Objective. To determine social factors associated with increased risk of hospital admission from diabetic ketoacidosis (DKA) or diabetic coma as well as risk of prolonged hospital stay.
Methods. A cohort of all children (≤21 years) with type 1 diabetes mellitus (DM) in the National Inpatient Sample admitted for DKA or diabetic coma during 1996 or 1997 was conducted. Patients’ age, race, gender, and insurance coverage were identified. Length of stay and charges were examined; prolonged length of stay was defined as ≥7 days.
Results. A total of 8443 children with a primary hospital diagnosis of DKA and 123 children with type 1 DM and coma were identified; 55% of the children were girls, 32% were nonwhite, 29% received Medicaid insurance, and 33% resided in areas of poverty. Children with prolonged hospital stay were significantly more likely to be of nonwhite race (odds ratio [OR]: 2.0; 95% confidence interval [CI]: 1.6–2.5), to receive Medicaid insurance (OR: 1.4; 95% CI: 1.1–1.7), to live in areas of poverty (OR: 1.3; 95% CI: 1.1–1.7), and to be of younger age.
Conclusions. When compared with state census data, nonwhite and poor children were more likely to be admitted with complications of DM and to have significantly prolonged and expensive hospital stays. These children should be targeted for intensive diabetes education and outpatient medical support both to improve their health and potentially to decrease total health care costs.
Diabetes mellitus (DM) is the second most common chronic disease of childhood. The prevalence has been estimated to be 120 cases per 100 000 population among children <14 years of age,1 with an annual incidence of approximately 20 new cases per 100 000 children per year.1,2 By 20 years of age, 0.3% of Americans have type 1 DM; this condition is as common in children as all cancers combined.3
There are approximately 4 hospital admissions per 100 000 children/year for diabetic ketoacidosis (DKA) in the United States.4 Current home monitoring techniques should allow most diabetic care to occur in an outpatient setting. However, DKA remains the most common cause of hospital admission for children with DM. One study reported that after an initial diagnosis of DM in children ages 8 to 13 years, 25% were rehospitalized within 2.5 years5; DKA was the most common reason for rehospitalization.6
The cost of care of a child with diabetes is high. Even families with health insurance can incur large out-of-pocket expenses for treatment of a child with diabetes. In 1 study of the financial impact of DM, 17% of families were found to have out-of-pocket expenses >10% of their household income.7 Therefore, poor families and families without insurance may have to make difficult choices between paying for health care expenses and other family needs.
Medicaid insurance is both a surrogate for low income and a method of payment that contains financial incentives; thus, children with chronic illness and Medicaid insurance may have different access to medical care or receive different care compared with other children. Children with Medicaid insurance and black children have been shown to receive less preventive asthma services and to use more emergency department and inpatient services for treatment of acute asthma exacerbations compared with children with other forms of insurance.8–10Because DM is an important chronic condition among children, we examined the relationship between race and Medicaid insurance with outcomes for children with DM treated for DKA or diabetic coma compared with other diabetic children.
National Inpatient Sample Database
The Agency for Healthcare Research and Quality makes the National Inpatient Sample (NIS) data set publicly available. The NIS contains inpatient stay information collected from state agencies, health associations, and other private organizations that maintain hospital discharge databases. It is designed to approximate a sample of United States nonfederal, short-term hospitals as defined by the American Hospital Association and is stratified according to geographic region, ownership, teaching status, and location.
The NIS includes information from 22 states. Demographic information includes patient race or ethnicity—white, black, Hispanic, Asian or Pacific Islander, and Native American—age, and gender. Diagnosis codes and procedure codes were examined, as well as length of stay (LOS) and survival. Insurance is divided into 4 primary payer groups: Medicaid, commercial or private health maintenance organization, self-pay, and “other” (Medicare, CHAMPUS, Title V, worker’s compensation, and no charge). Patients are grouped into 4 annual income categories based on zip code of residence: $0 to $25 000, $25 001 to $30 000, $30 001 to $35 000, and >$35 000.
