Secondary Logo

Journal Logo

Original Article

Reducing Diabetic Ketoacidosis Intensive Care Unit Admissions Through an Electronic Health Record-Driven, Standardized Care Pathway

Edholm, Karli; Lappé, Katie; Kukhareva, Polina; Hopkins, Christy; Hatton, Nathan D.; Gebhart, Benjamin; Nyman, Heather; Signor, Emily; Davis, Mikyla; Kawamoto, Kensaku; Johnson, Stacy A.

Author Information
doi: 10.1097/JHQ.0000000000000247



Diabetic ketoacidosis (DKA) is a common and costly acute complication of diabetes mellitus with wide variation in clinical practice and no clear guidelines for admission location (i.e., general medicine floor vs. intensive care unit [ICU]).1,2 Standardized protocols for DKA have previously been shown to improve clinical outcomes.3,4 In addition, multiple randomized, controlled clinical trials have demonstrated similar safety and efficacy of frequent doses (every 1–2 hours) of subcutaneous (SQ) rapid-acting insulin (i.e., lispro or aspart) compared with continuous insulin infusion in the treatment of nonsevere DKA.5-9 In light of this evidence, the American Diabetes Association's (ADA) consensus statement on hyperglycemic crises includes the use of SQ insulin for patients with nonsevere DKA.10

However, the mainstay of treatment for DKA continues to be the administration of low-dose regular insulin via continuous insulin infusion. In many institutions, including ours, continuous insulin infusion requires admission to an ICU. Intensive care unit utilization for DKA management varies widely across institutions and is associated with increased cost and invasive procedures, without demonstrated benefit to patients.2,11 In addition, unnecessary ICU utilization may result in missed opportunities for patient education, because this is not generally prioritized in an ICU setting. We aimed to decrease ICU utilization for DKA at our institution, as well as improve patient education and outpatient follow-up, through the implementation of an electronic health record (EHR)-driven, evidence-based care pathway that standardized DKA management according to illness severity.



We conducted a retrospective, observational preintervention to postintervention study to measure the impact of our EHR-driven standardized care pathway for DKA.


Could we implement an EHR-driven standardized care pathway for DKA, using SQ insulin, rather than a continuous insulin infusion, that would lead to decreased ICU admission for DKA, and improved patient education, without adversely affecting patient safety outcomes?

Ethical Approval

The Institutional Review Board classified this as a quality improvement project and did not require review and oversight.


Hospital admissions of patients 18 years and older who presented to the Emergency Department (ED) meeting laboratory criteria for DKA or who had an ICD-10-CM diagnosis code for DKA during their encounter (E08.1x, E09.1x, E10.1x) between January 1, 2016 and February 28, 2018 were included in the study. The period between January 1, 2016 and December 31, 2016 was considered preintervention, and February 1, 2017, to January 31, 2018, was considered postintervention. January 1, 2017, to January 31, 2017, was a “run-in” period in which EHR changes were put into place and physician and nursing education was completed. Data from this 1-month period was not included in the preintervention or postintervention analysis. Acknowledging that the laboratory criteria for DKA are nonspecific, and that ICD-10 codes may be inaccurate,12 a manual chart review was performed by two experienced attending physicians, who independently reviewed patient records to verify that each patient met prespecified criteria for a diagnosis of DKA. An independent adjudicator then resolved any discrepancies.

Patients were excluded if they were determined by chart review not to have DKA, left against medical advice, developed DKA during their hospital stay (i.e., not present on admission), were discharged from the ED without being admitted, or were directly admitted to the hospital from an outside ED, outside hospital, or outpatient clinic (Figure 1).

Figure 1.
Figure 1.:
Consort flow diagram. AMA = against medical advice.


A large academic medical center.


