Infant and late adolescent bimodal age distribution of pediatric tracheotomy.
Lewis CW, Carron JD, Perkins JA, Sie KCY, Feudtner C. Tracheotomy in Pediatric PatientsA National Perspective. Arch Otolaryngol Head Neck Surg. 2003;129(5):523-529. doi:10.1001/archotol.129.5.523
During the past 50 years, changes in the epidemiology of infectious diseases and the capabilities of medical technology have altered the indications for, and implications of, tracheotomy in children. Given the complexity of health care that these patients subsequently require, monitoring the performance of this procedure and patient outcomes across the diverse US health care system is warranted.
To characterize children who received tracheotomies in 1997 and to determine whether disposition and mortality vary by region or health care system attributes.
A nationally representative retrospective cohort drawn from an 80% sample of administrative hospital discharge records from all pediatric admissions in 22 states during 1997.
Patients aged 0 to 18 years who underwent tracheotomy.
The sampling scheme of the discharge records enabled the calculation of regional and national estimates and of age-stratified population-based rates of tracheotomies. Weighted descriptive statistical and Poisson analyses were performed.
The 2065 tracheotomy procedures recorded in the Kids' Inpatient Database yielded a national estimate of 4861 tracheotomies performed in 1997. The mean length of hospital stay was 50 days, with a mean total facilities charge exceeding $200 000. The rate of tracheotomy was highest among infants and varied significantly across regions of the United States. Adjusting for other patient and health care system attributes, patients who received their tracheotomy in a children's hospital had half the risk of dying during the admission compared with patients who were cared for in a non–children's hospital. Hospitals that performed more pediatric tracheotomies had significantly lower mortality rates than hospitals with lesser case volume. Among patients who survived to discharge, those cared for in the Northeast were discharged to long-term care facilities at twice the rate of patients in the West. Children cared for in children's hospitals or in teaching hospitals were significantly less likely to be discharged to a long-term care facility.
Pediatric tracheotomy is associated with significant variation in rates and outcomes across the United States and across different hospital types. Further research to clarify the reasons for these associations is warranted.
DURING THE PAST half century, the indications for—and implications of—tracheotomy among children have changed.1- 11 Formerly, acute infections such as diphtheria, croup, and epiglottitis were the main causes of airway compromise leading to tracheotomy. If the child survived the acute episode, the underlying process would resolve and the tracheotomy would no longer be needed. The advent of vaccines against Corynebacterium diphtheriae and Haemophilus influenzae, coupled with the increased use of endotracheal intubation, markedly reduced the use of tracheotomies for acute airway infections. However tracheotomies are now performed increasingly often for children who require prolonged mechanical ventilation or who have critical airway obstruction caused by an underlying chronic condition. For these children, tracheotomies often are required for months to years, with the consequently medically fragile children typically requiring complex tertiary care and, on discharge, labor-intensive home care.
To date, our knowledge of the epidemiology of tracheotomy has been limited to information garnered from case series from single institutions.1- 18 There are 2 reasons, however, to seek more broadly gathered information that would permit comparisons across institutions and across regions of the nation. First, comparison of the rates at which tracheotomies are performed in different regions may yield evidence of practice variation resulting from health care system characteristics. Second, comparison of outcomes, specifically disposition locale and mortality, across institution types may suggest hospital characteristics, such as patient volume, as specific sources of variation.
We therefore sought, by using a newly available, nationally representative hospital discharge database from 22 states, to describe the children who received tracheotomies and to determine whether tracheotomy rates or outcomes vary by region or health care system characteristics.
This study used data from the Healthcare Cost and Utilization Project (HCUP), a federal–state–private sector collaboration of inpatient discharge abstract data collection sponsored by the Agency for Healthcare Research and Quality, Rockville, Md. The HCUP Kids' Inpatient Database (KID) was designed to allow analyses of pediatric-specific discharge abstract data.19 This data set was released for the first time in the autumn of 2001. (Information on the KID database is available from http://www.ahcpr.gov/data/hcup/hcupkid.htm, accessed August 14, 2002.)
Data are collected and compiled in a uniform format from the 22 states that contributed inpatient discharge data to HCUP. These states are Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Iowa, Illinois, Kansas, Maryland, Massachusetts, Missouri, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Utah, Washington, and Wisconsin. The KID contains data from 2521 hospitals and includes 1.9 million unweighted pediatric discharges. In design of the KID database, 20% of uncomplicated births and 80% of all other pediatric discharges that occurred during 1997 were sampled. These discharges were weighted to obtain national estimates that are representative of 6.7 million pediatric hospital inpatient discharges from acute care community, academic, and specialty hospitals in the United States.
