CLINICAL PRACTICE

Hospital Management of DiabeticKetoacidosis: Are Clinical GuidelinesImplemented Effectively?R.K. Singh, P. Perros, B.M. Frier

Department of Diabetes, Royal Infirmary of Edinburgh, Edinburgh,UK

A study was undertaken in order to examine the quality of management of diabeticketoacidosis (DKA) in a teaching hospital and to assess whether the introduction of clinicalguidelines contributed to a satisfactory outcome. Data on presentation and managementof 71 cases of DKA admitted in one calendar year (1994) were collected and analysed.Comparing the data to standards set in guidelines, inadequacies of clinical managementwere identified including delay in initiation of intravenous fluid replacement (greater than30 min) and intravenous insulin (greater than 60 min) in 70 % and 69 % of cases,respectively; under-replacement with intravenous fluid in the first 24 h (less than 6.5 l) in70 % of cases, and insufficient intravenous potassium replacement (less than 70 mmol) inthe first 24 h in 70 % of cases. Suboptimal management of DKA may have contributedtowards death in one of the three fatalities, and to morbidity in other patients. In 22.5 %of cases (group 1) in whom the guidelines were alleged to have been followed, intravenousfluid, potassium, and insulin had been administered earlier and in larger quantitiescompared to the remaining cases (group 2). However, in most cases in group 1 thestandards set by the guidelines were unfulfilled and the incidence of hypokalaemia,hypoglycaemia, and duration of in-patient stay did not differ from group 2. The treatmentof DKA by non-specialist general medical staff in a large teaching hospital was frequentlyinadequate and was associated with significant mortality and morbidity. The introductionof guidelines had moderately influenced the process of managing DKA but not theoutcome, probably because of the low rate of their implementation by junior doctors. 1997 by John Wiley & Sons, Ltd.

No of Figures: 1. No of Tables: 3. No of Refs: 15.

KEY WORDS Insulin-dependent diabetes mellitus Diabetic ketoacidosis InsulinHypokalaemia Hypoglycaemia Clinical guidelines

Received 5 August 1996; revised 21 January 1997; accepted 25 January 1997

availability was complemented by lectures to juniorIntroductionmedical staff every 6 months as part of a post-graduateteaching programme. The specialist diabetes service isDiabetic ketoacidosis (DKA) is a life-threatening medical

emergency, the mortality from which declined in the not routinely involved in the management of mostepisodes of acute DKA admitted to our hospital, which1960s and 1970s because of earlier recognition, greater

patient education, and improved in-patient manage- is the primary responsibility of the general medicalreceiving team. Although a specialist on-call diabetesment.1–2 However the mortality remains unacceptably

high at 4.7–10.4%,3,4 and DKA is thought to be the most service has been available continuously for several yearsbefore the present survey, this had seldom been contactedcommon cause of death in diabetic patients under the

age of 20 years.5 Furthermore, some reports have by other physicians for the initial management of patientsadmitted with DKA.indicated that the incidence of DKA is not declining and

may even be rising.3,6 The aim of the present study was to document theincidence, causes, and complications of DKA over aOn admission to hospital, junior medical staff usually

initiate treatment and determine management protocols period of 12 months and to assess the quality ofmanagement of DKA in the hospital subsequent to thefor patients with DKA. To facilitate and improve the

management of DKA by junior medical staff, guidelines introduction of guidelines for non-specialist medical staff.based on a European consensus statement (EuropeanIDDM Policy Group)7 were produced for use in a large Patients and Methodsteaching hospital. The guidelines were distributed to allmedical, surgical and high dependency areas, and their Guidelines were formulated for the management of DKA,

based on a European consensus statement (EuropeanIDDM policy Group)7 for the use of junior doctors* Correspondence to: Dr B.M. Frier, Department of Diabetes, Royal

Infirmary of Edinburgh, Lauriston Place, Edinburgh EH3 9YW, UK. providing acute medical services in our hospital. This is

