Using type 2 diabetes population data in general practice may affect workflow and clinical outcomes

Systematically managing the use of data on people with type 2 diabetes (T2D) within clinic systems can be a complex and time-consuming task for general practitioners (GPs) [1]. In Denmark, general practice manages the treatment of approx. 80% of all individuals with T2D. Research shows that improved data use by GPs can enhance treatment quality by identifying patients eligible for specific treatments [2]. Similarly, comparing population data in forums where GPs share knowledge and pursue quality improvement can optimise clinical practice and reduce variation in treatment based on measures like glycated haemoglobin (HbA_1c) [3]. Therefore, an internet-based quality programme for GPs in Denmark (treatment pathways, in Danish: “forløbsplaner” (FLP)) [4] was developed. The FLP system provides general practice with a quick overview of the clinic’s T2D population and identifies patients who could benefit from a comprehensive assessment. Despite this initiative, many GPs do not meet process goals [5]. Danish studies have demonstrated that electronic feedback to GPs has effectively promoted adherence to treatment guidelines, making it a valuable tool for quality improvement [6, 7].

This feasibility study aimed to optimise diabetes care in general practice through increased use of T2D population data. It was based on existing tools and GPs’ experiences and wishes. Through a tailored implementation of data use, the study assessed whether it is possible to increase the use of the FLP population overview in general practice and determined whether increased use of population data can optimise clinical outcomes for T2D patients.

Methods

Design

This was a quality development project implemented as a collaborative effort between the Steno Diabetes Centre Copenhagen, the Department of Endocrinology and Nephrology at Nordsjællands Hospital and the Unit for Quality, Research and Patient Safety in General Practice in the Capital Region (KAP-H). General practices (N ≈ 100) from a restricted area of Northern Zealand, Denmark, were invited to participate in the study (Figure 1). Participants were invited by announcement in a monthly newsletter (PraksisNyt), followed by phone calls and oral presentations of the study in general practice medical forums. A total of 19 general practice clinics volunteered to participate in the intervention (IC). Afterwards, six control clinics (CC) were recruited similarly to compare the potential effects of the quality development interventions in the ICs. As the study was designed as a quality development feasibility study, it was impossible to gather data from the specific area of the Capital Region represented in the present study to assess the representativeness of the ICs and CCs. However, compared with data from all GPs in the capital region, the ICs and CCs reached fewer indicator goals [8]. The CCs were not matched to the ICs.

Each clinic participated with a primary study contact (GPs or nurses) to facilitate study contact, manage the practical aspects of study elements and disseminate information to the rest of the clinic.

Figure 1 summarises the recruitment process.

Data from the internet-based treatment programme

In Denmark, different electronic medical record (EMR) systems exist. All EMRs include a laboratory scheme (LAB) where FLP automatically gather real-time patient data. The GPs can digitally access the FLP at the individual and population levels and compare their data with those of other GPs. At the population level, the data (quality indicators) are categorised into organisational aspects (e.g., time since last blood pressure measurements), treatment-specific results (e.g., proportions of patients with abnormal blood pressure) and proportion of vulnerable patients (the FLP categorises vulnerable patients using a combination of several comorbidities, registered by the general practice). FLP indicators are based on national guidelines for T2D treatment and are presented as interactive charts and diagrams. The GPs can select a specific variable and retrieve a list of patients with the chosen variable. All quality indicators in the present study were based on definitions from the FLP [3], and data are from current patients with T2D in the ICs and CCs. Data were accessed only at the population level. Experienced endocrinologists among the authors selected the presented variables before analysis. Data from all variables are available by request.

