OBJECTIVE

To identify severe hypoglycemia events, defined as emergency department visits or hospitalizations for hypoglycemia, in patients with type 2 diabetes receiving care in a large health system and to identify patient characteristics associated with severe hypoglycemia events.

RESEARCH DESIGN AND METHODS

This was a retrospective cohort study from January 2006 to December 2015 using the electronic medical record in the Cleveland Clinic Health System (CCHS). Participants included 50,439 patients with type 2 diabetes receiving care in the CCHS. Number of severe hypoglycemia events and associated patient characteristics were identified.

RESULTS

The incidence proportion of severe hypoglycemia increased from 0.12% in 2006 to 0.31% in 2015 (P = 0.01). Compared with patients who did not experience severe hypoglycemia, those with severe hypoglycemia had similar median glycosylated hemoglobin (HbA1c) levels. More patients with severe hypoglycemia versus those without had a prior diagnosis of nonsevere hypoglycemia (9% vs. 2%, P < 0.001). Logistic regression confirmed an increased odds for severe hypoglycemia with insulin, sulfonylureas, increased number of diabetes medications, history of nonsevere hypoglycemia (odds ratio [OR] 3.01, P < 0.001), HbA1c <6% (42 mmol/mol) (OR 1.95, P < 0.001), black race, and increased Charlson comorbidity index. Lower odds of severe hypoglycemia were noted with higher BMI and use of metformin, dipeptidyl peptidase 4 inhibitors, and glucagon-like peptide 1 agonists.

CONCLUSIONS

In this retrospective study of patients with type 2 diabetes with severe hypoglycemia, patient characteristics were identified. Patients with severe hypoglycemia had previous nonsevere hypoglycemia diagnoses more frequently than those without. Identifying patients at high risk at the point of care can allow for change in modifiable risk factors and prevention of severe hypoglycemia events.

While optimal diabetes control can help prevent diabetes complications (17), hypoglycemia remains a significant risk of diabetes treatment and has been associated with mortality (810), and iatrogenic hypoglycemia related to diabetes treatment can limit ability to achieve glycemic control (11,12). Additional notable risks include decreased rating of health-related quality of life (13) and decreased work productivity (14) as well as accidents resulting in hospital visits (15). While documented or probable symptomatic hypoglycemia or asymptomatic hypoglycemia (16) may occur more commonly than severe hypoglycemia, the occurrence of severe hypoglycemia remains an important problem. Severe hypoglycemia is defined by the American Diabetes Association as “severe cognitive impairment requiring external assistance for recovery (17).” A recent reported estimate of severe hypoglycemia in a population of insured patients leading to an emergency department (ED) visit or hospitalization is 1.4–1.6 events per 100 person-years (18), with risk factors for severe hypoglycemia identified as older age, cardiovascular disease, congestive heart failure, chronic kidney disease, depression, higher glycosylated hemoglobin (HbA1c), and use of insulin or insulin secretagogues or β-blockers (18). Socioeconomic factors can also increase risk for hypoglycemia, with more frequent admissions for hypoglycemia reported in low- versus high-income populations (19). We sought to identify all severe hypoglycemia events, defined for this study as ED visits or hospitalizations for hypoglycemia, in patients with type 2 diabetes receiving care in the Cleveland Clinic Health System (CCHS) from 2006 to 2015 to identify patient characteristics associated with severe hypoglycemia events. We focused our study on patients with type 2 diabetes, as we sought to assess in this work, in addition to patient characteristics, whether specific medications used to treat type 2 diabetes were clearly associated with severe hypoglycemia in a large health system in order to inform future preventive efforts.

We conducted a retrospective cohort study using the enterprise-wide electronic medical record (EMR) in the CCHS. Our study sample was a cohort with type 2 diabetes, established using data from 2005 to 2015 in the CCHS that was created using a modified version of the Kho algorithm (20,21).

### Identification of Type 2 Diabetes and Study Cohort

The type 2 diabetes study cohort was created based upon ICD-9/-10 codes (250X, E11X). The algorithm was used to calculate the earliest date when a patient record contained any of the following combinations: 1) type 2 diabetes code and type 2 diabetes medication, 2) type 2 diabetes code and abnormal glucose, 3) type 2 diabetes code recorded twice and outpatient insulin, 4) type 2 diabetes medication and abnormal glucose, 5) insulin preceded by type 2 diabetes medication. Combinations preceded by a type 1 diabetes code (ICD-9 or ICD-10) were discarded. Codes for ketoacidosis were classified as type 1 diabetes codes and also discarded. The earliest date that any of the five conditions above were met was documented as the date in which the patient met the criteria for type 2 diabetes. To be included each year in the 10-year cohort for this study from January 2006 to December 2015, a patient must have had an outpatient visit in a primary care (internal medicine [IM] or family medicine [FM]) department or in the endocrinology department at least once every 2 years. Patients with a history of gastric bypass were excluded from the study population, given the possibility that hypoglycemia may occur at different rates or be treated differently in this population.

### Identification of Severe Hypoglycemia

Severe hypoglycemia events were identified based upon previous definitions (22,23). We identified ED visits and hospitalizations for hypoglycemia in the EMR using the following ICD-9 codes present in the encounter: 251.0 (hypoglycemic coma), 251.1 (other specified hypoglycemia), and 251.2 (hypoglycemia, unspecified). Additionally, the following ICD-10 codes were used: E08.641, E11.641, E11.649, E13.64, E13.641, E13.649, E16.0, E16.1, E15, and E16.2. During the course of the study, we verified with review of a subset of charts that our reported events captured these severe hypoglycemia events as intended.

### Identification of Diabetes Medications

We identified a diabetes medication class as active if it was on the outpatient medication list in the EMR. Medication classes include the following: metformin, sulfonylurea, meglitinide, insulin (basal or bolus), glucagon-like peptide 1 (GLP-1) agonist, dipeptidyl peptidase 4 (DPP-4) inhibitor, α-glucosidase inhibitor (AGI), sodium–glucose cotransporter 2 (SGLT2) inhibitor, and thiazolidinedione (TZD).

