Article 12 : Who Fails to Complete Tuberculosis Treatment in Haiti ’ s National Prison , and What Innovations Might Improve Success ?
“ Housing location ” at the prison was deemed non-identifiable because as many as 100 individuals may be in a multi-occupancy cell at any given time . Furthermore , pre-treatment housing assignments were not fixed , so therefore , it would not be feasible to identify a patient based on a past housing assignment .
Patients were categorized as new TB cases if they had no prior history of diagnosis / treatment , or as retreatment cases if they had a prior history of TB disease treated for at least one month .
TB types included pulmonary , either laboratory-confirmed or clinically diagnosed , and EPTB cases .
All data entered into the database were reviewed for errors and inconsistencies by checking outof-range values for variables , inconsistent spelling of word strings , and cross tabulations to ensure proper coding of variables .
Data analysis Univariate logistic regression was used to determine which , if any , covariates were significantly associated with incomplete treatment in prison . Odds ratios were considered significant using an alpha level of 0.05 . A final model was compiled using multivariate logistic regression and backwards elimination to determine risk factors for failing to complete treatment in prison .
Based on the results of the univariate logistic regression , the following covariates were considered for the full model : age , TB case category , TB type , and HIV status . These covariates were assessed for collinearity , effect modification , and confounding . Evidence suggested that age was a confounder ( i . e ., age alone , and not TB category or type , or HIV status , could affect the outcomes ), so age-adjusted and crude estimates were found using the final multivariate logistic regression model .
A secondary analysis , using a McNemar chi-squared test , was conducted to explore if mortality rates differed by TB type , TB case category , or HIV status . Cox regression analysis was used to determine if any covariates significantly affected the time to complete treatment in prison .
Institutional Review Board The Emory University Institutional Review Board reviewed and approved this study .
Results
Distribution of tuberculosis cases Out of the 661 patients , 556 ( 84.1 %) completed treatment while in prison and 105 ( 15.9 %) did not . New diagnoses comprised 536 out of 661 ( 81.0 %) cases and 124 ( 18.8 %) cases were determined as retreatment [ Table 1 ]. Eighteen out of 124 ( 14.5 %) retreatment cases failed to complete treatment in prison . Eighty-six out of 536 ( 16.0 %) new cases failed to complete treatment in prison . For the entire cohort , pre-treatment ages ranged from 18-71 years ( mean 31 years [ S . D . 8.0 ]). Pulmonary TB accounted for 535 ( 85.9 %) of the cases and EPTB for 88 ( 14.1 %) of the cases . Additionally , 554 patients ( 89.1 %) were HIV-negative and 68 ( 10.9 %) were living with HIV ; HIV status was missing for 39 ( 6.0 %) cases .
179