Abstract

Background

The number of uninsured children in the USA is increasing while the impact on children's health of being uninsured remains largely uncharacterized. We analyzed data from more than 23 million US children to evaluate the effect of insurance status on the outcome of US pediatric hospitalization.

Methods

In our analysis of two well-known large inpatient databases, we classified patients less than 18 years old as uninsured (self-pay) or insured (including Medicaid or private insurance). We adjusted for gender, race, age, geographic region, hospital type, admission source using regression models. In-hospital death was the primary outcome and secondary outcomes were hospital length of stay and total hospital charges adjusted to 2007 dollars.

Results

The crude in-hospital mortality was 0.75% for uninsured versus 0.47% for insured children, with adjusted mortality rates of 0.74 and 0.46%, respectively. On multivariate analysis, uninsured compared with insured patients had an increased mortality risk (odds ratio: 1.60, 95% CI: 1.45–1.76). The excess mortality in uninsured children in the US was 37.8%, or 16 787, of the 38 649 deaths over the 18 period of the study.

Conclusion

Children who were hospitalized without insurance have significantly increased all-cause in-hospital mortality as compared with children who present with insurance.

Introduction

Within the USA, the number of uninsured children has been increasing in recent years, both in absolute number and in proportion to the population. In 2006, an additional 1.4 million children were added to the ranks of the uninsured, bringing the total to 9.4 million, or 12.1% of all US children.1 This increase has been attributed to declines in employer coverage without commensurate increases in coverage provided through Medicaid or the State Children's Health Insurance Program (SCHIP).2,3 Uninsured adults4 are less likely to receive preventative care,5,6 are more severely ill when diagnosed,7,8 and have worse outcomes.7–12 In pediatric patients, insurance status has been shown to reduce unmet healthcare needs,13 improve access to health care,14,15 and improve utilization of health care.16 Furthermore, pediatric data from the National Inpatient Sample (NIS) has demonstrated an association between increased mortality and pediatric patients without insurance.17 Nevertheless, the global impact of insurance status on inpatient pediatric mortality remains largely uncharacterized.

Given the substantial policy debates regarding expanding SCHIP, a better understanding of the risks and benefits of providing improved insurance coverage to children is needed. The specific aim of this study was to use a collection of administrative databases in the USA from 1988 to 2005 to characterize the impact of insurance status on inpatient mortality and costs of care.

Methods

Databases

The analysis was based on a non-overlapping combination of 15 years of the NIS (1988–1996, 1998–1999, 2001–2002, 2004–2005) and 3 years of the Kids' Inpatient Database (KID) database (1997, 2000, 2003). Both databases are produced by the Healthcare Cost and Utilization Project (HCUP) of the Agency for Healthcare Research and Quality. The NIS is an all-payer database that annually contains information from up to 8 million inpatient discharges from approximately 1000 hospitals across the USA and represents a 20% stratified sample of all hospitals from 37 states.18

The KIDs’ database contains a sample of pediatric discharges from all the community, non-rehabilitation hospitals in states, which participate in the HCUP. Unlike the NIS, with hospital-level sampling, the KID samples patient discharges are then weighted to obtain national estimates. The algorithm involves systematic random sampling to select 10% of uncomplicated in-hospital births and 80% of complicated in-hospital births as well as other selected pediatric cases. The KID contains information from up to 36 states.19 Information collected on patients in both databases includes age at admission, gender, race, diagnosis and procedure information, insurance status, admission source, total hospital charges, disposition, length of stay (LOS) and descriptive hospital information (state, urbanicity, teaching status and bed size). Geographic regions were defined by US Census standards. Hospitals in metropolitan counties were considered urban, while hospitals in micropolitan or non-core areas were classified as rural. A hospital is considered to be a teaching hospital by the American Hospital Association if it has an American Medical Association-approved residency program, is a member of the Council of Teaching Hospitals or has a ratio of full-time equivalent interns and residents to beds of 0.25 or higher.20 Admission sources of the non-emergency department category included transfers from another acute care hospital, long-term care facility, outpatient clinics or correctional facilities.

Inclusion criterion for the analysis was an age at admission of less than 18 years. The primary-dependent variable was inpatient mortality while secondary-dependent variables were hospital LOS and total hospital charges in 2007 dollars using the medical care consumer price index.21 The primary exposure variable was insurance status. Patients were divided into uninsured and insured groups based on the recorded primary payer. Self-pay patients were defined as uninsured while individuals whose expected primary payer was Medicare, Medicaid or private insurance (including Blue Cross, commercial carriers, private HMOs and PPOs) were classified as insured. Unknown insurance status patients were excluded from comparative analyses.

Data management and analysis

A descriptive analysis was performed using variables of gender, race, inpatient mortality, geographic region (Northeast, Midwest, South, West), type of hospital (urban teaching, urban non-teaching and rural), and admission source (emergency department, non-emergency department). Summary statistics for LOS and inflation-adjusted total hospital charges were calculated.21 National estimates were produced by using the discharge weights to extrapolate from the NIS or KID data sets to the discharges from all US non-rehabilitation hospitals.

Unadjusted comparisons between groups were performed using Pearson's chi-square for categorical variables or the Wilcoxon Rank Sum test for non-normally distributed continuous variables. Multiple logistic regression was utilized to evaluate the impact of insurance status on inpatient mortality. To evaluate the impact of insurance status on LOS and charges, we used linear regression without transformation to facilitate interpretation. An interaction term between race and insurance among patients with available data was included in the regression analysis. Subset analysis was performed in patients aged 1 or older to eliminate the potential influence from inadequate prenatal care. Additional subset analyses were performed in subgroups across the entire spectrum of ICD-9 diagnosis codes, divided into subcategories by major disease groupings similar to that used by the National Hospital Discharge Survey (NHDS), the largest and longest running survey of hospital utilization in the USA.22 Simulation was performed by applying the adjusted death rates among insured hospitalizations, as calculated based on the overall multivariate model, to uninsured hospitalizations. An alternative simulation analysis was performed where uninsured patients were grouped with Medicaid patients for comparison with insured patients. Stata/Multi-Processor, version 10 (College Station, TX), was used for the analysis.

