TY - JOUR
AU - Winterstein, Almut, G
AB - Abstract Purpose Common inpatient hypoglycemia risk factor patterns (phenotypes) from an electronic health record (EHR)-based prediction model and preventive strategies were identified. Methods Patients admitted to 2 large academic medical centers who were in the top fifth percentile of a previously developed hypoglycemia risk score and developed hypoglycemia (blood glucose [BG] of <50mg/dL) were included in the study. Frequencies of all combinations of ≥4 risk factors contributing to the risk score among these patients were determined to identify common risk patterns. Clinical pharmacists developed clinical vignettes for each common pattern and formulated medication therapy management recommendations for hypoglycemia prevention. Results A total of 401 admissions with a hypoglycemic event were identified among 1,875 admissions whose hypoglycemic risk was in the top fifth percentile among all admissions that received antihyperglycemic drugs and evaluated. Five distinct phenotypes emerged: (1) frail patients with history of hypoglycemia receiving insulin on hospital day 1, (2) a rapid downward trend in BG values in patients receiving an insulin infusion or with a history of hypoglycemia, (3) administration of insulin in the presence of an active nothing by mouth order in frail patients, (4) repeated low BG level in frail patients, and (5) inadequate night-time BG monitoring for patients on long-acting insulin. The 5 themes jointly described 53.0% of high-risk patients who experienced hypoglycemia. Conclusion Five distinct phenotypes that are prevalent in patients at greatest risk for inpatient hypoglycemia were identified. hypoglycemia, prevention, risk factor, risk prediction KEY POINTS • An electronic health record–based automated risk score was used to characterize clinical risk factor patterns for hypoglycemia. • All possible combinations of 4 risk factors that made the top 5 contributions to the overall predicted risk for inpatient hypoglycemia were considered. • Five distinct risk factor patterns for hypoglyemia were identified, explaining more than half of hypoglycemia cases that occurred in admissions at greatest risk of hypoglycemia. For several decades, hypoglycemia has been one of the most prominent patient safety issues involving inpatient medication therapy management.1–4 Glucose control is challenging not only because of the narrow therapeutic range of antihyperglycemic agents but also due to the oftentimes rapidly changing risk factors for uncontrolled glucose concentrations in acute and intensive care settings and the significant impact of both hypoglycemia and hyperglycemia on patient outcomes. Hypoglycemia can result in immediate severe sequelae such as seizures and coma,5,6 but the subtler effects of hyperglycemia on morbidity and mortality7,8 have gained increasing appreciation. The resulting intensified insulin regimen may elevate the risk for hypoglycemia.9,10 As part of a national effort to reduce patient harm, hypoglycemia was selected as one of the top 3 adverse drug events (ADEs) to address.11,12 Hypoglycemia has been prioritized among patient safety initiatives by multiple institutions, including the Health Resources and Services Administration, the Centers for Medicare and Medicaid Services, and the National Quality Forum.13 The National Quality Strategy Priorities have identified opportunities for hypoglycemia prevention in inpatient settings, highlighting periodical risk assessment of individual patients as well as standardized and evidence-based use of antihyperglycemic agents.12 Indeed, previous studies have reported that the vast majority of severe hypoglycemic events occurred secondary to the use of antihyperglycemic medications, accompanied by inadequate management of caloric intake and other emerging risk factors (e.g., impaired renal function, rapid changes in BG levels).14–16 As such, hypoglycemia follows a comparably simple etiology involving an imbalance between drug dose and caloric intake, but accurate prediction of when and how this imbalance occurs has remained a significant clinical challenge. As part of a larger effort aimed at developing automated ADE risk scores based on electronic health record (EHR) data,17 we developed a risk score for drug-induced hypoglycemia in hospitalized patients.18 In reviewing risk-score results— specifically, the risk factor patterns that contributed to the highest risk scores—we noticed a distinct, recurrent risk factor pattern, which may be amenable to distinct clinical recommendations for intervention. Thus, in order to support both the recognition of high-risk patients and to offer practical recommendations for intervention, we aimed to (1) formally identify clinically meaningful clusters of hypoglycemia risk factors that could aid in recognizing patients or “risk factor phenotypes” in need of drug therapy management and (2) to provide targeted phenotype-specific recommendations for clinical interventions. Methods Study design and patient population. We conducted our analysis in a retrospective cohort of adult patients admitted to 2 large academic tertiary care hospitals affiliated with the University of Florida (UF) between January 1, 2012, and October 31, 2013 (UF Health Shands Hospital), and between March, 1, 2013, to October, 31, 2013 (UF Health Jacksonville). We included patients who experienced hypoglycemic events during hospital days 2–6 whose predicted risk for hypoglycemia was in the top fifth percentile among all patients included in risk scoring.18 In brief, the risk score was derived from a prediction model that uses 34 risk factors retrieved from discrete fields in the EHR to evaluate all adult inpatients who received any antihyperglycemic agent except α-glucosidase inhibitors. We defined a hypoglycemic event as a single blood glucose (BG) concentration of <50 mg/dL that was not followed by another concentration of >80 mg/dL within 10 minutes. This definition closely reflected “clinically significant hypoglycemia,” defined as a BG concentration of <54 mg/dL) in the classification of hypoglycemia recommended by the International Hypoglycemia Study Group.19 Admissions at high risk of hypoglycemia (the top fifth risk score percentile) represented 38.8% of all admissions with hypoglycemic events (487 of 1,256) that occurred in the entire risk-score cohort.18 The frequency of hypoglycemia observed in the top fifth percentile of the risk score was 16%. Risk factors that were considered in risk-model development included previous BG control and related trends, antihyperglycemic medication regimen and recent changes, exposure to other drugs with hypoglycemic