Healthcare utilization and cost of Stevens-Johnson syndrome and toxic epidermal necrolysis management in Thailand.

BACKGROUND: Stevens-Johnson syndrome (SJS) and Toxic Epidermal Necrolysis (TEN) are life-threatening dermatologic conditions. Although, the incidence of SJS/TEN in Thailand is high, information on cost of care for SJS/TEN is limited. This study aims to estimate healthcare resource utilization and cost of SJS/TEN in Thailand, using hospital perspective.
METHODS: A retrospective study using an electronic health database from a university-affiliated hospital in Thailand was undertaken. Patients admitted with SJS/TEN from 2002 to 2007 were included. Direct medical cost was estimated by the cost-to-charge ratio. Cost was converted to 2013 value by consumer price index, and converted to $US using 31 Baht/ 1 $US. The healthcare resource utilization was also estimated.
RESULTS: A total of 157 patients were included with average age of 45.3±23.0 years. About 146 patients (93.0%) were diagnosed as SJS and the remaining (7.0%) were diagnosed as TEN. Most of the patients (83.4%) were treated with systemic corticosteroids. Overall, mortality rate was 8.3%, while the average length of stay (LOS) was 10.1±13.2 days. The average cost of managing SJS/TEN for all patients was $1,064±$2,558. The average cost for SJS patients was $1,019±$2,601 while that for TEN patients was $1,660±$1,887.
CONCLUSIONS: Healthcare resource utilization and cost of care for SJS/TEN in Thailand were tremendous. The findings are important for policy makers to allocate healthcare resources and develop strategies to prevent SJS/TEN which could decrease length of stay and cost of care.

Introduction

Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are life-threatening dermatologic conditions, which predominantly involve the skin and mucous membranes. Both are rare but very severe conditions. SJS is characterized by high fever, malaise, and a rapidly developing blistering exanthema of macules and target-like lesion, accompanied by mucosal involvement, whereas TEN has similar presentations with even more extensive skin detachment and a higher mortality rate.[1] Although the incidence of SJS/TEN is as low as 0.4-6 persons per million populations each year,[123] their mortality rate has been estimated to be 1-50%.[4567] Moreover, about 30-70% of the surviving patients were suffering from long-term sequelae such as severe dry eye syndrome.[4567] There are many causes of SJS/TEN such as chemical exposures, mycoplasma pneumonia, viral infections, and immunizations but the most common causes are medications.[1]

The management of SJS/TEN requires substantial health care resources including prolonged hospitalization, aggressive fluid replacement, nutrition support, and antibacterial treatments.[1] Studies that have examined resource consimption due to SJS/TEN are limited. A recent study was conducted in India to investigate health care utilization and cost among 32 SJS/TEN cases.[8] The length of hospital stay (LOS) for SJS and TEN was reported to be 9.7 days [95% confidence interval (CI): 5.8-13.6] and 20.6 days (95% CI: 14.4-26.8), respectively. The average direct medical cost of care for SJS and TEN was approximately $38.4 and $123.6 per person, respectively.[8]

In Thailand, the incidence of drug-induced SJS/TEN and mortality rate of these conditions is high. The incidence of SJS/TEN are 10.35 per 1,000 persons while the mortality rate in the Thai population is 7-50%.[9] Since the incidence and mortality rate of SJS/TEN is high relative to other countries, it is important for clinicians and policymakers to understand the economic implications of these conditions. Such information on cost and health care resource utilization hence the present study was carried out.

Materials and Methods

Data source

A descriptive retrospective study was undertaken to estimate health care resource utilization and cost of care for SJS/TEN using an electronic database from a 1,000-bed regional hospital in northern Thailand from a hospital perspective. The database included inpatient database, pharmacy database, and charge database. The inpatient database contained information on age, gender, date of birth, health care insurance, and diagnosis code [International Classification of Diseases (ICD), 10th Revision], type of diagnosis (primary or secondary diagnosis), adjusted relative weight for each admission, date of admission, date of discharge, and discharge status. The pharmacy database was enriched with information on the generic names of drugs, regimen, and the days when these were supplied. The charge database included information on drugs, laboratory, medical service, and other service charges. The protocol was approved by Naresuan University Institutional Review Board (NUIRB) and the university-affiliated hospital and confidentiality of patients was maintained using unique identifiers.

