Hypoglycemia Symptoms are Reduced in Hospitalized Patients with Diabetes

Abstract

Aims

Hospitalized patients with diabetes are have an impaired ability to detect hypoglycemia events. The purpose of this study was to compare hypoglycemia symptom scores (HSS) in hospitalized patients with diabetes after a documented blood glucose (BG) <70 mg/dl with recalled HSS with outpatient hypoglycemia events.

Methods

Non-critically ill hospitalized patients with diabetes grouped as symptomatic (n = 23) or asymptomatic (n = 32) at time of index hypoglycemia completed a standardized HSS-Questionnaires (HSS-Q) related to the inpatient event and to recall of symptoms with outpatient hypoglycemia.

Results

After controlling for BG at time of index hypoglycemia (49.8 ± 11.4 vs. 57.4 ± 6.8 mg/dl, p=0.02), symptomatic patients reported higher HSS than asymptomatic patients with the inpatient event (11.6 ± 7.3 vs. 1.5 ± 3.4, p < 0.001) and in the outpatient setting (13.9 ± 8.6 vs. 10.1 ± 10.6, p < 0.01). Recurrent hypoglycemia was more frequent in asymptomatic patients (13% vs. 44%, p = 0.015) during the hospitalization.

Conclusions

Compared to symptomatic patients, asymptomatic patients had lower inpatient and outpatient HSS and more frequent recurrent hypoglycemia events. These results suggest modification of glycemic management strategies in high risk patients to reduce risk for hypoglycemia events.

Introduction

Hypoglycemia is a frequently observed and undesirable complication of glucose lowering therapies in hospitalized patients with diabetes (1-3). Hypoglycemic events are associated with prolonged hospital length of stay, as well as an increased risk for adverse events including mortality (3,4). These events represent a major barrier to achieving recommended levels of glycemic control in hospitalized patients (5,6). In an attempt to avoid these events, many hospitals have implemented nurse driven hypoglycemia treatment protocols (HTP) that prompt early intervention as a way of minimizing the duration and severity of hypoglycemia events (7,8). These HTP rely on routine point of care blood glucose (POC-BG) testing or patient report of hypoglycemia symptoms in order to prompt an intervention(7).

Many hospitalized patients do not reliably experience hypoglycemia symptoms. This can be attributed to pre-admission impaired awareness of hypoglycemia or factors related to the acute illness itself (9-11). Use of sedating medications can contribute to an impaired ability of patients to alert nursing personnel of a hypoglycemia event, increasing vulnerability for these events in the inpatient setting.

This frequency of undetected hypoglycemia in hospitalized patients has been demonstrated in studies using continuous glucose monitoring (CGM) devices which have revealed a greater frequency of hypoglycemia events above what is identified using POC-BG monitoring (9,12,13). This can lead to delays in recognition and treatment of hypoglycemic events with potential deterioration from mild to more severe hypoglycemia events.

There are several reported initiatives that rely on predictive modeling of data available in an electronic health record (EHR) to identify patients who may be at high risk for hypoglycemia in the hospital setting(14) (15). CGM devices have the potential to reduce frequency and time spent in hypoglycemia by providing alerts at levels above the hypoglycemia threshold that prompt early treatment (16,17). However, CGM devices are not currently approved for use in the inpatient setting, and to date few hospitals have implemented this technology (12,18). POC-care BG monitoring by nursing personnel performed prior to meals and bedtime in patients who are eating or every 4-6 hours in patients who not eating regular meals remains the standard of care for monitoring BG levels in hospitalized patients(5,6).

There is little data available regarding hypoglycemia symptoms in the hospitalized patient population with diabetes (10). The purpose of this pilot study was to compare hypoglycemia symptoms scores (HSS) in hospitalized patients following a hypoglycemia event (defined as BG < 70 mg/dl) and compare these to HSS from recalled outpatient events.

Research Design and Methods

This observational study was approved by the Institutional Review Board at the University of Pittsburgh. All patients provided written informed consent prior to participation.

Study Population

Mentally competent non-pregnant, non-critically ill hospitalized patients with established diabetes ages 18 to 90 years with an anticipated length of stay ≥3 days and life expectancy >1 year were recruited for this study. Patients were identified by computerized daily reports of hypoglycemic events (BG < 70 mg/dl) that occurred during the preceding 24 hour time period. A member of the study team confirmed that a hypoglycemic event had occurred through chart review. Exclusion criteria included patients with evidence of dementia or delirium in the EHR; primary admission diagnosis of hypoglycemia or diabetes ketoacidosis (DKA); or receiving total parenteral nutrition (TPN) or intravenous (IV) doses of narcotic medications.

