A quasi-experimental study was conducted to evaluate the effectiveness of a TST-based e-coach program on the knowledge, attitude, practice of patients self-administering insulin injections for the first time. The study adhered to the Transparent Reporting of Evaluations with Nonrandomized Designs statement. Participants provided informed consent during face-to-face interviews with a registered professional nurse for screening and the baseline assessment. Using PASS software (version 15; NCSS, LLC) to calculate the sample size, it was estimated that at least 86 participants were needed to give this study a 90% power at 2-sided 5% level of significance to detect an empirical effect size of 80.7% while allowing for an attrition rate of 20%. Between May 1 and December 1, 2022, 86 hospitalized patients who received their first insulin injection were recruited using convenience sampling at a level IIIA hospital in Suzhou, Jiangsu Province, China.

Repeated outcome measures were collected over 3 time points: before the intervention (T1), month 1 after the intervention (T2), and month 3 after the intervention (T3). Patients were included if they (1) were aged ≥18 years and met the diagnostic criteria of the guideline for the prevention and treatment of type 2 diabetes mellitus in China (2020 edition) [20]; (2) were required to use insulin for the first time under physicians’ orders; (3) had the ability to use smart devices and inject insulin by themselves; and (4) had clear consciousness, no communication barrier, and the willingness to provide informed consent and voluntarily participate in this study. Patients were excluded if they (1) had severe physical disorders or malignant tumors; (2) had severe heart, liver, kidney, or other organ lesions or neuropathy; (3) had mental disorders; (4) dropped out during the study period or did not complete the study for various reasons; or (5) failed to complete the educational course as required.

The trial was registered in the Chinese Clinical Trial Registry (ChiCTR2200058895) and conformed to the ethical guidelines of the 1975 Declaration of Helsinki. All study participants provided informed consent, and ethics approval was obtained from the Ethics Committee of the Second Affiliated Hospital of Soochow University (protocol code JD-LK-2022-025-01). It is worth noting that the registered sample size differs from the actual sample size due to variation in the sample size calculation tool version and the reference used. The sample size was reassessed using more recent tools, and the post hoc power analysis confirmed a statistical power exceeding 80%. Specifically, a power calculation was performed using PASS 15 software, based on a 2-sample t test, with the α set at .05 (2-tailed) and an effect size of Cohen d=0.426. Given 40 participants in the intervention group and 35 in the control group, the achieved statistical power was 80.7%, meeting the conventional threshold of >80% for statistical power. Furthermore, after the effectiveness of the experimental group’s protocol has been validated, the e-coach management program will be implemented on the control group based on their consent, ensuring the accuracy of the principle of fair compensation.

Using Microsoft Excel 2019 to establish a database and double-entry data, we numbered the study data and established different databases according to different scales to ensure the accuracy of the study data. SPSS software (version 26; IBM Corp) was used to statistically analyze the data, and the difference was considered statistically significant when P<.05. GraphPad Prism software (version 8.0; GraphPad Software) was applied to plot and analyze the repeated measures data.

Descriptive statistical analyses were performed on the general characteristics: among them, the measurement data conformed to normal distribution and were expressed using mean (SD); the count data were expressed using frequency and percentage. The baseline data of the 2 groups of patients were analyzed to verify whether they conformed to normal distribution and compared using t test and chi-square test.

Insulin attitude scores and insulin use knowledge-practice scores at all 3 time points were found to be normally distributed. Changes in the corresponding score outcome metrics and treatment effects across time points were then compared using repeated-measures ANOVA. In particular, given that the baseline blood glucose monitoring frequency differed significantly between the 2 groups, thereby compromising their comparability, a repeated-measures analysis of covariance (ANCOVA) was performed, with monitoring frequency entered as a covariate. This approach was adopted to enhance the scientific rigor of the findings and reduce potential bias.

