PEDIATRICS Vol. 107 No. 3 March 2001, pp. 499-504

Excess Risk of Severe Acute Illness in Children With Chronic Health Conditions

Nienke P. Dosa, MD^*, , Nancy M. Boeing, RN, NP, MS^*, and Robert K. Kanter, MD^*

From the ^* Department of Pediatrics, SUNY Upstate Medical University, Syracuse, New York; and the  Department of Pediatrics, University of Rochester School of Medicine and Dentistry, Rochester, New York.

	ABSTRACT

Top Abstract Methods Results Discussion References

We estimated the relative risk (RR) of an unscheduled pediatric intensive care unit (ICU) admission as a marker for severe acute illness in children with chronic health conditions, compared with previously healthy children. Potentially preventable events that lead to acute illness were identified to develop preventive strategies.

Methods.  Children with chronic conditions were defined as those who have a chronic physical, developmental, behavioral, or emotional condition and who also require health and related services of a type or amount beyond that required by children generally. RR was estimated from admissions and regional population data. Potentially preventable events that lead to ICU admission of chronically ill children were identified retrospectively from hospital records.

Results.  Children with chronic health conditions had an RR of 3.3 for an unscheduled ICU admission related to their chronic condition, compared with previously healthy individuals (95% confidence interval [CI] = 2.5-4.2). The risk of severe acute illness in the small subgroup that received technology-assisted care was much greater (RR = 373; 95% CI = 330-422) than for the large group of chronically ill patients who did not require technology-assisted care (RR = 2.3; 95% CI = 1.7-3.0), each expressed relative to previously healthy children. Acute illness related to chronic health conditions accounted for 45% of 251 unscheduled ICU admissions during the 1-year study. Thirty-two percent of admissions that were related to chronic conditions were judged to have been potentially preventable. Preventable events were more common for those who did not require technology-assisted care, occurring in 38% of admissions, compared with those who received technology-assisted care, for whom 19% of admissions involved a preventable event. Fifty-six percent of potentially preventable events involved the physical or social environment and decisions made by the family, whereas 64% could be attributed to health care system factors.

Conclusions.  Children with chronic health conditions account for a substantial share of severe acute illness in a region. Because their underlying conditions have already been identified, problems may be anticipated. The small number of children who receive technology-assisted care each have such a high risk of severe and unavoidable acute illness that individualized emergency care plans are justified. For the remainder of children with chronic conditions, investigation of health system strategies to improve families' ability to anticipate, minimize, or prevent related acute illness is warranted.

Children with chronic health conditions have a more than threefold excess rate of hospitalization compared with the general pediatric population.¹ The excess hospitalizations include acute illness that is related directly to the chronic condition, scheduled hospitalizations for elective procedures, and acute illness that is unrelated to the chronic condition. Little published data describing the health care burden of severe acute illness directly attributable to children's chronic conditions is available.

Diseases and their therapy can be described categorically for diagnosis-specific subgroups or noncategorically for larger populations with shared characteristics. Noncategorical analysis may reveal generic issues common to large populations that are not detectable in diagnosis-specific subgroups. In particular, children with chronic health conditions have many care requirements in common, regardless of their disease. Noncategorical analysis may reveal gaps in existing health services.²

The present study investigated the regional incidence of unscheduled pediatric intensive care unit (ICU) admission as a marker for severe acute illness. The excess risk of emergency admission to the ICU was evaluated for children with chronic health conditions, defined noncategorically, relative to that for previously healthy children. Severe acute illness that is directly attributable to children's chronic conditions is foreseeable and may be avoidable. Therefore, we identified and classified potentially preventable events that lead to ICU admission. An understanding of typical patterns of events that lead to severe acute illness in children with chronic health conditions might suggest strategies to prevent these illnesses or to minimize the severity of unpreventable illnesses when they occur.

	METHODS

Top Abstract Methods Results Discussion References

Design

We postulate that emergency, unscheduled ICU admissions provide a measure of the regional incidence of severe acute illness in a population. We performed a prospective study of admissions to a tertiary pediatric ICU that serves as the only such facility for a 17-county region. Using population data detailed below, we determined the incidence of unscheduled ICU admissions of children with chronic health conditions, whose severe acute illness was judged by the investigators to be related to their chronic condition. The incidence of unscheduled ICU admissions of previously healthy children served as the basis for comparison. Hospital records of the ICU admissions that were related to chronic conditions were examined retrospectively to identify potentially preventable illnesses.