risk, dietary restrictions including nothing by mouth (NPO) orders, and clinical characteristics such as renal function. Risk pattern identification. To identify risk factor patterns associated with hypoglycemia, we extracted risk score details for all admission-days that ranked in the top fifth percentile of all admissions and days for which a risk score was available and was followed by hypoglycemia. Using this subset of admission-days with high risk score and manifestation of hypoglycemia, we generated frequency distributions of all combinations of 4 risk factors that were among the top 5 risk factors that contributed to the total risk score. We did this separately for each of the first 5 hospitals days because we found that risk factor contributions to hypoglycemia risk may change over the course of a patient’s hospitalization. The risk factors that were considered for each hospital day are summarized in the appendix. Two clinical pharmacists reviewed all risk factor combinations with a frequency of at least 10, created patient vignettes to provide clinical context, and provided recommendations for clinical intervention. All data management and statistical analyses were performed using SAS version 9.4 (SAS, Cary, NC). The study was approved by the University of Florida institutional review and privacy boards. Results We collected EHR data from 65,518 admissions to UF Health Shands and 18,269 admissions to UF Health Jacksonville. During the first 5 hospital days, we identified 1,875 admissions (contributing to 3,078 days) during which an antihyperglycemic agent (except α-glucosidase inhibitors) was administered and were in the top fifth percentile of the risk score. Of these 1,875 admissions, we included 401 admissions with a hypoglycemic event (contributing to 487 event-days as one admission can have multiple hypoglycemic events). After each risk prediction day (hospital day 1–5), 96, 143, 117, 65, and 66 events occurred respectively. Most patients included in this high-risk group received long-acting insulin (81.9%) and had a history of hypoglycemia (75.6%). The median maximum BG value on the prediction day was 268 mg/dL (interquartile range [IQR], 177–374 mg/dL), and the median minimum BG value was 55 mg/dL (IQR, 43–73 mg/dL) (Table 1). The majority of patients had a preceding hospital day with at least 1 BG value of <70 mg/dL and 36% had type 1 diabetes mellitus. Table 1. Demographics and Hypoglycemia Risk Factors of Patients in the Top Fifth Hypoglycemia Risk Percentile Who Developed Hypoglycemia (n = 487)a Characteristicb Value Mean ± S.D. Median (Q1, Q3) Predicted probability of hypoglycemia, % 21.4 ± 12.7 16.6 (12.2, 26.7) Male, no. (%) 215 (44.2) Age, yr 51.8 ± 17.9 53.0 (39.0, 64.0) Most recent temperature, °C 36.7 ± 0.5 36.7 (36.4, 37.0) Weight, kg 72.5 ± 19.0 70.0 (61.0, 84.0) Admission BG conc., mg/dL 256.6 ± 166.3 218.0 (133.0, 348.0) BG conc. max, mg/dL 289.4 ± 147.1 268.0 (177.0, 374.0) BG conc. min, mg/dL 66.6 ± 41.7 55.0 (43.0,73.0) Ratio max:min BG conc. 5.4 ± 6.0 4.3 (2.9, 6.5) Mean BG conc. change/hr -1.6 ± 10.3 -0.4 (-3.8, 2.2) No. prior hospital days with BG conc. < 70mg/dL 1.3 ± 1.1 1.0 (0.0,2.0) Insulin infusion, no. (%) 66 (13.6) Long-acting insulin, no. (%) 399 (81.9) Long-acting insulin/no BG conc. check at PMc, no. (%) 10 (2.1) Subcutaneous insulin > 1 unit/kg/day, no. (%) 89 (18.3) Sulfonylurea, no. (%) 45 (9.2) No. high risk nonantidiabetic drugs 1.0 ± 1.0 1.0 (0.0, 2.0) New antidiabetic agent added, no. (%) 49 (10.1) Diabetic regimen change, no. (%) 107 (22.0) Dextrose 50% injection administered, no (%) 153 (31.4) Active NPO order, no. (%) 82 (16.8) NPO and long-acting insulin used, no. (%) 48 (9.9) Albumin conc., g/dLd 3.2 ± 0.7 3.2 (2.7, 3.7) CLcr, mL/min 61.0 ± 32.7 62.0 (29.6,100.0) Type 1 diabetes mellitus, no. (%) 178 (36.6) History of outpatient hypoglycemia, no. (%) 368 (75.6) Characteristicb Value Mean ± S.D. Median (Q1, Q3) Predicted probability of hypoglycemia, % 21.4 ± 12.7 16.6 (12.2, 26.7) Male, no. (%) 215 (44.2) Age, yr 51.8 ± 17.9 53.0 (39.0, 64.0) Most recent temperature, °C 36.7 ± 0.5 36.7 (36.4, 37.0) Weight, kg 72.5 ± 19.0 70.0 (61.0, 84.0) Admission BG conc., mg/dL 256.6 ± 166.3 218.0 (133.0, 348.0) BG conc. max, mg/dL 289.4 ± 147.1 268.0 (177.0, 374.0) BG conc. min, mg/dL 66.6 ± 41.7 55.0 (43.0,73.0) Ratio max:min BG conc. 5.4 ± 6.0 4.3 (2.9, 6.5) Mean BG conc. change/hr -1.6 ± 10.3 -0.4 (-3.8, 2.2) No. prior hospital days with BG conc. < 70mg/dL 1.3 ± 1.1 1.0 (0.0,2.0) Insulin infusion, no. (%) 66 (13.6) Long-acting insulin, no. (%) 399 (81.9) Long-acting insulin/no BG conc. check at PMc, no. (%) 10 (2.1) Subcutaneous insulin > 1 unit/kg/day, no. (%) 89 (18.3) Sulfonylurea, no. (%) 45 (9.2) No. high risk nonantidiabetic drugs 1.0 ± 1.0 1.0 (0.0, 2.0) New antidiabetic agent added, no. (%) 49 (10.1) Diabetic regimen change, no. (%) 107 (22.0) Dextrose 50% injection administered, no (%) 153 (31.4) Active NPO order, no. (%) 82 (16.8) NPO and long-acting insulin used, no. (%) 48 (9.9) Albumin conc., g/dLd 3.2 ± 0.7 3.2 (2.7, 3.7) CLcr, mL/min 61.0 ± 32.7 62.0 (29.6,100.0) Type 1 diabetes mellitus, no. (%) 178 (36.6) History of outpatient hypoglycemia, no. (%) 368 (75.6) aQ1 = quartile 1, Q3 = quartile 3, BG = blood glucose, max = maximum, min= minimum, NPO= nothing by mouth, CLcr = creatinine clearance. bIf not otherwise indicated, the risk factor was measured on the prediction day. cPatient received long-acting insulin after 8 a.m., and no BG value available after 6 p.m. dMean ± S.D. of albumin concentration was calculated after excluding missing values. Mean ± SD of CLcr was calculated after treating values ≥ 100 as 100. Most recent CLcr estimated via the Cockcroft-Gault equation; the mean serum creatinine concentration was used for calculation when several values were available for 1 day.18 View Large Table 1. Demographics and Hypoglycemia Risk Factors of Patients in the Top Fifth Hypoglycemia Risk Percentile Who Developed Hypoglycemia (n = 487)a Characteristicb Value Mean ± S.D. Median (Q1, Q3) Predicted probability of hypoglycemia, % 21.4 ± 12.7 16.6 (12.2, 26.7) Male, no. (%) 215 (44.2) Age, yr 51.8 ± 17.9 53.0 (39.0, 64.0) Most recent temperature, °C 36.7 ± 0.5 36.7 (36.4, 37.0) Weight, kg 72.5 ± 19.0 70.0 (61.0, 84.0) Admission BG conc., mg/dL 256.6 ± 166.3 218.0 (133.0, 348.0) BG conc. max, mg/dL 289.4 ± 147.1 268.0 (177.0, 374.0) BG conc. min, mg/dL 66.6 ± 41.7 55.0 (43.0,73.0) Ratio max:min BG conc. 5.4 ± 6.0 4.3 (2.9, 6.5) Mean BG conc. change/hr -1.6 ± 10.3 -0.4 (-3.8, 2.2) No. prior hospital days with BG conc. < 70mg/dL 1.3 ± 1.1 1.0 (0.0,2.0) Insulin infusion, no. (%) 66 (13.6) Long-acting insulin, no. (%) 