Healthcare Service Use Patterns Among Patients with Acid Sphingomyelinase Deficiency Type B: A Retrospective US Claims Analysis

Study Design and Data Source

This was a retrospective cohort study that used data from the IQVIA Open Claims patient-level pharmacy and medical database for the period 2010–2019. The IQVIA Open Claims database contains anonymized data from medical and pharmacy claims. As of December 2019, the database included data for 654 million patients. Patients with ASMD were identified on the basis of diagnosis codes associated with any claims for services that occurred between October 1, 2015 and October 31, 2019. The data used in this study to estimate healthcare service use patterns were derived from claims dated between January 1, 2012 and January 31, 2020.

Primary Analysis Cohort

The primary analysis cohort included patients with confirmed ASMD type B, defined as those who had at least two claims with an International Classification of Disease, 10th Revision, Clinical Modification (ICD-10-CM) diagnosis code associated with ASMD type B (E75.241, Niemann–Pick disease type B) and more claims with ASMD type B than other ASMD types (E75.240, Niemann–Pick disease type A; E75.242, Niemann–Pick disease type C; E75.243, Niemann–Pick disease type D). It should be noted that diagnosis of ASMD type B was not validated in these patients; the term “confirmed” refers to the fact that these patients had at least two claims with ICD-10 codes specific to ASMD type B. The use of at least two claims with the relevant diagnosis codes is commonly used for patient identification from claims databases. Patients with ASMD type A, confirmed by the relevant diagnosis code, evidence of neurological symptoms, and age at diagnosis of 3 years or less, were excluded from the primary analysis cohort. Eligible patients were required to have at least 6 months claims history before their first claim with the above diagnosis codes (lookback period).

Sensitivity Analysis Cohort

A large proportion of patients were coded under the unspecified or other ASMD disease codes and could not be classified into a specific type, as there was no clear delineation in the patient’s healthcare history where they could be easily categorized via applying clinical rules. Therefore, the IQVIA machine-learning algorithm was used to further classify and categorize these patients to identify those with a high probability of having ASMD type A, A/B, and B. The algorithm used information from a training set of patients to classify those with unspecified or other ICD-10-CM codes. The training set consisted of patients with confirmed ASMD types (Niemann–Pick A, B, C, and D) who had at least one claim activity beyond ICD-10-CM and a lookback period of at least 2 years (at least 2 years of claims history before their first claim with an ASMD diagnosis code). An XGBoost model was developed in which patients with ASMD type A or B (Niemann–Pick A or B) were defined as patients of interest (positive cohort), and other types (Niemann–Pick type C or D) as a negative cohort [14].

After training and evaluation, the model was applied to identify eligible patients with “unspecified” diagnosis codes between October 1, 2015 and October 31, 2019. The following criteria were applied: one or more claims with an ICD-10-CM diagnosis code for ASMD, but not previously categorized as Niemann–Pick type A, B, C, or D, at least one claim activity beyond ICD-10-CM, and an at least 2-year lookback period. Eligible patients were scored by the model according to their probability (high, medium, or low) of having ASMD type A, A/B, or B, and assigned to type B if they had no evidence of neurological symptoms or diagnoses. Patients were excluded if they had neurological symptoms or diagnoses, including stroke, cerebrovascular disease, abnormal electroencephalogram, abnormal gait, abnormal movements, ataxia, cognition abnormality, convulsions, difficulty walking, epilepsy, lack of coordination, neuropathy, palsy of conjugate gaze, Parkinson’s disease, and tremor. Excluding patients with neurological symptoms/diagnoses increased the likelihood of identifying patients with ASMD type B. The group of patients with a high probability of having ASMD type B according to this method made up the sensitivity analysis cohort, with no patients overlapping with the primary analysis cohort.

Index Date and Follow-up Period

The index date for patients in both cohorts was defined as the date of the earliest claim with an ICD-9-CM code of 272.7 (lipidoses) or ICD-10-CM diagnosis codes for ASMD type B, other ASMD, or unspecified ASMD; for the sensitivity analysis cohort, the date of the earliest claim with any ICD-10-CM code for ASMD (independent of type) could be used.