482 CCC 0742–3071/97/060482–05$17.50 1997 by John Wiley & Sons, Ltd. DIABETIC MEDICINE, 1997; 14: 482–486

CLINICAL PRACTICEa large teaching hospital with 990 beds, of which 288 I Urinalysis for glucose, ketones, and protein at least

once in the first 12 h.are designated for general medicine. The guidelines wereapproved by the Clinical Directorate of General Medicine I Nasogastric intubation in comatose patients.and distributed to all junior medical staff participating

Continuous ECG monitoring was recommended in allin the acute medical admissions rota, and also to allpatients. Administration of intravenous bicarbonate wasmedical wards, the Accident and Emergency Department,recommended only in severely acidotic patients (arterialand High Dependency areas.pH , 7,[H+] . 100 nmol−1) whose acidosis failed toThe cases of DKA which were diagnosed and treatedimprove after 1 h of therapy with intravenous saline andin the year after the introduction of the guidelines wereinsulin. The target plasma glucose concentration in theidentified by the Medical Records Department based onfirst 24 h was 12–15 mmol l−1. Avoidance of hypokalae-coding of discharge summary diagnoses. A computermia (serum potassium concentration less thansearch identified 84 possible entries. Two sets of case3.5 mmol l−1) and hypoglycaemia (plasma glucose lessnotes could not be found. Four additional cases werethan 3.5 mmol l−1) was advised for all patients. In theidentified serendipitously by the authors among patientsguidelines, the decision whether to administer antibioticsattending the Diabetic Clinic whose case summaries hador heparin was left to the discretion of the admittingbeen coded incorrectly on discharge from hospital. Theteam and depended upon their clinical assessment ofcase notes were reviewed and the diagnosis of DKA wasthe probability of underying infection and the riskaccepted if the following criteria were met:of thrombosis.

Serum electrolytes, plasma glucose and arterial blood1. hyperglycaemia (plasma glucose greater thangases were measured by the local Clinical Biochemistry11 mmol l−1);Laboratory. Total glycated haemoglobin (HbA1) was2. metabolic acidosis with a serum bicarbonate concen-assayed by HPLC (non-diabetic range less than 8 %).tration less than 15 mmol l−1 or arterial hydrogen ion

Results are expressed as the median values (and range).concentration higher than 48 nmol l−1 (pH , 7.31),Statistical analyses were performed using Wilcoxonassociated with ketonuria.Ranked Sum tests and chi-squared tests.

Fifteen patients were excluded because of failure to meetthe above diagnostic criteria for DKA. ResultsThe case notes were then reviewed for demographicdetails, previous episodes of DKA, clinical findings on

Baseline Characteristicspresentation, time elapsed from admission (i.e. from timefirst seen by medical staff at the Accident and Emergency

Clinical details of the patients with DKA at the time ofDepartment) to initiation of intravenous fluid and insulinadmission to hospital are shown in Table 1. Seventy-therapy, amounts of intravenous fluid, potassium andone episodes of DKA were recorded in 53 patients (31insulin administered, other interventions performed, bio-male and 22 female). Five patients who had episodes ofchemical parameters measured and the frequency ofrecurrent DKA were responsible for 16 events. Thebiochemical measurement in the first 24 h. In addition,median age of the entire group (n = 71) was 33 yearsmortality and other outcome measures (length of in-(range 17–80) years, and the median duration of diabetespatient stay, incidence of hypokalaemia, hypoglycaemia,prior to presentation was 10 (range 0–32) years. Eightand incidence of intercurrent illness including aspirationper cent were newly diagnosed IDDM. In 37 casespneumonia) and frequency of involvement of the on-(52 %) a previous history of DKA was identified. Forty-call diabetes service were recorded.four cases (56 %) had attended the outpatient diabeticThe management of each episode of DKA wasclinic of the same hospital previously, but 20 of thesecompared to the standards recommended in the guide-(45 %) were no longer attending regularly at the time oflines. These included:the acute illness. The result of HbA1 at the time ofadmission was available in 57 cases (80 %), in theI Time taken to start intravenous fluid replacement in

the Accident and Emergency Department majority of whom (88 %) the concentration was greaterthan 10 % (median 12 %, range 8.2–15 %), indicating(recommended less than 30 min).