The study was exempted by the Johns Hopkins Institutional Review Board.

Results

Demographic data of uninsured and insured patients

Data from 23 535 491 pediatric inpatient hospitalizations from 37 US states were analyzed. Of these, 5.4% (n = 1 262 452) were uninsured and 94.6% (n = 22 273 039) were insured (Table 1). Although statistically different, there was similar proportion of female patients in both the uninsured and insured groups (48.90 versus 48.67%, P < 0.001). Analysis of available race information demonstrated that Black (17.37 versus 15.72%) and Hispanic (26.38 versus 19.65%, P < 0.001) patients were more likely to be uninsured versus White patients (45.86 versus 56.82%, P < 0.001). Race data were not available for approximately 30% of identified pediatric patients as reporting is limited by hospitals and states.23 Similar proportions of uninsured children were under age 1 (66.83 versus 66.62%, P < 0.001) and 11–18 (16.56 versus 15.24%, P < 0.001).

Table 1

Demographics of study population and hospital characteristics

Characteristic Uninsured, N = 1 262 452 (%) Insured, N = 22 273 039 (%) P-value 
Gender    
 Male 643 733 (50.99) 11 400 733 (51.19) <0.001 
 Female 617 355 (48.90) 10 839 366 (48.67)  
 Unknown 1364 (0.11) 32 940 (0.15)  
Racea    
 White 413 047 (45.86) 8 912 760 (56.82) <0.001 
 Black 156 470 (17.37) 2 465 233 (15.72)  
 Hispanic 237 584 (26.38) 3 081 338 (19.65)  
 Other 93 576 (10.39) 1 225 742 (7.81)  
Age (in years)    
 <1 843 646 (66.83) 14 838 719 (66.62) <0.001 
 1–5 137 059 (10.86) 2 664 839 (11.96)  
 6–10 72 715 (5.76) 1 374 012 (6.17)  
 11–18 209 032 (16.56) 3 395 469 (15.24)  
Geographic region    
 Northeast 304 636 (24.13) 4 570 735 (20.52) <0.001 
 Midwest 298 308 (23.63) 4 776 434 (21.44)  
 South 396 722 (31.42) 7 179 995 (32.24)  
 West 262 785 (20.82) 5 745 874 (25.80)  
 Unknown ≤10 ≤10  
Type of hospital    
 Urban, teaching 600 339 (47.55) 10 638 658 (47.76) <0.001 
 Urban, non-teaching 484 827 (38.40) 9 025 059 (40.52)  
 Rural 170 956 (13.54) 2 543 238 (11.42)  
 Unknown 6 330 (0.50) 66 084 (0.30)  
Admission source    
 Emergency department 289 411 (22.92) 4 056 350 (18.21) <0.001 
 Non-emergency department 877 28 (69.49) 16 863 218 (75.71)  
 Unknown 95 755 (7.58) 1 353 471 (6.08)  
Characteristic Uninsured, N = 1 262 452 (%) Insured, N = 22 273 039 (%) P-value 
Gender    
 Male 643 733 (50.99) 11 400 733 (51.19) <0.001 
 Female 617 355 (48.90) 10 839 366 (48.67)  
 Unknown 1364 (0.11) 32 940 (0.15)  
Racea    
 White 413 047 (45.86) 8 912 760 (56.82) <0.001 
 Black 156 470 (17.37) 2 465 233 (15.72)  
 Hispanic 237 584 (26.38) 3 081 338 (19.65)  
 Other 93 576 (10.39) 1 225 742 (7.81)  
Age (in years)    
 <1 843 646 (66.83) 14 838 719 (66.62) <0.001 
 1–5 137 059 (10.86) 2 664 839 (11.96)  
 6–10 72 715 (5.76) 1 374 012 (6.17)  
 11–18 209 032 (16.56) 3 395 469 (15.24)  
Geographic region    
 Northeast 304 636 (24.13) 4 570 735 (20.52) <0.001 
 Midwest 298 308 (23.63) 4 776 434 (21.44)  
 South 396 722 (31.42) 7 179 995 (32.24)  
 West 262 785 (20.82) 5 745 874 (25.80)  
 Unknown ≤10 ≤10  
Type of hospital    
 Urban, teaching 600 339 (47.55) 10 638 658 (47.76) <0.001 
 Urban, non-teaching 484 827 (38.40) 9 025 059 (40.52)  
 Rural 170 956 (13.54) 2 543 238 (11.42)  
 Unknown 6 330 (0.50) 66 084 (0.30)  
Admission source    
 Emergency department 289 411 (22.92) 4 056 350 (18.21) <0.001 
 Non-emergency department 877 28 (69.49) 16 863 218 (75.71)  
 Unknown 95 755 (7.58) 1 353 471 (6.08)  

aNot all states collected information on race (29.58% unknown in insured and 28.66% in uninsured).

Hospital characteristics of uninsured and insured pediatric patients

A higher percentage of uninsured pediatric patients were seen in hospitals located in the Northeast (24.13 versus 20.52%, P < 0.001) and Midwest (23.63 versus 21.44%, P < 0.001) while a lower percentage were seen in states in the South (31.42 versus 32.24%) and West (20.82 versus 25.80%, P < 0.001) (Table 1). Pediatric patients from both groups were almost equally likely to be treated at urban teaching hospitals (47.55 versus 47.76%, P < 0.001) and urban non-teaching hospitals (38.40 versus 40.52%, P < 0.001). Uninsured pediatric patients compared with insured were more likely to present through the Emergency department (22.92 versus 18.21%, P < 0.001).

Inpatient hospitalization outcomes of uninsured and insured pediatric patients

Median LOS was 2 days (interquartile range: 1–3) for both the uninsured and insured groups (Table 2). Among patients who died during the admission, the uninsured pediatric patient had a shorter median LOS of 0 days (interquartile range: 0–2) versus insured patients with a median LOS of 1 day (interquartile range: 0–9). In contrast, among patients who survived their hospitalization, there was a smaller difference in LOS, with the uninsured having a median LOS of 2 days (interquartile range: 1–3) and the insured patients having a similar a median LOS of 2 days (interquartile range: 2–3).

