Management and Outcomes of Severe Hypoglycemia Treated by Emergency Medical Services in the U.S. Upper Midwest

Severe hypoglycemia is a serious but preventable adverse event in the management of diabetes (1). Severe, or level 3, hypoglycemia is defined as low blood glucose requiring assistance of another person for management and symptom recovery after treatment (2). It is associated with increased risk of cardiovascular events (3–7), mortality (3,5–13), decreased quality of life (14), disability (15), dementia (16–18), and high health care costs (15). Rates of severe hypoglycemia in real-world practice vary widely depending on how hypoglycemia is defined and ascertained, with further heterogeneity as a function of patient age, clinical complexity, and glucose-lowering treatment regimen (19). While many observational studies of severe hypoglycemia have focused on episodes requiring emergency department (ED) or hospital care (20), this approach overlooks hypoglycemic events that happen outside of this environment, including those treated by emergency medical services (EMS) without transport to the ED, resulting in missed opportunities to identify and intervene upon high-risk patients.

Approximately 1% of all 911 calls in the U.S. are related to hypoglycemia (21–24). Because protocols for hypoglycemia management and transport guidelines vary among EMS agencies (25), the rates of ED transport after treatment for severe hypoglycemia in previously published studies has ranged between 28 and 87% (21,23,24,26–31). Most of these studies were conducted using data that are a decade old or older, many relied on the primary impression of hypoglycemia or diabetic emergency or on glucose values alone without restricting analyses to patients with diabetes (therefore including a more heterogeneous population with different causes of hypoglycemia), or were conducted outside of the U.S., where patient evaluation and transport decision processes are different. Importantly, most studies were restricted to data available to the EMS team and as a result could not examine patient factors associated with successful treatment on scene nor could they quantify the impact of ED transport on the likelihood of severe hypoglycemia recurrence.

To address these knowledge gaps, we use contemporary EMS data from a large advanced life support ambulance service across two Midwest states, linked with electronic health record (EHR) data of an integrated health care delivery system serving an overlapping geographic area, to examine prehospital management of severe hypoglycemia among patients with known diabetes, identify patient- and treatment-level factors associated with ED transport, and quantify the frequency of recurrent severe hypoglycemia requiring EMS, ED, or hospital care among transported patients as compared with those who were treated and left on scene. Better understanding of outcomes following episodes of severe hypoglycemia requiring EMS care can help clinicians identify and proactively intervene upon at-risk patients and aid EMS agencies in risk-stratifying patients for ED transport.

We conducted a retrospective cohort study using EHRs of Mayo Clinic Ambulance, Mayo Clinic Rochester, and Mayo Clinic Health System (MCHS) practices. This study was approved by the Mayo Clinic Institutional Review Board and is reported in accordance with STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines for observational cohort studies.

We used Mayo Clinic Ambulance records to identify 911 encounters for hypoglycemia among adults (aged ≥18 years) between 1 January 2013 and 31 December 2019. Encounters for hypoglycemia were characterized by administration of glucose, dextrose, or glucagon. Nonemergent requests for service (i.e., interfacility transports) were excluded. Each record was linked to the Mayo Clinic EHR using the patient’s name, sex, date of birth, and address. We excluded individuals without a diagnosis of diabetes in the Mayo Clinic EHR and those who did not provide authorization for their data to be used for research (Supplementary Fig. 1). Diabetes status was identified using Healthcare Effectiveness Data and Information Set criteria (32) applied to the year prior to the index date. Diabetes type was ascertained using claims and medications data from the year prior to the index date as previously described (33).

Mayo Clinic Ambulance is an advanced life support provider and the primary response, treatment, and transport service for 14 locations throughout Minnesota and Western Wisconsin, covering 6,894 square miles of urban, suburban, and rural areas. Mayo Clinic Ambulance is staffed by paramedics and emergency medical technicians and responds to ∼100,000 requests for service, including 75,000 911 calls, annually.

