Prevalence of admission plasma glucose in 'diabetes' or 'at risk' ranges in hospital emergencies with no prior diagnosis of diabetes by gender, age and ethnicity

Abstract Aims To establish the prevalence of admission plasma glucose in 'diabetes' and 'at risk' ranges in emergency hospital admissions with no prior diagnosis of diabetes; characteristics of people with hyperglycaemia; and factors influencing glucose measurement. Methods Electronic patient records for 113 097 hospital admissions over 1 year from 2014 to 2015 included 43 201 emergencies with glucose available for 31 927 (74%) admissions, comprising 22 045 people. Data are presented for 18 965 people with no prior diagnosis of diabetes and glucose available on first attendance. Results Three quarters (14 214) were White Europeans aged 62 (43‐78) years, median (IQ range); 12% (2241) South Asians 46 (32‐64) years; 9% (1726) Unknown/Other ethnicities 43 (29‐61) years; and 4% (784) Afro‐Caribbeans 49 (33‐63) years, P < .001. Overall, 5% (1003) had glucose in the 'diabetes' range (≥11.1 mmol/L) higher at 8% (175) for South Asians; 16% (3042) were ‘at risk’ (7.8‐11.0 mmol/L), that is 17% (2379) White Europeans, 15% (338) South Asians, 14% (236) Unknown/Others and 11% (89) Afro‐Caribbeans, P < .001. The prevalence for South Asians aged <30 years was 2.1% and 5.2%, respectively, 2.6% and 8.6% for Afro‐Caribbeans <30 years, and 2.0% and 8.4% for White Europeans <40 years. Glucose increased with age and was more often in the 'diabetes' range for South Asians than White Europeans with South Asian men particularly affected. One third of all emergency admissions were for <24 hours with 58% of these having glucose measured compared to 82% with duration >24 hours. Conclusions Hyperglycaemia was evident in 21% of adults admitted as an emergency; various aspects related to follow‐up and initial testing, age and ethnicity need to be considered by professional bodies addressing undiagnosed diabetes in hospital admissions.


| INTRODUC TI ON
The current diabetes pandemic threatens both the health and economy of nations. 1 The prevalence is increasing year by year adding markedly to the cost of health care funded by governments or private healthcare organizations. Diabetes currently consumes over 10% of the UK National Health Service (NHS) budget. 2 The prevalence of type 2 diabetes and its complications vary by ethnicity with type 2 diabetes more prevalent in people of South Asian descent (six times) and in Africans and Afro-Caribbeans (three times) than White Europeans. 3,4 Symptoms of diabetes are not always evident until people consult a family doctor or are admitted to hospital. 5 Undiagnosed diabetes results in an eightfold increase in mortality for hospital admissions compared to those with normal glucose. 6 Admission glucose is strongly associated with mortality in acutely ill medical patients 7 with hyper/hypoglycaemia independent predictors of in-hospital mortality in patients not previously diagnosed with diabetes. 8,9 The American Diabetes Association guidance published in 2020 recommends measuring HbA1c in patients admitted with hyperglycaemia, defined as glucose >7.8 mmol/L. 6 Although HbA1c is used for diagnosis in the community in the UK following WHO guidance in 2011, 10 it is not currently requested routinely on hospital admission for this purpose. Data from an Irish hospital indicates that HbA1c could be used for follow-up testing 11 although it is not suitable for people with some haemoglobinopathies or altered red blood cell turnover. 10 This clinical audit reports on admission plasma glucose in emergency admissions with no prior diabetes coding using laboratory and demographic data from hospital electronic patient records over 1 year.
It aims to investigate people with glucose recorded on admission and no prior diabetes diagnosis, to categorize plasma glucose by 'at risk' and 'diabetes' ranges, and describe relationships to age, gender and ethnicity.

| Design
The clinical audit at Queen Elizabeth Hospital Birmingham was approved according to clinical governance (CARMS-12031), University Hospitals Birmingham NHS Foundation Trust. This hospital, located in the West Midlands of England with a culturally diverse catchment, is a major trauma centre for adults comprising >1200 beds with 100 critical care beds.

| Study data
Data were obtained from the electronic patient record system PICS (patient information and communication system) with the initial plasma glucose result measured by point-of-care testing or in the laboratory.

