The Case 

A 24-year-old, Arabic-speaking woman with a history of type 1 diabetes mellitus, gastroparesis, and severe erosive esophagitis presented to the emergency department (ED) with one day of worsening generalized abdominal pain, nausea, and vomiting. Her last dose of insulin was one day prior to presentation. She stopped taking her daily regimen of insulin because she was not tolerating any oral intake. Her past medical history was notable for a hospitalization two weeks earlier for the same complaints; at that time, her laboratory studies were normal, and she was diagnosed with severe erosive esophagitis by esophagogastroduodenoscopy.  

 Upon admission to the hospital, the patient was tachycardic (heart rate in 110’s) and had diffuse abdominal tenderness. Laboratory studies were significant for hyperglycemia (blood sugar level of 387 mg/dL), an elevated serum beta-hydroxybutyrate, and mildly abnormal venous blood gases (pH 7.4, PCO2 37 mm Hg, HCO3 18 mmol/L), as well as an anion gap of 23 (anion gap=[Na+] - [Cl + CO3]), which was elevated and indicative of metabolic acidosis.      

The admitting team thought the patient did not meet criteria for diabetic ketoacidosis (DKA) because the pH of 7.4 on the venous blood gas (VBG) was normal. The working diagnoses upon admission were severe esophagitis due to gastroparesis and poorly controlled diabetes. Dextrose-containing fluids were started as the patient was not tolerating any oral intake and an insulin sliding scale was ordered. Despite administration of increasing doses of subcutaneous insulin, maximal doses of prokinetics, and mucosal protective agents, she had persistent hyperglycemia in the 300-400 mg/dL range, abdominal pain, and vomiting. During the third night of hospitalization, the patient had a point-of-care fingerstick glucose measurement of 438 mg/dL. She reported having blurry vision, “not feeling well,” and fell into the nurse’s arms. The following day, she had a near fall, blurry vision, and change in mental status. An endocrinologist was consulted due to her poorly controlled blood glucose; the consultant diagnosed DKA. The patient was transferred to the intensive care unit (ICU) and an insulin drip was started, after which the patient’s metabolic derangements normalized and her symptoms resolved.  

The Commentary 

By Deborah Plante, MD, and Andrea Gonzalez Falero, MD 

This case presents two of the most common medical errors: misdiagnosis and delayed diagnosis.¹ On admission, the admitting physicians missed the diagnosis of diabetic ketoacidosis (DKA) in a patient with type 1 diabetes; the correct diagnosis was not established until 3 days later. Several factors may have contributed to this occurrence including misinterpretation of her laboratory tests on presentation, particularly her acid-base disorder, in the setting of type 1 diabetes. The clinicians failed to appreciate the importance of physiologic insulin replacement in patients with type 1 diabetes, and the consequences of withholding insulin when these patients are not eating. Insufficient or missed insulin doses are among several factors, including infection and illness, that can precipitate DKA in patients with type 1 diabetes.² At the time of admission, it was felt the patient was not experiencing DKA due to the normal pH on her venous blood gas (VBG), and her recent hospitalization for similar symptoms (although her laboratory tests were normal on that visit). But despite the normal pH, the patient had a high anion gap (AG) metabolic acidosis from DKA, with concurrent metabolic alkalosis from extensive vomiting. The diagnostic criteria for DKA center on high-AG metabolic acidosis with hyperglycemia; however, patients often present with more than one abnormality leading to mixed acid-base disorders, making interpretation more difficult. Initial evaluation of acid-base disorders always includes calculation of the anion gap,³ which represents the difference in concentration between the measured and unmeasured anions and cations. A normal AG represents the non-bicarbonate buffers, while an elevated AG suggests abnormal accumulation of endogenous or exogenous unmeasured anions such as ketones, as in this case.⁴ As we describe further below, mixed acid-base disorders can be identified by calculating the compensatory response to the primary abnormality and, when the values fall outside of what is expected, concluding that another primary disorder probably exists. 

The diagnosis may have been further delayed in this case due to the patient being Arabic-speaking, and therefore unable to communicate easily without an interpreter. Inadequate education about how to dose insulin on sick days and how to recognize early symptoms of DKA may have contributed to the initial delay in her presentation to the emergency department (ED). Specifically, she withheld insulin when she was unable to eat and may have not been able to communicate her symptoms of DKA properly to physicians without an interpreter.   

