Factors associated with glycemic control in children and adolescents with type 1 diabetes mellitus at a tertiary-care center in Thailand: a retrospective observational study

The study protocol was reviewed and approved by the institutional review board of Faculty of Medicine, Chulalongkorn University (certificate of approval no. 787/2014). Written informed consent was obtained from the parents or guardians of each patient, and the assent was obtained from each child. We enrolled 58 patients (28 boys and 30 girls; aged 13.6 ± 4.0 years) with T1D who had been diagnosed for >12 months and had been regularly followed up (at least every 3–4 months) at our pediatric diabetes clinic at King Chulalongkorn Memorial Hospital (KCMH), a tertiary-care university teaching hospital in Bangkok, Thailand, during 2014-2015. We excluded children who had intellectual deficits that may interfere with their psychosocial functioning and children who were not regularly followed up (seen in the clinic <3 visits over the past 12 months).

Patients were divided into 2 subgroups according to their average HbA_1c levels in the last 3 visits. In accordance with the International Society for Pediatric and Adolescent Diabetes guidelines, an HbA_1c cutoff of 7.5% was chosen to define excellent glycemic control for T1D. Thus, in the present study, “good glycemic control” (excellent to fairly good) was defined as average HbA_1c <8.0%, and “poor control” was defined as average HbA_1c ≥8.0%.

Data were obtained from each patient’s medical record at the time of their most recent visit, which included age, sex, onset characteristics, BMI, pubertal status, duration of diabetes, and current HbA_1c levels. Pubertal status was assessed by the method of Marshall and Tanner [9, 10]. Prepuberty was defined by using criteria of Tanner stage I for breasts in girls and testicular volume 3 mL or less for boys by Prader orchidometry at the study visit. Boys with testicular volume >3 mL, girls with Tanner breast stage of at least 2, and patients with a Tanner pubic hair stage of at least 2 were classified as having entered puberty. Details of diabetes management including insulin dosage, insulin regimen, and frequency of SMBG were collected. The patient and the parent or guardian who accompanied the child to the clinic were asked to complete a set of questionnaires during the most recent visit when the medical records of that visit were obtained.

Parents completed a set of questionnaires regarding their marital status, educational levels, and financial and employment statuses and a “Family relationship and functioning questionnaire” [11], which is a 7-item questionnaire that assesses the family’s ability to be flexible, supportive, communicative, and have enough income and time to support their child.

Youth with T1D were asked to complete a set of questionnaires including (1) “family APGAR scales” [12, 13] to assess the participant’s perception of family functioning by examining his or her satisfaction with 5 parameters of family relationships (namely, adaptability, partnership, growth, affection, and resolve); (2) “problem and conflict solving questionnaire” [14, 15], which is a 5-item questionnaire to assess the participant’s ability to manage conflict or solve problems; (3) psychological well-being and the diabetes-specified QoL scales were assessed by use of a modified World Health Organization-5 Well-being Index and; and QoL was measured by use of a modified Diabetes Attitudes, Wishes, and Needs (DAWN) survey [16, 17]. Patient participants were asked to rate how they felt during the last 2 weeks and to what extent they agreed with each statement.

All the questionnaires used in the present study had been translated and validated. Higher scores of these psychosocial indicators suggest stronger family support, good family relationship, better adaptive and problem-solving skills, and better QoL.

Data regarding adherence to the diabetes management were obtained from the record of data downloaded from a glucometer and the medical chart from 12 months before the child’s study visit. Four adherence measures were obtained: (1) the frequency of missing blood glucose checks in the week before the study visit (at least 4 checks per day were recommended for all children), (2) the number of missing insulin injections in the week before the study visit, (3) the number of the clinic visits that each child did not bring the blood glucose meter during the last 3 visits, and (4) the number of the clinic visits that each child did not bring the diabetes logbook during the last 3 visits. To obtain a single measure of adherence, we created a single composite score of the 4 adherence variables described earlier. We defined the adherence composite scores ranging from 0 to 8. We defined “good adherence” if the scores were ≥6 and “poor adherence” if the scores were <6. In addition, patient participants completed the “8-item Morisky Medication Adherence Scale (MMAS) questionnaire” [18], which is a self-rated inventory used as a quick screening measure to quantify medication adherence (on a scale of 0-8). Higher values of the adherence composite score and MMAS scores indicate better adherence.

