Temporal trends of medical cost and cost-effectiveness in sepsis patients: a Japanese nationwide medical claims database

Study setting and patients

We conducted a retrospective cohort study using the Japanese nationwide medical claims database obtained from the DPC system, which includes diagnoses, interventions, and comorbidities/complications during hospitalization. These classifications are used for medical service reimbursements during acute inpatient care. The DPC data were obtained from 1237 hospitals, which covered 71.5% of acute care facilities in 2017 [8]. All the registered patients were screened for sepsis between 2010 and 2017. Patients over the age of 20 were included in this study without any other exclusion criteria.

This study was approved by the Institutional Review Board of the Chiba University Graduate School of Medicine. The review board waived the need for written informed consent from participants or their guardians.

Data collection and definition

Using the claims database, the following information was collected: age, sex, length of hospital stay, chronic diseases (malignant tumor, hypertension, diabetes mellitus, heart failure, cerebrovascular disease, ischemic heart disease, chronic respiratory disease, and chronic renal failure), admission to the intensive care unit (ICU), discharge status (home, nursing facility, and inter-hospital transfer), medical cost, site of infection, medical procedures, laboratory tests, and admission diagnosis or complications during hospital stay. Primary diagnosis, comorbidities, and in-hospital complications were recorded as codes based on the International Statistical Classification of Diseases and Related Health Problems 10th revision (ICD-10) (Additional file 1: Table S1). The site of infection was determined according to the following recorded codes: respiratory (mouth, throat, nasal cavity, neck, lung, lower respiratory tract, chest cavity), urogenital (kidney, urinary tract, uterus, genital organs), abdominal (liver, gall bladder, intestine, peritoneal cavity, gastrointestinal system), bone and soft tissue (skin and soft tissue, bone and joint, lymph tissue, breast), meninges/brain/spinal cord, heart, blood, and unknown. Patients with missing data (n = 766,395) were excluded from the analysis: only data regarding the site of infection were missing. Multiple codes in the “site of infection” were replaced with the category “Multiple”.

We selected sepsis patients who had presumed serious infection and organ failure during hospital stay [4]. (Additional file 2: Fig. S1). Presumed serious infection was defined as initiation of new antibiotic treatment (intravenous) within ± 2 days followed by blood culture and antibiotic administration for at least consecutive 4 days. Because of unavailability of laboratory data in the database, data regarding organ dysfunction were extracted as follows: use of vasopressors, mechanical ventilation or oxygen supplementation, kidney injury extracted based on diuretics use, diagnostic codes related to kidney dysfunction or renal replacement therapy, liver injury extracted using the codes indicating liver dysfunction, thrombocytopenia, metabolic acidosis.

Medical cost

The total medical cost per hospitalization includes the fee of drugs, laboratory tests, radiological examinations, and medical procedures during the hospital stay. Although medical fees were mainly reimbursed on a bundled payment basis, we calculated medical costs based on the reference prices in the Japanese fee schedule, as described in a previous report [19]. Since the number of hospitals subject to DPC systems has been increasing over the years, yearly gross medical costs were adjusted by the number of registered patients in the DPC system. Daily medical cost per person was derived by dividing the total medical cost by the length of hospital stay. Furthermore, the value of medical costs was normalized with the consumer price index (CPI) in Japan in 2017 to compare the trend among different years (https://www.e-stat.go.jp/en/dbview?sid=0003427113). The CPI-adjusted costs were calculated using the following formula:

$$} - }\,} = } \times }\,}\,2017/}\,} \,}\,}\,}$$

Subsequently, the CPI-adjusted costs were converted into U.S. dollars in accordance with the latest exchange rate between U.S. dollars and Japanese yen as of February 3rd, 2022 (115.25 yen = $1 USD).

Calculation of cost-effectiveness

As a representative parameter to compare the trend of cost-effectiveness in healthcare settings throughout the study period, we calculated the effective cost per survivor from annual gross medical cost for all patients, including survivors and non-survivors, and number of survivors per year as follows [20, 21]:

$$}\,}\,}\,} = }\,}\,}\,}\,}\,}\,}/}\,}\,}\,}\,}$$

Statistical analysis

The primary outcome was the annual change in the effective cost per survivor in sepsis patients. The significance over different years was analyzed using a linear regression test, on the assumption that the regression plot was followed by linearity. As a sensitivity analysis, we conducted subgroup analyses for cost-effectiveness with regard to age, ICU admission, transfer to other hospitals, site of infection (single or multiple infection), mechanical ventilation, vasopressor therapy, and renal replacement therapy.

The secondary outcomes were the annual trend of medical cost per hospitalization and daily medical cost per person. We then conducted subgroup and multivariable regression analyses to investigate the association between the admission year (independent variable) and medical costs during hospital stay. For the subgroup analysis, sex, age, and site of infection were chosen as the target variables. The age subgroups were divided into adults (20–64 years), early elderly (65–74 years), and late elderly (≥ 75 years), as previously described [13]. Also, the subgroup analysis was conducted based on the site of infection (respiratory, urogenital, abdominal, bone and soft tissue, meninges/brain/spinal cord, heart, blood, multiple, and unknown). To remove baseline imbalances among subgroups, medical costs were adjusted for sex, age, number of chronic diseases, and site of infection using a generalized linear regression model. Using prediction probabilities for target variables, adjusted parameters were expressed with 95% confidence intervals (CI) as representative values in each subgroup. Also, we performed a multiple regression analysis to enhance the robustness of our study. We adjusted medical cost per hospitalization for the following possible confounders: age, sex, site of infection, number of chronic diseases, ICU admission, surgery, and length of hospital stay [16, 22, 23], which were listed on the demographics and clinical characteristics of patients in the cohort. After an evaluation of the association between medical costs and each variable, the variable with a p < 0.10 was retained in the regression model. Multicollinearity were measured to check the interaction and confounding among the variables. As a result of regression diagnostics, normal plots of the residuals displayed skewed distribution. Then, we developed a multiple regression analysis using log-transformed costs to better fit the normal distribution. Coefficients calculated from the regression analysis were converted into integer values in the results. Repeated admissions were excluded from mortality analysis.

Data were presented as means (standard deviation), medians (quartiles), or numbers and percentages, as appropriate. A two-tailed p-value < 0.05 was considered statistically significant. Data manipulation and statistical analyses were performed using SQL (mariadb v10.4.17), GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA), pandas (v1.0.5), scipy (v1.7.3), numpy (v1.21.4), seaborn (v0.11.2), matplotlib (v3.5.1), and statsmodels (v0.13.2) in Python (v3.9.0).

留言 (0)

沒有登入
gif