### Patient Characteristics

We identified the following patient demographic characteristics: age, sex, race, and income, defined as the 2011–2015 five-year estimates of median household income at the block group level obtained from the American Community Survey conducted by the U.S. Census Bureau (24). Clinical characteristics included BMI, HbA1c within the last 12 months, number of diabetes medications, and Charlson comorbidity score (25,26). We also captured a previous history of nonsevere hypoglycemia diagnoses, defined as ICD codes in an EMR encounter that were not related to an ED visit or hospitalization and occurred prior to the severe hypoglycemia event for the group that experienced an event and prior to the last IM, FM, or endocrinology appointment for those who did not have a severe hypoglycemia event.

The presence of patient comorbidities was captured using the following diagnoses and associated ICD-9/-10 codes: dementia, cognitive impairment, cardiovascular disease, congestive heart failure, chronic kidney disease, depression, other psychiatric diagnoses, and alcohol or substance abuse (Supplementary Table 1). These comorbidities were chosen based upon previous literature associating risk of severe hypoglycemia with comorbid conditions.

### Statistical Analysis

We identified the incidence proportion of severe hypoglycemia events in each year between 2006 and 2015 and reported these as counts and percentages. The denominator for the yearly proportions was the total number of patients with type 2 diabetes in the CCHS in each year who had either a primary care or endocrinology visit within 2 years. Specifically, the number of patients at risk was captured if the patient was present in the cohort between 1 January and 31 December for each year and the patient had an appointment with primary care or endocrinology in the time frame of 1 January from the year before through 31 December of the year reported. This was done to ensure that the denominator captured in this calculation included active patients in our health system. In reporting the yearly event numbers, we reported the numbers of events that represented the first severe hypoglycemia event in the data set for a patient, the second event, or events in addition to the second event. As our study is an open cohort, we calculated the proportions of severe hypoglycemia per year based on the denominator of patients with type 2 diabetes each year and did not use censoring to calculate the yearly incidence proportions. However, we also calculated and reported the number of patients who were present in the denominator from the year prior.

Patient characteristics were compared at the last primary care or endocrinology outpatient visit between patients who experienced a hypoglycemia event and those who did not. In the group of patients with a hypoglycemia event, the last visit captured was the last outpatient visit before the event, and for those without an event, the last recorded outpatient visit was used. Variables are reported as median with interquartile ranges (IQRs) for continuous variables or numbers and percentages for categorical variables. We reported medians with IQR because our continuous variables were not normally distributed. The Mann-Whitney U test was used for continuous variables and χ2 test was used for categorical variables with a significance level of 0.05. To adjust for the multiple comparisons, we used Bonferroni testing for variable significance. To further characterize the association of very low HbA1c (<6% [42 mmol/mol]) with severe hypoglycemia events, we compared frequency of HbA1c <6% (42 mmol/mol) in the groups with and without severe hypoglycemia.

In addition, to increase understanding for planning future interventions to prevent hypoglycemia, for those patients with a severe hypoglycemia event, patient characteristics were reported at the time of the first severe hypoglycemia event and compared with those at the time of the last recorded primary care or endocrinology outpatient visit, with the thought that an intervention could be implemented at the time of the outpatient assessment.

A multivariable logistic regression analysis was performed to control for potential confounding variables. All variables from the univariate analyses were included in the regression analysis with the following exceptions: 1) medication combination variables, which are reported in the univariate analyses for clinical relevance, were not included in the regression, as these variables would cause confounding if included as separate variables; 2) HbA1c <6% (42 mmol/mol) was included in the regression instead of HbA1c as a continuous variable, as only the former was significant in the univariate analysis and it may also be of significant clinical interest for hypoglycemia prevention efforts; and 3) for number of diabetes medications, we chose only the continuous variable in the regression, as both the categorization (one, two, and three or more medications) and continuous variables were significant in the univariate analysis and use of both variables would cause confounding. To handle missing variables, for the BMI and HbA1c variables we carried over the values from earlier records (if they existed). Otherwise, multiple imputation by chained equations was used. The model was constructed using patient characteristics, medication regimen, and comorbidity variables known at the time of the severe hypoglycemia event for those experiencing an event and at the last IM, FM, or endocrinology visit for those not experiencing an event. Statistical analyses were performed using the R statistical software, version 3.4.0 (https://cran.r-project.org/). This protocol was approved by the institutional review board at Cleveland Clinic.

Baseline characteristics of the 50,439 patients in the study cohort at time of entry with diabetes diagnosis are shown in Table 1. The highest percentage of patients who were taking a diabetes medication were taking metformin alone. The median age of the cohort was 61.0 years, and 51.7% were male. The majority of patients were Caucasian (74.1%), and 16.8% were black. The median HbA1c was 6.9% (52 mmol/mol) and median BMI was 32.4 kg/m2. A significant proportion of patients had a documented history of cardiovascular disease (32.9%), depression (14.8%), or other psychiatric diagnoses (21.8%). The study cohort had a mean follow-up time in the data set of 3.6 years (median 2.8 years [IQR 1.4, 5.3] [minimum 0 years and maximum 10 years]).

