Differing protocols of managing adult diabetic ketoacidosis outside of the intensive care unit make no difference to the rate of resolution of hyperglycaemia and acidosis

Diabetic ketoacidosis (DKA) is a life threatening, complex metabolic disorder complicating diabetes and is a common cause for admission to acute medical units. We have previously1 reported our 23-year experience of managing DKA at Auckland Hospital which extended observations from a previous report.2 This showed that there was a substantial reduction in length of stay (LOS) and need for intensive care unit (ICU) admission with low in-hospital mortality over the two decades of observation, although high rates of subsequent readmission with DKA has remained an issue.3 Over this time there was no change in the DKA management protocol, which remained “glucose centric”, in that decisions regarding insulin administration were made primarily based on current blood glucose rather than pH or ketone concentrations. The severity of acidosis and degree of hyperglycaemia at presentation was less in recent years,1 perhaps reflecting more widespread use of long-acting analogue insulins, better education of people with diabetes and easier access to care among other factors, which could have contributed to the reduced need for ICU admission and shorter LOS. However, the LOS and ICU admission rate for those with severe DKA (pH ≤7.1) also showed a trend downwards in recent years, suggesting factors other than the DKA management protocol itself have led to the improvement in outcomes, such as better immediate care in the emergency department.

In 2011, the Joint British Diabetes Societies Guideline for the Management of DKA was published4 with a shift away from a glucose-centric protocol to a more ketone-centric one, with the recommendation to commence a weight-based, fixed dose insulin infusion until ketosis (measured at the bedside) clears. The evidence that this approach results in better outcomes is however limited.5 A recent UK audit of the ketone-based protocol in managing 50 episodes of DKA6 showed that hypokalaemia occurred in 46% of people, but that the protocol for potassium replacement was not followed well. Forty percent of people had a hypoglycaemic episode at a median time of approximately 13 hours after the insulin infusion was started, despite 80% correctly receiving IV dextrose as per protocol. Moreover, the switch from IV to subcutaneous insulin was appropriately managed in only 34% of cases. This study suggests that the protocol may not be easy to follow and does not prevent hypokalaemia and hypoglycaemia, possibly due to a relatively high insulin infusion rate. New Zealand is currently attempting to develop a nationally agreed protocol of care for the management of DKA with some advocating the adoption of the UK protocol of care.

Following publication of the new UK DKA guideline, North Shore Hospital in Auckland changed their DKA management protocol from a glucose-centric one to the UK, more ketone-centric protocol in 2012. This allowed a comparison in outcome of people subsequently admitted with DKA treated at the two hospitals using two different protocols.

The optimal rate of correction of the metabolic abnormalities of DKA is unclear. Very rapid correction of hyperglycaemia can lead to hypoglycaemia and hypokalaemia, and the resulting rapid change in osmolality can increase the risk of cerebral oedema especially in children.7–9 We were thus interested in comparing the performance of the two protocols in correcting the metabolic abnormalities of people admitted with DKA to the two hospitals in the year following implementation of the new protocol at North Shore Hospital.

Methods

The management of all people aged 16 years and over admitted with DKA from 1 January 2013–31 December 2013 to both hospitals were compared. DKA was defined as a laboratory glucose ≥11.1mmol/l, and pH ≤7.30 and bicarbonate ≤15mmol/l and raised beta hydroxybutyrate ≥3mmol/l. Each episode of DKA was treated as a discrete event, and thus could include multiple episodes in an individual person. A beta hydroxybutyrate >8mmol/l is reported as “>8” at Auckland Hospital’s laboratory. The admissions of people who were pregnant, had end-stage renal disease, had type 2 diabetes, or who were transferred from another hospital or who were admitted to ICU (as they may have initially been treated by a different insulin infusion protocol) were excluded from the analysis. ICU admission criteria is similar between the two hospitals, with admission to ICU guidelines including significant haemodynamic or electrolyte abnormalities or pH <7.1 as noteworthy criteria.

