Economic justification for rehabilitative services has resulted in the need for outcome based research which could quantify success or failure in individual patients and formulate baseline variables which could predict outcomes. The purpose of this study is to investigate the utilization of baseline clinical, exercise test, and psychosocial variables to predict clinically relevant changes in exercise tolerance of cardiac patients who participated in early outpatient cardiac rehabilitation. Clinical records were analyzed retrospectively to obtain clinical, psychosocial and exercise test data for 94 patients referred to an early outpatient cardiac rehabilitation program at a large urban hospital in the Southeast US. All patients participated in supervised exercise training 3d/wk for 2-3 months. A standardized training outcome score STO) was devised to evaluate training effect by tabulating changes in patients predicted VO2, body weight and exercising heart rates after 8-12 weeks of exercise based cardiac rehabilitation. STO = Predicted VO2 change + BW change- HR change. The Multi-Factorial Analysis was applied to derive coefficients in the STO formula so that the STO scores reflected the independent effects of BW, HR and Predicted V02 changes on training outcome. Patients were classified into one of three possible outcome categories based on STO scores, i.e. improvement, no change, or decline. Thresholds for classifying patients were the following; STO scores greater than or equal to 3 SEM above the mean = improved, (N= 40: 41%), STO scores less than or equal to 3 SEM below the mean = decline, (N=34: 35%), STO scores within 3 SEM= no change, (N=23: 24%). Multiple logistic regression was used to identify patient attributes predictive of improvement, decline, or no change from measures routinely collected at the point of admission to rehabilitation. The model for prediction of improvement correctly classified 70% of patients as those who improved vs. those who did not (sensitivity 70%, specificity 71%). This model generated the following variables as having predictive capabilities; recent CABG, emotional status, social status, calcium channel blocker, recent angioplasty, maximum diastolic BP, maximum systolic BP and resting systolic BP. The model for predicting those who declined vs. those who did not decline demonstrated higher correct classification rate of 74% and specificity (84%). This model generated the following variables as having predictive capabilities; social status, calcium channel blocker, orthopedic limitation, role function, QOL score and Digitalis. However, these models may include certain bias because the same observations to fit the model were also used to estimate the classification errors. Therefore, cross validation was performed utilizing the single point deletion method; this method yielded somewhat lower fraction correct classification rates (66%,69%) and sensitivity rates (56%,44%) for improvement vs. no improvement and decline vs. no decline groups respectively. Conclusion A combined set of baseline clinical, psychosocial and exercise measures can demonstrate moderate success in predicting training outcome based on STO scores in hospital outpatient cardiac rehabilitation. In contrast psychosocial data seem to account for more of the variance in prediction of decline than other types of baseline variables examined in this study. Baseline blood pressure responses both at rest and during exercise were the greatest predictors of improvement. However, cross validation of these models indicates that these results could be biased eliciting overly optimistic predictive capabilities, due to the analysis of fitted data. These models need to be validated in independent sample with patients in similar settings. / Master of Science