Table 1

Baseline characteristics of study cohort

VariableNumber missingAll patients = 50,439
Diabetes medication
No medication  10,357 (20.5)
AGI  30 (0.1)
DPP-4 inhibitor alone  564 (1.1)
GLP-1 agonist alone  163 (0.3)
Metformin alone  18,694 (37.1)
Meglitinide alone  35 (0.1)
TZD alone  915 (1.8)
Insulin alone  2,201 (4.4)
Insulin + other diabetes medication  2,030 (4.0)
Insulin + sulfonylureas + other diabetes medication  1,869 (3.7)
No insulin and no sulfonylureas + 2 or more diabetes medications  3,077 (6.1)
SGLT2 inhibitor alone  22 (0)
Sulfonylureas alone  2,991 (5.9)
Sulfonylureas + other diabetes medication  7,491 (14.9)
Number of diabetes medications
0  10,357 (20.5)
1–2  34,291 (68)
≥3  5,791 (11.5)
Age at entering cohort (years)  61.0 (51.9, 69.9)
Race 1,222 (2.4)
Caucasian  37,386 (74.1)
Black  8,439 (16.8)
Asian/Pacific Islander  753 (1.5)
Other  2,639 (5.2)
Sex 1 (0)
Female  24,344 (48.3)
Male  26,094 (51.7)
Income (in thousands) 5,906 (12) 54.2 (39.4, 73.0)
BMI, kg/m2 12,687 (25) 32.4 (28.3, 37.8)
HbA1c, % [mmol/mol] 20,104 (40) 6.9 (6.5, 8) [52 (48, 64)]
Dementia  627 (1.2)
Cognitive impairment  810 (1.6)
Cardiovascular disease  16,605 (32.9)
Congestive heart failure  2,984 (5.9)
Chronic kidney disease  2,577 (5.1)
Depression  7,464 (14.8)
Other psychiatric diagnoses  10,984 (21.8)
Alcohol or substance abuse  4,604 (9.1)
Charlson comorbidity index  1 (1, 2)
History of hypoglycemia  412 (0.8)
VariableNumber missingAll patients = 50,439
Diabetes medication
No medication  10,357 (20.5)
AGI  30 (0.1)
DPP-4 inhibitor alone  564 (1.1)
GLP-1 agonist alone  163 (0.3)
Metformin alone  18,694 (37.1)
Meglitinide alone  35 (0.1)
TZD alone  915 (1.8)
Insulin alone  2,201 (4.4)
Insulin + other diabetes medication  2,030 (4.0)
Insulin + sulfonylureas + other diabetes medication  1,869 (3.7)
No insulin and no sulfonylureas + 2 or more diabetes medications  3,077 (6.1)
SGLT2 inhibitor alone  22 (0)
Sulfonylureas alone  2,991 (5.9)
Sulfonylureas + other diabetes medication  7,491 (14.9)
Number of diabetes medications
0  10,357 (20.5)
1–2  34,291 (68)
≥3  5,791 (11.5)
Age at entering cohort (years)  61.0 (51.9, 69.9)
Race 1,222 (2.4)
Caucasian  37,386 (74.1)
Black  8,439 (16.8)
Asian/Pacific Islander  753 (1.5)
Other  2,639 (5.2)
Sex 1 (0)
Female  24,344 (48.3)
Male  26,094 (51.7)
Income (in thousands) 5,906 (12) 54.2 (39.4, 73.0)
BMI, kg/m2 12,687 (25) 32.4 (28.3, 37.8)
HbA1c, % [mmol/mol] 20,104 (40) 6.9 (6.5, 8) [52 (48, 64)]
Dementia  627 (1.2)
Cognitive impairment  810 (1.6)
Cardiovascular disease  16,605 (32.9)
Congestive heart failure  2,984 (5.9)
Chronic kidney disease  2,577 (5.1)
Depression  7,464 (14.8)
Other psychiatric diagnoses  10,984 (21.8)
Alcohol or substance abuse  4,604 (9.1)
Charlson comorbidity index  1 (1, 2)
History of hypoglycemia  412 (0.8)

Data are n (%) or median (IQR).

The incidence proportions of severe hypoglycemia requiring an ED visit or hospitalization increased from 0.12% (8 events/6,491 patients with type 2 diabetes) in 2006 to 0.31% (101 events/33,114 patients with type 2 diabetes) in 2015 (P = 0.01) (Table 2). A greater proportion of severe hypoglycemia episodes were hospitalizations compared with ED visits alone within each calendar year (63% in 2006 and 88% in 2015). The number of ED sites represented increased from 4 in 2006 to 14 in 2015, while the number of hospital sites represented increased from 10 to 14 in the same time period. Most of the severe hypoglycemia events were the first such event for patients in the data set, although recurrent events were frequent.

Table 2

Annual incidence proportion of severe hypoglycemia events in a type 2 diabetes prevalence cohort from 2006 to 2015

Year
2006200720082009201020112012201320142015
Emergency sites 12 13 13 14 14
Hospitalization sites 10 10 10 11 11 13 13 13 14 14
Patients at risk 6,491 9,331 11,314 12,562 15,150 17,888 20,726 24,419 27,796 33,114
Patients at risk from previous year  5,608 7,991 8,797 11,236 13,549 16,052 18,771 21,839 24,758
SH  8 (0.12) 7 (0.07) 7 (0.06) 34 (0.27) 42 (0. 28) 45 (0.25) 72 (0.35) 79 (0.32) 92 (0.33) 101 (0.31)
1st SH in data set 26 28 31 56 58 68 81
2nd SH in data set   14 14 11 15 18 16
>2 SHs in data set
SH ED visits  3 (38) 2 (29) 2 (29) 4 (12) 3 (7) 4 (9) 7 (10) 5 (6) 4 (4) 12 (12)
SH hospitalizations 5 (63) 5 (71) 5 (71) 30 (88) 39 (93) 41 (91) 65 (90) 74 (94) 88 (96) 89 (88)
Year
2006200720082009201020112012201320142015
Emergency sites 12 13 13 14 14
Hospitalization sites 10 10 10 11 11 13 13 13 14 14
Patients at risk 6,491 9,331 11,314 12,562 15,150 17,888 20,726 24,419 27,796 33,114
Patients at risk from previous year  5,608 7,991 8,797 11,236 13,549 16,052 18,771 21,839 24,758
SH  8 (0.12) 7 (0.07) 7 (0.06) 34 (0.27) 42 (0. 28) 45 (0.25) 72 (0.35) 79 (0.32) 92 (0.33) 101 (0.31)
1st SH in data set 26 28 31 56 58 68 81
2nd SH in data set   14 14 11 15 18 16
>2 SHs in data set
SH ED visits  3 (38) 2 (29) 2 (29) 4 (12) 3 (7) 4 (9) 7 (10) 5 (6) 4 (4) 12 (12)
SH hospitalizations 5 (63) 5 (71) 5 (71) 30 (88) 39 (93) 41 (91) 65 (90) 74 (94) 88 (96) 89 (88)

Data are n or n (%). SH, severe hypoglycemia event.