Failure to consult or communicate with subspecialty providers early in the hospital course for glycemic management also contributed to further delay in diagnosis in this case. Many academic and some community hospitals now employ a specialized inpatient diabetes team and standard criteria for when to seek consultation, including blood glucose parameters such as two blood sugars greater than 180, all hyperglycemic emergencies such as DKA and hyperosmolar hyperglycemic syndrome (HHS), patients with type 1 diabetes, or patients on insulin pumps. Some hospital systems have alerts or run reports for patients with hyper- and hypoglycemia that generate automatic consultations for the inpatient diabetes team.    

Approach to Improving Safety & Patient Safety Target 

The process of admitting patients from the ED starts with the handoff and communication between ED staff and the inpatient team. Effective communication between medical teams can improve patient safety and decrease preventable errors.⁵ The handoff time is particularly vulnerable to deficiencies in communication. Closed-loop communication, which would have been helpful in this case,⁶ is the process of acknowledging the receipt of information and clarifying with the sender of the communicated message that the information received is the same as the original, intended information. In essence, it is the process of confirming and cross-checking information for accuracy.⁶  In this case, there was a missed opportunity at the ED-inpatient handoff to review the patient’s high-AG metabolic acidosis in the context of her type 1 diabetes and missed insulin doses.    

A stepwise approach to interpreting disturbances in acid-base and electrolyte values while correlating these findings with the underlying clinical condition (i.e., type 1 diabetes) is crucial. Handoff periods are often rushed, allowing providers insufficient time to discuss laboratory findings in correlation with the evolving history and physical exam. Mixed acid-base disorders are complex and can be intimidating to most physicians, particularly when initial laboratory values are not consistent with a simple acid-base abnormality and physicians need to recall equations to evaluate whether appropriate compensation is occurring. Therefore, having a practical stepwise approach to interpreting mixed acid-base disorders, such as the one Dr. Haber presented at UCSF Medical Grand Rounds (see below), can demystify the complexity.³     

Step 1: Look at pH to determine acidosis (<7.35) or alkalosis (>7.45). 
Step 2: Calculate anion gap (AG = Na-(Cl+HCO3) = 137-(96+18) = 23 in this case). If the anion gap is ≥20, there is a primary metabolic acidosis, regardless of pH or HCO3 level. 
Step 3: Calculate excess anion gap (delta gap): total anion gap minus the normal anion gap (12 mmol/L), and add this value to the measured bicarbonate concentration. If the sum is greater than a normal serum bicarbonate (23 mmol/L), there is an underlying metabolic alkalosis. If the sum is less than normal bicarbonate (<23 mmol/L), there is a non-anion gap metabolic acidosis.³ 

In this case, the delta gap (23-12 = 11 mmol/L) added to the measured bicarbonate (18 mmol/L) was 29 mmol/L, which was greater than the normal bicarbonate level (23mmol/L), suggesting a coexisting metabolic alkalosis.³ If the team had recognized this combination of high AG metabolic acidosis from DKA and metabolic alkalosis from vomiting, then appropriate treatment could have been initiated without delay. Therefore, taking a stepwise approach that included calculating the anion gap and excess anion gap would have prevented the correct diagnosis from being overlooked.    

It is very important for clinicians to be able to diagnose diabetic ketoacidosis (DKA) early, as it is a life-threatening yet preventable condition that is most commonly seen in patients with type 1 diabetes, but it can also occur in patients with type 2 diabetes. Common causes or contributing factors include stress from illness, decreased oral intake with variable insulin needs, skipped or missed insulin doses, and malfunction of an insulin pump.² Although mortality from DKA decreased from 2009 to 2014, hospitalizations have increased by 54% during this time period, causing significant morbidity and increased inpatient costs.⁷ The reasons for this increase in hospitalizations are not well understood; however, proposed explanations include: more admissions for mild DKA (criteria for which were established by the American Diabetes Association in 2001), recurrent admissions related to improper education about proper insulin administration and symptoms of DKA, and inadequate care for diabetes due to lack of access to care or the high costs of insulin in the United States.²  