I Whether 6.5 l of fluid had been infused intravenously poor glycaemic control.Infection was implicated as a precipitant of the DKAin the first 24 h (with provisions for adjusting the

volume of fluid infused according to the clinical state in only 15 % of cases and deliberate discontinuation ofinsulin therapy was suspected in 39 % of cases. Thisof the patient).

I Whether intravenous insulin had been started within group of patients with poor compliance was younger(median age 28.5, range 17–72 years vs 36, range 18–60 min. of admission.

I Replacement with at least 70 mmol of potassium in 80 years, p , 0.05), had poorer glycaemic control(median HbA1 12.3, range 9.2–15.0 % vs 10.6, rangethe first 24 h.

I Monitoring of biochemical parameters at least 5 times 8.2–14.3 %, p , 0.05) and had a higher default ratefrom the Diabetic Clinic (39 % vs 20 %, p , 0.05),in the first 24 h.

483DIABETIC KETOACIDOSIS AND CLINICAL GUIDELINES

CLINICAL PRACTICETable 1. Clinical details of patients with DKA at the time of Table 2. Management of DKA and comparison to rec-

ommended guidelinesadmission to hospital

Age (yr) (median, range) 33 (17–80) Median (range) % cases satisfyingrecommendedSex (F/M) 22/31

Duration of IDDM (yr) (median, 10 (0–32) standardrange)Patients with previous history of 52 Time from admission to 80 (0–330) 30a

DKA (%) i.v. fluid (min)Precipitating factors of DKA

Time from admission to 120 (2–360) 31b(number (%))

continuous i.v. insulinUnknown 27 (37)(min)Lack of compliance with insulin 28 (39)

therapy Volume of intravenous 4.8 (1–12) 30c

Infection fluid (l) infused in first 24 hUrinary tract 5 (8)

Amount of intravenous 60 (0–180) 30dGastroenteritis 3 (4)

potassium (mmol) infusedPneumonia 2 (3)in first 24 hNew diagnosis of IDDM 5 (8)

Myocardial Infarction 1 (1) Frequency of biochemical 4 (1–7) 4 (2–5)eAdmission source (number (%)) monitoring in first 24 h

GP 23 (33) (number of times)Accident and emergency 37 (54)

Nasogastric intubation in 2 33fDiabetic clinic 3 (4)

comatose patientsTransfer from other hospitals 6 (9)Other 2 (3) DKA guidelines followed 16 22.5

Conscious level (number (%))Referral to diabetes team 22 47Comatose (GCS 3) 6 (9)

Confused/agitated 19 (28)Adjunctive therapy (number (%))Alert 42 (63)

Antibiotics 20 (28)Vital signs (median (range))Electrocardiogram 43 (61)Tachycardia (. 100 beats.min−1) 107 (26–170)Urinary catheter 22 (31)Hypotension (systolic 15 (21)Intravenous bicarbonate 6 (8)BP , 100 mmHg)Central venous line 13 (18)Hypothermia (T , 35°C) 11 (15)Heparin 7 (10)Biochemical parameters (median

(range))aStarting intravenous fluid within 30 min of admissionPlasma glucose (mmol l−1) 38 (11.2–93.1)bStarting intravenous insulin within 60 min of admissionArterial hydrogen ion (nmol l−1) 69 (48–271)cMinimum of 6.5 l of fluid infused in first 24 hArterial pH 7.16 (6.57–7.32)dMinimum of 70 mmol in first 24 hArterial bicarbonate (mmol l−1) 8.5 (2.7–15)eFive times in first 24 hPlasma potassium (mmol l−1) 5.0 (3.5–8.0) fNasogastric intubation in all comatose patientsPlasma urea (mmol l−1) 7.5 (1.6–30.7)