Table 2

In-Hospital mortality, LOS and total hospital charges for uninsured and insured patients

Outcome Uninsured, N = 1 262 452 (%) Insured, N = 22 273 039 (%) P-value 
In-hospital mortality 
 Total death 9468 (0.75) 104 520 (0.47) <0.001 
Gender 
  Male 5546 (0.86) 58 320 (0.51) <0.001 
  Female 3908 (0.63) 46 080 (0.43) 
 Race 
  White 2539 (0.61) 35 039 (0.39) <0.001 
  Black 1911 (1.22) 17 102 (0.69) 
  Hispanic 1409 (0.59) 14 133 (0.46) 
  Others 661 (0.71) 6981 (0.57) 
Age (in years) 
  <1 7370 (0.87) 78 722 (0.53) <0.001 
  1–5 604 (0.44) 9778 (0.37) 
  6–10 286 (0.39) 4655 (0.34) 
  11–18 1208 (0.58) 11 365 (0.33) 
LOS (days) 
 Median (IQR) 2 (1–3) 2 (1–3) <0.001 
  Death 0 (0–2) 1 (0–9) 
  Alive 2 (1–3) 2 (2–3) 
Total hospital charges ($)a 
 Mean (no.) 9388 (1 218 077) 12 637 (21 374 281) <0.001 
 Median (IQR) 3034 (1419–7637) 3349 (1551–8852) 
  Death 8058 (1283–31 119) 20 951 (3527–83 958) 
  Alive 3024 (1420–7570) 3333 (1549–8778) 
Outcome Uninsured, N = 1 262 452 (%) Insured, N = 22 273 039 (%) P-value 
In-hospital mortality 
 Total death 9468 (0.75) 104 520 (0.47) <0.001 
Gender 
  Male 5546 (0.86) 58 320 (0.51) <0.001 
  Female 3908 (0.63) 46 080 (0.43) 
 Race 
  White 2539 (0.61) 35 039 (0.39) <0.001 
  Black 1911 (1.22) 17 102 (0.69) 
  Hispanic 1409 (0.59) 14 133 (0.46) 
  Others 661 (0.71) 6981 (0.57) 
Age (in years) 
  <1 7370 (0.87) 78 722 (0.53) <0.001 
  1–5 604 (0.44) 9778 (0.37) 
  6–10 286 (0.39) 4655 (0.34) 
  11–18 1208 (0.58) 11 365 (0.33) 
LOS (days) 
 Median (IQR) 2 (1–3) 2 (1–3) <0.001 
  Death 0 (0–2) 1 (0–9) 
  Alive 2 (1–3) 2 (2–3) 
Total hospital charges ($)a 
 Mean (no.) 9388 (1 218 077) 12 637 (21 374 281) <0.001 
 Median (IQR) 3034 (1419–7637) 3349 (1551–8852) 
  Death 8058 (1283–31 119) 20 951 (3527–83 958) 
  Alive 3024 (1420–7570) 3333 (1549–8778) 

IQR, interquartile range.

aAdjusted for medical care inflation to 2007 dollars.

Mean and median total hospital charges were lower in the uninsured ($9388, $3034) versus insured group ($12 637, $43 349). Among pediatric patients who died during their hospitalization, insurance status was significantly associated with total hospital charges, as median total hospital charges for uninsured patients who died were substantially lower than those who died and were insured ($8058 versus $20 951). Disparity remained between median total hospital charges of uninsured and insured children who survived their hospitalizations but was not as large ($3024 versus $3333).

In-hospital crude mortality was higher for uninsured children than insured (0.75 versus 0.47%). Crude mortality rates for uninsured were higher in every patient subgroup analyzed by gender, race, age, hospital region, or type of hospital (Table 3).

Table 3

Adjusted analysis for factors associated with inpatient mortality, 1988–2005

Mortality OR P-value 95% CI
 
Insurance status 
 Insured Ref.    
 Uninsured 1.60 <0.001 1.45 1.76 
Gender 
 Male Ref.    
 Female 0.85 <0.001 0.83 0.86 
Race 
 White Ref.    
 Black 1.50 <0.001 1.40 1.61 
 Hispanic 0.96 0.260 0.88 1.03 
 Others 1.19 0.001 1.07 1.31 
Age (in years) 
 <1 Ref.    
 1–5 0.62 <0.001 0.58 0.66 
 6–10 0.57 <0.001 0.54 0.61 
 11–18 0.63 <0.001 0.60 0.67 
Hospital region 
 Northeast Ref.    
 Midwest 1.25 0.046 1.00 1.55 
 South 1.31 <0.001 1.13 1.52 
 West 1.42 <0.001 1.20 1.68 
Type of hospital 
 Urban, teaching Ref.    
 Urban, non-teaching 0.37 <0.001 0.33 0.42 
 Rural 0.22 <0.001 0.18 0.27 
Calendar year     
 1988–92 Ref.    
 1993–97 0.91 0.221 0.79 1.05 
 1998–2002 0.87 0.088 0.74 1.02 
 2002–05 0.92 0.297 0.78 1.08 
Mortality OR P-value 95% CI
 
Insurance status 
 Insured Ref.    
 Uninsured 1.60 <0.001 1.45 1.76 
Gender 
 Male Ref.    
 Female 0.85 <0.001 0.83 0.86 
Race 
 White Ref.    
 Black 1.50 <0.001 1.40 1.61 
 Hispanic 0.96 0.260 0.88 1.03 
 Others 1.19 0.001 1.07 1.31 
Age (in years) 
 <1 Ref.    
 1–5 0.62 <0.001 0.58 0.66 
 6–10 0.57 <0.001 0.54 0.61 
 11–18 0.63 <0.001 0.60 0.67 
Hospital region 
 Northeast Ref.    
 Midwest 1.25 0.046 1.00 1.55 
 South 1.31 <0.001 1.13 1.52 
 West 1.42 <0.001 1.20 1.68 
Type of hospital 
 Urban, teaching Ref.    
 Urban, non-teaching 0.37 <0.001 0.33 0.42 
 Rural 0.22 <0.001 0.18 0.27 
Calendar year     
 1988–92 Ref.    
 1993–97 0.91 0.221 0.79 1.05 
 1998–2002 0.87 0.088 0.74 1.02 
 2002–05 0.92 0.297 0.78 1.08 

