Insulin Pump Alarms During Adverse Events: A Qualitative Descriptive Study

Abstract

Objective:

Explore alarm signals cited in insulin pump-associated adverse events (AEs), describe the clinical consequences and other root cause informing remarks that cooccurred with the alarm signals, and identify opportunities for improvements to patient education, instructional materials, and alarm systems to prevent future AEs.

Research Design and Methods:

We explored the type, frequency, and associated clinical consequences of alarm signals cited in a pre-coded data set of 2294 insulin pump-associated AEs involving the MiniMed 670G, MiniMed 630G, and t:slim X2. We also explored the clinical consequences and other root cause informing remarks that cooccurred with the top 10 most frequently cited alarm signals.

Results:

Overall, 403 AEs narratives cited at least one alarm signal. Of the 40 unique alarm signals cited, 42.5% were “alarms,” 25.0% were “alerts,” and 32.5% were not referenced in the instructional materials packaged with the corresponding pump. The top 10 most frequently cited alarm signals included two obstruction of flow alarms, which accounted for 49.9% of all AEs citing at least one alarm, and two unreferenced alarms. The most frequent cooccurring root cause informing remark varied across the top 10 alarm signals and revealed valuable insight into why these alarms may have occurred.

Conclusions:

Our findings demonstrate the value of analyzing alarm signals cited in insulin pump-associated AEs and reveal multiple opportunities for providers to educate patients on how to respond to alarm signals and manage their pumps to avoid AEs, and for insulin pump manufacturers to update instructional materials and improve alarm systems to support appropriate patient response.

Introduction

Insulin pumps, which can function as stand-alone devices or be paired with a continuous glucose monitor (CGM) as a sensor-augmented or closed-loop system, can help patients with diabetes improve hemoglobin A1c and reduce frequency of hypoglycemia. ¹ However, management of diabetes with an insulin pump requires constant vigilance and frequent user interaction with a complicated medical device. Software failures, hardware malfunctions, use errors, and situations that arise during routine insulin pump use (eg, infusion site disconnection during sleep) have the potential to expose patients on insulin pumps to life-threatening consequences, such as severe hyperglycemia, hypoglycemia, and diabetic ketoacidosis (DKA) if not recognized and addressed properly.^2-4 An effective insulin pump alarm system is critical to ensuring that patients who use insulin pumps are aware of these issues in a timely and actionable manner.

The purpose of an insulin pump alarm system is to safeguard patients by: (1) detecting changes in the patient’s health status (eg, low blood glucose [BG]) or other circumstances that could adversely affect the patient (eg, loss of device power) and (2) warning patients through an alarm, alert or notification (hereafter referred to as an alarm signal).^5,6 To be successful, the insulin pump alarm system must provide reliable and valid alarm condition detection and ensure the timeliness and perceptibility of the alarm signal, ⁷ and the patient must understand and take appropriate action. Ineffective alarm systems or patient response can result in adverse events (AEs).

Our prior review of a sample of 2429 insulin pump-associated AEs narratives found that 16.7% cited at least one alarm signal that occurred at the time of the event. ⁴ Identifying alarm signals that most frequently occur in insulin pump-associated AEs, along with any system and/or human performance breakdown that may have prevented the alarm system from achieving its safety purpose, can help prevent future AEs. Therefore, the aims of this qualitative analysis were to explore the type and frequency of alarm signals cited in insulin pump-associated AE narratives, describe the clinical consequences and other root cause informing remarks that cooccurred with the alarm signals during AEs, and identify opportunities for improvements to patient education, instructional materials, and alarm system design to prevent future AEs.

Materials and Methods

We leveraged our pre-coded insulin pump-associated AE data set for this analysis. This data set included insulin pump-associated AE narratives extracted from the Food and Drug Administration (FDA) Manufacturer and User Facility Device Experience (MAUDE) database that: (1) occurred between January 1 and June 30, 2020, (2) were classified as an injury or death, and (3) involved five insulin pump models marketed in the United States at that time: the MiniMed 670G Insulin Pump and MiniMed 630G Insulin Pump (Medtronic, Minneapolis, Minnesota), Omnipod Insulin Management System and Omnipod DASH Insulin Management System (Insulet Corp., Action Massachusetts), and t: slim X2 Insulin Delivery System (Tandem Diabetes Care, Inc., San Diego, California). In this data set, we applied qualitative template analysis to identify clinical consequences and potential root causes of the AEs. We provide a detailed description of the sampling and coding methods used to create this data set elsewhere. ⁴

