ORIGINAL ARTICLE

Morbidity and Mortality of Stevens-Johnson

Syndrome and Toxic Epidermal Necrolysis
in United States Adults
Derek Y. Hsu1, Joaquin Brieva1, Nanette B. Silverberg2 and Jonathan I. Silverberg3
Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are life-threatening disorders. Our study
objective was to describe the incidence, costs of care, length of stay, comorbidities, and mortality of SJS and
TEN in US adults. The Nationwide Inpatient Sample 2009e2012, containing a 20% sample of all US hospitalizations, was analyzed. We used a validated approach involving International Classification of Disease, 9th
edition, Clinical Modification codes to identify SJS, SJS/TEN, and TEN (n 2,591, n 502, and n 564,
respectively). The mean estimated incidences of SJS, SJS/TEN, and TEN were 9.2, 1.6, and 1.9 per million adults
per year, respectively. SJS/TEN was associated with nonwhite race, particularly Asians (odds ratio 3.27, 95%
confidence interval 3.02e3.54) and blacks (odds ratio 2.01, 95% confidence interval 1.92e2.10). Significantly prolonged length of stay and higher costs of care (SJS: 9.8  0.3 days, $21,437  $807; SJS/TEN: 16.5  1.0
days, $58,954  $5,238; TEN: 16.2  1.0 days, $53,695  $4,037) were observed compared with all other admissions
(4.7  0.02 days, $11,281  $98). Mean adjusted mortality was 4.8% for SJS, 19.4% for SJS/TEN, and 14.8% for TEN.
SJS, SJS/TEN, and TEN pose a substantial health care burden. Predictors of mortality included increasing age,
increasing number of chronic conditions, infection (septicemia, pneumonia, tuberculosis), hematological
malignancy (non-Hodgkins lymphoma, leukemia), and renal failure (P  0.03 for all). Further studies are needed
to confirm mortality findings to improve prognostication of SJS/TEN.
Journal of Investigative Dermatology (2016) 136, 1387e1397; doi:10.1016/j.jid.2016.03.023

INTRODUCTION
Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are rare but severe blistering disorders characterized by widespread epidermal necrosis of the skin and
mucosa and associated with substantial morbidity and mortality. Classification involves body surface area (BSA)
affected, with SJS involving less than 10% of BSA and TEN
involving greater than 30% of BSA (Bastuji-Garin et al.,
1993). SJS/TEN overlap syndrome describes skin detachment between 10% and 30% of BSA. The incidences of SJS
and TEN have been reported to be 1e7 and 1e2 cases per
million people, respectively (Roujeau, Guillame, et al., 1990;
Rzany et al., 1996; Schopf et al., 1991; Strom et al., 1991).
The epidemiology of SJS and TEN in the United States is not

Department of Dermatology, Northwestern University Feinberg School of

Medicine, Chicago, Illinois, USA; 2Department of Dermatology, MountSinai Icahn School of Medicine, New York, New York, USA; and
3
Departments of Dermatology, Preventive Medicine and Medical Social
Sciences, Northwestern University Feinberg School of Medicine, Chicago,
Illinois, USA
Correspondence: Jonathan I. Silverberg, MD, PhD, MPH, Department of
Dermatology, Suite 1600, 676 North St. Clair Street, 312-503-4985,
Northwestern University Feinberg School of Medicine, Chicago, IL 60611,
USA. E-mail: JonathanISilverberg@gmail.com
Abbreviations: BSA, body surface area; CI, confidence interval; ICD-9-CM,
International Classification of Disease, 9th edition, Clinical Modification;
HCUP, Healthcare Cost and Utilization Project; LOS, length of stay; NIS,
Nationwide Inpatient Sample; OR, odds ratio; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis
Received 24 January 2016; revised 8 March 2016; accepted 16 March 2016;
accepted manuscript published online 30 March 2016; corrected proof
published online 24 May 2016

well-defined. The rarity of SJS and TEN has precluded largescale epidemiological studies in the United States, although
large-scale studies have been performed in Europe
(Mockenhaupt et al., 2007).
Some clinical risk factors for SJS and TEN have been
identified, such as infection with HIV and mycoplasma
(Mittmann et al., 2012; Mulvey et al., 2007), older age (Oen
et al., 2015), and a seasonal predilection in early spring and
winter. However, the role of epidemiological and other
clinical factors are not well described. We hypothesized that
race/ethnicity play a role in the pathogenesis, as studies have
shown the HLA-B*1502 and HLA-B*5801 alleles to be
associated with SJS and TEN in East Asian populations,
especially with usage of carbamazepine and allopurinol
(Chung et al., 2004; Kaniwa et al., 2008). Mortality for SJS/
TEN is considerable and has been reported to be between
10% and 34% (Kim et al., 2012; Sekula et al., 2013); controversy exists as to whether newer treatments, such as
intravenous immunoglobulin, actually decrease mortality
(Schneck et al., 2008). To our knowledge, the burden of SJS
and TEN on the health care system in terms of cost and length
of stay (LOS) in the United States has not previously been
quantified. We sought to analyze SJS, SJS/TEN, and TEN in
adults in the US population to determine risk factors
including racial predilection, inpatient burden, and mortality.
RESULTS
Study characteristics

There were 23,009,584 discharges analyzed in the Nationwide Inpatient Sample (NIS) between 2009 and 2012 (HCUP
Nationwide Inpatient Sample (NIS), 2009e2012). Weighted

2016 The Authors. Published by Elsevier, Inc. on behalf of the Society for Investigative Dermatology.

www.jidonline.org 1387

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

frequencies, which more accurately represent the full spectrum of patients hospitalized in the United States, were 4,786
and 7,409 for a primary and secondary diagnosis of SJS;
1,558 and 815 for SJS/TEN; and 945 and 1,730 for TEN,
respectively. Crude frequencies per million people ranged
from 8.61e9.69 for SJS, 1.46e1.84 for SJS/TEN, and
1.58e2.26 for TEN. There were no significant differences in
the incidences of SJS, SJS/TEN, or TEN between 2009 and
2012 (analysis of variance, P > 0.05). Average age was 57.6
 0.4, 55.8  0.9, and 59.6  0.7 years for patients with
SJS, SJS/TEN, and TEN, respectively. The cohort was 58.7%
female.
Associations with SJS, SJS/TEN, and TEN

Demographic factors associated with SJS, SJS/TEN, and TEN,

reported as odds ratios (ORs) and 95% confidence intervals
(CIs) included race/ethnicity (survey logistic regression: OR
[95% confidence interval] for Asian 3.27 [3.02e3.54]; for
black 2.01 [1.92e2.10]; for Hispanic 1.20 [1.13e1.28];
for Native American 1.39 [1.16e1.68]; and for multiracial/
other 1.59 [1.44e1.74] compared with whites), female
sex (OR [95% CI] 1.27 [1.23e1.31]), insurance status
(OR [95% CI] for Medicare 1.16 [1.10e1.21]; for selfpay 1.11 [1.04e1.20] compared with private insurance),
age (OR [95% CI] for 40e59 years 0.83 [0.79e0.88]; for
60e79 years 0.67 [0.63e0.71]; for 80 years 0.47
[0.44e0.51]), season of admission (OR [95% CI] for winter
0.89 [0.85e0.94]), hospital location (OR [95% CI] for fringe/
metropolitan <1 million people 1.20 [1.16e1.25]; for
micropolitan 1.25 [1.17e1.33]), and increasing number of
chronic conditions (OR [95% CI] for 1e2 chronic
conditions 1.47 [1.32e1.63]; for 3e4 1.57 [1.41e1.74];
for 5 1.97 [1.78e2.19]) (see Supplementary Table S1
online and Table 1).
Comorbidities