399 (81.9) Long-acting insulin/no BG conc. check at PMc, no. (%) 10 (2.1) Subcutaneous insulin > 1 unit/kg/day, no. (%) 89 (18.3) Sulfonylurea, no. (%) 45 (9.2) No. high risk nonantidiabetic drugs 1.0 ± 1.0 1.0 (0.0, 2.0) New antidiabetic agent added, no. (%) 49 (10.1) Diabetic regimen change, no. (%) 107 (22.0) Dextrose 50% injection administered, no (%) 153 (31.4) Active NPO order, no. (%) 82 (16.8) NPO and long-acting insulin used, no. (%) 48 (9.9) Albumin conc., g/dLd 3.2 ± 0.7 3.2 (2.7, 3.7) CLcr, mL/min 61.0 ± 32.7 62.0 (29.6,100.0) Type 1 diabetes mellitus, no. (%) 178 (36.6) History of outpatient hypoglycemia, no. (%) 368 (75.6) Characteristicb Value Mean ± S.D. Median (Q1, Q3) Predicted probability of hypoglycemia, % 21.4 ± 12.7 16.6 (12.2, 26.7) Male, no. (%) 215 (44.2) Age, yr 51.8 ± 17.9 53.0 (39.0, 64.0) Most recent temperature, °C 36.7 ± 0.5 36.7 (36.4, 37.0) Weight, kg 72.5 ± 19.0 70.0 (61.0, 84.0) Admission BG conc., mg/dL 256.6 ± 166.3 218.0 (133.0, 348.0) BG conc. max, mg/dL 289.4 ± 147.1 268.0 (177.0, 374.0) BG conc. min, mg/dL 66.6 ± 41.7 55.0 (43.0,73.0) Ratio max:min BG conc. 5.4 ± 6.0 4.3 (2.9, 6.5) Mean BG conc. change/hr -1.6 ± 10.3 -0.4 (-3.8, 2.2) No. prior hospital days with BG conc. < 70mg/dL 1.3 ± 1.1 1.0 (0.0,2.0) Insulin infusion, no. (%) 66 (13.6) Long-acting insulin, no. (%) 399 (81.9) Long-acting insulin/no BG conc. check at PMc, no. (%) 10 (2.1) Subcutaneous insulin > 1 unit/kg/day, no. (%) 89 (18.3) Sulfonylurea, no. (%) 45 (9.2) No. high risk nonantidiabetic drugs 1.0 ± 1.0 1.0 (0.0, 2.0) New antidiabetic agent added, no. (%) 49 (10.1) Diabetic regimen change, no. (%) 107 (22.0) Dextrose 50% injection administered, no (%) 153 (31.4) Active NPO order, no. (%) 82 (16.8) NPO and long-acting insulin used, no. (%) 48 (9.9) Albumin conc., g/dLd 3.2 ± 0.7 3.2 (2.7, 3.7) CLcr, mL/min 61.0 ± 32.7 62.0 (29.6,100.0) Type 1 diabetes mellitus, no. (%) 178 (36.6) History of outpatient hypoglycemia, no. (%) 368 (75.6) aQ1 = quartile 1, Q3 = quartile 3, BG = blood glucose, max = maximum, min= minimum, NPO= nothing by mouth, CLcr = creatinine clearance. bIf not otherwise indicated, the risk factor was measured on the prediction day. cPatient received long-acting insulin after 8 a.m., and no BG value available after 6 p.m. dMean ± S.D. of albumin concentration was calculated after excluding missing values. Mean ± SD of CLcr was calculated after treating values ≥ 100 as 100. Most recent CLcr estimated via the Cockcroft-Gault equation; the mean serum creatinine concentration was used for calculation when several values were available for 1 day.18 View Large After reviewing all combinations of 4 or more risk factors, 5 distinct risk themes emerged (Table 2). These themes included (1) frail patients, defined as patients with low body weight and either low serum albumin level or low creatinine clearance (CLcr) value, with hypoglycemia in the year before hospitalization who received insulin on hospital day 1, (2) a rapid downward trend in BG values in patients receiving an insulin infusion or who had a history of hypoglycemia, (3) administration of insulin in the presence of an active NPO order in frail patients, (4) repeated low BG values in frail patients, and (5) inadequate nighttime BG value monitoring for patients receiving long-acting insulin. These 5 themes described jointly 53% of patients who experienced hypoglycemia after risk prediction between hospital days 1 and 5. The following sections further explore these 5 themes and provide representative patient cases. Table 2. Risk Factor Pattern and Prevalencea Clinical Theme Set of Risk Factors % Patients in the Top Fifth Risk Score Percentile Who Developed Hypoglycemia (n ) Frail patients with history of hypoglycemia receiving insulin on hospital day 1 (History of outpatient hypoglycemia OR use of 50% dextrose injection on hospital day 1) AND (low body weight OR low albumin conc. OR low CLcr) AND (use of long-acting insulin OR use of subcutaneous insulin > 1 unit/kg/day) on hospital day 1 60.4 (58 of 96 on hospital day1) Rapid downward trend in BG values in patients receiving insulin infusion or with history of hypoglycemia Insulin infusion AND (negative mean BG change per hour OR large ratio max:min BG prior day) on hospital day 1 OR (Negative mean BG conc. change per hour AND ≥1 prior days with BG conc. <70mg/dL) on hospital day 2–5 17.7 (17 of 96 on hospital day1) 14.3 (56 of 391 on hospital days 2–5) Administration of insulin in the presence of an active NPO order in frail patients (Administration of insulin during the active NPO order) AND (history of outpatient hypoglycemia OR use of long-acting insulin OR low body weight OR low CLcr) on hospital day 2 11.2 (16 of 143 on hospital day 2) Repeated low BG values in frail patients ≥1 prior days with BG conc. <70mg/dL AND (low body weight OR low CLcr) on hospital day 2–5 29.2 (114 of 391 on hospital days 2–5) Inadequate nighttime BG monitoring for patients on long-acting insulin Use of long-acting insulin after 8 a.m. and no BG value available after 6 p.m. AND (negative mean BG change per hour OR large ratio max:min BG conc. on prior day) on hospital day 2 3.5 (5 of 143 on hospital day 2) Clinical Theme Set of Risk Factors % Patients in the Top Fifth Risk Score Percentile Who Developed Hypoglycemia (n ) Frail patients with history of hypoglycemia receiving insulin on hospital day 1 (History of outpatient hypoglycemia OR use of 50% dextrose injection on hospital day 1) AND (low body weight OR low albumin conc. OR low CLcr) AND (use of long-acting insulin OR use of subcutaneous insulin > 1 unit/kg/day) on hospital day 1 60.4 (58 of 96 on hospital day1) Rapid downward trend in BG values in patients receiving insulin infusion or with history of hypoglycemia Insulin infusion AND (negative mean BG change per hour OR large ratio max:min BG prior day) on hospital day 1 OR (Negative mean BG conc. change per hour AND ≥1 prior days with BG conc. <70mg/dL) on hospital day 2–5 17.7 (17 of 96 on hospital day1) 14.3 (56 of 391 on hospital days 2–5) Administration of insulin in the presence of an active NPO order in frail patients (Administration of insulin during the active NPO order) AND (history of outpatient hypoglycemia OR use of long-acting insulin OR low body weight OR low CLcr) on hospital day 2 11.2 (16 of 143 on hospital day 2) Repeated low BG values in frail patients ≥1 prior days with BG conc. <70mg/dL AND (low body weight OR low CLcr) on hospital day 2–5 29.2 (114 of 391 on hospital days 2–5) Inadequate nighttime BG monitoring for patients on long-acting insulin Use of long-acting insulin after 8 a.m. and no BG value available after 6 p.m. AND (negative mean BG