Mayo Clinic is an integrated health care delivery system serving local, regional, national, and international patients each year with a hub in Rochester, MN. Mayo Clinic Rochester primary care practices (internal medicine, geriatrics, family medicine, and pediatrics) care for Mayo Clinic employees, their dependents, and local area residents. MCHS is a network of community-based clinics, hospitals, and health care facilities serving >70 communities in Minnesota and Wisconsin, delivering both primary and specialty care.

The primary outcome was transport to the ED, ascertained using Mayo Clinic Ambulance records. Secondary outcomes were the composite of repeat EMS (from Mayo Clinic Ambulance records), ED (from Mayo Clinic EHR), or hospital (from Mayo Clinic EHR) utilization for hypoglycemia as well as the individual components of this composite indicator within 3 days, 7 days, and 30 days of the index hypoglycemia-related EMS encounter. For patients who were hospitalized after the index EMS encounter, the period of susceptibility to recurrent events began on the day of hospital discharge. Hypoglycemia-related ED visits and hospitalizations were identified using ICD-9 and ICD-10 diagnosis codes (Supplementary Table 1) in the primary/first position on the ED or hospital claim. ED visits that resulted in hospital admissions were analyzed as a hospital visit.

Mayo Clinic Ambulance uses patient care medical guidelines developed and maintained by Medical Direction that apply to all Mayo Clinic Ambulance sites. For hypoglycemia-related EMS encounters, there are four potential outcomes: 1) transport to the ED; 2) no treatment or transport needed, permitted if the patient and EMS professionals agree that transport is not needed and the patient has capacity to make health care decisions; 3) treat and release, permitted if all of the following criteria are met: the care provided addressed the medical need or condition, the patient’s condition is stable with no apparent risk of deterioration, the patient and their care providers agree that treatment and no transport is appropriate for this situation, and the patient has capacity to make health care decisions; and 4) refusal of treatment and/or transport, used in situations in which the patient refused treatment, transport, or both and has capacity to make health care decisions.

All-cause mortality within 3, 7, and 30 days of the index hypoglycemia-related EMS encounter was a secondary outcome. Mortality was ascertained using the date of death available in the Mayo Clinic EHR.

Patient demographics were ascertained from Mayo Clinic EHR at the time of the index EMS encounter and included patient age, sex, race/ethnicity (non-Hispanic White vs. non-White; as self-reported by patients during registration), marital status (married or partnered vs. not married/partnered, which includes single, divorced, widowed, and unknown), and rurality of their primary residence. Rural status was categorized as rural versus urban using rural-urban commuting area codes in which metropolitan areas were classified as urban, while micropolitan areas, small towns, and rural areas were classified as rural (34,35). Patients’ insurance status could not be reported due to institutional guidelines, but Mayo Clinic and MCHS practices accept both private and public insurance plans.

Using Mayo Clinic EHR data for the year preceding the index hypoglycemia-related EMS encounter, we captured the presence of diabetes complications (retinopathy, nephropathy, neuropathy, cerebrovascular disease, cardiovascular disease, and peripheral vascular disease) (36), prior episodes of severe hypoglycemia and hyperglycemia requiring ED or hospital care, and select chronic health conditions that may affect odds of ED transport (i.e., hypertension, heart failure, chronic pulmonary disease, cancer, depression, dementia, alcohol abuse, and drug abuse) (37–39). We also assessed whether the patient had one or more of any of the above comorbidities (not including diabetes or prior severe hypoglycemia or hyperglycemia). Diabetes type was ascertained as previously described (33).

Orders for glucagon and glucose-lowering medications that were active in the ambulatory EHR medication list at the time of the index EMS encounter were identified. Treatment regimens were categorized into mutually exclusive groups based on their probability of causing hypoglycemia: bolus insulin (with or without basal insulin and noninsulin therapies), basal insulin (with or without noninsulin therapies), sulfonylurea (with or without other noninsulin therapies), other glucose-lowering therapies (metformin, glucagon-like peptide 1 receptor agonist, sodium–glucose cotransporter 2 inhibitor, dipeptidyl peptidase 4 inhibitor, thiazolidinedione, meglitinide, α-glucoside inhibitor, and amylin analog), and no active medications.