For this analysis, we explored the type, frequency, and associated clinical consequences of alarm signals cited in the 2294 insulin pump-associated AE narratives involving the MiniMed 670G, MiniMed 630G, or t:slim X2. We excluded the insulin pump-associated AEs involving the Omnipod and Omnipod DASH because their small sample sizes in the pre-coded data set yielded an inadequate volume of alarms for this analysis. We also explored the clinical consequences and other root cause informing remarks that cooccurred with the top 10 most frequently cited alarm signals. For this analysis, we used the following criteria to define key themes within each alarm signal: (1) most cited root cause informing remark across all categories and (2) ≥10% within-alarm signal frequency. We reviewed the most up-to-date insulin pump instructional materials for each pump model to support our analysis.^8-15

Results

Of the 2294 insulin pump-associated AE narratives reviewed in this analysis, 56.6% involved the MiniMed 670G, 22.5% involved the t:slim X2, and 20.9% involved the MiniMed 630G. Overall, 403 (17.6%) AEs cited at least one alarm signal. The largest proportion of reported alarm events involved the MiniMed 670G (58.3%), followed by the t:slim X2 (27.5%) and the MiniMed 630G (14.1%).

Clinical Consequences Mentioned in AEs Citing an Alarm Signal

Critical hyperglycemia, that is, BG > 400 mg/dL, was the most frequently mentioned clinical consequence (62.5%) in AE narratives citing at least one alarm signal, followed by critical hypoglycemia (ie, BG < 54 mg/dL; 14.6%). Furthermore, 10.2% of AE narratives citing at least one alarm signal also mentioned that the patient sought urgent medical care because of the event (ie, health care utilization; Table 1).

Table 1.

Frequency of Clinical Consequences Mentioned in Adverse Event Narratives Citing at Least One Alarm Signal (n = 403).

Clinical consequences	n	% of alarm signal AEs
BG value a	366	90.8
< 54 mg/dL	59	14.6
54-69 mg/dL	3	0.7
181-400 mg/dL	47	11.7
> 400 mg/dL	252	62.5
Both hypoglycemia and hyperglycemia	5	1.2
Health care utilization b	41	10.2
Hospital/ICU admissions	28	6.9
ER or hospital visit	7	1.7
Paramedics responded	6	1.5
Health outcomes	26	6.5
DKA	21	5.2
Skin inflammation, irritation, or infection	0	—
Loss of consciousness	5	1.2
Death	0	—
Seizure	0	—
Device-related outcomes	31	7.7
Reverted to manual injections or alternate pump	21	5.2
Needed a replacement pump	10	2.5
Secondary events	0	—
No specific consequences cited	24	6.0

Abbreviations: AE, adverse event; BG, blood glucose; DKA, diabetic ketoacidosis; ICU, intensive care unit.

^aBG value category consists of mutually exclusive subcodes based on the reported BG value at the time of the event.

^bHealth care utilization category consists of mutually exclusive subcodes based on the highest level of health care received.

Unique Alarm Signals Cited in AE Narratives

We found 40 unique alarm signals cited in the AE narratives reviewed (Table 2). According to the insulin pump user guides, alarm signals cited in insulin pump-associated AEs fell into two categories—“alarms” and “alerts”—with alarms described as the higher priority signals.^8,10,12-15 Of the unique alarms signals cited, insulin pump manufacturers categorized 42.5% as “alarms” and 25.0% as “alerts,” with three “alerts” appearing in top 10 most frequently cited alarm signals. Overall, “alarms” appeared in 77.2% and “alerts” appeared in 24.1% of the 403 insulin pump-associated AE narratives citing at least one alarm signal.

Table 2.

Frequency and Manufacturer Classification of Unique Alarm Signals Cited in Insulin Pump-Associated Adverse Events and Their Applicability to Specific Pump Models.