First, a multidisciplinary team including key stakeholders from the hospitalist group, ICU, ED, pharmacy, nursing, and information technology was formed. The group designed a clinical care pathway for the triage, evaluation, and management of patients admitted with DKA based on the ADA's consensus guidelines and relevant literature.5-10 Electronic health record decision support tools called best practice advisories (BPAs) were developed to alert and guide clinicians and nurses through the clinical care pathway (see Appendix 1, Supplemental Digital Content 1, for screenshots of all BPAs and EHR order sets). The process map, which provides a visual overview of the clinical care pathway from ED admission until the time of discharge (see Appendix 2, Supplemental Digital Content 2,, was posted in the ED, general medicine floor and ICU provider work areas for easy reference by physicians and nurses. In addition, all frontline providers (physicians, advance practice providers, residents, nurses, and pharmacists) received a single educational session before implementation of the pathway. The educational content primarily focused on the details and logistics of the clinical care pathway, including ED and acute care floor workflows, criteria for ICU and floor admission, and proper use of the BPAs and order sets. No formal ongoing education was performed because of the EHR-driven nature of the intervention. Nurses hired after implementation of the pathway received training as part of their new employee orientation. In addition, members of the DKA care pathway team from each department were available to assist providers and answer questions about the pathway once the DKA protocol was implemented. All insulin ordering and administration was performed in accordance with our hospital's preexisting policies and procedures, which are consistent with the “ISMP Guidelines for Optimizing Safe Subcutaneous Insulin Use in Adults.”13 Starting January 30, 2017, all adult patients were screened for entry into the pathway if a point-of-care blood glucose in the ED was >500 mg/dl. The first BPA (BPA #1) alerted the nurse to screen the patient for DKA by expediting a venous blood gas. Patients were entered into the pathway if labs were diagnostic of DKA, that is, blood glucose >250 mg/dl AND bicarbonate <18 mEq/L AND pH <7.3 (from venous or arterial blood gas) OR blood glucose >250 mg/dl AND bicarbonate <18 mEq/L AND anion gap >14 mEq/L (from basic or complete metabolic panel). If lab criteria for DKA was met, a second BPA (BPA #2) alerted the nurse and physician to the possible diagnosis of DKA and directed them to an order set to initiate DKA work-up and treatment, triage the patient to the appropriate admitting unit, and place a consult to the diabetes nurse educator. If the patient did not have DKA (i.e an alternative diagnosis explains their metabolic derangements), the provider could turn off subsequent BPA by selecting “Clinically not DKA.” As part of BPA #2, patients classified as having severe DKA based on ADA guidelines and our institutional ICU admission criteria (pH <7.0, stupor or coma, potassium <3.3 mEq/L, corrected sodium >150 mEq/L, end-stage renal disease on hemodialysis, systolic blood pressure <90 mm Hg after 2 L of intravenous fluids, other condition requiring ICU level care) were started on a continuous insulin infusion and admitted to an ICU. Patients with mild or moderate DKA, that is, those not meeting severe DKA or ICU admission criteria, were initiated on subcutaneous lispro insulin and admitted to a general medicine floor. A third BPA (BPA #3) alerted the nurse and physician when the patient's hyperglycemia resolved (defined as first blood glucose <250 mg/dl) and directed them to an order set to decrease the insulin dose and add dextrose 5% (D5) to the intravenous fluids. A fourth and final BPA (BPA #4) alerted the nurse and provider when the patient's anion gap closed (defined as first anion gap <12 mEq/L), and directed them to an order set to transition the patient to basal insulin and schedule outpatient follow-up.

Data Collection

We obtained data on patient and visit characteristics (age, gender, ethnicity, insurance status, weight, body mass index (BMI), Charlson comorbidity index, admission unit, admission and discharge dates and times, length of stay (LOS), 30-day ED return visit, 30-day hospital readmission, inpatient mortality, and diabetes nurse educator consultation), as well as laboratory and medication administration data from the enterprise data warehouse. Patient-specific cost data were obtained from our institutionally derived value driven outcomes tool.14,15


Our primary outcome, determined a priori, was ICU admission. Secondary outcomes were defined as treatment with an insulin drip; ED and hospital LOS; time to first insulin dose (any insulin given); time to resolution of hyperglycemia (first glucose <250 mg/dl); time to anion gap closure (first anion gap <12 mEq/L); time to initiation of basal insulin (first dose of glargine, NPH or 70/30 insulin); hemoglobin A1c ordered (obtained during admission or available in the EHR within the preceding 3 months); treatment-induced hypokalemia (potassium <3.3 mEq/L within 12 hours of first dose of insulin in a patient with an initial potassium >3.3 mEq/L); hypoglycemia (blood glucose <70 mg/dl); reopening of the anion gap (anion gap >14 mEq/L occurring after correction [<12 mEq/L] of an initially elevated anion gap [>14 mEq/L]); diabetes nurse educator consultation; 30-day ED return visit; 30-day hospital readmission; inpatient mortality; and total direct hospital cost. All time metrics were calculated using the ED admission date and time as the starting point. Time to initiation of basal insulin was missing for six visits, which were excluded from the analysis for that outcome only. Total direct hospital cost was not adjusted for inflation because Consumer Price Index was 0% between 2016 and 2017 and 2017 and 2018.16