We requested all cases for patients who were 21 years old and younger and had a primary diagnosis of DKA (International Classification of Diseases [ICD] code 250.11, 250.13) or diabetic coma in patients with type 1 DM (ICD-9 code 250.31 or 250.33)11 and who were in the data set in 1996 or 1997.
Poverty was defined as having a median annual zip code income of <$25 000. Poverty in this study was defined by using the proxy measurement of median income by postal code. Although this method is not a perfect measure of socioeconomic status, Soobader et al12 found that analysis of income by zip code provides a more robust correlation with individuals’ income than using smaller block groups or tracts. Prolonged stay was defined as LOS ≥7 days. This was approximately the upper 5th percentile of the sample LOS. Patients with a prolonged stay were compared with those without a prolonged stay.
Normally distributed continuous data were compared using the Student t test. The Kruskal-Wallis H test was used to compare skewed continuous data. Categorical data were examined using the χ2 test and odds ratio [OR]. The relationship of Medicaid insurance status and race with prolonged stay in children with diabetic ketoacidosis or diabetic coma was examined by calculating adjusted OR using multiple logistic regression (SPSS 7.5 for Windows, SPSS Inc, Chicago, IL) to control for potential confounding variables such as age and geographic area. The adjusted OR was reported with 95% confidence intervals [CI]. Mean values are reported with the standard deviation, and median values are reported with the 25th and 75th quartiles. Statistical significance was defined as P < .05.
There were 8443 children with DKA and 123 children with type 1 DM and coma in the data set (Table 1). Children who were admitted to a hospital with coma and type 1 DM had similar age, gender, and insurance status distribution when compared with children who were admitted with DKA. Fifty-five percent of patients were girls, and 32% of children were of nonwhite race. The average patient age was 14.3 years (±5.1). Twenty-nine percent had Medicaid insurance, 53% had commercial insurance, and 11% were self-pay. The remaining children had “other” insurance (7%) or insurance status was missing (0.3%). Thirty-four percent lived in areas of poverty (the median zip code income was <$25 000).
Seventeen children (0.2%) died. Children who were admitted with coma had a 39-fold greater risk of death compared with children who were admitted with DKA (OR: 39; 95% CI: 14–108). Although children who were admitted with diabetic coma had significantly higher median hospital charges ($7662 [$4249, $11 564]) compared with hospital charges for children with DKA ($4469 [$2871, $6969]; P < .001), their median length of hospital stay was similar.
Nonwhite children had significantly longer median hospital stay and greater charges compared with white children (3 days [2, 4] vs 2 days [2, 4]; P < .001; and $5499 [$3500, $8540] vs $4168 [$2730, $6419]; P < .001), respectively. Children who had Medicaid insurance also had significantly greater median charges compared with children who had other forms of insurance ($4785 [$3047, $7665] vs $4350 [$2826, $6741]; P < .001).
Patients were divided into those with or without a prolonged hospital stay (≥7 days). Four children in the data set did not have information on LOS and were excluded from further analysis. There were 8114 children who had LOS <1 week and 448 children with a prolonged LOS (Table 2). Children with prolonged LOS were significantly younger (P < .001) and were also more likely to be of nonwhite race (P < .001), to have Medicaid insurance (P < .001), and to reside in areas of poverty (P < .004). As expected, children with prolonged LOS had significantly increased hospital charges compared with children with shorter LOS. However, children with prolonged LOS were not more likely to die.
Table 3 contains a multivariate analysis for risk factors associated with prolonged LOS. After adjustment for living in poverty (median annual income <$25 000), patient age, Medicaid insurance coverage, hospital teaching status, and regions of the country, nonwhite children were twice as likely than white children to have prolonged LOS (OR: 2.0; 95% CI: 1.6–2.5). Children with Medicaid insurance were 1.4 times more likely to have prolonged LOS (OR: 1.4; 95% CI: 1.1–1.7) compared with children with other forms of medical insurance, and children living in areas of poverty were significantly more likely to have prolonged LOS (OR: 1.3; 95% CI: 1.1–1.7).