Adjusted analysis of factors affecting mortality

Multivariate analysis of factors that contribute to mortality is presented in Table 3. Children hospitalized while being uninsured had a 1.60 (95% CI: 1.45–1.76) times higher likelihood of death. Among the entire cohort of 23 million children, girls had decreased mortality compared with boys (odds ratio (OR): 0.85, 95% CI: 0.83–0.86). Black children had increased mortality (OR: 1.50 versus reference White, 95% CI: 1.40–1.61). Increasing age was protective against mortality, with OR of 0.62, 0.57, 0.63 for age groups 1–5, 6–10, 11–18, respectively, versus age less than 1.

Hospitals in the Midwest, South, and West demonstrated a higher mortality as compared with the Northeast (OR 1.25, 1.31, 1.42). Rural and non-teaching urban hospitals were protective (OR: 0.22 and 0.37 versus urban teaching hospitals). Analysis of mortality trends showed improvement in outcomes over time with OR of 0.91, 0.87, 0.92, for calendar years 1993–1997, 1998–2002, 2002–2005, respectively, as compared with 1988–1992. In subset analysis of patients aged 1 or older, mortality did not change significantly, with OR of 1.50 (95% CI: 1.36–1.64) for uninsured versus insured hospitalization. Interaction terms between race and insurance were found to be non-significant.

Additional multivariate analysis revealed that uninsured hospitalizations have 0.6 days shorter lengths of stay and $2342 lower total hospital charges. Adjusted mortality rates calculated based on the multivariate analysis were 0.74% for uninsured patients versus 0.46% for insured patients, corresponding to an excess mortality of 37.8% (Fig. 1).

Fig. 1

Adjusted mortality rates calculated based on the multivariate analysis for uninsured versus insured patients.

Fig. 1

Adjusted mortality rates calculated based on the multivariate analysis for uninsured versus insured patients.

Subset analysis of factors affecting mortality by first-listed diagnosis

Subset analysis for groups of patients separated by first-listed diagnosis is presented in Table 4. The largest categories of the first-listed diagnosis were complications relating to birth, pneumonia and asthma, a distribution that was similar to data from the NHDS. There was no quantitative difference across categories, in almost all groups uninsured patients were significantly more likely to die than uninsured patients.

Table 4

Number of discharges, deaths, and adjusted OR of death by first-listed diagnosis

Category of first-listed diagnosis ICD-9 code Total number Number of deaths Death rate (%) ORa 95% CI of OR
 
All conditions  23 535 491 113 988 0.48 1.60 1.45 1.76 
Infectious and parasitic diseases 001–139 573 686 1472 0.26 1.00 0.79 1.26 
 Septicemia 038 76 477 1432 1.87 0.76 0.54 1.07 
Neoplasms 140–239 9676 146 1.51 1.09 0.40 3.01 
 Malignant neoplasms 140–208, 230–234 83 084 3041 3.66 0.91 0.72 1.15 
  Malignant neoplasm of large intestine and rectum 153–154, 197.5 326 23 7.06 0.75 0.04 14.60 
  Malignant neoplasm of trachea, bronchus, and lung 162, 176.4, 197.0, 197.3 1,922 125 6.50 0.75 0.27 2.08 
 Benign neoplasms 210–229 34 144 51 0.15 NC NC NC 
  Benign neoplasm of uterus 218–219 93 0.00 NC NC NC 
Endocrine, nutritional and metabolic diseases, and immunity disorders 240–279 74 094 710 0.96 0.91 0.54 1.52 
 Diabetes mellitus 250 110 577 132 0.12 1.38 0.61 3.13 
 Volume depletion 276.5 369 598 263 0.07 1.03 0.53 2.01 
Diseases of the blood and blood-forming organs 280–289 104 386 214 0.21 0.48 0.16 1.43 
 Anemias 280–285 120 400 287 0.24 1.55 0.86 2.82 
Mental disorders 290–319 299 874 48 0.02 0.54 0.07 4.13 
 Psychoses 290–299 98 431 10 0.01 NC NC NC 
  Schizophrenic disorders 295 12 489 ≤10 0.02 NC NC NC 
  Major depressive disorder 296.2–296.3 147 526 12 0.01 3.46 0.32 37.81 
Diseases of the nervous system and sense organs 320–389 306 924 2328 0.76 1.44 1.14 1.82 
Diseases of the circulatory system 390–459 34 201 156 0.46 1.59 0.72 3.53 
 Essential hypertension 401 3027 ≤10 0.17 7.59 0.84 68.78 
 Heart disease 391, 392.0, 393–398, 402, 404, 410–416, 420–429 22 798 816 3.58 1.40 0.93 2.10 
  Acute myocardial infarction 410 481 66 13.72 1.04 0.18 6.14 
  Coronary atherosclerosis 414.0 155 ≤10 1.94 NC NC NC 
  Other ischemic heart disease 411–413, 414.1–414.9 455 ≤10 1.98 NC NC NC 
  Cardiac dysrhythmias 427 22 064 1045 4.74 2.22 1.60 3.09 
  Congestive heart failure 428.0, 428.2–428.4 9,023 579 6.42 1.13 0.68 1.88 
 Cerebrovascular disease 430–438 10 884 789 7.25 1.52 1.02 2.28 
 Diseases of the respiratory system 460–519 548 554 3357 0.61 0.76 0.60 0.97 
  Acute bronchitis and bronchiolitis 466 560 320 232 0.04 0.40 0.13 1.22 
  Pneumonia 480–486 727 209 1937 0.27 0.49 0.34 0.69 
  Chronic bronchitis 491 3775 20 0.53 2.08 0.23 18.50 
  Asthma 493 681 283 204 0.03 0.91 0.45 1.84 
 Diseases of the digestive system 520–579 289 048 1012 0.35 1.10 0.78 1.56 
  Appendicitis 540–543 311 970 39 0.01 1.01 0.24 4.31 
  Non-infectious enteritis and colitis 555–558 236 772 220 0.09 0.70 0.27 1.84 
  Intestinal obstruction 560 44 240 191 0.43 0.72 0.27 1.91 
  Diverticula of intestine 562 417 ≤10 0.48 NC NC NC 
  Cholelithiasis 574 19 075 14 0.07 NC NC NC 
  Acute pancreatitis 577.0 12 490 49 0.39 NC NC NC 
 Diseases of the genitourinary system 580–629 254 547 291 0.11 0.73 0.33 1.65 
  Calculus of kidney and ureter 592 13 255 ≤10 0.01 NC NC NC 
  Urinary tract infection 599.0 101 664 62 0.06 NC NC NC 
 Complications of pregnancy, childbirth, and the puerperium 630–677 854 127 79 0.01 3.69 1.70 7.99 
 Diseases of the skin and subcutaneous tissue 680–709 57 401 48 0.08 0.46 0.07 3.20 
  Cellulitis and abscess 681–682 142 111 26 0.02 NC NC NC 
 Diseases of the musculoskeletal system and connective tissue 710–739 196 643 209 0.11 1.12 0.50 2.50 
  Osteoarthrosis and allied disorders 715 432 0.00 NC NC NC 
  Intervertebral disc disorders 722 4117 ≤10 0.02 NC NC NC 
 Congenital anomalies 740–759 405 786 7854 1.94 1.55 1.33 1.79 
 Certain conditions originating in the perinatal period 760–779 565 037 12 702 2.25 1.64 1.46 1.85 
 Symptoms, signs, and ill-defined conditions 780–799 607 388 1376 0.23 1.15 0.83 1.58 
 Injury and poisoning 800–999 460 322 7337 1.59 1.77 1.58 1.98 
  Fractures, all sites 800–829 395 786 3099 0.78 1.84 1.57 2.15 
  Fracture of neck of femur 820 11 259 52 0.46 NC NC NC 
  Poisonings 960–989 145 161 355 0.24 1.81 1.26 2.59 
 Certain complications of surgical and medical care 996–999 172 449 1090 0.63 0.62 0.35 1.08 
 Supplementary classifications V01–V84 13 175 628 58 193 0.44 1.78 1.59 1.98 
  Females with deliveries V27 122 0.00 NC NC NC 
Category of first-listed diagnosis ICD-9 code Total number Number of deaths Death rate (%) ORa 95% CI of OR
 