Even if the patient in this case did not have DKA, the order for “sliding scale” insulin administration was not appropriate, given her underlying diagnosis of type 1 diabetes. Patients with type 1 diabetes are insulin deficient due to antibody-mediated destruction of beta cells, so they always require physiologic insulin replacement, which includes basal insulin to cover the body’s sugar production, nutritional insulin to cover carbohydrate intake, and a correction to account for any mismatch. The American Diabetes Association (ADA) guidelines suggest that physiologic insulin replacement should be used in hospitalized patients; the sole use of “sliding scale” insulin (i.e., doses starting at 0 units, based on point-of-care blood glucose values) is strongly discouraged for all patients with diabetes, and is never appropriate in patients with type 1 diabetes.⁸ In this case, the patient was treated with “sliding scale” insulin initially and endocrinology was not consulted until she had worsening hyperglycemia and clinical deterioration. A recent national survey showed that many hospitals do not employ order sets consistent with physiologic insulin in accordance with ADA guidelines.⁹  Leveraging the electronic health record (EHR) to include order sets for physiologic insulin, with clinical decision support (CDS) tools to guide providers in appropriate use of insulin, can prevent errors and reduce rates of hyper- and hypoglycemic events.¹⁰ In addition, consulting a specialized diabetes team within 24 hours of admission has been shown to decrease length of stay (LOS), costs, and readmission rates.¹¹ Having defined criteria for when endocrinology or a specialized diabetes team should be consulted, and even automating the consultation process, would be expected to reduce delays in diagnosis and effective treatment of hyperglycemic emergencies such as DKA or HHS.   

Finally, professional interpreters should be used to improve communication in clinical encounters. A systematic literature search performed in 2007 found that use of professional interpreters improved communication and comprehension, utilization, clinical outcomes, and satisfaction with care.¹² Although it was unclear how often a professional interpreter may have been used when communicating with the patient in this case, her lack of English proficiency likely contributed to the delay and errors in her diagnosis.   

Systems Change Needed/Quality Improvement Approach 

Given the complexity of mixed acid-base disorders, it is imperative that clinicians review laboratory results in their entirety and correlate them with underlying clinical conditions rather than assume the readmission diagnosis is the same as for the patient’s last visit. To support this process, calculating tools and system-wide alerts, such as an automatic anion gap calculator with an alert for elevated levels, should be embedded in the EHR. Closed-loop communication of abnormal test results and imaging studies has been shown to decrease medical errors while improving communication amongst interdisciplinary teams.¹³ Clinical decision support tools and corresponding order sets for physiologic insulin and hyperglycemic emergencies should be added to the EHR to improve inpatient insulin therapy and make it safer. For example, links to CDS tools offering criteria for mild, moderate, and severe DKA, with corresponding evidence-based insulin, fluid, and electrolyte replacement strategies, can be added at the top of the order entry screen.   

Education is one of the most important means through which to empower physicians, trainees, and nurses to improve safety associated with complex disorders and high-risk medications such as insulin. Allowing insulin infusion outside critical care units for patients who are medically stable simplifies medical decision making but requires education of the nursing staff involved. Our institution has designed online education modules for continuing medical education (CME) that coincide with CDS tools and order sets on physiologic insulin therapy to help educate all physicians and nursing staff. Furthermore, these modules were made to require a minimum pass rate to ensure that the learning objectives have been met. Educating trainees, nurses, and physicians regarding criteria for DKA and HHS, and management of type 1 diabetes—especially the need for basal insulin at all times —is imperative. Patients with type 1 diabetes should also be educated about management of their sick days, how to trouble-shoot pump malfunction, and how to develop a backup plan with their care providers.  

Hospitals should also employ dedicated multidisciplinary teams for inpatient diabetes care, to guide the hospital-wide effort to align processes and education with standards of care. These teams should include physicians, dieticians, pharmacists, nurses, and advanced practice providers. Creating a specialized diabetes team that is separate from the endocrinology consultation service, and defining a set of criteria for when to consult with that team, should result in shorter LOS (if consulted early), and decreased costs and readmission rates. And encouraging communication between teams, especially when glycemic management is not being met, or when clinical deterioration is noted, is vital.  