During 6 months preceding the hypoglycemia-related EMS encounter, we examined whether the patient had any office visits with a primary care provider, endocrinologist, certified diabetes care and education specialist, and medication therapy management pharmacist; ED visits for any cause; and hospitalizations for any cause. Primary care and endocrinology appointments were ascertained using scheduling data, while certified diabetes care and education specialist and pharmacy appointments were identified using Current Procedural Terminology/Healthcare Common Procedure Coding System codes detailed in Supplementary Table 2.

Baseline glycemic control was measured using the most recent hemoglobin A_1c (HbA_1c) available in the EHR during the 2 years prior to the index hypoglycemia-related EMS encounter. Half (53.7%) of HbA_1c results were within 3 months of the index event, 18.8% were between 3 and 6 months prior, 11.4% were 6–12 months prior, and 5.8% were 12–24 months prior. HbA_1c values were missing for 10.3% of patients and included in the models as a separate “missing” category.

Mayo Clinic Ambulance records were used to ascertain characteristics of the index hypoglycemia-related encounter: first documented blood glucose level; treatment detail (oral glucose, dextrose, or glucagon); and time of the day that the encounter took place (daytime, 7 a.m. to 7 p.m. vs. nighttime, 7 p.m. to 7 a.m., corresponding to EMS shifts). A glucose value was deemed to be most reliable if it was the earliest documented in the narrative of the report. Call logs of 129 hypoglycemia-related EMS calls with an initial glucose value ≥70 mg/dL were manually reviewed to verify that the documented glucose level was the first available prior to treatment administration. In 90 of 129 calls, an alternative value was identified through manual review of the event log.

Baseline patient characteristics were descriptively summarized at the patient level (for patients with multiple encounters, characteristics at the time of the first encounter were assessed) and the EMS encounter level and compared at the encounter level between those who were transported and those who were not transported using the Kruskal-Wallis test for continuous variables and the χ² test for categorical variables. Similar methods were used to compare patient status and clinical management at the time of the hypoglycemia-related EMS encounter between the two transport groups. To assess for differences in transport rates over time, we used the Cochran-Armitage trend test. For all analyses, two-sided P value <0.05 was considered to be significant.

Multivariate hierarchical logistic regression models were used to examine factors associated with ED transport following hypoglycemia-related EMS encounters, clustering at the patient level. Models included patient baseline characteristics (age, sex, race/ethnicity, rurality, marital status, whether any comorbidities were present, history of severe hypoglycemia and hyperglycemia, availability of glucagon prescriptions, diabetes treatment regimen, HbA_1c level, and prior health care utilization) and data from the EMS encounter (glucose level, time of the encounter, and how the hypoglycemia was treated). All variables were ascertained at the encounter level except for patient sex, race, and diabetes type (for which the first encounter date was used). Results were reported in the form of odds ratios (ORs), 95% CIs, and P values.

Rates of recurrent episodes of severe hypoglycemia requiring medical attention were descriptively reported for the 3, 7, and 30 days following the index hypoglycemia-related EMS encounter and were further compared across transport groups using the χ² test. For the composite outcome of severe hypoglycemia requiring medical attention, we conducted subgroup analyses as a function of patient age, baseline history of severe hypoglycemia, and whether patients were transported to the ED and discharged or admitted to the hospital. Furthermore, sensitivity analyses were performed excluding patients who had died during the follow-up window of relevance after the index hypoglycemia-related EMS encounter (i.e., those who died within 30 days of the index hypoglycemia-related EMS encounter were excluded when calculating rates of recurrent episodes of severe hypoglycemia within 30 days). Multilevel logistic regression was used to examine factors associated with experiencing a recurrent episode of severe hypoglycemia. The model included baseline patient characteristics (as above) and an indicator for whether or not the patient was transported after the index hypoglycemia-related EMS encounter. Finally, all analyses were repeated for patients with type 1 and type 2 diabetes separately.