Pediatric discharges were defined as all discharges of patients who were 18 years or younger at admission. The KID included a core set of information found in a typical discharge abstract, including such variables as admission age, source of admission, disposition status, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis (up to 15 possible) and procedure (up to 15 possible) codes, length of stay, and total charges. Charge data were only for the facility and exclude professional fees. Hospital-specific data included ownership or control, bed size, teaching status, children's hospital type, rural or urban location, and US geographic region.
Using ICD-9-CM procedure codes, we extracted all discharges from KID where temporary (31.1) or permanent (31.2) tracheotomy was performed either as a primary or other procedure. We did not include children who received only a tracheotomy revision. Underlying diagnoses were examined by Clinical Classification Software, a tool developed at the Agency for Healthcare Research and Quality for grouping patient diagnoses and procedures into a manageable number of clinically meaningful categories.20 We developed additional diagnostic categories for children who had any of the following ICD-9-CM codes as their discharge diagnoses: trauma or injury (800-995.81), with head injury (850-854) as a subgroup; congenital anomaly (740-759), with craniofacial congenital anomaly (744, 748.0-748.3, 749, 750.0-750.3, 755.55, 756.0) as a subgroup; prematurity (765), respiratory distress syndrome (767.0), or bronchopulmonary dysplasia (770.7); chronic upper airway abnormality (474, 476, 478.29, 478.3, 478.5, 478.74, 478.75, 519.1); acute infectious upper airway obstruction (463, 464, 475, 478.21-478.25, 528.3, 682.0, 682.1); pulmonary disorder: pneumonia, aspiration, infiltration, or insufficiency (466.11, 466.19, 480-518.9, 519.2-519.9); neurologic disorder (320-352.9, 357-359.9, 768); or neoplasm or malignancy (140-239).
Statistical analyses were performed with Stata 7.0 (Stata Corp, College Station, Tex). Counts were weighted by hospital-specific discharge weights to derive national estimates of number of discharges in which tracheotomy was performed. The US census population estimates for 1997 (available at http://www.census.gov/population/estimates/state/stats/st-99-12.txt, accessed November 25, 2001) were used to generate population-based incidence rates of tracheotomies. The survey commands in Stata 7.0 accounted for clustering and were used to generate weighted proportions and mean estimates and to conduct Poisson regression analyses. We constructed 2 distinct Poisson multivariate regression models to assess the outcomes of interest: death during hospitalization (compared with survival to discharge) or disposition to a long-term care facility (compared with discharge to home, home health care, or a short-term facility; patients who died were excluded), and each variable that we hypothesized a priori to be associated with the outcome variables. Predictor variables included in the model were primary payer (Medicaid as the reference category), age category (0-11 months as the reference category), number of diagnoses, whether the child received mechanical ventilation for longer than 96 hours during hospitalization (ICD-9-CM procedure code 96.72), whether there was an injury or poisoning diagnosis, whether the child had been transferred from another facility for this hospitalization, children's hospital type (non–children's hospital as the reference category), teaching hospital status, US geographic region (Northeast, Midwest, South, and West, which served as the reference category), and the annual tracheotomy case volume of each hospital, categorized into quartiles (with the lowest quartile serving as the reference category). Mean estimated number of tracheotomy discharges per hospital in 1997 in each of the 4 volume categories, from lowest to highest, were 4, 20, 39, and 72, respectively.
The 2065 tracheotomy procedures recorded in the KID database yielded a national estimate of 4861 tracheotomies performed in 1997. Among all age groups, the population-based incidence rate varied significantly across regions of the United States, with the highest rate in the Northeast at 7.5 tracheotomy-related discharges per 100 000 child-years (Table 1). Stratifying by age demonstrated that infants had the highest rates at 41.5 tracheotomy-related discharges per 100 000 child-years. Significant variation in incidence of tracheotomies existed between regions for infants, children, and younger adolescents, while rates for 15- to 18-years-olds did not vary significantly across regions.
Tracheotomies were predominantly performed at the extremes of the pediatric age range (Figure 1 and Table 2). Across all ages, lengths of stay averaged 50.2 days, with total charges for tracheotomy-associated hospitalizations exceeding $200 000. Infants had longer stays and higher hospital charges relative to other age groups. The proportion of children discharged to a long-term care facility increased with the age of the child, while the proportion of children receiving their tracheotomy at a children's hospital decreased.
Characteristics of tracheotomy-related hospitalizations varied significantly across US geographic region (Table 3). Medicaid was the primary payer for more tracheotomy hospitalizations in the South and West than in the Northeast and Midwest. Teaching institutions were the site of 93% of the tracheotomies performed in the Northeast, a substantially higher proportion than the national average of 72.3%. Only 5.4% of tracheotomies in the South were performed in a children's hospital. The proportion of patients who received mechanical ventilation for more than 96 hours during their hospitalization was highest in the West (54.6%).
