The Medical Costs of Gunshot Injuries in the United States FREE

In 1997, gunshots caused 31,636 fatal injuries and approximately 100,000 nonfatal injuries in the United States.^1- 2 In addition to the enormous human toll of gun violence, the cost of treating these injuries imposes a financial burden on society. While measuring medical costs is not as straightforward as counting the number of victims, valid cost estimates are important for at least 2 reasons. First, such estimates are relevant to evaluating gun violence–reduction programs. Second, reliable estimates for the financial burden that gun violence imposes on the medical care system may help guide reimbursement policies.

We present new estimates for the aggregate medical costs of gunshot injuries in the United States. Compiling such estimates requires a variety of data sets and a number of assumptions. The result is necessarily imperfect, but our estimates improve on previous published estimates in several ways. We used the most up-to-date and comprehensive sources available, including some that have not been used for this purpose before. In particular, our sample of hospitalized gunshot injuries is more current and more than twice as large as that used in previous studies,^3- 5 and our sample of cases treated in the emergency department (ED) is a vast improvement over the sample used in a previous estimate (800 cases vs 11 cases).^4- 5 We paid particular attention to the small proportion of serious gunshot injuries that account for a disproportionate share of lifetime medical costs by developing new estimates for gunshot-related spinal cord injuries using data from the National Spinal Cord Injury Statistical Center (NSCISC).⁶ We focused as much as possible on medical costs rather than payments or charges.⁷ Finally, how these costs are distributed may be at least as important as their overall magnitude for informing policy makers. Our study presents what we believe are the first nationally representative estimates for the sources of payment for medical costs that are specific to gunshot injuries.

In our approach, the national medical cost for each category of gunshot injuries is calculated as the product of 2 estimated magnitudes, the number of gunshot cases and the average cost per case in that category. As a rule, we were able to quantify the uncertainty about the number of cases, but had no way to quantify the uncertainty about the average cost. The primary sources of uncertainty with the average-cost estimates stem from questions about data representativeness rather than sampling variability. As a result, we present 95% confidence intervals (CIs) for the number of gunshot cases but not for average cost per case.

Table 1 outlines the key data sources that we used to estimate medical costs. Because the available data sources are focused on different levels of treatment, we organized our estimates according to the highest level of medical treatment the patient received—hospitalized, ED only, and "other." This last category includes untreated cases, both fatal and nonfatal, and nonfatal cases treated in physicians' offices. Our estimates start with average acute-care costs: we used data from New York and Maryland for hospitalized cases and from South Carolina for ED-only cases. Lifetime follow-up costs were calculated using a number of national data sources. The lifetime cost per case of each sort was then multiplied by the estimate for the number of gunshot cases of that sort.

Cases were classified in several ways, including (1) injury outcome (fatal or nonfatal); (2) the highest level of medical treatment received (hospitalized, ED only, or other); (3) victim's sex and age; and (4) diagnosis and external cause-of-injury codes from the International Classification of Diseases, Ninth Revision (ICD-9). The ICD-9 diagnosis codes allowed us to distinguish cases according to body part injured. The ICD-9 external cause-of-injury codes (E codes) identify the injury intent (self-inflicted, unintentional, assault), whether the injury was caused by a firearm, and, if so, the gun type (handgun, shotgun, rifle). Our incidence estimates were calculated separately for each of our disaggregated injury groups. Total national costs were thus defined as the sum across injury groups of average costs times incidence. One advantage of our method is that it allows us to more confidently project national costs because this procedure controls for the possibility that the gunshot case mix in the 3 states used to estimate acute-care costs may not be nationally representative.

In all of our calculations, we followed the recommendation of the Panel on Cost-Effectiveness in Health and Medicine, convened by the US Public Health Service in 1993, and converted lifetime costs into a present value using a 3% discount rate.⁸ Unless otherwise noted, we converted all medical costs into 1994 prices using the Consumer Price Index for Medical Care.⁹ We chose 1994 as our reference year because most of our medical cost data come from this period.

