Treatment costs for patients with Streptococcus suis infection in Northern Thailand: a hospital-based observational study of 14-year data

This was a cross-sectional study of the 14-year data at Chiang Mai University Hospital (CMUH), a 1400-bed tertiary teaching hospital during 2005–2018, the largest hospital in northern Thailand. The hospital’s electronic database was used to retrieve healthcare utilisation of S.suis positive patients admitted from January 2005 to December 2018 with confirmed positive S.suis either by cerebrospinal fluid (CSF) or hemoculture. The index date was the hospital admission date. S.suis cases with specimens taken from CSF or hemoculture were identified from the hospital laboratory records of microbiological proven S.suis patients with their hospital numbers (HNs). All confirmed S.suis cases with complete medical and healthcare resource utilization records who had been admitted from 1 May 2005 to 31 December 2018 were collected and reviewed. All costs were derived from hospital charges provided by the Office of Medical Records and Statistics, Chiang Mai University Hospital, to estimate the direct medical costs associated with treatment during the episode.

The study was approved by the Research Ethics Committee, Faculty of Medicine, Chiang Mai University (IRB no.010/2018) and Monash University Human Research Ethics Committee (MUHREC) (Project no.12225). Patients’ names, and personal and traceable information were omitted and treated as confidential in all study procedures. The informed consent was waived due to the retrospective nature of the study.

This study considered the healthcare provider’s perspective and included only direct medical costs associated with S.suis treatment which is generally fully paid by the provider in Thailand’s healthcare system. Direct medical costs included medication, laboratory tests, and others, e.g., room charges, meals, medical staff services, and medical devices during hospital admission. All costs were inflated to the 2019 value using Consumer Price Index with the medical care section [10]. The average market exchange rate in 2019 was US$ = 31.0470 Thai baht (THB) [11].

Descriptive statistics were used to summarize the data on patients’ characteristics, healthcare utilization and costs. Multiple imputations with predictive mean matching on log-transformed costs were employed to deal with missing data [12] which was Glasgow Coma Score (GCS) data. Due to the skewness of the cost data, Generalized linear models (GLMs) were used to forecast costs models and identify determinants of costs [13] associated with S.suis treatment. In addition, the models had the advantage of displaying both the mean and variance functions directly based on the original scale. Therefore, there would not be any issue concerning retransformation. Potential multicollinearity was examined before building the final model. The modified Parks test was adopted to determine the appropriate family for healthcare expenditures based on the relationship between raw scale means and variance function [13]. The interpretation for family was based on the coefficients for lnyhat: 0 Gaussian test lnyhat = 0, 1 Poisson test lnyhat = 1, 2 Gamma test lnyhat = 2, 3 Inverse Gaussian or Wald test lnyhat = 3 [14]. After the model development, the bootstrapping technique was applied to estimate bootstrap-corrected calibration coefficients for internal validation [15]. Significant predictors and costs were illustrated as marginal estimates for ease of interpretation. All analyses were executed using Stata/IC version 16.0 for Windows (StataCorp LP, College Station, TX, USA).

A total of 130 confirmed S.suis patients with complete medical records and hospital charges data were eligible based on the inclusion criteria. Among these, there were 89 males and 41 females. However, Glasgow Coma Score (GCS) data and length of stay were available only among 101 and 128 patients respectively. The mean age was 56.32 ± 13.77 years. The significant risk behaviors included alcohol drinking (50.77%) and eating raw pork (37.69%). Septicaemia was the most common clinical presentation (56.15%), followed by meningitis (36.92%) and infective endocarditis (IE) (26.15%) (Table 1). Approximately half of S. suis infected patients recovered from their infection (52.31%), while around one-third recovered with sequelae (33.08%). The average admission duration was 18 days. The characteristics of S.suis patients are presented in Table 1.

The cost data was right skewed with a long heavy right tail reflecting substantial more services required among severe cases and higher costs in relatively small number of patients (See Figure S1 in Online supplementary material). Total treatment cost of S.suis was estimated to be 16,207,676 (US$ 522,037.12) in 2019 (Table 2). The mean cost of medications, laboratory tests, X-ray, and others (room charges, meals, staff services, medical devices) for S.suis treatment paid by the hospital in 2019 value were 30,322 (US$ 977), 13,187 (US$ 425), 8,240 (US$ 265), and 72,926 (US$ 2,349), respectively. Majority of expenses were from “others” category. The average total treatment cost per episode was 124,675 (US$ 4,016). After removing patients who died at the hospital and rerunning the analysis, the similar trends were observed while the mean length of stay was 19.25 (SD = 17.62) days (See Tables S1 and S2 in Online supplementary material).

A bootstrap of 1000 sampling with replacement was performed for interval validation. The bootstrap-corrected calibration coefficient of the total cost model was 0.716 with a mean optimism equating to -0.0057 (95%CI -0.202 to 0.289), reflecting low bias and a good model.

Infective endocarditis (IE), GCS, length of hospital stay, and bicarbonate were associated with the high total treatment costs of S.suis infection (Table 3). Overall, marginal increases in IE and length of stay were strongly associated with total cost (standard error) increases of 132,443 THB (39,638 THB) and 5,490 THB (1,715 THB), respectively. In contrast, increases in GCS and bicarbonate level were related to total cost (standard error) decreases of 13,118 THB (5,026 THB) and 7,497 THB (3,430 THB), respectively. The average marginal estimates are presented in Table 4.