A total of 86 participants were enrolled in this study. Owing to the study procedures and various contributing factors, 3 of 43 (6.98%) participants in the intervention group and 8 of 43 (18.60%) participants in the control group were lost to follow-up. Overall, 11 of 86 participants were lost to follow-up, corresponding to an overall attrition rate of 12.79%. During the follow-up period, 5 participants were lost because of nonadherence to the prescribed insulin dosage (2 in the intervention group and 3 in the control group). In addition, 1 participant in the control group self-discontinued insulin therapy. An additional 5 participants were lost to follow-up because they could not be reached by telephone or WeChat (1 in the intervention group and 4 in the control group). Ultimately, 75 participants successfully completed the follow-up: 40 in the intervention group and 35 in the control group. To verify that loss to follow-up did not introduce bias into the final analysis, we first compared the baseline demographic and clinical characteristics between participants who completed the study and those lost to follow-up within each group separately and observed no statistically significant or clinically meaningful differences (Table S1 and Table S2 in Multimedia Appendix 1). Furthermore, comparisons of the numbers and rates of loss to follow-up between the groups showed no statistically significant difference (P=0.20), indicating that the participants lost to follow-up would not affect the study results and that comparative analyses were appropriate (Table 2). We also compared the characteristics of the 75 participants who completed the study across the 2 groups and found that the characteristics were well balanced, confirming that the final analytical sample remained representative of the original study population.

Table 3 presents the sociodemographic and health-related characteristics of all 75 participants. With the exception of the frequency of blood glucose monitoring, the 2 groups had comparable clinical characteristics and outcome measures. The comparison of the number of participants lost to follow-up and the attrition rate between groups showed no statistically significant difference (P=.05 to .97). This indicates that the differential impact of attrition on the study outcomes was not statistically significant, thereby supporting the validity of comparative analyses (Table 3).

It was tested whether the insulin knowledge-attitude-practice scores obey an approximate normal distribution using chi-square test (P>.05; see Table 3 for exact P values), and the levels of insulin perception scores at different time points of the 2 groups of patients were analyzed using repeated-measures ANOVA, and the results of the sphericity test did not satisfy the conditions of Mauchly’s sphericity hypothesis, and the Greenhouse-Geisser test was used. The results (Table 4) showed statistically significant group × time interaction effects on total insulin knowledge scores (P=.03), but the group × time interaction effects on total insulin practice scores were not statistically significant (P=.08)—that is, the insulin knowledge of the 2 groups of patients at each time point showed an unequal tendency to change over time (Figure 3), but this difference was not obvious in insulin injection practice (P=0.35; Figure 4). The group factor could influence the total insulin knowledge-practice scores of the 2 groups of patients; pairwise comparisons showed that the total insulin knowledge score of the intervention group was higher than that of the control group at 1 and 3 months after the intervention, and the difference was statistically significant (P<.001).

The between-group effect (P<.001) and time × group interaction effect (P<.001) on the total insulin attitude scores of the 2 groups of patients were statistically significant—that is, the insulin attitude scores of the 2 groups of patients at various time points showed unequal trends over time (Figure 5), and the group factor could influence the insulin perception scores of the 2 groups of patients.

The findings of this study focused on the effectiveness of a TST-based e-coach program on the knowledge, attitude, and practice of patients self-administering insulin injections for the first time. The results of this study showed that the e-coach program is effective in improving insulin knowledge and insulin attitude but was not proven effective for insulin practice. This discrepancy may be attributed to several factors: the weak predictive power of self-regulatory variables for behavioral practice within the temporal event construct of the TST; the strong habitual nature of behavioral prepotency, which necessitates a longer intervention period; and patient-related psychological factors. A meta-analysis showed that the correlation between self-regulatory variables in TST and actual behavior was extremely low (r=0.04, 95% CI −0.07 to 0), whereas intention (r=0.33) and behavioral prepotency (r=0.38) exhibited only moderate correlations. Behavioral prepotency, which reflects the temporal cumulative effects of past behavior and habits, had a moderate influence on current behavior (r=0.38) [25]. Habit restructuring typically requires an intervention period exceeding 3 months; within a short-term framework, established behavioral patterns tend to override the practical translation of newly acquired cognitive gains. It has also been emphasized that, before applying TST to a specific behavioral domain, verifying the association between its regulatory variables and the target behavior is essential; otherwise, discrepancies between theoretical predictions and actual behavioral outcomes are likely to occur [26]. Furthermore, this may be related to the stigmatization experienced by patients initiating insulin therapy because of their disease, which makes them reluctant to seek medical help or adhere to treatment. A survey by Arifin et al [27] demonstrated that all 210 enrolled patients reported experiencing stigmatization and discrimination, and the degree of stigma exerted a significant negative impact on their quality of life, ultimately impeding effective disease management and public health efforts. In this study, the standardized insulin injection behavior validated by the e-coach model focused on enhancing users’ self-regulatory skills; its theoretical basis offered relatively weak explanatory power for behavior. Moreover, restructuring the injection habit itself requires an intervention period extending beyond 3 months and is difficult to correct through remote management alone. In addition, patients initiating insulin therapy may develop negative attitudes such as shame and perceived disease-related stigma, which can affect their quality of life and treatment adherence, thereby compromising the standardization of insulin injection behavior. Taken together, these possible contributing factors led to a discrepancy between knowledge, attitude, and insulin injection behavioral outcomes in this study, wherein cognitive gains could not be readily translated into stable behavioral practice within 3 months.