The two hospitals serve similar sized populations of about 400,000 (although Auckland Hospital has a higher proportion of people of non-European ethnicity), are both teaching hospitals attached to Auckland Medical School, and share a pool of general medicine registrar trainees who rotate through each hospital for 6–12 months at a time. People domiciled in the catchment area of each hospital are admitted directly to that hospital. Substantial training of emergency department, medical and nursing staff in the use of the new protocol was undertaken during 2012 at North Shore Hospital, and 16 ketone meters for measuring ketones at the patient’s bedside were purchased (at a cost of $69 each). Ketone measurement at Auckland hospital was done in the main laboratory, but at North Shore only the admission ketone value was measured in the laboratory, with all subsequent tests being done using the point-of-care meter at the bedside.

The Auckland hospital protocol encourages the continuation of long-acting insulin in those already on such an insulin. The insulin infusion rate is based on the prevailing glucose concentration rather than on patient weight or ketone concentration. The usual default first scale insulin infusion rate typically starts at 6 to 12 units per hour. If the hourly measured capillary glucose levels are not falling, the rate of insulin infused per hour is rapidly escalated by protocol and on a variable scale. Insulin continues to be infused alone until the capillary glucose falls to <15mmol/L, when insulin is continued and 10% dextrose IV is added at 80ml per hour. Fluids containing varying concentrations of potassium are also infused according to the person’s renal function and potassium concentration. Once the person is eating and drinking and glucose values are stable, subcutaneous insulin is commenced and the insulin infusion is weaned to stop. While the protocol recommends “frequent” measurement of venous bicarbonate, potassium and pH, there is no requirement to repeat the measurement of plasma beta-hydroxybutyrate after the initial diagnosis of DKA.

The North Shore protocol states that people should be given subcutaneous long-acting insulin (0.25units/kg) on the evening of admission or the next morning, and the IV insulin infusion commenced at 0.1 unit per kg of body weight/hr as soon as possible. Adjustments to the rate of infusion (1 unit per hour, every hour) are based on a combination of hourly measurements of ketones and glucose using point-of-care meters at the bedside. Venous blood gas analysis is recommended five times in the first 24 hours. When the glucose falls to <14mmol/l, 10% dextrose is commenced at 100ml/hr and both infusions are continued until the ketones have cleared and the patient is eating and drinking.

People admitted to both hospitals with DKA are seen first either by the emergency department staff before being transferred to the general medical “team of the day”, or they can be admitted directly to general medicine if the GP calls ahead of their arrival. They are managed in the acute assessment areas until stable, and then transferred to a general medical ward (or ICU). Specialist diabetes nurse or diabetologist involvement in patients’ care is by referral from the general medical team and recommended in all cases within 24 hours of admission but is only available during the week (Monday to Friday) and within normal working hours (0800–1700). Acute admissions at Auckland Hospital were allocated at 0800 each morning between four teams that were on a roster to receive all acutely admitted general medical patients (including those with DKA) admitted within the previous 24 hours. Of the 30 physicians contributing to the roster at that time, six had specific advanced training in diabetes; people with DKA were thus not specifically directed to a team with a diabetologist. At North Shore Hospital there were 15 general medical teams, with four including a physician with specific advanced training in diabetes; newly admitted patients were allocated between three teams at 0800 each day.

All people in New Zealand have a unique National Health number, which facilitates retrieval of their medical records. All records and investigations at Auckland hospital are stored electronically, but hard copy notes are still in use at North Shore (although letters and all investigations are also stored electronically). The notes (hard copy or electronic) of people admitted with DKA were examined and all relevant data extracted in retrospect.

The primary end points of interest were rate of correction of acidosis and hyperglycaemia with secondary end points being incidence of hypoglycaemia, hypokalaemia and length of stay (LOS).

Audit studies such as this are approved by the local hospital ethics committees.

Results

Data was compared using t, Chi squared, or Mann-Whitney tests and results are presented as mean ±SD (range) or when not normally distributed, as median (95% CI or range). Forty-one admissions of 35 people admitted to Auckland Hospital met all the criteria for inclusion in the study. A further 17 admissions were excluded: two were transferred from another hospital during the admission, six were admitted to ICU and nine had type 2 diabetes. Thirty admissions of 26 people were included at North Shore Hospital. A further 13 admissions were excluded: six were transferred from another hospital and seven were admitted to ICU.

