Criteria abstracted from The Users' Guides to the Medical Literature series in JAMA
Diabetic ketoacidosis: Predictors of outcome in a pediatric intensive care unit of a developing country.
Jayashree M, Singhi S.
Pediatr Crit Care Med. 2004 Sep;5(5):427-33..[abstract]Reviewed By: Paul Bauer MD, Medical College of Wisconsin
Review posted September 11, 2008
I. What is being studied?
The study objective:
To study the outcome and predictors of mortality in children with diabetic ketoacidosis in a developing country.
The study design:
Retrospective case series.
II. Are the results in the study valid?
Was a representative group of patients completely followed up?
Yes. The charts of sixty-eight consecutive patients admitted with DKA to a PICU in India were studied retrospectively for physiologic and laboratory parameters associated with mortality.
Was follow-up sufficiently long and complete?
Yes and no. There was no reason to look beyond hospital stay for parameters associated with hospital mortality. However, it is not clear that the entire hospital course for each patient was reviewed, beyond the events recorded in the Pediatric ICU. It is not specifically stated, but it may be considered unlikely for a patient to survive the PICU only to die in the hospital; this was not identified in this study.
Were all potential predictors included?
The authors were limited by the data available from clinical care. A number of characteristics were compared between the group that survived and the group that did not. The characteristics were compared using a Mann-Whitney U Test. They included age, duration of polyuria, blood glucose at presentation and at 6-12 hours, pH, serum sodium, serum potassium, serum osmolality, and administration of sodium bicarbonate. As complications of the primary process and its treatment go, the authors also examined the sorting of patients who had cerebral edema, septic shock, pulmonary edema, hypokalemia, and hypoglycemia into the column of survivors and no survivors. When looking at predictors of mortality among patient characteristics, it may have been instructive to use the anion gap as a measure of acidosis because it considers the anion space better than the isolated consideration of sodium, potassium, and bicarbonate. This could have been calculated from the existing electrolyte data for each patient. To that effect, it may also have been helpful to measure the serum albumin in order to more accurately assess the anion space (1) . However, if that was not part of the routine protocol for assessment of patients in DKA in the study ICU, this would not have been possible. Finally, serum lactate, if it was measured, could have been used as a marker of regional hypoperfusion and acidosis.
Did the investigators test the independent contribution of each predictor variable?
Yes. Survivors and nonsurvivors were compared using univariate analysis to identify variables having a significant association with mortality. Also, multiple logistic regression analysis was used on all variables, identifying high osmolality on admission to have the most significant association with mortality.
Were outcome variables clearly and objectively defined?
Yes. Mortality was the chief outcome variable. Nine (13.2%) patient s died , the causes of death listed as septic shock (n = 4), cerebral edema (n = 2), cerebral edema with pulmonary edema (n = 2), and hypokalemia with ventricular tachycardia (n = 1) . The other outcome was cerebral edema. Nine (13.2%) patients developed cerebral edema without remarkable differences in glucose concentrations, age, pH, serum osmolality compared to those who did not have cerebral edema. The single difference between those with cerebral edema and those without was a greater rate of correction in serum sodium to 140 mEq/L
III. What are the results?
What is (are) the prediction tool(s)?
The results of their univariate analysis pointed to an association between mortality and: a) higher serum osmolality ( 315 mOsm/L (269â€“351) vs. 303 mOsm/L (265â€“369 ) , b) severe acidosis at admission ( 6.98 (6.80â€“7.30) vs. 7.07 (6.80â€“7.38) , and persistent hyperglycemia at 6-12 hours after admission ( 334 mg/dL (152â€“1600) vs. 253 mg/dL (18â€“600 ) and acidosis at 6â€“12 hrs ( 7.15 (6.80â€“7.25) vs. 7.30 (7.0â€“7.50 ) after initiation of the DKA protocol. They also found that cerebral edema and complicating sepsis were associated with high mortality. Four (44%) of the those with cerebral edema died and 4 (100%) of those with sepsis died.