Populations, Definitions, and Data Collection

Chronic conditions were defined noncategorically by the criteria of the Federal Maternal Child Health Bureau as those who have a chronic physical, developmental, behavioral, or emotional condition and who also require health and related services of a type or amount beyond that required by children generally.^1,3 By this definition, it has been estimated that 18% of all children who are <18 years old have a chronic health condition¹ and that the remaining children are free of a chronic condition; for the purposes of the present study, 82% of the population was considered to represent the previously healthy individuals. Patients with unscheduled ICU admissions were classified as previously healthy or having a chronic health condition. Patients were considered to have a chronic condition if a preexisting disorder, eg, as a congenital defect, was newly diagnosed on the current admission.

Subpopulations of children with chronic health conditions were analyzed further according to their need for technology-assisted care. A previous study in our region determined that 4.7 per 10 000 children <18 years old received technology-assisted care.^4,5 For purposes of the present study, technology-assisted care was considered to include mechanical ventilation, intravenous medications or nutrition, daily device-based respiratory or nutritional support (tracheostomy, oxygen, tube feeding), implanted cardiac pacemakers, and implanted pumps for administration of medication. We did not classify apnea monitors, urinary catheters, or colostomies as technology-assisted care.

Data were collected during a 1-year period (April 1998 to April 1999) to provide a seasonally unbiased sample of acute illnesses. Demographic data were collected from hospital charts during the patients' hospitalizations, including ICU day 1 mortality probability.⁶ Patients with scheduled admissions were excluded from analysis because they do not represent typical regional acute illness associated with chronic conditions; rather, they reflect unique patterns of services and procedures that are available in a particular hospital. The patients with chronic conditions and whose ICU admission could not be attributed to their chronic conditions also were excluded from the analysis of risk associated with chronic illness. In addition, patients who were admitted from counties outside our usual 17-county referral region and those who were 18 years and older were excluded from analysis of regional pediatric population risks.

Charts of all unscheduled ICU admissions that were related to chronic health conditions were reviewed and abstracted retrospectively to investigate the events that lead to severe acute illness. Events that preceded admission were classified by the authors into categories adapted from McConnochie et al.⁷ Category 1 events describe biological factors, some of which are present for all children with chronic health conditions, by definition. Biological factors include the natural course of illness and anticipated complications of diseases and procedures. For purposes of the study, category 1 events are considered unpreventable by existing health care services. Events in the categories 2 to 6 are considered potentially preventable. Category 2 events involve the relationship between the child's biological condition and decisions made by the family, as well as factors in the social or physical environment. Delays in symptom recognition or seeking medical attention and deficiencies in carrying out existing management plans are included in this group. Category 3 events include interactions between the chronic health care services and the child-family unit. Diagnostic delays, chronic management deficiencies, incomplete family health education, and obstacles in access to the chronic health care provider are considered to be in this category. Category 4 events imply deficiencies in episode-specific interaction between the health care system and the child-family unit. These may include diagnostic delays, management decisions, and difficulties in gaining access to the episode-specific provider. Category 5 events describe interactions between chronic and episode-specific health care providers. Finally, category 6 events are those in which it seems that an illness might have been preventable but insufficient information is contained in the chart for accurate classification. Some patients had >1 category of potentially preventable event.

The region served by our ICU includes several small cities and an extensive rural area. US Census data indicate a total population of 472 000 children who are <18 years old in the 17-county region.⁵

The study ICU is a 7-bed multidisciplinary unit in an academic health center. Pediatric medical and surgical patients and some neonatal surgical patients are admitted to this unit. The unit serves as the regional pediatric trauma and cardiac surgical center. Pediatric burn patients with smoke inhalation generally are admitted to this ICU, although older children with isolated burn injuries sometimes are admitted to a separate adult burn unit. ICU admission criteria include any of the following: 1) requirement for at least 1 nurse per 2 patients, 2) vital signs assessment at least hourly, 3) fluid resuscitation or vasoactive drugs for shock, 4) acute mechanical ventilation or impending respiratory failure, and 5) treatment for intracranial hypertension. Decisions to admit unscheduled patients all were made by ICU physicians who were board-certified in pediatric critical care medicine. Because a separate intermediate unit was available to monitor physiologically stable patients who need chronic ventilation or other technological assistance, ICU care was reserved for patients who were judged to be physiologically unstable.