Modelos de regressão para variáveis resposta binárias são muito comuns em diversas áreas do conhecimento. O modelo mais utilizado nessas situações é o modelo de regressão logística, que assume que o logito da probabilidade de ocorrência de um dos valores da variável resposta é uma função linear das variáveis preditoras. Quando essa suposição não é razoável, algumas possíveis alternativas são: realizar transformação das variáveis preditoras e/ou inserir termos quadráticos ou cúbicos no modelo. O problema dessa abordagem é que ela dificulta bastante a interpretação dos parâmetros do modelo e, em algumas áreas, é fundamental que eles sejam interpretáveis. Assim, uma abordagem muitas vezes utilizada é a categorização das variáveis preditoras quantitativas do modelo. Sendo assim, este trabalho tem como objetivo propor duas novas classes de métodos de categorização de variáveis contínuas em modelos de regressão para variáveis resposta binárias. A primeira classe de métodos é univariada e busca maximizar a associação entre a variável resposta e a covariável categorizada utilizando medidas de associação para variáveis qualitativas. Já a classe de métodos multivariada tenta incorporar a estrutura de dependência entre as covariáveis do modelo através da categorização conjunta de todas as variáveis preditoras. Para avaliar o desempenho, aplicamos as classes de métodos propostas e quatro métodos de categorização existentes em 3 bases de dados relacionadas à área de risco de crédito e a dois cenários de dados simulados. Os resultados nas bases reais sugerem que a classe univariada proposta têm um desempenho superior aos métodos existentes quando comparamos o poder preditivo do modelo de regressão logística. Já os resultados nas bases de dados simuladas sugerem que ambas as classes propostas possuem um desempenho superior aos métodos existentes. Em relação ao desempenho computacional, o método multivariado mostrou-se inferior e o univariado é superior aos métodos existentes. / Regression models for binary response variables are very common in several areas of knowledge. The most used model in these situations is the logistic regression model, which assumes that the logit of the probability of a certain event is a linear function of the predictors variables. When this assumption is not reasonable, it is common to make some changes in the model, such as: transformation of predictor variables and/or add quadratic or cubic terms to the model. The problem with this approach is that it hinders parameter interpretation, and in some areas it is fundamental to interpret the parameters. Thus, a common approach is to categorize the quantitative covariates. This work aims to propose two new classes of categorization methods for continuous variables in binary regression models. The first class of methods is univariate and seeks to maximize the association between the response variable and the categorized covariate using measures of association for qualitative variables. The second class of methods is multivariate and incorporates the predictor variables correlation structure through the joint categorization of all covariates. To evaluate the performance, we applied the proposed methods and four existing categorization methods in 3 credit scoring databases and in two simulated cenarios. The results in the real databases suggest that the proposed univariate class of categorization methods performs better than the existing methods when we compare the predictive power of the logistic regression model. The results in the simulated databases suggest that both proposed classes perform better than the existing methods. Regarding computational performance, the multivariate method is inferior and the univariate method is superior to the existing methods.

Modelos de regressão para variáveis resposta binárias são muito comuns em diversas áreas do conhecimento. O modelo mais utilizado nessas situações é o modelo de regressão logística, que assume que o logito da probabilidade de ocorrência de um dos valores da variável resposta é uma função linear das variáveis preditoras. Quando essa suposição não é razoável, algumas possíveis alternativas são: realizar transformação das variáveis preditoras e/ou inserir termos quadráticos ou cúbicos no modelo. O problema dessa abordagem é que ela dificulta bastante a interpretação dos parâmetros do modelo e, em algumas áreas, é fundamental que eles sejam interpretáveis. Assim, uma abordagem muitas vezes utilizada é a categorização das variáveis preditoras quantitativas do modelo. Sendo assim, este trabalho tem como objetivo propor duas novas classes de métodos de categorização de variáveis contínuas em modelos de regressão para variáveis resposta binárias. A primeira classe de métodos é univariada e busca maximizar a associação entre a variável resposta e a covariável categorizada utilizando medidas de associação para variáveis qualitativas. Já a classe de métodos multivariada tenta incorporar a estrutura de dependência entre as covariáveis do modelo através da categorização conjunta de todas as variáveis preditoras. Para avaliar o desempenho, aplicamos as classes de métodos propostas e quatro métodos de categorização existentes em 3 bases de dados relacionadas à área de risco de crédito e a dois cenários de dados simulados. Os resultados nas bases reais sugerem que a classe univariada proposta têm um desempenho superior aos métodos existentes quando comparamos o poder preditivo do modelo de regressão logística. Já os resultados nas bases de dados simuladas sugerem que ambas as classes propostas possuem um desempenho superior aos métodos existentes. Em relação ao desempenho computacional, o método multivariado mostrou-se inferior e o univariado é superior aos métodos existentes. / Regression models for binary response variables are very common in several areas of knowledge. The most used model in these situations is the logistic regression model, which assumes that the logit of the probability of a certain event is a linear function of the predictors variables. When this assumption is not reasonable, it is common to make some changes in the model, such as: transformation of predictor variables and/or add quadratic or cubic terms to the model. The problem with this approach is that it hinders parameter interpretation, and in some areas it is fundamental to interpret the parameters. Thus, a common approach is to categorize the quantitative covariates. This work aims to propose two new classes of categorization methods for continuous variables in binary regression models. The first class of methods is univariate and seeks to maximize the association between the response variable and the categorized covariate using measures of association for qualitative variables. The second class of methods is multivariate and incorporates the predictor variables correlation structure through the joint categorization of all covariates. To evaluate the performance, we applied the proposed methods and four existing categorization methods in 3 credit scoring databases and in two simulated cenarios. The results in the real databases suggest that the proposed univariate class of categorization methods performs better than the existing methods when we compare the predictive power of the logistic regression model. The results in the simulated databases suggest that both proposed classes perform better than the existing methods. Regarding computational performance, the multivariate method is inferior and the univariate method is superior to the existing methods.