The results of this study showed that the reason for the low degree of improvement of insulin injection practice in the control group may be due to the lack of knowledge enhancement and time perspective [28]. Injection technique practice, as one of the factors affecting the blood glucose control of patients with diabetes with insulin injection,is vital in health education [29]. According to the TST, in addition to the social cognitive variables and self-regulation ability described by the TPB, TST should also include the time perspective (that is, the inconsistency in the length of time for which the intervention remains effective for a given individual leads to different value evaluation of individual practice) [28]. Studies have shown that many patients receiving insulin treatment, especially those who take insulin for the first time, tend to have a special fear of injection and lack of knowledge about diabetes prevention and management [30]. Due to the short duration of intervention in this study, it may not reach the degree of cognitive variables and time perspective potency of individual intention to change practice, that is, the improvement in the level of knowledge of patients in the control group may not be enough to change the direction of future practice, which led to the low improvement in insulin injection practice of patients. The prevention and control of chronic diseases requires early action before the symptoms of the disease appear; however, is it possible to adjust one’s practice in the appropriate direction only when they recognize the long-term consequences of current bad practice and try to reduce the intention-practice gap [31].

Furthermore, the results of item-specific scores for insulin injection behavior at different time points showed that the between-group differences were significant for item 3 (disinfecting an area >5 cm in radius centered on the injection site with alcohol) and item 5 (keeping the needle in place for more than 10 seconds after injection before withdrawal), and the interaction effect for item 5 was also significant. No statistically significant differences were observed for the remaining items. These findings suggest that the behavior of keeping the needle in place for more than 10 seconds after injection may be jointly influenced by the intervention and time factors, reflecting a dynamic change in patients’ mastery of this key point, which warrants reinforced emphasis during education. The degree of improvement in other behavioral aspects remains to be further enhanced, indicating that in patients receiving insulin injections for the first time, key weaknesses exist in procedural details such as preinjection assessment and injection site rotation. Merely reinforcing selected operational details is insufficient to avert all risks; individualized training must be implemented to strengthen these weaknesses [32]. In particular, for first-time insulin users, establishing a complete and standardized behavioral pattern at an early stage can prevent long-term complications such as lipohypertrophy, thereby further improving treatment adherence and quality of life.

Furthermore, the results of this study demonstrated that, after controlling for baseline covariates, the between-group differences in the temporal trends of insulin attitude scores across time points were not statistically significant. The TST underscores the predictive roles of domain specificity and temporal perspective on future health behaviors; however, the present findings did not reveal a significant trend of change in insulin attitudes over time among first-time insulin users. Studies [33,34] have shown that the most common reasons for the poor self-management ability of insulin-injecting patients are the lack of sufficient knowledge and motivation, and the lack of knowledge leads to nonstandard injection practice, which leads to complications. It is suggested that when guiding the insulin injection practice among patients initiating insulin therapy, medical staff should implement behavioral prepotency interventions targeting cognitive deficiencies as soon as possible after delivering the corresponding knowledge in the early stage of health education. The intervention contents may include the significance and benefits of health risk practices (misunderstanding of insulin injection process and consequences of nonstandard insulin injection) and health promotion practice (correctly standardizing insulin injection practice). To help patients adhere to the belief of connectivity between health practice and disease prevention, reduce the intention-practice gap, and narrow the gap between knowledge and health practice.