change per hour OR large ratio max:min BG conc. on prior day) on hospital day 2 3.5 (5 of 143 on hospital day 2) aFrail patients include those with low body weight AND (low albumin OR low creatinine clearance [CLcr]). BG = blood glucose, max= maximum, min = minimum, NPO= nothing by mouth, CLcr= creatinine clearance. View Large Table 2. Risk Factor Pattern and Prevalencea Clinical Theme Set of Risk Factors % Patients in the Top Fifth Risk Score Percentile Who Developed Hypoglycemia (n ) Frail patients with history of hypoglycemia receiving insulin on hospital day 1 (History of outpatient hypoglycemia OR use of 50% dextrose injection on hospital day 1) AND (low body weight OR low albumin conc. OR low CLcr) AND (use of long-acting insulin OR use of subcutaneous insulin > 1 unit/kg/day) on hospital day 1 60.4 (58 of 96 on hospital day1) Rapid downward trend in BG values in patients receiving insulin infusion or with history of hypoglycemia Insulin infusion AND (negative mean BG change per hour OR large ratio max:min BG prior day) on hospital day 1 OR (Negative mean BG conc. change per hour AND ≥1 prior days with BG conc. <70mg/dL) on hospital day 2–5 17.7 (17 of 96 on hospital day1) 14.3 (56 of 391 on hospital days 2–5) Administration of insulin in the presence of an active NPO order in frail patients (Administration of insulin during the active NPO order) AND (history of outpatient hypoglycemia OR use of long-acting insulin OR low body weight OR low CLcr) on hospital day 2 11.2 (16 of 143 on hospital day 2) Repeated low BG values in frail patients ≥1 prior days with BG conc. <70mg/dL AND (low body weight OR low CLcr) on hospital day 2–5 29.2 (114 of 391 on hospital days 2–5) Inadequate nighttime BG monitoring for patients on long-acting insulin Use of long-acting insulin after 8 a.m. and no BG value available after 6 p.m. AND (negative mean BG change per hour OR large ratio max:min BG conc. on prior day) on hospital day 2 3.5 (5 of 143 on hospital day 2) Clinical Theme Set of Risk Factors % Patients in the Top Fifth Risk Score Percentile Who Developed Hypoglycemia (n ) Frail patients with history of hypoglycemia receiving insulin on hospital day 1 (History of outpatient hypoglycemia OR use of 50% dextrose injection on hospital day 1) AND (low body weight OR low albumin conc. OR low CLcr) AND (use of long-acting insulin OR use of subcutaneous insulin > 1 unit/kg/day) on hospital day 1 60.4 (58 of 96 on hospital day1) Rapid downward trend in BG values in patients receiving insulin infusion or with history of hypoglycemia Insulin infusion AND (negative mean BG change per hour OR large ratio max:min BG prior day) on hospital day 1 OR (Negative mean BG conc. change per hour AND ≥1 prior days with BG conc. <70mg/dL) on hospital day 2–5 17.7 (17 of 96 on hospital day1) 14.3 (56 of 391 on hospital days 2–5) Administration of insulin in the presence of an active NPO order in frail patients (Administration of insulin during the active NPO order) AND (history of outpatient hypoglycemia OR use of long-acting insulin OR low body weight OR low CLcr) on hospital day 2 11.2 (16 of 143 on hospital day 2) Repeated low BG values in frail patients ≥1 prior days with BG conc. <70mg/dL AND (low body weight OR low CLcr) on hospital day 2–5 29.2 (114 of 391 on hospital days 2–5) Inadequate nighttime BG monitoring for patients on long-acting insulin Use of long-acting insulin after 8 a.m. and no BG value available after 6 p.m. AND (negative mean BG change per hour OR large ratio max:min BG conc. on prior day) on hospital day 2 3.5 (5 of 143 on hospital day 2) aFrail patients include those with low body weight AND (low albumin OR low creatinine clearance [CLcr]). BG = blood glucose, max= maximum, min = minimum, NPO= nothing by mouth, CLcr= creatinine clearance. View Large Theme 1: Frail patients with history of hypoglycemia receiving insulin on hospital day 1. Maintaining normoglycemia with insulin has been associated with improvements in morbidity, mortality, and hospital length of stay.20,21 Choosing the preferred agent and dose can be challenging on admission. Changing caloric intake, NPO status, acute reductions in renal function, fluid balance, fluctuating BG patterns, and several other factors must carefully be considered when dosing insulin. In addition, high insulin doses used to achieve strict BG control have been associated with increased severe hypoglycemia when compared with those used with less-aggressive BG control.22 This association is especially true in frail, lower-weight, insulin-dependent patients with type 1 diabetes who have little ability to compensate for insulin overdoses. Thus, glucose control in frail patients with a history of hypoglycemia warrants careful review of the prior-to-admission (PTA) regimen, considering all novel factors imposed by the acute care admission. Representative case. On the morning of hospital day 2, risk scoring using all risk factors of the previous day indicates that a patient with type I diabetes is at high risk for hypoglycemia. Key risk factors contributing to the high-risk score include a history of hypoglycemia, weight of 49 kg, albumin of 2.6, use of long-acting insulin, and use of high-dose (>1 unit/kg/day) subcutaneous insulin on hospital day 1 (Figure 1). The pharmacist reviews the patient’s chart and discovers that the patient’s long-acting (basal) insulin regimen was continued at the PTA dose of 45 units every evening. In addition, corrective short-acting insulin was ordered, with 2 doses administered after postmeal BG spikes in addition to the bedtime dose on hospital day 1. The pharmacist noted that this is a very “basal heavy” regimen (1.2 units/kg) with no complementary mealtime (prandial) insulin doses. The pharmacist recommends a basal insulin dose reduction of 30–40% and the addition of a low prandial dose (5–7 units) of short-acting insulin with subsequent dose adjustments based on response. In addition, the use of a corrective insulin dose at bedtime was discouraged. Figure 1. View largeDownload slide Representative patient case details for frail patients with history of hypoglycemia receiving insulin on hospital day 1 (theme 1). BG = blood glucose, max= maximum, min= minimum, CLcr = creatinine clearance. Figure 1. View largeDownload slide Representative patient case details for frail patients with history of hypoglycemia receiving insulin on hospital day 1 (theme 1). BG = blood glucose, max= maximum, min= minimum, CLcr = creatinine clearance. Theme 2: Rapid downward trend in BG levels in patients on insulin infusion or with prior history of hypoglycemia. Hypoglycemic risk may be underappreciated in patients admitted with highly