Healthcare Utilization and Costs of Systemic Lupus Erythematosus by Disease Severity in the United States

MATERIALS AND METHODS

Study design. In this study, we conducted a retrospective analysis of EHR data (utilization analysis) and integrated claims and EHR data (cost analysis) for 2 cohorts of patients with SLE in the US. We quantified the following: (1) incidence and prevalence of SLE; (2) healthcare resource use patterns of patients with SLE, including hospitalizations/inpatient visits, outpatient visits including emergency department (ED) and office or clinic visits, and prescription drug use; and (3) annual direct medical and pharmacy costs incurred, stratified by SLE disease severity. Institutional review board approval was not required because this study utilized anonymized administrative claims and EHR data.

Data source. We used Optum administrative claims and EHR data from Humedica Inc. The EHR data contain information for 70 million patients, of whom 50 million are associated with integrated delivery networks (IDN), groups that provide a spectrum of healthcare services. These data include laboratory results; vital signs; body measurements; lifestyle observations; biomarkers; inpatient and outpatient treatments, including written prescriptions; inpatient-administered medications; and provider notes. The Optum claims database includes medical and pharmacy claims from approximately 40 million commercially insured patients and 6.1 million Medicare Advantage enrollees. For the purposes of this study, a project-specific data extract was generated. It included all EHR and integrated claims data for patients who had ≥ 1 SLE-related healthcare encounter, defined by the presence of International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) codes 710.0 or M32.X, respectively, in their EHR. Both the EHR and claims datasets use billing nomenclature to code healthcare services provided. Diagnosis status and procedures performed are coded using ICD-9-CM and ICD-10-CM. Medication use was identified by the National Drug Codes and Healthcare Common Procedure Coding System, J codes.

Patient identification. Patients with SLE were identified in EHR data using algorithms that required patients to be associated with an IDN, to mitigate for the effects of unobserved services, and have ≥ 2 SLE-related encounters ≥ 60 days apart and ≥ 1 pharmacy claim for an SLE-related medication; or ≥ 2 SLE-related encounters ≥ 60 days apart, of which ≥ 1 was a visit to a rheumatologist¹¹. An SLE-related healthcare encounter was defined by the assignment of an ICD-9-CM code 710.0 or ICD-10-CM code M32.X to the patient’s EHR for hospitalization, office or clinic visit, ED visit, or other outpatient service visit. SLE-related medications included 4 drug classes used to manage SLE: antimalarial medications, immunosuppressive agents, systemic corticosteroids, and biologics.

Disease severity (mild, moderate, or severe) was determined for each patient by a healthcare utilization–based algorithm derived from the severity of diagnoses listed on claims, prescription drug claims, and conditions associated with SLE severity available in nonlaboratory claims (e.g., renal impairment was considered moderate and endstage renal disease was considered severe)⁹. Disease severity was determined for each patient during each year of the study, and the severity assigned to a particular year was the highest severity category attained by the patient during that year (see Supplementary Table 1, available with the online version of this article, for full description of the SLE severity algorithm).

Incidence and prevalence of SLE. We identified 2 retrospective cohorts, incident and prevalent, from the EHR data to determine annual incidence and prevalence of SLE from 2011 to 2015. An incident case of SLE referred to both newly diagnosed patients and patients whose SLE condition was dormant during the preestimation year. To be considered an incident or a newly diagnosed SLE case in a given year, patients had to meet the SLE case definition between January 1 and December 31 of the year of estimation. Incident patients also had to be included in EHR data for ≥ 365 days (with no SLE-related codes) before their first SLE-related encounter. To be considered a prevalent SLE case, newly diagnosed or patients with existing SLE had to meet the case definition for SLE and have information in the database during the year of estimation. Patients who did not have information in the database for a given year were not counted during that year. The denominator used for determining yearly incidence and prevalence was the number of patients who had ≥ 1 encounter during the year.

Patterns of healthcare utilization for patients with SLE. Healthcare utilization was determined for the years 2011 to 2015. Patients were followed from their index dates (earliest SLE diagnosis date observed in the database) until the month of their last observed activities in the EHR database or date of death, whichever was first. All study participants were required to be part of the EHR database during the calendar year of estimation to assess annual healthcare utilization patterns and to have ≥ 1 calendar year of data after the index date.

All-cause healthcare utilization was measured in the following categories: hospitalizations, outpatient visits (ED and office or clinic visits), total drug prescriptions, and SLE-related drug prescriptions [antimalarial medications, immunosuppressive agents (azathioprine, mycophenolate mofetil, cyclosporine, methotrexate, leflunomide, cyclophosphamide, chlorambucil, nitrogen mustard), systemic corticosteroids, and biologics (RTX, belimumab)]. Disease severity was not measured in this cohort because corticosteroid dose could not be determined for all patients.

Direct medical and pharmacy costs associated with SLE. Healthcare cost analyses were limited to data from 2012 to 2015 in a separate cohort. To be included in the cost analysis, patients had to (1) be identified in the EHR database; (2) be enrolled in an Optum-participating health insurance plan with both medical and pharmacy benefits on the index date; and (3) have ≥ 1 calendar year of data after the index date.

SLE-related outcomes evaluated were annual direct medical costs for patients with ≥ 1 service-type encounter, stratified by service type (e.g., hospitalizations, ED visits, office/clinic visits, and other patient services); annual prescription drug costs, including both outpatient- and inpatient-administered drugs; and annual direct medical costs associated with SLE. Study outcomes were also stratified by service type and SLE disease severity (mild, moderate, or severe)⁹. All costs were converted to 2016 US$ using the Consumer Price Index annual averages for medical care¹².

Statistical analyses. Patient demographics were characterized using descriptive statistics. Healthcare utilization for 2011 to 2015 was stratified by year and evaluated using descriptive statistics. Healthcare costs for 2012 to 2015 were categorized and reported annually for hospitalizations, outpatient visits, ED visits, clinic or office visits, outpatient services, and medications. Annual costs were also reported by SLE disease severity. All analyses were performed using SAS version 9.4 (SAS: SAS Institute Inc.).