Although injuries due to falls from buildings and other structures represent a small proportion of injuries to children and adolescents, many of these injuries are preventable through simple environmental modifications.^1–4 Falls from height, originally documented as a problem in the northeastern US, particularly New York City,^5–7 are now recognized as a cause of serious pediatric trauma on several continents.^{8,9,10,11,12,13,14,15} Falls from buildings, the majority of which are reported to involve a fall out of a window, may result in death or serious lifetime cognitive and physical disabilities.^8,16–18 Traumatic brain injury, spinal cord injuries, major fractures, and psychological trauma are not uncommon.^19,20 These injuries have been reported to cluster in resource limited urban neighborhoods where parental educational and income levels are frequently below the national average.^1,5,16

Environmental interventions aimed at children were implemented in the cities of New York and Boston, resulting in a decrease in injuries due to this mechanism.^1,3,4 Injury reductions of up to 96% were reported after introduction of prevention strategies that included multifaceted interventional efforts with parental education, legislation requiring mandatory window guards in residences and common space of children aged 10 years and younger, and enforcement of that legislation.^1,21 In this study, we examine demographic risk factors and characteristics of injuries as a result of falls from buildings and structures (which includes window falls) across the pediatric and adolescent age span, first in a national population hospitalized for this cause in the US, and then in New York City where legislated and enforced environmental modifications are in effect for children 10 years and younger.

METHODS

Data source(s)

The Kids Inpatient Database (KID-HCUP), provided the national sample of statewide acute care hospital discharges from 27 states for calendar year 2000.^22,23 KID-HCUP contains demographic, socioeconomic (median income for zip code of residence), diagnostic, treatment, and hospital outcome data on admissions. The cross sectional data set contains unweighted hospital discharge data on 2 516 833 pediatric admissions to community, non-rehabilitation hospitals (patients aged 19 years and less) and 7 291 032 weighted discharges. We excluded people aged 19 years of age because not all states reported on people older than 18 years. In addition, 61 routine, elective, or scheduled admissions for falls from buildings or structures were excluded as unlikely incident injury. For KID-HCUP, two states did not have data on month of discharge (Florida and Connecticut) and six states reported all races as other/unknown (Georgia, Kentucky, Maine, Oregon, Washington, and West Virginia). These states are excluded from incidence calculations by race and ethnicity, but appear in totals not stratified by race and ethnicity.

The New York Statewide Planning and Research Cooperative System (SPARCS), provided data at the zip code level for New York State residents hospitalized for the year 2001. Unlike KID-HCUP, SPARCS data, obtained from the New York State Department of Health, enabled us to examine the annual injury incidence specifically for New York City, which has a widely enforced window guard law. Approximately 40% of the children living in New York State reside in one of the five boroughs of New York City. The US census, conducted on the entire population decennially in the US, provided data on the population at risk by age, sex, and race/ethnicity and was used for the denominators in incidence calculations. The census collected number of units in housing structures using a categorical variable ranging from 1 to 50 or more. We used residence in a structure with 10 or more units as a proxy for exposure to multistorey dwellings.²³

We defined acute injury to include urgent or emergent hospital admissions or admission through an emergency department resulting from falls from buildings or structures. We used external cause of injury (E-codes) from ICD-9-CM as well as classifications derived from the Clinical Classification Software (CCS) included as variables in the KID-HCUP data.²⁴ ICD-9-CM code E882.0, used to identify falls from buildings or structures, excludes people who jump or fell from a building or structure to escape fire. In KID-HCUP, income is provided as a categorical variable at the neighborhood level using median income (in US dollars) for the zip code of patient residence. The definition of preadult disorders includes ICD-9-CM based codes for disturbances of emotions and conduct specific to or originating in childhood and adolescence.²⁴

