Improving diabetic kidney disease screening in a VA resident primary care clinic: a quality improvement initiative

Abstract

Background

Annual screening for diabetic kidney disease (DKD) with serum creatinine and albuminuria is recommended by multiple clinical guidelines but screening rates remain low nationwide (<50%). The Veterans Health Administration has implemented quality improvement (QI) interventions that resulted in a significantly higher national DKD screening rate of 63%. However, the rate in the resident primary care clinic at the Iowa City Veterans Affairs (ICVA) was only 41%, suggesting that these interventions may not be as effective in resident clinics. Residents have unique schedules and time demands, thus QI initiatives targeting this transient physician population need to be appropriately tailored.

Methods

An interprofessional working group aimed to increase the screening rate for DKD in the resident clinic from 41% to 65% over 8 months using Lean methodology. This non-randomised, prospective QI project was conducted at the resident primary care clinic at the ICVA, which serves approximately 3000 Veterans (including 750 with diabetes). Two sequential interventions were implemented: (1) a national electronic health record reminder and (2) a tailored, resident-specific, multifaceted approach involving incentives, gamification, workflow improvements, audit and feedback, and benchmarking. The primary outcome was monthly percentage of Veterans with both a urine albumin-to-creatinine ratio and serum creatinine tested in the past year. The process measure was weekly percentage of eligible Veterans tested for albuminuria within 1 day before or 7 days after a primary care appointment.

Results

The percentage of compliant Veterans in the resident primary care clinic increased from 41% (239/590) in June 2023 to 66% (395/603) by March 2024. Weekly screening rates of eligible Veterans rose from 16% to 48%. The biggest improvement was seen after the resident-specific interventions.

Conclusions

This project demonstrates the need for and effectiveness of tailored, resident-specific implementation strategies to achieve QI aims among these physicians.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Annual screening for diabetic kidney disease is recommended by multiple national and international guidelines, but screening rates remain low in clinical practice. Residents represent a unique, transient physician population that may benefit from tailored interventions to improve screening rates.

WHAT THIS STUDY ADDS

• This project describes the effectiveness of a tailored, resident-specific intervention that harnessed multiple implementation strategies, including provider-level feedback, incentives and improved workflow.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This project shows that resident physicians can be receptive to individualised performance feedback. Incentives and audit/feedback at the resident-level should be explored as a means of optimising primary care delivery.

Introduction

Problem description and specific aims

Diabetes is a major risk factor for chronic kidney disease (CKD) and the leading cause of end-stage kidney disease in the USA.¹ Diabetic kidney disease (DKD)—defined as having diabetes in combination with an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² and/or albuminuria >30 mg/day—affects approximately 600 000 of the estimated 2 million Veterans with diabetes.² It is associated with substantial rates of morbidity and mortality, which the VA spends billions of dollars annually to manage.³

One of the strongest predictors of disease progression within DKD is the degree of albuminuria.⁴ Albuminuria levels in patients with DKD direct the prescribing of several guideline-recommended medications that slow disease progression, including ACE inhibitors (ACEI), angiotensin receptor blockers (ARB), sodium glucose cotransporter-2 inhibitors (SGLT-2i), glucagon-like peptide-1 receptor agonists and non-steroidal mineralocorticoid receptor agonists (MRA).⁵ Timely prescribing of these medications is essential to improve patient outcomes and save costs.⁶ Albuminuria levels are also an important component of risk estimators such as the Kidney Risk Failure Equation, which Kidney Disease Improving Global Outcomes (KDIGO) recommends using to guide nephrology referrals and other management.⁷ Therefore, in addition to screening for DKD with a serum creatinine, routine screening for albuminuria with a urine albumin to creatinine ratio (uACR) is a key component of screening and managing patients with diabetes that is recommended by multiple organisations including the American Diabetes Association, KDIGO and the VA/Department of Defense.^8–10