All data management and analyses were carried out using SAS 9.4 (SAS Institute, Cary, NC).

We identified 1,977 EMS encounters for hypoglycemia between January 2013 and December 2019 among 1,028 adults with diabetes. At the time of the first hypoglycemia-related EMS encounter, patients were, on average, 63.5 years old (SD 17.7), 55.2% were men, 87.4% were non-Hispanic White, 43.6% were married/partnered, and 42.4% lived in a rural area (Table 1). HbA_1c results were available for 89.6% of the cohort, with mean HbA_1c 7.8% (SD 1.8). The vast majority (91.2%) of patients had at least one comorbidity in addition to diabetes and 36.6% had experienced an ED visit and 34.1% experienced at least one hospitalization in the prior 6 months.

Patients with type 1 diabetes comprised 25.6% of the cohort (N = 263), but accounted for 43.1% of EMS calls (N = 852). Their baseline characteristics are detailed in Supplementary Table 3, while characteristics of patients with type 2 diabetes (N = 765) and the 1,125 EMS calls they made are detailed in Supplementary Table 4. Patients with type 1 diabetes, when compared with patients with type 2 diabetes, were generally younger, more often non-Hispanic White, and more often lived in urban areas. Mean HbA_1c was 8.2% (SD 1.9) among patients with type 1 diabetes and 7.6% (SD 1.7) among patients with type 2 diabetes.

The majority of patients who experienced a hypoglycemia-related EMS encounter were being treated with bolus insulin at the time of their event (70.2%), while 13.9% were treated with basal insulin regimens, 8.0% with sulfonylurea regimens, and 1.9% with other noninsulin/nonsulfonylurea drugs; there were no glucose-lowering medications documented in the EHR for 6.0% of patients (Table 1). Among patients with type 2 diabetes, specifically, 60.1% were treated with bolus insulin regimens, 18.6% with basal insulin regimens, 10.7% with sulfonylurea regimens, and 2.5% with other glucose-lowering drugs (Supplementary Table 4). Nineteen percent of patients had an active glucagon prescription (42.2% of patients with type 1 diabetes and 11.1% of patients with type 2 diabetes). Upon EMS arrival, blood glucose levels were <54 mg/dL in 75.8% of patients (Table 2). One-third were treated with oral glucose, while 74.2% received D50 and 7.5% received glucagon. These data are presented separately for patients with type 1 and type 2 diabetes in Supplementary Tables 5 and 6.

Patients were transported to the ED in 917 of 1,977 (46.4%) hypoglycemia-related encounters, and 412 patients were ultimately hospitalized (20.8% of all patients or 44.9% of transported patients), while 977 (49.4%) were treated and released on scene, 35 (1.8%) did not require treatment or transport, and 48 (2.4%) refused treatment and/or transport. Rates of transport increased between 2013 (45.3%) and 2017 (55.3), but then declined through 2019 (46.6%) (P = 0.03). Tables 1 and 2 compare patient and EMS encounter characteristics, respectively, between transported and nontransported patients. Transported patients were, generally, older (mean age 63.4 [17.1] vs. 58.7 [16.3] years; P < 0.001), non-White (13.9 vs. 9.4%; P = 0.002), urban residents (56.7 vs. 51.4%; P = 0.02), and with comorbidities (94.0 vs. 87.3%; P < 0.001). Patients with type 1 diabetes were transported less frequently than patients with type 2 diabetes (31.1 vs. 58.0%). Nearly all examined comorbidities were more prevalent in transported than nontransported patients. Fourteen percent of transported patients had a prior episode of severe hypoglycemia compared with 10.8% of nontransported patients. Transported patients had greater prevalence of prior ED visits (48.4 vs. 37.7%; P < 0.001) and hospitalizations (43.3 vs. 28.6%; P < 0.001). They were more often treated with glucose-lowering regimens that did not include bolus insulin, less frequently had active glucagon prescriptions (24.5 vs. 28.8%), and had higher blood glucose levels during the EMS encounter compared with those who were not transported.