Study population

All patients who were admitted to the study hospital from 2002 to 2007 with a diagnosis of SJS or TEN were included. The SJS and TEN were defined by ICD-10 as L51.1 for SJS and L51.2 for TEN. The patients who met the inclusion criteria were tracked during the hospitalization episode of SJS/TEN which was from the date of admission to the date of discharge.

The data of all included patients were collected as follows: demographic data, comorbidities, length of stay, discharge status, treatment patterns for SJS/TEN, and cost of SJS/TEN care.

Outcome measures

The primary outcome measure was direct medical cost of care for SJS, TEN, and SJS/TEN. It was determined using the cost-to-charge ratio (CCR) technique as a primary analysis. The direct medical cost was calculated based on CCR technique using the equation 1.

Σ[(charge × CCR) × CPIx/CPI2013]      —Equation 1

Number of patients

where CCR stands for ratio of cost-to-charge in each year

CPIx stands for consumer price index for the year of data used

CPI2013 stands for consumer price index for the year 2013

In a sensitivity analysis, diagnosis-related group (DRG) technique was used as an alternative method to estimate the direct medical cost using equation 2.

Σ[(BR × AdjRW) × CPIx/CPI2013]      —Equation 2

Number of patients

where BR stands for base rate for a hospitalization which is ฿10,000 (at year 2007)

AdjRW stands for adjusted relative weight

CPIx stands for consumer price index for the year of data used CPI2013 stands for consumer price index for the year 2013

Briefly, the CCR technique is used to estimate cost of disease management when the actual cost is not available. Estimated cost is calculated by multiplying charge, which is usually recorded for reimbursement purpose with CRR. The CRR is usually calculated by the hospital for administrative purpose. In our study, CCR was calculated by the hospital and was recorded each year while charges of each health care section (such as drug, laboratory, medical service, and other service charges) were recorded in the charge database. Charge in our study means a payment that the hospital will charge payers with if the payment mechanism is fee-for-service. We calculated cost of care for SJS/TEN by multiplying charges recorded in the charge database with CCR of each year, which was provided by hospital staff. The CCR was 0.77 for 2008 and 0.90 for 2007, 2009, and 2010. For example, a total charge of an admission in 2008 was ฿20,000. A total cost of the admission was ฿15,400.

For DRG technique, we calculated cost of care for SJS/TEN by multiplying adjusted relative weight (adjRW) of each DRG with base rate of an admission, which was an amount of money that payers pay for one adjRW. In our study, we used ฿10,000 for one adjRW. AdjRW for each admission was calculated by hospital administrators and recorded in the inpatient database using DRG technique. For example, adjRW of an admission was 2.301. Cost of the admission was ฿23,010.

All costs were converted to the value in the year 2013 using an average consumer price index in the medical and personal care sector and converted to $US using an average exchange rate of ฿31/$US.[10] Secondary outcome measures are length of stay (LOS), discharge status, and patterns of medication use for treatment.

Data analysis

Descriptive statistics were used to describe patients’ demographics and outcome measures. Quantitative variables such as age, direct medical cost, and frequency of resource utilization were tested for normal distribution by visualized histograms. The means and standard deviations were used when data were normally distributed while medians and interquartile ranges (IQRs) were used when the data were not normally distributed. Categorical variables such as gender, comorbidity, and discharge status were described by their frequencies and percentages. Subgroup analyses were also performed by separation of SJS and TEN patients, and the types of diagnoses were classified as primary diagnosis and secondary diagnosis of SJS/TEN. Primary diagnosis was identified as a diagnosis contributing the most to the whole hospital stay of the patient while secondary diagnoses were other important diagnoses in the hospital stay such as comorbidities or complications. STATA package version 11.0 (StataCorp®, College Station, TX, USA) was used for all analyses.