In this study, nonwhite children and those who had Medicaid insurance and were hospitalized for either DKA or diabetic coma had significantly longer hospital stay and greater charges. They were more likely to have a hospital stay ≥1 week, or a “prolonged” hospital stay. We also found that among hospitalized children with complications of DM, nonwhite children and those with Medicaid insurance were overrepresented compared with the general population.
In 1996, Medicaid insured approximately 21% of American children.13 We found that 29% of all children who were admitted with type 1 DM and or coma and 39% of those with a prolonged stay had Medicaid insurance. This is consistent with the finding by Parker and Schoendorf14 that nearly one third of all children discharged from the hospital in their sample had Medicaid insurance. Likewise, we found a disproportionate number of nonwhite children (32%) admitted with acute complications of DM, and an even greater number (47%) had a prolonged hospital stay. The proportion of nonwhite people who live in the 18 states that reported race to the NIS data set in 1997 was 21.5%.15 We cannot determine whether a patient had multiple hospital admissions during the 2 years, so the overrepresentation of nonwhite and Medicaid patients may be skewed by a group of very poorly controlled children with diabetes and repeat episodes of DKA or diabetic coma. Alternatively, nonwhite children and those who have Medicaid insurance may experience a worse average level of glycemic control compared with other children with diabetes. However, the reported incidence of DM in black children is lower than for white children.16
The reasons for the prolonged LOS among Medicaid patients are unclear and probably multifactorial. We did not have information on severity of illness. Because mortality for DKA is low, we cannot infer that severity of illness at presentation in the 2 groups was similar. In addition, we do not have information on the patients’ severity of hyperglycemia or acidosis at presentation to the hospital or the HgbA1C, a measure of previous glucose control. Furthermore, we do not know whether a patient had a new diagnosis of DM and needed intensive diabetes education that potentially prolonged the hospital stay.
Past studies have shown that children with Medicaid insurance, although faring better than poor children with no insurance, receive different health care than their privately insured counterparts.8,17,18 Children with Medicaid are more likely to receive their care at health centers or clinics, hospital outpatient clinics, or emergency departments.17 They are also more likely to be hospitalized for conditions that potentially are treatable out of the hospital.14,17,18 In addition, among children with acute surgical conditions such as acute appendicitis, children with Medicaid insurance had greater risk of perforation and abscess formation and longer length of hospital stay compared with children with commercial or health maintenance organization insurance.19 These findings suggest that barriers may exist to acute hospital care such that children with Medicaid are more severely ill at the time of hospitalization.
Our study found that living in poverty was a significant independent risk factor for prolonged hospital stay. However, in Spain, where there is universal access to medical care, poverty was not an important factor in hospital admissions for illnesses considered to be sensitive to ambulatory care.20 In addition, being from a single-parent home has been associated with higher hemoglobin A1C even after controlling for race, age, family socioeconomic status, and adherence to medical care as reported by the mother.21 Thus, poverty per se may not be the important cause for prolonged admissions. However, one might hypothesize that the perception of the child’s access to care and ability to have appropriate posthospital care may affect the treating physician’s decisions regarding discharge and lead to lengthened hospital stay in the United States.
The reason that children who were hospitalized in the Midwest and Northeast had significantly greater risk of prolonged stay is uncertain. Regional differences in health care utilization have been reported previously and may be related to physician features as well as differences in health care delivery systems.22–26
The study has other limitations to consider. The diagnosis codes for DM are defined as insulin-dependent (type 1) and non-insulin-dependent (type 2) disease. Children with type 2 DM often have ketosis even when they are initially diagnosed with DM. The data set included 120 children who were coded with non-insulin-dependent hyperosmolar coma and were excluded from the analysis of DKA. Finally, the data provided on race are limited, and the number of children who were Hispanic, Native American, and Asian/Pacific Islander was small. Four of the states, comprising 17% of the data set, provide no information on race and thus were excluded from the analysis of race. Therefore, it is possible that our estimates of the effect of nonwhite race may not reflect the entire United States population.