In multivariate analysis (Table 3), the strongest predictors of transport to the ED were diabetes type (OR 0.44 [95% CI 0.31–0.62] for type 1 vs. type 2 diabetes), having had a prior ED visit (OR 1.38 [95% CI 1.03–1.85]), having a noninsulin treatment regimen, and having a glucose level >54 mg/dL. Patients who received glucagon from EMS were more likely to be transported (OR 2.51 [95% CI 1.57–4.02]). Among those with type 1 diabetes (Supplementary Table 7), the presence of comorbidities other than diabetes (OR 2.91 [95% CI 1.04–8.14]) was the only patient characteristic associated with ED transport. Other factors were related to the event itself; specifically, having a higher initial blood glucose level (70–100 mg/dL: OR 7.65 [95% CI 2.08–28.17] vs. <54 mg/dL or “low”) and treatment with glucagon (OR 3.77 [95% CI 1.89–7.54]). In contrast, among patients with type 2 diabetes (Supplementary Table 8), the odds of transport were associated with patient age (65–74 years: OR 3.46 [95% CI 1.72–6.95]; ≥75 years: OR 2.84 [95% CI 1.46–5.54] vs. 18–44 years), being unmarried (OR 1.63 [95% CI 1.10–2.43]), being treated with medications other than insulin, higher glucose level upon arrival (54–69 mg/dL: OR 2.45 [95% CI 1.51–3.98]; 70–100 mg/dL: OR 2.72 [95% CI 1.09–6.82] vs. <54 mg/dL or “low”), and the event occurring during the day as compared with at night (OR 1.51 [95% CI 1.11–2.04]).

Recurrent severe hypoglycemia was common, particularly among patients who were not transported to the ED (Fig. 1). Within 3, 7, and 30 days of the index encounter, transported patients experienced recurrent severe hypoglycemia requiring medical attention 2.8, 5.2, and 10.6% of the time, respectively, compared with 7.4, 11.2, and 22.8% of the time among patients who had not been transported; P < 0.001 for each. Rates of recurrent EMS, ED, and hospitalization events among patients with type 1 and type 2 diabetes are detailed in Supplementary Tables 9–10. In multivariable analysis, patients who were transported had an OR of 0.58 (95% CI 0.42–0.80) for recurrent severe hypoglycemia within 30 days (Supplementary Table 11). In subgroup analyses among patients with type 1 diabetes (Supplementary Table 12) and type 2 diabetes (Supplementary Table 13), the odds of recurrent severe hypoglycemia among patients who were transported, as compared with those who were not transported, were 0.64 (95% CI 0.42–0.96) and 0.42 (95% CI 0.24–0.75), respectively.

We further examined the rates of recurrent severe hypoglycemia (composite of EMS, ED, and hospitalization events) in subgroups of patient age (Supplementary Table 14) and prior history of severe hypoglycemia (Supplementary Table 15), in which transported patients were also less likely to experience recurrent severe hypoglycemia. Finally, we compared rates of recurrent severe hypoglycemia as a function of whether patients were left on scene, transported to the ED but discharged, or ultimately admitted to the hospital (Fig. 1 and Supplementary Table 16). Rates of recurrence decreased significantly with each higher level of care received.

All-cause mortality was high among patients experiencing severe hypoglycemia treated by EMS, particularly among those who required transport (Supplementary Table 17). Within 30 days of the index EMS call, 58 transported (6.3%) and 6 nontransported (0.6%) patients died. Excluding patients who had died from analyses of recurrent severe hypoglycemia did not alter those results (Supplementary Table 18).