Data collection for the study was approved by the Institutional Review Board for the Protection of Human Subjects, and the requirement for consent was waived for collection of epidemiologic information.

Analysis

Characteristics of populations were described as the median and 90th percentile, or numbers in groups. Comparisons between the groups were analyzed by Mann-Whitney rank sum test for non-normally distributed continuous data, or ² tests for categorical data. Differences were considered to be statistically significant at P < .05. The excess risk of unscheduled ICU admissions for chronically ill children relative to previously healthy populations was expressed as the risk ratio and 95% confidence intervals.⁸

	RESULTS

Top Abstract Methods Results Discussion References

During the 1-year study period, 411 patients accounted for 444 admissions to the ICU. Patients who were excluded from analysis of regional risks accounted for 185 electively scheduled admissions, 27 admissions of children from outside the usual referral area, and 1 admission of a patient who was older than 17 years. Thus, 251 unscheduled pediatric ICU admissions from the region occurred during the study period (Fig 1). A total of 248 patients were analyzed for the relative risk (RR) of admission that was related to a chronic condition and identification of preventable events. Three patients with chronic conditions were excluded from the analysis because their acute illnesses were unrelated to the chronic condition. Ninety-seven patients were admitted 112 times for severe acute illnesses that were related to their chronic condition, and 136 previously healthy children were admitted a single time.

View larger version (32K): [in this window] [in a new window]   Fig. 1.   Unscheduled ICU admissions classified according to chronic condition related or unrelated to acute illness, or previously healthy. For chronic conditions with related acute illness, subgroups are also indicated.

The groups of chronically ill and previously healthy children differed in the following ways (Table 1). Those with a previous chronic illness were significantly older; 42% had reached school age (5 years or older). Children with a previous chronic illness were more likely to be receiving publicly funded health insurance.

Children with chronic health conditions had more than a threefold excess risk that a severe acute illness would result in an unscheduled ICU admission, compared with previously healthy children (Table 2). The excess risk exceeded 300 for the subgroup of children who required technology-assisted care. Although the number of children who required technology-assisted care was estimated to be only 222 in a regional population of 472 000,^4,5 their 36 admissions accounted for 14% of all 251 unscheduled admissions to the ICU. The remainder of the children with chronic conditions who did not receive technology-assisted care in the community were only approximately twice as likely to require ICU admission for severe acute illness, compared with the previously healthy population. Although risk for chronically ill children who did not require technology-assisted care is only slightly increased individually, their 76 admissions accounted for 30% of 251 emergency ICU admissions.

Populations also were examined by age. The excess risk of a severe acute illness in chronically ill infants (0-2 years) was greater (RR = 6.2) than that for older children (RR = 2.7), relative to previously healthy children.

Subgroups of chronically ill children with related severe acute illness were classified according to chronic technology assistance, primary chronic organ system involved, and congenital or acquired cause (Fig 1). Neurologic disorders were the most common chronic organ derangement associated with acute illness, resulting in 15% of unscheduled admissions. Chronic neurologic disorders included seizures, congenital malformations of the nervous system, and brain tumors. Acute illnesses that required ICU admission and that were associated with underlying neurologic conditions included status epilepticus, aspiration pneumonia, and viral lower respiratory infections. Chronic cardiovascular conditions included congenital heart disease and cardiomyopathy in 9% of unscheduled admissions. In this group, acute decompensation of congestive heart failure, viral lower respiratory infections, and emergency surgical procedures that were related to the cardiac lesions accounted for their unscheduled ICU admissions. Seven percent of unscheduled admissions were due to chronic respiratory conditions, including exacerbation of chronic asthma and bronchopulmonary dysplasia. An additional 14% of emergency ICU admissions were associated with a variety of less common chronic conditions. Chronic conditions classified by congenital cause rather than by organ system account for 16% of unscheduled admissions. Congenital disorders include single gene, chromosomal, multigene, multifactorial, and unclassified anomalies.⁹

Events that lead to related severe acute illness in chronically ill children are summarized in Table 3. One third of admissions were potentially preventable. Potentially preventable events occurred significantly more frequently for chronically ill children who were not receiving technology-assisted care than for those who required technology-assisted care (P < .05).

