide an opportunity for the clinician to discuss responses during administration of the instrument. However, there are several advantages to the use of the DBRS as a research instrument. The time for administration of the revised DBRS is ∼7 min, a substantially reduced subject burden compared with the DSMP, and because the DBRS is self-administered, there is no need for extensive interviewer training. Also, included in the DBRS but absent from the DSMP are questions about the interaction between the patient and the diabetes health care team (16). The reliability and validity of the revised DBRS make it a very reasonable substitute for the DSMP.

There are several disadvantages to the current study. The sample includes patients from three different sites and patients that were randomly assigned to participate in an intervention. Methods for assessing A1C and for administering the adherence instruments (in person versus by telephone) varied across sites. Although this may increase the generalizability of these findings, there might be some concern that differences across sites and across intervention and control groups biased the results. The significant effect of method of administration for youth-reported DSMP deserves further investigation. However, variations across sites are more likely to mask significant relationships than to enhance them, and when site, method of administration, and group assignment were statistically controlled there was little effect on the results. Finally, some of the criticisms of the DSMP and the DBRS were addressed with standardized interviewer training and modifications to the items. It is possible that comparisons between the original DSMP and the original DBRS would have provided different results.

Both the revised DSMP and the revised DBRS are suitable measures of conventional and flexible self-management of type 1 diabetes. The evidence from this study indicates that the revised DSMP provides an effective standardized system for interviewing patients about adherence. The revised DBRS provides a measure of diabetes regimen adherence that can be self-administered with youth as young as age 11 years with good internal consistency and criterion validity. Investigators and clinicians should consider their goals and resources in deciding which instrument best suits their needs.

The research reported in this article was funded by the National Institute of Child Health and Human Development.

We thank Sandy Cook for providing the original version of the DBRS, Tim Wysocki for providing the original version of the DSMP, and Lani Greening for administering the revised DBRS to families during its development and for providing helpful feedback.