Intervention

Visits in general practice

The ICs were offered support to optimise the use of their T2D population data for treatment purposes. An experienced general practitioner and organisational advisor from the KAP-H planned a stepwise quality development process with each IC (Figure 2). The three visits lasted 1-2 hours and contained technical data management, exploration of current T2D care procedures and advice on relevant quality development areas related to the specific use of the population data chosen by the primary contact. The ICs received targeted advice on interpreting and applying data insights to individual patient care. The organisational advisor highlighted the national treatment guidelines, the improvement guide [9] and the “Plan, Do, Study, Act” (PDSA) principle as tools for quality improvement [10]. Changes in FLP variables were measured at each visit. Data consisted of process and results indicators: vulnerable patients, HbA_1c, blood pressure, cholesterol, kidney function, comorbidities and medical treatment and measurement statistics.

At follow-up visits, each IC was asked to estimate the approximate time used to work with the population data. The CCs continued their usual work routines and contributed with FLP data at baseline and one follow-up (after 8-9 months) during the same period as the ICs.

Assessment of the clinic’s organisation

The primary contact in each clinic answered a questionnaire about their internal organisation and T2D treatment procedures (e.g., type of practice, staff involved in T2D treatment, internal workflow routines), routine use of the FLP and experienced barriers related to data-driven quality development. Responses were recorded on a 1-5 Likert scale (arbitrary units (AU) (1: ”not at all”, 2: ”to a minor degree”, 3: “to some degree”, 4: “to a high degree”, 5: “to a very high degree”, and 0: “don’t know/not relevant” OR 1: “never”, 2: “less than once a month”, 3: “at least once a month”, 4: “every week”, 5: “every day”, and 0: “don’t know/not relevant”) at baseline and after 6 and 12 months for IC and at follow-up for CC.

Ethical considerations

The study was registered as “approval not required” with the Regional Ethics Committee of the Capital Region (F-22073139). Clinic owners gave written consent to participate and approved using their T2D population data for evaluation purposes on an aggregated level.

Statistics

As the study was a quality developmental project, no a priori power calculation was made. Available case data were included in the analysis. Within-group differences from baseline to the 12-month follow-up were analysed with a paired samples t-test. Between-group differences were analysed with a univariate general linear model with delta values as the dependent variable, group as a fixed factor and baseline value and T2D patient prevalence as covariates. Data are presented as mean (± standard deviation) or change score (95% confidence interval (CI)). The significance level was at 0.05 (IBM SPSS Statistics 28 for Windows, IBM Corp.).

Results

Four ICs left the study due to other priorities (Figure 1). Table 1 presents baseline characteristics. The ICs reported an average time spent working with FLP data and their chosen focus areas of 4.5 hours (2-14 hours) in the first six months and 2.2 hours (1-4 hours) in the second period. In four of 15 ICs, the work was done solely by GPs. In the remaining 11 ICs, the tasks were done primarily by nurses with supervision.

The practicality of using population data from the internet-based treatment programme regularly

The ICs reported reductions in experiences of inadequate knowledge of FLP and use of population data. We observed between-group differences in the degree of follow-up barriers related to experiences of FLP being too technically challenging, demanding specialist T2D knowledge, demanding FLP knowledge and useless in favour of the ICs (Table 2).

In-clinic organisation

The ICs increased organisational aspects from “less than once a month” to “at least once a month”: the number of in-clinic conferences with 1.1 AU (95% CI: 0.2-2.0), p = 0.018 from 2.3 (± 1.3) to 3.4 (± 1.3), use of T2D population data with 0.7 AU (95% CI: 0.2-1.1), p = 0.007 from 2.1 (± 1.0) to 2.7 (± 1.0), use of individual data with 0.8 AU (95% CI: 0.1-1.5), p = 0.028, from 2.2 (± 0.9) to 3.0 (± 0.9). There was a between-group difference in the regular use of population data of 0.9 AU (95% CI: 0.1-1.8), p = 0.041 in favour of ICs (95% IC: 2.7 AU (± 1.0) (“at least once a month” versus CC: 1.8 AU (± 0.4) ("less than once a month”)).

