| Popis: |
The first paper uses a series of probit models to examine patient demand for isoniazid preventative therapy (9INH) for treatment of latent tuberculosis infection (LTBI) in an ethnically diverse clinic population. I produce an economic model of LTBI adherence that describes the value of immunity via acquired (treatment with 9INH therapy) or innate (no therapeutic intervention) means. The value of acquired versus innate immunity to TB depends on four factors: TB contact potential, immuno-compromised status, duration and acculturation status. This theoretical model is tested with clinic data. Results reveal that patients who have a greater contact potential are less likely to complete LTBI therapy (p < 0.05), while patients with diabetes are more likely to complete (p < 0.05). Immigration status negatively affects completion, while acculturation positively affects adherence (p < 0.05). The second paper uses a series of selection models to examine the costs to treat LTBI. I use a series of generalized Tobit models with random effects to control for individual heterogeneity to isolate the contribution of genetic effects to the monthly costs to treat patients who experience side effects to 9INH. I am able to estimate the differences in 9INH treatment costs between ethnic groups that feature distinct genetic predispositions to suffer side effects from 9INH. Results reveal that self-selection is present in this clinic data, although the classic Heckman two-step estimator is inappropriate to correctly capture this effect. I use a semi-parametric selection estimator and compare results with the Heckman estimator. Predictive capacity modeling shows that an ordered probit selection rule with parametric estimator is the optimal methodology to account for selection. The final paper uses a series of multilevel semi-parametric selection models to examine the comparative-effectiveness and absolute-effectiveness of isoniazid therapy (9INH) for treatment of LTBI. The comparative models incorporate random effects designed to control for individual heterogeneity and semi-parametric selection estimators to isolate the contribution of genetic effects to the total costs to treat. To analyze absolute effectiveness, I simulate cost-effectiveness ratios of an alternative therapy consisting of four months of rifampin (4RIF), a drug known to cause few side effects. I compare these ratios with bootstrapped confidence intervals of the cost-effectiveness ratios for 9INH within each ethnic group. Results reveal wide variations in the comparative effectiveness of 9INH across ethnic groups and biased results when the commonly employed case deletion methodology is used. For absolute effectiveness, 4RIF is cost-effective across all simulation scenarios for all patient groups, except for Asians. |