Statistical analysis was performed using IBM SPSS Statistics for Windows, version 20.0. Normally distributed data were expressed as mean ± standard deviation (SD), whereas non-normally distributed data were expressed as median and interquartile range (IQR; Q1, Q3). Comparisons between 2 groups were analyzed using a Fisher’s exact or chi-square test for categorical data and an unpaired t test or Mann–Whitney U test for continuous data. Receiver operating characteristic curves were used to analyze the reasonable cutoff values of selected psychosocial indices for predicting poor glycemic control. Univariate logistic regression analysis was used to assess differences between the good control and poor control groups in various characteristics. P < 0.05 was considered significant in tests of statistical inference.

We enrolled 58 patients (28 boys and 30 girls) in the present study. Their demographic and clinical characteristics are shown in Table 1. Thirty-nine (67%) patients were on a basal-bolus insulin regimen with 4 or more daily injections, 12 (21%) patients were on twice-daily injections with premixed insulin, 2 patients (3%) were on twice-daily injections, and 5 patients (9%) were on thrice-daily injections with split mixed human insulin (RI/NPH) (modified conventional regimen) (Table 2). There were no patients using an insulin pump in the present study. Mean HbA_1c levels were

9.2 ± 1.9%. By the definition in the present study, there were 13 (22%) patient participants considered to have good control (average HbA_1c 7.0 ± 0.7%) and 45 (78%) patient participants in the group considered to have poor control (average HbA_1c 9.9 ± 1.6%). Age, sex, duration of diabetes, and the pubertal status were not different between the group with good control and the group with poor control. Weight did not differ significantly between groups, but those in the group with poor control tended to have a higher median BMI-standard deviation score (BMI-SDS) than those in the group with good control (Table 1).

The insulin regimen and the frequency of SMBG did not differ between patient participants in the group with good control and those in the group with poor control. Patient participants in the group with poor control required a significantly higher insulin dose than those in the group with good control (Table 2).

There was no significant difference between the groups with poor or good control in terms of the frequency of missing

SMBG or insulin injections in 1 week or the number of routine clinic visits in the last 3 visits to which patients had not brought glucometers or log books. Neither the composite adherence score nor the MMAS score was different between groups (Table 3).

Family income was significantly different between the groups with poor or good glycemic control. Eleven of 13 (85%) patients in the group with good control and 15 of 45 (33%) patients in the group with poor control had family income ≥30,000 Thai baht (THB)/month (P = 0.01). The average monthly household income in Thailand in 2015 (National Statistical Office of Thailand) was 26,915 THB or 786 U.S. dollars (USD; 1 USD equivalent to 34.241 THB, U.S. Federal Reserve Bank G5.A annual average rate 2015). There were no significant differences in financial self-sufficiency, parental age, parents’ education, or marital status between the groups (data not shown).

The median scores (IQR) of family relationship and functioning as reported by primary caregivers did not differ significantly between the groups with good and poor control (Table 4). Most of the psychosocial variables assessed by the child were significantly different between the groups. The family APGAR scores were higher in the good control group than those in the poor control group, and the scores for QoL were also higher in the good control group. Participants in the good control group appeared to have higher scores for problem and conflict solving than those in the group with poor glycemic control, but the difference was not significant, possibly because of the small sample size.

After performing the univariate analysis, significant factors (P < 0.05) were entered into a univariate logistic regression model to identify those most importantly associated with poor

Minimum and maximum scores of each questionnaire are given in the brackets. Higher scores of these indicators generally suggest stronger family support and better relationships. Values are presented as median (IQR; Q1, Q3). P-value corresponds to a Mann–Whitney U test. *P < 0.05 HbA_1c, glycated hemoglobin; IQR, interquartile range; QoL, quality of life

glycemic control. The only remaining significantly associated factor of poor glycemic control was a QoL score <25 (Table 5).