Results

Demographic data

A total of 157 patients with 56.1% males were included. Average age of the patients was 45.3 ± 23.0 years. One hundred and forty-six patients (93.0%) were diagnosed as SJS and the remaining patients were diagnosed as TEN. Of those patients, 110 patients were primarily diagnosed as SJS and 8 patients were diagnosed as TEN while 36 patients were secondarily diagnosed as SJS, and 3 patients were secondarily diagnosed as TEN. Infectious, endocrine, and cardiovascular diseases are the most common co-morbidities seen in the patients [Table 1].

Table 1

Patient demographics

Demographic data	Numbers (%)
	SJS (N = 146; primary diagnosis = 110)	TEN (N = 11; primary diagnosis = 8)	Total (SJS/TEN) (N = 157)
Age (years, mean±SD)	44.5±22.5	58.1±29.2	45.5±23.2
<18	19 (13.0)	1 (9.1)	20 (12.7)
18–60	85 (58.2)	4 (36.4)	89 (56.7)
>60	42 (28.8)	6 (54.5)	48 (30.6)
Gender
Male	79 (54.1)	5 (45.5)	84 (53.5)
Female	67 (45.9)	6 (54.6)	73 (46.5)
Health insurance
CSMBS	21 (14.4)	2 (18.2)	23 (14.7)
UC	113 (77.4)	9 (81.8)	122 (77.7)
SSS	11 (7.5)	0 (0)	11 (7.0)
Unknown	1 (0.7)	0 (0)	1 (0.6)
Comorbidities
Infectious diseases	54 (37.0)	2 (18.2)	56 (35.7)
Neoplasm	5 (3.4)	1 (9.1)	6 (3.8)
Blood diseases	15 (10.3)	3 (27.3)	18 (11.5)
Endocrine diseases	31 (21.2)	2 (18.2)	33 (21.0)
Mental diseases	4 (2.7)	1 (9.1)	5 (3.2)
Nervous diseases	13 (8.9)	0 (0)	13 (8.3)
Cardiovascular diseases	29 (19.9)	1 (9.1)	30 (19.1)
Respiratory diseases	20 (13.7)	1 (9.1)	21 (13.4)
Gastrointestinal diseases	9 (6.2)	1 (9.1)	10 (6.4)
Skin diseases	19 (13.0)	1 (9.1)	20 (12.7)
Musculoskeletal diseases	12 (8.2)	1 (9.1)	13 (8.3)
Genitourinary diseases	26 (17.8)	2 (18.2)	28 (17.8)

CSMBS = Civil servant medical benefit schemes, SD = Standard deviation, SSS = Social security schemes, UC = Universal coverage, SJS = Stevens–Johnson syndrome, TEN = Toxic epidermal necrolysis, IQR = Interquartile range

Treatment patterns

The medications used during the treatment of SJS/TEN can be categorized into two groups: Supportive and specific treatments. For supportive treatment, acid-suppressing drugs [H2-blockers, proton pump inhibitor (PPI)], antianxiety drugs, antihistamine drugs, anesthetic drugs, and analgesic drugs were used. The most common drugs for supportive treatment were antihistamine (91.1%) and analgesic drugs (73.2%). Various types of steroids were used as specific treatment. Most patients (83.4%) received systemic corticosteroids. Almost half of the patients also received ophthalmic steroids (43.3%) [Table 2]. Prednisolone was commonly used as an oral medication while dexamethasone was usually used as an injection medication. Moreover, dexamethasone was the most commonly used among ophthalmic steroids.

Table 2

Medication utilization for SJS/TEN

Drug therapy	Number of patients (%)
	SJS (N = 146)	TEN (N = 11)	Total (SJS/TEN) (N = 157)
Supportive drugs
Acid-suppressing drugs	63 (43.2)	5 (45.5)	68 (43.3)
Antianxiety drugs	29 (19.9)	1 (9.1)	30 (19.1)
Antihistamine drugs	133 (91.1)	10 (90.9)	143 (91.1)
Anesthetic drugs	88 (60.3)	5 (45.5)	93 (59.2)
Analgesic drugs	105 (71.9)	10 (90.9)	115 (73.2)
Specific drugs
Systemic steroids	121 (82.9)	10 (90.9)	131 (83.4)
Oral drugs (tablet)	92 (63.0)	4 (36.4)	96 (61.1)
IV injection drugs	105 (71.9)	10 (90.9)	115 (73.2)
Local steroids
Eye drugs	61 (41.8)	7 (63.6)	68 (43.3)
Topical drugs (cream)	3 (2.1)	0 (0)	3 (1.9)
Oral paste	12 (8.2)	0 (0)	12 (7.6)