Repeated-measures ANOVA to test the effects of the e-coach health education management program on the insulin attitude of patients self-administering insulin injections for the first time showed significant differences in the intergroup effects and interaction effects between the 2 groups, indicating that group factors could affect the insulin attitude scores of the 2 groups. The insulin attitude scores of the 2 groups were not equal at each time point and showed the trend of changing with time. The TST emphasizes the predictive effect of domain-specific variables and time perspective on future health practice. The results of this study also show the changing trend of insulin attitude with time in patients self-administering insulin injections for the first time, which suggests that medical staff should comprehensively consider the 2-way influence of attitude cognition and time when paying attention to improving patients’ self-management ability so as to avoid fixed and stereotyped educational thinking. Health management attitudes regarding short-term and long-term benefits should be developed to enhance the influence of intention on future health practice [35]. As seen in Figure 5, the insulin attitude score of the test group showed an upward trend with time, while that of the control group was relatively stable, and the overall degree of change was not high. This indicates that the attitude of patients receiving insulin injections for the first time did not significantly improve since the end of routine care, and this information received less attention in daily life. It is recommended that nurses promote the concept of patients acquiring insulin knowledge and attitudes through future education to enhance their domain-specific knowledge and corresponding cognitive abilities.

The influence of patient age in this study should not be overlooked. The intervention population of the e-coach health education management program was restricted to first-time insulin users, whose overall age fell within the young and middle-aged range, making remote operation and management relatively easy. Consequently, the improvements in knowledge and attitude levels were reasonably favorable within the short-term intervention period. However, given the overall diabetes population, the proportion of older adults cannot be underestimated, and the extent to which their insulin-related knowledge, attitudes, and behaviors improve still requires further investigation. Previous results from the e-coach program have demonstrated that integrating one-on-one counseling with virtual group meetings reduced health coaches’ working hours by 40% without compromising the quality of implementation or the health benefits for older adults, thereby significantly enhancing project scalability and sustainability [36]. Moreover, home-based exercise programs remotely supervised via videoconferencing have been proven effective in overcoming mobility barriers, with systematic reviews supporting their feasibility and effectiveness in increasing physical activity participation among older adults [37,38]. For patients with chronic conditions such as obesity, diabetes, and hypertension, the e-coach intervention strategy, which is grounded in the “7 Steps to Health” approach and integrates self-determination theory, can further support the behavior change process [39]. Collectively, these findings indicate that the potential of e-coach health education program management in geriatric diabetes care merits further exploration. However, although older adults are progressively adapting to digital technologies, they still encounter difficulties in operating electronic health tools; therefore, intervention designs should emphasize age-friendly features such as simplified interfaces and voice interaction [32,40].

This study has limitations. It is worth further expanding the research on the intervention population. At present, the intervention population of the e-coach health education management program is limited to patients self-administering insulin injections for the first time, who are generally in the middle-aged and young age groups, and the remote operation and management is less difficult. Compared with the overall chronic disease population in China, the proportion of older patients remains substantial and should not be underestimated. Therefore, the extension and application of this intervention in the older patient population merit further exploration. In terms of intervention duration, it would be worthwhile to further validate the effects of interventions over 6 months or more to produce more accurate results.

The e-coach program based on the TST was effective in improving insulin knowledge and insulin attitude but was not proven effective for insulin practice. Medical staff should comprehensively consider the 2-way influence of attitude cognition and time when paying attention to improving patients’ self-management ability, so as to avoid fixed and stereotyped educational thinking. Health management attitudes regarding short-term and long-term benefits should be developed to enhance the influence of intention on future health practice. With regard to the duration of the intervention, the effects may be more pronounced when the intervention lasts for 6 months or longer; it would be worthwhile to further investigate the extent of these long-term effects in the future.