Despite these clinical guidelines, national rates of albuminuria screening remain low, with US Renal Data System 2023 data showing that screening of patients with diabetes for albuminuria was <50% among Medicare beneficiaries in 2021.¹ Within the Veterans Health Administration (VHA), the largest integrated healthcare system in the USA, the percentage of Veterans with diabetes who had been screened for albuminuria was 63% as of June 2023. This overall higher screening rate may reflect national efforts within VHA to improve albuminuria screening rates in patients with diabetes by using tools such as a clinical dashboard with screening metrics and an electronic clinical reminder.¹¹ Unfortunately, these national efforts were not as effective within the resident primary care clinic at Iowa City VA, where screening rates were significantly lower at 41% (239/590). We hypothesised that previous national interventions had not been as effective in our resident clinic because they did not target residents specifically. We further hypothesised that effective implementation strategies would have to be tailored to meet residency-specific challenges, including residents’ status as trainees, the inherent transient nature of residency and the unique time demands resident physicians face. We formed an interprofessional working group with the aim of increasing the screening rates of DKD in the resident primary care clinic at Iowa City Veterans Affairs (ICVA) from 41% to 65% over 8 months.

Available knowledge and rationale

On chart review of patients with diabetes, it was discovered that Veterans seen in the resident primary care clinic had serum creatinine checked multiple times a year but were failing to be screened for albuminuria. Thus, the improvement team concluded that improving albuminuria screening with uACR testing would increase the clinic’s DKD screening rate and lead to better guideline-recommended medication prescribing habits by the residents.

Using Lean methodology, stakeholder interviews were conducted to create a process map and gap analysis to investigate why uACR rates were lower than desired (online supplemental figures 1 and 2). This showed that screening for DKD with uACR is a multifaceted process with several areas for potential improvement. A driver diagram identified three primary drivers to low uACR testing: the identification of eligible Veterans, the ordering of the lab, and the completion of the test (online supplemental figure 3). Many drivers were determined to be addressable by educational interventions as well as changes to the electronic health record (EHR).

Quality improvement (QI) projects performed elsewhere to address this issue have had success with EHR-based clinical decision support tools in combination with reminders, feedback and/or educational sessions.^12–15 Out of the over 70 implementation strategies identified in supporting practice changes, education is considered very feasible but only average importance,¹⁶ likely due to its low fidelity when compared with other more enduring interventions, such as automated order sets. However, previous QI projects have shown that education of residents can have a significant impact on multiple quality metrics likely due to residents being more open to education.^{17 18} Consequently, the team decided to include resident education as a key intervention, with the use of multiple implementation strategies to complement the impact of educational sessions. These strategies included (1) audit and feedback, (2) benchmarking, (3) incentives, (4) dynamic training (case-based learning), (5) developing educational materials and (6) distribution of educational materials.

Finally, special considerations were given to the unique aspects of implementing a QI project within resident primary care clinic, as compared with an attending-only clinic. To start, because resident physicians practice under the purview of an attending physician, clinical dashboards often do not drill down to resident-level data, which limits individualised feedback. As trainees, residents are not able to receive performance-based financial incentives, which have been used to improve performance in attending primary care clinics locally. Additionally, residency is transient, with new residents rotating in (and senior residents graduating) each July, thus QI initiatives targeted at residents ideally need to occur within these academic year cycles. Finally, residents face unique time constraints and have limited time in the clinic. However, residents at our institution do have protected learning time during the week they are in clinic to receive educational sessions around providing evidence-based primary care and may be more open to educational interventions when compared with attendings.

Methods

Context

This QI project was conducted at the resident primary care clinic at ICVA, which includes medical support assistants, 5 licensed practical nurses (LPN), 5 registered nurses (RN) case managers, 10 attending physicians and 40 internal medicine residents. The residents are grouped into five cohorts, with each cohort rotating through the clinic once every 5 weeks. Together, they manage the care of roughly 3000 Veterans, approximately 25% of whom have been diagnosed with diabetes. In addition to the primary care clinic staff, multiple institutional leaders, clinical applications coordinators, laboratory technicians and nephrologists participated in the planning and implementation of this project. Their input was solicited during stakeholder interviews, which were used in process mapping and gap analysis to identify potential targets for intervention. Additionally, they played a key role in a specific intervention—changing of the order menu.

