Article Text
PDF
Abstract
Objective To examine the impact of multidisciplinary team input and intensive insulin therapy on glycaemic control in children and adolescents with diabetes over a 13-year period.
Design Two statistical approaches were used to interrogate the dataset. First a matched pair analysis to compare insulin treatment-type effect (pump vs multiple daily injections (MDIs)), followed by panel data regression to assess the impact of intensive re-education on glycated haemoglobin (HbA1c), in addition to treatment type.
Setting A large tertiary paediatric diabetes centre using a prospectively maintained database of clinical encounters from 2007 to 2020.
Main outcome measures Difference in HbA1c between treatment types (matching methodology) and expected change in HbA1c with treatment type and re-education (panel data).
Results Compared with MDI, matched pump patients had a lower HbA1c 6 months after pump commencement (ΔHbA1c=-0.53%, CI -0.34% to -0.72%; n=106). This effect was robust in controlling for socioeconomic deprivation (ΔHbA1c=-0.74%, CI -0.40% to -1.08%; n=29). Panel data analysis demonstrated a -0.55% reduction in HbA1c with pump therapy compared with MDI therapy (CI -0.43% to -0.67%). Patients who had intensive re-education had recorded an HbA1c of 0.95% (CI 0.85% to 1.05%) greater than otherwise identical patients prior to re-education. Following these sessions, HbA1c dropped by a mean -0.81% (CI -0.68% to -0.95%) within 6 months. These were also robust in controlling for socioeconomic factors.
Conclusions Compared with matched peers on MDI regimens, patients on pump therapy have lower expected HbA1c, an effect sustained for up to 8 years. Intensive re-education is associated with a significant drop in previously elevated HbA1c levels.
- Paediatrics
- Endocrinology
- Statistics
Data availability statement
Data are available upon reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Intensive insulin therapy is associated with improved glycaemic control but requires a multidisciplinary approach, which is resource heavy.
WHAT THIS STUDY ADDS
The positive association between glycaemic control of insulin pump therapy is sustained relative to multiple daily injections, but both treatment types are associated with improved control with intensive education sessions.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Direct intensive education is strongly associated with improved glycaemic control in patients who are struggling. This suggests that extra resourcing to complement intensive diabetes therapy and education may be justified by the better outcomes seen in both the short and long terms.
Introduction
The last decade has seen various changes in the management of paediatric type 1 diabetes (T1D)1–4 with an increasing use of multiple daily injections (MDIs, basal bolus regimens) and insulin pump therapy further facilitating proactive management.5–7 These forms of intensive insulin therapy (IIT) have largely replaced fixed dosing regimens such as two times per day (BD) or three times a day dosing8 as the preferable treatments, with consequent positive impact on diabetes control and quality of life.9 The successful implementation of IIT is dependent on adequate resource allocation including input from multidisciplinary team (MDT) members (eg, clinical nurse specialists, dietitians and social workers).8
Multiple studies have been performed using electronic databases to assess factors which may impact on diabetes outcomes, with varied conclusions. While most of the research points towards superior benefits of pump therapy, many studies are limited by short follow-up intervals and challenges in conducting high-quality clinical trials.10 11 One Australian study from 1999 to 2016 demonstrated sustained improved glycaemic control with pump therapy compared with age and body mass index-matched MDI patients,12 whereas a shorter UK randomised trial showed no significant difference between pump and MDI patients over a 5.5-year period.13 In the latter study, authors concluded that it is not pump therapy per se that improves glycaemic control but rather increased MDT input. There is a paucity of studies however, that specifically analyse the impact of MDT input itself on diabetes management over long periods.14 This study examines the specific impact of MDT input and pump therapy over a longer duration (13-year period).
Our study is based on data and experience from the largest paediatric diabetes centre in Ireland, with a long-standing history of improved access to pump therapy complemented by MDT support. At the end of our sample, there were 519 patients attending the service with an age range from 14 months to 19 years of whom 469 have a diagnosis of type 1 diabetes mellitus (T1DM); of these, 422 (90.0%) are on IIT and 237 (50.5%) used pumps. Since 2007, all outpatient interactions have been recorded electronically, including routine clinic visits, annual reviews, intensive nurse led re-education sessions, telephone calls, clinical psychology and dietetics consultations.