In multivariate survey logistic regression models controlling

for sociodemographic factors, ophthalmologic complications
were observed for SJS and for SJS/TEN and TEN, including
ocular infection or inflammation (OR [95% CI] for SJS
24.54 [21.02e28.64]; for SJS/TEN and TEN 33.84
[26.93e42.52]), affecting 8.1% and 10.6% of the cohort,
respectively. However, blindness/vision impairment was
associated only with the SJS cohort (OR [95% CI] 2.40
[1.86e3.08]) (Figure 1).
SJS was also associated with multiple malignancies,
particularly multiple myeloma (OR [95% CI] 2.91
[1.60e5.28]), leukemia (OR [95% CI] 2.67 [1.97e3.62]),
non-Hodgkins lymphoma (OR [95% CI] 2.01
[1.38e2.94]), and central nervous system cancer (OR [95%
CI] 3.79 [2.32e6.21]). Moreover, SJS was associated with
numerous infections, including mycoplasma infection (OR
[95% CI] 10.22 [3.27e31.97]), septicemia (OR [95% CI]
4.10 [3.69e4.59]), HIV/AIDS (OR [95% CI] 2.62
[1.70e4.03]), any skin infection (OR [95% CI] 2.94
[2.58e3.35]), any viral infection (OR [95% CI] 3.90
[3.10e4.89]), any bacterial infection (OR [95% CI] 2.56
[2.28e2.89]), any fungal infection (OR [95% CI] 5.76
[5.08e6.54]), infectious hepatitis (OR [95% CI] 1.88
[1.53e2.32]), and pneumonia (OR [95% CI] 1.67
[1.45e1.91]). There were also associations with a number of
1388 Journal of Investigative Dermatology (2016), Volume 136

other disorders, including epilepsy (OR [95% CI] 2.47

[2.15e2.85]), systemic lupus erythematosus (OR [95% CI]
5.34 [2.64e10.80]), acute kidney injury (OR [95% CI]
1.78, [1.36e2.33]), renal failure (OR [95% CI] 3.18
[2.86e3.53]), chronic kidney disease (OR [95% CI] 2.00
[1.78e2.24]), and liver disease/damage (OR [95% CI] 1.56
[1.34e1.82]). Similar associations were seen with the
SJS/TEN and TEN cohort, although the associations with
central nervous system cancer, mycoplasma infection, infectious hepatitis, blindness or vision impairment, systemic
lupus erythematosus, and acute kidney injury did not remain
significant. There were no cases of mycoplasma infection in
patients with SJS/TEN or TEN.
Among patients with a secondary diagnosis of SJS, the
most common primary diagnoses included HIV/AIDS (rank
#8) and various subtypes of septicemia, including septicemia not otherwise specified (#1), methicillin-resistant
Staphylococcus aureus septicemia (#14), methicillinsensitive Staphylococcus aureus septicemia (#12), and
Escherichia coli septicemia (#18) (see Supplementary
Table S2 online). Among patients with a secondary diagnosis of SJS/TEN or TEN, the most common primary diagnoses also included septicemia and its various subtypes,
including septicemia NOS (#1), methicillin-resistant
Staphylococcus aureus septicemia (#6), Pseudomonas species septicemia (#12), E. coli septicemia (#13), and Gramnegative bacteria septicemia (#17).
Hospital course and disposition

The mean LOS for patients with SJS, SJS/TEN, and TEN
was 9.82  0.24, 16.50  0.97, and 16.20  1.00 days,
respectively, compared with 4.70  0.02 days for
those without these disorders. There were no significant
differences in mortality for SJS, SJS/TEN, or TEN between
2009 and 2012 (P > 0.05). The strongest shared
predictorswhen analyzing SJS, SJS/TEN, and TEN
individuallyfor increased LOS was increasing number of
chronic conditions (linear regression, P < 0.05 for all)
(Table 2).
Approximately 1 in 7 patients with SJS, SJS/TEN, and
TEN underwent mechanical ventilation (2450; 14.2%, 95%
CI 12.6e15.7%), including ventilation for less than 96
hours (863; 5.0%, 95% CI 4.2e5.7%) and greater than 96
hours (1607; 9.3%, 95% CI 8.0e10.6%) (see
Supplementary Table S3 online). One in 11 patients underwent dialysis (1498; 8.7%, 95% CI 7.6e9.7%). Skin grafts
and physical therapy were performed in only a small subset
of patients (for skin grafts: 339; 2.0%, 95% CI 1.4e2.5%;
for physical therapy: 438; 2.5%, 95% CI 1.8e3.3%).
Feeding procedures were performed in 1498; 8.7% (95%
CI 7.8e9.7%) of patients (nasogastric tube, total parenteral
nutrition, or percutaneous gastrostomy).
Patients were most frequently routinely discharged to
home or other self-care (7973; 46.1%, 95% CI
44.4e47.9%). However, a large subset were transferred to
other facilities (3871; 22.5%, 95% CI 20.9e23.8%),
received home health care (2542; 14.7%, 95% CI
13.4e16.0%), transferred to a short-term hospital (1340;
7.7%, 95% CI 6.8e8.7%), or left against medical advice
(90; 0.5%, 95% CI 0.3e0.8%).

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Table 1. Adjusted predictors of SJS, SJS/TEN, or TEN in adult patients

SJS
Variable

OR [95% CI]

SJS/TEN
P

OR [95% CI]

TEN
P

OR [95% CI]

SJS, SJS/TEN, TEN

P

OR [95% CI]

Age, years
18e39

Ref

Ref

Ref

40e59

0.81 [0.76e0.86]

<0.0001

0.73 [0.64e0.83]

<0.0001

1.14 [0.98e1.31]

Ref

0.09

0.83 [0.79e0.88]

<0.0001

60e79

0.62 [0.58e0.67]

<0.0001

0.65 [0.56e0.75]

<0.0001

0.96 [0.82e1.13]

0.6

0.67 [0.63e0.71]

<0.0001

80

0.43 [0.40e0.47]

<0.0001

0.53 [0.44e0.64]

<0.0001

0.70 [0.57e0.84]

0.0002

0.47 [0.44e0.51]

<0.0001

Winter

0.85 [0.81e0.90]

<0.0001

0.97 [0.85e1.11]

<0.0001

Spring

Ref

Ref

Season
0.7

Summer

1.02 [0.96e1.08]

0.5

1.00 [0.87e1.14]

0.9

Fall

0.96 [0.91e1.02]

0.2

1.35 [1.19e1.53]

<0.0001

0.89 [0.85e0.94]

Ref

Ref

1.03 [0.98e1.08]

0.3

1.01 [0.96e1.06]

0.7

Sex
Female
Male

1.30 [1.24e1.35]

<0.0001

1.56 [1.42e1.72]

<0.0001

1.27 [1.23e1.31]

<0.0001

Ref

Ref

Ref

Ref

Race
White

Ref

Ref

Ref

Ref

Black

1.87 [1.77e1.97]

<0.0001

3.00 [2.67e3.37]

<0.0001

1.97 [1.76e2.20]

<.0001

2.01 [1.92e2.10]

<0.0001

Hispanic

1.19 [1.10e1.28]

<0.0001

1.53 [1.29e1.82]

<0.0001

1.05 [0.89e1.25]

0.6

1.20 [1.13e1.28]

<0.0001

Asian

2.84 [2.57e3.14]

<0.0001

6.70 [5.62e7.98]

<0.0001

2.77 [2.22e3.44]

<.0001

3.27 [3.02e3.54]

<0.0001

Native American

1.55 [1.26e1.90]

<0.0001

1.57 [0.99e2.49]

0.06

1.39 [1.16e1.68]

0.0005

Other

1.24 [1.09e1.40]

0.001

3.69 [3.04e4.47]

<0.0001

1.71 [1.36e2.15]

<.0001

1.59 [1.44e1.74]

<0.0001

Income quartile
1st

0.93 [0.87e0.99]

0.02

1.12 [0.97e1.29]

0.1

1.09 [0.95e1.25]

0.2

0.97 [0.92e1.03]

0.3

2nd

0.87 [0.82e0.92]

<0.0001

0.99 [0.85e1.15]

0.9

1.09 [0.95e1.25]

0.2

0.92 [0.87e0.97]

0.001

3rd

0.93 [0.88e0.99]

0.02

1.19 [1.04e1.37]

0.01

1.23 [1.08e1.40]

<.0001

1.00 [0.95e1.06]

0.9

4th

Ref

Ref

Ref

Ref

Medicare

1.23 [1.16e1.30]

<0.0001

0.84 [0.73e0.95]

0.01

1.15 [1.02e1.30]

0.02

1.16 [1.10e1.21]

<0.0001

Medicaid

1.01 [0.94e1.08]

0.88 [0.76e1.03]

0.1

1.18 [1.02e1.37]

0.03

1.01 [0.96e1.07]

Ref

Ref

0.95 [0.78e1.17]

0.9

1.11 [1.04e1.20]

0.004

Insurance

Private insurance
Self pay

Ref
1.17 [1.08e1.27]

0.8

0.0003

Ref

1.00 [0.83e1.20]

0.9

0.7

No charge

0.74 [0.52e1.04]

0.1

1.45 [0.87e2.41]

0.2

0.55 [0.23e1.32]

0.8

0.83 [0.63e1.08]

0.2

Other

0.91 [0.81e1.03]

0.1

1.32 [1.06e1.65]

0.01

0.96 [0.74e1.25]