Statistical analyses

Cumulative incidence (per 100 000 persons per year) was calculated for falls from buildings and structures for the US using weighted KID-HCUP (numerators), for New York City/New York State using SPARCS data (numerators), and the US census 2000 (denominators).²³ For analyses specific to New York City, we grouped the five boroughs of New York City (Bronx, Brooklyn, Kings, New York, and Richmond Counties), all of which are covered by window guard legislation. Multivariable logistic regression was used to identify independent factors and odds ratios for factors associated with falls from buildings.²⁵ The “cases” for logistic regression were injuries resulting from falls from buildings/structures and “non-cases” were children hospitalized for other reasons. The χ² test was used in univariate analyses of categorical variables with statistical significance defined as less than or equal to 0.05. Means are presented with standard deviations or with 95% confidence intervals. We used SAS 9.0 for processing the large databases.²⁶

RESULTS

Of the 2 163 402 people age 18 years and younger who were discharged from a US hospital during the year 2000, 1161 discharges were classified as acute injury due to falls from buildings and structures. Of these, 70 were classified as intentional falls or jumps and 1091 discharges were categorized as unintentional incident injury due to falls from buildings or structures.

Intentional self inflicted injury

Of the 70 children and adolescents classified as having an intentional fall or jump, 27 (38.6%) were whites, 10 (14.3%) were blacks, 11 (15.7%) were Hispanics, and 22 (31.4%) were of other or unknown race or ethnicity. The age distribution of persons attempting suicide included 50 (71.4%) people aged 15–18 years, and 16 (22.9%) aged 10–13 years. The majority were in households earning incomes of $35,000 or more (55.2%). Comorbid mental health or alcohol or substance abuse diagnoses were reported for 57 (81.4%) of those attempting suicide by this mechanism. Mental illness diagnoses included affective disorder in 16 (22.9%), substance abuse in 12 (17.1%), and other conditions in 44 (62.8%). Of the 31 attempted suicides where place of injury was reported, nearly half (45.2%) occurred at home and two thirds occurred during weekdays. Injuries categorized as attempted suicide are excluded from further analysis.

Unintentional injury

There were race/ethnic differences in the age distribution, income levels and place of injury for the 1091 admissions for unintentional falls from buildings (table 1). Among blacks and Hispanics, nearly half of hospitalizations occurred in the youngest age group, compared with one third in whites (table 1). Being male was associated with increased incidence of injury accounting for more than two thirds of all falls from buildings and structures. Falls among blacks clustered in communities with zip codes below the national median while most falls among whites were associated with incomes higher than the national median (table 1). Nearly 90% of very young children aged 0–4 years for whom place of injury was reported fell at home (fig 2). The number of falls at home showed a linear decrease with age.

Race and ethnic differences in acute injury incidence due to falls from buildings

The cumulative incidence of injury due to falls from buildings is shown by age and race/ethnicity in table 2. Among the total population, children aged 0–4 years had a higher incidence than older children, although the pattern tended to be U-shaped across the age groups examined. Incidence among minority children aged 0–4 years was approximately double that of similarly aged white children. Twelve patients (1.1%) who lived to be admitted to the hospital died before discharge. Inhospital mortality did not differ by race/ethnicity.

Incident injury in interventional v non-interventional areas

New York City had a higher proportion of the population residing in multifamily dwellings with 10 or more units compared to the nation (53.8% v 12.6%, p<0.0001), but the incidence of injury from falls from buildings was nearly half that observed in the US. Of the 30 falls involving children residing in New York City, approximately half occurred outside the age range covered by window guard legislation. Although incidence for the State of New York was lower than the national average (2.47 v 2.81 per 100 000 people), this was due to lower incidence in New York City as incidence rates in areas of the state not covered by legislation were slightly higher than the national average (fig 3). Boston, Massachusetts and Chicago, Illinois implemented non-legislative based intervention programmes. Although KID-HCUP did not contain data for Illinois nor separate data for Boston, unintentional falls from buildings and structures for people aged 0–18 years in the State of Massachusetts tended to be higher than for New York State (3.00 v 2.47 per 100 000) and the US (3.00 v 2.81 per 100 000) (fig 3).