The univariate analyses of characteristics of patients with a severe hypoglycemia event compared with those without a severe hypoglycemia event captured at the last primary care or endocrinology visit are reported by hypoglycemia group in Supplementary Table 2 and by variable group in Supplementary Table 3. Compared with patients who did not experience a severe hypoglycemia episode, those with severe hypoglycemia included a greater proportion of patients who were older (71.9 vs. 64.1 years, P < 0.001) and had a lower BMI (29.2 vs. 31.7 kg/m2, P < 0.001), but median HbA1c was similar (6.8% [51 mmol/mol] vs. 6.7% [50 mmol/mol], P = 0.925). A significantly higher proportion of patients treated with insulin alone, a sulfonylurea alone, insulin with a sulfonylurea in addition to another diabetes medication, or a sulfonylurea with another diabetes medication (other than insulin) had severe hypoglycemia. A greater percentage of patients treated with three or more diabetes medications in any class had severe hypoglycemia (26.5% vs. 18.5%, P < 0.001). A significant proportion of black patients experienced severe hypoglycemia compared with the prevalence of black race at baseline—32.2% of patients with severe hypoglycemia were black vs. 60.4% Caucasian (P < 0.001)—and those with severe hypoglycemia had a lower median income ($48,000 vs.$54,000, P < 0.001). There was a significantly greater prevalence of comorbidities in patients with severe hypoglycemia, including dementia, cognitive impairment, cardiovascular disease, congestive heart failure, and chronic kidney disease, and a higher Charlson comorbidity score (4 vs. 2, P < 0.001). There was a similar prevalence of depression, other psychiatric diagnoses, and alcohol or substance abuse in those with and those without severe hypoglycemia. Nine percent of patients who experienced severe hypoglycemia had a documented previous history of nonsevere hypoglycemia in the EMR prior to the severe hypoglycemia episode compared with 2.2% documented in patients who did not experience severe hypoglycemia prior to their last IM, FM, or endocrinology visit in the data set (P < 0.001).

In the subset of patients with severe hypoglycemia (N = 366), 16.1% (N = 59) had HbA1c <6% (42 mmol/mol), while in patients without severe hypoglycemia (N = 50,073), only 12.0% (N = 5,990) had HbA1c <6% (42 mmol/mol) (P for difference = 0.01).

The multivariable logistic regression analysis based upon variables obtained at the last IM, FM, or endocrinology visit for patients without hypoglycemia and at the time of the severe hypoglycemia event for those with severe hypoglycemia is shown in Table 3. This confirms an increased odds for severe hypoglycemia event with the use of insulin (odds ratio [OR] 2.77, P < 0.001), with the use of sulfonylureas (OR 2.49, P < 0.001), with an increased number of diabetes medications (OR 1.56, P < 0.001), with history of nonsevere hypoglycemia (OR 3.01, P < 0.001), and with HbA1c <6% (42 mmol/mol) (OR 1.95, P < 0.001), black race (OR 2.55, P < 0.001), cardiovascular disease (OR 1.68, P 0.001), congestive heart failure (OR 1.33, P 0.04), and increased Charlson comorbidity index (OR 1.15, P < 0.001). Other comorbidities including dementia, cognitive impairment, chronic kidney disease, depression, other psychiatric diagnoses, and alcohol/substance abuse were not significantly associated with severe hypoglycemia in this model. Lower odds of severe hypoglycemia were noted with higher BMI and use of metformin, DPP-4 inhibitors, or GLP-1 agonists. The model was adjusted for age, sex, and median income.

Table 3

Regression model of severe hypoglycemia events to identify important patient characteristics associated with severe hypoglycemia

VariablesOR (95% CI)P
Intercept 0.001 (0, 0.004) <0.001
Insulin 2.773 (1.975, 3.892) <0.001
Sulfonylureas 2.487 (1.919, 3.223) <0.001
AGI 2.049 (0.699, 6.009) 0.19
DPP-4 inhibitor 0.514 (0.354, 0.744) 0.001
GLP-1 agonist 0.225 (0.081, 0.623) 0.004
Metformin 0.431 (0.321, 0.578) <0.001
Meglitinide 1.181 (0.492, 2.833) 0.71
SGLT2 inhibitor 0.001 (0, 1.8e+14) 0.71
TZD 0.725 (0.433, 1.215) 0.22
Number of diabetes medications 1.558 (1.357, 1.789) <0.001
History of nonsevere hypoglycemia diagnosis 3.01 (2.087, 4.343) <0.001
Age 1.009 (0.998, 1.02) 0.10
BMI 0.962 (0.945, 0.98) <0.001
HbA1c <6% (42 mmol/mol) 1.952 (1.436, 2.652) <0.001
Median income 1.002 (0.997, 1.006) 0.41
Male sex  0.672 (0.532, 0.849) 0.001
Race
Caucasian Reference
Black  2.553 (1.962, 3.322) <0.001
Asian/Pacific Islander 1.221 (0.38, 3.928) 0.73
Other 0.798 (0.389, 1.638) 0.54
Dementia 0.773 (0.487, 1.226) 0.27
Cognitive impairment 1.398 (0.949, 2.057) 0.09
Cardiovascular disease 1.684 (1.242, 2.283) 0.001
Congestive heart failure 1.331 (1.013, 1.75) 0.04
Chronic kidney disease 1.266 (0.949, 1.688) 0.11
Depression 1.081 (0.815, 1.434) 0.59
Other psychiatric diagnoses 0.924 (0.712, 1.199) 0.55
Alcohol or substance abuse 1.045 (0.765, 1.428) 0.78
Charlson comorbidity index 1.153 (1.096, 1.212) <0.001
VariablesOR (95% CI)P
Intercept 0.001 (0, 0.004) <0.001
Insulin 2.773 (1.975, 3.892) <0.001
Sulfonylureas 2.487 (1.919, 3.223) <0.001
AGI 2.049 (0.699, 6.009) 0.19
DPP-4 inhibitor 0.514 (0.354, 0.744) 0.001
GLP-1 agonist 0.225 (0.081, 0.623) 0.004
Metformin 0.431 (0.321, 0.578) <0.001
Meglitinide 1.181 (0.492, 2.833) 0.71
SGLT2 inhibitor 0.001 (0, 1.8e+14) 0.71
TZD 0.725 (0.433, 1.215) 0.22
Number of diabetes medications 1.558 (1.357, 1.789) <0.001
History of nonsevere hypoglycemia diagnosis 3.01 (2.087, 4.343) <0.001
Age 1.009 (0.998, 1.02) 0.10
BMI 0.962 (0.945, 0.98) <0.001
HbA1c <6% (42 mmol/mol) 1.952 (1.436, 2.652) <0.001
Median income 1.002 (0.997, 1.006) 0.41
Male sex  0.672 (0.532, 0.849) 0.001
Race
Caucasian Reference
Black  2.553 (1.962, 3.322) <0.001
Asian/Pacific Islander 1.221 (0.38, 3.928) 0.73
Other 0.798 (0.389, 1.638) 0.54
Dementia 0.773 (0.487, 1.226) 0.27
Cognitive impairment 1.398 (0.949, 2.057) 0.09
Cardiovascular disease 1.684 (1.242, 2.283) 0.001
Congestive heart failure 1.331 (1.013, 1.75) 0.04
Chronic kidney disease 1.266 (0.949, 1.688) 0.11
Depression 1.081 (0.815, 1.434) 0.59
Other psychiatric diagnoses 0.924 (0.712, 1.199) 0.55
Alcohol or substance abuse 1.045 (0.765, 1.428) 0.78
Charlson comorbidity index 1.153 (1.096, 1.212) <0.001