The primary diagnosis assigned to children who received a tracheotomy differed depending on the age of the child (Table 4). Infants had the greatest variety in the primary diagnosis listed, the most frequent of which was congenital anomalies, prematurity, or pulmonary disorders. Only 4.2% of children younger than 1 year had an ICD-9-CM code for injury, while 60.3% had an ICD-9-CM code for a congenital anomaly among their list of diagnoses (Table 2). In contrast, in late adolescence injury was the primary diagnosis in 77.2% of tracheotomy discharges, with one third of these being head injuries. Acute infectious upper airway obstructive disorders (such as epiglottitis) accounted for approximately 1% of all primary diagnoses; they were most frequent in the 1- to 9-year-old age group, in whom they were listed as the primary diagnosis in 2% of cases.
Accounting for multiple patient characteristics, we found that the rate of death in children's hospitals was half that of patients who had their tracheotomies performed in non–children's facilities (Table 5). The mortality rate was significantly associated with pediatric tracheotomy volume, with hospitals in the largest case-volume quartile (estimated mean number of tracheotomy discharges per hospital in 1997, 72) having only 5% of the mortality risk observed among hospitals in the smallest case-volume quartile (estimated mean number of tracheotomy discharges per hospital in 1997, 4). There was a significant declining linear trend in mortality risk across all volume quartiles (P<.001). Children in the 1- to 9-year-old age group (compared with infants) and those who had an injury ICD-9-CM diagnosis (compared with no injury) were also significantly less likely to die during their hospitalization. Factors associated with an adjusted increased risk of death included a greater number of diagnoses or mechanical ventilation for more than 96 hours. No significant difference in the mortality risk existed among US regions, by payer group or teaching hospital status.
Among patients who survived to discharge, the 10- to 14-year-old and 15- to 18-year-old age groups had significantly higher rates of disposition to a long-term care facility than infants, again adjusting for various patient characteristics (Table 5). Discharges from hospitals in the Northeast or Midwest relative to those in the West were associated with a significantly higher rate of disposition to a long-term care facility. Furthermore, those with more diagnoses, who had received mechanical ventilation for more than 96 hours, and who had an injury ICD-9-CM diagnosis were significantly more likely to be discharged to a long-term care facility. Factors associated with a significantly lower rate of long-term care facility discharge included being discharged from a children's hospital, from a teaching facility, or from a hospital with a larger case volume.
This national study showed that pediatric tracheotomies were a relatively infrequent procedure, occurring at a rate of 6.6 children per 100 000 child-years, and associated with only 0.07% of all pediatric discharges nationally. Tracheotomy rates were highest in the 2 extremes of the pediatric age range. Although discharge diagnoses differed markedly across ages, chronic conditions and injury predominated. Acute infections of the airway were an associated diagnosis in only a small number of children. Children who received tracheotomies in 1997 used a substantial amount of health care resources, with mean length of hospital stay of 50 days and total hospital facility charges averaging more than $200 000. The long hospital stays and high costs were more likely due to the underlying conditions that gave rise to the need for tracheotomy, rather than to the tracheotomy per se. Estimated hospital charges for all children who received tracheotomies in 1997 exceeded $952 million. Of patients who survived to discharge, 47% were sent to either home health care or a long-term care facility where additional health care costs would have been incurred but could not be quantified in this study. Across regions of the United States and across hospital types there were significant variations in incidence, mortality, and discharge disposition for children who received tracheotomies in 1997.
Use of this newly available pediatric-specific administrative database offers the unique advantage of providing a national perspective on pediatric tracheotomy rates and outcomes. However, several important potential limitations associated with administrative data bear mentioning. Other variables that could have increased our understanding of pediatric tracheotomies in the United States were not available. For example, indication for tracheotomy was not included in the database. Chart audit was impossible given the anonymous nature of the data. We therefore had to rely on the first diagnosis listed as a proxy, even though this may not have reflected the diagnosis that led to tracheotomy placement. Location of the hospital was not provided in the database to protect the identity of the hospitals, thus limiting our ability to examine smaller area variations. Nor were patient-specific identifiers provided, prohibiting our ability to determine whether a child had more than 1 discharge associated with placement of a tracheotomy in 1997 (although we did omit tracheotomy revisions). We were also unable to determine, from the information contained in this database, whether a child was actually discharged with a tracheotomy in place or receiving mechanical ventilation. Finally, although data are compiled in a uniform fashion by HCUP, there may be differences between institutions in how discharge data are coded. However, given that tracheotomy placement is designated by only 2 ICD-9-CM codes, we would not expect our case definition to have been influenced by coding variation. Furthermore, there is little ambiguity in either the procedure or the outcomes of interest (eg, mortality), therefore making it unlikely that systematic bias would affect our primary hypotheses or our results.