SJS = Stevens–Johnson syndrome, TEN = Toxic epidermal necrolysis

Health care resource utilization and cost of care

Most of the patients recovered after the treatment (72.42%) while 17.8%, 1.9%, and 8.3% of the patients were referred to other treatments, referred against the treatment, and died, respectively. The average LOS was 10.1 ± 13.2 days. The average LOS was 10.2 ± 13.6 days in patients with SJS, while it was 8.5 ± 6.2 days in patient with TEN. The case-fatality ratio of patients with TEN (18.2%) was higher than that of patients with SJS (7.5%) [Table 3]. The average LOS in patients with primary diagnosis as SJS/TEN was 6.8 ± 4.8 days while that in patients with secondary diagnosis as SJS/TEN was 20.2 ± 22.5 days. The case-fatality ratio of patients with primary and secondary diagnoses was 3.4% and 23.1%, respectively [Table 3].

Table 3

Length of stay and discharge status based on types of diagnoses

Measures	Types of diagnosis (SJS/TEN)		Types of diagnosis (1º or 2º diagnosis)		Total (SJS/TEN) (N =157)
	SJS (N = 146)	TEN (N = 11)	Primary diagnosis (N = 118)	Secondary diagnosis (N = 39)
Length of stay; day
Mean±SD	10.2±13.6	8.5±6.2	6.8±4.8	20.2±22.5	10.1±13.2
(Median, IQR)	6 (4-11)	6 (4-13)	5 (4-9)	12 (4-27)	6 (4-11)
Discharge status (N, %)
Cure	107 (73.3)	6 (54.7)	95 (80.5)	18 (46.2)	113 (72.0)
Against treatment	1 (0.7)	2 (18.2)	2 (1.7)	1 (2.6)	3 (1.9)
Refer	27 (18.5)	1 (9.1)	17 (14.4)	11 (28.2)	28 (17.8)
Dead	11 (7.5)	2 (18.2)	4 (3.4)	9 (23.1)	13 (8.3)

IQR = Interquartile range, SD = Standard deviation, SJS = Stevens-Johnson syndrome, TEN = Toxic epidermal necrolysis

Overall, based on CCR technique the average cost of care for SJS/TEN was $1,064 ± $2,558 with the median cost of $369 [IQR; $194-870]. The average cost estimated by CCR was higher than the cost estimated by DRG ($1,064 ± $2,558 vs $617 ± $1,224) [Table 4].

Table 4

Direct medical cost of care for SJS/TEN management by diagnosis (US$)

Cost (US$)	Cost-to-charge ratio			Diagnosis-related group
	SJS (N = 146)	TEN (N = 11)	Total (N =157)	SJS (N = 146)	TEN (N = 11)	Total (SJS/TEN) (N =157)
Mean (SD)	1,019 (2,601)	1,660 (1,887)	1,064 (2,558)	631 (1,265)	438 (327)	617 (1,224)
Median (IQR)	357 (189–836)	831 (263–3,331)	369 (194–870)	293 (241–524)	317 (238–395)	294 (241–516)

IQR = Interquartile range, SD = Standard deviation, SJS = Stevens-Johnson syndrome, TEN = Toxic epidermal necrolysis

The average cost of care for patients with SJS was $1,019 ± $2,601 while that for patients with TEN was $1,660 ± $1,887 [Table 4]. The average cost of care for cases with SJS/TEN as secondary diagnosis ($2,600 ± $4,658) was higher than that for primary diagnosis ($556 ± $797) [Table 5]. A major driver of cost of SJS/TEN care was medication cost, which accounted for 41% of the direct medical cost. The service cost, diagnosis cost, and other costs were 22%, 20%, and 17%, respectively.