Characteristics of patients captured at their last primary care or endocrinology outpatient visit compared with the time of the severe hypoglycemia event are shown in Table 4. The visits reported represent visits with 41 unique endocrinology providers and 208 unique IM or FM providers. The time elapsed between the last primary care or endocrinology visit and the severe hypoglycemia event was a mean of 3.3 and median of 1.5 months (IQR 0.5, 3.5). A significantly greater proportion of patients were taking insulin, either alone or with other diabetes medications, were taking three or more diabetes medications in any class, had a diagnosis of cardiovascular disease, had a trend toward diagnoses of chronic kidney disease, and had a higher Charlson comorbidity score at the time of the severe hypoglycemia event as compared with their last outpatient IM, FM, or endocrinology visit. There was no significant difference between HbA1c values at the last visit and at the time of the severe hypoglycemia episode (median HbA1c 6.8% [51 mmol/mol] vs. 6.7% [50 mmol/mol], P = 0.41).

Table 4

Patients with severe hypoglycemia: characteristics at last IM, FM, or endocrinology visit compared with those at the hypoglycemia event

VariableCharacteristic at last visit (N = 366)Characteristic at severe hypoglycemia event (N = 366)P
Diabetes medication*   <0.001
No medication 20 (5.5) 15 (4.1)
AGI alone 2 (0.5) 1 (0.3)
DPP-4 inhibitor alone 3 (0.8) 4 (1.1)
GLP-1 agonist alone 0 (0) 0 (0)
Metformin alone 24 (6.6) 14 (3.8)
Meglitinide alone 0 (0) 0 (0)
TZD alone 3 (0.8) 0 (0)
Insulin alone 88 (24) 107 (29.2)
Insulin + other diabetes medication 23 (6.3) 36 (9.8)
Insulin + sulfonylureas + other diabetes medication 41 (11.2) 87 (23.8)
No insulin and no sulfonylureas + 2 or more diabetes medications 4 (1.1) 0 (0)
SGLT2 inhibitor alone 0 (0) 0 (0)
Sulfonylureas alone 59 (16.1) 45 (12.3)
Sulfonylureas + other diabetes medication 99 (27) 57 (15.6)
Number of diabetes medications   0.002
0 20 (5.4) 15 (4.1)
1–2 249 (68) 210 (57.4)
≥3 97 (26.5) 141 (38.5)
History of nonsevere hypoglycemia diagnosis 33 (9) 42 (11.5) 0.27
Age at visit, years 71.85 (61.35, 80.54) 72.04 (61.48, 80.75) 0.74
BMI, kg/m2 29.2 (25.0, 34.1) 29.22 (24.68, 34.38) 0.77
HbA1c, % [mmol/mol] 6.8 (6.1, 8) [51 (43, 64)] 6.7 (6, 7.8) [50 (42, 62)] 0.41
Dementia 25 (6.8) 33 (9) 0.27
Cognitive impairment 36 (9.8) 47 (12.8) 0.20
Cardiovascular disease 260 (71) 284 (77.6) 0.04
Congestive heart failure 121 (33.1) 144 (39.3) 0.08
Chronic kidney disease 134 (36.6) 160 (43.7) 0.05
Depression 96 (26.2) 104 (28.4) 0.51
Other psychiatric diagnoses 131 (35.8) 153 (41.8) 0.10
Alcohol or substance abuse 61 (16.7) 71 (19.4) 0.34
Charlson comorbidity index 4 (3, 7) 5.5 (3, 7) <0.001
VariableCharacteristic at last visit (N = 366)Characteristic at severe hypoglycemia event (N = 366)P
Diabetes medication*   <0.001
No medication 20 (5.5) 15 (4.1)
AGI alone 2 (0.5) 1 (0.3)
DPP-4 inhibitor alone 3 (0.8) 4 (1.1)
GLP-1 agonist alone 0 (0) 0 (0)
Metformin alone 24 (6.6) 14 (3.8)
Meglitinide alone 0 (0) 0 (0)
TZD alone 3 (0.8) 0 (0)
Insulin alone 88 (24) 107 (29.2)
Insulin + other diabetes medication 23 (6.3) 36 (9.8)
Insulin + sulfonylureas + other diabetes medication 41 (11.2) 87 (23.8)
No insulin and no sulfonylureas + 2 or more diabetes medications 4 (1.1) 0 (0)
SGLT2 inhibitor alone 0 (0) 0 (0)
Sulfonylureas alone 59 (16.1) 45 (12.3)
Sulfonylureas + other diabetes medication 99 (27) 57 (15.6)
Number of diabetes medications   0.002
0 20 (5.4) 15 (4.1)
1–2 249 (68) 210 (57.4)
≥3 97 (26.5) 141 (38.5)
History of nonsevere hypoglycemia diagnosis 33 (9) 42 (11.5) 0.27
Age at visit, years 71.85 (61.35, 80.54) 72.04 (61.48, 80.75) 0.74
BMI, kg/m2 29.2 (25.0, 34.1) 29.22 (24.68, 34.38) 0.77
HbA1c, % [mmol/mol] 6.8 (6.1, 8) [51 (43, 64)] 6.7 (6, 7.8) [50 (42, 62)] 0.41
Dementia 25 (6.8) 33 (9) 0.27
Cognitive impairment 36 (9.8) 47 (12.8) 0.20
Cardiovascular disease 260 (71) 284 (77.6) 0.04
Congestive heart failure 121 (33.1) 144 (39.3) 0.08
Chronic kidney disease 134 (36.6) 160 (43.7) 0.05
Depression 96 (26.2) 104 (28.4) 0.51
Other psychiatric diagnoses 131 (35.8) 153 (41.8) 0.10
Alcohol or substance abuse 61 (16.7) 71 (19.4) 0.34
Charlson comorbidity index 4 (3, 7) 5.5 (3, 7) <0.001