Adjustment Variables

To account for differences in baseline patient characteristics between the preintervention and postintervention groups, we included age, gender, ethnicity, insurance status, BMI, Charlson comorbidity index, and DKA severity in statistical models. Charlson comorbidity index was calculated according to the algorithm specified by Quan et al17 using all patient diagnoses from previous visits and the index visit identified from the facility billing system. There were two values missing for BMI. A single stochastic regression imputation based on age and weight was used to fill them in.

Data Analysis

Categorical variables are reported as count and percentage. Continuous patient characteristics are reported as mean ± SD and continuous outcome variables are reported as mean ± SE. Categorical variables were compared using chi squared tests and continuous variables were compared using t tests. To adjust for variation in the patient characteristics, generalized linear models with gamma distribution and log transformation were used for continuous outcomes and logistic regression models were used for categorical outcomes. p values <0.05 were considered significant. We used SAS version 9.4 statistical software (SAS Institute. Inc., Cary, NC) and R version 3.4.3 for data analysis.


A total of 687 admissions met inclusion criteria (laboratory criteria for DKA or an encounter ICD-CM diagnosis code for DKA) during the study period. Of these, 277 admissions were eliminated based on exclusion criteria (Figure 1), leaving 410 admissions with possible DKA. Manual chart review identified an additional 197 admissions that were not clinically consistent with DKA; these were excluded from analysis. The remaining 214 admissions (171 unique patients) comprised the study cohorts; 106 admissions in the preimplementation group and 108 admissions in the postimplementation group. Baseline patient characteristics and admission laboratory data are shown in Table 1. Baseline characteristics were well matched except for mean age, which was approximately 4 years older in the post-intervention group (36.42 ± 13.80 years vs. 40.77 ± 16.28 years, p = .04).

Table 1. - Visit Characteristics Preintervention and Postintervention
Study period
Characteristica Preintervention Postintervention pb
No. of admissions 106 108
No. of unique patients 85 88
Patient characteristics
 Age, yr 36.42 ± 13.80 40.77 ± 16.28 .04
 Female gender 50 (47.2%) 49 (45.4%) .79
 Hispanic/Latino ethnicity 18 (17.0%) 28 (25.9%) .11
 Insured 81 (76.4%) 80 (74.1%) .69
 Weight, kg 77.94 ± 24.50 73.54 ± 20.13 .15
 Body mass index (BMI) 26.59 ± 7.47 25.38 ± 7.04 .22
 Charlson comorbidity index 3.16 ± 2.81 3.30 ± 2.64 .72
 Hemoglobin A1cc 11.89 ± 3.03 11.78 ± 3.14 .82
 Required mechanical ventilation 5 (4.7%) 4 (3.7%) .71
 Required vasopressors 6 (5.7%) 6 (5.6%) .97
 Severe diabetic ketoacidosis 46 (43.4%) 37 (34.3%) .17
 Admission (first) glucose 532.70 ± 201.49 535.00 ± 224.37 .94
 Admission (first) pH 7.22 ± 0.14 7.24 ± 0.15 .23
 Admission (first) bicarbonate 11.26 ± 5.05 12.07 ± 5.28 .25
aValues expressed as n (%) for categorical variables and as mean ± SD for continuous variables.
bp-values are based on χ2 tests for categorical variables and on t tests for continuous variables.
cMean hemoglobin A1c is calculated based on 182 visits where hemoglobin A1c was available.