The active intervention programme (AIP) is a diabetes nurse specialist-led service offered to any patient with poor or deteriorating glycaemic control, warranting extra support between scheduled medical appointments. AIP sessions are arranged separately from regular clinics to allow an intensive dedicated one-to-one approach to any issues contributing to poor glycaemic control. They may be arranged for several reasons such as persistently poor control, concerns about patient disengagement from the service or poor practice reported at clinic. The content of each AIP session is tailored to the specific patients. A typical encounter will involve nurse specialist-led assessment and re-education with other MDT input as deemed necessary. These are initially provided on a single session basis, but further AIP sessions may be arranged if necessary.
Methods
Data
Data were retrieved from the centre’s electronic database (Diamond) from January 2007 to February 2020 inclusive. Analysis stopped in March 2020 due to the COVID-19 pandemic. This dataset consisted of 25 865 encounters for 1359 patients. Encounters here are defined as any in-person interaction; a single encounter may include interaction with multiple members of the MDT. In order to control for socioeconomic effects, the database is cross referenced with the national socioeconomic deprivation index, 2016 Pobal HP Deprivation Index, a national index used for evaluation of social deprivation.15 16 Patients are assigned a socioeconomic quintile based on this index. To facilitate a quantitative analysis, only encounters with a recorded glycated haemoglobin (HbA1c) were included. Telephone consultations were not included in the analysis. Direct intensive education is strongly associated with improved glycaemic control in patients who are struggling. This suggests that extra resourcing to complement intensive diabetes therapy and education may be justified by the better outcomes seen in both the short and long terms. The Children's Health Ireland Research Ethics Committee endorsed this project as a Service Evaluation.
Statistical methods
This study uses a ‘matching methodology’ and a ‘panel data regression’ to gain maximum information. Combining these two methodological approaches aims to use available data (1) to investigate for any difference in glycaemic control (HbA1c) in pump versus MDI therapy and (2) by using a panel data regression approach, to determine the expected difference in glycaemic control associated with AIP and other variables.
Matching
Every patient ever on a pump was extracted from the database along with demographic details: date of birth, date of diagnosis and HbA1c. Potential matches are determined subject to the following criteria: sex, being within 1 year of age, diabetes diagnosis within 1 year of each other, HbA1c within 0.5% of each other at the time of pump commencement and being in the same socioeconomic quintile. If there is more than one potential match, then the matching patient with the most data is selected. The socioeconomic criteria do reduce the number of pairs, so the results both with and without controlling for socioeconomic status are included.
Once each pair is determined, the HbA1c for each member of the pair is aligned relative to the time of pump commencement, time zero (t=0). The HbA1c for each member of the pair is recorded at 6 months prior to commencement of the pump therapy, at the time of pump therapy and subsequently 6 months, 1 year and 2, 3, 4, 5, 6, 7 and 8 years post commencement.
Three categories of glycaemic control were used for subgroup analysis, similar to Burckhardt et al.12 These were poor control (HbA1c >8.5%), suboptimal control (HbA1c ≥7.5% and ≤8.5%) and optimal control (HbA1c <7.5%) at the time of pump commencement.
Panel regression
The panel data regression models the relationship between variables using longitudinal data. Our database is used to determine the impact on HbA1c of time since diagnosis, participation in AIP, sex, age, socioeconomic quintile and insulin type. The hypothesised model took HbA1c as linearly related to these variables. For dichotomous variables (ie, insulin regimen type, participation in AIP and sex), a dummy variable was used to capture the impact of the presence or absence of that characteristic on HbA1c. Insulin regimens are quantified relative to a patient on a BD insulin regimen. The model in this study assessed the expected HbA1c in the 6 months prior to and after AIP sessions to quantify their impact on HbA1c. Estimation of the models used MATLAB’s fitlme function. Detailed explanation of the model including structure and assumptions is included in the online supplemental material.
Supplemental material
Results
Only patients with T1D were included, leading to a total 17 354 episodes for the 1249 patients (figure 1). After a review of all electronic clinic notes and admission records, 254 patients did not have a reliable date of diagnosis and were excluded. Of all 17 354 encounters of patients with T1D with at least one recorded HbA1c, 15 284 had an associated date of diagnosis, giving an 88% retention rate of those with T1DM or 59% of the 17 354 episodes. For each encounter the date, anonymised patient ID, HbA1c, mode of insulin delivery, sex, age, time since diagnosis and type of encounter (eg, routine clinic and annual review) were extracted. Descriptive statistics are given in table 1.
Flow diagram showing exclusion criteria from database. HbA1c, glycated haemoglobin; T1DM, type 1 diabetes mellitus.