elevated glucose levels or diabetic ketoacidosis (DKA). Aggressive use of insulin infusion resulting in steep decline in glucose levels and glucose “overcorrection,” improper fluid management, and inappropriate transition to subcutaneous insulin puts these patients at increased risk. In addition, providers may be inclined to start patients with DKA on their PTA insulin doses, which may be too high. A major cause for DKA is nonadherence with prescribed insulin, which emphasizes the need for careful evaluation, including potential reduction of PTA insulin doses when transitioning from an insulin infusion to a subcutaneous maintenance regimen. Pharmacists can play a key role in reducing hypoglycemia in these patients by advising on insulin infusion adjustments and transitions to subcutaneous insulin regimen based on rate of glucose decline and anion gap closure. Their involvement in fluid and electrolyte management is equally important. After initial fluid resuscitation, pharmacists can help ensure that patients are appropriately transitioned to dextrose-containing fluids as their BG normalizes and AG closes to prevent hypoglycemia. In addition, pharmacists can evaluate the patient’s use of insulin before hospitalization and use that information to guide an appropriate subcutaneous insulin regimen when transitioning off insulin infusion. Representative case. On the morning of hospital day 2, risk scoring indicates that a patient admitted for DKA is at high risk for hypoglycemia (Figure 2). As expected in a DKA patient, individual risk factors contributing to the high-risk score include a high BG maximum concentration (697 mg/dL), low BG minimum concentration (64 mg/dL), high maximum-to-minimum BG ratio (11), large mean decrease in BG concentration (-56.7 mg/dL/hr), use of insulin infusion, and history of hypoglycemia. In addition, it is discovered that the patient’s most recent glycosylated hemoglobin A1c (HbA1C) value is 9.1%, indicating potential nonadherence with home antidiabetic regimen. Pharmacist interventions in this case may include recommendation to use conservative insulin doses when transitioning the patient from a continuous to subcutaneous insulin regimen. Doses listed on a PTA medication list will often prove to be too aggressive for the nonadherent DKA patient. When transitioning off of an insulin infusion in these patients, a reduced-dose, long-acting insulin should be considered with a complementary prandial dose plus correction dose with a short-acting insulin analog. Figure 2. View largeDownload slide Representative patient case details for rapid downward trend in blood glucose (BG) levels in patients on insulin infusion or with prior history of hypoglycemia (theme 2). Max= maximum, min= minimum, NPO = nothing by mouth, CLcr = creatinine clearance. Figure 2. View largeDownload slide Representative patient case details for rapid downward trend in blood glucose (BG) levels in patients on insulin infusion or with prior history of hypoglycemia (theme 2). Max= maximum, min= minimum, NPO = nothing by mouth, CLcr = creatinine clearance. Theme 3: Administration of insulin in the presence of an active NPO order in frail patients. Changes in caloric or carbohydrate intake (e.g., NPO, enteral or parenteral nutrition) that occur during hospitalization make maintaining normal glucose levels a challenge. Patients are frequently made NPO for surgery or procedures, and some are placed on reduced carbohydrate diets or diets for specific diseases, such as congestive heart failure or renal disease. Even when a special diet is not ordered, hospitalized patients’ dietary needs and intake will likely differ considerably when compared with their dietary habits before admission. Dietary challenges can be compounded in patients who are already malnourished, patients who have renal insufficiency, and those with known prior hypoglycemic episodes. Dietary intake must be carefully evaluated, and future changes must be anticipated in order to appropriately adjust insulin doses. A reduced-dose basal plus corrective insulin regimen is the preferred treatment for patients with poor oral intake or those who have NPO orders. Providers may overlook additional reduction in the long-acting basal insulin which may be required. Conversely, some providers may incorrectly discontinue all insulin when the patient is made NPO or has poor oral intake, which can have severe consequences in the insulin-dependent patient. An additional challenge is determining exactly what and how much the patient is eating due to inconsistent documentation of dietary intake in the medical record. Representative case. On hospital day 3, risk scoring identifies a patient with type 1 diabetes at high risk for hypoglycemia due to presence of a long-acting insulin with an active NPO diet order (Figure 3). In addition, the patient is also using 2 nondiabetic medications associated with hypoglycemia (ibuprofen, gabapentin)23–25 and has a history of hypoglycemia. Upon review, the pharmacist discovers that an NPO order was placed the previous day in preparation for a complex surgery that morning. It is noted that the patient’s listed home dose of long-acting insulin was continued on admission and remains active. The surgical service informs the pharmacist that the patient will remain NPO following surgery and will not likely resume oral intake for 24–48 hours. The pharmacist recommends a 20 percent reduction in the patient’s long-acting insulin dose, along with close BG monitoring and continued coverage with low-dose correction insulin. The pharmacist informs the team that the long-acting insulin dose will need to be gradually increased along with the use of a mealtime dose of short-acting insulin as diet is escalated. Other potential pharmacist interventions in this case include recommending discontinuation of the patient’s ibuprofen. Figure 3. View largeDownload slide Representative patient case details for administration of insulin in the presence of an active nothing by mouth (NPO) order in select patients (theme 3). BG = blood glucose, max= maximum, min = minimum, CLcr= creatinine clearance. Figure 3. View largeDownload slide Representative patient case details for administration of insulin in the presence of an active nothing by mouth (NPO) order in select patients (theme 3). BG = blood glucose, max= maximum, min = minimum, CLcr= creatinine clearance. Theme 4: Repeated low BG values in frail patients. Hospitalized patients may display a pattern of mild low to low-normal BG values over several days. These patients may go unrecognized if their BG values never decrease below the threshold for true hypoglycemia or manifest common