Table 5

Direct medical cost of care for SJS/TEN management by types of cost estimation (US$)

Cost (US$)	Cost-to-charge ratio		Diagnosis-related group
	Primary diagnosis (N = 118)	Secondary diagnosis (N = 39)	Primary diagnosis (N = 118)	Secondary diagnosis (N = 39)
Mean (SD)	556 (797)	2,600 (4,658)	390 (280)	1,304 (2,295)
Median (IQR)	314 (173–572)	1,242 (347–2,115)	245 (241–512)	594 (255–1,154)

IQR = Interquartile range, SD = Standard deviation

Discussion

To the best of our knowledge, this study is one of the largest studies describing health care utilization and cost for SJS/TEN management. Previous studies[81112] have been conducted in less than 50 patients while this study was undertaken in more than 150 patients. The results in this study showed crucial information to health policymakers so as to allocate health care resources to manage SJS/TEN at the population level efficiently. Moreover, the findings raised awareness of clinicians on the importance of SJS/TEN and its associated substantial economic loss.

The average LOS found in this study was about 10 days with case-fatality ratio of 8%. These findings were lower than those reported in a previous Thai study, which revealed that the average LOS was 19.2 days with case-fatality ratio of 14%.[12] The difference was likely due to the severity and complexity of included patients. It is a fact that TEN is more severe than SJS. Thus, patients with TEN are likely to have a longer length of stay than patients with SJS. This finding might be affected by a smaller group of patients, as only 7% were diagnosed with TEN in this study while it was 15.7% in the previous study. There was another Thai study reporting the average LOS of patients with SJS to be 5.3 days.[11] Even though the study used a national database from the National Health Security Office, only 15 patients with SJS were determined. Thus, the average LOS from the study might be inaccurate and imprecise. In addition, our finding indicated longer LOS in patients with SJS than patients with TEN. This could be explained by the fact that patients with TEN have higher case-fatality ratio and may have died in short period of time, leading to shorter average LOS compared to that of patients with SJS. To further address this issue, we performed a post hoc subgroup analysis to explore the LOS of patients who were cured from SJS or TEN. We found that LOS of SJS survivors was 9.16 ± 9.92 days while the LOS of TEN survivors was 9.33 ± 8.09 days. The LOS of TEN patients was longer than that of SJS patients. This supported our discussion point that the reason for shorter LOS of TEN patients compared to SJS patients might be due to higher case-fatality ratio.

Our findings revealed patterns of SJS/TEN treatment. The majority of patients, which was more than 80% of hospitalized patients, received systematic corticosteroids, along with supportive care. This was similar to a previous Thai study, which indicated that the majority of patients (76.5%) admitted to a university hospital received intravenous corticosteroids.[12] It is important to note that there remains a lack of evidence demonstrating the clear benefits of using corticosteroids for SJS/TEN management.[6] A number of observational studies found no beneficial effect of systemic corticosteroid compared to supportive care management.[1314151617] To date, several specific medications (such as intravenous immunoglobulin, thalidomide, cyclosporine, and cyclophosphamide) have been used to treat SJS/TEN without strong supporting evidence.[6]

In this study, we observed $1,000 spending per admission. The spending was 19.4% of gross domestic product per capita, which could be considered relatively high.[18] Costs in patients with TEN and patients with secondary diagnosis were higher than those patients with SJS and patients with secondary diagnosis. The accuracy of cost estimate is important, especially when it is applied to health economic evaluation. Sources of data and the number of patients included in the study reflected the accuracy of health care cost estimation. There were three previous cost-effectiveness studies related to SJS/TEN in Thailand that used different sources of cost for SJS/TEN management.[192021] First, a study from Rattanavipapong et al.[19] estimated the cost of carbamazepine (CBZ)-induced SJS/TEN in 15 adult patients using a multicenter retrospective database analysis. The authors found that the average direct medical cost of CBZ-induced SJS/TEN was ฿25,868 ($834). The second study was conducted by Tiamkao et al., which included 51 SJS/TEN patients.[21] The cost of SJS/TEN management in the study was obtained from reimbursement records of a university hospital. They found that the average cost of SJS/TEN management was ฿53,858 ($1,737). The last study was conducted by Saokaew et al., which included 151 patients.[20] The cost of SJS/TEN was estimated using a retrospective database analysis from a tertiary care hospital (1,000-bed capacity). They found that the average direct medical cost was ฿15,440 ($498). The first two studies[1921] provided insufficient description of cost estimation. Most importantly, the costs were obtained from a limited sample size. The last study[20] used the cost, which was calculated based on the same source of this study. However, the description of cost estimation was not comprehensive.