Database description
Figure 2 shows the time series of glycaemic control (HbA1c) for the full sample with both socioeconomically matched and unmatched samples. An alternate matching method using the mean of all matching patients is presented in online supplemental figure 1. For the full sample (figure 2), there was no significant difference in prepump commencement HbA1c, but immediately after commencement of pump therapy, there was a sharp and sustained improvement in HbA1c for the pump cohort, evident within 6 months with a fall in HbA1c of −0.53% at 1 year relative to non-pump patients (7.47% vs 8.00%) (CI −0.34% to −0.72%). This improvement was sustained with a −0.52% difference at 8 years (7.80% vs 8.32%) (CI −0.03% to −1.01%). A similar effect exists for patients matched based on socioeconomic status though less persistence with significance at 5 years −0.61% (CI -0.07% to −1.15%). In the subgroup analysis (figure 3), the most significant and sustained improvement was in the suboptimal control group (figure 3B). By 1 year post commencement, pump patients had a -0.50% lower HbA1c than their matched peers (CI −0.28% to −0.71%). This remained significant with a difference of −0.84% (CI −0.25% to −1.43%) at 8 years. Improved control was also observed in the poor control group (figure 3A) up to −0.79% at 1 year (CI −0.30% to −1.28%). A sustained effect was not evident beyond 4 years, likely due to a reduced sample size and resulting loss of power. Optimal control patients showed a weaker but qualitatively similar pattern (figure 3C).
Glycaemic control of pump and non-pump cohorts both with and without matching for socioeconomic status. Underlying table presents average MDI cohort HbA1c minus average pump cohort HbA1c with calculated p value for null hypothesis of no difference using paired t-tests. Pairs with data presented for each time period. Data on the age at pump start are included for each time period and the percentage of patients assigned male. HbA1c, glycated haemoglobin; MDI, multiple daily injection.
Glycaemic control of pump and non-pump cohorts split by initial control. (B) Average non-pump cohort HbA1c minus average pump cohort HbA1c with calculated p value for null hypothesis of no difference. Pairs with data presented for each time period. HbA1c, glycated haemoglobin; MDI, multiple daily injection.
Results from the panel data regression are presented in table 2. Each individual coefficient can be interpreted as the impact of changes in the variable on HbA1c, assuming all other variables remain unchanged. The ‘pre-AIP’ dummy variable is strongly significant, indicating that in the 6 months prior to an AIP session being held, a patient’s HbA1c is on average 0.95% (CI 0.86% to 1.05%) higher than otherwise identical patients. This is consistent with expectations, as a deteriorating HbA1c is a common indication for an AIP session in the first place. The variable ‘post AIP’ shows that in the 6 months after an AIP session, HbA1c is expected to drop -0.81% (CI 0.68% to 0.95%). We provide alternative models in the supplemental material which account for possible ‘mean reversion’ effects, and AIP remains associated with an expected fall in HbA1c (online supplemental tables 1-6). The impact of socioeconomic status is significant, with the highest quintile status patients having an expected HbA1c −0.35% (CI −0.19% to −0.52%) lower than otherwise identical lowest quintile patients.
Estimated fixed effects of panel data regression
Given that IIT was considered the ‘new’ treatment for the purpose of this study, interpretation of differing insulin regimens was made with respect to BD dosing (‘old’ treatment), all else being equal. Comparisons also can be made between other regimens, for example, when comparing MDI patients to otherwise identical BD counterparts they are expected to have HbA1c -0.15% less than the BD cohort (CI −0.29% to −0.001%). The pump therapy cohort shows the best expected glycaemic control of all regimens with an HbA1c -0.70% lower than the BD cohort (CI −0.57% to −0.81%). Pump therapy HbA1c is expected to be -0.55% lower than that for MDI therapy (CI −0.43% to −0.68%).
Discussion
This is a novel study using two distinct statistical methodologies to explore a large dataset of paediatric diabetes encounters: (1) the impact of intensive insulin therapies and (2) the role of an AIP in their provision. This study contributes to the current evidence by using a new approach to support the thesis that pump therapy improves glycaemic control in addition to providing new evidence strongly suggestive of a clinically significant impact of intensive re-education sessions on glycaemic control. Our results are consistent with the consensus in the literature that pump therapy is associated with improved glycaemic control. These findings are robust to socioeconomic factors using a high granularity national index of social deprivation. In addition, our unique panel data regression model gives results which are both internally validating and interesting in the context of the recent study by Blair et al,13 who argue that pump therapy is not as important as ongoing education. Our results suggest that both pump therapy and additional support resources have a role to play in improving glycaemic control and that the expected effect of AIP sessions is independent of the clinical benefit of pump therapy.
