Article Text

Download PDFPDF

Identifying targets to reduce the incidence of diabetic ketoacidosis at diagnosis of type 1 diabetes in the UK
  1. K Lokulo-Sodipe1,
  2. R J Moon1,
  3. J A Edge2,
  4. J H Davies1
  1. 1Department of Paediatric Endocrinology, University Hospital Southampton, Southampton, UK
  2. 2Department of Paediatric Diabetes, Oxford Children's Hospital, Oxford, UK
  1. Correspondence to Dr Justin H Davies, Department of Paediatric Endocrinology, University Hospital Southampton, Tremona Road, Southampton, SO16 6YD, UK; Justin.Davies{at}


Background Diabetic ketoacidosis (DKA) is the leading cause of mortality in childhood diabetes, and at diagnosis might represent delayed presentation. The extent and reasons for delays are unclear, but identifying and targeting factors associated with DKA could reduce this incidence.

Objective To compare the patient pathway before diagnosis of type 1 diabetes mellitus (T1DM) in children presenting with DKA and non-acidotic hyperglycaemia.

Design, setting and patients Over a 3-month period, children newly diagnosed with T1DM were identified on admission to UK hospitals. Parents and medical teams completed a questionnaire about events before diagnosis.

Results Data were available for 261 children (54% male), median age 10.3y (range 0.8–16.6 y). 25% presented with DKA, but more commonly in children <2y (80% vs 23%, p<0.001). Fewer children with DKA reported polyuria (76% vs 86%) or polydipsia (86% vs 94%) (both p<0.05), but more reported fatigue (74% vs 52%) and weight loss (75% vs 54%) (both p<0.01). 24% of children had multiple healthcare professional (HCP) contacts, and these children had lower pH on admission. 46% of children with a delayed presentation to secondary care had non-urgent investigations. 64% of parents had considered a diagnosis of diabetes, and these children were less likely to present with DKA (13% vs 47%, p<0.001).

Conclusions Multiple HCP contacts increased risk of presentation in DKA, whereas, parental awareness of diabetes was protective. Improved public and health professional education targeting non-classical symptoms, awareness of diabetes in under 2 y, and point-of-care testing could reduce DKA at diagnosis of diabetes.

  • children
  • Diabetes
  • presentation
  • diagnosis
  • diabetic ketoacidosis

Statistics from

What is already known on this topic

  • The incidence of DKA at diagnosis of T1DM in the UK is 25% and is relatively unchanged from reports over the past 20 years.

  • Risk factors for presentation at diagnosis with DKA include younger age, ethnic minority background, delayed diagnosis, diagnostic error and delayed treatment.

  • Protective factors against DKA may include higher caregiver education level, family history of T1DM, higher background incidence of T1DM and higher gross domestic product.

What this study adds

  • Multiple healthcare professional contacts prior to diagnosis, lead to delayed referral, lower pH on admission and increased need for intravenous insulin.

  • Fewer children present with DKA when parents consider the diagnosis of diabetes, and this is more likely when polydipsia and/or polyuria are present.

  • Parents report symptoms of fatigue and weight loss more frequently in children presenting with DKA compared with hyperglycaemia alone.


Approximately 26 per 100 000 children are diagnosed with type 1 diabetes mellitus (T1DM) each year, although this figure continues to rise by approximately 4% per year.1 ,2 Despite improvements in diabetes care leading to an increasing life expectancy, the mortality rate for children with T1DM remains higher than the general population, and the leading cause of death is diabetic ketoacidosis (DKA).3 In the UK, approximately 25–30% of children will present with DKA.4–7 However, globally, there is a large variation in the incidence of DKA at presentation, ranging from 13% to 80%.8 Children presenting with DKA are at higher risk of life-threatening complications including cerebral oedema, cerebral ischaemia and hypoxic brain injury.9 They are more likely to require admission to an intensive care unit and contribute to greater healthcare spending.10