Study Design

Once chart review confirmation of the hypoglycemia event was obtained, a member of the study team approached the patient to review the study and obtain written informed consent. Once consent procedure was obtained, participants were asked to complete two standardized HSS Questionnaires (HSS-Q): one to quantitate autonomic (A-HSS), neuroglycopenic (N-HSS) and Total HSS (T-HSS) related to the index event and one based on recall of symptoms experienced with hypoglycemia in the outpatient setting (11,19,20). Results were quantified using a LIKERT Scale ranging from 0 (none) to 5 (very severe) for A-HSS (sweating, palpitations, tremor, hunger, anxiety, irritability) and N-HSS (difficulty in thinking, faintness, dizziness, tingling, blurred vision) symptoms. Patients were also asked to complete an eight-item modified questionnaire to obtain information regarding the frequency and awareness of hypoglycemia events at home (21).

The following demographic and clinical data were obtained from chart review and patient interviews: age, sex, type and duration of diabetes, reason for admission, most recent HbA1c, blood pressure (BP), heart rate, diabetes-related macrovascular (CAD, CVA, PVD), microvascular (retinopathy, neuropathy and nephropathy) complications, medications (including narcotics, beta blockers and steroids), transaminase levels, creatinine levels, history of kidney disease or dialysis therapy, albumin, hematocrit, length of stay (LOS), all bedside point of care blood glucose (POC-BG) values in 48 hour period prior to and following the index event, treatment used for the hypoglycemic event, as well as glucose lowering medications. A Charlson Comorbidity Index (CCI) was calculated for all participants (22). Monofilament (MF) testing was performed as a measure of peripheral neuropathy.

Data Analysis / Definitions

Patients were grouped as having symptomatic or asymptomatic hypoglycemia based on their response to the question: “Are you aware that you experienced a hypoglycemia event?” Patients who answered yes were identified as having symptomatic hypoglycemia and those who reported only being aware when treatment was provided as having asymptomatic hypoglycemia. HSS were determined as A-HSS, N-HSS, and T-HSS for both the index event and recall of symptoms in the outpatient setting. The maximum T-HSS was 55 with 30 for autonomic symptoms and 25 for neuroglycopenic symptoms. Higher scores represented more severe symptoms. The modified Clarke Questionnaire identified patients in both groups with ≥ 4 responses as having impaired awareness of hypoglycemia (21).

Statistical Analysis

Continuous variables were described by means and standard deviations. Categorical variables were described by frequencies and percentages. For comparisons between symptomatic and asymptomatic patients, the nonparametric Mann-Whitney U test was used for continuous variables, and either the chi-squared or Fisher’s exact test was used for categorical variables, as appropriate. For analysis of symptom scores, regression analysis was performed for T-HSS, A-HSS, and N-HSS, adjusting for index BG. IBM SPSS 22 was used for all statistical analysis (IBM Corp., Armonk, NY). Statistical significance was set at two-tailed p-value < 0.05.

Results

Of 265 patient charts that were reviewed for the study, 105 patients met inclusion criteria and 55 patients agreed to participate. Among consented participants, 23 (42%) experienced symptomatic and 32 (58%) asymptomatic hypoglycemia (Table 1). There were no differences between symptomatic and asymptomatic patients for age, sex, ethnicity, basal insulin dose, HbA1c, duration or type of diabetes, BP, heart rate, creatinine, albumin, hematocrit, CCI, or LOS (Table 1). The reasons for hospital admission are included in Supplemental Table 1. T Blood glucose levels at time of the index hypoglycemia event were higher in asymptomatic patients (49.8 ± 11.4 vs. 57.4 ± 6.8 mg/dl, p=0.02). (Table 1).

Table 1.