0.4

0.97 [0.88e1.07]

0.5

Number of chronic conditions

0

Ref

Ref

Ref

Ref

1e2

1.28 [1.13e1.44]

<0.0001

3.31 [2.38e4.60]

<0.0001

3.23 [2.35e4.45]

0.01

1.47 [1.32e1.63]

<0.0001

3e4

1.38 [1.23e1.55]

<0.0001

2.42 [1.73e3.37]

<0.0001

5.47 [4.00e7.48]

<.0001

1.57 [1.41e1.74]

<0.0001

1.75 [1.56e1.96]

<0.0001

3.50 [2.53e4.86]

<0.0001

6.15 [4.51e8.38]

<.0001

1.97 [1.78e2.19]

<0.0001

Hospital location
Metropolitan 1 million

Ref

Ref

Ref

1.25 [1.19e1.31]

<0.0001

1.13 [1.02e1.24]

Micropolitan

1.37 [1.28e1.48]

<0.0001

Not metropolitan or micropolitan

1.17 [1.07e1.28]

0.001

Fringe/metropolitan <1 million

Ref

0.02

1.20 [1.16e1.25]

<0.0001

1.02 [0.87e1.20]

0.8

1.25 [1.17e1.33]

<0.0001

0.87 [0.71e1.07]

0.2

1.07 [0.99e1.16]

0.1

Abbreviations: CI, confidence interval; OR, odds ratio; Ref, reference; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis.
Boldface indicates result is a statistically significant finding.

Cost of care

The mean inflation-adjusted cost of care for patients with SJS,

SJS/TEN, and TEN was $21,437  $806; $58,954  $5,238;
and $53,695  $4,037, respectively. In contrast, patients
without these diagnoses had a mean cost of care of $11,284
 $98. The cumulative cost of care from 2009e2012 was
$249,296,785; $130,567,975; and $134,123,463 for all
patients with SJS, SJS/TEN, and TEN, respectively. Predictors
of cost of care were analyzed for each disorder separately;
shared significant predictors of increased cost of care
included nonwhite race in SJS or SJS/TEN patients and

increasing number of chronic conditions for SJS, SJS/TEN, or

TEN (linear regression, P < 0.05 for all) (Table 3).
Mortality

The mean annual age- and sex-adjusted in-hospital mortality

rates were 4.8% (range 3.7e7.6%) for SJS, 19.4% (range
15.7e22.3%) for SJS/TEN, and 14.8% (range 7.7e19.0%)
for TEN. Patients with a secondary diagnosis of SJS and
SJS/TEN had higher rates of mortality than those with a primary diagnosis; mean adjusted mortality rates were 3.1%
(range 2.6e3.6%) and 5.9% (range 5.3e6.5%) for a
www.jidonline.org 1389

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Figure 1. Association between SJS or SJS/TEN and TEN and health comorbidities. Survey logistic regression models were constructed with SJS or SJS/TEN
and TEN as the independent variable and the respective comorbidity as the dependent variable. Models included race/ethnicity, age and sex as covariates.
Adjusted odds ratios, 95% confidence intervals, corrected P-values, and forest plots of the adjusted odds ratios are presented.

1390 Journal of Investigative Dermatology (2016), Volume 136

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Table 2. Predictors of length of stay in patients with SJS, SJS/TEN, or TEN

SJS
Variable

LSM

Adj Beta [95% CI]

18e39

1.90

0 [ref]

40e59

1.90 0.00084 [e0.076 to 0.077]

SJS/TEN
P

LSM

TEN
P

Adj Beta [95% CI]

LSM

Adj Beta [95% CI]

2.4

0 [ref]

Age, years

0.9

2.25

0 [ref]

2.47

0.21 [0.20 to 0.22]

<0.0001 1.98 e0.43 [e0.54 to e0.32] <0.0001

60e79

1.86

e0.039 [e0.12 to 0.046]

0.4

2.27

0.019 [0.015 to 0.024]

<0.0001 1.98 e0.43 [e0.56 to e0.31] <0.0001

80

1.83

e0.065 [e0.16 to 0.033]

0.2

2.11

e0.14 [e0.16 to e0.13]

<0.0001 1.99 e0.42 [e0.55 to e0.29] <0.0001

Season
Winter

1.92

0.074 [0.0019 to 0.15]

0.04

2.18

e0.090 [e0.095 to e0.086]

Spring

1.85

0 [ref]

2.27

0 [ref]

<0.0001 2.04 0.011 [e0.062 to 0.085]

Summer

1.87

0.023 [e0.049 to 0.094]

0.5

2.40

0.13 [0.12 to 0.14]

Fall

1.86

0.016 [e0.054 to 0.087]

0.6

2.26 e0.0073 [e0.011 to e0.0039] <0.0001 2.17

0.1

2.30

0.038 [0.034 to 0.042]

2.26

0 [ref]

2.14

0 [ref]

2.03

<0.0001 2.12

0.8

0 [ref]

0.094 [e0.011 to 0.20]

0.08

0.14 [0.022 to 0.26]

0.02

Gender
Female

1.85 e0.045 [e0.098 to 0.0090]

Male

1.90

0 [ref]

<0.0001 2.01 e0.16 [e0.24 to e0.074]

2.17

0 [ref]

2.00

0 [ref]

0.0002

Race
White

1.78

0 [ref]

Black

1.84

0.063 [e0.0041 to 0.13]

0.07

Hispanic

1.82

0.040 [e0.053 to 0.13]

Asian

0.96

0.18 [0.052 to 0.30]

2.14 e0.0041 [e0.0065 to e0.0017]

0.009

2.04 0.034 [e0.015 to 0.084]

0.4

2.36

0.19 [0.19 to 0.20]

<0.0001 2.03

0.006

2.27

0.12 [0.11 to 0.14]

<0.0001 2.08 0.083 [e0.0031 to 0.17]

Native American

2.00

0.22 [e0.050 to 0.49]

0.1

NE

NE

Other

1.85

0.069 [e0.080 to 0.22]

0.4

2.50

0.36 [0.36 to 0.36]

NE

2.33

<0.0001 2.05

0.026 [e0.098 to 0.15]

0.2
0.4
0.06

0.33 [0.086 to 0.57]

0.008

0.053 [e0.26 to 0.36]

0.05

Income quartile
1st

1.89

0.0045 [e0.077 to 0.086]

0.9

2.30

0.024 [0.012 to 0.037]

0.017 [e0.13 to 0.16]

0.8

2nd

1.85

e0.033 [e0.11 to 0.045]

0.4

2.19

e0.087 [e0.099 to e0.074]

<0.0001 2.02

e0.10 [e0.25 to 0.043]

0.2

3rd

1.88 e0.0063 [e0.085 to 0.072]

0.9

2.34

0.060 [0.049 to 0.071]

<0.0001 2.08

e0.040 [e0.20 to 0.12]

0.6

4th

1.88

0 [ref]

2.28

0 [ref]

Medicare

1.84

0.015 [e0.055 to 0.084]

0.7

2.21

0.082 [0.078 to 0.085]

Medicaid

1.94

0.11 [0.016 to 0.20]

0.02

2.28

0.16 [0.15 to 0.16]

0.0002 2.14

2.12

0 [ref]

Insurance

<0.0001 1.95 e0.044 [e0.17 to 0.080]

<0.0001 2.35

1.99

0.5

0.35 [0.26 to 0.44]

<0.0001

0 [ref]

Private insurance

1.83

0 [ref]

2.12

0 [ref]

Self-pay

1.77

e0.058 [e0.15 to 0.035]

0.2

1.91

e0.22 [e0.23 to e0.20]

No charge

2.02

0.19 [e0.14 to 0.53]

0.3

2.83

0.71 [0.69 to 0.72]

<0.0001 2.36

0.37 [0.24 to 0.50]

<0.0001

Other

1.85

0.025 [e0.14 to 0.19]

0.8

2.32

0.19 [0.19 to 0.20]

<0.0001 2.30

0.31 [0.089 to 0.52]

0.006

<0.0001 1.59 e0.41 [e0.51 to e0.31] <0.0001

Number of chronic
conditions
0

1.59

0 [ref]

1.77

0 [ref]

0 [ref]

1e2

1.77

0.19 [0.074 to 0.30]

0.001

2.35

0.57 [0.51 to 0.64]

<0.0001 2.04

1.39

0.65 [0.49 to 0.81]

<0.0001

3e4

1.93

0.34 [0.24 to 0.44]

<0.0001 2.41

0.64 [0.57 to 0.70]

<0.0001 2.38

0.99 [0.86 to 1.13]