A recent systematic review including data from more than 30 countries demonstrated an increased risk of DKA at presentation of T1DM in younger children, ethnic minority groups, and those without private healthcare insurance.11 Furthermore, a country's gross domestic product and background incidence of T1DM are inversely associated with the percentage of children that present with DKA8 ,12 suggesting that healthcare accessibility and diabetes awareness are key contributing factors. This is also supported by a lower incidence of DKA at presentation in children who have a first-degree relative with T1DM.11

It is not clear whether DKA at diagnosis represents a delay in presentation or a more aggressive course of the disease,13 but the National Institute of Health and Care Excellence guidance published in 2004 recommended that children with suspected diabetes should be referred on the same day to the local paediatric diabetes team.14 Despite this, the incidence of DKA at diagnosis in the UK has remained virtually unchanged for 20 years,4–6 and is nearly twice as high as that observed in Sweden.8 It is possible that differences in healthcare provision prevent the UK achieving these lower rates. We undertook this study to obtain a contemporary estimate of DKA incidence at diagnosis of T1DM, and to develop a greater understanding of demographic and healthcare factors that are associated with a longer duration of symptoms and/or DKA at diagnosis in the UK. It is hoped that this information could contribute to the development of future interventions to improve the patient pathway to diagnosis of T1DM.


A national survey was carried out through the Regional Paediatric Diabetes Networks with the support of NHS Diabetes, the Association of Children's Diabetes Clinicians (ACDC), and the British Society for Paediatric Endocrinology and Diabetes (BSPED). Children presenting with T1DM in England, Wales and Northern Ireland between December 2011 and February 2012 were included. Data was collected by local paediatric diabetes teams at the time of diagnosis.


Following initial stabilisation, a questionnaire was completed by the child's primary caregiver during the hospital admission. Questions included their child's gender, date of birth, postcode, ethnic background, family history of diabetes, number of family members and caregiver education level. Information was also collected regarding the duration of non-specific parental concerns about their child, specific symptom duration, healthcare professional (HCP) contacts, advice given and investigations performed before referral to hospital.

Local paediatric diabetes teams completed the second part of the questionnaire, recording the investigations prior to referral, method of referral, and biochemical results on admission. Questionnaires were sent without identifying data to the researchers for analysis.

As the audit was entirely anonymous, advice was taken from Caldicott guardian that consent was not required and was implicit by the completion of the voluntary questionnaire.


DKA was defined as a blood glucose >11 mmol/L, ketonuria or ketonaemia, and venous pH <7.3 or bicarbonate <15 mmol/L.15 A delayed diagnosis was defined as >24 h from presentation to primary or secondary services to referral to the diabetes multidisciplinary team.14 ,16

Deprivation index scores were calculated using the participant's postcode and the English 2010 Index of Multiple Deprivation from the UK Office of National Statistics, which has been demonstrated to serve as a surrogate measure of poverty within a geographical area.17 For participants from Wales and Northern Ireland, postcodes were converted to deprivation indices using the Welsh Index of Multiple Deprivation 201118 and the Northern Ireland Multiple Deprivation Measure 2010,19 respectively, and converted to UK-equivalent values.

Statistical analysis

Data was checked for normality of distribution, and where possible, non-normally distributed data was log transformed. Continuous outcomes were compared using independent samples t test or Mann–Whitney U test for normal and non-normally distributed variables, respectively. One-way analysis of variance with posthoc Bonferroni correction was used to determine differences in outcome when three or more groups were compared. Correlations were examined using Pearson's correlation coefficient or Spearman's rank correlation coefficient. Categorical outcomes were analysed using χ2 test. Results are presented as mean±SD, unless otherwise stated. All analyses were performed using SPSS V.19. P<0.05 was considered statistically significant.


Characteristics of study participants

Questionnaires were returned for 261 children treated at 75 hospitals (hospital name provided in 233 questionnaires) (median 3 children per hospital (range 1–11)): 140 (54%) children were male, and the median age at diagnosis of T1DM was 10.3 years (range 0.8–16.6 years). Of the 252 children for whom biochemical data was available, 63 (25%) presented in DKA.

The median age of children presenting with DKA was similar to those presenting with non-acidotic hyperglycaemia (HG), although for children under the age of 2 years, significantly more presented with DKA (table 1); in this age group, 80% presented with DKA, compared with 23% of children older than 2 years.