Intervention development

Intervention 1: EHR clinical reminder

In late September 2023, ICVA volunteered to become a pilot site for the national automated clinical reminder with the goal of improving DKD screening. This national automated clinical reminder in the EHR provides a standardised way for LPNs to identify Veterans eligible for screening. This reminder automatically triggered when LPNs checked in Veterans with diabetes who were due for DKD screening. LPNs had the option to defer or resolve the reminder. To resolve the reminder, they needed to document outside records or order DKD screening labs. Interviews with frontline providers revealed lack of consistent completion of the reminder for various reasons (online supplemental figure 1). Therefore, an RN champion created an educational PowerPoint for LPNs and RNs in the resident clinic in December 2023 on how to complete the new automated reminder.

Intervention 2: multifaceted intervention

The team decided to focus on tailored, resident-specific interventions designed to enhance the clinical reminder rollout—incorporating education about the clinical reminder into the multifaceted intervention in November 2023. This consisted of incentives, gamification, improved workflow, audit and feedback, benchmarking, dynamic training (case-based learning), developing educational materials and distribution of educational materials. To incentivise and gamify screening, a competition was announced to residents who rotated in the primary care clinic by email and flyer postings in the clinic workspace. Residents were informed that prizes would be given to those with the highest screening rates by February 2024. The prizes included a kidney ‘plush’ (stuffed animal) for the top performer in each clinic cohort and a group awarded-sponsored dinner for the cohort with the highest screening rates.

In December 2023, the EHR lab order menu was updated to make the order for albuminuria, the test needed to screen for DKD, easier to find. This change involved multiple meetings with institutional leaders, clinical applications coordinators, laboratory technicians and nephrologists to identify preferred terminology and placement of the order prior to rolling out the menu change to the entire ICVA. Ultimately, the order was moved to a more prominent location, added to a new ‘urine’ test sub-menu and renamed to more accurately describe the test (urine albumin/creatinine ratio (microalbumin)) which prioritised the more descriptive test name but kept the previous term ‘microalbumin’ to ease the transition (online supplemental figure 4).

Educational sessions were conducted December 2023 through January 2024. These sessions were 15 min and given in-person to all residents rotating in the clinic. Each session included a presentation, handout and practice cases. Residents were also shown their current screening rates compared with their peers and the national screening rate. Residents were also shown how to find the order in the EHR.

Measures, study of the interventions, analysis and ethical considerations

This was a non-randomised, prospective QI project. The primary outcome measure was the monthly percentage of Veterans in the resident primary care clinic with diabetes who were up-to-date on annual DKD screening. This outcome was autocorrelated because when a Veteran is up-to-date on screening, they remain in the numerator until a year has passed from their last resulted serum creatinine or uACR. Serum creatinine was measured by enzymatic method in mg/dL with male reference range 0.60–1.30 and female reference range 0.50–1.10 (maximum analytical measuring range 0.10–37.00), with eGFR calculated using the 2021 CKD Epidemiology Collaboration creatinine equation.¹⁹ Urine albumin was measured in mg/L with no reference range (maximum analytical range 5–2000). Urine creatinine was measured in mg/dL with reference range 22.0–328.0 (maximum analytical measuring range 2.6–740.0). This measure was collected from the electronic quality measurement (eQM) dashboard, transferred to Excel (Microsoft, Redmond, Washington), and analysed using QI Macros (KnowWare International, Denver, Colorado, USA).^{20 21} eQM inclusion criteria included Veterans who had at least two encounters for diabetes or who had been dispensed insulin or other hypoglycaemic agents during the year prior. eQM exclusion criteria included age <18, age >85, receipt of hospice or palliative care services in the year prior, and Veterans with evidence of end-stage renal disease (ESRD), among others (online supplemental table 1). Statistical analysis involved creating a run chart with the centre line representing the median and relying on visual analysis only, as the measure was autocorrelated.²²