<0.0001

2.21

0.62 [0.52 to 0.72]

<0.0001 2.58

0.81 [0.74 to 0.87]

<0.0001 2.55

1.16 [1.03 to 1.30]

<0.0001

Hospital location
Metropolitan >1 million 1.88
Fringe/metropolitan
<1 million

1.89

0 [ref]
0.019 [e0.049 to 0.086]

0.6

2.39

0 [ref]

2.32

e0.075 [e0.083 to e0.067]

2.24

<0.0001 2.20

0 [ref]

e0.040 [e0.11 to 0.30]

0.3

Micropolitan

1.87

e0.0072 [e0.11 to 0.96]

0.9

2.49

0.095 [0.088 to 0.10]

<0.0001 2.03 e0.21 [e0.28 to e0.13] <0.0001

Not metropolitan or
micropolitan

1.86

e0.018 [e0.13 to 0.098]

0.8

1.91

e0.48 [e0.49 to e0.47]

<0.0001 1.90 e0.34 [e0.41 to e0.26] <0.0001

Year
2009

1.96

0 [ref]

2.34

0 [ref]

2010

4.90

e0.061 [e0.16 to 0.038]

0.2

2.28

e0.060 [e0.061 to e0.060]

2.08

2011

1.85

e0.11 [e0.20 to e0.019]

0.02

2.30

e0.041 [e0.047 to e0.036]

<0.0001 1.97

e0.10 [e0.22 to 0.010]

0.08

2012

1.79

e0.17 [e0.25 to e0.093] <0.001

2.19

e0.14 [e0.15 to e0.14]

<0.0001 2.06

e0.14 [e0.11 to 0.083]

0.8

<0.0001 2.24

0 [ref]

0.17 [0.067 to 0.27]

0.001

Abbreviations: Adj, adjusted; CI, confidence interval; LSM, least squares means; NE, not estimated; OR, odds ratio; ref, reference; SJS, Stevens-Johnson
syndrome; TEN, toxic epidermal necrolysis.
Boldface indicates result is a statistically significant finding.

primary and secondary diagnosis of SJS, 14.3% (range

12.2e16.3%) and 29.5% (range 24.5e34.6%) for SJS/TEN,
and 17.8% (range 15.0e20.7%) and 15.0% (range
12.5e17.6%) for TEN, respectively. In multivariate logistic

regression models controlling for all sociodemographic

factors and comorbidities, predictors of mortality for patients
hospitalized with SJS, SJS/TEN, or TEN included increasing
age (adjusted OR [95% CI] for 40e59 years 1.46
www.jidonline.org 1391

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Table 3. Predictors of cost of care in patients with SJS, SJS/TEN, or TEN

SJS
Variable
Age, years
18e39
40e59
60e79
80
Season
Winter
Spring
Summer
Fall
Gender
Female
Male
Race
White
Black
Hispanic
Asian
Native American
Other
Income quartile
1st
2nd
3rd
4th
Insurance
Medicare
Medicaid
Private insurance
Self-pay
No charge
Other
Number of chronic
conditions
0
1e2
3e4
5
Hospital location
Metropolitan
>1 million
Fringe/metropolitan
<1 million
Micropolitan
Not metropolitan
or micropolitan
Year
2009
2010
2011
2012

LSM

SJS/TEN
P

Adj Beta [95% CI]

LSM

Adj Beta [95% CI]

TEN
P

LSM

Adj Beta [95% CI]

9.30
0 [ref]
9.34 0.043 [e0.062 to 0.15]
9.31 0.0096 [e0.099 to 0.12]
9.16 e0.14 [e0.27 to e0.013]

0.4
0.9
0.03

10.15
0 [ref]

10.2
0 [ref]

10.36
0.21 [0.20 to 0.22]
<0.0001 9.74 e0.44 [e0.82 to e0.062] 0.02
10.12 e0.025 [e0.031 to e0.019] <0.0001 9.71 e0.48 [e0.89 to e0.066] 0.02
9.62 e0.53 [e0.56 to e0.49] <0.0001 9.52 e0.67 [e1.13 to e0.20] 0.005

9.34 0.071 [e0.016 to 0.16]

9.27
0 [ref]
9.22 e0.048 [e0.14 to 0.048]
9.27 e0.0027 [e0.088 to 0.082]

0.1

0.3
0.9

9.78 e0.34 [e0.35 to e0.34] <0.0001

10.12
0 [ref]

10.3
0.18 [0.17 to 0.19]
<0.0001
10.06 e0.062 [e0.080 to e0.044] <0.0001

9.25 e0.062 [e0.13 to 0.0048]

9.31
0 [ref]

0.07

10.05
10.08

e0.36 [e0.42 to e0.30]

0 [ref]

<0.0001

9.10
9.15
9.24
9.30
9.50
9.37

0.2
0.003
0.003
0.04
0.01

9.87
9.76
10.2
10.08
NE
10.41

0 [ref]
e0.11 [e0.12 to e0.11]
0.33 [0.32 to 0.34]
0.21 [0.18 to 0.25]
NE
0.55 [0.54 to 0.55]

9.74
0 [ref]
<0.0001 9.65 e0.089 [e0.35 to 0.18]
<0.0001 9.55 e0.19 [e0.60 to 0.23]
<0.0001 9.9
0.15 [e0.37 to 0.67]
NE
9.84 0.096 [e1.03 to 1.22]
<0.0001 10.1
0.31 [e0.21 to 0.83]

9.21 e0.16 [e0.26 to e0.060]

9.24 e0.13 [e0.23 to e0.033]
9.30 e0.078 [e0.18 to 0.021]
9.37
0 [ref]

0.002
0.01
0.1

9.99 e0.083 [e0.11 to e0.057] <0.0001 9.77 e0.23 [e0.56 to 0.092] 0.2
10.09 0.021 [e0.0044 to 0.046]
0.1
9.66 e0.35 [e0.69 to e0.010] 0.04
10.1
0.027 [e0.0041 to 0.058]
0.1
9.72 e0.28 [e0.61 to 0.049] 0.1
10.07
0 [ref]

10.00
0 [ref]

9.28 e0.039 [e0.13 to 0.055]

9.39 0.067 [e0.048 to 0.18]
9.32
0 [ref]
9.13 e0.19 [e0.32 to e0.066]
9.20
e0.12 [e0.66 to 0.43]
9.35
0.031 [e0.15 to 0.21]

0.4
0.3

0.003
0.7
0.7

9.95 0.011 [0.0019 to 0.020]

9.91 e0.030 [e0.040 to e0.019]
9.94
0 [ref]
9.57 e0.37 [e0.38 to e0.37]
10.54
0.60 [0.59 to 0.61]
10.47
0.53 [0.52 to 0.55]

0 [ref]
0.056 [e0.034 to 0.15]
0.15 [0.049 to 0.25]
0.21 [0.070 to 0.34]
0.41 [0.028 to 0.79]
0.28 [0.061 to 0.50]

8.89
9.18
9.29
9.74

0 [ref]
0.29 [0.15 to 0.43]
0.39 [0.26 to 0.52]
0.85 [0.71 to 0.98]

9.39

0 [ref]

9.23
<0.0001 10.06
<0.0001 10.28
<0.0001 10.69

10.19

0 [ref]
0.82 [0.79 to 0.86]
1.04 [1.01 to 1.08]
1.46 [1.41 to 1.50]

9.77
9.72
9.81
9.86

0.045 [e0.25 to 0.34]

0 [ref]
0.091 [e0.22 to 0.40]
0.14 [e0.17 to 0.45]

0.8

0.6
0.4

9.67 e0.23 [e0.44 to e0.020] 0.03

9.91
0 [ref]

0.5
0.4
0.6
0.9
0.2

0.02
9.60 e0.28 [e0.60 to 0.039]
<0.0001 10.14 0.27 [e0.068 to 0.60]

9.88
0 [ref]
<0.0001 9.08 e0.80 [e1.25 to e0.35]
<0.0001 10.06
0.18 [e0.41 to 0.78]
<0.0001 9.98
0.10 [e0.37 to 0.58]

0.09
0.1

0.0005
0.5
0.7

8.97
<0.0001 9.60
<0.0001 10.21
<0.0001 10.38

0 [ref]
0.62 [e0.21 to 1.46]
1.23 [0.42 to 2.05]
1.40 [0.60 to 2.20]

0.1
0.003
0.0006

0 [ref]

0 [ref]

10.1

9.25 e0.14 [e0.23 to e0.063]

0.0006 10.01

e0.17 [e0.17 to e0.17]