Alarm signals (n = 403) a	Manufacturer classification b	MiniMed 670G	MiniMed 630G	t:slim X2	n	% of all alarms
Insulin flow blocked alarm	Alarm	x	x		134	33.3
Occlusion alarm	Alarm			x	67	16.6
Change sensor alert	Alert	x	x		36	8.9
Calibration not accepted alert	Alert	x	x		34	8.4
Insulin pump error alarm	Alarm	x	x		28	6.9
Critical pump error alarm	Alarm	x	x		27	6.7
Hardware low-level failure alarm	Unreferenced	x	x		17	4.2
Malfunction alarm	Alarm			x	15	3.7
Sensor updating alert	Alert	x	x		12	3.0
Minimum fill notification	Unreferenced c			x	8	2.0
Stuck button alarm	Alarm	x	x		6	1.5
Cartridge alarm	Alarm			x	6	1.5
Loss of communication	Unreferenced	x			6	1.5
Altitude alarm	Alarm			x	5	1.2
Software error detected alarm	Unreferenced	x			5	1.2
Auto-off alarm	Alarm			x	5	1.2
CGM error 42	Unreferenced d			x	5	1.2
Failed battery alarm	Alarm	x	x		5	1.2
Power loss alarm	Alarm	x	x		5	1.2
No delivery alarm	Unreferenced	x	x		5	1.2
Post-reset RAM CRC alarm	Unreferenced	x	x		3	0.7
Cannot find sensor signal alert	Alert	x	x		3	0.7
Lost sensor signal alert	Alert	x	x		3	0.7
Low battery alert	Alert	x			3	0.7
Replace battery now alarm	Alarm	x	x		3	0.7
Bolus not delivered alarm	Alert e	x	x		3	0.7
Broken force sensor alarm	Unreferenced	x			1	0.2
Button alarm	Alarm			x	1	0.2
Mechanical error	Unreferenced	x			1	0.2
Temperature alarm	Alarm			x	1	0.2
Sensor signal not found alert	Alert	x	x		1	0.2
Threshold suspend alarm	Unreferenced	x			1	0.2
Power error detected alarm	Alarm	x	x		1	0.2
Power source alert	Alert			x	1	0.2
Replace battery alert	Alert	x	x		1	0.2
Empty cartridge alarm	Alarm			x	1	0.2
No reservoir detected alarm	Alarm	x	x		1	0.2
Delivery stopped alarm	Unreferenced	x			1	0.2
Over delivery alarm	Unreferenced	x			1	0.2
Under delivery alarm	Unreferenced	x			1	0.2

Abbreviations: AE, adverse event; CGM, continuous glucose monitor.

^aWe cite the language most frequently used to describe the alarm signal in the AE narratives.

^bWe only designate an alarm signal as “unreferenced” when we could not confidently link the reported alarm signal to an alarm or alert described in the instructional material packaged with the corresponding pump.

^cMinimum fill notification was not described in the instructional material packaged with the pump but found on the manufacturer’s product support website.

^dAlthough two of the four t:slim X2 user guides describe a “CGM error,” none mentioned a “CGM error 42.” For this analysis, we categorized this alarm signal as “unreferenced.”

^eAlthough AE narratives referred to this as an “alarm,” chapter 13 of the MiniMed 670G and MiniMed 630G user guides entitled “Alarms, alerts, and messages” refers to it as an “alert.” For this analysis, we also categorized this alarm signal as an “alert.”

Of the unique alarm signals, 32.5% were not listed in the instructional materials of corresponding pump, and seven of these unreferenced alarms were cited in more than one AE narrative. Furthermore, two unreferenced alarms appeared in the top 10 most frequently cited alarm signals—hardware low-level failure alarm and minimum fill notification. Overall, unreferenced alarms appeared in 13.6% of the 403 insulin pump-associated AE narratives citing at least one alarm signal.

Top 10 Most Frequently Cited Alarm Signals

Tables 3 shows the clinical consequences and Tables 4–6 show the other root cause informing remarks that cooccurred in AE narratives that cited the top 10 alarm signals. In the sections below, we organize the key results from all four tables by each alarm signal.

Table 3.

Cooccurrence of Critical Hyperglycemia, Critical Hypoglycemia, and Hospital/Intensive Care Unit Admissions Within Adverse Events Citing the Top Ten Alarm Signals.