Data are n (%) or median (IQR).

*Bonferroni-adjusted pairwise χ2 test for insulin + sulfonylureas + other diabetes medication versus sulfonylureas alone, as well as for insulin + sulfonylureas + other diabetes medication versus sulfonylureas + other diabetes medication, is significant.

†Bonferroni-adjusted pairwise χ2 test for one to two medications versus three or more medications is significant.

In this retrospective study of patients with type 2 diabetes in a large health system, we found that the incidence proportion of severe hypoglycemia defined as an ED visit or hospitalization significantly increased over 10 years from 2006 to 2015 from 0.12% to 0.31%, with most events resulting in hospitalization. The median HbA1c was <7% (53 mmol/mol) in the cohort and was not significantly different in those who developed severe hypoglycemia and those who did not. Our regression model demonstrated an increased odds for severe hypoglycemia with a history of nonsevere hypoglycemia, HbA1c <6% (42 mmol/mol), black race, and increased comorbidities, as well as with the use of insulin, sulfonylureas, and increased number of diabetes medications.

Previous work has identified black race, recent hospital discharge, and use of medication from five or more therapeutic classes (27), as well as dementia and cognitive impairment (28), as risk factors for severe hypoglycemia in older adults, and cardiovascular disease, congestive heart failure, chronic kidney disease, and depression have been associated with severe hypoglycemia (18). Intensive treatment of diabetes in clinically complex patients has also been associated with increased risk of severe hypoglycemia (29). However, hospitalizations for severe hypoglycemia have been shown to remain high in adults over age 75 years and in black patients covered by Medicare fee-for-service insurance (30). Our study supports these previous findings of increased proportions of patients with severe hypoglycemia in the groups on multiple antidiabetes drugs, those with multiple comorbidities, and those of black race.

In our study, the median HbA1c was <7% (53 mmol/mol) both in patients with and in patients without severe hypoglycemia but with a greater proportion of those with HbA1c <6% (42 mmol/mol) in the group with severe hypoglycemia. The increased odds of severe hypoglycemia with a very low HbA1c was also demonstrated in our regression model. This is different from a previous report that found higher HbA1c to be associated with severe hypoglycemia (18). However, consideration of deintensification of diabetes treatment in patients at risk for hypoglycemia has previously been advocated (31) and may be beneficial in a population similar to our study cohort with multiple comorbidities. The possibility of glycemic overtreatment in patients at high risk for severe hypoglycemia has also been a target for awareness in the Veterans Health Administration as part of the Choosing Wisely initiative (32,33). Tseng et al. (34) demonstrated that a significant proportion of patients with risk factors for severe hypoglycemia cared for in the Veterans Health Administration were receiving treatment with insulin and or sulfonylureas and had evidence of intensive treatment, suggesting possible overtreatment in these high-risk groups.

In our study of real-world practice EMR data, the fact that 9% of patients experiencing severe hypoglycemia had a previous history of documented nonsevere hypoglycemia is a significant observation, and our regression model also shows a significantly higher odds (3.01) of severe hypoglycemia in patients with this history. The ability to capture documented nonsevere hypoglycemia in the EMR presents an additional opportunity to use our EMR systems to identify patients who may be at increased risk for severe hypoglycemia based upon this history in addition to demographic characteristics and comorbidities previously identified as associated with severe hypoglycemia. Identifying patients at risk for severe hypoglycemia using data readily available in the EMR can promote appropriate strategies for either deintensification of diabetes treatment or a change in treatment choices at the point of care for prevention of severe hypoglycemia events. Our finding that a greater proportion of patients were taking insulin or were taking three or more diabetes medications and had more comorbidities at the time of the severe hypoglycemia event as compared with their last outpatient primary care or endocrinology visit leads us to question whether treatment may have actually been inappropriately intensified at that last visit, especially in patients at risk for comorbidities. The EMR may be used to reduce missed opportunities at the time of the outpatient visit to identify at-risk patients and to make changes in modifiable risk factors, such as switching to medications that may have lower risk of hypoglycemia, to prevent a severe hypoglycemia event.

While hypoglycemia can limit the ability for patients with type 2 diabetes to reach a target treatment goal and is associated with additional risks that can affect quality of life, the health care costs of hypoglycemia in type 2 diabetes can also be significant, with a mean cost previously estimated at $17,564 for a hospital admission and$1,387 for an ED visit (35). Thus, identifying patients who may be at risk for severe hypoglycemia and implementing prevention strategies is of utmost importance to both improve health outcomes of patients with type 2 diabetes and reduce costs.

Our study is strengthened by our use of a shared EMR across our entire health system, but a limitation is the documentation of hypoglycemia as well as comorbidities using ICD codes, which relies on provider coding practices. It may be that improved coding for hypoglycemia as a diagnosis in an ED or hospital encounter could have explained some of the increase in rate of severe hypoglycemia events captured in the time frame of our study, specifically, between 2008 and 2009. However, the number of patients with type 2 diabetes in our data set also steadily increased during the time period of our study, as did the number of sites in our health system. Our model did not adjust for clustering of severe hypoglycemia by site or for recurrent events, given the low number of events per site, but this will be explored in future work as we improve our ability to capture severe hypoglycemia in the EMR through more accurate coding practices. In addition, the mean follow-up time in our cohort was 3.6 years. This may have limited our ability to model factors associated with severe hypoglycemia because of survivorship bias. However, as we took care to limit our population to patients who had evidence of obtaining regular care in our health system, our findings reflect real-world health system data and thus have implications for preventive efforts using our current EMR systems.