Unadjusted clinical outcomes preintervention and postintervention are shown in Table 2. The primary outcome of ICU admissions decreased from 67.0% to 41.7% (p < .001). Diabetes nurse educator consultations increased from 45.3% to 63.9% (p = .006), and time to initiation of basal insulin increased from 18.19 ± 1.25 hours to 22.47 ± 1.76 hours (p = .05). There were no differences in other time metrics (time to first insulin dose, time to resolution of hyperglycemia, and time to anion gap closure). Reopening of the anion gap increased from 4.7% to 13.9% (p = .02). However, this was not associated with increased hospital LOS (p = .87). There were no significant changes in other safety outcomes (ED LOS, hypoglycemia, treatment-induced hypokalemia, inpatient mortality, and 30-day hospital readmission) or total direct hospital cost. Thirty-day ED return visit decreased from 12.3% to 2.8% (p = .008).

Table 2. - Clinical Outcomes Preimplementation and Postimplementation of an EHR-Driven Standardized Care Pathway for Diabetic Ketoacidosis
Study period
Characteristics Preintervention Postintervention pa
No. of admissions 106 108
No. of unique patients 85 88
ICU admission 71 (67.0%) 45 (41.7%) <.001
Treatment with an insulin drip 87 (82.1%) 47 (43.5%) <.001
ED LOS, h 5.47 ± 0.36 5.68 ± 0.38 .70
Hospital LOS, d 3.39 ± 0.29 3.33 ± 0.28 .87
Time to first insulin dose, h 3.32 ± 0.22 3.67 ± 0.29 .34
Time to resolution of hyperglycemia, h 6.05 ± 0.53 7.25 ± 0.72 .18
Time to anion gap closure, h 8.81 ± 0.55 9.98 ± 0.80 .23
Time to initiation of basal insulin, h 18.19 ± 1.25 22.47 ± 1.76 .05
Hemoglobin A1c ordered 81 (76.4%) 101 (93.5%) <.001
Treatment-induced hypokalemia 13 (12.3%) 12 (11.1%) .79
Hypoglycemia (any glucose <70 mg/dl) 34 (32.1%) 27 (25.0%) .25
Reopening of the anion gap 5 (4.7%) 15 (13.9%) .02
Diabetes nurse educator consultation 48 (45.3%) 69 (63.9%) .006
30-day ED return visit 13 (12.3%) 3 (2.8%) .008
30-day hospital readmission 11 (10.4%) 12 (11.1%) .86
Inpatient mortality 1 (0.9%) 3 (2.8%) .32
Relative total direct cost 1.00 ± 0.09 1.01 ± 0.11 .95
Note: Values expressed as n (%) for categorical variables and mean ± SE for continuous variables.
ap-values are based on χ2 tests for categorical variables and on t tests for continuous variables.
ED = emergency department; EHR = electronic health record; ICU = intensive care unit; LOS = length of stay.

After adjustment for age, gender, ethnicity, insurance status, BMI, Charlson comorbidity index, and DKA severity at admission, primary and secondary outcomes were similar to the unadjusted outcomes (Table 3).