Types of unintentional injury

Types of injury sustained during falls from buildings were markedly different among the age groups (table 3). Multiple trauma was more common among teenagers aged 15–18 years than among young children and adolescents (p<0.001). Very young children had high incidence of intracranial injuries and fractures of the skull and face, but had relatively infrequent spinal cord injury and fracture of the extremities in comparison with older children. Upper extremity fractures were less frequent in very young children but doubled or tripled in older children and adolescents. Lower extremity fractures were relatively uncommon in very young children, but increased with age, quadrupling in adolescents aged 15–18 years compared with children aged 0–4 years. Fewer blacks had diagnoses of fracture than whites or Hispanics (p = 0.002). Mental health diagnoses were associated with higher incidence of spinal cord injury (p = 0.0001), and tended to be higher in those experiencing multiple trauma (p = 0.057).

Seasonal unintentional injuries

Injury due to falls from buildings showed seasonal gradients that varied across race/ethnic and age groups. In the total population, injuries began to increase slightly in early spring and began to decline in September (fig 4A). Injuries were lowest between the cold weather months of November and February (fig 4A). Seasonal patterns in very young children showed low frequency of injury in the winter (fig 4B). In contrast, older children and teenagers showed a more flattened seasonal distribution (fig 4C). There were notable race/ethnic differences, with Hispanics showing a tendency toward an extended “season of risk”. In logistic models stratified by race/ethnicity for children aged 0–9 years, increased odds ratios were observed for an extended season of “warm weather” months that began in April and extended through September.

Psychiatric, alcohol, and substance abuse diagnoses in those with unintentional injury

Race/ethnic, socioeconomic, and age differences were noted for mental health, alcohol abuse, and substance abuse diagnoses. Mental health diagnoses were uncommon in very young age groups, but increased with age. Mental health diagnoses, including substance abuse, were highest in older adolescents aged 15–18 years (p<0.0001). Concomitant mental health or substance abuse diagnoses occurred twice as frequently in whites than in blacks or Hispanics (p = 0.015). Alcohol abuse diagnoses were an independent predictor of injury due to unintentional falls from buildings only in white adolescents (OR = 2.50, 95% CI 1.60 to 3.90).

Race/ethnicity and socioeconomic status

The distribution of socioeconomic status as measured using median household income for the patient’s zip code of residence is shown in table 1. In an attempt to disentangle socioeconomic status and race/ethnicity, separate models were developed for whites, blacks, and Hispanics. Among black children and adolescents, 35.5% of whom lived in households earning 150% of the poverty level or less, a model of independent predictors of falls from buildings showed lower neighborhood income to be a significant risk factor only for the lowest income group in reference to the highest (p = 0.038). However, in separate models for whites and Hispanics, where only 3% and 10%, respectively, lived at this income level, the neighborhood income results were mixed.

Cost of hospital care associated with falls from buildings

Median per person hospital charges for injuries from falls from buildings or structures were $8968 (mean $17,020 (SD $28,576)) (in year 2000 US dollars). Hospital charges were similar across age groups except in the oldest group where there was an increased tendency toward multiple trauma. Median charges for people with multiple trauma versus without were $20,312 (mean $40,388 (SD $54,393)) and $8485 (mean $15,253 (SD $24,760)), respectively. Length of hospital stay varied similarly, with hospital stay for those with versus without multiple trauma experiencing longer hospital stays (6.9 (SD 6.7) days v 3.3 (SD 5.6) days).

DISCUSSION

The risk of unintentional falls from buildings was bimodal, with the highest incidence in very young children aged 0–4 years and in teenagers. Among minority children, both Hispanics and blacks were at greater risk for injury from falls from buildings than white children. Risk factors varied significantly over the age span, with alcohol abuse/psychiatric diagnoses rising in importance in white adolescents and teenagers.