hypoglycemic symptoms (e.g., sweating, chills, dizziness, headache, fatigue). Risk scoring helps to identify these patients through evaluation of minimum BG value for previous days and a history of hypoglycemia before the current hospitalization. Conditions such as low body weight or poor renal function further elevate these patients’ risk. Representative case. On hospital day 6, risk scoring indicates that a patient with type 2 diabetes is at high risk of developing hypoglycemia. Risk factors contributing to the score include a BG value of <70 mg/dL on 4 previous hospital days, high maximum:minimum BG ratio (5:1) indicating large fluctuations in BG values, use of long-acting insulin, impaired renal function, and a history of hypoglycemia (Figure 4). When this type of theme emerges, the pharmacist should assess daily BG fluctuations and determine when minimum values are occurring. Early morning BG values prior to breakfast that are at or slightly below 70 mg/dL in the presence of elevated daytime BG values suggest a need to redistribute insulin dosing by lowering the basal insulin dose and providing higher prandial doses. Adjustments to the timing of insulin administration and diet consumption can be made to stave off repeated low values and, in turn, decrease the likelihood of severe hypoglycemia. Figure 4. View largeDownload slide Representative patient case details for repeated low blood glucose (BG) values in frail patients (theme 4). Max= maximum; Min= minimum; SQ= subcutaneous; HR = high-risk; NPO= nothing per oral; CLcrl= creatinine clearance. Figure 4. View largeDownload slide Representative patient case details for repeated low blood glucose (BG) values in frail patients (theme 4). Max= maximum; Min= minimum; SQ= subcutaneous; HR = high-risk; NPO= nothing per oral; CLcrl= creatinine clearance. Theme 5: Inadequate nighttime BG monitoring for patients on long-acting insulin. Hospitalized patients with diabetes on general wards should have their BG value checked a minimum of 4 times daily (i.e., before each meal and at bedtime). A risk pattern that emerged as part of this analysis was a lack of BG monitoring after 6 p.m. in patients who received long-acting insulin after 8 a.m. Patients in this subset who had large decreases in BG values or large BG fluctuations on the preceding day were at even higher risk. Bedtime BG assessment is important in that it can stop a continuing downward trend with a light snack befoe sleep, reducing the risk of nocturnal or early morning hypoglycemia. In addition, it is important to avoid routine use of correction factor at bedtime. If glucose readings are high (>180–200 mg/dL), a reduced bedtime dose may be considered, but, in general, bedtime doses of short-acting insulin should be discouraged. Representative case. On hospital day 3, the risk score identifies a patient with type 2 diabetes who is at high risk for hypoglycemia. Risk factors include use of long-acting insulin, large BG fluctuations, failure to monitor BG after 6 p.m., and a history of hypoglycemia (Figure 5). The patient is receiving a basal-bolus-correction insulin regimen. The pharmacists note that BG levels have been on the normal-to-high side throughout the day but consistently drop every morning (68 and 62 mg/dL the previous 2 mornings, respectively). It is discovered that the patient has an order for glucose checks three times a day before meals, but no check at bedtime. It is also noted that the patient is not eating anything after 6 p.m. The pharmacist recommends modifying the BG monitoring orders to include a check at bedtime. In addition, the pharmacist recommends a light bedtime snack to prevent low BG values in the early morning. The pharmacist advises against routine use of bedtime correction insulin or, alternatively, a reduced-dose bedtime correction dose of insulin. Figure 5. View largeDownload slide Representative patient case details for inadequate nighttime blood glucose (BG)monitoring for patients on long-acting insulin (theme 5). Max= maximum; Min= minimum;HR = high-risk;; CLcr = creatinine clearance; DM = diabetes mellitus. Figure 5. View largeDownload slide Representative patient case details for inadequate nighttime blood glucose (BG)monitoring for patients on long-acting insulin (theme 5). Max= maximum; Min= minimum;HR = high-risk;; CLcr = creatinine clearance; DM = diabetes mellitus. Discussion To our knowledge, this is the first study to utilize an automated risk score to characterize clinical risk factor patterns for hypoglycemia and to describe practical examples for intervention based on these patterns. This study applied the information gleaned from a novel predictive risk score, where determination of risk factors is fully automated and operationalized from discrete EHR fields for daily prediction of hypoglycemia.18 All 5 phenotypes that we identified involved well-known and prevalent risk factors for hypoglycemia, such as history of previous hypoglycemia, large BG fluctuations, and low renal function.26 Taken individually, none of these risk factors will have high specificity in identifying patients who will develop hypoglycemia during an inpatient stay. But when viewed in specific combinations, these risk phenotypes explained the etiology of hypoglycemia for more than half of the patients in the top fifth risk score percentile. Our 5 phenotypes appeared in 53% of patients who experienced hypoglycemia after risk prediction. The rate of hypoglycemia in this high-risk group was 16%;18 suggesting that clinicians would see 1 hypoglycemia event in every 7 high-risk admissions. While the patterns and interventions described here were developed from information gathered in our institution, clinicians may recognize the case vignettes are common risk factor constellations in their patients. Yet, local practice patterns and interventions may differ as treatment protocols and available resources are not always comparable. Compared with manual chart review, which is usually confined to extraction of a small number of risk factors, we were able to take advantage of a great volume of information from discrete fields in electronic health records. The application of a previously developed risk score allowed us then to interpret the contribution of each risk factor in the context of predicted probability of hypoglycemia, enhancing our ability to identify prominent risk factor pattern. Finally, our pattern analysis across several hundred hypoglycemia cases that were correctly flagged as “high risk” by the risk score allowed distilling phenotypes that appear common in our institution. We hope that both the description of these phenotypes as well as our description of