The length of follow-up in this database is a strength of the study when compared with those in the existing literature which studied considerably shorter duration of follow-up.1 17–20
The overall results of the two methodologies are consistent with our hypothesis that pump therapy is associated with improved glycaemic control, rather than the move from BD to intensive injection therapy (ie, MDI). This is an interesting observation, as the education input required for a switch to MDI from BD regimens would not be considerably different from that required for the switch from MDI to pump.
This panel data approach allows us to simultaneously estimate the effect of multiple factors including insulin regimen, sex, age, time since diagnosis, socioeconomic status and AIP participation on HbA1c. It allows us to study the impact of multiple interventions such as moving the same patient from BD to MDI therapy or to pump therapy, something which is difficult to perform in a randomised controlled trial. This methodology also provides a clinically meaningful estimate of changes in diabetes treatment, that is, expected change in HbA1c.
Our data are representative of a large tertiary paediatric diabetes centre, the largest nationally, which participates in the international benchmarking programme SWEET ('SWEET' is an acronym derived from 'Better control in Pediatric and Adolescent diabeteS: Working to crEate CEnTers of Reference').21 We believe our data are representative of many paediatric diabetes centres both nationally and internationally, where new technologies are being embraced, alongside a recognition that intensive diabetes education is required to optimise their impact.
Limitations
We recognise that HbA1c is not an isolated marker for diabetes control. Other factors which may potentially influence risk of long-term complications, such as time in range, use of sensors, new fast-acting and basal insulins, and hybrid closed loop technologies, were beyond the scope of this study. We have not explored other outcomes, such as severe hypoglycaemia and rates of diabetic ketoacidosis in established diabetes, on glycaemic outcomes. Existing research indicates severe hypoglycaemia is less associated with pump therapy,20 22–25 and our experience is similar anecdotally. We note that increasingly, our patients are using continuous glucose monitoring (CGM), both in the form of sensor augmented pump, but more with stand-alone CGM devices. The interaction of CGM with pump therapy and its impact on outcomes warrant further study, and our methodology will be applicable to examine new treatments. The fact that this study was based on a relatively ethnically homogeneous cohort may be a limitation of the study but currently represents our population. The methodology, however, lends itself well to replication and expansion to multiple centres with a wider range of ethnicities.
Conclusions
As we move towards increasing use of more sophisticated diabetes technologies, more support and education from the diabetes MDT and continuing professional development will be required. The foundations for new evolving technologies have been set up in our service with a now proven success in our education programme. These are encouraging findings as the delivery of successful nurse-delivered education relies on intensive education to optimise the capabilities of evolving technologies.
The AIP is a resource-intensive programme which is clearly associated with significant improvement in HbA1c. The improvements are of a magnitude that justifies the continued and potentially increased resourcing of this and similar services to mitigate the long-term health and economic effects of poor diabetes control. We can therefore speculate that that ongoing routine education built into clinic visits will also have a positive effect on glycaemic control and future outcomes beyond the adolescent period.
Our findings provide evidence for the current standard of care of a combination of advanced insulin delivery regimens along with ongoing fully resourced intensive education and support teams. As we evolve towards the use of more complex technologies on the trajectory towards closed-loop technology for our patients, it is important that we have evidence to justify the investment in our MDTs and intensive education programmes.
Data availability statement
Data are available upon reasonable request.
Ethics statements
Patient consent for publication
Acknowledgments
The authors wish to acknowledge the assistance of Ms Debra Doherty, medical social worker at CHI Crumlin in the socio-economic grouping analysis.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Correction notice This article has been corrected since it was first published. The open access licence type has been changed to CC BY-NC. 14 Sep 2023.
Contributors JF: initial research proposal, data processing, statistical analysis including coding, formatting, interpreting and presenting statistical results, and primary authoring of first draft of paper. AE: clinical nurse specialist, primary designer of the active intervention programme, and contributed to design of project and drafting of the written article. ES: diabetes database management, collection and preparation of data for study, including addition of historical data corrections, input on research design and contribution to written article. SMO’C: refinement of research proposal and design, conducted literature review as part of the drafting process, co-ordination of work between team members, and significant redrafting or original draft with ongoing input. JF is the guarantor.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.