Table 3. - Impact of Intervention on Outcomes
Outcome variables Preintervention Postintervention Intervention Effecta pb
Admitted to ICU Unadjusted 66.98% ± 4.57% 41.67% ± 4.74% −65% (−80% to −39%) <.001
Adjusted 69.51% ± 4.97% 41.35% ± 5.39% −69% (−84% to −41%) <.001
Treated with insulin drip Unadjusted 82.08% ± 3.73% 43.52% ± 4.77% −83% (−91% to −69%) <.001
Adjusted 85.44% ± 3.55% 44.59% ± 5.49% −86% (−93% to −72%) <.001
ED LOS, h Unadjusted 5.47 ± 0.28 5.68 ± 0.29 4% (−10% to 20%) .62
Adjusted 5.31 ± 0.25 5.49 ± 0.25 3% (−9% to 18%) .63
Hospital LOS, d Unadjusted 3.39 ± 0.23 3.33 ± 0.23 −2% (−19% to 18%) .84
Adjusted 3.27 ± 0.20 3.07 ± 0.19 −6% (−21% to 12%) .49
Time to first insulin given, h Unadjusted 3.32 ± 0.21 3.67 ± 0.23 11% (−8% to 32%) .27
Adjusted 3.28 ± 0.20 3.46 ± 0.21 5% (−11% to 25%) .56
Time to resolution of hyperglycemia, h Unadjusted 6.05 ± 0.53 7.25 ± 0.63 20% (−6% to 53%) .14
Adjusted 5.93 ± 0.50 6.68 ± 0.55 13% (−11% to 42%) .32
Time to anion gap closure, h Unadjusted 8.81 ± 0.60 9.98 ± 0.67 13% (−6% to 37%) .19
Adjusted 8.68 ± 0.58 9.71 ± 0.64 12% (−7% to 35%) .25
Time to initiation of basal insulin, h Unadjusted 18.19 ± 1.05 22.47 ± 1.26 24% (6% to 45%) .009
Adjusted 17.79 ± 0.96 21.62 ± 1.14 22% (4% to 41%) .01
Hemoglobin A1c ordered Unadjusted 76.42% ± 4.12% 93.52% ± 2.37% 345% (83% to 982%) <.001
Adjusted 80.65% ± 4.26% 94.98% ± 2.09% 354% (79% to 53%) .001
Hypoglycemia (any glucose <70 mg/dl) Unadjusted 32.08% ± 4.53% 25.00% ± 4.17% −29% (−61% to 28%) .25
Adjusted 31.38% ± 4.95% 21.36% ± 4.19% −41% (−69% to 16%) .12
Treatment-induced hypokalemia Unadjusted 12.26% ± 3.19% 11.11% ± 3.02% −11% (−61% to 106%) .79
Adjusted 10.37% ± 3.13% 8.80% ± 2.85% −17% (−66% to 105%) .69
Re-opening of the anion gap Unadjusted 4.72% ± 2.06% 13.89% ± 3.33% 226% (14% to 832%) .03
Adjusted 2.72% ± 1.50% 10.01% ± 3.42% 299% (27% to 151%) .02
Seen by diabetes nurse educator Unadjusted 45.28% ± 4.83% 63.89% ± 4.62% 114% (24% to 270%) .007
Adjusted 44.47% ± 5.14% 65.00% ± 4.79% 132% (29% to 318%) .005
ED return visit within 30 days Unadjusted 12.26% ± 3.19% 2.78% ± 1.58% −80% (−94% to −26%) .02
Adjusted 9.18% ± 3.13% 2.17% ± 1.36% −78% (−94% to −18%) .02
Hospital readmission within 30 days Unadjusted 10.38% ± 2.96% 11.11% ± 3.02% 8% (−55% to 157%) .86
Adjusted 9.14% ± 2.86% 10.24% ± 2.97% 13% (−54% to 178%) .78
Inpatient mortality Unadjusted 0.94% ± 0.94% 2.78% ± 1.58% 200% (−69% to 831%) .34
Adjusted 0.21% ± 0.31% 0.48% ± 0.66% 126% (−85% to 404%) .56
Relative total direct cost Unadjusted 1.00 ± 0.07 1.01 ± 0.07 1% (−17% to 23%) .94
Adjusted 1.00 ± 0.06 0.95 ± 0.06 −5% (−20% to 13%) .55
Note: ICU utilization for DKA decreased by 65% in the postintervention period compared with the preintervention period.
aIntervention effect represents relative change in odds for categorical variables and in amount for continuous variables, and was calculated as exponentiation of the beta parameter for the variable minus one. Minus (−) sign represents decrease in odds or quantity.
bp-values are based on generalized linear models. Outcomes adjusted for age, gender, ethnicity, insurance status, BMI, CCI, and DKA severity at admission.
DKA = diabetic ketoacidosis; ED = emergency department; ICU = intensive care unit; LOS = length of stay.