All conditions  23 535 491 113 988 0.48 1.60 1.45 1.76 
Infectious and parasitic diseases 001–139 573 686 1472 0.26 1.00 0.79 1.26 
 Septicemia 038 76 477 1432 1.87 0.76 0.54 1.07 
Neoplasms 140–239 9676 146 1.51 1.09 0.40 3.01 
 Malignant neoplasms 140–208, 230–234 83 084 3041 3.66 0.91 0.72 1.15 
  Malignant neoplasm of large intestine and rectum 153–154, 197.5 326 23 7.06 0.75 0.04 14.60 
  Malignant neoplasm of trachea, bronchus, and lung 162, 176.4, 197.0, 197.3 1,922 125 6.50 0.75 0.27 2.08 
 Benign neoplasms 210–229 34 144 51 0.15 NC NC NC 
  Benign neoplasm of uterus 218–219 93 0.00 NC NC NC 
Endocrine, nutritional and metabolic diseases, and immunity disorders 240–279 74 094 710 0.96 0.91 0.54 1.52 
 Diabetes mellitus 250 110 577 132 0.12 1.38 0.61 3.13 
 Volume depletion 276.5 369 598 263 0.07 1.03 0.53 2.01 
Diseases of the blood and blood-forming organs 280–289 104 386 214 0.21 0.48 0.16 1.43 
 Anemias 280–285 120 400 287 0.24 1.55 0.86 2.82 
Mental disorders 290–319 299 874 48 0.02 0.54 0.07 4.13 
 Psychoses 290–299 98 431 10 0.01 NC NC NC 
  Schizophrenic disorders 295 12 489 ≤10 0.02 NC NC NC 
  Major depressive disorder 296.2–296.3 147 526 12 0.01 3.46 0.32 37.81 
Diseases of the nervous system and sense organs 320–389 306 924 2328 0.76 1.44 1.14 1.82 
Diseases of the circulatory system 390–459 34 201 156 0.46 1.59 0.72 3.53 
 Essential hypertension 401 3027 ≤10 0.17 7.59 0.84 68.78 
 Heart disease 391, 392.0, 393–398, 402, 404, 410–416, 420–429 22 798 816 3.58 1.40 0.93 2.10 
  Acute myocardial infarction 410 481 66 13.72 1.04 0.18 6.14 
  Coronary atherosclerosis 414.0 155 ≤10 1.94 NC NC NC 
  Other ischemic heart disease 411–413, 414.1–414.9 455 ≤10 1.98 NC NC NC 
  Cardiac dysrhythmias 427 22 064 1045 4.74 2.22 1.60 3.09 
  Congestive heart failure 428.0, 428.2–428.4 9,023 579 6.42 1.13 0.68 1.88 
 Cerebrovascular disease 430–438 10 884 789 7.25 1.52 1.02 2.28 
 Diseases of the respiratory system 460–519 548 554 3357 0.61 0.76 0.60 0.97 
  Acute bronchitis and bronchiolitis 466 560 320 232 0.04 0.40 0.13 1.22 
  Pneumonia 480–486 727 209 1937 0.27 0.49 0.34 0.69 
  Chronic bronchitis 491 3775 20 0.53 2.08 0.23 18.50 
  Asthma 493 681 283 204 0.03 0.91 0.45 1.84 
 Diseases of the digestive system 520–579 289 048 1012 0.35 1.10 0.78 1.56 
  Appendicitis 540–543 311 970 39 0.01 1.01 0.24 4.31 
  Non-infectious enteritis and colitis 555–558 236 772 220 0.09 0.70 0.27 1.84 
  Intestinal obstruction 560 44 240 191 0.43 0.72 0.27 1.91 
  Diverticula of intestine 562 417 ≤10 0.48 NC NC NC 
  Cholelithiasis 574 19 075 14 0.07 NC NC NC 
  Acute pancreatitis 577.0 12 490 49 0.39 NC NC NC 
 Diseases of the genitourinary system 580–629 254 547 291 0.11 0.73 0.33 1.65 
  Calculus of kidney and ureter 592 13 255 ≤10 0.01 NC NC NC 
  Urinary tract infection 599.0 101 664 62 0.06 NC NC NC 
 Complications of pregnancy, childbirth, and the puerperium 630–677 854 127 79 0.01 3.69 1.70 7.99 
 Diseases of the skin and subcutaneous tissue 680–709 57 401 48 0.08 0.46 0.07 3.20 
  Cellulitis and abscess 681–682 142 111 26 0.02 NC NC NC 
 Diseases of the musculoskeletal system and connective tissue 710–739 196 643 209 0.11 1.12 0.50 2.50 
  Osteoarthrosis and allied disorders 715 432 0.00 NC NC NC 
  Intervertebral disc disorders 722 4117 ≤10 0.02 NC NC NC 
 Congenital anomalies 740–759 405 786 7854 1.94 1.55 1.33 1.79 
 Certain conditions originating in the perinatal period 760–779 565 037 12 702 2.25 1.64 1.46 1.85 
 Symptoms, signs, and ill-defined conditions 780–799 607 388 1376 0.23 1.15 0.83 1.58 
 Injury and poisoning 800–999 460 322 7337 1.59 1.77 1.58 1.98 
  Fractures, all sites 800–829 395 786 3099 0.78 1.84 1.57 2.15 
  Fracture of neck of femur 820 11 259 52 0.46 NC NC NC 
  Poisonings 960–989 145 161 355 0.24 1.81 1.26 2.59 
 Certain complications of surgical and medical care 996–999 172 449 1090 0.63 0.62 0.35 1.08 
 Supplementary classifications V01–V84 13 175 628 58 193 0.44 1.78 1.59 1.98 
  Females with deliveries V27 122 0.00 NC NC NC 