plasma glucose by a laboratory method and of serumpotassium concentrations in the initial 24 h after admis-than the remaining patients. In 37 % of episodes no

precipitating factor could be identified (Table 1). sion was less than the minimum recommended numberof at least five measurements in 41 (58 %) cases (medianfrequency of plasma glucose and potassium estimationsManagementwas 4, range 1–7 times). Arterial blood gas estimationswere performed twice or less in the first 24 h in 61 %The time from admission to the initiation of effective

therapy varied widely. Intravenous fluids were not started of cases (median frequency of 2, range 0–8 times). Onlytwo of six comatose patients had a nasogastric tubeuntil a median of 80 (range 0–330) min and in 70 % of

cases were delayed beyond 30 min (Table 2). The time inserted. In seven patients (10 %) urinalysis within 12 hof admission was not recorded in the case notes, althoughdelay from admission to the institution of intravenous

insulin infusion was also excessive (greater than 60 min) it might have been analysed. An electrocardiogram wasperformed in only 43 (61 %) of patients and in 8 ofin 69 % of cases (median 120, range 2–360 min). In

70 % of patients the total amount of intravenous fluid these (19 %), changes suggestive of electrolyte abnormali-ties were noted.infused in the first 24 h was generally less than the

recommended volume of 6.5 l (median fluid volume The DKA guidelines had allegedly been followed bythe admitting team in only 16 (24 %) cases. The number4.8, range 1–12 l). The amount of potassium chloride

administered in the first 24 h was also inadequate (less of episodes of DKA that were referred subsequently (after24 h) by the general medical team to the specialistthan 70 mmol) with a median value of 60 (range 0–180)

mmol in 70 % of cases. The frequency of monitoring of diabetes service was only 22 (47 %) out of a possible 47.

484 R.K. SINGH ET AL.

CLINICAL PRACTICEinpatient stay in group 1 was similar to group 2 (medianOutcomesstay 3.5, range 1.5–12 days vs 4, range 0.25–60days, p = NS). The incidence of hypoglycaemia andThe mortality rate was 4.2 % (Table 3). The causes of

death were myocardial infarction, septicaemia secondary hypokalaemia in the first 24 h of admission were similarin the two groups. Two patients presenting in coma into pneumonia, and acute pancreatitis; and the ages of

these patients (56–82 years) will have contributed towards whom the guidelines had not been used, developedaspiration pneumonias (Table 4). The standards set bythe poor outcome. Two comatose patients who had not

received a nasogastric tube each developed aspiration the guidelines were met in a larger proportion of patientsin Group 1 than in Group 2 (Figure 1).pneumonia. Thirteen patients (18 %) had a central venous

line inserted and in two cases the procedure was Twenty-one cases of DKA were treated by the diabetesteam who treated patients with more intravenous fluidcomplicated by the development of pneumothoraces,

one of whom required assisted ventilation. During the (median 5.5, range 1.6–11 l vs 4.5, range 0.5–12 l,p , 0.05), and initiated intravenous insulin therapy earlierfirst 24 h of treatment hypokalaemia (serum potassium

concentration lower than 3.5 mmol l−1) occurred in (median 100, range 2–250 min vs 130, range 20–330 min, p , 0.05) than other teams.22 cases (31 %), and hypoglycaemia (plasma glucose

concentration lower than 3.5 mmol l−1) in 14 of patients,with symptomatic hypoglycaemia occurring in 9 % of Discussioncases. The duration of ketonuria after admission wasdocumented in only 15 patients (median 70, range 24– In the present survey a thorough assessment was made

of the causes, clinical findings, and management of 71124 h). The median duration of intravenous therapy was42 h (range 0–192 h), and the median length of inpatient emergency medical admissions with DKA. The total

number is likely to be an underestimate of the truestay was 4 days (range 6 h–59 days). A minority of cases(30 %) had follow-up appointments arranged, or prior number of admissions with DKA, as identification of

cases relied upon discharge coding by the Medicalappointments with the Diabetes clinic confirmed. Only4 (6 %) patients were referred for further education to Records Department, which has been shown previously