<0.0001

9.89 e0.23 [e0.47 to 0.012] 0.06

9.21 e0.19 [e0.32 to e0.055]

9.24 e0.16 [e0.34 to 0.016]

0.005
0.08

10.34
9.72

0.15 [0.15 to 0.15]

e0.47 [e0.47 to e0.46]

<0.0001
<0.0001

9.7 e0.41 [e0.78 to e0.044] 0.03

9.45 e0.66 [e1.10 to e0.22] 0.003

9.32
0 [ref]
9.30 e0.016 [e0.14 to 0.11]
9.30 e0.023 [e0.15 to 0.10]
9.19 e0.13 [e0.23 to e0.028]

0.8
0.7
0.01

10.12
0 [ref]

10.07 e0.051 [e0.081 to e0.020] 0.001

10.14
0.027 [0.026 to 0.028]
<0.0001
9.93 e0.18 [e0.19 to e0.18] <0.0001

9.68
9.96
9.77
9.74

0 [ref]
0.28 [e0.019 to 0.58]
0.086 [e0.22 to 0.39]
0.056 [e0.25 to 0.36]

0.07
0.6
0.7

Abbreviations: Adj, adjusted; CI, confidence interval; LSM, least squares means; OR, odds ratio; ref, reference; SJS, Stevens-Johnson syndrome; TEN, toxic
epidermal necrolysis.
Boldface indicates result is a statistically significant finding.

[1.14e1.88]; for 60e79 years 2.51 [1.93e3.25]; for 80

years 2.60 [1.92e3.42]), insurance status (adjusted OR
[95% CI] for Medicaid 2.21 [1.81e2.70]; for Medicare
1392 Journal of Investigative Dermatology (2016), Volume 136

1.47 [1.24e1.74]), non-Hodgkins lymphoma (adjusted

OR [95% CI] 5.25 [2.15e11.02]), leukemia (adjusted OR
[95% CI] 2.19 [1.60e3.02]), septicemia (adjusted

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Table 4. Predictors of mortality in patients with SJS,

SJS/TEN, or TEN
Variable
Age, years
18e39
40e59
60e79
80
Insurance
Medicare
Medicaid
Private insurance
Self-pay
No charge
Other
Number of chronic conditions
0
1e2
3e4
5
Hematological malignancy
Leukemia
Non-Hodgkins lymphoma
Infection
Septicemia
Any skin infection
Tuberculosis
Pneumonia
Various
Renal failure
Blindness/vision loss

Adj OR [95% CI]

Ref
1.46 [1.14e1.88]
2.51 [1.93e3.25]
2.60 [1.92e3.42]

0.01
<0.0001
<0.0001

1.47 [1.24e1.74]
2.21 [1.81e2.70]
Ref
1.20 [0.83e1.75]
NE
1.21 [0.80e1.83]

0.0002
<0.0001

0.4
0.9
0.5

Ref
5.24 [2.11e13.02]
2.35 [0.95e5.83]
3.29 [1.34e8.08]

0.003
0.1
0.03

2.19 [1.60e3.02]
5.25 [4.01e6.88]

<0.0001
<0.0001

7.80
0.48
10.12
1.57

[6.85e8.85]
[0.40e0.58]
[6.09e16.84]
[1.38e1.80]

3.92 [3.47e4.44]
0.17 [0.08e0.36]

<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0001

Abbreviations: Adj, adjusted; CI, confidence interval; NE, not estimated;

OR, odds ratio; Ref, reference; SJS, Stevens-Johnson syndrome; TEN, toxic
epidermal necrolysis.
Boldface indicates result is a statistically significant finding.

OR [95% CI] 7.80 [6.85e8.85]), tuberculosis (adjusted OR

[95% CI] 10.12 [6.09e16.84]), pneumonia (adjusted OR
[95% CI] 1.57 [1.38e1.80]), and renal failure (adjusted
OR [95% CI] 3.92 [3.47e4.44]) (Table 4).
DISCUSSION
The present study found that SJS, SJS/TEN, and TEN were
a significant health care burden, with a mean cost of hospitalization of greater than $20,000 for SJS and greater than
$50,000 for SJS/TEN and TEN; a total 4-year cost of
$511,245,822 (mean $128 million per year); and up to
4-fold longer duration and 5-fold higher mean cost of hospitalization compared with that of an average hospital
admission. In comparison, pemphigus, an autoimmune blistering disorder associated with significant morbidity and
mortality, had a mean hospitalization cost of only $14,521
and a mean LOS of 7.2 dayssignificantly less than that of
SJS, SJS/TEN, or TEN (Hsu, Brieva, and Silverberg, 2016).
There was a higher crude incidence of SJS (9.2 cases per
million people) but similar incidence of TEN (Naldi et al.,
1990; Roujeau, Chosidow, et al., 1990) in the US adult
population compared with previous studies. Incidences were
higher than previously reported in Germany (1.89 cases of
SJS and TEN per 1 million people) (Rzany et al., 1996). One

of the strongest risk factors for SJS, SJS/TEN, and TEN was an
increasing number of chronic conditions, suggesting that
polypharmacy and poor health status are major contributors.
Other predictors included female sex, which is consistent
with previous studies (Sekula et al., 2013), and younger age
in adults. There were significant racial disparities, such that
nonwhites compared with whites had higher rates of SJS, SJS/
TEN, and TEN, with longer LOSs and increased mortality.
Although studies in the past have linked certain human
leukocyte antigen haplotypes to SJS/TEN in Asian populations
(Chung et al., 2004; Kaniwa et al., 2008), to our knowledge,
no studies examined rates of SJS/TEN in different racial
groups. We found that Asians and blacks had the highest rates
of SJS, SJS/TEN, and TEN, although Hispanics, Native
Americans, and those who were multiracial/other had
smaller but significant increases. Future studies are needed to
identify the reasons for such racial disparities, including
common genetic factors, differences of access to care, and
medication use.
The mean age- and sex-adjusted mortality of SJS (4.8%),
SJS/TEN (19.4%), and TEN (14.8%) were lower than those in
previous reports. A systematic review including 20 studies of
outcomes for TEN managed in burn centers found a mean
mortality of 30% (range 10e80%). However, these patients
may have had more severe disease because they were cared
for in a burn unit. Nevertheless, 16 of 20 studies included
less than 50 patients (Mahar et al., 2014). The two largest
studies from the United States, which included only 109
(Imahara et al., 2006) and 199 patients (Palmieri et al., 2002),
reported in-hospital mortality rates of 20% and 32%,
respectively. Most of these studies were limited because most
of their patients were referred; presumably more severe or
refractory cases were referred, which may not reflect the
entire population of TEN patients. The results of the present
study are likely generalizable to the entire US population. On
the other hand, it may be possible that a small fraction of the
patients we analyzed did not have SJS/TEN, which may lead
to an underestimation of the mortality rate. Compared with
pemphigus, a severe blistering disorder, mortality for SJS/TEN
(range 4.8e19.4%) was substantially higher than that
of inpatients with pemphigus (range 1.60e3.20%) (Hsu,
Brieva, Sinha, et al., 2016). The present study found similar
or even worse outcomes for SJS/TEN than for TEN, with
comparable rates of ventilation, physical therapy, dialysis,
feeding, cost of care, LOS, and mortality; however, TEN
was associated with higher rates of skin grafts. Together, the
results suggest that once the affected BSA surpasses 10%,
there is no incremental increase in morbidity and mortality
with increasing affected BSA. This is consistent with one
of the elements of the SCORe of Toxic Epidermal Necrolysis
score (detached or compromised body surface >10%), a
severity-of-illness score designed to predict mortality in SJS/
TEN (Bastuji-Garin et al., 2000). We found significantly
higher mortality in patients with older age, increasing number of chronic conditions, associated hematological malignancy, renal failure, septicemia, pneumonia, and
tuberculosis. Confirmatory studies are needed to determine
whether hematological malignancies, rather than all malignancies in general, are associated with mortality, because
only non-Hodgkins lymphoma and leukemia were
www.jidonline.org 1393