NC, not calculable.

aOR (uninsured versus insured) adjusted for gender, race, age category, hospital region, type of hospital and calendar year interval.

Another subset analysis was performed by race, and similarly confirmed the deleterious effects of uninsurance in all racial groups (data not shown).

Preventable death projections

Applying the adjusted mortality rate among insured hospitalizations (0.46%) to the uninsured hospitalizations, our simulation estimated that 3535, or 37.8%, of the deaths among uninsured hospitalizations might have been prevented by insurance coverage, assuming no other confounding factors (Fig. 1). Extrapolating data to the entire US population using the weights provided in the NIS database, the total number of hospitalizations for the time period of the study in the US was 117 million (Fig. 2). Of these, 5 995 395 were hospitalized with no insurance at admission and 38 649 of this group went on to die during that same hospitalization. Provision of insurance may potentially have allowed 16,787 children to survive over the 18 years period of the study according to the simulation.

Fig. 2

Data showing number of hospitalizations with regard to uninsured patients.

Fig. 2

Data showing number of hospitalizations with regard to uninsured patients.

Preventable death projections: alternative classification of the uninsured

Since Medicaid patients are heterogeneous with a proportion maintaining long-term coverage while others might gain retroactive coverage upon discharge, we decided to perform the analysis utilizing an alternative classification where patients who received insurance retroactively were classified into the uninsured group. Using this alternative categorization, the adjusted mortality rate among the insured was 0.44% as compared with 0.52% among the uninsured/Medicaid population. Of the 241 209 deaths that occurred in the uninsured/Medicaid population, the simulation suggests that 39 323 deaths may have been prevented by providing insurance. This sensitivity analysis confirmed that our model does not change qualitatively regardless of whether Medicaid patients were classified as insured or uninsured.

Discussion

Main finding of this study

Analyzing data from 23 million admissions over 18 consecutive years, our study demonstrates that US children who were hospitalized without insurance had significantly increased all-cause mortality as compared with children with insurance. The adjusted mortality of uninsured children was approximately 60% higher than insured children. Extrapolating data to national figures, we estimate that 16 787 deaths of 5 995 395 uninsured hospitalizations might have been prevented over this time period assuming lack of insurance was the driving factor.

Although the overall LOS of insured and uninsured children was the same, insured children who died were likely to be hospitalized longer by 1 day, which may be an indirect indicator of disease severity of the uninsured at presentation. Similarly, total hospital charges for insured children who died were substantially higher at $20 951 per patient as compared with uninsured children at $8058. This may be due to the higher disease severity of uninsured children, which precludes them from receiving additional care as they likely expired sooner than the group of insured children who also died in the hospital. This presumption of increased disease severity of uninsured children is partially supported by the noted difference in resources being expended on uninsured and insured children who live during their hospitalization which is much less.

While this type of study cannot establish a causal relationship between lack of insurance and increased mortality in children, the findings are consistent with similar associations found in adults.24–26 The lack of insurance may be a surrogate for socioeconomic status especially impacting pre-natal care. Similar to analysis of adult populations,4 parents of children without insurance may have limited access to health services and be more likely to use the Emergency department as their primary source of care,5,6 have more advanced disease when they present for care7,8 and have a higher rate of avoidable hospital stays.27,28 However, when we limited our analysis to patients 1 year and older, in an attempt to exclude the potential negative impact of poor pre-natal care, results were similar.

What is already known on this topic

While most developed countries provide universal health insurance, the US does not. Many health policy experts believe that the US should provide health care regardless of its ability to pay.29–33 The lack of insurance contributes to worse health outcomes and increased costs of care in adults.7–11,27,28,33,34 Our study provides evidence that the lack of insurance in children may have similar consequences, with a 60% increased risk of mortality.