case vignettes along with recommendations for intervention will improve the detection of high-risk patients in clinical practice, even if no sophisticated prediction methods are available, and allow pharmacists greater involvement in optimal BG management during admission. Some limitations inherent in risk score development and pattern analysis need to be noted. The estimates calculated by our risk score reflect risk factors “as present” in our two tertiary hospitals. The individual variables used to calculate the risk score will inevitably be different across institutions because of differences in drug formularies, laboratory reporting practices, institution-specific protocols, and other local practices that affect hyperglycemia management. In addition, variables change over time as new therapies emerge, diagnostic advances are made, and the evidence evolves. We also note that there are numerous risk factors that can interact in multiple additional ways to put a patient at high risk. We have presented 5 common patterns and clinical scenarios identified in our study population, which should not distract from vigilance when other risk factor patterns arise. Lastly, the effectiveness of the suggested preventive interventions was not evaluated in this study, and examples are presented only to illustrate interventions that could be implemented by pharmacists and that are consistent with current glucose management guidelines.27–29 Future research that develops and evaluates explicit interventions for the identified phenotypes may further enhance BG control during hospitalization. Conclusion Five distinct phenotypes that are prevalent in patients at greatest risk for inpatient hypoglycemia were identified. Disclosures This study was funded in part by the ASHP Research and Education Foundation. The authors have declared no potential conflicts of interest. References 1. Cullen DJ , Sweitzer BJ , Bates DW et al. Preventable adverse drug events in hospitalized patients: a comparative study of intensive care and general care units . Crit Care Med . 1997 ; 25 : 1289 – 97 . Google Scholar Crossref Search ADS PubMed 2. Cullen DJ , Bates DW , Leape LL. Prevention of adverse drug events: a decade of progress in patient safety . J Clin Anesth . 2000 ; 12 : 600 – 14 . Google Scholar Crossref Search ADS PubMed 3. Bates DW , Boyle DL , Vander Vliet MB et al. Relationship between medication errors and adverse drug events . J Gen Intern Med . 1995 ; 10 : 199 – 205 . Google Scholar Crossref Search ADS PubMed 4. Murad MH , Coto-Yglesias F , Wang AT et al. Clinical review: drug-induced hypoglycemia: a systematic review . J Clin Endocrinol Metab . 2009 ; 94 ( 3 ): 741 – 5 . Google Scholar Crossref Search ADS PubMed 5. Turchin A , Matheny ME , Shubina M et al. Hypoglycemia and clinical outcomes in patients with diabetes hospitalized in the general ward . Diabetes Care . 2009 ; 32 : 1153 – 7 . Google Scholar Crossref Search ADS PubMed 6. Finfer S , Liu B , Chittock DR et al. Hypoglycemia and risk of death in critically ill patients . N Engl J Med . 2012 ; 367 : 1108 – 18 . Google Scholar Crossref Search ADS PubMed 7. Pasquel FJ , Spiegelman R , McCauley M et al. Hyperglycemia during total parenteral nutrition: an important marker of poor outcome and mortality in hospitalized patients . Diabetes Care . 2010 ; 33 : 739 – 41 . Google Scholar Crossref Search ADS PubMed 8. King JT , Goulet JL , Perkal MF et al. Glycemic control and infections in patients with diabetes undergoing noncardiac surgery . Ann Surg . 2011 ; 253 : 158 – 65 . Google Scholar Crossref Search ADS PubMed 9. Varghese P , Gleason V , Sorokin R et al. Hypoglycemia in hospitalized patients treated with antihyperglycemic agents . J Hosp Med . 2007 ; 2 : 234 – 40 . Google Scholar Crossref Search ADS PubMed 10. Hulkower RD , Pollack RM , Zonszein J. Understanding hypoglycemia in hospitalized patients . Diabetes Manag . Mar 2014 ; 4 : 165 – 76 . Google Scholar Crossref Search ADS 11. Winterstein AG , Hatton RC , Gonzalez-Rothi R et al. Identifying clinically significant preventable adverse drug events through a hospital’s database of adverse drug reaction reports . Am J Health-Syst Pharm . 2002 ; 59 : 1742 – 9 . Google Scholar PubMed 12. U.S. Department of Health and Human Services, Office of Disease Prevention and Health Promotion . National action plan for adverse drug event prevention . (accessed 2017 Feb 15) 13. Rodriguez-Gutierrez R , Ospina NS , McCoy RG et al. Inclusion of hypoglycemia in clinical practice guidelines and performance measures in the care of patients with diabetes . JAMA Intern Med . 2016 ; 176 : 1714 – 6 . Google Scholar Crossref Search ADS PubMed 14. Deusenberry CM , Coley KC , Korytkowski MT , Donihi AC. Hypoglycemia in hospitalized patients treated with sulfonylureas . Pharmacotherapy . 2012 ; 32 : 613 – 7 . Google Scholar Crossref Search ADS PubMed 15. Dendy JA , Chockalingam V , Tirumalasetty NN et al. Identifying risk factors for severe hypoglycemia in hospitalized patients with diabetes . Endocr Pract . 2014 ; 20 : 1051 – 6 . Google Scholar Crossref Search ADS PubMed 16. Milligan PE , Bocox MC , Pratt E et al. Multifaceted approach to reducing occurrence of severe hypoglycemia in a large healthcare system . Am J Health-Syst Pharm . 2015 ; 72 : 1631 – 41 . Google Scholar Crossref Search ADS PubMed 17. Winterstein AG , Staley B , Henriksen C et al. Development and validation of a complexity score to rank hospitalized patients at risk for preventable adverse drug events . Am J Health-Syst Pharm . 2017 ; 74 : 1970 – 84 . Google Scholar Crossref Search ADS PubMed 18. Winterstein AG , Jeon N , Staley B et al. Development and validation of an automated algorithm for identifying patients at high risk for drug-induced hypoglycemia . Am J Health-Syst Pharm . 2018 ; 75 : 1714 – 28 . Google Scholar Crossref Search ADS PubMed 19. American Diabetes Association . Glycemic targets . Diabetes Care . 2018 ; 41 ( suppl 1 ): S55 – 64 . http://care.diabetesjournals.org/content/41/Supplement_1/S55 20. Umpierrez GE , Isaacs SD , Bazargan N et al. Hyperglycemia: an independent marker of in-hospital mortality in patients with undiagnosed diabetes . J Clin Endocrinol Metab . 2002 ; 87 : 978 – 82 . Google Scholar Crossref Search ADS PubMed 21. Frisch A , Chandra P , Smiley D et al. Prevalence and clinical outcome of hyperglycemia in the perioperative period in noncardiac surgery . Diabetes Care . 2010 ; 33 : 1783 – 8 . Google Scholar Crossref Search ADS PubMed 22. Finfer S , Chittock DR , Su SY et al. Intensive versus conventional glucose control in critically ill patients . N Engl J Med . 