The severity of biochemical derangement and clinical features of the people admitted to ICU was very similar between the hospitals; pH 7.04±0.1 Auckland vs 7.02±0.14 North Shore, glucose 35.3±8.8 vs 41.5±21mmol/l, anion gap 32±5.3 vs 32±10 and lactate 4.2±1.95 vs 3.2±1.8mmol/l. This suggests that the criteria for admission to ICU were very similar and that patients looked after on the general medical wards at the two hospitals were of similar acuity.

On admission, the degree of hyperglycaemia, pH and bicarbonate were not different between the two cohorts, but the anion gap was larger at North Shore, suggesting a greater degree of ketosis (Table 1). A similar proportion of people were admitted after-hours or at the weekend. There were more people of European ethnicity admitted with DKA to North Shore than to Auckland Hospital (Table 1), reflecting the different ethnic makeup of the two catchment areas. In the 12 months before this study, 11 (31%) of the people admitted to Auckland Hospital had had 22 admissions with DKA and nine (36%) of the people admitted to North Shore Hospital had had 12 admissions with DKA (data not shown).

Table 1: Details of admissions of people with type 1 diabetes presenting with DKA to Auckland City and North Shore Hospitals in 2013.

‍

Data are mean ± SD (± range) or median (95% CI of median). # = in those with known diabetes at presentation.

Insulin omission/error or non-adherence was the cause of DKA admission in 41% of people at Auckland Hospital and 27% at North Shore, infection in 20% and 37% and alcohol or drug excess in 10% and 13% respectively. A variety of miscellaneous causes of DKA was identified in the remaining people.

Treatment of DKA—Table 2 and Figures 1 and 2

Table 2:Treatment and outcome of people with type 1 diabetes admitted to Auckland and North Shore hospitals with DKA in 2013.

Data are mean ± SD (range) or median (95% CI). # after the initial admission baseline measurement.

Figure 1: Glucose, pH, bicarbonate and potassium concentrations of people with DKA admitted to Auckland City Hospital (n=41) and North Shore Hospital (n=30) and treated with an intravenous insulin infusion.  

Data are mean ± SD.

The number of bedside glucose measurements done in the first 24 hours was higher at the North Shore site but slightly more venous blood gases were done at the Auckland site. By protocol, very few ketone measurements beyond those obtained at baseline were done at Auckland Hospital. A total of 385 point of care tests for ketones were done in the 30 DKA admissions at North Shore Hospital. As seen in Figure 2, the concentration of ketones in people admitted to North Shore fell rapidly to the target of <0.5mmol/l.

Figure 2: Kaplan Meier graph showing time for ketone concentration to reach <0.5mmol/l in 30 episodes of DKA at North Shore hospital.

Data are mean ± 95%CI.

In some people, there was considerable delay in starting IV insulin after presenting to hospital at both sites, and especially at the Auckland site. Review of the admissions with the longest delays revealed a variety of reasons, including misjudging the severity of DKA in a person whose glucose was “only 18mmol/l”, not considering DKA in two people with newly presenting diabetes admitted with “gastroenteritis”, and delay in considering DKA in a person with known type 1 diabetes as “the patient looked well”. For some admissions there was no clear reason for the delay and this may have reflected the busyness of the emergency department at the time.

The volume of saline infused in the first four hours was similar between the two hospitals but more saline was infused in the subsequent 20 hours at the North Shore site. The amount of IV insulin infused was greater at North Shore in the 4–24-hour period, consistent with the protocol, but the duration of insulin infusion was very similar between the two sites. In the first 24 hours of admission, 20 of 41 people at Auckland had a protocol scale change (minor adjustment up or down) and three people had three or more scale changes. At North shore only three people had their initial infusion rate changed in the first 24 hours.

