The clinical profile and associated mortality in people with and without diabetes with Coronavirus disease 2019 on admission to acute hospital services

Abstract Introduction To assess if in adults with COVID‐19, whether those with diabetes and complications (DM+C) present with a more severe clinical profile and if that relates to increased mortality, compared to those with diabetes with no complications (DM‐NC) and those without diabetes. Methods Service‐level data was used from 996 adults with laboratory confirmed COVID‐19 who presented to the Queen Elizabeth Hospital Birmingham, UK, from March to June 2020. All individuals were categorized into DM+C, DM‐NC, and non‐diabetes groups. Physiological and laboratory measurements in the first 5 days after admission were collated and compared among groups. Cox proportional hazards regression models were used to evaluate associations between diabetes status and the risk of mortality. Results Among the 996 individuals, 104 (10.4%) were DM+C, 295 (29.6%) DM‐NC and 597 (59.9%) non‐diabetes. There were 309 (31.0%) in‐hospital deaths documented, 40 (4.0% of total cohort) were DM+C, 99 (9.9%) DM‐NC and 170 (17.0%) non‐diabetes. Individuals with DM+C were more likely to present with high anion gap/metabolic acidosis, features of renal impairment, and low albumin/lymphocyte count than those with DM‐NC or those without diabetes. There was no significant difference in mortality rates among the groups: compared to individuals without diabetes, the adjusted HRs were 1.39 (95% CI 0.95–2.03, p = 0.093) and 1.18 (95% CI 0.90–1.54, p = 0.226) in DM+C and DM‐C, respectively. Conclusions Those with COVID‐19 and DM+C presented with a more severe clinical and biochemical profile, but this did not associate with increased mortality in this study.


| INTRODUC TI ON
Coronavirus disease 2019 (COVID-19) has impacted on morbidity and mortality of people across the globe. People with diabetes mellitus (DM) represents one group particularly adversely affected. 1 Recent studies have demonstrated people with DM have higher risks of more severe COVID-19 outcomes, including higher mortality, as well as the presence of other co-morbidities including cardiovascular disease (CVD) and chronic kidney disease (CKD). 2,3 Recent focus has been on in-hospital morbidity and mortality.
Within UK hospital settings, DM represents a significant proportion of all COVID-19 cases and deaths. 3,4 In-hospital mortality risks are increased in both type 1 and type 2 DM, however the former possesses the highest risk. 4 As DM represents a highly prevalent and heterogeneous population, there is a need to identify additional DM sub-groups who may be at higher risk of severe COVID-19 related outcomes. One area for examination is the sub-groups of DM with complications (DM+C) and DM with no complications (DM-NC).
Secondly, whilst recognized COVID-19 risk factors, including the presence of co-morbidities, provide useful information on those at higher risk of acquiring COVID-19, these risk factors do not account for initial clinical status or severity on initial admission to hospitals. 2 A novel approach could be to examine service level hospital data, including blood tests and other measures, which clinicians use for initial patient assessment and progress. The results may aid strategies for early clinical risk stratification for treatment or alternatively, guide prevention strategies (eg vaccinations).
The aim of this study is to examine whether DM+C patients with COVID-19 present with more adverse clinical and biochemical profiles and increased mortality compared to people with COVID-19 and DM-NC or without DM in an extensively phenotype adult cohort presenting to a large urban UK hospital.

| Study setting
This retrospective cohort study, using prospectively collected data, was conducted in Queen Elizabeth Hospital Birmingham (QEHB), a large teaching hospital within University Hospitals of Birmingham (UHB). It is situated in the city of Birmingham, West Midlands, UK and has over 1200 beds. West Midlands is a multi-ethnic region with 30% of residents classified as Black, Asian and Minority Ethnic (BAME), of whom South Asians (18.9%) and Blacks (6.0%) are the most prevalent minorities. 5

| Data sources
We constructed the data using the Patient Administration Database (PAS) and the Electronic Medical Record system, known as the Patient Information and Communication System (PICS). The PAS database record information on age, gender, ethnicity, address (post code), primary reason for admission, discharge diagnostic codes, inpatient death, and discharge destination. Admission is defined as the time spent by an individual from recorded time of entry to recorded time of exit from the hospital. The PAS database was linked using unique patient identifiers (hospital number) to the PICS. It is a purpose-designed system which records all in-hospital prescriptions, laboratory results and electronic observations and generates alerts to reduce prescription errors and notify abnormal blood results. 7 The linked PAS-PICS databases have been used for multiple diabetes related research. 8-10

| Study population
All adult patients (>16 years old) who presented to QEHB from 20th of March to 9th of June 2020 with a confirmed positive swab specimen result for COVID-19 were included in the analysis.

