Data source
For this retrospective cohort study, we used data derived from the Diagnosis Procedure Combination database.5 All 82 academic hospitals in Japan are obliged to participate in the database. Each year, data are collected on about eight million hospitalized patients of all ages, which is equivalent to about 50% of all acute-care hospitalizations in Japan. The database includes the following information: unique identifiers of hospitals; baseline patient characteristics; and diagnosis at admission, comorbidities at admission, and complications after admission recorded as text data in Japanese and using International Classification of Diseases, Tenth Revision (ICD-10) codes. The database also contains information on medical procedures and treatments (including drug administrations, device use, and surgical and nonsurgical procedures) based on the original Japanese codes; length of stay; discharge status; and medical costs during hospitalization. The attending physicians have to confirm the recorded patient data for all diagnoses and comorbidities with reference to medical records. Furthermore, because accurate reporting is linked with the payment system in Japan, the attending physicians and hospitals are required to accurately report diagnoses and comorbidities. The present study was approved by the Institutional Review Board of The University of Tokyo (approval number: 3501-(3); December 25, 2017). The requirement for informed consent was waived because of the anonymous nature of the data.
Participants
We used the database to identify patients who were diagnosed with KD (ICD-10 code: M303) from April 2012 to March 2020. We included patients whose first IVIG treatment was started within 3 days of hospitalization and who received at least 2 g/kg of IVIG within 3 days of starting their initial IVIG treatment. We checked the Japanese text describing the detailed diagnoses in each case to include patients with atypical KD and exclude patients with “suspected” KD. We also excluded patients aged > 6 years because the appropriate IVIG dose for older patients with KD remains controversial. We further excluded patients weighing < 3 kg, those with missing data, and those with inadequate treatment (< 2 g/kg IVIG). The included children with KD in the acute phase were divided into two groups depending on whether they had a 5% or 10% IVIG preparation as their initial treatment. We excluded patients who received both 5% and 10% IVIG preparations.
Outcomes
The primary outcome was the occurrence of CAAs at the time of discharge. CAAs were identified using the recorded diagnosis of CAAs (ICD-10 code: I254) and/or the appearance of “CAAs” in the Japanese language text data. The secondary outcomes were the proportion of patients with IVIG resistance, length of stay, and medical costs. IVIG resistance was defined as the use of IVIG at a total dose of ≥ 4.0 g/kg and/or in combination with any steroid, infliximab, cyclosporine, and/or plasma exchange that was not performed at the same time as the initial IVIG treatment. We also examined adverse events occurring with IVIG, including hyperviscosity syndrome (ICD-10 code: R70.1) and congestive heart failure (ICD-10 code: I50.0).
Covariates
The included baseline characteristics included sex, age, weight, height, hospital days of illness at initial IVIG, type of hospital, complex chronic conditions,6 activities of daily living at admission, Japan Coma Scale score at admission, transportation by ambulance, transportation from another hospital, fiscal year, additional treatments, and hospital volume. Assessments using the Japan Coma Scale have previously been shown to be highly correlated with assessments using the Glasgow Coma Scale.7 Hospital volume was defined as the annual number of all patients with KD at each hospital. We separated the included patients into hospital volume tertiles so that the numbers of patients in the groups were almost equal. Additional treatment was defined as the use of any steroid, infliximab, and/or cyclosporine at the same time as or before the start of the initial IVIG treatment.
Propensity score-matched analysis
We performed 4:1 propensity score-matched analyses to compare the outcomes between the two IVIG concentration groups. We estimated the propensity score using a logistic regression model in which 10% IVIG use was the dependent variable, with the covariates described above. For the propensity score matching, nearest-neighbor matching without replacement was performed. The caliper width was set at ≤ 0.2 of the pooled standard deviation of the evaluated propensity scores. We estimated the covariate balance between the two groups before and after propensity score matching using the absolute standardized difference. An absolute standardized difference of > 10% was regarded as imbalanced.8
Instrumental variable analysis
Propensity score-matched analyses cannot exclude the effects of unmeasured confounders such as symptoms of KD, laboratory data, duration of illness, duration of fever from KD onset, duration of IVIG administration, and duration of observation after initial IVIG treatment. Thus, to confirm our propensity score-matched analyses, we performed instrumental variable analyses. The key assumptions for instrumental variable analysis are as follows: the instrumental variable (i) is highly correlated with the treatment assignment, (ii) is not correlated with other confounders, and (iii) does not affect patient outcomes except through the treatment.9
In instrumental variable analyses, physician prescribing preference is commonly used as an instrumental variable. This variable reflects the notion that a physician’s prescribing decision depends not only on the patient’s characteristics but also on the physician’s preference for each treatment. Physician’s preference is mostly independent of patient characteristics and outcomes and can therefore be used as an instrumental variable. In the present study, we used prescription for previous patient as the instrumental variable.10 When the prescription for the previous patient in the same institution was 10% IVIG, the current patient was assumed to be more likely to receive 10% IVIG. When the prescription for the previous patient in the same institution was 5% IVIG, the current patient was assumed to be more likely to receive 5% IVIG. Prescription for the previous patient is assumed to be independent of the current patient’s characteristics and not directly related to their outcome. Therefore, this variable was considered to meet the three key requirements for an instrumental variable described above.
In this study, we employed a two-stage residual inclusion method for both continuous and binary outcome variables.11,12 In the first-stage model, we determined the row residual for each patient by calculating the difference between the model-predicted probability of receiving the treatment choice and the actual treatment received. These residuals were included as an additional covariate in the second-stage model, where the association between treatment choice and outcome was estimated in an unbiased manner after adjustment for covariates. We used a multivariable linear regression model for continuous outcome variables and a multivariable logistic regression model for binary outcome variables. All instrumental variable analyses were performed using robust standard errors. To assess the validity of the instrumental variable, we tested its association with our main predictor of actual treatment choice using the F-statistic (F-statistic > 10 is considered to reflect a valid instrumental variable).13 We also determined that the instrumental variable was not associated with the patient background characteristics or outcomes. A two-sided p < 0.05 was considered significant. All statistical analyses were conducted using Stata, version 17.0 (StataCorp LP, College Station, TX, USA).
Statistical analysis
Categorical variables, which are shown as numbers and percentages, were compared using Fisher’s exact test. Continuous variables are shown as means and standard deviations or medians and interquartile ranges. The Mann–Whitney U test was used to compare non-normally distributed variables between the two groups. Specifically, this test was used to compare length of stay and medical costs.
To allow for the meaningful interpretation of variables with skewed distributions, we calculated median differences with 95% confidence intervals (95% CIs) using the Hodges–Lehmann estimator. The chi-square test was used to compare the proportions of patients with CAAs and IVIG resistance between the two groups. We also estimated the risk differences with 95% CIs.
As a sensitivity analysis, CAAs were defined as a diagnosis of CAAs and the use of anticoagulants such as warfarin or clopidogrel, or a diagnosis of CAAs and cardiac catheterization. The diagnosis of CAAs was defined as the presence of a descriptive diagnosis of CAAs (ICD-10 code: I254) or the use of “CAAs” to describe the case in the Japanese text data.