There was another study formally reporting direct medical cost of patients with SJS.[11] The direct medical cost ranged $242.8-674.7 based on health insurance. The previous direct medical cost was lower than the one found in this study, which indicated that the average direct medical cost was $1,064. It is important to note that the result from the previous study was based on only 15 patients with SJS; its accuracy and precision were questionable.

In our study, we descriptively compared the different approaches, which were the CCR and DRG approaches to determine cost. Findings indicated the cost derived by CCR was higher than that derived by DRG. The CCR should be applied when cost estimation is based on a hospital perspective. This is because the CCR reflects better the average cost incurred in a hospital than DRG. However, DRG should be applied when cost estimation is based on the payer's perspective because DRG was developed for reimbursement purpose. DRG reflects better the cost paid by the government than the CCR. Thus, to select an appropriate method for cost analysis studies, the study perspective should be carefully considered.

Our subgroup analysis revealed longer LOS (6.8 ± 4.8 days vs 20.2 ± 22.5) in patients with secondary diagnosis. Furthermore, our finding indicated higher cost in patients with secondary diagnosis as SJS/TEN than that in patients with primary diagnosis as SJS/TEN ($556 ± $797 vs $2,600 ± $4,658). As per the definition used in our study, primary diagnosis was the major diagnosis of the admission or visit while secondary diagnosis was comorbidity or complications. This implies that patients with SJS/TEN as secondary diagnosis would have other diseases, which might contribute to the LOS. The conditions of patients with secondary diagnosis as SJS/TEN were likely to be more complicated than of patients with primary diagnosis as SJS/TEN. This leads to longer LOS and higher cost in patients with secondary diagnosis as SJS/TEN than in patients with primary diagnosis as SJS/TEN.

This study has several limitations. First, our study was conducted in a regional hospital; generalizability should be limited to hospitals with characteristics similar to this tertiary study setting. Diagnostic testing strategies or treatment patterns for SJS/TEN management of this hospital might be different from those of other settings, which may increase the cost of SJS/TEN management. It is important to note that most patients with SJS/TEN are less likely to be hospitalized in a secondary or primary care setting. Second, misclassification might have occurred since these findings were based on only electronic databases. Third, the estimated costs in this study were based on costs incurred during the hospitalization of patients with SJS/TEN. Other treatment costs that patients might have sought were not captured. In addition, the study was undertaken from a hospital perspective; direct nonmedical cost and indirect cost, especially the loss of productivity were not included. Further studies determining the total cost of SJS/TEN from the societal perspective are still required. Lastly, the estimated costs and health care utilization were the overall estimates. There might be some differences in a different level of care such as the intensive care unit and general wards. However, we could not estimate the costs and health care utilization by the level of care because of the limitation on data availability. Further studies determining such costs by the level of care are still required.

Conclusion

In conclusion, this study demonstrates the significance of the SJS/TEN using Thailand as a case study. There remains a need for estimating the costs of SJS/TEN management in other developing countries, especially in countries where the prevalence of SJS/TEN is high. Our findings should be used as a key input parameter for any future cost-effectiveness studies related to SJS/TEN in Thailand. Moreover, policymakers should be concerned and use these findings as an important information for efficient resource allocation. In addition, these findings should be promoted among clinicians to enhance awareness of cost-related SJS/TEN and to operate effective actions to prevent SJS/TEN from adverse drug events.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.