It was a prospective study. All patients having known or unknown diabetes hospitalized at JSS Medical College Hospital, Mysuru, Karnataka, India with a diagnosis of DKA (blood glucose more than 250 mg/dl, arterial pH less than 7.3, bicarbonate less than 15 mEq/L and moderate degree of ketonuria) during the period November 2007 to October 2009 were recruited for the study. Total 110 patients admitted with the above diagnosis were recruited and completely evaluated, in whom standard protocol was followed for treatment of DKA. Ethical clearance was given by the Ethical Committee of JSS Medical College, Mysuru. Informed consent was taken by the subjects and attendants in case of unconscious patients.

Pregnant women, patients with pre-existing chronic illnesses like Congestive Cardiac Failure (CCF), liver & renal impairment, type 2 respiratory failure and those not treated according to standard protocol were excluded.

Total WBC count, blood urea, serum creatinine, serum sodium, serum potassium, arterial blood pH, arterial blood bicarbonates (HCO₃), Aspartate Aminotransferase (AST), Alanine Aminotransferase (ALT), serum total bilirubin, other routine and relevant investigations were noted in search of precipitating factors.

Presence of DMS, defined as any level of sensorium other than fully conscious and oriented, pulse rate, blood pressure, body temperature, capillary blood glucose, units of regular insulin administered, litres of normal saline administered and urine output were noted.

A master chart was prepared using the following parameters like presenting symptoms and precipitating and mortality factors like age more or less than 65 years, presence or absence of fever (defined as temperature above 98.9°F in the morning and 99.9°F in the evening), DMS. (This included drowsy, delirious, obtunded, stuporous, semi-comatose and comatose patients). Shock (defined as blood pressure <90 mm of Hg systolic + pulse rate >100 beats per minute + urine output less than 0.5 ml/kg/hour), RBS (more or less than 300 mg/dL after 12 hours after treatment using standard protocol, blood pH more or less than 7.2, blood HCO₃ more or less than 15 mmol/L, blood Urea more or less than to 100 mg%, serum creatinine more or less than to 1.6 mg%, serum sodium more or less than to 150 meq/l, serum potassium more or less than to 3.5 meql/l, total WBC count more or less than to 15,000/mm³, Fluid requirement more or less than 6 litres in 24 hours, serum bilirubin more than or less than 1.5 mg%, ALT and AST more or less than 80 IU, insulin requirement more or less than 50 units in 12 hours to bring serum glucose to less than 300 mg%.

The data in the proforma of all the 110 patients enrolled in the study was entered into the EPINFO version 7.0.5 software. A p-value less than 0.05 was considered as statistically significant. Less than 0.01 was considered as very significant and less than 0.001 was considered as very highly significant.

Scoring methodology was used to identify the association between the number of statistically significant adverse variables present in an individual and the outcome. A score was defined as number of adverse parameters presenting in an individual in the first 24 hours. A cut off score (i.e., the number of adverse parameters identified) of 5 was utilized to study the association with death as outcome. A similar scoring system was followed by Esjtathiou SP et al., [11].

Multiple logistic regression analysis was done using all variables with p-value <0.5 after univariate analysis to find out the variables which are independent predictors of mortality in DKA.

The average age was 42.33 years, with 90 patients (81.81%) less than 65 years. There were 60 males (54.54%) with male: female ratio of 1.2: 1. Out of 110 patients with DKA, 10 patients died, with the mortality rate in the hospital being 9.09%. The most common presenting symptom was found to be vomiting seen in almost 50% (55 cases) of individuals, followed by generalized weakness and abdominal pain. The details of other symptoms are mentioned in [Table/Fig-1].

The most common precipitating factor in our study was found to be infections which accounted for 52% (57 cases). Other causes like stroke, binge alcohol and emotional disturbances were seen in 2.7% of cases. In 10% (11 cases) cases no cause could be recognized in spite of search for common precipitants through proper history, clinical examination and lab investigation. Details are in [Table/Fig-2].

Among the infections, Urinary Tract Infection (UTI) was seen in 27% of cases. The details of the rest of the infections are mentioned in [Table/Fig-3].

At the end of univariate analysis the following statistically significant variables were identified as probable predictors of mortality:

A scoring methodology was adopted to study the association of the statistically significant adverse parameters in a particular individual with DKA with mortality as mentioned in [Table/Fig-13].

There were eight patients out of 14 (57.14%) who had a score of more than 5. All of them died, which was statistically highly significant with a p-value of 0.001.

Multiple logistic regressions analysis was done using EPINFO software on statistically significant univariate factors to identify the independent factors associated with outcome. Results indicated that no factor stood independently associated with DKA mortality. This clearly indicates each factor has relation with other factors in explaining the variation in the outcome.

The average age of patients admitted with DKA is between 40 to 50 years [12-14] with male to female ratio of 1:1 to 1.4:1 [15-17]. In this study the average age was 42.33 years, with 81.81% less than 65 years and male: female ratio being 1.2:1.