We calculated the incidence of fatal gunshot injuries that occurred in the United States in 1994 from the 1994 Vital Statistics census of deaths.¹⁰ The 1994 Vital Statistics multiple-cause-of-death file includes ICD-9 diagnosis codes and E codes, the victim's sex and age at death, and whether the victim died as a hospital inpatient, an ED patient (which under our definition includes dead-on-arrival cases and those for whom hospitalization status was unknown), or outside of the ED or hospital (such as dead at the scene).

Estimates for the number of nonfatal gunshot injuries come from the National Electronic Injury Surveillance System (NEISS), a national probability sample of hospital EDs.² Case studies suggest that NEISS coders record 92% to 97% of gunshot injuries treated in sampled EDs.^11- 12 While NEISS uses a system different than the ICD-9 to provide injury characteristics, we have developed a system that matches NEISS injury codes with ICD-9 diagnosis and E codes. We adjusted the NEISS estimates for the fact that some gunshot victims receive emergency medical treatment outside of hospital EDs (authors' unpublished data, 1999).

The detailed injury and victim data available from the Vital Statistics and NEISS allow us to estimate the national incidence of fatal and nonfatal gunshot injuries within each of the disaggregated injury groups described above.

With slight variations, the lifetime cost per hospitalized gunshot survivor was computed using the formulas in equations 1 and 2:

(1) Short-term=(Hospital Costs) × (1+ Fee Ratio) ×
(Readmission) × (1+Postdischarge);
and
2) LC=(Short-term) × (Long-term) × (1 + Claims).

In the equations, Short-term indicates costs during the first 6 months following the injury; Hospital Costs, hospital costs for initial hospitalization; Fee Ratio, ratio of inpatient professional fees to hospital costs;Readmission, average number of hospital admissions per gunshot injury; Postdischarge, ratio of costs in the first 6 months after discharge to acute inpatient care costs; LC, lifetime medical cost per gunshot survivor; Long-term, ratio of total lifetime costs to Short-term; and Claims, ratio of claims administration costs to total lifetime medical costs. Hospital costs for gunshot injuries are based on a census of hospital discharges in Maryland in 1994-1995 (2852 cases) and New York in 1994 (3835 cases). Around 7% of victims in each state were deceased at discharge. We focused on hospital discharge data from Maryland and New York because these states require that hospitals provide E codes for all injury-related hospital discharges. More importantly, regulatory agencies in these states ascertained costs of care by hospital service, and, as a result, the charges reported in these hospital-discharge data sets can be converted to reasonably accurate estimates of actual hospital costs.

The ratio of professional fees to hospital costs (Fee Ratio) was calculated at the diagnosis-code level using the 1992-1994 Civilian Health and Medical Program of the Uniformed Services (CHAMPUS) data, which are based on payments (including co-payment) in an insured population of 2 million military dependents and non-Medicare retirees.¹³ While CHAMPUS is the best available data source for estimating professional fees, the lack of E codes meant that we had to calculate professional fee ratios by diagnosis regardless of whether the injury was by gunshot or other means.

The ratio of ancillary and follow-up costs in the first 6 months to all acute inpatient care costs (Postdischarge) was calculated from the 1987 National Medical Expenditure Survey (NMES).^4- 5 These follow-up costs include emergency transport, prescriptions, medical supplies (such as crutches), home health care, and follow-up physician visits (but not follow-up hospitalizations). Because NMES contains only 397 total hospital-admitted injuries, we could not produce estimates for Postdischarge that are specific to individual ICD-9 diagnosis or E codes. Instead, we used NMES data to estimate that Postdischarge is equal to 11.8% for all injuries, regardless of the body part injured or method of injury.

Some gunshot victims will have repeat hospitalizations in the short term that are not captured by Postdischarge. To adjust for this, we used Missouri hospital-discharge data from 1994 that include individual patient identification numbers to estimate the average number of hospital admissions for each gunshot victim during the first year following the injury (Readmission). We multiplied the Maryland and New York estimates of costs per admission by our Missouri estimate of total first-year hospital admissions per victim to obtain an estimate of hospital costs per gunshot injury case.