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

associated with increased mortality. Although we were unable to assess actual serum glucose levels, presence of diabetes or proxies for elevated blood glucose (e.g.,
hyperosmolar hyperglycemic nonketotic syndrome) were not
significant predictors of mortality. Elevated serum glucose
level and heart rate can be proxies for severe infection;
however, in our analysis, only a few infections predicted for
mortality. Further confirmatory studies are needed to determine whether the strongest infectious predictors of death are
septicemia, pneumonia, and tuberculosis.
Mortality was higher among patients with a secondary
diagnosis of SJS or SJS/TEN versus those with a primary
diagnosis. This may be because patients with SJS or SJS/TEN
listed as a secondary diagnosis had other serious medical
problems that warranted admission and iatrogenically
developed SJS or SJS/TEN during their admission. Alternatively, they may have delayed seeking care for their skin
condition until complications developed, such as septicemia.
These conditions may be listed as the primary reason for
admission. Indeed, septicemia and its various subtypes
accounted for most the top 20 most common primary diagnoses when SJS, SJS/TEN, or TEN were secondary diagnoses. Future studies are warranted to clarify how many
inpatients develop SJS, SJS/TEN, and TEN as a result of
treatment for another health disorder. In turn, it may be
possible to develop interventions to decrease the risk
of developing and improve early detection of SJS, SJS/TEN,
and TEN.
The harmful effects of SJS, SJS/TEN, and TEN on mortality
extend beyond the inpatient setting. An international registry
study of 460 patients with SJS, SJS/TEN, and TEN found that
mortality rates increased over time to 1-year mortality rates of
24%, 43%, and 49%, respectively (Sekula et al., 2013). These
patients may require additional interventions to reduce their
excess mortality risk even on discharge from the hospital.
This is particularly important given that 44.9% of patients
were transferred to other facilities and short-term hospitals
or received home health care. Future studies are needed
to identify the best interventions to reduce inpatient and
postdischarge mortality for SJS/TEN.
Several malignancies were associated with SJS and SJS/TEN
or TEN, including multiple myeloma, leukemia, non-Hodgkins lymphoma, and central nervous system cancer. Various
case reports have linked SJS to mogamulizumab treatment in
adult T-cell leukemia/lymphoma (Ishida et al., 2013).
Whether the association of malignancy with SJS/TEN is
disease- or medication-related is not completely understood
(Gravante et al., 2007; Rosen et al., 2014). The association
between SJS and central nervous system cancer has been
previously described and is likely due to use of dexamethasone and antiepileptic drugs in the setting of cerebral edema
from brain tumors (Mockenhaupt et al., 2007). However, the
present study suggests that there is an association of SJS/TEN
with hematologic malignancies in particular and not malignancy in general. These findings may help guide future
research into the causative agents of SJS/TEN. Infectious
comorbidities that were significantly associated with SJS/TEN
included mycoplasma, HIV/AIDS, and tuberculosis, which
confirms previous reports (Jung et al., 2015; Mittmann et al.,
2012; Mulvey et al., 2007; Tay et al., 1996). In particular,
1394 Journal of Investigative Dermatology (2016), Volume 136

mycoplasma was associated with SJS but not with SJS/TEN

and TEN. However, mycoplasma was not associated with
mortality, suggesting that it may be associated with a better
prognosis. Herpes infection was not significantly associated
with either SJS or SJS/TEN and TEN. HIV/AIDS was the eighth
most common primary diagnosis for patients with a secondary diagnosis of SJS/TEN. This is consistent with previous
studies and may be due to nevirapine use (Metry et al., 2001)
and/or immune dysregulation. Tuberculosis may be associated with SJS/TEN with the use of antituberculosis medications (Wirima and Harries, 1991). Sepsis, bacterial, fungal,
and viral infections were associated with SJS/TEN. These are
well-established complications and the most common causes
of death in SJS/TEN (Letko et al., 2005). On the other hand,
antibiotics used to treat these infections may be the causative
agent of SJS/TEN. Systemic lupus erythematosus was significantly associated with SJS as well. It is unknown whether this
is medication-induced or a misdiagnosis, or if an immune
dysregulation state may predispose to SJS/TEN. However, the
association with SJS/TEN and systemic lupus erythematosus
has been previously reported (Bellakhal et al., 2015; Lee
et al., 2011; Torchia et al., 2012). Lastly, acute kidney
injury (2.3%), renal failure (28.2%), and chronic kidney
disease (21.8%) were also observed as significant comorbid
health conditions with SJS. Although renal insufficiency may
predispose to SJS through impaired excretion of drugs (Chung
et al., 2015; Hung et al., 2009), SJS may also lead to acute
kidney injury and renal failure through contributing factors
such as fluid/electrolyte loss and severe infection. A previous
retrospective case series of 234 patients in Taiwan also
showed that chronic kidney disease and renal failure were
associated with SJS/TEN (Hung et al., 2009). Chronic kidney
disease and renal failure had similar associations with
SJS/TEN and TEN, although acute kidney injury was not
significant.
Strengths of this study include an analysis of a nationally
representative sample of all-payer data over a period of
4 years with over 21 million records that included several
thousand cases of SJS, SJS/TEN, and TEN using a validated
algorithm for case selection. Given the severe, painful, and
mortal nature of SJS/TEN, virtually all patients with these
disorders present for care and are admitted to the hospital for
evaluation and management. Thus, this inpatient cohort is
likely sufficient to describe the epidemiology of these disorders. Nevertheless, it is possible that the prevalence of these
disorders is slightly underestimated. Limitations of this study
include the inability to perform temporal analysis of comorbidities, because date of diagnosis was not present. Moreover, identification of comorbidities was performed using
International Classification of Disease, 9th edition, Clinical
Modification (ICD-9-CM) (http://www.cdc.gov/nchs/icd/
icd9cm.htm) codes and not verified by review of the health
record or cross-referencing with cancer or other health registries. Finally, we were unable to assess the medications
used preceding the diagnosis of SJS/TEN or for the treatment
of SJS/TEN. Thus, we were unable to determine whether SJS/
TEN is related to a particular disease or its medications. LOS
may have been overestimated, because we excluded patients
with LOSs shorter than 3 days in our case definition. Controversy exists about whether current therapeutic approaches

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

for SJS/TEN and TEN, such as systemic corticosteroids and

intravenous immunoglobulin, actually reduce mortality.
Although we were unable to determine the treatments used,
the present study suggests that mortality for SJS and TEN remains substantial. Future studies are needed to answer these
questions.
In conclusion, the incidence of SJS was higher than previously reported. Significant racial disparities were observed,
with increased incidence, mortality, cost, and LOS in
nonwhite patients. Mortality was associated with a number of
previously unrecognized factors. Future studies are needed to
improve prognostication and reduce the incidence and
mortality of SJS/TEN.
METHODS
Data source
The 2009e2012 Nationwide Inpatient Sample (NIS) was analyzed
(HCUP, 2009e2012). The NIS is sponsored by the HCUP of the
Agency for Healthcare Research and Quality. Each year of the NIS
contains an approximately 20% stratified representative sample of
all hospitalizations in the United States. Sample weights were
created by the NIS that factored in the sampling design of hospitals.
These sample weights were needed to provide representative estimates of hospital discharges across the United States. All data were
de-identified, and no attempts were made to identify any of the individuals in the database. Patient consent was not obtained because
the databases were received de-identified. All parties with access to
the HCUP were compliant with the HCUPs formal data use
agreement. This study was deemed to not require institutional review
board approval.

Selection of SJS, SJS/TEN, and TEN

The databases were searched for a primary and/or secondary
discharge ICD-9-CM codes of 695.13, 695.14, and 695.15 corresponding to SJS, SJS/TEN, and TEN, respectively. We precluded
analysis of any years before 2009 because ICD-9-CM codes specific
for SJS, SJS/TEN, and TEN were not introduced until October 2008.
Furthermore, based on a previously validated algorithm, patients
who had a LOS shorter than 3 days were excluded to improve the
positive predictive value of the diagnostic codes (Davis et al., 2015).
In the validation study performed by Davis et al., in comparison to
patients not hospitalized, the likelihood of case status increased with
increasing LOS (OR [95% CI] for 3e5 days 1.78 [1.60e38.28]; for
6e13 days 7.67 [1.42e41.56]; and for >13 days 63.9
[8.2e499.1]). Patients with a concomitant diagnosis of any bullous
dermatoses were also excluded (ICD-9-CM codes 694.0e694.9)
because certain bullous disorders, such as linear IgA bullous dermatoses, have been reported to masquerade as TEN (Kakar et al.,
2013). We also excluded any patients with a concomitant diagnosis of erythema multiforme major.