to be unreliable in this hospital and restricts accuracythe hospital-based diabetes service.The cases of DKA were subdivided into those in whom of case ascertainment.8 Lack of compliance with insulin

therapy was identified in a significant proportion ofthe guidelines had been used by the admitting medicalteam (group 1, n = 16) and those in whom they were patients, who tended to be young, had poor glycaemic

control and a history of default from previous routinenot (group 2, n = 55). The median age and initialbiochemical parameters in the two groups were similar, review at the diabetic clinic. This experience is similar

to that from other centres,9,10 and deliberate discontinu-with the exception of serum bicarbonate, which waslower in group 1 (median 7.3, range 2.7–14 mmol l−1 vs ation of insulin therapy appeared to account for some

of these episodes of DKA. Many patients who present10, range 3–14 mmol l−1). Group 1 received moreintravenous fluid (6.9, range 4–12 l vs 4.2, range 0.5– acutely to hospital emergency departments with severe

hypoglycaemia have similar clinical characteristics11,1211 l, p , 0.05), earlier onset of therapy (median timefrom admission 25, range 0–85 min vs 85, range 2– suggesting that they represent a subgroup of patients

who have chaotic self-management of diabetes. The330 min, p , 0.05), and more intravenous potassiumreplacement (80, range 20–180 mmol vs 50, range 0– proportion of cases of DKA caused by untreated newly130 mmol, p , 0.05), than group 2. The duration of

Table 3. Outcome measures in cases of DKA

Number of cases(%)

Fatal 3 (4.5)Aspiration pneumoniaa 2 (3)Pneumothoraxb 2 (3)Hypokalaemia in first 24 h 22 (31)Hypoglycaemia in first 24 h 10 (14)Duration of ketonuria (hr) 70 (24–120)Median inpatient stay (days) 4 (0.25–59)Follow-up arranged or confirmed 22 (30)Copy of discharge summary sent to Dia- 33 (45)

Figure 1. Proportion of cases of DKA meeting the managementbetes clinicstandards set by the guidelines. Group 1 (open columns) wereFurther diabetic education arranged 4 (6)patients who were alleged to have been managed accordingto the guidelines, whereas group 2 (closed columns) comprisedaOccurred in comatose patients who were not treated with naso-the remaining patients. Asterisks denote statistically signifi-gastric intubation.

bResulted from central venous line insertion. cant differences

485DIABETIC KETOACIDOSIS AND CLINICAL GUIDELINES

CLINICAL PRACTICEdiagnosed IDDM was in marked contrast to figures as a relatively uncommon medical emergency and junior

medical staff may seldom have the opportunity to managehigh as 30 % reported in earlier studies13 which may bethis serious metabolic problem. Hospital medical staffattributable to better management of IDDM in general andshould recognize that DKA is often being managedincreased clinical awareness of the risk of development ofsuboptimally and where this is confirmed should pursueDKA in general practice.a solution suitable to the prevailing local circumstances.The present study has revealed that the management

of DKA was suboptimal in a significant proportion ofpatients being treated in a large teaching hospital despitethe introduction of clinical guidelines. Disconcerting

Acknowledgementfindings in the present study include the considerabledelay in initiating intravenous fluid and insulin therapyby medical staff, and the high incidence of iatrogenic The authors wish to thank the Medical Records Depart-hypokalaemia and hypoglycaemia in the initial 24 h after ment of the Royal Infirmary of Edinburgh for their co-admission to hospital. The guidelines were claimed to operation with this study.have been followed by medical staff in only 22.5 % ofcases (group 1), and careful objective assessment revealedthat even in these cases the recommended standardshad not been achieved in the majority. Outcomes in Referencesboth groups (with and without guidelines) were similar,which is likely to be related to the low rate of 1. Soler NG, Bennet MA, Fitzgerald MG, Malins JM. Intensive

care in the management of diabetic ketoacidosis. Lancetimplementation of the guidelines even among cases in1973; 1: 951–954.group 1. This occurred despite regular educational

2. Gale EAM, Dornan TL, Tattersall RB. Severely uncontrolledsessions and widespread dissemination of the writtendiabetes in the over fifties. Diabetologia 1981; 21: 25–28.

document to junior medical staff and clinical areas where 3. Basu A, Close CF, Jenkins D, Krentz AJ, Nattrass M,patients with DKA receive treatment. This experience is Wright AD. Persisting mortality in diabetic ketoacidosis.