2009 ; 360 : 1283 – 97 . Google Scholar Crossref Search ADS PubMed 23. Sone H , Takahashi A , Yamada N. Ibuprofen-related hypoglycemia in a patient receiving sulfonylurea . Ann Intern Med . 2001 ; 134 : 344 . Google Scholar Crossref Search ADS PubMed 24. Penumalee S , Kissner PZ , Migdal SD. Gabapentin-induced hypoglycemia in a long-term peritoneal dialysis patient . Am J Kidney Dis . 2003 ; 42 :E3-5. 25. Li J , Zhang N , Ye B et al. Non-steroidal anti-inflammatory drugs increase insulin release from beta cells by inhibiting ATP-sensitive potassium channels . Br J Pharmacol . 2007 ; 151 : 483 – 93 . Google Scholar Crossref Search ADS PubMed 26. Yun JS , Ko SH. Risk factors and adverse outcomes of severe hypoglycemia in type 2 diabetes mellitus . Diabetes Metab J . 2016 ; 40 : 423 – 32 . Google Scholar Crossref Search ADS PubMed 27. Umpierrez GE , Hellman R , Korytkowski MT et al. Management of hyperglycemia in hospitalized patients in non-critical care setting: an endocrine society clinical practice guideline . J Clin Endocrinol Metab . 2012 ; 97 : 16 – 38 . Google Scholar Crossref Search ADS PubMed 28. American Diabetes Association . Diabetes care in the hospital . Diabetes Care . 2016 ; 39 ( suppl 1 ): S99 – 104 . Crossref Search ADS PubMed 29. Moghissi ES , Korytkowski MT , DiNardo M et al. American Association of Clinical Endocrinologists and American Diabetes Association consensus statement on inpatient glycemic control . Endocr Pract . 2009 ; 15 : 353 – 69 . Google Scholar Crossref Search ADS PubMed Appendix Risk factors considered for identification of risk patterns. To identify risk factor patterns associated with hypoglycemia, we extracted risk score detail for all admission-days that ranked in the top fifth percentile of all admissions and days for which a risk score was available and that were followed by hypoglycemia. We generated frequency distributions of all combinations of four risk factors that were among the top five risk factors that contributed to an admission’s total risk score. We did this separately for each of the first five hospitals days (1, 2, or 3–5) because we found that risk factor contributions (marked as “x”) to hypoglycemia risk may change during the course of a patient’s admission. BG = blood glucose, NPO = nothing by mouth. Risk Factor Day1 Day2 Day3–5 Admission BG conc. (mg/dL) x x BG maximum conc. (mg/dL) x BG minimum conc. (mg/dL) x BG conc. fluctuation (mg/dL), log transformed x x x BG conc. trend x x x No. prior days with BG conc. < 70mg/dL x x Insulin infusion x x x Long-acting insulin x x x Insufficient BG monitoring for long-acting insulin regimen x High-dose subcutaneous insulin x x x Sulfonylureas x x x No. nondiabetic agent(s) with hypoglycemic effects x Initiation of new antihyperglycemic drug x Change in diabetes regimen x Use of 50% dextrose injection x x NPO order x Insulin and presence of an active NPO order x Serum albumin conc. (mg/dL) x x Creatinine clearance (mL/min) x x x Type 1 diabetes mellitus x x History of hypoglycemia x x x Body weight (kg) x x x Body temperature (°C) x x Male x Risk Factor Day1 Day2 Day3–5 Admission BG conc. (mg/dL) x x BG maximum conc. (mg/dL) x BG minimum conc. (mg/dL) x BG conc. fluctuation (mg/dL), log transformed x x x BG conc. trend x x x No. prior days with BG conc. < 70mg/dL x x Insulin infusion x x x Long-acting insulin x x x Insufficient BG monitoring for long-acting insulin regimen x High-dose subcutaneous insulin x x x Sulfonylureas x x x No. nondiabetic agent(s) with hypoglycemic effects x Initiation of new antihyperglycemic drug x Change in diabetes regimen x Use of 50% dextrose injection x x NPO order x Insulin and presence of an active NPO order x Serum albumin conc. (mg/dL) x x Creatinine clearance (mL/min) x x x Type 1 diabetes mellitus x x History of hypoglycemia x x x Body weight (kg) x x x Body temperature (°C) x x Male x View Large Risk factors considered for identification of risk patterns. To identify risk factor patterns associated with hypoglycemia, we extracted risk score detail for all admission-days that ranked in the top fifth percentile of all admissions and days for which a risk score was available and that were followed by hypoglycemia. We generated frequency distributions of all combinations of four risk factors that were among the top five risk factors that contributed to an admission’s total risk score. We did this separately for each of the first five hospitals days (1, 2, or 3–5) because we found that risk factor contributions (marked as “x”) to hypoglycemia risk may change during the course of a patient’s admission. BG = blood glucose, NPO = nothing by mouth. Risk Factor Day1 Day2 Day3–5 Admission BG conc. (mg/dL) x x BG maximum conc. (mg/dL) x BG minimum conc. (mg/dL) x BG conc. fluctuation (mg/dL), log transformed x x x BG conc. trend x x x No. prior days with BG conc. < 70mg/dL x x Insulin infusion x x x Long-acting insulin x x x Insufficient BG monitoring for long-acting insulin regimen x High-dose subcutaneous insulin x x x Sulfonylureas x x x No. nondiabetic agent(s) with hypoglycemic effects x Initiation of new antihyperglycemic drug x Change in diabetes regimen x Use of 50% dextrose injection x x NPO order x Insulin and presence of an active NPO order x Serum albumin conc. (mg/dL) x x Creatinine clearance (mL/min) x x x Type 1 diabetes mellitus x x History of hypoglycemia x x x Body weight (kg) x x x Body temperature (°C) x x Male x Risk Factor Day1 Day2 Day3–5 Admission BG conc. (mg/dL) x x BG maximum conc. (mg/dL) x BG minimum conc. (mg/dL) x BG conc. fluctuation (mg/dL), log transformed x x x BG conc. trend x x x No. prior days with BG conc. < 70mg/dL x x Insulin infusion x x x Long-acting insulin x x x Insufficient BG monitoring for long-acting insulin regimen x High-dose subcutaneous insulin x x x Sulfonylureas x x x No. nondiabetic agent(s) with hypoglycemic effects x Initiation of new antihyperglycemic drug x Change in diabetes regimen x Use of 50% dextrose injection x x NPO order x Insulin and presence of an active NPO order x Serum albumin conc. (mg/dL) x x Creatinine clearance (mL/min) x x x Type 1 diabetes mellitus x x History of hypoglycemia x x x Body weight (kg) x x x Body temperature (°C) x x Male x View Large © American Society of Health-System Pharmacists 2019. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Common inpatient hypoglycemia phenotypes identified from an automated electronic health record–based prediction model
JF - American Journal of Health-System Pharmacy
DO - 10.1093/ajhp/zxy017
DA - 2019-01-25
UR - https://www.deepdyve.com/lp/oxford-university-press/common-inpatient-hypoglycemia-phenotypes-identified-from-an-automated-mLZm6ivK0Y
SP - 166
VL - 76
IS - 3
DP - DeepDyve
ER -