What this study adds

We used comprehensive data that covers up to 37 states over the 18 year period, thus minimizing institutional and regional variations present in smaller datasets and case series. Additionally, our study included all pediatric patients admitted to a hospital rather than patients with a specific disease or having a specific procedure. Our methodology in categorizing uninsured versus insured patients is one of the more stringent of previous approaches. Only children under 18 years of age who had insurance information available in the database at the time of discharge were included in the study. Children labeled as ‘self-pay’ were placed into the uninsured category whereas patients with Medicaid or private insurance were placed in the insured category. However, many children labeled as having Medicaid insurance for the purposes of our study may have initially presented to the hospital without insurance but then received coverage retroactively after being enrolled in the SCHIP program. Using discharge data, we cannot separate children who are admitted with Medicaid versus those who received it retroactively through SCHIP. As such, patients with Medicaid are often labeled or grouped with the uninsured.25,35–37 Thus, we kept patients who carried the Medicaid designation under the insured category for the present analysis. In order to further address this issue of differing criteria of uninsured versus insured, we performed an alternative analysis categorizing both uninsured and Medicaid patients together to compare them against insured patients. This analysis demonstrated that 39 323 deaths may have been preventable due to uninsured patient status.

Limitations of this study

There are several limitations to our study. First, the present study design cannot establish a causal relationship that demonstrates being uninsured leads to higher mortality rates. Second, because we cannot link patients over time through these databases, the unit of analysis for this study is the hospitalization rather than the patient. Third, insurance status may be a marker for other factors that impact mortality such as socioeconomic status. This is especially true in evaluating infant mortality. However, as previously discussed, our findings were similar in children higher than 1 year of age. Fourth, we do not know how long patients had or did not have insurance prior to hospital admission, and this could affect patient access to preventative care which, in turn, could affect inpatient mortality when they are hospitalized. A patient could be labeled as insured even if he was uninsured for long periods of time before that hospitalization. However, it is equally likely for a patient to lose insurance just prior to hospitalization and thus be classified as uninsured. Therefore, the risk of misclassifying is equal among uninsured and insured, which would theoretically bias the study toward finding no (or less) association between insurance status and mortality.38 Fifth, since our data are limited to inpatient records, outcomes would be unknown for patients who never had care and died at home. Sixth, the databases do not contain adequate information to allow for adjustment for disease severity or co-morbid disease, which are predictors for mortality. Besides lack of information, there are no accepted methods to adjust for comorbid disease in children, unlike adults.39,40 However, the average number of secondary diagnosis codes, a crude method to evaluate comorbid disease, was similar between uninsured and insured groups (1.5 versus 1.3, P = 0.15).

In conclusion, children who were hospitalized without insurance have significantly increased all-cause in-hospital mortality as compared with children who present with insurance. Given these results and the increasing number of uninsured children, governmental and societal priorities should shift to ensure that the most vulnerable members of our society have health insurance.

Authors contributions

F.A. had full access to all the data in the study and takes responsibility of the data and the accuracy of the data analysis. F.A. and D.C.C. organized the study concept and design. F.A., D.C.C. and Y.Z. acquired, analyzed and interpreted the data. F.A., T.L., K.C., P.J.P. and P.M.C. drafted the manuscript. Y.Z. performed the statistical analysis. M.B. performed background literature research.

Funding

The authors would like to extend their appreciation to the Robert Garrett Fund for the Surgical Treatment of Children for support of this study.

Acknowledgements

The authors would like to extend their appreciation to Misty Burbank and Mindy Bathurst for assistance with preparation of the manuscript.

References

2
DeNavas-Walt
C
Proctor
BD
Hill Lee
C
U.S. Census Bureau, Current Population Reports, P60-229, Income, Poverty, and Health Insurance Coverage in the United States: 2004. U.S. Government Printing Office
 
Washington, DC, 2005
3
Holahan
John
Cook
Allison
Why did the number of uninsured continue to increase in 2005?
 , 
2006
 
Kaiser Commission on Medicaid and Uninsured
4
Hadley
J
Sicker and poorer–the consequences of being uninsured: a review of the research on the relationship between health insurance, medical care use, health, work, and income
Med Care Res Rev
 , 
2003
, vol. 
60
 (pg. 
3S
-
75S
)
5
Hadley
J
Insurance coverage, medical care use, and short-term health changes following an unintentional injury or the onset of a chronic condition
JAMA
 , 
2007
, vol. 
297
 (pg. 
1073
-
84
)
6
Schoen
C
DesRoches
C
Uninsured and unstably insured: the importance of continuous insurance coverage
Health Serv Res
 , 
2000
, vol. 
35
 (pg. 
187
-
206
)
7
Lee-Feldstein
A
Feldstein
PJ
Buchmueller
T
, et al.  . 
The relationship of HMOs, health insurance, and delivery systems to breast cancer outcomes
Med Care
 , 
2000
, vol. 
38
 (pg. 
705
-
18
)
8
Ayanian
JZ
Kohler
BA
Abe
T
, et al.  . 
The relation between health insurance coverage and clinical outcomes among women with breast cancer
N Engl J Med
 , 
1993
, vol. 
329
 (pg. 
326
-
31
)
9
Canto
JG
Zalenski
RJ
Ornato
JP
, et al.  . 
Use of emergency medical services in acute myocardial infarction and subsequent quality of care: observations from the National Registry of Myocardial Infarction 2
Circulation
 , 
2002
, vol. 
106
 (pg. 
3018
-
23
)
10
Young
GJ
Cohen
BB
Inequities in hospital care, the Massachusetts experience
Inquiry
 , 
1991
, vol. 
28
 (pg. 
255
-
62
)
11
Roetzheim
RG
Pal
N
Gonzalez
EC
, et al.  . 
Effects of health insurance and race on colorectal cancer treatments and outcomes
Am J Public Health
 , 
2000
, vol. 
90
 (pg. 
1746
-
54
)
12
Ayanian
JZ
Weissman
JS
Schneider
EC
, et al.  . 
Unmet health needs of uninsured adults in the United States
JAMA
 , 
2000
, vol. 
284
 (pg. 
2061
-
9
)
13
Feinberg
E
Swartz
K
Zaslavsky
A
, et al.  . 
Family income and the impact of a children's health insurance program on reported need for health services and unmet health need
Pediatrics
 , 
2002
, vol. 
109
 pg. 
E29
 