Clinical Characteristics of Study Participants

	Symptomatic Hypoglycemia	Asymptomatic Hypoglycemia	P Value
Number of Patients	23	32
Age (years)	60.8 ± 2.3	62.2 ± 2.1	0.65
Sex (% Male)	48%	41%	0.60
Ethnicity (% white)	82.6	84.4	1.0
DM Duration (years)	23.0 ± 2.5	18.8 ± 2.4	0.15
Type 2 DM (%)	87%	88%	1.0
HbA1c (%)	7.5 ± 0.3	8.3 ± 0.4	0.27
HbA1c (mmol/mol)	58 ± 3.3	67 ± 4.4
Index BG (mg/dl)	49.8 ± 11.4	57.4 ± 6.8	0.02
Basal Insulin Dose(units/kg)	0.35 ± 0.04	0.50 ± 0.10	0.34
Insulin regimen			0.62
Basal bolus	22%	19%
Basal + correction insulin	70%	78%
Correction insulin alone	8%	3%
Microvascular Complications (#)	1.9 ± 0.9	1.3 ± 1.0	0.024
Macrovascular Complications (#)	1.0 ± 1.0	0.8 ± 0.8	0.41
DBP (mm Hg)	64.3 ± 2.8	67.7 ± 2.5	0.34
Heart Rate	77.0 ± 2.2	83.5 ± 2.8	0.20
Creatinine (mg/dl)	1.88 ± 1.6	1.67 ± 1.2	0.60
HCT (%)	31.2 ± 1.0	30.6 ± 0.9	0.4
Albumin g/dl	3.3 ± 0.1	3.3 ± 0.1	0.99
LOS (days)	14.8 ± 8.8	14.6 ± 15.6	0.96
CCI	5.8 ± 0.50	5.8 ± 0.49	0.7
Education Level	28.6%	6.7%	0.05
% ≥ College Degree

BG = Blood glucose; CCI = Charlson Comorbidity Index; DBP = diastolic blood pressure; SBP = systolic blood pressure; HCT = hematocrit; LOS = Length of stay

Among patients who reported experiencing hypoglycemia events at home prior to admission (23 symptomatic vs. 26 asymptomatic), there was a trend toward a higher percentage of asymptomatic patients who met criteria for impaired awareness of hypoglycemia (13% vs. 38%, p = 0.054). More symptomatic patients (55% vs. 27%, p = 0.051) reported that they experienced hypoglycemia on a daily or weekly basis in the outpatient setting. The majority of patients in each group (91% vs. 73%) reported that they always or often felt symptoms of hypoglycemia when their BG was low, with a minority of patients in each group (18% vs.33%) reporting that they do not experience symptoms until their BG is < 54 mg/dl.

The majority of patients in both groups were receiving basal insulin in combination with correction sliding scale insulin (SSI) (23,24). There were two asymptomatic patients who received non-insulin therapies. One patient received glimepiride 5 mg in combination with correction SSI and one patient received repaglinide 0.5 mg before meals. Correction SSI was prescribed for 87% of patients in both groups, with low dose correction SSI being prescribed most frequently for both symptomatic (60%) and asymptomatic (54%) patients.

Symptomatic patients had more microvascular (p = 0.024) but not macrovascular complications (Table 1). There were no significant group differences in prevalence of individual microvascular or macrovascular complications (Supplemental Table 2). There was no difference in the percentage of symptomatic or asymptomatic patients with impaired or absent MF testing (47.4 vs. 27.6%, p = 0.16). A similar percentage of patients in each group were receiving glucocorticoid therapy (21.7 vs. 18.8 %, p = 1.0) or beta blockers (69.6 vs. 65.6%, p = 0.76).

The majority of hypoglycemic events were attributed to inappropriate timing of insulin administration with meals (46%), followed by increases in serum creatinine (12%), stacking of insulin doses (6%), error in insulin dosing (2%), or discontinuation of oral intake (2%). No cause was identified for the remaining 17 patients. The majority of events were treated according to the hospital HTP with oral caloric nutrition (82%) (7). Three patients (1 symptomatic, 2 asymptomatic) received intravenous dextrose. The therapeutic intervention for the hypoglycemia event was not documented for seven participants.

Adjusting for index BG, symptomatic patients reported higher T-HSS (11.6 ± 7.3 vs 1.5 ± 3.4), A-HSS (7.1 ± 7.0 vs. 0.5 ± 1.7), and N-HSS (4.4 ± 4.5 vs. 1.0 ± 2.1, all p < 0.001) associated with the index hypoglycemic event than asymptomatic patients (Figure 1).

Figure 1:

Consort diagram for patient recruitment

These differences in symptom scores remained significant when examining patients according to an index BG < 55 mg/dl and ≥ 55 mg/dl (25) (Table 2). T-HSS and A-HSS but not N-HSS were significantly higher in symptomatic patients with BG values < 55 mg/dl. T-HSS, A-HSS, and N-HSS were all higher in symptomatic patients with an index BG ≥ 55 mg/dl and < 70 mg/dl.

