Targeting insulin stacking to address overnight hypoglycaemia in hospitalised patients with diabetes

Abstract

Inpatient hypoglycaemia is a significant concern in patients with diabetes due to its association with increased mortality. At the Veterans Affairs Greater Los Angeles Healthcare System, we developed a project to reduce overnight hypoglycaemia in hospitalised patients with diabetes by addressing insulin stacking, defined as insulin dosed within 4 hours of each other. By delaying the timing of bedtime correctional insulin administration in the electronic health record, we achieved a 28% reduction in the proportion of patients experiencing insulin stacking after one year. This led to significant decreases in overnight hypoglycaemia.

Introduction

Diabetes is one of the most common diagnoses among hospitalised patients in the USA, affecting one in four inpatients.¹ Hypoglycaemia in these patients is a major concern as it leads to worse clinical outcomes including increased mortality.² Consequently, the American Diabetes Association’s Standards of Care recommends a glycaemic target of 140–180 mg/dL in most hospitalised patients but will allow for blood glucose values as low as 110 mg/dL in those who are able to achieve this without hypoglycaemia.³ Most hypoglycaemic events occurring in the hospitalised patients are iatrogenic; consequently, careful attention is needed to eliminate these serious and avoidable events.^{4 5}

In the Veterans Affairs Greater Los Angeles Healthcare System, frequent overnight hypoglycaemia was anecdotally noted. Overnight hypoglycaemia is particularly concerning as it can go without detection as sleeping patients may not awaken to communicate their symptoms. Further investigation revealed that insulin stacking of the dinner nutritional insulin and bedtime correctional insulin, which was often being administered at intervals less than 4 hours, may be a potential driver of these low blood sugars. Insulin stacking results in hypoglycaemia because multiple doses of insulin are administered within a short period, often before the effects of previous doses have fully manifested.⁶ There is no standard time definition for insulin stacking, but this study used less than 4 hours based on the known insulin pharmacokinetics.

The aim of this project was to decrease the percentage of low blood sugars experienced overnight by hospitalised Veterans with diabetes mellitus by at least 25% (outcome measure) from April 2022 to June 2023 by decreasing the rates of insulin stacking (process measure).

Methods

A multidisciplinary team of endocrinologists and hospitalists and the chief resident in quality and safety created a process map (figure 1) for insulin administration in patients with diabetes, defined as having an A1C value ≥6.5 any time between 1 year prior to their admission or up to 7 days after their admission. Based on our process map, we noted that the electronic health record (EHR) order for bedtime blood sugar check and insulin sliding scale administration was set to 21:00, while the dinner sugar check and sliding scale administration was set to 17:00. This is exactly 4 hours separation. Nurses are allotted +/- 1 hour- grace period to administer medications; consequently, in our pre-intervention data, 42% of patients in our hospital were receiving these insulin doses <4 hours apart. Thus, we worked with pharmacy, informatics and nursing to change the timing of the bedtime sliding scale insulin administration from 21:00 to 22:00 in the EHR, which created a forcing function, a strong action for driving change based on the Institute for Healthcare Improvement action hierarchy. This change went into effect in June 2022. This change was communicated to the nursing supervisors to disseminate to the rest of the nursing staff in the entire hospital.

Figure 1. This process map details nurse (RN) workflow for insulin administration which demonstrates the possibility for insulin stacking based on the current state electronic health record insulin dosing schedule.

To evaluate the short and long-term clinical impact of this intervention, we used descriptive statistics and χ2 analysis to compare process and outcomes measures in the pre-intervention period (April to June 2022), to immediately postintervention (June to September 2022) and 1-year postintervention (April to June 2023). This study was found to be exempt by the local Institutional Review Board.

Results

Our intervention reduced the proportion of patient experiencing insulin stacking by 23% in the quarter immediately postintervention and 28% in the quarter a year postintervention (table 1). This resulted in a reduction in blood sugars below target (110 mg/dL) at both time points. True hypoglycaemia (blood glucose<70 mg/mL) was rare but reduced by 50% at a year post-intervention (table 1). Blood sugars in the range of 70–99 mg/dL, which can often predict future hypoglycaemia, reduced by 26% and 54%, respectively.⁷

Table 1. Clinical impact of change in nightly blood glucose (BG) check and insulin administration.

	1 April 2022 to 21 June 2022	22 June 2022 to 22 September 2022	1 April 2023 to 21 June 2023
	Pre-intervention	Immediately postintervention	1 year postintervention
	n=2148	n=2214	n=2293
Process measures
Mean time of bedtime BG check	20:53	21:09	21:10
Mean time of bedtime insulin sliding scale administration	21:08	21:31	21:35
Mean time of last dinner insulin administration	17:02	17:01	17:03
Patients receiving insulin <4 hours apart	42.60%	32.88% (p<0.001)	30.48% (p<0.001)
Patients receiving insulin >4 hours apart	57.40%	67.12% (p<0.001)	69.52% (p<0.001)
Outcome measures
Below target (less than 110 mg/dL)	4.28%	3.03% (p=0.027)	2.70 (p=0.004)
Overnight BG between 70 and 99 mg/dL	2.7%	1.99% (p=0.119)	1.22% (p<0.001)
Overnight hypoglycaemia BG <70 mg/dL	0.56%	0.54% (p=0.0941)	0.26% (p=0.120)

P value presents a significant difference between current period and pre-intervention period.

Conclusion

We demonstrated that careful evaluation of long-standing health system practices can reveal problematic processes, which once intervened on can result in clinically meaningful impact. The change in the timing of insulin orders did not eliminate all insulin stacking. This has informed our next steps to collaborate with frontline nursing staff to better understand how dosing insulin greater than 4 hours apart may impact their workflows. Widespread nursing education regarding the intervention is important because the order in the EHR might change, but their workflow may not. This will guide additional strategies to reduce iatrogenic low blood sugars in the hospital related to insulin stacking.