It is also likely that we have underestimated the occurrence of nonsevere hypoglycemia by solely using ICD codes to identify the events. Adding the use of natural language processing of “free-text” progress notes may allow for better capture of documented hypoglycemia (both severe and nonsevere episodes) (36) in future work. In addition, while our focus in this work was on severe hypoglycemia episodes, which are more likely to be documented than less severe hypoglycemia, we have not captured episodes of severe hypoglycemia that may have been treated outside our health system. For example, patients may have been assessed and treated for severe hypoglycemia by calling an emergency medical service to their home and subsequently not transferred to an ED if symptoms had resolved. Finally, our record of medication use was based upon EMR documentation and not on pharmacy data documenting that the medication prescription was indeed filled. However, even with pharmacy data, it may not be possible to ascertain compliance with a medication regimen in a retrospective study.

In this retrospective study of patients with type 2 diabetes in a large, integrated health system, we identified characteristics of patients experiencing severe hypoglycemia. Importantly, many patients with severe hypoglycemia had previously documented nonsevere hypoglycemia. Future work should focus on identification of patients at high risk at the point of care to allow for change in modifiable risk factors and prevention of severe hypoglycemia events.

Funding. A.D.M.-H. receives funding from the Agency for Healthcare Research and Quality (grant no. K08HS024128).