Table 2:

Total, Autonomic, and Neuroglycopenic Hypoglycemia Symptom Scores in Patients Grouped According to Index Hypoglycemia BG < 55 mg/dl and ≥ 55 mg/dl

Patients with index BG < 55 mg/dl
	Symptomatic N = 13	Asymptomatic N = 12	P Value
Total HSS	9.31 ± 6.51	2.00 ± 2.90	0.002
Autonomic HSS	5.92 ± 4.46	0.58 ± 1.5	<0.001
Neuroglycopenic HSS	3.38 ± 4.07	1.42 ± 2.11	0.247
Patients with index BG > 55 and < 70 mg/dl
	Symptomatic N = 10	Asymptomatic N = 20	P Value
Total HSS	14.50 ± 7.44	1.25 ± 3,70	<0.001
Autonomic HSS	8.70 ± 4.96	0.80 ± 2.12	<0.001
Neuroglycopenic HSS	5.80 ± 4.07	1.42 ± 2.11	0.007

When comparing within group differences between inpatient and recalled outpatient HSS, symptomatic patients reported higher outpatient A-HSS (9.3 ± 5.9, p = 0.055), with similar N-HSS (4.6 ± 4.5) and T-HSS (13.9 ± 8.6) (Figure 1). Asymptomatic patients reported higher outpatient A-HSS (5.3 ± 6.4), N-HSS (4.8 ± 5.2) and T-HSS (10.1 ± 10.6, all p < 0.01) compared to the inpatient setting (Figure 1).

There were no significant group differences in mean, nadir or peak BG during the 48 hour period preceding and following the index hypoglycemia event (Table 3). More asymptomatic patients experienced ≥ 1 recurrent hypoglycemic event in the 48 hour time period following the index hypoglycemic event (13% vs. 44%, p = 0.015). The range of low BG during this 48 hour time period was 57-63 mg/dl in symptomatic patients and 27-68 mg/dl in asymptomatic patients.

Table 3.

Mean Blood Glucose Values Preceding and Following Hypoglycemia

	Relation to Index Event	Symptomatic Patients	Asymptomatic Patients	P
Mean BG (mg/dl)	48 hours before	183 ± 57	163 ± 56	0.17
Mean BG (mg/dl)	48 hours after	169 ± 211	153 ± 40	0.23
Nadir BG (mg/dl)	48 hours before	98 ± 32	91 ± 45	0.24
Nadir BG (mg/dl)	48 hours after	77 ± 20	72 ± 25	0.25
Peak BG (mg/dl)	48 hours before	246 ± 89	227 ± 93	0.24
Peak BG (mg/dl)	48 hours after	245 ± 92	238 ± 93	0.75

BG = Blood glucose

Discussion

The majority of patients in this report experienced asymptomatic hypoglycemia while hospitalized. These patients reported lower A-HSS, N-HSS and T-HSS than what was observed in symptomatic patients with the index hypoglycemia event. Differences in HSS remained when examining patients with index hypoglycemia BG values <55 mg/dl and ≥ 55 mg/dl (Table 2). Hypoglycemia symptoms were experienced with a lower intensity among both groups while hospitalized, with more profound differences in those with asymptomatic inpatient hypoglycemia. These findings support prior suggestions that hospitalized patients may experience hypoglycemia differently in the inpatient compared to the outpatient setting (9,10,24). To our knowledge, this is the first report documenting significant differences in HSS between inpatient and outpatient settings. The lower inpatient A-HSS in both groups, and lower N-HSS in the asymptomatic group suggests that hospitalization or acute illness itself may be a risk factor for impaired awareness of hypoglycemia as has been previously suggested (9) (10).

A small percentage of patients met criteria for impaired awareness of hypoglycemia using a modified version of a previously validated questionnaire (21). When questions on this survey were examined individually, the majority of patients in both groups reported that they “always” or “often” feel symptoms of hypoglycemia at home. However, 27% of patients in the asymptomatic group reported that they did not reliably perceive symptoms at the time of a low BG, and 33% reported that they did not experience symptoms with BG values < 54 mg/dl (25).

These results may provide support for implementing a process that identifies patients at high risk for hypoglycemia events at time of hospitalization (14,15,17). One approach would be to use a simplified or modified HSS-Q that could potentially prompt de-intensification of inpatient glycemic goals toward the higher range of current recommendations as a way of minimizing risk for these events (5,6). Others have used predictive modeling methods using existing patient data in an EHR (14).