Our study has several limitations. First, this quality improvement initiative was performed at a single academic medical center, which may limit its generalizability to other practice settings. Second, the intervention was multifaceted, so we cannot discern which aspects were most responsible for the observed effects. Next, as this was not a prospective or randomized controlled study, the results could be due to secular trends. However, we are not aware of any other systemic changes that occurred during the project time period that would have affected care for patients with DKA. In addition, since we were embarking on a quality improvement initiative meant to be a “real-world” experience, we did not track protocol adherence. Third, although serum ketones are included in the ADA diagnostic criteria for DKA, the lab takes 1–3 days to result in our institution, so we were unable to incorporate them into our care pathway. This resulted in a large number of patients who triggered the care pathway but who had other clinical explanations for their laboratory derangements (i.e., out of hospital cardiac arrest, septic shock, etc.). Providers were easily able to opt out of the pathway for these patients, but it required us to perform a chart review for the purposes of this study to determine which patients actually had DKA. Institutions with the ability to perform real-time serum ketone checks can improve the specificity of DKA triggers in making the diagnosis of DKA when implementing a similar pathway. Finally, although our postintervention time period was 1 year, we do not have data beyond that to monitor for sustainability of the results.


Our study suggests that the implementation of a novel, EHR-driven, standardized DKA care pathway using SQ insulin for the treatment of mild and moderate DKA significantly decreases the number of ICU admissions for continuous insulin infusion. In addition, we observed a significant increase in the completion of diabetes nurse educator consultations and a reduction in return visits to the ED within 30 days. These potential benefits were noted without significant increases in clinically meaningful adverse events including hypoglycemia, treatment-induced hypokalemia, hospital readmission, or mortality.

To ensure our intervention (which required more up-front treatment in the ED) did not adversely affect ED throughput, we monitored ED LOS, which did not increase. We also measured different time points in the care pathway to monitor patients' clinical response to our intervention. We saw no change in the length of time for initiation of treatment (first dose of insulin given), resolution of hyperglycemia, or anion gap closure.

We were successful in our attempt to improve patient education, as measured by the increase in patients who had a consult with the diabetes nurse educator before discharge (from 45.3% to 63.9%, p = .006). Our intervention included a process to schedule outpatient follow-up before discharge. However, because of constraints of the EHR, we were unable to reliably determine which patients were scheduled for outpatient follow-up before discharge in the preintervention group. Therefore, we were unable to assess this metric preintervention and postintervention. We believe this outpatient follow-up, as well as the education patients received by the diabetes nurse educator, likely explains the significant decrease in 30-day ED return visits (12.3%–2.8%, p = .008). However, 30-day hospital readmission did not change as a result of our intervention.

After intervention, initiation of basal insulin following the resolution of DKA took approximately 4 hours longer. This may be explained by the fact that more patients in the postintervention group were treated on a general medicine floor. Upon closure of the anion gap, a BPA (BPA #4) fired to the nurse and physician prompting the physician to cancel the current insulin orders and order basal insulin therapy. On the general medicine floor, there is a higher patient to nurse ratio which can delay recognition of new BPAs by nurses, thereby delaying notification to the physician to complete the basal insulin order set. In addition, physician night coverage is structured so that physicians do not routinely open an inpatient chart unless called by nursing for an acute change. As a result, there could be a physician delay in seeing a BPA noting closure of the anion gap and need to initiate basal insulin. Finally, BPA #4 provided specific instructions to the provider recommending initiation of basal insulin between the hours of 6 am and 12 pm or 6 pm and 12 am only as this is the time of day when basal insulin is generally administered. Therefore, physicians may have chosen to delay basal insulin order entry until these prespecified times.

We also observed an increase in reopening of the anion gap (which we defined as anion gap >14 mEq/L occurring after correction [<12 mEq/L] of an initially elevated anion gap [>14 mEq/L]). Of the 15 patients in whom this occurred, 10 had been treated with continuous insulin infusion in the ICU and 5 had been treated with SQ insulin on the general medicine floor. Three of the patients had other diagnoses (i.e., not recurrent DKA) accounting for their gap reopening. Only one patient was restarted on continuous insulin infusion, and the remaining patients either had no change in treatment or had a titration of SQ insulin. Based on the chart review and the fact that hospital LOS did not increase after the intervention, we do not believe the increase we observed resulted in patient harm. Hypoglycemia, treatment-induced hypokalemia, and inpatient mortality remained unchanged.

We were surprised at not being able to demonstrate a cost savings, despite dramatically decreasing ICU admission rates. This is likely because there was no decrease in hospital LOS, which would have likely been associated with a more dramatic cost reduction.