Top 10 alarm signals	Critical BG outcomes				Hospital/ICU admissions
	< 54 mg/dL		> 400 mg/dL
	n	% within alarm	n	% within alarm	n	% within alarm
Insulin flow blocked alarm (n = 134)	4	3.0	101	75.4	9	6.7
Occlusion alarm (n = 67)	0	—	44	65.7	6	9.0
Change sensor alert (n = 36)	14	38.9	19	52.8	0	—
Calibration not accepted alert (n = 34)	14	41.2	18	52.9	1	2.9
Insulin pump error alarm (n = 28)	5	17.9	16	57.1	2	7.1
Critical pump error alarm (n = 27)	4	14.8	20	74.1	1	3.7
Hardware low-level failure alarm (n = 17)	3	17.6	13	76.5	4	23.5
Malfunction alarm (n = 15)	3	20.0	6	40.0	0	—
Sensor updating alert (n = 12)	7	58.3	3	25.0	0	—
Minimum fill notification (n = 8)	2	25.0	3	37.5	1	12.5

Abbreviations: BG, blood glucose; ICU, intensive care unit.

Table 4.

Cooccurrence of a Device Component Issue Mentioned Within Adverse Events Citing the Top 10 Alarm Signals.

Top 10 alarm signals	Pump or pod reservoir or cartridge		Infusion set or site		CGM		Battery or power		Screen or display		Internal electronic components		Alarms or alerts		Buttons or keypad
	n	% within alarm a	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm
Insulin flow blocked alarm (n = 134)	5	3.7	16	11.9	1	0.7	2	1.5	0	—	1	0.7	1	0.7	3	2.2
Occlusion alarm (n = 67)	1	1.5	0	—	2	3.0	1	1.5	0	—	0	—	2	3.0	0	—
Change sensor alert (n = 36)	0	—	1	2.8	3	8.3	1	2.8	0	—	0	—	0	—	0	—
Calibration not accepted alert (n = 34)	0	—	1	2.9	3	8.8	1	2.9	0	—	0	—	0	—	0	—
Insulin pump error alarm (n = 28)	2	7.1	1	3.6	0	—	1	3.6	1	3.6	3	10.7	0	—	0	—
Critical pump error alarm (n = 27)	3	11.1	1	3.7	0	—	9	33.3	1	3.7	15	55.6	0	—	0	—
Hardware low-level failure alarm (n = 17)	2	11.8	0	—	0	—	1	5.9	0	—	0	—	0	—	10	58.8
Malfunction alarm (n = 15)	0	—	0	—	0	—	1	6.7	0	—	0	—	0	—	0	—
Sensor updating alert (n = 12)	0	—	0	—	0	—	0	—	0	—	0	—	0	—	0	—
Minimum fill notification (n = 8)	0	—	0	—	0	—	0	—	0	—	0	—	0	—	0	—

Abbreviation: CGM, continuous glucose monitor.

^aPercentages highlighted in gray designate the most cited device component issue within each alarm that also met the ≥ 10% within-alarm threshold.

Table 5.

Cooccurrence of an Insulin Pump Task Mentioned Within Adverse Events Citing the Top 10 Alarm Signals.

Top 10 alarm signals	Bolus programming		Changing the pod/infusion set, including reservoir/cartridge		Inserting the sensor		Changing or charging the battery		Preparing for imaging
	n	% within alarm a	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm
Insulin flow blocked alarm (n = 134)	0	—	3	2.2	0	—	0	—	1	0.7
Occlusion alarm (n = 67)	1	1.5	6	9.0	0	—	1	1.5	0	—
Change sensor alert (n = 36)	0	—	0	—	0	—	0	—	0	—
Calibration not accepted alert (n = 34)	0	—	0	—	1	2.9	0	—	1	2.9
Insulin pump error alarm (n = 28)	0	—	0	—	0	—	0	—	0	—
Critical pump error alarm (n = 27)	0	—	0	—	0	—	0	—	0	—
Hardware low-level failure alarm (n = 17)	0	—	0	—	0	—	0	—	0	—
Malfunction alarm (n = 15)	0	—	0	—	0	—	1	6.7	0	—
Sensor updating alert (n = 12)	0	—	0	—	0	—	0	—	0	—
Minimum fill notification (n = 8)	0	—	8	100.0	0	—	0	—	0	—

^aPercentages highlighted in gray designate the most cited pump task within each alarm that also met the ≥ 10% within-alarm threshold.

Table 6.

Cooccurrence of a Use Error Mentioned Within Adverse Events Citing the Top 10 Alarms, Alerts, and Notifications.