DKA is considered an important ambulatory care–sensitive condition as defined by Parker and Shoendorf, ie, a condition for which ambulatory care may reduce, although not eliminate, the need for hospitalization.14 Previous studies have shown that the populations that are most likely to be readmitted are black children and adolescents.27 Poor glycemic control, as reflected by elevated hemoglobin A1C, is also predictive of readmission.27 Black youths have been shown to have a 4-times greater risk of poor glycemic control compared with white youths.28 However, outpatient programs that closely monitor individuals with diabetes decrease the need for admission for DKA in these high-risk groups.6
In North Carolina, increased access for Medicaid patients to 24-hour-a-day primary care via the Carolina Access program was shown recently to decrease the number of all visits and nonurgent visits to the emergency department.29 Thus, programs targeted to give consistent primary care for poor children with DM might decrease their need for hospitalization and the future consequences of poorly controlled diabetes. Intensive diabetic care programs that improve glycemic control among adult patients have been shown to decrease total health costs by decreasing complications.30 Among adult patients with DM, treatment of DKA has been estimated to consume >25% of health care dollars and up to 50% among patients with recurrent episodes.31 This suggests that if intensive outpatient pediatric programs improved glucose control, then the children would have fewer episodes of DKA and the programs in all likelihood would be cost-effective.
Nonwhite children and those with Medicaid insurance are at increased risk for a hospital admission and prolonged hospital stay when compared with white children and those with other insurance. Living in poverty and young age were also significant independent predictors of prolonged hospital stay. Improved access to intensive diabetic care programs for nonwhite children and children with Medicaid insurance might potentially decrease the need for hospital admission and provide a better quality of life for these children yet decrease total health care costs.
- ↵Blanchard JF, Ludwig S, Wajda A, et al. Incidence and prevalence of diabetes in Manitoba, 1986–1991. Diabetes Care.1996;19 :807– 811
- ↵LaPorte R, Cruickshanks K. Incidence and Risk Factors for Insulin-Dependent Diabetes. Washington, DC: US Government Printing Office; 1985. Report No. NIH Publ. No. 85
- ↵Glasgow AM, Weissberg-Benchell J, Tynan WD, et al. Readmissions of children with diabetes mellitus to a children’s hospital. Pediatrics.1991;88 :98– 104
- ↵Songer TJ, LaPorte R, Lave JR, Dorman JS, Becker DJ. Health insurance and the financial impact of IDDM in families with a child with IDDM. Diabetes Care.1997;20 :577– 584
- ↵International Classification of Diseases, Ninth Revision Clinical Modification: ICD-9-CM. 4th ed. Los Angeles, CA: Practice Management Information Corporation; 1994
- ↵Medical Expenditure Panel Survey (MEPS) 1996. Available at: http://www.meps.ahrq.gov/papers/99–0042/99–0042.htm#table 4. Accessed August 2000
- ↵Centers for Disease Control and Prevention. Census Projections. Available at: http://wonder.cdc.gov. Accessed February 20, 2001
- ↵Diabetes Epidemiology Research International Group. Geographic patterns of childhood insulin-dependent diabetes mellitus. Diabetes.1988;37 :1113– 1119
- ↵Bratton SL, Haberkern CM, Waldhausen JH. Acute appendicitis risks of complications: age and Medicaid insurance. Pediatrics.2000;106 :75– 78
- ↵Thompson SJ, Auslander WF, White NH. Influence of family structure on youths with diabetes. Health Soc Work.2001;26 :7– 14
- ↵Palta M, LeCaire T, Daniels K, Shen G, Allen C, D’Alessio D. Risk factors for hospitalization in a cohort with type 1 diabetes. Wisconsin Diabetes Registry. Am J Epidemiol.1997;146 :627– 636
- ↵Delamater AM, Shaw KH, Applegate EB, et al. Risk for metabolic control problems in minority youth with diabetes. Diabetes Care.1999;22 :700– 705
- ↵Javor KA, Kotsanos JG, McDonald RC, Baron AD, Kesterson JG, Tierney WM. Diabetic ketoacidosis charges relative to medical charges of adults with type I diabetes. Diabetes Care.1997;20 :349– 354
- Copyright © 2002 by the American Academy of Pediatrics