There is growing interest in the use of CGM systems for patients identified as being at high risk for hypoglycemia (12,17). Several small studies comparing CGM with POC BG testing in the inpatient setting demonstrated improved detection rates and reduced time spent with hypoglycemia in high-risk patients, especially nocturnal hypoglycemia (17). The accuracy of CGM devices is higher with BG levels between 70 and 250 mg/dl, and lower with BG levels outside these ranges. Setting alarms for BG at levels above hypoglycemia ranges to 80 mg/dl has been demonstrated to reduce the frequency of hypoglycemia events as well as time spent in hypoglycemia (17). Potential barriers to the widespread use of these devices include higher initial cost associated with implementation and educational requirements for hospital personnel (12,18).

Both symptomatic and asymptomatic participants in this study had longstanding diabetes with evidence of chronic kidney disease, both of which are known contributors to risk for hypoglycemia (Table 1). In an attempt to identity risk factors for asymptomatic hypoglycemia in this study, no group differences were observed for age, sex, HbA1c, serum creatinine, duration or type of diabetes, CCI, basal insulin dose, or LOS. These results differ from what has been observed in the outpatient setting where intensive glycemic management, lower HbA1c, and longer duration of diabetes have all been associated impaired awareness of hypoglycemia (14,26,27). Basal insulin doses >0.6 units/kg day have also been identified as a risk factor for inpatient hypoglycemia (14,27). The basal insulin doses observed in this study were lower than what has been reported to pose risk for inpatient hypoglycemia (Table 1). Of interest, more symptomatic patients held a college degree, possibly suggesting that these patients may have been more knowledgeable regarding symptoms of hypoglycemia (28,29)

A prior history of hypoglycemia occurring during hospitalization has been associated with both recurrent hypoglycemia and hyperglycemia preceding and following the event (personal observations) (23). Both hyperglycemia and hypoglycemia occurred frequently among participants in this study during the 48 hour time period following the hypoglycemia event. Recurrent hypoglycemia was observed in 13% symptomatic and 44% of asymptomatic patients. There were no group differences in 24 or 48 hour nadir, peak or mean BG levels preceding or following the index event in this study (Table 3). Of note, the nadir BG was lower in both groups in the 48 hour time period following index hypoglycemia when compared to those preceding the event (Table 3).

There are important limitations to this study. One is that the sample size is small which makes the ability to generalize these findings to the larger inpatient population with diabetes. However, this report confirms the high prevalence of asymptomatic hypoglycemia observed among hospitalized patients in an earlier report (10). Another limitation is that information about the specific doses of oral carbohydrate or parenteral glucose used to treat the hypoglycemia event was not collected. The questionnaires used to obtain HSS relied on patient recall of events that occurred up to 24 hours following the index event and even longer following recalled home events which may have impacted the accuracy of the patient reports. The specific timing of hypoglycemia events occurring in the outpatient setting was not obtained which could have led to inaccurate reporting of symptoms relating to outpatient events. Variability in accuracy of patient reporting of hypoglycemia symptoms has been previously described (30). However, this limitation was equally distributed between the two groups and cannot alone explain the significantly lower inpatient and outpatient scores obtained in asymptomatic patients. The average BG defining index hypoglycemia was higher in asymptomatic patients which may have contributed to the lower HSS in this group.

In summary, this study provides information regarding the inability to rely on patient reports of symptoms to be able to detect and treat hypoglycemia events in a timely manner in the inpatient setting. Differences in HSS between inpatient and outpatient settings, particularly for A-HSS emphasize the need for identification of patients at risk for hypoglycemia as well as vigilance for these events among all insulin treated patients (31). More than 50% of patients did not recognize that they experienced hypoglycemia, putting them at high risk for prolonged and recurrent events, the latter which was also observed in this report. CGM may offer a potential way of addressing these concerns in the future (12,17). However, until these devices are approved for inpatient use and more widely available, POC BG testing coupled with heightened vigilance for patients determined as being at high risk for hypoglycemia remain the currently available best practice strategies for minimizing these events.

Figure 2:

Hypoglycemia Symptom Scores in Symptomatic and Asymptomatic Patients

* P < 0.01 vs corresponding scores in symptomatic group (unadjusted data)

Study Highlights:

• Hypoglycemia is a frequently observed and undesirable complication of glucose lowering therapies in hospitalized patients with diabetes

• Patients in the hospital are reported to experience hypoglycemia symptoms differently than what they experience in the outpatient setting

• This reports describes reduced autonomic hypoglycemia symptoms in patients with symptomatic and asymptomatic hypoglycemia following an inpatient hypoglycemia event < 70 mg/dl

• Recurrent hypoglycemia was observed more frequently among patients experiencing asymptomatic hypoglycemia during a hospital stay

• Modification of glucose lowering therapies may reduce recurrent hypoglycemia among patients identified as being at high risk for these events