The strength of our study is that it demonstrates a real-world example of the use of a subcutaneous insulin protocol rather than continuous insulin infusion for nonsevere DKA. This approach has been studied in randomized controlled clinical trials, but has not been widely adopted, or as far as we know, been implemented into a protocol to manage patients with DKA. Despite multiple prior studies demonstrating safety and efficacy of frequent doses of short-acting insulin compared with a continuous insulin infusion in mild or moderate DKA, the mainstay of DKA treatment has remained continuous insulin infusion.5-9 The reason for this is likely multifactorial. One reason may be lack of awareness. Other reasons could be the perceived ease and comfort with use of a continuous insulin infusion, as well as general concern about adopting randomized controlled trial data into routine clinical practice. Other studies have demonstrated the safety and efficacy of continuous insulin infusion outside of the ICU.18,19 However, many institutions, including ours, do not allow use of continuous insulin infusion outside of the ICU because of the intensity of nursing care required.


Implementation of an EHR-driven, standardized care pathway for DKA that incorporated general medicine floor (non-ICU) admission and treatment with subcutaneous insulin for patients with mild or moderate DKA led to decreased ICU utilization and improved patient education.


It is our hope that this study will encourage other institutions with high ICU utilization for DKA to rethink their management and consider either adjusting hospital policies regarding ICU admission, or implementing a similar care pathway using subcutaneous insulin. This study can provide a framework for other institutions aiming to improve their DKA care or decrease their ICU utilization for DKA.

Authors' Biographies

Karli Edholm, MD, FACP, was previously an Assistant Professor of Medicine, the Inpatient Medical Director for the Division of General Internal Medicine, and the Director of Quality Improvement for the hospitalist group at the University of Utah (Salt Lake City, UT). During her tenure at the University of Utah, she led and implemented multiple quality improvement initiatives. She recently transitioned to working as a hospitalist at Bozeman Health Deaconess Hospital (Bozeman, MT).

Katie Lappé, MD, FACP, is an Internal Medicine Hospitalist at the University of Utah Hospital and Assistant Professor of Medicine at the University of Utah School of Medicine and George E. Wahlen Veterans Affairs Hospital. Dr. Lappé is an Associate Program Director for the Internal Medicine residency program with a focus on assessment. She has participated in multiple hospitalist-led quality improvement initiatives.

Polina Kukhareva, PhD, MPH, is a research associate in the Department of Biomedical Informatics at the University of Utah. She collaborates with physician-investigators and contributes her expertise in statistical methods, data extraction, and study design to several projects that leverage health information technology (IT) to improve patient care.

Christy Hopkins MD, MPH, MBA, FACEP, is an Associate Professor in the Division of Emergency Medicine and the Emergency Services Medical Director at University of Utah Health. She also serves as the chief value officer (CVO) for Emergency Services.

Nathan D. Hatton, MD, MS, is an Associate Professor of Medicine at the University of Utah. He specializes in Pulmonary and Critical Care Medicine. He previously served as the Co-director of the Medical Intensive Care Unit and current serves as the Co-director of the Pulmonary Hypertension/Dyspnea clinic at the University of Utah.

Benjamin Gebhart, PharmD, is a Clinical Pharmacy Specialist in the Medical Intensive Care Unit at University of Utah Health (Salt Lake City, UT).

Heather Nyman, PharmD, is a clinical pharmacist in acute care internal medicine at the University of Utah Hospital. She is an Associate Professor (Clinical) in Pharmacotherapy at the University of Utah College of Pharmacy where she serves as Vice Chair of Clinical Practice.

Emily Signor, MD, is a Chief Medical Resident for the Internal Medicine Residency Program at the University of Utah.

Mikyla Davis, BSN, RN, is a Clinical Staff Nurse Educator for University of Utah Health.

Kensaku Kawamoto, MD, PhD, MHS, is an Associate Professor and Vice Chair for Clinical Informatics in the Department of Biomedical Informatics at the University of Utah, as well as the Associate Chief Medical Information Officer for University of Utah Health. He is also a Fellow of the American College of Medical Informatics, Board Member of Health Level Seven International (HL7), a Member of the U.S. Health Information Technology Advisory Committee (HITAC), and Co-Chair of the HITAC Interoperability Standards Priorities Task Force.