Fifty-six percent of potentially preventable events that lead to severe acute illness involved family and environmental factors (category 2 events), such as delays in seeking medical attention (n = 6), medication noncompliance (n = 5), and inappropriate parental supervision (n = 3). Environmental factors included exposure to smoking and emotional stresses leading to self-destructive behavior for children with mental illness.

Health system deficiencies contributed to 64% of potentially preventable severe acute illnesses (categories 3-5). These included deficiencies in chronic care (category 3), such as inadequate care coordination (n = 5), failure to provide mental health services (n = 3), and end-of-life care for a patient who might have more appropriately been referred to hospice care (n = 1). The acute health care system was implicated in 12 potentially preventable admissions (category 4). The majority of these were related to diagnostic decisions. Overall, health care system deficiencies contributed to 21% of all unscheduled ICU admissions of children with chronic conditions; chronic and acute care services were implicated equally.

	DISCUSSION

Top Abstract Methods Results Discussion References

Children with chronic health conditions, considered noncategorically with respect to their diagnoses, have approximately a threefold excess risk that severe acute illness will require an unscheduled ICU admission when compared with previously healthy individuals. Children with chronic health conditions account for nearly half of unscheduled ICU admissions. Most important, one third of their admissions may be preventable.

Estimated risks of severe acute illness that is associated with chronic conditions depend on the assumed size of the chronically ill population. Our estimate is based on the 18% that require special care or services.¹ If a chronic condition is defined as the risk for needing special services at a future time, then an unknown additional number would be involved. Other definitions identify smaller proportions of the pediatric population¹: 6.7% have limitations in their school or play activities; 0.2% have a limitation in eating, bathing, or dressing; and 0.047% receive technology-assisted care.^4,5

Biases in the populations sampled might influence our findings. Similar predicted mortality risk was observed for the chronic and previously healthy groups (Table 1), suggesting no bias in the application of ICU admission criteria. The RR of severe acute illness that is associated with chronic conditions might be overestimated if children with chronic disorders were preferentially admitted to our tertiary ICU while previously healthy patients received intensive care at other locations in or outside the region. Those with chronic conditions might be more likely to be transferred to a center within the region because of their greater medical complexity, previous relationship with the regional center, and the incentive for community hospitals to avoid nonrevenue-generating care for a severe acute illness in a chronically ill child. Previously healthy patients are not likely to have a previous relationship with a regional center; therefore, those who reside in counties that border an adjacent referral region might be more likely to be referred to an ICU outside the region. Examination of data by selected counties suggests that this may occur. For rural counties near Syracuse (Oswego, Oneida, Cortland, and Madison counties), the ICU admission rate was more than twice as high (4.5 per 10 000 healthy population) than for counties in our region that border adjacent regions (Cayuga, Tompkins, Chemung, Herkimer, Otsego, and Delaware counties; 1.9 per 10 000 healthy population). However, if the hypothetical "missing" ICU admissions of previously healthy children from distant counties are added at a rate equal to that from counties near Syracuse,⁵ then only 29 patients would be added. RR of ICU admission for chronically ill children still would exceed the previously healthy group by a factor of 2.7.

Conversely, risks of severe acute illness in patients with chronic health conditions would be underestimated if chronic patients tended to be transferred to hospitals outside our region for their acute care. This might occur if they had a previous relationship with specialists at another center outside the region. Emergency ICU admissions usually require stabilization at the nearest pediatric center within the region. The vast majority of patients within the region reside closer to Syracuse than to ICUs in adjacent regions, so such a bias in the data is unlikely. Any study of severe acute illness in a region may involve patient observations lost beyond the border of the region, to unsampled hospitals within the region, or as the result of prehospital deaths and deaths occurring in emergency departments. Because our center serves as the only tertiary pediatric ICU for the region, the population sampled is likely to represent regional events with greater accuracy than if such a study were conducted in a region that is served by multiple pediatric ICUs.

Overestimation of risk for chronic patients also might occur if ICU admission was arranged for patients who required chronic ventilator assistance and who were hospitalized but not severely ill. This bias was avoided in the present study because stable patients who were assisted by a ventilator were admitted to a separate intermediate care unit and excluded from analysis. Finally, risk reported for the subgroup that requires technology assistance may be overestimated if the population of patients who require technology assistance and currently live in the region exceeds the assumed number estimated from a 1992 survey.⁴ No recent data document national trends regarding the prevalence of community-based technology-assisted care.