The mortality rate in patients with DKA is less than 5% in experienced centers, 5%-15% in community hospitals and 30% of all intensive care cases [11,17]. In this study, the hospital mortality rate was 9.09%.

Usually DKA most commonly presents with nausea and vomiting, polyuria and polydypsia, abdominal pain and dizziness and altered sensorium [16-19].

In this study, the most common presenting symptom was found to be vomiting, seen in almost 50% of individuals, followed by generalized weakness, abdominal pain, symptoms of hyperglycaemia, fever, DMS and breathlessness.

Precipitating factors of DKA in the descending order of frequency includes infections (52%), skipping of anti diabetic treatment (21%), undiagnosed diabetes (11%). Among infections, urinary tract infection is the commonest (27%) followed by cellulitis and respiratory infections. In most of the studies, the precipitating factors are infection, skipping of medication, undiagnosed diabetes and comorbid conditions [14-17,20,21]. The earlier studies showing various precipitating factors in DKA are shown in [Table/Fig-14] [12,22-26].

Old age is an independent mortality predictor of DKA [17,19]. In this study 25% of patients with more than 65 years died with a p-value of 0.006 which is statistically very significant.

Fever in DKA is primarily due to the precipitating factor or due to elevated counter regulatory hormones [27]. Fever in the first 24 hours of onset of DKA is an independent risk factor for mortality in DKA [27]. In this study, 31.81% of the patients had fever, out of which 22.9% expired with a p-value of <0.002 which is statistically very significant.

DMS depends on extent of dehydration, presence of co morbid diseases; type of precipitating factor [27]. Presence of this variable at presentation or within first 24 hours is an independent risk factor for mortality in DKA [28,29]. In this study, 28% of the patients had DMS in the first 24 hours, out of which 26.5% expired with a p-value of 0.001, which is statistically very highly significant.

Shock in DKA is primarily hypovolemic resulting from severe osmotic diuresis as a result of glycosuria and also due to loss of water through hyper ventilation [30]. Our study showed that 30.9% of patients had shock in the first 24 hours, out of which 20.6% expired with a p-value of 0.01. Presence of shock was found to be a statistically significant predictor of mortality.

None of the earlier studies have shown shock as significant predictor of mortality in DKA, probably because in the earlier studies all co morbidities were included which might have contributed to death nullifying the effect of shock as a mortality predictor.

Glucose more than 300 mg/dL after 12 hour of standard protocol treatment is an independent risk factor for mortality [11].

In this study, patients with RBS more than 300 mg% at or after the first 12 hours were 32.7% of which 22.2% had resistant RBS and expired. A p-value was found to be 0.002 which is statistically very significant.

The degree of acidosis was assumed to be a predictor of mortality and was statistically proven as an independent mortality predictor [11]. In this study, patients with severe acidosis had increased mortality rates. 39.09% of the patient had pH less than 7.2, of which 18.6% expired with a p-value of 0.013, which is statistically significant.

In this study, 43.6% cases had HCO₃ less than 15 mmol/L, of which 16.7% expired. 3.2% of the patients with HCO₃ more than 15 mol/l expired. The p-value was found to be 0.02 and statistically just significant. This is similar to a study done by Trikia S et al., where mortality was high in patients who had a low bicarbonate level [17]. But in another study, the effect of HCO₃ was not significant [11].

Insulin requirement is dependent on presence of precipitating or stress factor and dehydration [29]. Mortality prediction studies showed that the administration of more than 50 units of regular insulin within first 12 hours is itself an independent predictor of mortality [11].

The insulin requirement in the first 24 hours was more than 50 units in 40% of patients, out of which 20.5% expired. A p-value was found to be 0.001 which is a very significant predictor of mortality.

Fluid requirement more than 6 litres in 24 hours is a significant predictor of mortality as per mortality prediction studies [11]. In this study, 31.8% of patients required fluids more than 6 litres in 24 hours, of which, 22.9% expired. The p-value was found to be 0.002 which is statistically a very significant predictor of mortality.

Other parameters at admissions like serum sodium, serum potassium, blood urea, serum creatinine, total WBC count, serum bilirubin, AST and ALT were shown to have no statistically significant values in predicting mortality in DKA which is being further supported by earlier studies [11].

It was observed that patients with adverse parameters more than five died in our study. Statistically very high significant association was observed with eight patients dying in this group with a p-value of 0.001. This is similar to the study done by Efstathiou SP et.al, where the mortality was high in cases where the independent predictors were more than 5 [11].

The main limitation of this study is the small number of patients recruited for the study. Similar type of studies can be conducted in a larger group, may be multicentric, which will help in the risk stratification of patients with DKA at the time of admission.

The most common presenting complaints in DKA are vomiting and generalized weakness. The most common precipitating factors are infection and non-compliance to anti diabetic treatment. On the first day of hospitalization risk stratification of patients with DKA is possible from simple clinical and laboratory variables. Patients with higher score of adverse parameters, especially more than 5 are more prone for mortality and such patients need to be monitored carefully to prevent such outcomes.