Consistently strong peaks in numbers of injuries have been reported to occur during warm weather months, generally reflecting increased falls from windows left open for home temperature regulation.^12,19 Our findings mirror those of other studies when all ages are considered, however, significant differences emerge when age groups are examined separately. Our observation showed age related risks with higher warm weather risks for very young children and higher year round risk in adolescents. Seasonal differences were most pronounced in very young children, but flattened with increasing age, reflecting an increased tendency toward year round injury among 15–18 year olds. This pattern was also associated with a gradual shift in place of occurrence and risk factors for injury from falls from buildings. Changes in the age profile are probably multifactorial with older adolescent injury likely reflecting increased social influences, mental illness, and substance abuse. Although falls from buildings in younger versus older children likely represent a different sequelae of events requiring a different approach to effective prevention, the presence of window guards may deter both.

Our study found high psychiatric illness in intentional falls and a high proportion of substance or alcohol abuse in white teenagers experiencing unintentional falls. These findings are consistent with the findings of Gill, whose toxicological investigation of fatal falls from height in New York City documented similarly high incidence of use of psychiatric medication in suicide attempts and found that 36% of “accidental” fatal injuries due to falls from height were associated with use of alcohol or illicit substance.²⁷ Although it appears likely that our unintentional falls may include some undiagnosed suicide attempts, unintentional injury is also a plausible explanation for falls secondary to impaired judgment associated with alcohol and drug abuse. Among suicide attempts by jumping from a building or other structure, nearly half occurred at home suggesting the possibility that window guards might also deter or divert home suicides through increased obstacles to use of this mechanism.

A window guard law, with annual enforcement, has been effective in New York City where exposure to high rises associated with multifamily dwellings is higher than the national average, but where incidence of injury due to falls from buildings is far lower. Despite declines in the early seventies, falls from height continued to be reported as a problem in New York in 1975 with one report noting as many as 159 falls and 19 deaths due to falls.^28,29 In 2001, 30 children were hospitalized as a result of falls from buildings and structures in New York City, but there were no reported window fall deaths and many falls were among older adolescents not covered by the law.^30,31 Although declines were noted following introduction of legislation, steeper declines were noted with enforcement of that legislation which included legal action with criminal charges against landlords failing to comply with the window guard law and with continued efforts to educate parents and consumers through a programme called “Children Can’t Fly”.^1,3,6 Although prevention efforts have produced improvements in the absence of well designed legislation with concomitant enforcement, such areas report that nearly 40% of injuries resulting from falls from heights continue to be from window falls. In one state where educational campaigns were introduced in their largest city without mandatory legislation, we observed state level hospitalization incidence of falls from buildings and structures to be similar or higher than the national average although the city reports lower injuries from windows falls.⁴ We did not have information that allowed us to differentiate injury as a result of window falls from falls from buildings or to identify areas of this state reached by educational campaigns.

This study had limitations. Most importantly, the ICD-9-CM code used to identify falls from buildings and structures includes non-window falls from buildings, such as falls from fire escapes. Although the data set was designed for such purposes, we made national estimates based on data from 27 states. Bias may have been introduced if the 27 states were not representative of the US population. For New York State, incidence produced using KID-HCUP and SPARCS were virtually identical. We were unable to include in our incidence calculations injuries in people dying before hospital admission, not seeking treatment, failing to receive the appropriate E-code, or those treated and released from an emergency department. Furthermore, our data include zip code for residence and not place of injury for those not injured at home. Although it is likely that falls in very young children occurring in the home were the result of window falls, among older children this is less likely as fewer than half of the falls in older adolescents occurred at home, leaving us with inadequate information with which to design or assess interventional strategies for this group. ICD-9-CMs do not provide sufficient information to enable one to tailor interventions across the adolescent age span. There is a need to improve documentation and completeness of injury coding in medical records and administrative data sets.

In summary, in New York City, where window guard legislation is enforced and exposure to multifamily dwellings is higher, injury from falls from buildings and structures is half that observed nationally. The finding that warm weather risks begin earlier in the spring and extend later in the fall season than previously reported has important prevention implications for both developing and developed societies. Window guards have potential to provide affordable, effective prevention of injuries in developing societies where the population is disproportionately young and living in warm climates, and where high rise and multiple storey buildings are being used increasingly to alleviate overcrowding.