| Measurements
Blood glucose was measured at point-of-care on the wards in random capillary blood on glucose meters or arterial/venous whole blood samples on gas machines with results reported as plasma; Roche Cobas Inform II glucose meters (CV 5%) and Roche Cobas b221 blood gas machines (CV 4%). Blood was collected into fluoride oxalate vacutainers for measurement of plasma glucose in the central hospital laboratory using Roche Cobas 8000 analysers (CV 2%).
Internal quality control and external quality assurance were overseen by the hospital blood sciences laboratory. The performance of the point-of-care equipment was compared with the laboratory analysers and found to be acceptable. Power calculations were performed using sample sizes for people in ethnic groups in Table 2 with a significance level of .05/6 used to adjust for multiple comparisons. Due to unequal sample size, the power to detect a difference depended on whether the lower proportion related to the smaller or larger group.
F I G U R E 1 Flow chart for emergency admissions to a UK hospital located in a multi-ethnic region over 1 y: 2014-2015

| Availability of glucose
One third of all emergency admissions stayed in hospital for <24 hours and two thirds for >24 hours; 58% of emergency admissions with a duration <24 hours had glucose available and 82% of emergency admissions with duration >24 hours (Table 1). Admissions with plasma glucose reported (31 927) were 2 years older than the total cohort (43 201), median, IQ range for age, 62 (43-77) vs 60 (41-77) years, P < .001 (Table 1), TA B L E 1 Admission plasma glucose in emergency admissions over 1 y from 2014 to 2015

| Timing of admission
Glucose was more likely to be measured on admission to Queen Elizabeth Hospital Birmingham at the weekend (Saturday/Sunday), 76% vs 73%, P < .001, or during the week between 6 pm and 6 am, 76% vs 71%, P < .001.

| Glucose and glycaemic status on admission
Glucose was 6.4 (5.4-8.0) mmol/L in 74% (31 927) of admissions with glucose available (Table 1) Over 20% of the people admitted as an emergency had hyperglycaemia; 5% had glucose in the 'diabetes' range and 16% in the 'at risk' range, Table 2 and Figure 2 with a higher proportion of South Asians (8%) than White Europeans (5%) in the 'diabetes' range, P < .001. The proportion of White Europeans (17%) and South Asians (15%) in the 'at risk' range was higher than for Afro-Caribbeans, (11%) P < .001 and P = .010. Some guidance specifies age limits below which people should not be tested for undiagnosed diabetes. 15

| Ethnicity of people admitted
Three quarters of people admitted without prior diabetes coding with glucose available were White European, median age 62 years (

| Admission glucose by age, gender and ethnicity
Glucose was higher as the age of the people admitted increased ( Figure 3B,C), and varied by ethnicity. South Asian men aged >21 years and women aged >37 years had higher glucose than the White Europeans. In terms of overall plasma glucose levels, White Europeans and South Asians had slightly higher median glucose on admission at 6.2 mmol/L than Afro-Caribbeans, 6.0 mmol/L, and Unknown/Others 5.9 mmmol/L, P < .001. There were significant differences in glucose over the age distribution depending on people's gender and ethnicity (Table 3). Increases in glucose with age were greater for South Asian men and women than for White Europeans of the same gender; also for men whose ethnicity was Unknown/ Other and Afro-Caribbean women (Table 3).
When relating ethnicity to glucose ranges, South Asian men were more prevalent than expected in the 'diabetes' range, >11.0 mmol/L, and South Asian women in the lowest range, <5.0 mmol/L ( Figure 4A).
A similar analysis with age showed that South Asian women aged <30 years were most prevalent in the lowest glucose range, with South Asian men and women aged >70 years more prevalent in the 'diabetes' and 'at risk' ranges ( Figure 4C). White European men aged 50-69 years old were prevalent in the 'diabetes', 'at risk' and 'prediabetes' ranges ( Figure 4B).

| Diabetes diagnosed during hospital admission
Diabetes was diagnosed following routine protocol during the index admission in people without a prior diagnosis before