Regional variations in hospitalization and surgical procedure rates are well-documented phenomena.21- 23 Birkmeyer and colleagues22 have argued that the greatest variability in surgical procedure rates occurs when there is uncertainty regarding the best means to manage certain conditions. This explanation may apply to pediatric tracheotomy. Among older adolescents—in whom less regional variation was seen—head and neck injury, or other serious injury requiring prolonged mechanical ventilatory support, was the predominant primary diagnosis associated with tracheotomy. By contrast, among younger patients, the variety in associated primary diagnoses was greater; a consequently broader array of management options for these varied conditions may have led to the regional variation noted in this study. Furthermore, differences in how the values of patients, families, and providers are used to guide medical decision making may also lead to variation in procedure rates across regions.22 Such differences are more likely to cause variation of rates in the younger age groups, among whom injury is less common and decisions about tracheotomies are more likely to be made on a nonemergent basis.
Children who received tracheotomies in the Northeast and Midwest were discharged to long-term care facilities at significantly higher rates than those in the West. Having adjusted for available proxies of disease severity (prolonged mechanical ventilation status, number of diagnoses, and presence of injury), this finding is less likely to represent a difference in the actual need for institutional care. Rather, regional differences in the availability of long-term care facilities and home health care may have contributed to this finding. Historically, more pediatric long-term care facilities were developed and maintained in the Northeast and mid-Atlantic states than in the West (e-mail communication, John M. Neff, MD, January 4, 2002), and this legacy may have affected providers' beliefs and practice habits regarding appropriate disposition for patients with tracheotomies. Children cared for in teaching institutions or children's hospitals or units also had lower rates of discharge to long-term care facilities. The reason underlying this finding is uncertain, as we are unaware of any study of patient or health care system characteristics that predict where pediatric patients with tracheotomies are likely to be discharged (one published study focused on adults, noting that younger patients were more likely to be discharged to home24). We postulate, though, that teaching institutions and children's hospitals place a greater emphasis on or have more resources for discharge coordination and planning, thus allowing more children to be discharged to their homes.
Children who received their tracheotomy in children's hospitals had twice the rates of survival to discharge when compared with children who received their tracheotomy in non–children's hospitals. In our analysis we adjusted for potential confounders that could affect mortality risk. For example, we adjusted for transfer status, since it was possible that some children may have been transferred to a children's hospital after they had been medically stabilized in another facility, thus potentially lowering the mortality risk in children's hospitals, since more stable patients would survive to be transferred. We did not find an effect of transfer status on mortality. Furthermore, one might expect that patients in children's hospitals would likely have more serious and complicated conditions, and therefore have a higher expected mortality rate after tracheotomy. If so, our observation that adjusted mortality is halved in children's hospitals is especially compelling.
Independent of the children's hospital effect, larger case volume of pediatric tracheotomies per hospital was associated with significantly lower mortality rates. Similar inverse volume-mortality effects have been documented in other settings, such as pediatric cardiac surgery, pediatric intensive care, and liver transplantation.25- 27 Previous investigators have attributed the volume-mortality relationship to greater clinical experience of physicians and staff at higher-volume institutions.
We undertook this study to better describe the population of children undergoing tracheotomy and examine differences in rates of this procedure and its outcomes, specifically disposition locale and mortality, across regions of the country and across different hospital types. Our findings have several implications. First, regional variation of tracheotomy rates among the younger age groups, and in particular the infants, warrants further investigation. We cannot, within the confines of this study, determine what would be the appropriate rate; rather, the variation observed across regions suggests that factors aside from the health status of individual patients may be affecting the decision to perform a tracheotomy, and this effect may or may not be in the patients' best interest. Second, the observation that mortality risk varies by hospital types and by the volume of pediatric tracheotomies performed at a given hospital raises the possibility that the quality of care, and subsequently the outcomes, could be improved in some hospitals. Finally, the differences noted regarding where the children and adolescents were sent after discharge from the hospital underscore the importance not only of long-term follow-up studies to understand the implications of various disposition arrangements but also of viewing patients with complex chronic conditions as being served by a system of health and social supportive services that extend far beyond the walls of hospitals.
Corresponding author: Charlotte W. Lewis, MD, MPH, Child Health Institute, University of Washington, Box 354920, Seattle, WA 98195-4920 (e-mail: email@example.com).
Accepted for publication September 24, 2002.
Dr Lewis is supported by grant K23 DE14062-02 from the National Institute of Dental and Craniofacial Research, Bethesda, Md. Dr Feudtner is supported by grant K08 HS0000-02 from the Agency for Healthcare Research and Quality, Rockville, Md.
We thank Fred Rivara, MD, MPH, for his advice and guidance and Ann Nykamp for her administrative support.