Asthma in Families Facing Out-of-pocket Requirements With Deductibles

Background and Significance Asthma is one of the most common serious chronic diseases of adults and children in the United States. Despite guidelines and evidence, adherence to recommended asthma controller medications is low. Cost is an important barrier to non-adherence to asthma medications. Employers are increasingly adopting high-deductible health plans (HDHPs) particularly those that qualify for Health Savings Accounts (HSAs), which subject most medications to deductibles rather than copayments as in traditional coverage. HSA-HDHPs can thus lead to forgone care due to cost, including clinically appropriate services such as asthma medications. As a response, value-based insurance designs (VBID) have been proposed to promote high-value care by reducing or eliminating cost-sharing for these services. One common example is a Preventive Drug List (PDL) that can accompany HSA-HDHPs, which exempts certain chronic medications from the deductible to promote adherence. Many PDLs include asthma controller and rescue medications. With the increasing prevalence of HSA-HDHPs, PDLs have the potential to improve controller medication adherence for both adults and children.

Study Aims: The goal of this research is to provide needed evidence on the impact of two important developments in the health insurance landscape, HSA-HDHPs and PDLs, and whether PDLs can mitigate cost barriers associated with HSA-HDHPs and improve patient-centered outcomes for adults and children with asthma. The Aims of this research are:

1. To understand health care decision making and experiences of families with asthma with HSA-HDHPs and PDLs.
2. To examine the impact of HSA-HDHPs with and without PDLs on use of asthma controller and rescue medications, and on adverse clinical outcomes (asthma-related ED visits and hospitalizations), overall and for vulnerable subgroups (low-income and racial/ethnic minority patients).
3. To examine the extent to which the response to HSA-HDHPs and PDLs is affected by the presence of other family members with asthma or other chronic conditions.
4. To examine the impact of HSA-HDHPs with and without PDLs on OOP costs for patients and families with asthma.

Sources of Data

Aim 1. This study will collect patient-reported data through in-depth qualitative phone interviews

Aims 2-4 will use 14 years (2004-2017) of enrollment and claims data from a large U.S. commercial health plan

Study Population

The Aim 1 study population will consist of adults with asthma and parents of children with asthma. Eligible participants will be those who are currently enrolled in employer-sponsored health insurance plans (high-deductible plans with and without a PDL, or traditional plans).

The health plan population will be identified through enrollment and claims data from Harvard Pilgrim Health Care (HPHC). Eligible patients will include adults aged 18-64 and children aged 4-17 years with a diagnosis of asthma. Patients will be selected if they have been enrolled in an employer-sponsored HSA-HDHP with or without a PDL, or a traditional plan without a high deductible, for the prior year. Among those eligible in each sub-group, the investigators will randomly select 9 from each cell to send a recruitment mailing, for a total of 81 patients or parents invited to participate.

A population from the Asthma and Allergy Foundation of America (AAFA) will be recruited through postings to AAFA's Asthma Online Community, Educational Support Group, email list serve, Facebook page, and newsletters.

For Aims 2-4, the study population will consist of adults aged 18-64 years and children aged 4-17 years with asthma. Study subjects must have spent a year in a traditional health plan with no or low deductibles prior to the switch to an HSA-HDHP without a PDL, or in a HSA-HDHP without a PDL prior to switching to an HSA-HDHP with PDL, and then remain enrolled for at least one year post-switch. Control group members spend a year in a traditional plan or HDHP without PDL then remain in that plan for at least another 12 months.

Enrollees over age 64 who are eligible for Medicare will be excluded. The investigators will identify members with asthma during the baseline year using the same claims-based algorithm in which an asthma diagnosis will be defined as having at least one inpatient or two outpatient claims in the prior year with a diagnosis of asthma based on International Classification of Diseases (ICD) 9 and 10 codes for asthma.

Eligible employers are those that offer only one plan type in a given benefit year: 1) traditional Health Maintenance Organization/Preferred Provider Organization/Point of Service plans (deductibles <$1000, copayments of <$50 for most services, tiered copayments for medicines); 2) HSA-HDHPs without PDLs; or 3) HSA-HDHPs with PDLs. The investigators will select "full-replacement" employers that replace a traditional plan or HSA-HDHP without PDL with an HSA-HDHP without PDL or HSA-HDHP with PDL, respectively, for all employees at a given point in time; matched comparison employers will include those that keep all employees in their prior plan.

Outcomes

Aim 1 qualitative interviews will assess patient and family experiences across a number of domains related to asthma health care decision making and outcomes in HDHPs.

The primary study outcomes for Aims 2-4 are the claims-based measures of asthma medication use, outcomes, and OOP costs listed below, measured at the individual level. See section on primary outcomes for details,

Predictors and Covariates

The primary predictor variables include study period and insurance type. Study period indicates the one-year period before or up to three years after the index date. Insurance type includes: 1) HSA-HDHP without PDL; 2), HSA-HDHP with PDL; and 3) traditional plan.