Duality of Interest. This work was funded by a Merck Investigator Studies Program grant (MERCK1702AMH). A.D.M.-H., X.J., A.M., K.M.C., J.M.B., and M.W.K. report receiving research funding from Novo Nordisk and Merck. K.M.P. reports receiving research funding from Novo Nordisk and Merck; receiving consulting fees from Novo Nordisk, Sanofi, Eli Lilly, and Merck; and receiving honoraria from Merck, AstraZeneca, Sanofi, and Novo Nordisk for speaking/educational activities. R.S.Z. reports receiving research funding from Novo Nordisk and Merck, receiving speaker honoraria from Merck and Johnson & Johnson, and receiving consulting fees from Novo Nordisk and Merck. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.D.M.-H. created the study protocol, contributed to the analysis plan, wrote the first draft of the manuscript, and revised the manuscript to create the final version. K.M.P., X.J., A.M., T.D., K.M.C., J.M.B., M.W.K., and R.S.Z. contributed to the design of the study protocol and reviewed the manuscript. X.J., T.D., and M.W.K. contributed to the analysis plan. X.J. performed the data analysis. A.M. was responsible for data extraction. A.D.M.-H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
The DCCT Research Group
.
The Diabetes Control and Complications Trial (DCCT): Design and methodologic considerations for the feasibility phase
.
Diabetes
1986
;
35
:
530
545
[PubMed]
2.
Nathan
DM
,
Cleary
PA
,
Backlund
JY
, et al.;
Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) Study Research Group
.
Intensive diabetes treatment and cardiovascular disease in patients with type 1 diabetes
.
N Engl J Med
2005
;
353
:
2643
2653
[PubMed]
3.
UK Prospective Diabetes Study (UKPDS) Group
.
Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33)
.
Lancet
1998
;
352
:
837
853
[PubMed]
4.
Holman
RR
,
Paul
SK
,
Bethel
MA
,
Matthews
DR
,
Neil
HAW
.
10-year follow-up of intensive glucose control in type 2 diabetes
.
N Engl J Med
2008
;
359
:
1577
1589
[PubMed]
5.
Duckworth
W
,
Abraira
C
,
Moritz
T
, et al.;
.
Glucose control and vascular complications in veterans with type 2 diabetes
.
N Engl J Med
2009
;
360
:
129
139
[PubMed]
6.
,
Patel
A
,
MacMahon
S
,
Chalmers
J
, et al
.
Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes
.
N Engl J Med
2008
;
358
:
2560
2572
[PubMed]
7.
Action to Control Cardiovascular Risk in Diabetes Study Group
,
Gerstein
HC
,
Miller
ME
,
Byington
RP
, et al
.
Effects of intensive glucose lowering in type 2 diabetes
.
N Engl J Med
2008
;
358
:
2545
2559
[PubMed]
8.
Bonds
DE
,
Miller
ME
,
Bergenstal
RM
, et al
.
The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study
.
BMJ
2010
;
340
:
b4909
[PubMed]
9.
McCoy
RG
,
Van Houten
HK
,
Ziegenfuss
JY
,
Shah
ND
,
Wermers
RA
,
Smith
SA
.
Increased mortality of patients with diabetes reporting severe hypoglycemia
.
Diabetes Care
2012
;
35
:
1897
1901
[PubMed]
10.
Zoungas
S
,
Patel
A
,
Chalmers
J
, et al.;
.
Severe hypoglycemia and risks of vascular events and death
.
N Engl J Med
2010
;
363
:
1410
1418
[PubMed]
11.
Cryer
P
.
Hypoglycemia during therapy of diabetes
. In
Endotext
.
De Groot
LJ
,
Chrousos
G
,
Dungan
K
, et al
., Eds.
South Dartmouth, MA
,
MDText.com, Inc.
,
2000
12.
Cryer
PE
.
The barrier of hypoglycemia in diabetes
.
Diabetes
2008
;
57
:
3169
3176
[PubMed]
13.
Alvarez-Guisasola
F
,
Yin
DD
,
Nocea
G
,
Qiu
Y
,
Mavros
P
.
Association of hypoglycemic symptoms with patients’ rating of their health-related quality of life state: a cross sectional study
.
Health Qual Life Outcomes
2010
;
8
:
86
[PubMed]
14.
Brod
M
,
Christensen
T
,
Thomsen
TL
,
Bushnell
DM
.
The impact of non-severe hypoglycemic events on work productivity and diabetes management
.
Value Health
2011
;
14
:
665
671
[PubMed]
15.
Signorovitch
JE
,
Macaulay
D
,
Diener
M
, et al
.
Hypoglycaemia and accident risk in people with type 2 diabetes mellitus treated with non-insulin antidiabetes drugs
.
Diabetes Obes Metab
2013
;
15
:
335
341
[PubMed]
16.
Seaquist
ER
,
Anderson
J
,
Childs
B
, et al
.
Hypoglycemia and diabetes: a report of a workgroup of the American Diabetes Association and the Endocrine Society
.
Diabetes Care
2013
;
36
:
1384
1395
[PubMed]
17.
International Hypoglycaemia Study Group
.
Glucose concentrations of less than 3.0 mmol/L (54 mg/dL) should be reported in clinical trials: a joint position statement of the American Diabetes Association and the European Association for the Study of Diabetes
.
Diabetes Care
2017
;
40
:
155
157
[PubMed]
18.
Pathak
RD
,
Schroeder
EB
,
Seaquist
ER
, et al.;
SUPREME-DM Study Group
.
Severe hypoglycemia requiring medical intervention in a large cohort of adults with diabetes receiving care in U.S. integrated health care delivery systems: 2005–2011
.
Diabetes Care
2016
;
39
:
363
370
[PubMed]
19.
Seligman
HK
,
Bolger
AF
,
Guzman
D
,
López
A
,
Bibbins-Domingo
K
.
Exhaustion of food budgets at month’s end and hospital admissions for hypoglycemia
.
Health Aff (Millwood)
2014
;
33
:
116
123
[PubMed]
20.
Kho
AN
,
Hayes
MG
,
Rasmussen-Torvik
L
, et al
.
Use of diverse electronic medical record systems to identify genetic risk for type 2 diabetes within a genome-wide association study
.
J Am Med Inform Assoc
2012
;
19
:
212
218
[PubMed]
21.
Pantalone
KM
,
Hobbs
TM
,
Wells
BJ
, et al
.
Clinical characteristics, complications, comorbidities and treatment patterns among patients with type 2 diabetes mellitus in a large integrated health system
.
BMJ Open Diabetes Res Care
2015
;
3
:
e000093
[PubMed]
22.
Whitmer
RA
,
Karter
AJ
,
Yaffe
K
,
Quesenberry
CP
Jr.
,
Selby
JV
.
Hypoglycemic episodes and risk of dementia in older patients with type 2 diabetes mellitus
.
JAMA
2009
;
301
:
1565
1572
[PubMed]
23.
Ginde
AA
,
Blanc
PG
,
Lieberman
RM
,
Camargo
CA
Jr
.
Validation of ICD-9-CM coding algorithm for improved identification of hypoglycemia visits
.
BMC Endocr Disord
2008
;
8
:
4
[PubMed]
24.
U.S. Census Bureau. American Community Survey (ACS) [Internet]. Available from https://www.census.gov/programs-surveys/acs. Accessed 11 August 2017
25.
Charlson
ME
,
Pompei
P
,
Ales
KL
,
MacKenzie
CR
.
A new method of classifying prognostic comorbidity in longitudinal studies: development and validation
.
J Chronic Dis
1987
;
40
:
373
383
[PubMed]
26.
Charlson
M
,
Szatrowski
TP
,
Peterson
J
,
Gold
J
.
Validation of a combined comorbidity index
.
J Clin Epidemiol
1994
;
47
:
1245
1251
[PubMed]
27.
Shorr
RI
,
Ray
WA
,
Daugherty
JR
,
Griffin
MR
.
Incidence and risk factors for serious hypoglycemia in older persons using insulin or sulfonylureas
.
Arch Intern Med
1997
;
157
:
1681
1686
[PubMed]
28.
Feil
DG
,
Rajan
M
,
Soroka
O
,
Tseng
C-L
,
Miller
DR
,
Pogach
LM
.
Risk of hypoglycemia in older veterans with dementia and cognitive impairment: implications for practice and policy
.
J Am Geriatr Soc
2011
;
59
:
2263
2272
[PubMed]
29.
McCoy
RG
,
Lipska
KJ
,
Yao
X
,
Ross
JS
,
Montori
VM
,
Shah
ND
.
Intensive treatment and severe hypoglycemia among adults with type 2 diabetes
.
JAMA Intern Med
2016
;
176
:
969
978
[PubMed]
30.
Lipska
KJ
,
Ross
JS
,
Wang
Y
, et al
.
National trends in US hospital admissions for hyperglycemia and hypoglycemia among Medicare beneficiaries, 1999 to 2011
.
JAMA Intern Med
2014
;
174
:
1116
1124
[PubMed]
31.
Kerr
EA
,
Hofer
TP
.
Deintensification of routine medical services: The next frontier for improving care quality
.
JAMA Intern Med
2016
;
176
:
978
980
[PubMed]
32.
U.S. Department of Veterans Affairs. Quality, safety & value [Internet]. Available from http://www.qualityandsafety.va.gov/ChoosingWiselyHealthSafetyInitiative/HypoglycemiaSite/For_Clinicians.asp. Accessed 5 April 2016
33.
Aron
DC
,
Lowery
J
,
Tseng
CL
,
Conlin
P
,
Kahwati
L
.
De-implementation of inappropriately tight control (of hypoglycemia) for health: protocol with an example of a research grant application
.
Implement Sci
2014
;
9
:
58
[PubMed]
34.
Tseng
CL
,
Soroka
O
,
Maney
M
,
Aron
DC
,
Pogach
LM
.
Assessing potential glycemic overtreatment in persons at hypoglycemic risk
.
JAMA Intern Med
2014
;
174
:
259
268
[PubMed]
35.
Quilliam
BJ
,
Simeone
JC
,
Ozbay
AB
,
Kogut
SJ
.
The incidence and costs of hypoglycemia in type 2 diabetes
.
Am J Manag Care
2011
;
17
:
673
680
[PubMed]
36.
Nunes
AP
,
Yang
J
,
L
, et al
.
Assessing occurrence of hypoglycemia and its severity from electronic health records of patients with type 2 diabetes mellitus
.
Diabetes Res Clin Pract
2016
;
121
:
192
203
[PubMed]
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.