Stacy A. Johnson, MD, has been an Internal Medicine Hospitalist at the University of Utah Hospital for the past 10 years, and is an Associate Professor of Medicine at the University of Utah School of Medicine. He also serves as the Medical Director for the University of Utah Health Thrombosis Service. In this role he is active in patient outcomes research with a focus on thromboembolism and bleeding events.


1. Tieder JS, McLeod L, Keren R, et al. Variation in resource use and readmission for diabetic ketoacidosis in children's hospitals. Pediatrics. 2013;132(2):229-236.
2. Gershengorn HB, Iwashyna TJ, Cooke CR, Scales DC, Kahn JM, Wunsch H. Variation in use of intensive care for adults with diabetic ketoacidosis. Crit Care Med. 2012;40(7):2009-2015.
3. Thuzar M, Malabu UH, Tisdell B, Sangla KS. Use of a standardised diabetic ketoacidosis management protocol improved clinical outcomes. Diabetes Res Clin Pract. 2014;104(1):e8–e11.
4. Bull SV, Douglas IS, Foster M, Albert RK. Mandatory protocol for treating adult patients with diabetic ketoacidosis decreases intensive care unit and hospital lengths of stay: Results of a nonrandomized trial. Crit Care Med. 2007;35(1):41-46.
5. Umpierrez GE, Latif K, Stoever J, et al. Efficacy of subcutaneous insulin lispro versus continuous intravenous regular insulin for the treatment of patients with diabetic ketoacidosis. Am J Med. 2004;117(5):291-296.
6. Umpierrez GE, Jones S, Smiley D, et al. Insulin analogs versus human insulin in the treatment of patients with diabetic ketoacidosis: A randomized controlled trial. Diabetes Care. 2009;32(7):1164-1169.
7. Umpierrez GE, Cuervo R, Karabell A, Latif K, Freire AX, Kitabchi AE. Treatment of diabetic ketoacidosis with subcutaneous insulin aspart. Diabetes Care. 2004;27(8):1873-1878.
8. Karoli R, Fatima J, Salman T, Sandhu S, Shankar R. Managing diabetic ketoacidosis in non-intensive care unit setting: Role of insulin analogs. Indian J Pharmacol. 2011;43(4):398-401.
9. Ersoz HO, Ukinc K, Kose M, et al. Subcutaneous lispro and intravenous regular insulin treatments are equally effective and safe for the treatment of mild and moderate diabetic ketoacidosis in adult patients. Int J Clin Pract. 2006;60(4):429-433.
10. Kitabchi AE, Umpierrez GE, Miles JM, Fisher JN. Hyperglycemic crises in adult patients with diabetes. Diabetes Care. 2009;32(7):1335-1343.
11. Chang DW, Shapiro MF. Association between intensive care unit utilization during hospitalization and costs, use of invasive procedures, and mortality. JAMA Intern Med. 2016;176(10):1492-1499.
12. O'Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton CM. Measuring diagnoses: ICD code accuracy. Health Serv Res. 2005;40(5 pt 2):1620-1639.
13. Institute for Safe Medication Practices. ISMP Guidelines for Optimizing Safe Subcutaneous Insulin Use in Adults. Accessed November 22, 2019.
14. Kawamoto K, Martin CJ, Williams K, et al. Value driven outcomes (VDO): A pragmatic, modular, and extensible software framework for understanding and improving health care costs and outcomes. J Am Med Inform Assoc. 2015;22(1):223-235.
15. Lee VS, Kawamoto K, Hess R, et al. Implementation of a value-driven outcomes program to identify high variability in clinical costs and outcomes and association with reduced cost and improved quality. JAMA. 2016;316(10):1061-1072.
16. CPI Adjustment Tables—HRSA. Accessed June 6, 2018.
17. Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173(6):676-682.
18. Nemecek BD, Hermayer KL, Arnold PC, Bohm NM. Evaluation of ward management of diabetic ketoacidosis. Clin Diabetes. 2014;32(3):100-104.
19. Smiley D, Rhee M, Peng L, et al. Safety and efficacy of continuous insulin infusion in noncritical care settings. J Hosp Med. 2010;5(4):212-217.

diabetic ketoacidosis; clinical care pathway; electronic health record; value-based care

Supplemental Digital Content

© 2020 National Association for Healthcare Quality