Table 1

Characteristics of the children included in the survey, separated by presentation with either diabetic ketoacidosis (DKA) or non-acidotic hyperglycaemia (HG)

Children who presented with DKA were of similar sex, ethnicity and family structure, and to children who presented with HG (table 1). Social deprivation scores did not differ between the two groups (p>0.05 for all).

Symptom duration

Polyuria and/or polydipsia were reported for at least 1 day prior to diagnosis in 93% patients. In children with polyuria and/or polydipsia, the median durations of these symptoms were 14 days (range 1–182 days) and 14 days (1–152 days), respectively. Overall, fewer children reported enuresis (32%), fatigue (55%), and weight loss (57%). No associations were identified between duration of symptoms and social deprivation score, caregiver education level or the number of parents living in the household (p>0.05 for all). However, symptom duration was significantly shorter in children who had four or more siblings compared with those with no siblings (polydipsia 6±3 vs 18±3 days, p<0.01 and polyuria 7±3 vs 20±3 days, p<0.05).

Overall, children who presented in DKA were less likely to report polyuria and/or polydipsia, however, when only children older than 2 years were included, the number of children reporting these symptoms was similar (table 2). Parents of children presenting with DKA were more likely to report weight loss (p=0.004) and fatigue (p=0.002) than those with children who presented with HG (table 2). In all children, the duration of fatigue and weight loss, and the total length of parental concern were shorter in those presenting with DKA (table 2), despite a similar HbA1c at present (109±21 mmol/mmol (12.9±4.1%) vs 104±27 mmol/L (11.7±4.6%), p=0.32). HbA1c at diagnosis was moderately positively correlated with reported duration of polyuria, polydipsia, weight loss and parental concern in children who presented with HG, but displayed no significant correlations with reported symptom duration in those who presented with DKA (table 3).

Table 2

Frequency and duration of symptoms at presentation of type 1 diabetes mellitus in children presenting with diabetic ketoacidosis in comparison to non-acidotic hyperglycaemia

Table 3

Correlation coefficients for symptom duration with HbA1c in children with newly diagnosed T1DM, and according to mode of presentation with either diabetic ketoacidosis (DKA) or non-acidotic hyperglycaemia (HG)

Health professional contacts

Totally, 248 (95%) parents responded to the question detailing the number of HCP encounters related to their concerns, prior to hospital admission. Of these, 188 (76%) were admitted following the first HCP contact. Where there was delayed admission, the median delay was 5 days (range 1–152 days). Explanations for this included being given an alternative diagnosis (n=19; 33%), receiving advice to have blood or urine investigations (n=26; 46%) and referral to another HCP (n=2; 4%).

Children aged below 2 years were more likely to have multiple HCP contacts prior to admission than older children (69% vs 23%, p=0.007). All children under 2 years who had been seen on more than one occasion were initially given an alternative diagnosis, which was most commonly an infection. There were no differences identified in ethnicity, social deprivation score, or caregiver educational level between children who had single and multiple HCP contacts (data not shown), however, children from single-parent families were more likely to be admitted following first HCP contact (88% vs 73%, p=0.04).

There were 33% of children who presented in DKA and had been seen by multiple HCPs prior to admission, compared with 21% of those with HG, which was of borderline statistical significance (p=0.058). However, children who had multiple HCP contacts had a lower pH on admission (median 7.32 (IQR 7.15–7.43) vs median 7.36 (IQR 7.29–7.39), p=0.03) and a higher percentage were treated with intravenous insulin (39% vs 24%, p=0.04).

No child who presented with DKA had been advised to have routine investigations, whereas, 25% of those with HG did.

Parental consideration of T1DM as a possible diagnosis

Overall, 64% of parents reported having considered diabetes as the cause for their child's symptoms, and 80% of those parents raised their concerns with a HCP. If polyuria and/or polydipsia were present, more parents considered the diagnosis (68% vs 20%, p<0.001).

Parents of children who have a first-degree relative with diabetes were more likely to consider T1DM as a possible diagnosis (80% vs 62%, p=0.021), and although fewer presented with DKA, this did not reach statistical significance (16% vs 27%, p=0.13).