The prevalence of diabetes recorded in people in hospital in
Birmingham is nearly double that of the community at 22% vs 12%. 16 Early diagnosis of diabetes is important as people can be advised to alter their diet, exercise regimen and lifestyle, or blood glucose lowering treatment can be introduced when necessary.
Hyperglycaemia on admission to hospital is defined as 'at risk' of diabetes, that is 7.9-11 mmol/L or in the 'diabetes' range, that is >11 mmol/L. Immediate action is required on the ward when people present with very high glucose, for example 25/30 mmol/L 15 to prevent/diagnose life-threatening conditions such as diabetic ketoacidosis.
Undiagnosed diabetes may be the cause of hyperglycaemia on admission but its diagnosis should be confirmed by additional testing with HbA1c. In this audit, 5% of White Europeans and 8% of South Asians and Afro-Caribbeans had glucose in the 'diabetes' range but there is little evidence on how many cases of diabetes would be confirmed on HbA1c testing. People in this study below the age limits specified in some guidance on additional testing had glucose in the abnormal ranges. On emergency admission, 94% of those admitted to this hospital with a previous diabetes diagnosis had glucose recorded, but the figure for those not previously diagnosed was 74%. How this performance compares with other UK hospitals could be assessed by national inpatient audit programs. This figure may be related to the length of stay in hospital. One third of all emergency admissions were for <24 hours with 58% of these having glucose measured compared to 82% with a duration of >24 hours (Table 1). Those without glucose available were younger and more likely to be White European. The time/day of admission did not markedly influence the availability with only small differences observed possibly reflecting the hospital organization.
The number of people in the 'at risk' range was much higher than those in the 'diabetes' range involving 16% of people admitted with glucose measured. However, this audit is limited by the length of time the various ethnic groups have resided in the West Midlands. 17 As South Asians and Afro-Caribbeans present with diabetes at a younger age than White Europeans, it is vital to consider the age cut-offs for further testing quoted in some cases as 30 years for the former groups and 40 years for latter. 15 In routine practice, we have identified South Asian males presenting in their 20s with very high glucose and diabetic ketoacidosis-the reason why this audit was generated.
As there were fewer Afro-Caribbeans in the audit, the power to

F I G U R E 4
Glucose levels in people admitted to hospital as an emergency without prior diabetes diagnosis by gender, ethnicity and age. M, male; F, female. Blue-frequency observed significantly more than expected with positivity increasing with depth of colour; redless frequently; grey-nonsignificant residuals. Boxes-areas proportional to difference in observed and expected frequencies; dashed grey baseline is expected count; above baseline greater than expected frequencies; below fewer than expected frequencies. Height is proportional to contribution of Pearson's residuals; width is proportional to square root of expected counts detect differences between proportions of 5% and 8% in this group was lower at 54%-78% than for the other ethnic groups, 87%-100%.
The data presented here suggest that there should be no lower limit for follow-up testing in adults ( Figure 2). This is important as diagnosis impacts on both disease progression and the risk of developing complications which are apparent several years before diagnosis. This evidence will help guideline writers to assess the workload

| CON CLUS IONS
A significant number of people admitted as an emergency but not previously diagnosed with diabetes had hyperglycaemia within the 'diabetes' (5%) and 'at-risk' ranges (16%) (Figure 2). South Asians were admitted at a younger age than White Caucasians with their admission glucose higher and South Asian men particularly affected ( Figure 4A).
This audit highlighted various issues regarding the availability of glucose on admission (75%), readmission rate as hyperglycaemia increased with the number of admissions, whether age limits should be employed for additional HbA1c testing to confirm diagnosis as people below limits specified had admission glucose in the abnormal ranges, and the effect of the length of time the various ethnic groups have resided in the West Midlands. Further investigation into the efficacy, procedures and cost of diagnosis in emergency admissions is required-this will involve reflex HbA1c testing and algorithms linking hospital and primary care. Liaison between public health, diabetes organizations and researchers is required to address these issues.

ACK N OWLED G EM ENTS
Laboratory measurements were produced by the biomedical scientists in Clinical Laboratory Sciences at Queen Elizabeth Hospital Birmingham. We would like to thank G. Gill (University Hospitals Birmingham NHS Foundation Trust) who performed the audit and R.
A. Round (University Hospitals Birmingham NHS Foundation Trust) for their assistance with the manuscript.

CO N FLI C T O F I NTE R E S T
There are no conflicts of interest for the authors. The study sponsor, University Hospitals Birmingham NHS Foundation Trust, was not involved in the design of the study; the collection, analysis and interpretation of data; writing the report; or the decision to submit the report for publication.

All authors qualify for authorship based on the International
Committee of Medical Journal Editors criteria. All authors take full responsibility for content of manuscript. SG designed and organized the clinical audit and reviewed data and manuscript. SEM was responsible for data analysis and writing the manuscript. I.A-P. created the database. PGN and JAW analysed data and edited the manuscript. IMS and GVG reviewed the data analyses and the manuscript. RS contributed to data analysis and interpretation, and writing the manuscript. JW reviewed clinical aspects and also the manuscript. SDL contributed to data analysis and interpretation, and reviewed the manuscript. GAR and WH contributed to the overall audit process and reviewed clinical aspects of the paper. SG is the guarantor of this work.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets generated during and/or analysed during the study are not publicly available. The data set contains clinical data which cannot be shared publicly due to UK data protection legislation.