Differences in healthcare expenditures for inflammatory bowel disease by insurance status, income, and clinical care setting

Data

We performed a longitudinal analysis on data from 1996 to 2011 in the Household Component of MEPS, a nationally representative database conducted by the Agency for Healthcare Research and Quality. MEPS collects data on healthcare utilization and expenditures, health status, health insurance coverage, income, employment, and socio-demographic characteristics for the civilian, non-institutionalized population. 15,000 new individuals are sampled each year and followed for two years with in-person interviews, with response rates ranging from 54 to 78% (Agency for Healthcare Research and Quality, 2013a). MEPS utilizes a complex sampling methodology that includes stratification, clustering, multistage selection, and oversampling of certain subgroups including racial/ethnic minorities (Machlin, Yu & Zodet, 2005). Survey weights allow for nationally representative data analyses and the weighting process includes adjustments for nonresponse over time along with calibration to independent population figures from the U.S. Census Bureau’s Current Population Survey (Machlin, Yu & Zodet, 2005; Agency for Healthcare Research and Quality, 2010).

MEPS defines inpatient, emergency, and outpatient visits as occurring in a hospital setting or a facility connected with a hospital (Agency for Healthcare Research and Quality, 2009). Outpatient visits are defined as not requiring overnight hospitalization, as opposed to inpatient visits. Office-based events do not occur in a hospital or hospital-connected facility, but can occur in a variety of settings including doctor’s or group practice office, medical clinic, surgical center, community health center, walk-in urgent care centers, or laboratory/X-ray facilities (Agency for Healthcare Research and Quality, 2009). Thus, both outpatient and office-based care may include general primary care, and both may involve same-day procedures.

Self-reported expenditure data are validated with information from healthcare and pharmaceutical providers. Self-reported medical conditions are mapped by professional coders to International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes (Agency for Healthcare Research and Quality, 2013b).

Study population and variables

Individuals ages 3–90 with ICD-9-CM codes of 555.x or 556.x were included in this study a priori and defined as having CD or UC, respectively. Individuals lacking person-level weights were excluded.

Demographic data included age, sex, race/ethnicity, and poverty status. Race/ethnicity was encoded as non-Hispanic white (subsequently abbreviated to “white”) or non-white, which included black, Hispanic, Asian, Native American, and mixed-race individuals. Poverty status was measured as a binary variable comparing poor patients to not poor patients, with poverty defined as having a family income less than 100% of the federal poverty line (FPL) defined by the U.S. Census Bureau’s Current Population Survey.

The health-related quality of life comorbidity index (HRQL-CI) was used to adjust for comorbid conditions. The HRQL-CI is a validated risk adjustment index that outperforms the Charlson comorbidity index when external validation was assessed in MEPS (Mukherjee et al., 2011; Ou et al., 2012). To form the HRQL-CI, Mukherjee et al. (2011) selected 44 adult, gender-neutral, chronic conditions, then identified those significantly associated with the Short Form-12 physical component summary and mental component summary. The resulting two subsets of conditions comprise the HRQL-CI, consisting of a physical component score and a mental component score.

Insurance status was measured as a series of binary variables comparing private, public, and no insurance, for individuals who maintained the same insurance category for a full year. The definition of public insurance in MEPS included Medicaid, Medicare, Tricare (U.S. Department of Defense Military Health System), State Children’s Health Insurance Program (SCHIP), and other public hospital/physician programs (Agency for Healthcare Research and Quality, 2013b). Private insurance was non-public insurance that covered hospital and physician care. Individuals only covered by single-service plans (e.g. drug, dental, or vision plans) were considered uninsured.

IBD-related medications were identified using pharmacy-reported prescription names. We identified immunomodulators—that is, thiopurines (6-mercaptopurine and azathioprine) and methotrexate—anti-tumor necrosis factor (anti-TNF) agents (adalimumab), 5-aminosalicylate agents, prednisone, antibiotics (metronidazole and ciprofloxacin), and other IBD-related medicines (e.g., laxatives, anti-diarrheals, proton pump inhibitors, and histamine H2 receptor antagonists) as identified by gastroenterology-specific clinical judgment.

Statistical analyses

The primary dependent variables were health expenditures—in total and subcategorized into prescribed medicines or mutually-exclusive clinical care settings (inpatient, outpatient, emergency, and office-based). The primary independent variables were insurance status and poverty status. In calculating standard errors, we accounted for the complex sampling design of MEPS using Stata version 12 (Statacorp, College Station, TX). Sampling variances were estimated using Taylor series linearization (delta method).

Means and proportions were used to produce summary statistics. Multivariate logistic regression models examined the likelihood of incurring annual per capita expenditures above the mean for each respective category (total expenditures, prescribed medicines, or specific clinical care settings). The covariates were age, sex, race/ethnicity, and comorbidities as measured by the HRQL-CI.

Characteristics of IBD patients

We identified 238 individuals with CD, 95 with UC, and 276,369 individuals without IBD (Table 1). MEPS only collects information on conditions associated with medical events, so treated prevalence for CD was 0.17% when weighted to the U.S. population and 0.07% for UC. Unless noted, all subsequent values also refer to nationally representative estimates. The mean age was 47 for CD and 44 for UC, and 43% of CD patients and 59% of UC patients were female.