The process measure was the weekly percentage of eligible Veterans seen in clinic who received testing for albuminuria within 1 day prior or 7 days after a primary care appointment. Eligible Veterans were those with diabetes on their problem list and with no albuminuria testing completed in the prior year. This measure was collected monthly from the VHA’s Corporate Data Warehouse, transferred into Excel and analysed using QI Macros.^{20 21} To analyse data over time, a p-chart was used as it can identify variations in data as interventions are sequentially introduced. The three-sigma upper and lower control limits were calculated using a standard p-chart formula $\stackrel{-}{p}\pm 3\sqrt{\frac{\stackrel{-}{p}\left(1-\stackrel{-}{p}\right)}{n}}$ bound between 0% and 100%, where p̄ represents the overall proportion of events and n represents the subgroup denominator.²¹ The Institute for Healthcare Improvement rules for special cause variation were used to determine if the interventions resulted in significant change.²³ All data were stored behind the VA firewall. Standards for Quality Improvement Reporting Excellence (SQUIRE) 2.0 guidelines were used to report findings.²⁴ This project did not meet the regulatory definition of human subjects research and did not require review by the University of Iowa Institutional Review Board because this was a QI project to improve patient care.

Patient and public involvement

Patients and/or members of the public were not involved in the design, conduct, reporting or dissemination plans of this study.

Results

With the start of this project in October 2023, there were 777 Veterans in the resident primary care clinic at the Iowa City VA with diabetes on their problem list. The average age was 70% and 98% were male. Of these patients, 604 met the eQM inclusion criteria (online supplemental table 1) used for the primary outcome measure. The primary outcome measure was graphed monthly on a run chart (figure 1). The median percentage of Veterans with diabetes up-to-date on screening for DKD the 12 months prior to the academic year (July 2022 to June 2023) was 38%. This median performance level was fixed with the baseline data. The run chart shows a modest increase in the percentage of Veterans with diabetes up to date on DKD screening to 44% (268/604) by the end of October 2023. However, there is a significant up-trend starting November 2023, achieving the target of 65% as of March 2024. Overall, there was an absolute increase of 25% from June 2023 to March 2024, from 41% (239/590) to 66% (395/603).

Figure 1. This run chart shows the percentage of Veterans with diabetes assigned to the resident clinic with both a serum creatinine and uACR in the prior year. Numbered black arrows indicate (1) electronic health record reminder, (2) multifaceted intervention and (3) conclusion of all interventions. CL, centre line; DKD, diabetic kidney disease; uACR, urine albumin to creatinine ratio.

The process measure was graphed and analysed using a statistical process control chart (p-chart) (figure 2). Initially, 20 baseline data points preceding the interventions were analysed and determined to be in statistical process control with a mean screening rate of 16% (74/477) from May 2023 to September 2023. During the multifaceted interventions, multiple special cause signals were noted starting on 11/6/2023. All interventions were fully implemented by the week of 19 February 2024 and special cause signals remained present. To analyse the new state of DKD screening, split limit analysis was used for the same statistical process control chart starting at the implementation of the national reminder and at the start of the multifaceted intervention. This showed a new average screening rate of 48% (171/359). One special cause signal was still present from the week of 25 December 2023 to 8 January 2024 (two out of three consecutive points beyond 2 sigma deviations).

Figure 2. This statistical process control p-chart shows the weekly percentage of eligible Veterans seen in clinic who received testing for albuminuria within 1 day prior or 7 days after a primary care appointment. Limits were recalculated after the two interventions, resulting in split control limits. Numbered black arrows indicate (1) electronic health record reminder, (2) multifaceted intervention and (3) conclusion of all interventions. CL, centre line; LCL, lower control limit; PCP, primary care provider; uACR, urine albumin to creatinine ratio; UCL, upper control limit.