14
Slifkin
RT
Freeman
VA
Silberman
P
Effect of the North Carolina State Children's Health Insurance Program on Beneficiary Access to Care
Arch Pediatr Adolesc Med
 , 
2002
, vol. 
156
 (pg. 
1223
-
9
)
15
Szilagyi
PG
Dick
AW
Klein
JD
, et al.  . 
Improved access and quality of care after enrollment in the New York State Children's Health Insurance Program (SCHIP)
Pediatrics
 , 
2004
, vol. 
113
 (pg. 
e395
-
e404
)
16
Kempe
A
Beaty
BL
Crane
LA
, et al.  . 
Changes in access, utilization, and quality of care after enrollment into a state child health insurance plan
Pediatrics
 , 
2005
, vol. 
115
 (pg. 
364
-
71
)
17
Caskey
RN
Davis
MM
Differences associated with age, transfer status, and insurance coverage in end-of-life hospital care for children
J Hosp Med
 , 
2008
, vol. 
3
 (pg. 
376
-
83
)
18
HCUP Databases
Healthcare Cost and Utilization Project (HCUP)
 , 
2005
Agency for Healthcare Research and Quality
 
19
HCUP Kids' Inpatient Database (KID)
Healthcare Cost and Utilization Project (HCUP). Agency for Healthcare Research and Quality
 , 
2003
 
20
Health Forum (Corporate Author)
AHA Hospital Statistics
 , 
2007
American Hospital Association
 
1 Pap/Cdr edition ed
21
U.S.Bureau of Labor Statistics
Consumer Price Indexes (CPI)
2007
 
Division of Consumer Prices and Price Indexes
22
Kozak
LJ
DeFrances
CJ
Hall
MJ
National hospital discharge survey: 2004 annual summary with detailed diagnosis and procedure data
Vital Health Stat
 , 
2006
, vol. 
13
 (pg. 
1
-
209
)
23
Agency for Healthcare Research and Quality
Introduction to the HCUP Nationwide Inpatient Sample (KID)
 , 
2003
 
24
Franks
P
Clancy
CM
Gold
MR
, et al.  . 
Health insurance and subjective health status: data from the 1987 National Medical Expenditure survey
Am J Public Health
 , 
1993
, vol. 
83
 (pg. 
1295
-
9
)
25
Hadley
J
Steinberg
EP
Feder
J
Comparison of uninsured and privately insured hospital patients. Condition on admission, resource use, and outcome
JAMA
 , 
1991
, vol. 
265
 (pg. 
374
-
9
)
26
Sorlie
PD
Johnson
NJ
Backlund
E
, et al.  . 
Mortality in the uninsured compared with that in persons with public and private health insurance
Arch Intern Med
 , 
1994
, vol. 
154
 (pg. 
2409
-
16
)
27
Kozak
LJ
Hall
MJ
Owings
MF
Trends in avoidable hospitalizations, 1980–1998
Health Aff (Millwood)
 , 
2001
, vol. 
20
 (pg. 
225
-
32
)
28
Pappas
G
Hadden
WC
Kozak
LJ
, et al.  . 
Potentially avoidable hospitalizations: inequalities in rates between US socioeconomic groups
Am J Public Health
 , 
1997
, vol. 
87
 (pg. 
811
-
6
)
29
Herzlinger
RE
Parsa-Parsi
R
Consumer-driven health care: lessons from Switzerland
JAMA
 , 
2004
, vol. 
292
 (pg. 
1213
-
20
)
30
Luft
HS
Universal health care coverage: a potential hybrid solution
JAMA
 , 
2007
, vol. 
297
 (pg. 
1115
-
8
)
31
Woolhandler
S
Himmelstein
DU
Angell
M
, et al.  . 
Proposal of the physicians' working group for single-payer national health insurance
JAMA
 , 
2003
, vol. 
290
 (pg. 
798
-
805
)
32
Altman
SH
Doonan
M
Can Massachusetts lead the way in health care reform?
N Engl J Med
 , 
2006
, vol. 
354
 (pg. 
2093
-
5
)
33
Kuttner
R
Market-based failure—a second opinion on U.S. health care costs
N Engl J Med
 , 
2008
, vol. 
358
 (pg. 
549
-
51
)
34
Mitka
M
Forecast for US uninsured remains gloomy
JAMA
 , 
2004
, vol. 
291
 (pg. 
2305
-
6
)
35
Todd
J
Armon
C
Griggs
A
, et al.  . 
Increased rates of morbidity, mortality, and charges for hospitalized children with public or no health insurance as compared with children with private insurance in Colorado and the United States
Pediatrics
 , 
2006
, vol. 
118
 (pg. 
577
-
85
)
36
McWilliams
JM
Zaslavsky
AM
Meara
E
, et al.  . 
Health insurance coverage and mortality among the near-elderly
Health Aff (Millwood)
 , 
2004
, vol. 
23
 (pg. 
223
-
33
)
37
Durbin
DR
Giardino
AP
Shaw
KN
, et al.  . 
The effect of insurance status on likelihood of neonatal interhospital transfer
Pediatrics
 , 
1997
, vol. 
100
 pg. 
E8
 
38
Olson
LM
Tang
SF
Newacheck
PW
Children in the United States with discontinuous health insurance coverage
N Engl J Med
 , 
2005
, vol. 
353
 (pg. 
382
-
91
)
39
Romano
PS
Roos
LL
Jollis
JG
Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives
J Clin Epidemiol
 , 
1993
, vol. 
46
 (pg. 
1075
-
9
)
40
Needham
DM
Scales
DC
Laupacis
A
, et al.  . 
A systematic review of the Charlson comorbidity index using Canadian administrative databases: a perspective on risk adjustment in critical care research
J Crit Care
 , 
2005
, vol. 
20
 (pg. 
12
-
9
)