On multiple regression analysis, the authors report that only serum osmolality on presentation was independently predictive of mortality.
A large number of the patients studied received bicarbonate .(Their protocol followed indicated the administration of sodium bicarbonate for serum pH less than 7.2.) In this study there was no greater burden of mortality associated with administration of sodium bicarbonate. A study conducted by Marcin et al in 2002 (2) , shows no direct association between cerebral edema and the administration of bicarbonate to correct acidosis but only 2 patients out of 61 (3%) were submitted to it, so the power to identify an association was low. In the current study 51 out of 68 patients (75%) received it and no association was found with its use and cerebral edema. It had been administered equally to those with cerebral edema (seven of nine patients) and to those without cerebral edema (forty-four of fifty-nine patients). This contrasts findings by Glaser et al in 2001 (3), a retrospective multicenter study over a 15 year period that found an independent association with cerebral edema and administration of bicarbonate; where 23 of 61 patients with cerebral edema (38%) had been treated with sodium bicarbonate as opposed to 43 of 174 matched controls (25%).
How well does the model categorize patients into different levels of risk?
There is the problem of cutoffs for risk stratification because of the broad range of values reported for serum osmolality on admission. Among survivors, the range for serum osmolality on admission was 265 to 369 mOsm/L. For nonsurvivors, the range was from 269-351 mOsm/L. Another way of stating this is that there were survivors who had a higher serum osmolality than the worst serum osmolality among nonsurvivors. This makes it difficult to categorize patients in risk groups based on serum osmolality. On their multiple regression analysis, serum osmolality at admission was the only variable associated with death with an odds ratio of 1.08 and a 95% confidence interval of 1.002â€“1.155; statistically significant at p < 0.04). T he risk of mortality increases by 8% for each unit increase in osmolality â€“ or to be precise, by 1.08 times.
How confident are you in the estimates of risk?
The risk associated with increasing osmolality might be from 2% to as high as 16% for each unit increase.
IV. Will the results help me in caring for my patients?
Does the tool maintain its prediction power in a new sample of patients?
This was not tested.
Are your patients similar to those patients used in deriving and validating the tool(s)?
No. There are significant differences in the patient population studied and those admitted to our ICU in Milwaukee on the basis of nutritional status on admission and late presentation. In the study patients, many had not presented until after 4 weeks of polyuria a nd polydipsia and many were chronically malnourished .
Will the results lead directly to selecting or avoiding therapy?
Not directly. The study does generate more important questions however. A similar prospective study would get at questions of predictability more effectively. Other useful questions are proposed in the discussion section of the paper: Is the slow improvement in pH and blood glucose related to under-fluid-resuscitation of the DKA patient? Is the degree of dehydration underestimated? Is the prevalence of treatment-related hypokalemia because of malnutrition or inadequate potassium replacement? What are the anthropologic characteristics of the patient population around concern for early symptoms of DKA like polyuria and polydipsia. What limits access to health care for the patients described in the paper ?
Are the results useful for reassuring or counseling patients?
The results of the study are primarily useful in describing a very high rate of mortality in patients admitted with DKA in a developing country. As stated above, the results of the study generate many more useful questions about why the mortality rate is so high and point to unequal outcomes in a resource-poor area for a common disease process . .
References (if any)
- Balasubramanyan, N., P.L. Havens, and G.M. Hoffman, Unmeasured anions identified by the Fencl-Stewart method predict mortality better than base excess, anion gap, and lactate in patients in the pediatric intensive care unit.[see comment]. Critical Care Medicine, 1999. 27(8): p. 1577-81.
- Marcin, J.P., et al., Factors associated with adverse outcomes in children with diabetic ketoacidosis-related cerebral edema. The Journal of Pediatrics, 2002. 141(6): p. 793-797.
- Glaser, N., et al., Risk Factors for Cerebral Edema in Children with Diabetic Ketoacidosis. N Engl J Med, 2001. 344(4): p. 264-269.
Last Updated: September 11, 2008