Data source
The Japanese Diagnosis Procedure Combination database is a nationwide administrative database of claims and discharge abstract data that has been previously described in detail [8]. The database includes data for approximately 7 million inpatients per year from more than 1000 participating hospitals and covers approximately 90% of all tertiary-care hospitals in Japan. A previous validation study confirmed that the recorded procedures had high sensitivity and specificity [9]. The database includes the following information: age, sex, body mass index (BMI), diagnosis recorded using the International Classification of Diseases, 10th revision (ICD-10) codes and free-text data in Japanese, level of consciousness at admission, and discharge status. Additionally, the database includes data on procedures performed and drugs used during hospitalization [8].
This study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 2013, and approved by the institutional review board of The University of Tokyo (approval no. 2018030NI). The need for informed consent was waived due to the anonymity of the data.
Study design and population
Using the database, we identified patients aged 1–15 years who underwent percutaneous kidney biopsy under intravenous sedation or general anesthesia between July 1, 2010 and March 31, 2019. We defined the use of intravenous sedation only by the absence of the record of general anesthesia with intubation use but by the presence of intravenous sedation use during biopsy; we defined the use of general anesthesia by the presence of the record of general anesthesia with intubation use during biopsy. Patients scheduled for admission with a definite or suspected diagnosis of kidney disease (ICD-10 codes: D690, N00-08, N11, N12, N14, N158, N159, N16, N17, N18, N19, N25, N289, N391, N392, R31, or R80) who underwent a kidney biopsy within 4 days after admission were included. We excluded patients with impaired consciousness at admission; those who received kidney replacement therapy, mechanical ventilation, vasopressors, or a surgical procedure under general anesthesia before the kidney biopsy; those admitted to the intensive care unit; or those with a transplanted kidney (ICD-10 codes: Z940 or T861) or malignancy (C or D0). We excluded patients with transplanted kidney because the location of the transplanted kidney may differ from that of the naive kidney, and therefore the risk of bleeding at the time of biopsy can vary according to a previous study. [10] Patients who received blood transfusion or invasive hemostasis before kidney biopsy and those with missing BMI data were also excluded.
Study variables
We extracted data on comorbidities at admission, including metabolic disease (ICD-10 codes starting with E), mental disease (F), neurological disease (G), cardiovascular disease (I), respiratory disease (J), musculoskeletal disease (M), and congenital disease (Q). We also extracted data on the etiology of kidney disease recorded in the database (chronic nephritis or nephrotic syndrome), use of corticosteroids or immunosuppressants before kidney biopsy, use of albumin infusion before kidney biopsy, use of tranexamic acid on the day of kidney biopsy (because a randomized controlled trial showed its effect on reducing bleeding events [11]), type of anesthesia used during kidney biopsy (general anesthesia or intravenous sedation only), and history of previous kidney biopsy during the study period. We also obtained information on hospital characteristics and the year when the biopsy was performed because these factors may be associated with physician preference for application of a certain sedation method for biopsy, and possibly complication occurrence [12, 13]. We defined the hospital volume for pediatric kidney biopsies as the average annual number of pediatric patients who underwent percutaneous kidney biopsy and made a binary variable indicating whether the hospital was an academic hospital.
Age was categorized into three groups: 1–5, 6–11, and 12–15 years. BMI was categorized based on the BMI standard deviation score (BMI-SDS) for Japanese children [14]: underweight (BMI-SDS of < -1.28), normal weight (BMI-SDS of -1.28 to 1.279), and overweight (BMI-SDS of ≥ 1.28). Hospital volume was categorized as low, medium, or high based on the rounded tertiles of the included hospitals.
Outcomes
The primary outcome was the occurrence of bleeding complications. Severe bleeding complications were defined as either red cell transfusion within 7 days after kidney biopsy or invasive hemostasis (transcatheter arterial embolization or nephrectomy). Non-severe complications were defined as diseases occurring after hospitalization according to ICD-10 codes denoting hemorrhagic events or hematoma formation, such as acute hemorrhagic anemia. The codes included hemorrhagic anemia (D500), retroperitoneal hemorrhage (K661), retroperitoneal hematoma (K668), renal or perirenal hemorrhage (N288), hemorrhagic shock (R571), hemorrhage occurring following the procedure (R58), renal hematoma (S3700), perirenal hematoma (T140), post-procedure hemorrhage or hematoma (T810), and hemorrhagic shock (T794), as defined in a previous study examining post-biopsy complications [13].
Statistical analysis
We summarized the patient characteristics in the intravenous sedation and general anesthesia groups. Patient characteristics were compared using Fisher’s exact test for categorical variables and Student’s t-test for continuous variables. We also described the length of hospital stay and total hospitalization costs for each group. The total hospitalization costs were converted into US dollars ($), assuming that 120 Japanese yen was equivalent to $1.
We used propensity scores to minimize confounding caused by indications. When we estimated the propensity scores for receiving intravenous sedation only, we used a logistic regression model, setting the dependent variable as the receipt of intravenous sedation and the independent variables as all the above-listed covariates. Based on the calculated propensity scores, we conducted an analysis using overlap weights to adjust for the differences [15]. Overlap weights were defined as 1 − propensity score among patients receiving intravenous sedation only and as the propensity score among patients receiving general anesthesia. Thereafter, we used weighted generalized linear models with binomial distribution and log link functions to obtain the relative risks. To calculate the relative risks with overlap weights in both groups, we used robust variance estimators to calculate confidence intervals (CIs) as used in weighted analyses [16]. We also calculated risk differences in the primary analysis using a generalized linear model with a Gaussian distribution, identity link function, and robust variance.
Stratified analysis
We performed analyses stratified by age and sex with and without overlap weights. In the age-stratified analyses, we used the age in months to adjust for individual age.
Sensitivity analysis
Two sensitivity analyses of the outcomes were performed. First, we performed propensity score matching instead of overlapping weights. Using this propensity score-matched cohort, we used generalized estimating equations with binomial outcome distribution, log link function, exchangeable working correlation model, and sandwich variance estimator with each hospital set as a unit of a cluster; this method can adjust for the effects of hospital clustering [17, 18]. Second, we used the instrumental variable method to confirm causal inference. Instrumental variable analysis can theoretically adjust for both measured and unmeasured confounders between two groups [19, 20]. Each hospital’s preference for intravenous sedation only was selected as an instrumental variable because the choice of intravenous sedation presumably depends only on physician preference [19]. We used two-stage residual inclusion estimation for the instrumental variable analysis [21, 22].
We used a two-sided significance level of 0.05 and performed all statistical analyses using Stata version 17 (StataCorp, College Station, TX, USA).
