Skip to main content
. Author manuscript; available in PMC: 2016 Apr 1.
Published in final edited form as: Med Care. 2016 Apr;54(4):406–413. doi: 10.1097/MLR.0000000000000502

TABLE 2.

Multivariate Cox Proportional Hazards Analysis of Variable Association With Time to Discontinuation of Rosiglitazone After May 21, 2007 Safety Advisory in Black and White Medicare Fee-for-Service Enrollees (n = 45,274)

Variables Total (N) Cases *(N) Follow-up Time
(Person-Years)
Crude Difference in Median
Time to Discontinuation (d)
Adjusted HR
(95% CI)
Race
  Black (reference) 7862 5386 4387
  White 37,412 25,105 20,787 12 0.92 (0.88, 0.95)
Personal income
  Not low (reference) 27,451 19,403 15,374
  Low 17,823 11,088 9799 31 0.84 (0.81, 0.86)
SES index score quintile
  Quintile 5, lowest (reference) 9560 6348 5340
  Quintile 1, highest 7784 5506 4337 −30 1.05 (1.00, 1.11)
  Quintile 2 8281 5589 4578 −15 0.98 (0.88, 1.03)
  Quintile 3 10,390 6939 5773 −7 0.98 (0.96, 1.03)
  Quintile 4 9259 6109 5146 −1 0.97 (0.97, 1.02)
Sex
  Female (reference) 28,149 18,970 15,676
  Male 17,125 11,521 9497 8 0.96 (0.93, 0.99)
Age group (y)
  65–69 (reference) 14,300 9989 8096
  70–74 12,009 8228 6763 6.5 0.96 (0.93, 1.00)
  75–79 9161 6185 5094 0.5 0.94 (0.91, 0.98)
  80–84 5866 3789 3185 15.5 0.92 (0.88, 0.97)
  ≥ 85 3938 2300 2035 34.5 0.83 (0.78, 0.88)
US Census division
  New England (reference) 2144 1472 1186
  Middle Atlantic 6854 4932 3826 −19.5 1.10 (1.02, 1.18)
  East North Central 6312 4151 3498 28.5 0.93 (0.87, 1.01)
  West North Central 3346 2014 1859 67.5 0.79 (0.73, 0.86)
  South Atlantic 9758 6740 5438 2.5 0.99 (0.92, 1.07)
  East South Central 4266 2695 2377 34.5 0.86 (0.79, 0.94)
  West South Central 4815 3210 2665 14.5 0.94 (0.86, 1.02)
  Mountain 2338 1559 1295 32.5 0.95 (0.87, 1.04)
  Pacific 4200 2859 2328 4.5 0.98 (0.90, 1.06)
Cardiovascular (CV) disease
  History of CV disease (reference) 10,247 7044 5633
  No history of CV disease 35,027 23,447 19,540 24 0.90 (0.87, 0.94)
Medication adherence
  Low adherence (reference) (MPR <1) 25,903 18,141 14,575
  High adherence (MPR ≥ 1) 19,371 12,350 10,599 38 0.88 (0.85, 0.91)
Health service utilization
  ≤ 2 Outpatient visits (reference) 27,359 18,084 15,180
  > 2 Outpatient visits 17,915 12,407 9993 −26 1.05 (1.02, 1.09)
Receipt of non-TZD DM medication
  Yes (reference) 29,730 20,504 16,597
  No 15,544 9987 8576 26 0.89 (0.86, 0.92)
Provider type
  Endocrinology/ Cardiology (reference) 2198 1588 1228
  Other 43,076 28,903 23,945 25.5 0.95 (0.89, 1.02)
*

Cases defined as enrollees not censored who discontinue rosiglitazone before the end of study period.

We fit Cox-proportional hazard models to adjust for the selected covariates; confidence intervals that do not cross 1 are significant at P < 0.05; hazard ratio (HR) < 1 implies later discontinuation than comparator group and HR > 1 implies earlier discontinuation than comparator group.

Receipt of at least 1 claim for a nonthiazolidinedione oral diabetic medication in the baseline period before the May 21, 2007 advisory.

CI indicates confidence interval; d, days; DM, diabetes mellitus; HR, hazard ratio; MPR, medication possession ratio; non-TZD, a medication not in the thiazolidinedione class; SES, socioeconomic status; y, years.