Other co-variates include baseline asthma severity, co-morbidity, presence of other chronic conditions, self-reported and geocoded race/ethnicity data, neighborhood education and poverty levels, sex, age, state, individual versus family plan, baseline number of outpatient visits, presence of an inpatient hospitalization, total expenditures, employer size, average employer baseline expenditures per capita, number of family members, mean age of children in the family, mean age of adults, baseline mean family morbidity score, number of family members with asthma, number of asthma and other medications used by the family, and number of family ED visits and hospitalizations in the baseline year.

Analysis plan

Aim 1

The study team will analyze qualitative data in iterative cycles of content analysis in the manner described by Patton. In the second, deductive phase of analysis, investigators will consider data code-by-code to identify areas of convergence and divergence by insurance type.

Aim 2

Analyses will compare changes in outcomes from baseline to up to three years of follow-up among 1) asthma patients switched to HSA-HDHPs without PDLs from traditional plans vs. matched patients whose employers remain in traditional plans; and 2) patients switched to HSA-HDHPs with PDLs from HSA-HDHPs without PDLs vs. matched controls remaining in HSA-HDHPs without PDLs. The investigators will use separate regression models to compare year-to-year changes for each intervention group relative to its matched control group, rather than including all patients in a single model with multiple interaction terms.

The analyses will use both Interrupted Time Series (ITS) and Difference-in-Differences (DiD) frameworks. Controlling for potential confounders, analyses will use generalized linear models (GLMs) to model the independent effect of switching to each of the two types of HSA-HDHPs (with or without PDLs) on the likelihood of each outcome, assessed by interacting insurance type and study period variables in models. Extended GLMs - generalized estimating equations (GEE) and generalized linear mixed models (GLMM) - are appropriate methods to adjust for correlation within families and to examine changes in outcomes between baseline and follow-up.

This study will examine controller medication adherence using ITS. Investigators will test the statistical significance of level or trend changes following insurance plan type switch using GLM models, adjusting for seasonality and first-order autocorrelation between sequential monthly measurements using the empirical sandwich estimator.

For analyses of rescue medication use, investigators will focus on albuterol and levalbuterol inhaler users. The standardized number of rescue inhalers dispensed will be modeled as count data in difference-in-differences models. ITS models will be used to model changes in level and trend of monthly rates of rescue inhalers dispensed, as in previous studies. Analyses will also model the ratio of controller medications to total asthma medications. For analyses of asthma-related ED visits and hospitalizations, outcomes can be binary, counts, or continuous. The investigators will use logistic GEE models to estimate the effect of switching to each type of HSA-HDHP on binary outcomes such as any asthma-related hospitalization. Negative binomial regression will be used to model the effect for count outcomes such as ED visits. Investigators will select the conditional mean and variance functions based on the actual data, using a log link with a Gamma error distribution.

To determine the impact of HSA-HDHP with and without PDLs on vulnerable populations and test for heterogeneity of treatment effects among vulnerable populations, the investigators will first perform stratified analyses, comparing outcomes between intervention and control subgroups defined by the binary measures of the risk factors of interest (low income, minority race/ethnicity, moderate-severe asthma, presence of other chronic conditions, high morbidity based on the Johns Hopkins Adjusted Clinical Groups (ACG) system). Analyses will use three-way interaction terms (insurance type * study period * subgroup) to test for statistical differences between subgroups in the impact of the change to an HSA-HDHP with PDL vs. remaining in a traditional plan. In Aim 2 analyses, the investigators will include both adults and children together, with age included as a co-variate.

Aim 3

Analyses will use the same population, outcomes, study group comparisons, and modeling strategies as Aims 1 and 2 except that stratified analyses will be performed, comparing outcomes between intervention and control groups stratified by adult/child status. To statistically test for heterogeneity of treatment effect for adults vs. children, analyses will use three-way interaction terms (insurance type * study period * adult/child) to test for statistical differences between adults and children in the impact of the change to an HSA-HDHP with PDL vs. remaining in a HSA-HDHP without PDL. Analyses will be done at the individual level, but will include family-level variables as predictors of interest. Separately for adults and children, the investigators will test the extent to which having another family member with asthma, another chronic condition, or high baseline family ACG morbidity modifies the impact of HDHPs and PDLs on study outcomes for an individual asthma patient. The primary predictor of interest will be the interaction between the family-level variable, study period, and study group.

Aim 4

Analyses will be similar to those of Aim 2, but for OOP cost outcomes. The primary analyses of changes in OOP costs will use a DiD analytic framework. The investigators will follow the same approaches as used in analyzing ED visits and hospitalizations, employing two-part models/zero-inflated negative binomial models to account for zero costs. GEE or GLMM models will be used to examine changes in outcomes between baseline and follow-up to model the independent effect of switching to each of the two types of HSA-HDHPs (with or without PDLs) on the likelihood of having financial burden (OOP cost greater than 5% of income).