Top 10 alarm signals	Did not remove air bubbles from tubing or reservoir/cartridge		Did not disconnect infusion set from body		Did not remove the pump		Used incompatible insulin		Data entry error		Did not change supplies or extended supply use
	n	% within alarm a	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm	n	% within alarm
Insulin flow blocked alarm (n = 134)	0	—	0	—	1	0.7	0	—	0	—	1	0.7
Occlusion alarm (n = 67)	1	1.5	0	—	0	—	7	10.4	1	1.5	3	4.5
Change sensor alert (n = 36)	0	—	0	—	0	—	0	—	0	—	0	—
Calibration not accepted alert (n = 34)	0	—	0	—	1	2.9	0	—	0	—	0	—
Insulin pump error alarm (n = 28)	0	—	0	—	0	—	0	—	0	—	0	—
Critical pump error alarm (n = 27)	0	—	0	—	0	—	0	—	0	—	0	—
Hardware low-level failure alarm (n = 17)	0	—	0	—	0	—	0	—	0	—	0	—
Malfunction alarm (n = 15)	0	—	0	—	0	—	0	—	0	—	0	—
Sensor updating alert (n = 12)	0	—	0	—	0	—	0	—	0	—	0	—
Minimum fill notification (n = 8)	0	—	1	12.5	0	—	0	—	0	—	0	—

^aPercentages highlighted in gray designate the most cited use error within each alarm that also met the ≥ 10% within-alarm threshold.

Insulin flow blocked alarm (MiniMed 670G and 630G)

Critical hyperglycemia was mentioned in 75.4% of AE narratives citing an insulin flow blocked alarm (Table 3). An issue with the infusion set or site emerged as the most frequent cooccurring root cause informing remark in AEs citing an insulin flow blocked alarm (11.9%; Tables 4–6), with 81.3% of these AE narratives specifically mentioning a bent cannula (13/16). On the other hand, an insulin flow blocked alarm only cooccurred in 14.0% of all AEs that mentioned a bent cannula (13/93).

Occlusion alarm (t:slim X2)

Critical hyperglycemia was mentioned in 65.7% of AE narratives citing an occlusion alarm (Table 3). The error of using incompatible insulin emerged as the most frequent cooccurring root cause informing remark in AEs citing an occlusion alarm (10.4%; Tables 4–6), with Admelog, Apidra, Fiasp, and residual insulin from previously used cartridges mentioned as the incompatible insulins.

Change sensor alert (MiniMed 670G and 630G)

In AE narratives citing a change sensor alert, 52.8% mentioned critical hyperglycemia and 38.9% mentioned critical hypoglycemia (Table 3). No root cause informing remarks cooccurred in ≥ 10% of AEs citing a change sensor alert (Tables 4-6).

Calibration not accepted alert (MiniMed 670G and 630G)

In AE narratives citing a calibration not accepted alert, 52.9% mentioned critical hyperglycemia and 41.2% mentioned critical hypoglycemia (Table 3). No root cause informing remarks cooccurred in ≥ 10% of AEs citing a calibration not accepted alert (Tables 4-6).

Insulin pump error alarm (MiniMed 670G and 630G)

Critical hyperglycemia was mentioned in 57.1% of AE narratives citing an insulin pump error alarm (Table 3). An issue with the internal electronic components emerged as the most frequent cooccurring root cause informing remark in AEs citing an insulin pump error alarm (10.7%; Tables 4–6).

Critical pump error alarm (MiniMed 670G and 630G)

Critical hyperglycemia was mentioned in 74.1% of AE narratives citing a critical pump error alarm (Table 3). An issue with the internal electronic components emerged as the most frequent cooccurring root cause informing remark in AEs citing a critical pump error alarm (55.6%; Tables 4–6), with 80.0% of these AE narratives specifically mentioning moisture damage or corrosion found on the electronic, motor, and/or battery tube assembly (12/15).

Hardware low-level failure alarm (MiniMed 670G and 630G)

Critical hyperglycemia was mentioned in 76.5% of AE narratives citing a hardware low-level failure alarm, and 23.5% of AEs citing this alarm signal resulted in a hospital/intensive care unit (ICU) admission (Table 3). An issue with the buttons or keypad emerged as the most frequent cooccurring root cause informing remark in AEs citing this alarm signal (58.8%; Table 4), with all AE narratives mentioning a “broken trace on the keypad assembly.” Content of AE narratives revealed that a hardware low-level failure alarm is also referred to as a “pump error code 63,” and this alarm can occur after running a self-test on the pump. Although the MiniMed 670G and 630G user manuals describe the self-test function, the hardware low-level failure alarm is not mentioned.