The hospitalization of a child with a chronic health condition may be viewed as an opportunity to identify potentially correctable health system processes. We assumed that multiple factors contribute to acute illness. Although it often is possible to identify a caregiver or provider who was directly involved in an adverse event, such events may be better understood and prevented in relation to underlying health care system processes.¹⁰ Our categorization of preventable events that precede ICU admission is a preliminary effort to apply a conceptual framework on preventable morbidity to a regional population of children with chronic conditions. We did not attempt to validate categories because of limitations imposed by retrospective data collection. Prospective investigation will allow for more rigorous analysis.

Assignment of preventable events into the categories "chronic" or "episode-specific" health care system events (category 3 or 4) is somewhat arbitrary. Chronic versus acute providers and underlying versus acute illnesses sometimes are difficult to distinguish. Because data on events that lead to hospitalization were obtained retrospectively, observations may be incomplete and probably represent underestimated rates of preventable events. In particular, it is likely that interactions between chronic and episode-specific providers (category 5) are underreported in a retrospective sample from hospital charts. Despite the insensitivity of a retrospective chart review, we identified potentially preventable events before one third of unscheduled ICU admissions that were related to chronic illness.

Our observations identify 2 different subgroups whose excess risks suggest distinct preventive and management strategies. For 81% of the admissions for the group that required technology assistance, no potentially preventable event was found. Thus, careful attention to preparing for unavoidable emergencies is warranted.^11,12 Although not detected by our observations, other children may demonstrate a high risk that is unrelated to technology-assisted care, and individual emergency care plans may be necessary for them as well.¹³

The much larger group of children who have chronic conditions and who do not require technology-assisted care have only slightly increased risk individually, but collectively, they account for a substantial regional burden of acute illness. Potentially preventable events were identified for more than one third of admissions in this group. Most of these can be attributed to family decision and the patients' physical and social environment. However, in many cases, a caregiver's role might have been more effective in preventing or minimizing acute illness if the health system had better prepared the family for their responsibilities. Although families have a central role in providing complex care, previous observations have shown that they often have an inadequate understanding of their child's chronic illness.^14,15 Improvements in the health system's ability to collaborate with family caregivers must be developed. Because almost half of the patients with chronic conditions that required emergency ICU admission were of school age, it may be useful to include school health providers in preventive and emergency care planning.

Preventive efforts that are generalizable to large populations include professional education. Development of the generalist medical discipline of pediatric chronic disease management may contribute to service and educational improvement.^16-18 Because chronically ill patients make up nearly half of the unscheduled admissions to a pediatric ICU, closer attention to the relationship between intensive care and community-based chronic care may be warranted in critical care training.

Chronically ill children who were admitted to the ICU were more likely to have health insurance (especially publicly funded insurance) than were previously healthy children. However, costs of hospitalization for chronic conditions are incompletely covered by most types of health insurance. As a result of potentially preventable severe acute illness, substantial costs are imposed on families,¹⁹ hospitals,²⁰ and providers,²¹ as well as on public resources.²² Modest reallocation of resources to prevention and preparation might reduce total costs of acute illness that are associated with chronic conditions.

This study contributes to a profile of health care use among children with chronic conditions by providing data that population-based studies do not detail. We assume that the need for intensive care represents the most severe end of a spectrum of morbidity that constitutes chronic disease. It is notable that excess risk for intensive care in this regional study parallels national estimates for other indicators of acute illness for children with chronic conditions; 3.1 times as many bed days as a result of illness, 2.6 times the number of school absences as a result of illness, 3.4 times as many hospitalizations, and 2.5 times the physician contacts, each relative to children without special health care needs.¹

The diversity of patients illustrates one of the inherent challenges posed by childhood chronic conditions, which are characterized by their heterogeneity. Whereas adult-onset disability involves a few relatively common conditions, pediatric conditions, particularly congenital conditions, often present as rare or even singular diagnostic or management challenges for the community health care system. The large number of pediatric chronic conditions justifies a noncategorical approach to analyzing and planning their acute and chronic care, to find common management elements among diverse disorders. Efforts to promote collaboration among families, home and primary care providers, subspecialists, and emergency and acute care services may be particularly beneficial for medically vulnerable children.

	FOOTNOTES

Received for publication Jan 21, 2000; accepted Jun 29, 2000.