Of the children whose parents considered diabetes as a diagnosis in their child, fewer presented with DKA than those where the diagnosis was not considered (13% vs 47%, p<0.001). However, the incidence of DKA at presentation was no different whether or not the parent(s) discussed these concerns with a HCP (13% vs 24%, p=0.24).

Of the 26 children referred by their GP for routine blood or urine investigations, 21 (81%) of the parents had considered diabetes as a possible diagnosis, and 18 (86%) had raised these concerns with the HCP. None of these children presented with DKA.


This survey provides contemporary information regarding the patient journey prior to diagnosis of T1DM, and explored potential factors for presentation with DKA. We have identified a number of key findings: first, the incidence of DKA at presentation of T1DM remains unchanged from previous UK studies over the last 25 years,4–6 with one-quarter of children presenting in this way. Second, the majority of children do report classical symptoms of polyuria and/or polydipsia, although there is marked variation in the duration of these symptoms. However, fatigue, weight loss and enuresis are also common symptoms, and more frequently, short histories of these symptoms were reported in children who presented with DKA. Third, multiple HCP contacts and delayed presentation to secondary care were common and were associated with increased incidence of DKA at presentation. Many children had undergone unnecessary investigations prior to referral, and while none of these children progressed to DKA in this cohort, these delays could lead to greater morbidity and mortality. Finally, we identified a number of factors which were associated with presentation with DKA. Parental consideration of diabetes, and having a relative with diabetes, were protective, and highlight the need for education of the general public.

There are a number of limitations to this survey. We invited all centres in England, Wales and Northern Ireland to take part in the audit. Although this is the largest survey of its kind, only 42% of paediatric units contributed data to the survey. Therefore, we cannot be certain that the DKA rate would have been the same across the country, or that presentations may not have varied between regions. However, the incidence of DKA was remarkably similar (and alarmingly high) to that previously reported, so it did not appear that there was any reporting bias. Secondly, the data on duration of symptoms and HCP contacts is subject to recall bias and parental perception. It is possible that symptoms parents attributed to T1DM were caused by an alternative diagnosis, and this is particularly likely in a small number who reported symptoms for several months to years prior to diagnosis. Similarly, we did not seek to confirm the reported number of HCP contacts, or that the diagnosis of T1DM was definitely missed. Furthermore, we relied on the local diabetes team to provide accurate data on biochemical findings and management at initial presentation rather than obtaining copies of the case-notes. This approach was important to enable collection of countrywide data. Importantly however, as the data were collected prospectively over a three-month period, the results should not be influenced by temporal changes in healthcare-seeking behaviour, referral patterns or changes in management.

The pattern of presenting symptoms differed in children with DKA and HG, such that those with DKA were less likely to report classical symptoms, but more frequently reported the less specific symptoms of fatigue and weight loss. However, when children aged below 2 years at diagnosis were excluded from the analysis, the proportions of children in the two groups reporting polyuria and polydipsia were similar. It is well recognised that younger children are more likely to present with DKA,11 and difficulty in recognising these symptoms in this age group might contribute to greater morbidity at presentation. Reported symptom duration was shorter in children with DKA, consistent with reports of some,20–22 but not all23 previous studies. This could represent either a more aggressive disease process with more rapid decline in pancreatic insulin production and/or metabolic decompensation, or parental failure to recognise symptoms. The concept of a more aggressive disease is supported by studies which have demonstrated a shorter honeymoon period in children who present in DKA.24 However, in this study, we identified a number of factors which would support a theory of delayed parental recognition of symptoms. First, HbA1c at presentation was similar in children with DKA and HG, and while symptom duration displayed positive correlation with HbA1c in children with HG, this was not evident in those with DKA. Second, fewer parents of children with DKA reported considering diabetes as a possible diagnosis, although this could be due to a more rapid onset and less opportunity to explore potential causes themselves (eg, searching the internet or discussions with friends/family). Children who had four or more siblings had shorter symptom duration, possibly suggesting that experienced parents had greater ability to detect abnormal symptoms, although incidence of DKA at presentation did not differ. Third, similar to previous studies,11 we found that fewer children who have a first-degree relative with diabetes presented with DKA. Together, these findings suggest that some knowledge or experience of diabetes is protective and, therefore, highlight the need for education of the general public with regards to symptoms, signs and need for early presentation in suspected diabetes.