When responding to a call for severe hypoglycemia, EMS providers face the decision of whether to transport patients to the ED, as this may provide for opportunities to engage patients, identify reasons for the hypoglycemic event, deliver education, and modify treatment regimens to reduce the likelihood of hypoglycemia recurrence. However, ED visits are costly to the patient and the health care system and as such should be limited to situations in which higher levels of care are truly necessary and beneficial to the patient. There is substantial variability in the published transport rates of patients treated for severe hypoglycemia (21,23,24,26–31), calling for the examination of transport practices and their implications on recurrent severe hypoglycemic events. Our ability to leverage and link EMS and clinic EHR systems allowed us to identify patients with severe hypoglycemia requiring antihypoglycemic therapy (rather than the primary impression of hypoglycemia, which has been shown to lack both sensitivity and specificity [23]), focus our analyses on patients with known diabetes, contextualize events with patient-level clinical information, and examine longitudinal severe hypoglycemia recurrence.

In our analysis of 1,977 EMS encounters for hypoglycemia between January 2013 and December 2019 made by 1,028 adults with diabetes, 46.4% were transported to the ED. This transport rate is lower than observed in recent U.S.-based studies, in which the majority of patients were ultimately transported (21,23,24,26). In both Minnesota and Wisconsin, there are no standard statewide patient care guidelines, such that transport decisions are made by the Medical Direction of each EMS agency. The Mayo Clinic Ambulance hypoglycemia patient care guideline allows for patients to be treated and not transported if the patient’s hypoglycemia is successfully treated, the patient is stable with no apparent risk of deterioration, the patient and their care providers agree that no transport is appropriate, and the patient has capacity to make health care decisions. When Rostykus et al. (25) examined the guidelines of 185 ambulance services, they found only 49 allowed for nontransport following the correction of blood glucose. Additionally, the Centers for Medicare & Medicaid Services reimbursement incentivizes transport over nontransport. This practice may change with the Emergency Triage, Treat, and Transport payment model, which has been implemented in select ambulance agencies across the U.S. over the past year, though it was not in place at Mayo Clinic Ambulance when this study was conducted.

Factors associated with greater likelihood of transport to the ED were not surprising. Patients with type 1 diabetes were less likely to be transported than patients with type 2 diabetes, as were patients treated with medications known to cause hypoglycemia, while patients with prior history of ED utilization were more likely to be transported. This suggests that patients with unexplained causes for hypoglycemia are more likely to be transported for further evaluation and management, while patients in whom the hypoglycemic event is “expected” may choose to stay home. This decision is therefore likely influenced by multiple factors, including the perception of transport not adding value to their treatment or health as well as the desire to avoid the costs associated with EMS transport and an ED visit. Younger patients, who are more often uninsured or underinsured than older patients, may therefore decline transport more often.

When patients with type 1 and type 2 diabetes were examined separately, additional informative patterns emerged. Among patients with type 1 diabetes, the presence of comorbidities other than diabetes and having higher (70–100 mg/dL) glucose levels upon EMS arrival were the only patient characteristics associated with ED transport, increasing the odds by ∼3- and 7-fold, respectively. This again suggests that otherwise healthy patients with type 1 diabetes, among whom severe hypoglycemia is likely to be the result of mismatches among administered insulin, carbohydrate intake, and physical activity, may not see the benefit of going to the ED once their hypoglycemia is adequately treated on scene. Patients who call EMS at a higher glucose level and those who have underlying comorbidities may seek ED care for reasons beyond the correction of hypoglycemia, such as management of the factors that precipitated the hypoglycemic event in the first place.