Healthcare service use patterns were evaluated over each patient’s follow-up period, defined as the period between, and including, the index date and the last day of the month of the last observed claim (assumes that insurance coverage, if ending, would extend through the end of the month; Fig. 1). To account for lack of information on enrolment in the IQVIA Open Claims database, claims activity was used as a proxy and patients in the primary analysis and sensitivity cohorts were required to have claims in at least two different months.

Fig. 1figure 1Healthcare Service Use Measures

Demographic and clinical characteristics (based on codes for diagnosis and ASMD-related healthcare services/procedures) were reported for the primary and sensitivity analysis cohorts according to patient age (< or ≥ 18 years) at the index date. Exact dates of initial diagnosis or use of healthcare services before diagnosis were not available for this analysis. A data quality assessment was performed on claims for patients in the primary analysis and sensitivity analysis cohorts. Rules were established to identify questionable claims, including potential duplicates and those with negative charge or payment amounts. All questionable claims and claims for services that occurred 3 days before and after the questionable claims were manually examined and subsequently resolved, which could include editing or deletion.

The following healthcare service use measures were described overall and for major health complications of ASMD: number of claims and annualized event rates for office visits, outpatient visits, emergency department (ED) visits, and inpatient hospitalizations. Major health complications included bleeding; cardiovascular; liver- and spleen-related disorders; respiratory disorders/lung disease; cerebrovascular; cognitive, developmental, and/or emotional problems; convulsions and/or epilepsy; Parkinson’s disease, dystonia, tremors, and/or other movement disorders; hyperlipidemia; osteoporosis; and thrombocytopenia (Supplementary material Table 1).

Multiple claims on the same day for office visits and outpatient visits were counted as one health encounter and ED visits were defined by dates of admission and discharge. Unique hospitalizations were defined by a series of hospital claims without a gap between service dates. ED visits on the day before or day of a hospitalization claim were considered part of the series of claims for one hospitalization. The length of hospital stay for major health complications of ASMD was defined by the number of consecutive days in an uninterrupted series of hospital claims.

ASMD-related services included liver transplant, lung transplant, splenectomy, coronary bypass graft surgery, stent placement, repair of heart valves, other cardiac surgery, oxygen use, and physical therapy (Supplementary material Table 2). Medication use was based on numbers of prescriptions for angiotensin-converting enzyme (ACE) inhibitors, aldosterone antagonists, antibiotics, anticoagulants, beta-blockers, bisphosphonates, influenza vaccine, pneumococcal pneumonia vaccine, respiratory medications, statins/antihyperlipidemic agents, and vitamin D (Supplementary material Table 3).

Statistical Analyses

Statistical analyses were conducted using SAS 9.4. Data for each study measure were analyzed for each patient group, i.e., the primary analysis cohort with confirmed ASMD type B and the sensitivity analysis cohort with high probability ASMD type B.

Categorical variables were summarized by their counts and proportions. Continuous variables were summarized by their means, standard deviations (SD), medians, and interquartile ranges (IQRs). Annualized event rates (with 95% confidence intervals [CI]) for office, outpatient, and ED visits, and hospitalizations were calculated as 365.25*\(\left(\sum__/\sum__\right)\) where \(x\) is the number of events experienced by a patient, \(d\) is the number of days in a patient’s follow-up period, and \(i\) indexes patients. Annualized event rates were presented as events per 1000 patient years (PY). Mean length of hospital stay (with 95% CI) was calculated as \(\sum__/\sum__\), where \(l\) is a patient’s total number of days spent in the hospital during the follow-up period, \(h\) is a patient’s total number of hospitalizations during the follow-up period, and \(i\) indexes patients.

Ethical Approval

This study only involved the use of retrospective, anonymized data from medical and pharmacy claims, and researchers did not have access to named or identifiable patient information. This study was therefore exempt from institutional review board review and was not conducted in accordance with the declaration of Helsinki. Informed consent was not required.

留言 (0)

沒有登入
gif