Discussion

Summary, interpretation and limitations

This project demonstrates improvements in both an outcome measure (percentage of Veterans with diabetes up-to-date on DKD screening) and a process measure (percentage of eligible Veterans screened within 7 days of a primary care appointment) for Veterans seen in a resident primary care clinic. The improvement team successfully achieved their goal of improving the percentage of Veterans with diabetes up-to-date on DKD screening from 41% June 2023 to 66% by March 2024, while also tripling the weekly screening rates of eligible Veterans from 16% to 48%.

In both the outcome measure and process measure, there was little to no impact of the national reminder on DKD screening in the resident primary care clinic until after the implementation of the multifaceted intervention. The p-chart (figure 2) in particular highlights the significant change in weekly screening rates after the multifaceted intervention was introduced, with sustained special cause signals even after all interventions were completed—establishing a new baseline three times higher than before (48% compared with 16%). These findings suggest that the incentives, changing the order menu, audit and feedback and education were key factors in driving changes and maintaining them in the resident primary care clinic. This is consistent with previous QI projects, which have shown that education of residents as a part of QI projects can have impact on multiple quality metrics^{17 18} and suggest that a different approach may be needed when conducting QI projects with residents, compared with standard QI projects in non-resident clinics.

In addition to the increased DKD screening rates that were achieved, the results highlight the potential utility in sharing target feedback with resident physicians, who do not have access to individualised screening rates and are not eligible for performance-based financial incentives due to their status as trainees. This was residents’ first opportunity to directly see how they were performing individually, compared with their peers, and compared with the VHA more broadly, as well as engage in friendly competition with their peers—receiving prizes for higher rates of screening. The VHA trains 45 000 resident physicians annually,²⁵ so including them on performance dashboards could allow for audit and feedback or gamification opportunities aimed at improving patient care. Of note, providing residents with data on quality metrics and benchmarks is recommended by the Accreditation Council for Graduate Medical Education Clinical Learning Environment Review Pathways—a tool used by clinical sites to optimise the clinical learning environment for physician trainees in the USA.²⁶

As noted in the results, there was a special cause signal from 24 December 2023 to 8 January 2024. The lower screening rate the week of 24 December 2023 could be related to the shortened clinic week with fewer patient scheduled, higher proportion of acute care visits and residents out for vacation.

There are several limitations to this project. First, this project was conducted in a specific environment (a resident primary care clinic at our institution), so results cannot be generalised to other clinics. Additionally, this project lacked balancing or secondary outcome measures, such as prescribing rates of renal-protective medications or referrals to nephrology, which would better evaluate downstream clinical impacts. This was mainly due to inability to collect these data. However, there have been studies that show screening for albuminuria correlates with prescribing of guideline recommended medications, including ACEI/ARB, MRA and SGLT-2i.⁶ Next steps include continuation of education yearly as new residents are introduced and expansion to other primary care clinics within the VA Iowa City Healthcare System, which combined serve approximately 7000 Veterans with diabetes.

Conclusions

This project improved DKD screening in a local resident primary care clinic by using a multifaceted QI intervention and highlights that EHR-based QI interventions can be augmented by other targeted interventions, such as provider-level feedback, incentives and improved workflow. It also underscores how multiple implementation strategies can be deployed to make educational sessions more effective, especially at the resident level. Incentives and audit/feedback at the resident-level should be explored as a means of optimising primary care metrics. As mentioned previously, resident physicians practice under the purview of an attending physician and often clinical dashboards do not drill down to resident-level data. This project shows that resident physicians can be receptive to individualised performance feedback. The VHA trains 45 000 resident physicians annually,²⁵ so including them on performance dashboards could lead to improved Veteran care across the nation.

The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the US government.

Data availability statement

Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.