Comorbidities and procedures

A comorbidity was considered present with either a primary or
secondary diagnosis of the disease. The NIS lists one primary diagnosis and up to 24 secondary diagnoses. Some comorbidities were
precoded according to the NIS Clinical Classification Software
provided by AHRQ (Rockville, MD). All comorbidities were determined through ICD-9-CM codes. Comorbidities analyzed were
selected based on previous associations with SJS/TEN in the literature or to explore for novel associations. ORs and 95% CIs were
plotted on forest plots through Open Meta Analyst, an open-source

software supported by Brown University and the Agency for

Healthcare Research and Quality. Post hoc correction for multiple
dependent tests was performed by minimizing the false discovery
rate with the approach of Benjamini and Hochberg and corrected Pvalues are presented (Benjamini and Hochberg, 1995). ICD-9-CM
codes were also used to determine procedures, such as mechanical ventilation, skin grafts, and hemodialysis. A list of the top 20
most common primary diagnoses for patients with a secondary
diagnosis of SJS and for SJS/TEN and TEN was also generated.

Mortality
Annual inpatient mortality was calculated for patients with no, a
primary, or a secondary diagnosis of SJS, SJS/TEN, or TEN. Mortality
rates were adjusted for sex and race based on the composition of the
US population corresponding to that year based on data from the US
Census Bureau (United States Census Bureau, 2013). Calculation of
mortality excluded any patients who were transferred to another
hospital, which included 7.7% of the overall cohort. Mortality risk
and loss of function severity were determined by the All Patient
Refined Diagnosis Related Group, which includes severity of illness
and risk of mortality subclasses. The All Patient Refined Diagnosis
Related Group is assigned by software (3M APR DRG Software)
developed by 3M Health Information Systems (St. Paul, MN) based
on patient diagnosis-related group.

Statistical analysis
All data processing and statistical analyses were performed using
SAS version 9.4 (SAS Institute, Cary, NC). Analyses were performed
using SURVEY procedures that adjusted for survey weighting,
sampling clusters, and strata. Weighted incidences of a primary
and/or secondary diagnosis of SJS, SJS/TEN, and TEN were determined. Calculation of incidences excluded patients who were
transferred to another hospital. The cost for inpatient care was
calculated based on the total charge of the hospitalization and the
cost-to-charge ratio estimated by HCUP. All costs were adjusted
for inflation to the year 2014 according to the Consumer Price
Index from the United States Bureau of Labor Statistics (United
States Department of Labor, 2014). Summary statistics were
generated for LOS and estimated inflation-adjusted cost of care,
including sum, mean, standard deviation, minimum, maximum,
and median.
Associations of SJS/TEN and associations of mortality were
examined for the following variables: age (18e39 [reference],
40e59, 60e79, and 80 years), sex (male [reference], female), race/
ethnicity (white [reference], black, Hispanic, Asian, Native American, multiracial/other), median annual income of the hospital zip
code (quartiles [reference first quartile]), health insurance
coverage (Medicare, Medicaid, private [reference], self-pay, no
charge, other), number of chronic conditions (0 [reference], 1e2,
3e4, 5), season of admission (winter, spring [reference], summer,
autumn), hospital location (metropolitan [at least 1 urban cluster of
population 50,000], 1 million [reference], fringe/metropolitan
<1 million, micropolitan [at least 1 urban cluster of population
10,000e49,999], not metropolitan or micropolitan; Northeast,
Midwest, South, and West [reference]), teaching status (yes, no
[reference]), bed size (small [reference], medium, large), and patient
disposition at discharge. Race/ethnicity was reported by the
participating hospitals and coded by HCUP. Micropolitan counties
are defined as an urban cluster population between 10,000 and
49,999 people (Centers for Disease Control and Prevention, 2013).
The aforementioned covariates were chosen to determine if there are
www.jidonline.org 1395

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

racial, socioeconomic, and geographic disparities in SJS/TEN. A
chronic condition is defined as a condition that lasts for at least
12 months and meets at least one of the following criteria: it places
limitations on self-care or requires continuous treatment/therapy
(Healthcare Cost and Utilization Project, 2016). Identification of
chronic conditions is based on ICD-9-CM codes and on the methodology of Hwang et al. (2001), by which all ICD-9-CM diagnoses
codes were manually determined to be acute or chronic. Multivariate survey logistic regression models were constructed, and the
control group included all adult patients without any diagnosis of
SJS, SJS/TEN, or TEN, yielding a representative cohort of hospitalized
patients in the United States. Cases of normal pregnancy/birth were
excluded from the control group. To determine the predictors of cost
of care and LOS, multivariate linear regression models with the
abovementioned covariates were constructed with log-transformed
cost of care and LOS as the dependent variables. Cost of care and
LOS were log-transformed because they were not normally distributed. To determine the predictors of inpatient mortality among patients with SJS, SJS/TEN, or TEN, multivariate logistic regression
models with stepwise selection were constructed with mortality as
the dependent variable. Mortality analyses included only patients
with a diagnosis of SJS, SJS/TEN, and TEN. All aforementioned
sociodemographic factors and all comorbidities analyzed were
included in the mortality selection models. Analysis of variance was
used to compare incidence and mortality rates for SJS, SJS/TEN, or
TEN over time. Post hoc comparisons were performed between
2009, 2010, 2011, and 2012 using Tukeys method. Complete case
analysis was performed. A two-sided P-value less than 0.05 was
considered statistically significant.
CONFLICT OF INTEREST
The authors state no conflict of interest.

ACKNOWLEDGMENTS

Bellakhal S, Ben Kaab B, Teyeb Z, Souissi A, Derbel F, Douggui MH. Lupus

erythemateux systemique se presentant comme un syndrome de StevensJohnson [Systemic lupus erythematosus presenting as Stevens-Johnson
syndrome]. Pathologie-biologie 2015;63:222e3 [in French].
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and
powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol
1995;57:289e300.
Centers for Disease Control and Prevention. 2013 NCHS urbanerural
classification scheme for counties, http://www.cdc.gov/nchs/data_access/
urban_rural.htm; 2013 (accessed 1 November 2015).
Chung W-H, Hung S-I, Hong H-S, Hsih M-S, Yang L-C, Ho H-C, et al. (2004)
Medical genetics: a marker for Stevens-Johnson syndrome. Nature
2004;428:486.
Chung WH, Chang WC, Stocker SL, Juo CG, Graham GG, Lee MH, et al.
Insights into the poor prognosis of allopurinol-induced severe cutaneous
adverse reactions: the impact of renal insufficiency, high plasma levels of
oxypurinol and granulysin. Ann Rheum Dis 2015;74:2157e64.
Davis RL, Gallagher MA, Asgari MM, Eide MJ, Margolis DJ, Macy E, et al.
Identification of Stevens-Johnson syndrome and toxic epidermal necrolysis
in electronic health record databases. Pharmacoepidemiol Drug Saf
2015;24:684e92.
Gravante G, Delogu D, Marianetti M, Esposito G, Montone A. Toxic
epidermal necrolysis and Steven-Johnson syndrome in oncologic patients.
Eur Rev Med Pharmacol Sci 2007;11:269e74.
Healthcare Cost and Utilization Project. Chronic Condition Indicator (CCI) for
ICD-9-CM, http://www.hcup-us.ahrq.gov/toolssoftware/chronic/chronic.
jsp; 2016 (accessed 1 November 2015).
HCUP Nationwide Inpatient Sample (NIS). Healthcare Cost and Utilization
Project (HCUP). Agency for Healthcare Research and Quality, Rockville,
MD, www.hcup-us.ahrq.gov/nisoverview.jsp; 2009e2012 (accessed
November 1, 2015).
Hsu D, Brieva J, Silverberg JI. Costs of care for hospitalization for pemphigus
in the United States [e-pub ahead of print]. Arch Dermatol 2016; http://dx.
doi.org/10.1001/jamadermatol.2015.5240 (accessed 1 December 2015).
Hsu DY, Brieva J, Sinha AA, Langan S, Silverberg JI. Comorbidities and
inpatient mortality for pemphigus in the United States [e-pub ahead of
print]. Br J Dermatol 2016; http://dx.doi.org/10.111/bjd.14463 (accessed 1
December 2015).

The results of this study were presented in part at the 2016 Society
for Investigative Dermatology annual meeting in Scottsdale, Arizona. This
publication was made possible with support from the Agency for Healthcare
Research and Quality, grant number K12HS023011, and the Dermatology
Foundation.

Hung CC, Liu WC, Kuo MC, Lee CH, Hwang SJ, Chen HC. Acute renal failure
and its risk factors in Stevens-Johnson syndrome and toxic epidermal
necrolysis. Am J Nephrol 2009;29:633e8.

Disclaimer

Imahara SD, Holmes JHT, Heimbach DM, Engrav LE, Honari S, Klein MB,
et al. SCORTEN overestimates mortality in the setting of a standardized
treatment protocol. J Burn Care Res 2006;27:270e5.