The rapid growth in online learning opportunities and online courses in K-12 education is well documented in the literature. Studies conducted by various researchers that have focused on the K-12 population of online learners demonstrate that certain online learner characteristics and online learning environment characteristics may impact the likelihood of students passing or failing online courses. Research has produced models that predict online course success with measurable degrees of accuracy. This descriptive study examines characteristics of students enrolled in online high school courses provided by a virtual learning program administered by a single Virginia public school district. The study determined that students’ prior academic success; confidence in their technology skills and access to technology; confidence in their ability to achieve; and strong beliefs in their organizational skills proved to have a significant statistical relationship with online course success. The study developed a model with these factors that predicted success in online courses with a high degree of accuracy and predicted failure with a moderate degree of accuracy.The study has policy implications for public school leaders in Virginia as they implement recent state legislation requiring students to successfully complete a virtual course to graduate from public high school. The study indicates that additional research is warranted to further delineate learner and learning environment characteristics producing a model that more accurately predicts failure in online courses. Additional research is warranted with larger samples from single district virtual programs.

Submitted by Aelson Maciera (aelsoncm@terra.com.br) on 2017-08-16T20:00:07Z No. of bitstreams: 1 DissDMBS.pdf: 821487 bytes, checksum: 497fc9b102478d03042a1c3d10a45c19 (MD5) / Approved for entry into archive by Ronildo Prado (bco.producao.intelectual@gmail.com) on 2018-01-29T18:10:09Z (GMT) No. of bitstreams: 1 DissDMBS.pdf: 821487 bytes, checksum: 497fc9b102478d03042a1c3d10a45c19 (MD5) / Approved for entry into archive by Ronildo Prado (bco.producao.intelectual@gmail.com) on 2018-01-29T18:10:17Z (GMT) No. of bitstreams: 1 DissDMBS.pdf: 821487 bytes, checksum: 497fc9b102478d03042a1c3d10a45c19 (MD5) / Made available in DSpace on 2018-01-29T18:14:39Z (GMT). No. of bitstreams: 1 DissDMBS.pdf: 821487 bytes, checksum: 497fc9b102478d03042a1c3d10a45c19 (MD5) Previous issue date: 2017-06-13 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Regression models for binary response variables are very common in several areas of knowledge. The most used model in these situations is the logistic regression model, which assumes that the logit of the probability of a certain event is a linear function of the predictors variables. When this assumption is not reasonable, it is common to make some changes in the model, such as: transformation of predictor variables and/or add quadratic or cubic terms to the model. The problem with this approach is that it hinders parameter interpretation, and in some areas it is fundamental to interpret the parameters. Thus, a common approach is to categorize the quantitative covariates. This work aims to propose two new classes of categorization methods for continuous variables in binary regression models. The first class of methods is univariate and seeks to maximize the association between the response variable and the categorized covariate using measures of association for qualitative variables. The second class of methods is multivariate and incorporates the predictor variables correlation structure through the joint categorization of all covariates. To evaluate the performance, we applied the proposed methods and four existing categorization methods in 3 credit scoring databases and in two simulated cenarios. The results in the real databases suggest that the proposed univariate class of categorization methods performs better than the existing methods when we compare the predictive power of the logistic regression model. The results in the simulated databases suggest that both proposed classes perform better than the existing methods. Regarding computational performance, the multivariate method is inferior and the univariate method is superior to the existing methods. / Modelos de regressão para variáveis resposta binárias são muito comuns em diversas áreas do conhecimento. O modelo mais utilizado nessas situações é o modelo de regressão logística, que assume que o logito da probabilidade de ocorrência de um dos valores da variável resposta é uma função linear das variáveis preditoras. Quando essa suposição não é razoável, algumas possíveis alternativas são: realizar transformação das variáveis preditoras e/ou inserir termos quadráticos ou cúbicos no modelo. O problema dessa abordagem é que ela dificulta bastante a interpretação dos parâmetros do modelo e, em algumas áreas, é fundamental que eles sejam interpretáveis. Assim, uma abordagem muitas vezes utilizada é a categorização das variáveis preditoras quantitativas do modelo. Sendo assim, este trabalho tem como objetivo propor duas novas classes de métodos de categorização de variáveis contínuas em modelos de regressão para variáveis resposta binárias. A primeira classe de métodos é univariada e busca maximizar a associação entre a variável resposta e a covariável categorizada utilizando medidas de associação para variáveis qualitativas. Já a classe de métodos multivariada tenta incorporar a estrutura de dependência entre as covariáveis do modelo através da categorização conjunta de todas as variáveis preditoras. Para avaliar o desempenho, aplicamos as classes de métodos propostas e quatro métodos de categorização existentes em 3 bases de dados relacionadas à área de risco de crédito e a dois cenários de dados simulados. Os resultados nas bases reais sugerem que a classe univariada proposta têm um desempenho superior aos métodos existentes quando comparamos o poder preditivo do modelo de regressão logística. Já os resultados nas bases de dados simuladas sugerem que ambas as classes propostas possuem um desempenho superior aos métodos existentes. Em relação ao desempenho computacional, o método multivariado mostrou-se inferior e o univariado é superior aos métodos existentes.