Compared to the overall population, patients with IBD were more likely to be white (88% vs. 73%) and less likely to be black and Hispanic (6% and 4% respectively vs. 13% and 13%) (Table 1). Those with IBD were also more likely to be in the highest income bracket of ≥400% FPL (48% vs. 39% of the overall population), and more likely to hold private insurance all year (47% vs. 33% of the overall population). The proportions of IBD patients holding public and no insurance were comparable to the overall population. Mean HRQL-CI scores were 2.06 for IBD patients (SE 0.16) and 1.78 for all respondents (SE 0.01).

Direct cost burden of IBD by clinical care setting

Annual per capita mean expenditures for CD, UC, and all IBD were $10,364, $7,827, and $9,528, respectively, each significantly higher than non-IBD expenditures ($4,314, p < 0.05) by $3–6 K more per year (Table 2). Among IBD patients, inpatient mean expenditures ($3,392, SE 578) composed the highest proportion of direct costs, above outpatient, office-based, emergency, or prescribed medicines (p < 0.05) and nearly double the next closest subcategory of office-based expenditures ($1,705, SE 163) (Table 2; Fig. 1). In contrast, emergency expenditures ($252, SE 53) composed the lowest proportion of direct costs (p < 0.05).

In terms of out-of-pocket (OOP) costs, annual per capita mean expenditures for all IBD were again significantly higher than for non-IBD ($1,061 vs. $597, p < 0.05) (Table 2). Although inpatient costs contributed the greatest amount to total IBD expenditures as described above, when considering OOP costs, inpatient (mean $48, SE 17) contributed less than outpatient, office-based, and prescribed medicine costs. The greatest OOP contribution came from office-based (mean $219, SE 28) and prescribed medicine costs (mean $150, SE 18), while emergency costs contributed the least to OOP expenditures (mean $29, SE 9).

Direct cost burden of publicly vs. privately insured IBD patients

When examining the effect of insurance status on annual per capita mean expenditures, publicly insured IBD patients had the highest direct costs by over $10 K ($18,067), over double that of privately insured ($8,014, p < 0.05) and uninsured patients ($5,129, p < 0.05) (Table 3). For those publicly insured patients, the vast majority of their high expenditures derived from inpatient costs, at 5x or $7.8 K more than the next closest subcategory of office-based costs (mean $9,790 vs. $1,941, p < 0.05) (Table 3; Fig. 2A). For privately insured or uninsured patients, however, inpatient costs were not significantly greater than any other subcategories.

When comparing mean expenditures between private and public insurance in each subcategory, only the inpatient subcategory exhibited a significant difference. Publicly insured patients spent 4.5x or $7.6 K more than the privately insured (mean $9,790 vs. $2,174, p < 0.05) (Table 3; Fig. 2A). All other clinical settings and prescribed medicine costs were comparable between IBD patients with public and private insurance.

Effects of no insurance and race/ethnicity on IBD expenditures

Due to the small sample size of uninsured IBD patients, mean expenditures by clinical care setting showed little statistical significance against publicly or privately insured patients (Table 3). Office-based visits, however, showed that the uninsured spent significantly less (mean $529, SE 152) than either the privately insured (mean $1801, SE 256, p < 0.05) or the publicly insured (mean $1941, SE 435, p < 0.05) by factors of 3.5 and 3.7, respectively (Table 3; Fig. 2A).

No relationships were found between mean expenditures for IBD patients and race/ethnicity when comparing white to black patients, white to Hispanic patients, or white to non-white patients.

Disproportionate spending on acute vs. non-acute care by insurance status and income in multivariate analyses

Multivariate logistic regressions paralleled annual per capita mean expenditure trends when examining the effect of insurance status on IBD expenditures. Figure 2B shows that for total expenditures, publicly insured patients had significantly higher odds of spending above the mean than privately insured (OR 2.13, CI [1.08–4.19]) and uninsured patients (OR 4.94, CI [1.26–19.47]). IBD patients with public insurance were more likely to spend more for acute care, defined as inpatient and emergency visits, compared to private or no insurance. Just as was seen with mean expenditures, the increased spending seen with public insurance was disproportionately due to high inpatient spending (public vs. private OR 2.82, CI [1.30–6.10]; public vs. uninsured OR 2.95, CI [1.02–8.54]). Emergency spending was also more likely to be above the mean with public insurance compared to private insurance (OR 2.50, CI [1.23–5.06]).

In contrast, privately insured IBD patients were more likely to spend more for non-acute care, defined as outpatient visits, office-based visits, and prescribed medicines. For outpatient and office-based care, privately insured patients were significantly more likely to spend above the mean than the uninsured (outpatient OR 7.02, CI [1.39–35.40]; office OR 9.69, CI [1.78–52.67]), with no significant relationship to public insurance (Fig. 2B). For prescribed medicines, private insurance was more likely to spend above the mean than public insurance (OR 2.05, CI [1.08–3.88]).

Table 4 shows that poor IBD patients (<100% FPL) were less likely to spend more for non-acute care, compared to not poor IBD patients. Poor patients were significantly less likely to spend above the mean for outpatient (OR 0.38, CI [0.15–0.95]) and office-based care (OR 0.21, CI [0.07–0.62]). With a low n = 41 for poor patients, no other significant differences were found between poor and not poor IBD patients for inpatient, emergency, prescribed medicine, or total expenditures.

No relationships were found in multivariate analyses comparing IBD patients’ expenditures to race/ethnicity when comparing white to black patients, white to Hispanic patients, or white to non-white patients. No significant relationships were found regardless of whether poverty was included as a covariate or not.