The sponsor had no role in the design and conduct of the study; collection,
management, analysis and interpretation of data; preparation, review, or
approval of the manuscript; or decision to submit the manuscript for
publication.

AUTHOR CONTRIBUTIONS
JIS had full access to all the data in the study and takes responsibility for the
integrity of the data and accuracy of the data analysis. JIS and DH were
responsible for the study concept and design. JIS and DH were responsible for
acquisition of data. JIS, DH, NBS, and JB were responsible for analysis and
interpretation of data. JIS, DH, NBS, and JB were responsible for drafting the
manuscript. JIS, DH, NBS, and JB critically reviewed the manuscript for
important intellectual content. JIS and DH were responsible for statistical
analysis. JIS obtained funding.

SUPPLEMENTARY MATERIAL

Hwang W, Weller W, Ireys H, Anderson G. Out-of-pocket medical spending

for care of chronic conditions. Health Aff (Millwood) 2001;20:267e78.

Ishida T, Ito A, Sato F, Kusumoto S, Iida S, Inagaki H, et al. Stevens-Johnson

Syndrome associated with mogamulizumab treatment of adult T-cell
leukemia/lymphoma. Cancer Sci 2013;104:647e50.
Jung SO, Cho MJ, Park DI, Kim SY, Kim JO, Jung SS, et al. Stevens-Johnson
Syndrome and hypothermia associated with anti-tuberculosis medication
in a patient with heart failure. Chinese Med J 2015;128:2837e8.
Kakar R, Paugh H, Jaworsky C. Linear IgA bullous disease presenting as
toxic epidermal necrolysis: a case report and review of the literature.
Dermatology 2013;227:209e13.
Kaniwa N, Saito Y, Aihara M, Matsunaga K, Tohkin M, Kurose K, et al. HLA-B
locus in Japanese patients with anti-epileptics and allopurinol-related
StevenseJohnson syndrome and toxic epidermal necrolysis. Pharmacogenomics 2008;9:1617e22.

REFERENCES

Kim H-I, Kim S-W, Park G-Y, Kwon E-G, Kim H-H, Jeong J-Y, et al.
Causes and treatment outcomes of Stevens-Johnson syndrome and
toxic epidermal necrolysis in 82 adult patients. Korean J Int Med 2012;
27:203e10.

Bastuji-Garin S, Fouchard N, Bertocchi M, Roujeau JC, Revuz J,

Wolkenstein P. SCORTEN: a severity-of-illness score for toxic epidermal
necrolysis. J Invest Dermatol 2000;115:149e53.

Lee HY, Tey HL, Pang SM, Thirumoorthy T. Systemic lupus erythematosus
presenting as Stevens-Johnson syndrome and toxic epidermal necrolysis: a
report of three cases. Lupus 2011;20:647e52.

Bastuji-Garin S, Rzany B, Stern RS, Shear NH, Naldi L, Roujeau JC. Clinical
classification of cases of toxic epidermal necrolysis, Stevens-Johnson
syndrome, and erythema multiforme. Arch Dermatol 1993;129:92e6.

Letko E, Papaliodis DN, Papaliodis GN, Daoud YJ, Ahmed AR, Foster CS.
Stevens-Johnson syndrome and toxic epidermal necrolysis: a review of the
literature. Ann Allergy Asthma Immunol 2005;94:419e36.

Supplementary Material is linked to the online version of the paper at www.

jidonline.org, and at http://dx.doi.org/10.1016/j.jid.2016.03.023.

1396 Journal of Investigative Dermatology (2016), Volume 136

DY Hsu et al.

Morbidity and Mortality of SJS and TEN

Mahar PD, Wasiak J, Hii B, Cleland H, Watters DA, Gin D, et al. A systematic
review of the management and outcome of toxic epidermal necrolysis
treated in burns centres. Burns 2014;40:1245e54.
Metry DW, Lahart CJ, Farmer KL, Hebert AA. Stevens-Johnson syndrome caused
by the antiretroviral drug nevirapine. J Am Acad Dermatol 2001;44:354e7.
Mittmann N, Knowles SR, Koo M, Shear NH, Rachlis A, Rourke SB. Incidence
of toxic epidermal necrolysis and Stevens-Johnson syndrome in an HIV
cohort: an observational, retrospective case series study. Am J Clin Dermatol 2012;13:49e54.
Mockenhaupt M, Viboud C, Dunant A, Naldi L, Halevy S, Bavinck JNB, et al.
Stevens-Johnson syndrome and toxic epidermal necrolysis: assessment
of medication risks with emphasis on recently marketed drugs. The EuroSCAR-Study. J Invest Dermatol 2007;128:35e44.
Mulvey JM, Padowitz A, Lindley-Jones M, Nickels R. Mycoplasma pneumoniae associated with Stevens Johnson syndrome. Anaesth Intensive Care
2007;35:414e7.
Naldi L, Locati F, Marchesi L, Cainelli T. Incidence of toxic epidermal
necrolysis in Italy. Arch Dermatol 1990;126:1103e4.
Oen IM, van der Vlies CH, Roeleveld YW, Dokter J, Hop MJ, van Baar ME.
Epidemiology and costs of patients with toxic epidermal necrolysis: a 27-year
retrospective study. J Eur Acad Dermatol Venereol 2015;29:2444e50.
Palmieri TL, Greenhalgh DG, Saffle JR, Spence RJ, Peck MD, Jeng JC, et al.
A multicenter review of toxic epidermal necrolysis treated in U.S. burn
centers at the end of the twentieth century. J Burn Care Rehabil 2002;23:
87e96.

Rzany B, Mockenhaupt M, Baur S, Schroder W, Stocker U, Mueller J, et al.

Epidemiology of erythema exsudativum multiforme majus, StevensJohnson syndrome, and toxic epidermal necrolysis in Germany (19901992): structure and results of a population-based registry. J Clin Epidemiol
1996;49:769e73.
Schneck J, Fagot J-P, Sekula P, Sassolas B, Roujeau JC, Mockenhaupt M.
Effects of treatments on the mortality of Stevens-Johnson syndrome
and toxic epidermal necrolysis: A retrospective study on patients
included in the prospective EuroSCAR Study. J Am Acad Dermatol
2008;58:33e40.
Schopf E, Stuhmer A, Rzany B, Victor N, Zentgraf R, Kapp JF. Toxic epidermal
necrolysis and Stevens-Johnson syndrome. An epidemiologic study from
West Germany. Arch Dermatol 1991;127:839e42.
Sekula P, Dunant A, Mockenhaupt M, Naldi L, Bouwes Bavinck JN, Halevy S,
et al. Comprehensive survival analysis of a cohort of patients with
StevenseJohnson syndrome and toxic epidermal necrolysis. J Invest
Dermatol 2013;133:1197e204.
Strom BL, Carson JL, Halpern AC, Schinnar R, Snyder ES, Shaw M, et al.
A population-based study of Stevens-Johnson syndrome: Incidence and
antecedent drug exposures. Arch Dermatol 1991;127:831e8.
Tay Y-K, Huff JC, Weston WL. Mycoplasma pneumoniae infection is
associated with Stevens-Johnson syndrome, not erythema multiforme (von
Hebra). J Am Acad Dermatol 1996;35:757e60.
Torchia D, Romanelli P, Kerdel FA. Erythema multiforme and Stevens-Johnson
syndrome/toxic epidermal necrolysis associated with lupus erythematosus.
J Am Acad Dermatol 2012;67:417e21.

Rosen AC, Balagula Y, Raisch DW, Garg V, Nardone B, Larsen N, et al. Lifethreatening dermatologic adverse events in oncology. Anti-cancer Drug
2014;25:225e34.

United States Census Bureau. CPS Data on Age and Sex, https://www.census.
gov/population/age/data/cps.html; 2013 (accessed November 1, 2015).

Roujeau JC, Chosidow O, Saiag P, Guillaume JC. Toxic epidermal necrolysis

(Lyell syndrome). J Am Acad Dermatol 1990;23:1039e58.

United States Department of Labor. Consumer price index, http://www.bls.

gov/cpi/; 2014 (accessed 1 November 2015).

Roujeau JC, Guillaume JC, Fabre JP, Penso D, Flechet ML, Girre JP. (1990b)
Toxic epidermal necrolysis (Lyell syndrome). Incidence and drug etiology
in France, 1981-1985. Arch Dermatol 1990;126:37e42.

Wirima JJ, Harries AD. Stevens-Johnson syndrome during anti-tuberculosis

chemotherapy in HIV-seropositive patients: report on six cases. East Afr
Med J 1991;68:64e6.

www.jidonline.org 1397