Previous studies have shown diagnostic error to be associated with increased risk of presentation with DKA25 ,26 and to occur frequently in younger children.26 Our findings support this as those who had multiple HCP contacts had lower blood pH on admission and were more likely to present in DKA. It has previously been suggested that increased medical consultations27 and delayed referral5 are associated with DKA, although this is not supported by all studies.12 ,28 We particularly found that children under 2 years of age were more likely to have multiple HCP contacts. DKA rates are highest in this age group, therefore, it is important to highlight that opportunities to make an earlier diagnosis might have been missed and could have prevented progression to DKA. Nearly half of all children who had a delay in admission of greater than 24 h had been advised to undergo routine blood and/or urine investigations, yet nearly 70% of the parents of these children had discussed a possible diagnosis of diabetes with the advising HCP. Although none of these children progressed to DKA, point-of-care blood sugar testing might have enabled an earlier diagnosis in these children.

Three intervention studies aiming to reduce the incidence of DKA at diagnosis have had varying success.7 ,29 ,30 In Italy, Vanelli et al identified that secondary enuresis and nocturia can be early and frequent symptoms of T1DM and, therefore, used this in a poster campaign targeted towards HCPs, teachers and parents. Primary care paediatricians were also provided with equipment for measuring capillary blood glucose and guidelines for the diagnosis of T1DM.29 In comparison with a neighbouring region, the incidence of DKA, duration of symptoms and HbA1c at diagnosis decreased.29 In our study, only a third of children reported enuresis compared with 89% in the Italian study after the campaign. Therefore, specifically targeting this symptom in the UK might prevent a large proportion of children with symptoms being missed. A similar campaign in Australia using HCP education sessions, posters, radio and newspaper adverts and provision of blood glucose testing equipment to HCPs also reported a significant decrease in DKA at diagnosis.30 By contrast, a Welsh campaign using posters in primary care surgeries and radio broadcasts, but without provision of equipment or education, failed to reduce DKA incidence.7 It is possible that the differences in healthcare systems contributed to the variable success of these campaigns; Italy has primary care paediatricians, and in Australia, patients do not have a named primary care practitioner. The results would also suggest that increasing awareness in HCP by significant amounts of education is important to the success of such campaigns.

In conclusion, this survey has identified a number of factors which are related to having DKA at diagnosis of T1DM: older age, a first-degree relative with diabetes, and parental consideration of diabetes as a diagnosis were all protective, whereas, lacking classical symptoms of polyuria and/or polydipsia, and multiple HCP contacts increased the risk. These might represent important targets to reduce the incidence of DKA at diagnosis in the UK, and clearly, future campaigns should include general public awareness and HCP education. In particular, educational programmes for HCPs should highlight the importance of non-specific symptoms (eg, weight loss and fatigue), in addition to the classical symptoms. Awareness that T1DM does occur in young children needs to be increased, as children under the age of 2 years were likely to have multiple HCP contacts and present in DKA. In this age group, polyuria and polydipsia are less likely to be recognised and, therefore, consideration needs to be given towards other symptoms. Furthermore, an emphasis needs to be placed on point-of-care blood glucose testing, as many children were referred for routine investigations, which further contribute to diagnostic delays.


The authors would like to thank NHS Diabetes, the Association of Children's Diabetes Clinicians, the British Society for Paediatric Endocrinology and Diabetes, and participating local diabetes teams across the UK for facilitating this study. All authors had full access to all the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis.


View Abstract


  • Contributors KLS performed the literature search, extracted the data, performed data analysis, and wrote the first draft of the manuscript. RJM performed statistical analysis, reviewed and edited the manuscript. JHD and JE developed and designed the survey, reviewed and edited the manuscript. JHD and JE are the guarantors of the work.

  • Funding RJM is supported by an NIHR academic clinical fellowship.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.