Margaret K. Pasquale, PhD; Anthony M. Louder, PhD, RPh; Michael C. Deminski, MS, RPh; Richard B. Chambers, MSPH; and Seema Haider, MSc
Claims data for Medicare members prescribed oral linezolid and discharged from an inpatient stay for complicated or uncomplicated SSTI or pneumonia were examined for benefit design, out-of-pocket costs, reversals, rehospitalizations, and healthcare costs. Visual inspection was used to distinguish between copay and coinsurance because the medical claims did not contain an explicit indicator for copay or coinsurance. All integer values (ie, $6, $10, $25, $30, $100, $150) and known noninteger values of Medicare Part D copays (ie, $5.60, $6.30) were classified as copay, while all remaining noninteger values (ie, $57.43, $238.59) were classified as coinsurance. For all descriptive analyses, means were compared using 2-sample t tests and count variables were compared using x2
tests. Demographic and clinical characteristics included age, sex, geographic region, low-income status and dual eligibility (Medicare and Medicaid eligibility), characteristics of the initial broader window typically necessary for medical claims–based comorbidity scores such as the Charlson Comorbidity Index score.14
The impact of the reversal on postindex healthcare costs (plan payment plus member medical cost) was modeled using a generalized linear model (GLM) with a gamma distribution as its probability distribution and log link as its link function. Generalized linear models are commonly utilized to account for distributions of highly skewed data, characteristic of medical expenditures.15
The log link function for the GLM is specified below, where μ refers to postindex healthcare costs:
log u = reversal + Out-of-PocketCategory + age
+ gender + region + LowIncomeSubsidy-DualEligible
+ RxRiskV + ICUstay + surgery + pre-indexhealth
Covariates included in the model were age, sex, geographic region, low-income subsidy/dual eligibility, baseline RxRisk-V score, a surgical procedure or an intensive care unithospitalization such as surgery or intensive care unit stay, and the RxRisk-V comorbidity score, a surgical procedure or an intensive care unit stay during the initial hospitalization, out-of-pocket costs, and preindex healthcare costs (per $1000). In order to include data on members with no costs during the follow-up period, $1 was added to all member costs. The models provided adjusted mean costs and 95% confidence intervals for the prescription fill and reversal groups.RESULTS
A total 1062 Medicare members were available for analysis. Among the total sample, 16.5% of the members reversed their prescription for oral linezolid. Demographic and clinical characteristics by fill versus reversal groups indicated there were no statistical differences in age, sex, or geographic region (Table 1
). However, a higher percentage of the members filling their linezolid prescription had low-income subsidy/ dual-eligibility status compared with members reversing their linezolid prescription (P
<.0001). A majority of the characteristics of the initial hospitalization were similar, with the exception that a statistically higher percentage of reversal members were hospitalized for complicated or uncomplicated SSTI (P
= .0148). That corresponded to a higher percentage of fill members hospitalized for pneumonia (P
= .0106). The RxRisk-V comorbidity scores and preindex total healthcare costs were not statistically different between the 2 groups.
shows that as out-of-pocket costs increased the percentage of members reversing their prescriptions also increased, with out-of-pocket costs above $100 resulting in a reversal rate as high as 27% (P
<.0001). It is assumed that patients with out-of-pocket costs of more than $100 were subject to a coinsurance benefit versus a copay for most patients with out-of-pocket costs of less than $100. Mean (standard deviation) out-of-pocket costs for members with a copay were $7.05 ($14.89); for members with coinsurance the mean out-of-pocket cost was $466.52 ($574.67).
Among members who reversed their prescriptions for oral linezolid, 73% filled a prescription for an alternative antibiotic during the 30 days after discharge, while 27% received no antibiotic. Among patients with prescriptions for alternative antibiotics, a small percentage (9%) were treated with oral or parenteral vancomycin, and 0.6% were treated with parenteral tigecycline. Other, more common antibiotics with prescriptions filled during the 30 days after discharge were amoxicillin (10.5%), ciprofloxacin (15.7%), clindamycin (11.0%), doxycycline (10.5%), and levofloxacin (7.0%) (complete list provided upon request).
As shown in Table 2
, infection-related (complicated or uncomplicated SSTI or pneumonia) rehospitalizations were 14 percentage points higher (P
<.0001) and all-cause rehospitalizations were 10 percentage points higher (P
= .0027) in the reversal group versus the fill group. For each type of infection (complicated or uncomplicated SSTI, pneumonia, or both SSTI and pneumonia), no significant difference was detected between the fill and reversal groups. When grouping prescription fill versus reversal groups by the percentage readmitted for the same infection as the one diagnosed at initial hospitalization, a higher percentage of the reversal group was hospitalized for the same infection: 20% reversal group versus 9% fill group (P
<.0001) for complicated or uncomplicated SSTI and 3% reversal group versus 1% fill group (P
= .0268) for pneumonia.
Unadjusted postindex prescription drug, medical, and total healthcare costs are reported in Table 3
. Whereas postindex prescription drug costs were significantly lower for members with a reversal (P
<.0001), postindex medical costs were significantly higher for these members (P
= .0013) compared with members with a fill. The combined total unadjusted healthcare costs were not statistically different between the 2 groups (P
Adjusted prescription drug, medical, and total healthcare costs are reported in Table 4
. After adjusting for demographic and clinical characteristics, differences in the prescription drug and medical costs remained statistically significant between the fill and reversal groups. Notably, with adjustment, the difference in total healthcare costs between the fill and reversal groups became statistically significant (P
= .0349), with the mean healthcare cost for the reversal group being $1280.93 higher than that for the fill group.
PDF is available on the last page.