| Data collection and variable definitions
Patient demographics and clinical data were collected from PAS and PICS. Clinician confirmed co-morbidities were available from PAS and PICS, complemented by in-hospital prescription data and diagnostic codes derived from previous hospital admissions. The PICS encodes diagnoses using NHS Digital SNOMED CT browser alongside and mapped on to ICD-10 codes allowing for the presentation and inclusion of historically entered ICD-10 codes. The composite CVD, hypertension, severe renal diseases, dementia, chronic obstructive pulmonary disease, cancer, asthma, atrial fibrillation, were defined by the combination of ICD-10 codes and PAS-PICS encoded diagnoses, whichever was available at study entry. The composite CVD was defined as one of the following presentations: myocardial infarction, peripheral vascular disease, heart failure, cerebrovascular infarction, stroke, transient ischaemic attack and ischaemic heart disease.

| Clinical assessments
All patients underwent nasopharyngeal and oropharyngeal swab specimen (miniature absorbent pads) testing for COVID-19. These were processed in accordance with NHS guidance within UHB NHS laboratories. 11 The swab specimens were measured for COVID-19 using either real-time reverse transcription polymerase chain reaction or transcription mediated amplification methods on one of three assays: Abbott M2000, Cepheid GeneXpert or a Hologic Panther. Co-efficient of variation values were based on calibrations and therefore varied between individual runs.
Venepuncture was conducted to ascertain venous blood for routine metabolic blood tests; the first blood tests were taken before administration of any intravenous fluids. An arterial blood gas was performed to assess for acid-base status and estimated partial pressure of oxygen and carbon dioxide levels. Physiological assessments included measurement of respiratory rate and pulse rate via a pulse auxometer, systolic and diastolic blood pressures and temperature.
All swab specimens, blood tests and physiological assessments were performed by trained healthcare professionals following standard operating procedures.
All physiological and laboratory measurements were categorized based on clinically meaningful thresholds and the earliest available measurement was used in the analysis. Missing data were presented as a missing category for all measurements.

| Definitions of DM, DM+C and glycaemic categories for analysis
DM was defined as those who with an ICD-10 record of DM or its complications, or who were recorded to have been prescribed any of the DM drugs using a previously published algorithm. 9 DM com-

| Covariates
Ethnicity was self-reported by the patient or their family members on admission to hospital. Body mass index (BMI) was categorized based on the World Health Organisation Criteria: normal weight (BMI of < 25 kg/m 2 ), overweight (BMI of 25 kg/m 2 to < 30 kg/m 2 ), obesity (BMI of 30 kg/m 2 to < 35 kg/m 2 ), obesity II & III (BMI of ≥35 kg/m 2 ). Charlson Comorbidity Index (CCI) was calculated (DM and DM complication score were removed from the equation) using ICD-10 code and was categorized into four groups (0, 1, 2, and ≥3). 12 In the non-diabetes, DM-NC and DM+C groups, we looked at the trends of available physiological and laboratory measurements in the first 5 days after admission. These included measures of metabolic acidosis and compensation (anion gap, partial pressure of carbon dioxide (pCO2), bicarbonate (HCO3-) and hydrogen ions), indicators of underlying presence of inflammation (serum C-reactive protein, CRP), measures of immune response (lymphocyte count), serum electrolytes and renal function (Na+, K+, urea, estimated glomerular filtration rate, eGFR) and other clinically useful physiological and laboratory measurements (partial pressure of oxygen (pO 2 ), heart rate, temperature and serum albumin). All these measurement across three groups are presented visually as mean (standard error, SE) or median (IQR) for symmetrical and skewed continuous variables, respectively.

| Follow-up and outcome
All eligible patients were followed-up from hospital admission until the earliest of any censoring event (patient discharged, death, study end date) in hours. A small proportion (3.2%) of patients were not discharged at study end-date.

| Statistical analyses
Baseline characteristics for the total population and DM sub-groups are presented as mean ± standard deviation (standard deviation, SD) or median (interquartile range, IQR) for symmetrical/ skewed continuous variables and as frequency (percentage) for categorical variables. All physiological and laboratory measurements were compared across non-diabetes, DM-NC, and DM+C groups using ANOVA or Kruskal-Wallis test depending on data distribution. For categorical variable comparisons, Chi-square test was applied.

| Regression analysis
Cox proportional hazard regression models were used to calculate crude and adjusted hazard ratios (aHRs), together with their corre-

| Physiological and laboratory measurements at presentation and in the first 5 days
Individuals with DM+C were more likely to present with a pH level < 7.3 and a higher anion gap than in those with DM-NC or those without DM, p = .001 and p < .001 respectively ( Table 2).
Features of renal impairment (high urea, raised K+ and lower eGFR) were more common at presentation in the DM+C group than in the DM-NC group or those without DM, which could be related to underlying CKD, dehydration or acute kidney injury associated with acute COVID-19. In particular, eGFR < 30ml/min/1.73m 2 was more common in people with DM+C (54.8%) than those with DM-NC (18.6%) and those without DM (12.9%), p < .001. People with DM+C also had lower serum albumin and lymphocyte count.
People with DM+C had lower levels of serum CRP, heart rate or temperature compared to people with DM-NC and those without DM.
Where measurements were available, these observations largely persisted in the first 5 days after admission ( Figures 1&2).