Malfunction alarm (t:slim X2)

In AE narratives citing a malfunction alarm, 40.0% mentioned critical hyperglycemia and 20.0% mentioned critical hypoglycemia (Table 3). No root cause informing remarks cooccurred in ≥ 10% of AEs citing a malfunction alarm (Tables 4-6).

Sensor updating alert (MiniMed 670G and 630G)

Critical hypoglycemia was mentioned in 58.3% of AE narratives citing a sensor updating alert (Table 3). No root cause informing remarks cooccurred in ≥ 10% of AEs citing a sensor updating alert (Tables 4-6).

Minimum fill notification (t:slim X2)

In AE narratives citing a minimum fill notification, 37.5% mentioned critical hyperglycemia and 25.0% mentioned critical hypoglycemia. Furthermore, 12.5% of AEs citing this alarm signal resulted in a hospital/ICU admission (Table 3). All AE narratives citing a minimum fill notification described the patient engaged in the task of changing the infusion set, including cartridge (Table 5). Most patients (87.5%) reported filling their cartridge with 200 to 300 units of insulin, which exceeds the minimum amount of “95 units” required according to the t:slim X2 user guides.^12-15 Although minimum fill notification is not referenced in the t:slim X2 user guides, we were able to find a description of this notification and how to address it through a search of Tandem’s t:slim X2 product support website. ¹⁶

Discussion

In this analysis exploring the type and frequency of alarm signals, associated clinical consequences, and cooccurring root cause informing remarks cited in insulin pump-associated AE narratives, we found that 17.6% of all AEs reviewed cited an alarm signal with 40 unique alarm signals mentioned. Among these AEs, critical hyperglycemia (62.5%) and critical hypoglycemia (14.6%) were the most reported clinical consequences, and 10.2% of patients reported the need for urgent medical care because of the AE. The frequency of obstruction of flow alarms cited in insulin pump-associated AE narratives, including both insulin flow blocked alarms (MiniMed 670G and 630G) and occlusion alarms (t:slim X2), was notably higher than any other alarm signal reported.

Our analysis identified key information that diabetes care providers can use to help patients prevent future insulin pump-associated AEs. For the t:slim X2 insulin pump, we found that occlusion alarms most commonly cooccurred with the use of incompatible insulins in AE narratives. Providers can reinforce the potential risks associated with using incompatible insulin and ensure that patients who use the t:slim X2 are only prescribed Novolog or Humalog U-100 insulins. In addition, we found that 12.5% of t:slim X2 AEs that cited a minimum fill notification resulted in a hospital/ICU admission. Providers can emphasize the need to include between 95 and 300 units of insulin when filling the cartridge, explain the potential consequences of filling it with too much or too little insulin, and point patients to the Tandem website to learn more about the minimum fill notification, since it is not included in the t:slim X2 user guide. ¹⁶

For the MiniMed 670G and 630G insulin pumps, we found that insulin flow blocked alarms did not commonly cooccur with bent cannulas in AE narratives. Providers can emphasize to patients that a bent cannula may not trigger an insulin flow blocked alarm, so that, they should check their insertion site and change their infusion set if they experience persistent, unexplained hyperglycemic episodes, even in the absence of an insulin flow blocked alarm. In addition, we found three MiniMed 670G and MiniMed 630G CGM-related “alerts” in the top 10 most frequently cited alarm signals. Providers can advise patients that CGM “alerts” may require a more immediate response to prevent AEs. Finally, we found that critical pump error alarms often cooccur with the identification of moisture damage or corrosion found on the electronic, motor, and/or battery tube assembly of the MiniMed 670G and 630G. Providers can explain the association between moisture damage or corrosion of the internal electronic components and the occurrence of a critical pump error alarm and emphasize the importance of protecting the insulin pumps from excessive moisture.

Our analysis identified opportunities for insulin pump manufacturers to improve instructional materials pertaining to alarm signals. We found that 32.5% of the unique alarm signals cited in AE narratives were not mentioned in the insulin pump instructional materials packaged with the corresponding pump. Furthermore, two unreferenced alarms emerged in the top 10 most frequently cited alarm signals and had the highest within-alarm proportions of ICU/hospital admissions (23.5% and 12.5% respectively). Patients may not understand the severity of these alarm signals and/or how to troubleshoot the insulin pump when they occur. Opportunely, manufacturers can leverage our findings to revise user guides and other instructional materials to ensure the inclusion of all potential alarms signals and communicate the appropriate patient responses to prevent future AEs.