Diabetic Med 1993; 10: 282–284.common in other areas of clinical management14 and4. Lebovitz HE. Diabetic ketoacidosis. Lancet 1995; 345:throws considerable doubt on the effective implemen-

767–772.tation of clinical guidelines which are being widely 5. Tunbridge WMGT. Factors contributing to deaths ofpromoted as a method of standardizing treatment sched- diabetics under 50 years of age. Lancet 1981; 2: 569–573.ules for many medical disorders. Continuity of medical 6. Ellemann K, Soerensen JN, Pedersen L, Edsberg B, Ortved

Andersen O. Epidemiology and treatment of diabeticcare may have been a key factor in determiningketoacidosis in a community population. Diab Care 1984;compliance with guidelines, as by our experience it is7: 528–532.

not uncommon for up to three different junior doctors 7. European IDDM Policy Group. Consensus guidelines forto be involved in the initial management of patients with the management of insulin-dependent (Type 1) diabetes.

Diabetic Med 1993; 10: 990–1005.DKA. No evaluation was made of the standards of8. Leslie PJ, Patrick AW, Hepburn DA, Scougal IJ, Frier BM.clinical management of DKA in the hospital before

Hospital in-patient statistics underestimate the morbiditythe introduction of clinical guidelines (to assess their associated with diabetes mellitus. Diabetic Med 1992; 9:influence more precisely) but the low rate of implemen- 379–385.tation would suggest that the previous quality of clinical 9. Herskowitz-Dumont R, Jacobson AM, Cole C, Hauser ST,

Wolfsdorf JI, Willett JB, Milley JE, Wertlieb D. Psychosocialmanagement was probably no different from that observedpredictors of acute complications of diabetes in youth.in the present survey.Diabetic Med 1995; 12: 612–618.

We suspect that the present observations of the local 10. Thompson CJ, Cummings F, Jung RT, Newton RW. Thedeficiencies of management of DKA in hospital may effects of an integrated education programme on the

management of DKA. Prac Diab Int 1995; 12: 235–237.reflect the current practice in many hospitals in the UK11. Collier A, Steedman DJ, Patrick AW, Nimmo GR, Matthewsproviding acute medical services, many of which are

DM, Little K, Clarke BF. Comparison of intravenousunder severe pressure.15 Possible means of addressing glucagon and dextrose in treatment of severe hypoglyce-these problems include the establishment of a computer- mia in an Accident and Emergency Department. Diabbased decision support system (currently being developed Care 1987; 10: 712–715.

12. Potter J, Clarke P, Gale EAM, Dave SH, Tattersallin our centre), the introduction of a mandatory protocolRB. Insulin-induced hypoglycaemia in an accident and(which requires monitoring and could be regarded asemergency department: the tip of an iceberg? Br Med J

impinging upon clinical freedom), and devolving the 1982; 285: 1180–1182.management of DKA principally to physicians specializ- 13. Faich GA, Fishbein HA, Ellis SE. The epidemiology of

DKA: a population-based study. Am J Epidemiol 1983;ing in diabetes. The latter is being implemented locally117: 551–558.but is impractical for centres where an out of hours

14. Haines A, Jones R. Implementing findings of research. Brspecialist service is unavailable, and may potentially Med J 1994; 308: 1488–1492.compromise junior doctors’ training in acute general 15. Christie B. Scottish bed crisis leads to review on planning

services. Br Med J 1996; 312: 204.medicine. In many centres in the UK, DKA has become

486 R.K. SINGH ET AL.