In the first 24 hours of admission after DKA, 25 of 41 people at Auckland and all 30 people at North Shore received a long acting subcutaneous injection of insulin (usually glargine). A similar amount of potassium was infused in the first 24 hours at both sites. The numbers of patients with documented hypokalaemia (<3.5mmol/l) was similar between the two sites. Although very few were <3mmol/l, the proportion of patients with a potassium at any time <3.5mmol/l, was relatively high at both sites suggesting the protocol for infusion of potassium needs updating. Two patients had a glucose <4mmol/l at any one time at Auckland and three at North shore. All were minor and easily corrected.

The time when 10% dextrose was started after admission and the duration of the infusion was no different between the sites. The volume and rate of infusion of 10% dextrose was however higher at North Shore by protocol. The rate of fall of glucose, and rise of pH and bicarbonate were nearly identical at the two sites (Figure 1).

The number of people with DKA seen as an inpatient by a specialist diabetologist was significantly higher at the Auckland site. This may account for the trend seen in weekend discharge rates between the hospitals as the inpatient review by the diabetes team only occurs during working hours Monday–Friday.

The LOS was similar at the two sites. One patient was readmitted to Auckland hospital within 48 hours of discharge (not with DKA) and there were five readmissions of four patients in days 3–28 at Auckland and four readmissions of four patients at North Shore, with a variety of medical illnesses (data not shown). No patients were readmitted with DKA within 28 days.

One patient died from overwhelming sepsis 40 hours after their admission at North Shore.

Discussion

Previous studies have shown that following protocols of care for complex disorders such as DKA improves outcome and many different protocols for management of DKA have been published.10–15 There is consensus that relatively low-dose IV insulin infusion, along with vigorous fluid resuscitation and close monitoring of patients is important. Recent studies from the UK6,16 have shown, however, that standard protocols for the management of DKA are in fact poorly followed. The recent UK Guidelines 4 have recommended that emphasis should shift away from glucose concentration-driven protocols to ketone and pH-driven considerations, using frequent bedside ketone and glucose testing to inform when insulin infusions can be safely changed to subcutaneous insulin. While the protocol may have merit, we are not aware of any large and robust randomised controlled trials examining if this has advantages for patient outcomes that matter—mortality and length of stay.

There were few overall differences in the outcome in the first 24 hours of people admitted with DKA between the two hospitals in our study. Both hospitals have considerable room for improvement in the time to start an insulin infusion. At Auckland Hospital, 40% of people did not receive a subcutaneous injection of long-acting insulin in the first 24 hours of admission despite the protocol encouraging its use. There were no patients on sodium–glucose co-transporter inhibitors at the time of this study to account for possible insidious presentations of DKA. The admission weight of people with DKA was not routinely measured (16 of 30 admissions North Shore, 18 of 41 admissions Auckland) and thus the dose of insulin infused per hour at North Shore was based only on a bedside estimate of weight. Both protocols saw a number of patients develop hypokalaemia but this was mild and easily corrected. There were very few episodes of hypoglycaemia at both sites, and all were mild. The protocol at North Shore hospital stipulates that the rate of infusion of 10% dextrose is 100ml/hr, but this was not followed in many patients (Table 2) for unknown reasons. This may have led to some confounding of results.

By protocol the North Shore patients had many bedside point-of-care measurements of ketones done in the first 24 hours (mean 13 per patient). There was no evidence this resulted in any advantage in the improvement in metabolic parameters in the first 24 hours between the two sites. At both sites patients had multiple measurements of glucose by finger prick performed (nearly every hour) as well as frequent venous samples for blood gases and potassium, with likely resultant disruption on their ability to rest or sleep. The additional measurement of ketones invariably adds to nursing time as well as incurring added expense (just over $2.00 per ketone strip—New Zealand Scientific and Medical supplier, personal communication).

Limitations of this study include its retrospective nature and the number of exclusions of admissions due to transfer of patient care from outlying hospitals, or ICU admission. The model of funding in New Zealand is that people where possible need to be cared for in the hospital nearest to their residence. A similar number of people were cared for in ICU but we excluded these admissions, as the ICUs have their own protocols for management of DKA.

Conclusion

This study has shown very similar rates of resolution of DKA in the first 24 hours after admission, despite the use of two different protocols, with little evidence of any added value in the frequent bedside measurement of ketones. Furthermore, there was no difference in the length of stay between the two hospitals.

‍