AJPB

A Linezolid Prior Authorization Program: Clinical and Economic Outcomes | Page 2

Published Online: April 16, 2014
Catherine I. Starner, PharmD, BCPS; R. Scott McClelland, PharmD; Yang Qiu, MS; Richard A. Zabinski, PharmD; Nancy Cotter, PharmD; and Patrick P. Gleason, PharmD, BCPS, FCCP
During the period of January 1, 2011, through June 30, 2011, each member’s earliest linezolid pharmacy claim that was rejected (for the intervention group) and paid (for the comparison group) was identified and defined as the index claim. All pharmacy claims were identified using the assigned Generic Product Identifi er (GPI, Medi- Span, Indianapolis, Indiana), starting with 1623. Member characteristics from the index claim included age, gender, and zip code–derived median income and education. Charlson Comorbidity Index (CCI) score was calculated in the pre-6–month period.17,18 The baseline number of outpatient visits, hospitalizations, and emergency department (ED) visits were also collected in the 6 months prior to the index linezolid claim. Members’ infectious organism(s) were collected using International Classification of Diseases, Ninth Revision (ICD-9) codes found in the 6 months prior to the index claim in any field of the member’s medical claims. Infectious organisms were hierarchically ordered and categorized as follows: (1) infections with drug-resistant organisms specific to vancomycin, MRSA pneumonia, MRSA septicemia, or MRSA carrier status; (2) infections due to undefi ned organisms; and (3) all other infectious organisms.

To assess the drug-use patterns following the index claim, all pharmacy and medical claims were queried during the 30 days after the index date for the presence of drug supply for linezolid or any other antibiotic agent. Linezolid medical claims were identified using the Healthcare Common Procedure Coding System (Level II) J2020 code. For evaluating drug supply presence on specific follow-up dates, an end date for each claim (ie, fill date plus days of supply on the claim) was created. On each day evaluated after the index date (ie, day 30 and 60), presence of linezolid and any other antibiotic was reported in the following hierarchical order: (1) linezolid with or without another antibiotic; (2) other antibiotics (without linezolid); and (3) no antibiotic therapy. Hospitalizations and ED visits were identifi ed in the medical claims using revenue codes or place-of-service codes when revenue codes were not available. Outpatient visits were also identifi ed using medical claims Current Procedural Terminology codes (99201 through 99205 and 99211 through 99215). Economic outcomes were evaluated in the follow-up periods using all pharmacy and medical claims and total paid amounts (member share plus health insurer payment). Costs specific to linezolid and all other antibiotics are reported separately. Linezolid and all other antibiotic costs from both the pharmacy and medical benefi t were identifi ed and reported. Total cost included all pharmacy plus all medical costs.

Unadjusted comparisons between groups were performed with the χ² test for categorical variables, Fisher’s exact test for those with counts less than 5, Wilcoxon rank-sum test for counts, and the likelihood ratio test for expenditures. A logistic regression model was used to test hospitalization and ED visit differences between the intervention and comparison groups, with adjustment for: age group, male gender, Charlson Comorbidity Index score categories (0 as the reference, 1-2, and greater than or equal to 3), zip code–derived education dichotomized to those with a bachelor’s degree or above and all others, zip code–derived income dichotomized to $0 to $49,999 and greater than or equal to $50,000, existence of baseline hospitalization or ED visit, and hierarchical presence of specific ICD-9–coded infectious organisms (Table 1) with Group 2 as the reference. The logistic regression fit was assessed using the C-statistic and Hosmer-Lemeshow goodness-of-fit statistic. Cost analyses were performed using the generalized linear model with Gamma distribution and adjusted for the same covariates listed above. The overall fit of the generalized linear model was assessed using the scaled deviance and Pearson χ² goodness- of-fi t statistic. All statistical testing was performed using SAS version 9.2 (Cary, North Carolina). All P values were 2-sided with an a priori alpha of .05. A sensitivity analysis was performed repeating all analyses using a 60- day follow-up period, post index date.

RESULTS

The Figure shows the analysis fl ow for the intervention and comparison health plan members. Of the 1,167,888 eligible members exposed to the PA in the intervention group, 217 (2 per 10,000) had a rejected linezolid index claim between January 1, 2011, and June 30, 2011. The prevalence of members who were continuously enrolled 6 months prior to their index claim was 185 (85.3%) of the 217 members with a linezolid rejected claim. The comparison group had 77 (1 per 10,000) members with a paid linezolid claim, and 69 (89.6%) met analysis criteria. A χ² test was performed to test the difference in rate of linezolid prescribing between the 2 groups, and the intervention group had a significantly higher linezolid prescribing rate compared with the comparison group. (P <.001). This difference refl ects national antibiotic prescribing trends, which show higher rates in the South.11

Table 1 shows that all baseline characteristics were similar, except that the intervention group had a significantly lower percentage of members with a zip code–derived median income of >$50,000 (P = .032). In addition, directional differences existed between the categories of infectious organisms (category 1: infections with drug resistant organisms specific to vancomycin, MRSA pneumonia, septicemia, or carrier status was found in 27.0% vs 39.1% of members; and category 2: infections due to undefined organisms found in 61.1% vs 50.7% of members, among intervention and comparison groups, respectively).

The unadjusted number of offi ce visits, hospitalizations, and ED visits are shown in Table 2. During the 30-day follow-up, the intervention group had an average of 3.0 office visits (standard deviation [SD] 2.1) and the comparison group 2.0 (SD 1.4), P = .332. There were no statistical differences in unadjusted hospitalizations (14.6% vs 17.4%) or ED visits (18.4% vs 14.5%) between the intervention and comparison groups, respectively.

PDF is available on the last page.

Issue: March/April 2014
More on AJMC.COM