Our analysis also revealed multiple opportunities for insulin pump manufacturers to improve alarm systems. First, our findings support the existing calls to improve alarm systems in detecting occlusions and warning patients in sufficient time to prevent AEs.^7,17,18 For example, we found that only a small proportion of the AEs involving the MiniMed 670G and 630G pumps that cited a bent cannula also mentioned an insulin flow blocked alarm (14%). A bent cannula prevents proper insulin delivery and may result in severe clinical consequences, such as DKA; thus, insulin pump manufacturers should investigate ways to improve alarm systems in detecting this type of occlusion to prevent related AEs.^19,20

Second, there is a need to optimize insulin pump “alerts” as the frequency of “alerts” cited in insulin pump-associated AEs suggests that some “alerts” may require a more immediate patient response to prevent AEs. Patients may be perceiving “alerts” as less urgent and not responding appropriately, or they may be experiencing alarm fatigue due to an unnecessary number of unique “alerts” used to communicate similar information.^19,20 Manufacturers can conduct human factors evaluations to investigate patient perceptions and behavioral responses to existing insulin pump “alerts” to identify the most appropriate improvements, such as recategorizing some “alerts” as “alarms” and/or reducing the number of unique “alerts” presented to the patient.

Finally, our analysis revealed opportunities to improve the investigation and reporting of alarm signals that occur during insulin pump-associated AEs. First, there is a need for more accurate reporting of the alarm signal name. The 13 unreferenced alarm signals identified in our analysis may have been due to the reporter’s inability to recall the specific name of the alarm signal and/or their use of general terms to describe the alarm signal received. Manufacturers can leverage the pump history to find the specific alarm signal name. Second, there is a need for additional investigation into why the alarm occurred. The limited amount of information provided in AE narratives inhibits the ability to identify the root cause and most appropriate corrective action to address these AEs. Accurate reporting of alarm signals that occurred at the time of insulin pump-associated AE, along with the circumstances leading up to the alarm, would support future improvements to patient education, instructional materials, and alarm system design to prevent AEs.

Limitations

Our analysis was limited by the same challenges to analyzing AE data encountered in prior analyses, such as the lack of sufficient detail about the event and the inclusion of unverified circumstances and potentially biased information in AE narratives.^4,17,21 In addition, the findings of this analysis only pertain to three insulin pump models—MiniMed 670G, MiniMed 630G, t:slim X2. Although Medtronic discontinued the sale of the MiniMed 670G in Feb 2021, a review of AEs in MAUDE revealed 779 reported injuries involving the MiniMed 670G occurred in the first six months of 2023, including 127 AEs citing least one alarm signal (16.3%). This finding demonstrates that patients are still using the MiniMed 670G and therefore remain exposed to the same risks described in this study.

Our analysis was also limited by incomplete investigation and/or reporting of alarm signals in insulin pump-associated AE narratives. As a result, the actual frequency of alarm signals that occurred during insulin pump-associated AEs reported in MAUDE may be greater than described here. Furthermore, some unreferenced alarms cited in AE narratives may have corresponded to actual alarms on the insulin pump, but there was not enough information in the narrative to confidently make this assumption. Finally, the AE narratives lacked information as to why the alarm signals did not prevent the injury from occurring (eg, late alarm, user failing to act on the alarm, etc). Despite these limitations, our analysis demonstrates the value of exploring alarm signals cited in insulin pump-associated AE narratives in MAUDE to inform strategies to prevent future AE. Future research could investigate the type and frequency of alarm signals and cooccurring root cause informing remarks in AEs involving an injury or death that did not cite an alarm to explore trends in circumstances where alarm signals did not occur, and in AEs classified as “malfunctions” to explore trends in circumstances where alarm signals were successful in preventing injuries.

Conclusion

Findings from our qualitative descriptive analysis exploring the alarm signals cited in insulin pump-associated AEs revealed multiple opportunities for providers to educate patients on how to respond to insulin pump alarms to avoid serious clinical consequences, and for insulin pump manufacturers to update instructional materials and improve insulin pump alarm systems to support appropriate patient response. Our findings demonstrate the value of analyzing alarm signals cited in insulin pump-associated AEs reported in the FDA’s MAUDE database while supporting the need to improve the investigation and reporting of alarm signals to inform improvements to patient education, instructional materials, and alarm system design to prevent future AEs.