Except for catastrophic spinal cord injuries and traumatic brain injuries, we estimated long-term follow-up medical costs by calculating for each ICD-9 diagnosis code the fraction of lifetime medical payments for treating an injury that is incurred during the first 6 months and then converting it to a ratio (Long-term). Our estimate for Long-term comes from the National Council on Compensation Insurance's Detailed Claims Information (DCI) data set, which provides a sample of 452,000 injury cases (including 138,000 hospitalized cases) that occurred from 1979 through 1987.¹³ The advantage of the DCI data set is that it is one of the few sources of lifetime medical costs. The disadvantages of the DCI are that it provides information only at the diagnosis level and not at the E-code level, only includes injuries that occur at work, is somewhat dated, and in some states excludes injuries that involve fewer than 3 to 9 days of work loss.

As an example of how these components combine to form average costs per case, consider an unintentional gunshot injury to the abdomen with initial hospitalization costs (Hospital Costs) equal to $10,000 and professional payments equal to 30% of the hospitalization costs (Fee Ratio=0.3). Suppose that on average patients with such injuries are hospitalized 1.1 times during the first year (Readmission=1.1), that other follow-up costs during the first 6 months equal 70% of hospitalization costs (Postdischarge=0.7), and that the DCI data suggest that lifetime medical costs for injuries within this ICD-9 diagnosis code are twice the total costs incurred during the first 6 months (Long-term=2.0). Lifetime costs then equal ($10,000 × 1.3 × 1.1 × 1.7 × 2.0)=$48,620.

The calculation is slightly different for traumatic brain injuries, for which we made some additional adjustment for lifetime custodial-care costs that are unlikely to be fully captured by the Long-term factor. The Bureau of the Census¹⁴ reports an annual cost of $84,285 (inflated from 1993 to 1994 dollars using the consumer price index–all items) for custodial care in a public facility for individuals with mental retardation. We followed previous studies¹⁵ in assuming that traumatic brain injury care in an intensive care facility costs twice the Census Bureau's custodial-care figure.

We produced separate estimates for the lifetime medical costs for spinal cord injuries because these relatively rare catastrophic cases account for a disproportionately large share of the overall costs of treating gun injuries. We estimated lifetime medical charges using data from the NSCISC on 820 gunshot-related spinal cord injuries treated between 1988 and 1992 in 24 model spinal cord injury treatment centers.⁶ We converted estimates for inpatient hospital charges into costs using Medicare cost-to-charge ratios for the spinal cord injury hospitals.

Finally, we controlled for the possibility that Maryland and New York medical prices may not be nationally representative by deflating our cost estimates using the Health Care Financing Administration's Hospital Wage Index values for October 1, 1994.¹⁶ This hospital wage index should closely approximate overall variation in medical costs in light of a recent case study suggesting that 80% of total hospital operating expenses come from employee costs.¹⁷ We also multiplied lifetime medical costs by 1.03 or 1.04 (depending on the patient's primary payment source) to reflect claims administration costs.^4- 5

We estimated medical charges for gunshot survivors admitted to the ED but not admitted to the hospital by means of a census of ED discharges from South Carolina for 1997. The South Carolina data set provides information on 796 ED-only gunshot injuries and includes those who were dead on arrival at the ED. While these data are not from our reference year (1994) and have charge rather than cost data, they are the best available data on this subject. We estimated follow-up medical costs using the ancillary and follow-up factors described in equation 1.

No data set provides reliable information on nonfatal gunshot injuries each year in which the victim does not seek medical treatment at the ED or hospital. Since these cases are likely to account for only a very small share of total costs, we ignored them. For fatal non-ED gunshot injuries (the victim dies on the spot) we assumed that the medical costs equal the cost of emergency transport to the medical examiner, equal to $175,⁵ plus the costs of the medical examiner's or coroner's examination, equal to $443.¹⁸

The primary hospital discharge data set used to calculate these estimates is the Agency for Health Care Policy and Research's Healthcare Cost and Utilization Project (HCUP) for 1994,¹⁹ which includes an E-coded sample of discharges from California, Connecticut, Massachusetts, Maryland, New Jersey, New York, Washington, and Wisconsin. (We did not use the HCUP discharge data to estimate costs themselves because only the Maryland and New York data provide information on costs rather than charges). For nonfatal cases we also used the 226 gunshot cases from the National Hospital Discharge Survey (NHDS) for 1996,²⁰ a nationally representative sample of discharges from 500 hospitals with 63% of cases E coded. For fatal hospitalized injuries, we replicated the HCUP estimates using the National Mortality Followback Survey (NMFS),²¹ a nationally representative sample of US decedents in 1993 that includes 2764 gunshot victims.