Clinical data

We observed within-group differences for ICs in process and result indicators. We observed between-group differences in favour of ICs for the increases in process indicators: percentage of patients with annual control, active use of the FLP, HbA_1c measurement, low-density lipoprotein (LDL) measurements and estimated glomerular filtration rate (eGFR) measurements; and result indicators: increase in patients with LDL < 2.5 mmol/l and reduction in patients with ischaemic heart disease (IHD) and no treatment with glucagon-like peptide 1/sodium-glucose cotransporter 2 inhibitor (Table 2).

Discussion

This study supports the applicability of quality optimisation in general practice based on FLP population data. Using population data and adopting a structured quality improvement approach with appropriate adaptations may improve clinical data and clinical outcomes in T2D care.

This study showed that compared to CCs, ICs experienced a 22% reduction in the proportion of patients with an elevated systolic blood pressure (SBP) and no blood pressure treatment, an 11% reduction in patients with high LDL and no lipid-lowering treatment and a 22% reduction in patients with diagnosis of IHD and no secondary prevention treatment. By planning quality development from prioritised treatment variables, GPs could implement actions and workflow routines to optimise the treatment of the T2D population through comprehensive knowledge of the population. This aligns with a recent review showing that an incorporated action plan aligned with recommended guidelines is an essential feature in electronic audit and feedback systems. The ability to work with real-time feedback on a patient list from a specific population was critical for frontline users [11]. Our study also aligns with that by Schroll et al., who showed the potential of data collection and electronic feedback of quality indicators for diabetes care in general practice [6]. The study by Schroll et al. focused on electronic feedback and found reductions in the proportion of patients without the recommended treatment. However, the reduction was smaller than found in our study, which focused on quality development and applied recommendations to workflow routines, specific actions based on population data and regular in-clinic follow-up. This suggests that combining electronic feedback with a structured quality intervention will enhance value and boost the likelihood of success. The implementation of quality development was generally not regarded as time-consuming by the GPs.

While ICs found the initiative easy and meaningful, the recruitment process proved challenging. This suggests that implementing the current FLP initiative in the future may be difficult, potentially limiting its full potential.

Strengths and limitations

The strength of the present feasibility study is that the quality development intervention was based on an already-designed tool. Therefore, the intervention was close to real life, making further implementation more relevant. It is also a strength that the intervention was tailored to each IC, making success more probable. However, tailoring the chosen focus areas to each IC made pinpointing the most critical part of the intervention difficult. The main limitation of this study is its non-randomised design and the fact that the participating clinics volunteered to participate. Selection bias cannot be discarded due to the risk that GPs in the ICs may have been keener to participate in data-inspired quality development than other GPs in the region. While the baseline FLP variables of the ICs and CCs were not different (Table 1), indicating that the general practices were similar in this respect, some differences were seen in organisational characteristics and in-clinic workflow routines. In this aspect, the small number of CCs is a limitation.

Future studies should address this limitation by employing randomised designs or adjusting for baseline differences. Additionally, challenges such as data integration, practitioner engagement and time constraints warrant further exploration.

Conclusions

The present intervention, a data-driven quality development process, may represent a practical solution for creating workflow routines through an optimised use of population data and improved data overview in general practice, leading to an optimised in-clinic organisation of treatment of the T2D population. The results indicate that a low-cost intervention can improve diabetes-related process indicators and clinical variables, benefiting people with diabetes. While the results are promising, potential bias underscores the importance of designing robust methodologies to ensure generalisability. Scaling this approach to a broader population could further validate its effectiveness and sustainability in chronic disease management.

Implications for practice and future research

Hopefully, the results reported herein may motivate the diabetes quality organisation in Denmark to implement the quality development process further. Moreover, the methods applied are transferrable to other patient populations with chronic diseases, enabling GPs to implement workflow routines for many patients. It is also possible to implement the quality development process in other countries where electronic medical records are currently in place. Future studies should focus on the most critical barriers for implementing the FLP.