In contrast, among patients with type 2 diabetes, the odds of transport were higher with increasing patient age, being unmarried, being treated with medications other than insulin, as well as higher glucose levels upon arrival. These additional factors may reflect the patients’ and EMS providers’ concerns about frailty, ability to self-manage, or underlying causes of the hypoglycemic event that cannot be readily reversed on scene. Hypoglycemia in older patients and patients with multiple comorbidities is more likely to be caused by factors other than glucose-lowering therapy. They are more likely to experience infectious and ischemic events, be ill, or experience prolonged inadequate oral intake, all of which can contribute to hypoglycemia that is not readily reversed with one-time treatment and elicit concern among EMS providers, patients, and caregivers. The higher rate of transport among single patients, who are less likely to have a caregiver available in the home, underscores the importance of caregiver support for ensuring patient safety if left on scene.

We were reassured to find no significant association between patients’ race or rural residence with the odds of ED transport in any of the multivariable analyses (in the overall population or when subset by diabetes type), suggesting that the decision to transport patients to the ED does not appear to be influenced by these characteristics. Nevertheless, non-White patients made up a higher proportion of the transported (13.9%) than nontransported (9.4%) group. This may be driven by less access to diabetes care, greater reliance on the ED as the primary site of medical care, or lack of adequate diabetes self-management education to empower patients to modify their own treatment regimens or have a clinician to reach out to for guidance in the event of a severe hypoglycemic event. Such lack of resources among racial and ethnic minority groups has been noted in prior research and underscores the pervasive impacts of structural racism in the health care system. EMS clinicians may also advise transport to patients they believe not have adequate follow-up or ability to self-manage, and the higher rate of transport among non-White patients may reflect this bias. Rural patients also made up a higher proportion of the nontransported (48.6%) than transported (43.3%) group, suggesting that patients living in rural areas are more likely to call EMS to help treat their hypoglycemic episode, perhaps because they lack easy access to care elsewhere, but then may decline transport to an ED that is potentially far from their residence.

Patients with an active glucagon prescription were transported to the ED less often than patients without one (24.5 vs. 28.8%), which is not surprising, as the goal of prescribing glucagon to at-risk patients is to allow their caregivers to treat severe hypoglycemia. However, in multivariate analysis, the availability of a glucagon prescription was not independently associated with reduced risk of ED transport. It is likely that this variable is confounded by other factors, including prior history of severe hypoglycemia that had prompted the patient’s health care provider to prescribe glucagon in the first place. Rates of glucagon prescribing were much higher in our population than in recently reported national studies of patients experiencing severe hypoglycemia (40,41) (42.2% of patients with type 1 diabetes and 11.1% of patients with type 2 diabetes), which is reassuring considering that all patients in our study had experienced a severe hypoglycemic event and were presumably at high risk for such events. Prospective studies will be needed to examine the impact of glucagon availability on rates of severe hypoglycemic events requiring EMS, ED, and hospital-level care.

Importantly, transported patients were half as likely to experience recurrent severe hypoglycemia. The higher rates of recurrent hypoglycemia among nontransported patients were apparent in patients with both type 1 and type 2 diabetes, though the magnitude of effect was stronger in patients with type 2 than type 1 diabetes (OR 0.42 vs. 0.64). The ED encounter may provide an opportunity for treatment modification both directly, during the ED visit, and indirectly, because in our health system, primary care providers are automatically notified when their patients are seen in the ED, prompting them to follow up with their patients after such events occur. Such treatment modification, whether change in the medication regimen, prescription of glucagon, or delivery of diabetes self-management education, is critical for the prevention of recurrent hypoglycemia.

A systemic notification strategy is important because patients generally do not report their hypoglycemic event to their health care providers (42,43), and providers do not consistently screen their at-risk patients for hypoglycemia (44). Multiple studies (41,44) have demonstrated low rates of treatment modification after severe hypoglycemic events, which increases the likelihood of their recurrence. Efforts are currently underway to examine, in detail, treatment modifications that may have occurred after the index hypoglycemic event among both transported and nontransported patients in order to elucidate what factors, exactly, were responsible for lower rates of hypoglycemia recurrence among transported patients. While this investigation was outside the scope of our current work, our findings underscore the importance of patients informing their health care providers about their hypoglycemic events in a timely manner (given the high rate of hypoglycemia recurrence even within 3 days of the initial event) and clinicians intervening to prevent hypoglycemia recurrence. Similarly, health care systems should develop systems to follow up with patients after all acute care utilization events, including EMS encounters (if these can be identified by the health care system), ED visits, and hospitalizations.