Estimated payment sources for ED-only cases come from the 1997 ED discharge data from South Carolina. We also estimated the distribution of primary payers for ED-only cases using pooled data from the 1992-1996 National Hospital Ambulatory Medical Care Survey (NHAMCS).²² The NHAMCS provides information on primary payment source, but not on total costs or charges. The 5 years of pooled NHAMCS data contain 129 gunshot cases.

The average costs for acute-care treatment ($14,757 vs $14,497) and total lifetime costs ($36,685 vs $34,420) of persons hospitalized with nonfatal gunshot injuries in Maryland and New York are remarkably similar, as seen in Table 2. In both Maryland and New York, nonfatal self-inflicted gunshot injuries have higher lifetime costs than unintentional injuries or assaults.

As is shown in Table 2, the medical costs of fatal gunshot injuries in which the individual was hospitalized $12,691 in Maryland and $12,191 in New York, calculated as the sum of acute care costs and 6-month ancillary and follow-up costs are, on average, less than half those for nonfatal cases. The difference between fatal and nonfatal hospitalized gunshot injury cases highlights the magnitude of follow-up treatment costs. For the nonfatal gunshot injury cases shown in Table 2, the majority of medical treatment costs come after the patient has been discharged from the hospital (54% of costs in Maryland and 58% in New York).

As seen in Table 3, the lifetime charges associated with ED-only gunshot injuries in South Carolina average $1321 for nonfatal cases and $2394 for fatal cases.

Table 4 presents information on the primary expected payment source for gunshot injuries by severity of the injury. Government programs are the primary payers for 40% to 50% of hospitalized gunshot injury cases. For hospital-admitted survivors, the average of the 1994 HCUP and 1996 NHDS estimates suggest that government is the primary payer in 44% of cases. Government is the primary payer in a smaller number of less-expensive injuries—only 7% for fatal cases from the South Carolina ED data and averaging 18% for nonfatal cases between the South Carolina ED data and the 1992-1996 NHAMCS data. On the other hand, government is the primary payer for a greater share of costs of more expensive gunshot injuries. For example, the NSCISC data suggest that government programs are the primary payers of acute-care costs for 62.5% of spinal cord injuries due to gunshots and 88.6% of spinal injury cases after initial hospitalization.

Private insurance is the primary payer for 18% of hospitalized and ED-only cases. Costs that are classified as "self pay" may increase charges for other patients because some of these victims will never pay their medical bills. These payer distributions are roughly similar for assault, unintentional, and self-inflicted injuries (authors' unpublished data, 1999).

While these estimates show the distribution of primary payers for gunshot cases, the bottom half of Table 4 suggests that primary payers are responsible for the large majority of total costs, at least for serious injuries. Data from the NSCISC and NMFS suggest that the distribution of all sources of payment is quite similar to the distribution of primary payers. Presumably, the same pattern holds for less-serious injuries.

We estimated that there were a total of 134,445 gunshot injuries in the United States in 1994 (95% CI, 109,465-159,425), as shown in the first row of Table 5. The estimates for nonfatal gunshot injuries come from the 1994 NEISS data and, as noted above, exclude gunshot injury patients who do not seek professional medical treatment. Of the 95,860 nonfatal gunshot cases (95% CI, 70,880-120,840), most occurred as the result of an assault. Our figure for the number of fatal gunshot injuries, 38,585, comes from the Vital Statistics census of deaths. For both fatal and nonfatal cases, we assume the same distribution across intent categories (unintentional, self-inflicted, and assaults) for cases in which intent is unknown as for cases in which intent is known.