Address correspondence to Robert K. Kanter, MD, Department of Pediatrics, 750 E Adams St, Syracuse, NY 13210. E-mail: kanterr{at}upstate.edu

	ABBREVIATIONS

ICU, intensive care unit; RR, relative risk.

	REFERENCES

Top Abstract Methods Results Discussion References

1. Newacheck PW, Strickland B, Shonkoff JP, An epidemiologic profile of children with special health care needs. Pediatrics 1998; 102:117-121 [Abstract/Free Full Text]
2. Perrin JM Universality, inclusion, and continuity: implications for pediatrics. Pediatrics 1999; 103:859-863 [Free Full Text]
3. McPherson M, Arango P, Fox H, A new definition of children with special health care needs. Pediatrics 1998; 102:137-140 [Free Full Text]
4. Ciota R, Singer N. Steering Committee on Home Care for Chronically Ill Children, 1987-1992, Final Report. Syracuse, NY: Central New York Health Systems Agency, Inc; 1992
5. US Bureau of the Census. County and City Data Book, 1994. Washington, DC: US Government Printing Office; 1994
6. Pollack MM, Ruttimann UE, Getson PR Pediatric risk of mortality (PRISM) score. Crit Care Med 1988; 16:1110-1116 [Medline]
7. McConnochie KM, Roghmann KJ, Kitzman HJ, Liptak GS, McBride JT Ensuring high-quality alternatives while ending pediatric inpatient care as we know it. Arch Pediatr Adolesc Med 1997; 151:341-349 [Abstract]
8. Ahlbom A, Norell S. Introduction to Modern Epidemiology. Chestnut Hill, MA: Epidemiology Resources; 1984:63-73
9. Baird PA, Anderson TW, Newcombe HB, Lowry RB Genetic disorders in children and young adults: a population study. Am J Hum Genet 1988; 42:677-693 [Medline]
10. Bogner MS, ed. Human Error in Medicine. Hillsdale, NJ: Lawrence Erlbaum Associates; 1994
11. Sacchetti A, Gerardi M, Barkin R, Emergency data set for children with special health care needs. Ann Emerg Med 1996; 28:324-327 [CrossRef][Medline]
12. Emergency Medical Services for Children, Task Force on Children with Special Health Care Needs EMS for children: recommendations for coordinating care for children with special health care needs. Ann Emerg Med 1997; 30:274-280 [CrossRef][Medline]
13. Sherman JM, Capen CL The Red Alert Program for life-threatening asthma. Pediatrics 1997; 100:187-191 [Abstract/Free Full Text]
14. Carraccio CL, Dettmer KS, duPont ML, Sacchetti AD Family member knowledge of children's medical problems: the need for universal application of an emergency data set. Pediatrics 1998; 102:367-370 [Abstract/Free Full Text]
15. Levine C The loneliness of the long term caregiver. N Engl J Med 1999; 340:1587-1590 [Free Full Text]
16. Liptak GS, Burns CM, Davidson PW, McAnarney ER Effects of providing comprehensive ambulatory services to children with chronic conditions. Arch Pediatr Adolesc Med 1998; 152:1003-1008 [Abstract/Free Full Text]
17. Goldberg AI, Gardner HG, Gibson LE Home care: the next frontier of pediatric practice. J Pediatr 1994; 125:686-690 [CrossRef][Medline]
18. American Academy of Pediatrics, Committee on Children With Disabilities Guidelines for home care of infants, children, and adolescents with chronic disease. Pediatrics 1995; 96:161-164 [Abstract/Free Full Text]
19. Perrin JM, Shayne MW, Bloom SR. Home and Community Care for Chronically Ill Children. New York, NY: Oxford University Press; 1993:87-99
20. Silber JH, Gleeson SP, Zhao H The influence of chronic disease on resource utilization in common acute pediatric conditions. Arch Pediatr Adolesc Med 1999; 153:169-179 [Abstract/Free Full Text]
21. Colby DC Medicaid physician fees, 1993. Health Aff 1994; 13:255-263 [Abstract]
22. Ireys HT, Anderson GF, Shaffer TJ, Neff JM Expenditures for care of children with chronic illnesses enrolled in the Washington State Medicaid program, fiscal year 1993. Pediatrics 1997; 100:197-204 [Abstract/Free Full Text]