Ambulance services rarely, if ever, have access to EHR systems of the sites where their patients receive medical care. This creates two important challenges. First, EMS providers have limited knowledge of the patient’s medical history, risk factors for hypoglycemia, and clinical resources available to them to prevent hypoglycemia recurrence. EMS providers also cannot document within the patient’s EHR, which prevents them from informing the patients’ treating clinicians of the hypoglycemic episode, leaving the responsibility of notifying the care team to the patients. As a result, there is little opportunity for altering medication regimens or providing education focused on preventing another episode. Thus, patients would benefit from efforts to support the interoperability of EHR systems of ambulance agencies and health systems, which would enable for informatics tools that allow EMS teams to inform the patients’ care teams that an event occurred, triggering them to reach out to patients and work with them to improve glycemic control and self-management.

While this is one of the largest contemporary studies of severe hypoglycemia requiring EMS care that was further enriched by linking longitudinal EMS and clinic EHR data, it has important limitations. Mayo Clinic Ambulance may not have been the first responder on scene (e.g., the fire department or a bystander could have responded first) and as such, the documented glucose value and treatment may not have been the first ones to occur. We sought to minimize this limitation by manually reviewing the call reports for events in which the index glucose level was in the euglycemic range (≥70 mg/dL), but may have missed other instances when the documented glucose value was <70 mg/dL. We could not tell if patients resided in a skilled nursing facility or another congregate setting at the time or after the index hypoglycemic event. Residents of a facility may be treated by local staff for their hypoglycemic event without requiring EMS care, resulting in “missed” events. We also could not examine the rates of hypoglycemic events that were treated by family or bystanders and did not culminate in an EMS, ED, or hospital encounter. While we were not able to capture ED visits and hospitalizations outside of Mayo Clinic or MCHS, because Mayo Clinic Ambulance is the sole EMS provider in the catchment area, the number of missed clinical encounters is likely very small. Finally, this is not a population estimate, and findings may not generalize to other settings and populations. Nevertheless, our findings are representative of the upper Midwest, including both urban and rural settings.

Nearly half of severe hypoglycemic events treated by Mayo Clinic Ambulance service resulted in transport to the ED. Transport rates were highest among patients with type 2 diabetes. Among patients with type 1 diabetes, presence of comorbidities was the strongest predictor of transport to the ED. In contrast, among patients with type 2 diabetes, factors associated with transport included older age, being single (i.e., without caregiver support at home), and with features suggestive of endogenous precipitants of hypoglycemia. Recurrent severe hypoglyce-mic events were common in this population, particularly among patients who were not transported to the ED after their initial event. ED transport nearly halved the likelihood of severe hypoglycemia, underscoring the importance of clinical evaluation in the aftermath of severe hypoglycemia to provide an opportunity for timely patient education and treatment modification.

Acknowledgments. The authors thank M. Carson Rogerson IV (Mayo Clinic Ambulance) for assistance with accessing Mayo Clinic Ambulance data and Theo Herrin (formerly of Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery) for assistance accessing Mayo Clinic data.

Funding. This study was funded by National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases grants K23DK114497 and R03DK127010 (R.G.M.) and the Mayo Clinic K2R Research Award.

Study contents are the sole responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. In the last 36 months, R.G.M. has consulted with Emmi on the development of patient education materials related to prediabetes and diabetes. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.A.M. and R.G.M. designed the study, interpreted the data, and wrote the manuscript. K.M.S. analyzed the data and reviewed and edited the manuscript. A.E.G. assisted with data analyses and reviewed and edited the manuscript. L.A.M. and R.G.M. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.