| DISCUSS ION
Our study shows that people admitted with symptomatic COVID-19 and DM were more likely to be men, from a BAME background, and had higher BMI and more CVD, and more ESRD compared to those without DM. In addition, patients with DM+C had higher BMI, CVD and more ESRD compared to DM-NC, as would have been expected. Patients with DM+C had higher anion gap, urea, potassium, and lower pH, lymphocytes, albumin, compared to DM-NC. In addition, the DM+C group had lower heart rate, higher BP, less tachypnoea, lower Hb compared to patients with DM-NC. Patients in the DM+C group had a 39% higher mortality rates than people without DM or with DM-NC, but this did not reach significance. Other predictors of higher mortality included age, higher CCI, men and BAME groups.
The relatively higher mortality observed in people with DM compared to those without DM in this study is consistent with that reported previously in other COVID-19 studies and other studies showing higher mortality in patients with DM in relation to influenza, SARS and MERS. 13,14 This increases risk of adverse outcomes from these viral infections is likely due to multiple mechanisms including impaired immune response within a hyperglycaemic environment and reduced cellular expression of angiotensin-converting enzyme (ACE) 2, leaving cells prone to damage through inflammation. [14][15][16] The mortality risk was non-significantly greater in people with DM+C than DM-NC or those without DM. This in part could be due to differences in BMI, ethnicity, CCI, CVD and ESRD, which have been reported previously to be associated with increased risk of adverse COVID-19 outcomes. 2 . Diabetes autonomic neuropathy (DAN), which is common in people with DM, especially in the presence of complications, might also contribute to the increased mortality considering the established associations between DAN, CVD, CKD and mortality in DM. [17][18][19] This is supported by the results showing differences in heart rate and respiratory rate be- DM+C is usually associated with longer duration of DM and more adverse glycaemic control, which in turn may impact on pathological mechanisms affecting the response to COVID-19. Furthermore, the hyperglycaemic complication itself could leave the body more prone to a more adverse outcome where it impacts on ACE-2 cell expression. A recent study found worse glycaemic control was a risk factor for increased mortality. 3 The results of this study need to be considered in the context of its limitations. This study was conducted from a single centre which might affect the external validity of the findings. However, this single centre is a large, tertiary and receives patients form a large population beyond its localities, especially during the COVID-19 crisis. The sample size of our study is relatively small, as was the follow-up period of three months, which was reflected in the some of the 95% CIs reported in the study, and is reflected by the 39% increased risk of mortality in COVID-19 DM+C patients compared to non-diabetes not reaching significance. For the same Abbreviations: CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; HCO3-, bicarbonate; pCO2, partial pressure of carbon dioxide.

TA B L E 2 (Continued)
duration were not available in our analysis, although those with complications are likely to have higher HbA1c and longer diabetes duration.
The strengths of this study include the in-depth phenotyping which was made possible with the presence of the appropriate data management systems in the hospital trust. Also, cases of COVID-19 were laboratory confirmed. In addition, our study population included multiple ethnicities and in proportions mirroring those of the West Midlands county. 5 Finally, a recent mortality risk score has been developed for the general population, which utilizes total number of co-morbidities as one parameter for scoring, without considering presence/ F I G U R E 1 Measures of metabolic acidosis, I inflammation and immune response after hospital admission. Data presented for mean or median values over time. Key: HCO 3 , bicarbonate level; pCO 2 , partial pressure of carbon dioxide; K, potassium level; CRP, C-reactive protein level; Na, sodium level; Ur, urea level; pO 2 , partial pressure of oxygen absence of diabetes separately. 20 During development of the risk score, biochemical test variables assessed for inclusion did not include measurements of acid-base status including pH, bicarbonate or anion gap. It is not known if this could add further benefits to a risk score.

| CON CLUS ION
In this multi-ethnic cohort of adults with COVID-19 presenting to hospitals, we found clinical and biochemical profiles were adverse in people with DM+C.
F I G U R E 2 Renal function, electrolytes, and physiological and laboratory measurements after hospital admission. Data presented for mean or median values over time. Key: HCO 3 , bicarbonate level; pCO 2 , partial pressure of carbon dioxide; K, potassium level; CRP, C-reactive protein level; Na, sodium level; Ur, urea level; pO 2 , partial pressure of oxygen