We estimated that gunshot injuries in the United States in 1994 will cost $2.3 billion in lifetime medical costs (95% CI, $2.1 billion-$2.5 billion) or, on average, about $17,000 per injury ($2.3 billion per 134,445 injuries). Of these costs, 74% are accounted for by injury due to assaults.

Government programs pay about half of the total lifetime costs of treating gunshot injuries, with private insurance and victims covering another fifth each. Our estimates for the payment sources for gunshot injuries, shown in the bottom row of Table 5, were calculated by a weighted average of the fraction of each component of the total costs of gunshot injuries paid for by government, private insurance, victims, and other sources.

Our study suggests that the lifetime costs of treating all US gunshot injuries in 1994 was $2.3 billion. Of these costs, we estimated that $1.1 billion was paid by government. Our data leave some uncertainty about who pays for the $0.4 billion that the records indicate is to be paid by the victims themselves, although data from several Massachusetts hospitals suggest that self-pay cases are nearly 6 times as likely as others to become bad debts and ultimately be covered by other sources.²³

In Table 6, our results are compared with those found in previous research. We focus on hospitalized gunshot survivors because these are the injuries examined in most previous studies and because our estimates suggest that these cases account for 85% of the total lifetime medical costs of treating all gunshot injuries. As seen in Table 6, previous estimates for the acute-care costs for hospitalized gunshot injury survivors range from $11,023 to $21,324 (all in 1994 dollars). Part of the discrepancy between our results and those in previous studies is that we estimated costs while 5^25- 29 of the 9 previous studies estimate charges. Furthermore, most studies were limited to data from trauma centers rather than all hospitals. In any event, our estimates for acute-care treatment costs fall in the middle of the range of previous research.

As seen in Table 6, only 2 previous studies^3- 5 estimated the lifetime costs of medical treatment. Our preferred estimate of $35,367 per nonfatal hospitalized gunshot injury is approximately 14% higher than 1 estimate of $30,969, and 2.3 times the estimate of $15,602³ (all in 1994 prices). These differences are probably accounted for by variations in the data sources used for estimating medical costs and our different treatment of the small minority of catastrophic cases that account for a large share of total costs.

Taken together, the estimates presented in this study substantially improve the quality of information on the medical costs of gunshot injuries in the United States. Nevertheless, these findings are subject to several qualifications stemming from limitations of the available data. Our estimates for the national costs of hospital treatment come from data from only 2 states. Unfortunately, no other states both require hospitals to E code discharge records and provide information that can identify actual medical costs. Our estimates for follow-up medical costs come from NMES and DCI data that do not allow us to specifically identify costs for gunshots distinct from other types of injuries. The DCI data are further restricted to workplace injuries. Finally, our E-coded ED-discharge data from South Carolina provide information on charges, not costs.

Our study's most important potential limitation concerns the lifetime medical costs for treating gunshot injuries resulting in permanent disability, which account for a large share of total medical costs. While we developed separate estimates for 1 type of catastrophic case that may result from gunshots—spinal cord injuries—there are other types of costly disabilities that may also be caused by gunshot injuries. We estimated the lifetime medical costs for these other injuries using somewhat dated NMES and DCI data that did not allow us to distinguish gunshot injuries from other injuries within the same ICD-9 diagnosis code. Our estimates will understate the actual lifetime costs of treating gunshot injuries if gunshots are more likely than other injuries with the same diagnosis to result in long-term disability. Our attempts to develop separate lifetime-cost estimates for other types of disabilities were hampered by limited charge, payment, or cost information for long-term care from available data sources. Improving surveillance and cost data for gunshot injuries resulting in long-term disability is the highest priority in any effort to further refine the estimates presented here.

The estimated medical costs of treating the gunshot injuries received during 1994 in the United States was $2.3 billion. The average medical cost of a gunshot injury was approximately $17,000, of which 49% was borne by taxpayers, 18% by private insurance, and 33% by other sources. While medical costs are a relatively small component of the total burden imposed on society by gun violence,^3- 5 they represent a substantial cost to the medical care system.