Take-Home Points 

• Calculated AG should be integrated into laboratory reports of metabolic panels, with an alert for abnormal values. 
• Closed-loop communication should be utilized to decrease medical errors while improving communication within and between interdisciplinary teams. 
• Physicians should be educated on the use of a stepwise practical approach to diagnosing acid-base disorders. 
• Institute a specialized diabetes team for inpatient management along with established criteria for when (early) to consult with the team’s members. 
• Create order sets and CDS tools to guide physiologic insulin replacement, in accordance with ADA standards of care. 
• Educate patients, nursing staff, and physicians on risk factors for DKA and its symptoms in patients with type 1 diabetes. 

Deborah Plante, MD 
Clinical Professor of Medicine 
Director, Inpatient Glycemic Team 
Division of Endocrinology, Diabetes and Metabolism, Department of Medicine 
UC Davis Health  

Andrea Gonzalez Falero, MD  
Endocrine Fellow 
Department of Endocrinology, Division of Internal Medicine 
UC Davis Health 

References  

1. Carver N, Gupta V, Hipskind JE. Medical Error. [Updated 2020 Oct 5]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2020 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK430763/ 

2. Vellanki P, Umpierrez GE. Increasing Hospitalizations for DKA: A Need for Prevention Programs. Diabetes Care. 2018;41(9):1839-1841. doi:10.2337/dci18-0004 PubMed citation

3. Haber RJ. A practical approach to acid-base disorders. West J Med. 1991;155(2):146-151. PubMed citation

4. Tsapenko MV. Modified delta gap equation for quick evaluation of mixed metabolic acid-base disorders. Oman Med J. 2013;28(1):73-74. doi:10.5001/omj.2013.18 PubMed citation

5. Abbott RL, Weber P, Kelley B. Medical professional liability insurance and its relation to medical error and healthcare risk management for the practicing physician. Am J Ophthalmol. 2005;140(6):1106-1111. doi:10.1016/j.ajo.2005.07.020 PubMed citation

6. Agency for Healthcare Research and Quality (AHRQ). TeamSTEPPS Fundamentals Course: Module 3. Communication (July 2018). Available at: https://www.ahrq.gov/teamstepps/instructor/fundamentals/module3/slcommunication.html. Accessed April 19, 2021.  

7. Centers for Disease Control and Prevention (CDC). National Diabetes Statistics Report, 2020. Atlanta, GA: Center for Disease Control and Prevention, U.S. Dept of Health and Human Services; 2020. Available at https://www.cdc.gov/diabetes/library/features/diabetes-stat-report.html

8. American Diabetes Association. 15. Diabetes Care in the Hospital: Standards of Medical Care in Diabetes-2019. Diabetes Care. 2019;42(Suppl 1):S173-S181. doi:10.2337/dc19-S015 PubMed citation

9. DelCarpini H. U.S. Survey of Healthcare Professionals: Inpatient Diabetes Care and Glycemic Management. Focus on Insulin Therapy and Hypoglycemia. Published February 1, 2018. Available at: https://glytecsystems.com/resource/u-s-survey-of-healthcare-professionals-inpatient-diabetes-care-and-glycemic-management-focus-on-insulin-therapy-and-hypoglycemia/. Accessed April 5, 2021. 

10. Mathioudakis N, Jeun R, Godwin G, et al. Development and Implementation of a Subcutaneous Insulin Clinical Decision Support Tool for Hospitalized Patients. J Diabetes Sci Technol. 2019;13(3):522-532. doi:10.1177/1932296818798036 PubMed citation

11. Bansal V, Mottalib A, Pawar TK, et al. Inpatient diabetes management by specialized diabetes team versus primary service team in non-critical care units: impact on 30-day readmission rate and hospital cost. BMJ Open Diabetes Res Care. 2018;6(1):e000460.  doi:10.1136/bmjdrc-2017-000460 PubMed citation

12. Karliner LS, Jacobs EA, Chen AH, Mutha S. Do professional interpreters improve clinical care for patients with limited English proficiency? A systematic review of the literature. Health Serv Res. 2007;42(2):727-754. doi:10.1111/j.1475-6773.2006.00629.x PubMed citation

13. Lacson R, Prevedello LM, Andriole KP, et al. Four-year impact of an alert notification system on closed-loop communication of critical test results. AJR Am J Roentgenol. 2014;203(5):933-938. doi:10.2214/AJR.14.13064 PubMed citation