Orientador: Célio F.B. Haddad / Resumo: Na última década houve um grande desenvolvimento nos Modelos de Distribuição de Espécies (MDE), com diversas aplicações na conservação da biodiversidade. No entanto, apesar dos avanços recentes, a seleção de variáveis preditoras tem sido relativamente negligenciada na construção dos MDE. Este procedimento deveria ser um dos passos cruciais do processo de modelagem, já que as variáveis preditoras estão relacionadas diretamente à capacidade dos modelos de capturar os requisitos ambientais das espécies. Neste contexto, os anfíbios são excelentes organismos modelo para avaliar a importância da seleção de variáveis preditoras ecologicamente significativas no MDE. Isto pode trazer avanços para a biogeografia e biologia da conservação, uma vez que os anfíbios são usados como bioindicadores da qualidade ambiental e da integridade de hábitat. A presente tese de doutorado teve como objetivo principal verificar o efeito da utilização de variáveis preditoras ecologicamente significativas no processo de modelagem dos anfíbios e posteriormente aplicar parte deste conhecimento na comunidade de anfíbios do Estado de São Paulo, visando verificar o potencial desta metodologia para identificar áreas de alto valor de riqueza de anfíbios e verificar também o potencial de invasão de Eleutherodactylus jonhstonei, uma espécie de anfíbio invasora registrada para o Estado de São Paulo. No primeiro capítulo avaliamos a importância da seleção de variáveis essenciais ao MDE usando os anfíbios como estudo... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: In the last decade there has been a great development in the Species Distribution Models (SDM), with several applications in conservation planning. However, despite recent advances, the selection of predictor variables has been relatively neglected in the construction of SDM. This methodological approach should be one of the critical steps of the modeling process, as the predictor variables are directly related to the ability of models to capture the environmental requirements of the species. In this context, amphibians are excellent model for assessing the importance of selecting ecologically meaningful variables in the SDM. This methodology may lead to advances in biogeography and conservation biology, since amphibians are used as bioindicators of environmental quality and habitat integrity. The aim of the work was to verify the effect of the use of ecologically meaningful variables in the amphibian modeling process and to apply part of this knowledge to the amphibian community of São Paulo state, checking the potential of this methodology to identify areas of high amphibian richness value and to verify the potential invasion of Eleutherodactylus jonhstonei, an invasive amphibian species registered in São Paulo state. In the first chapter we evaluated the importance of selecting essential variables in SDM using amphibians as a case study. The second chapter deals specifically with the amphibian modeling protocol of São Paulo state. The central focus of this chapter has been... (Complete abstract click electronic access below) / Doutor

No description available.

Abstract Purpose: The United States has the highest rate of teen pregnancy of any industrialized nation. Adolescents who have their first sexual intercourse at a young age are at increased risk for teen pregnancies and acquiring a sexually transmitted disease. This study examines predictors of early onset sexual intercourse in male and female residents of the United States. Methods: A nationally representative sample of N=7,643 females and N=4.928 males ages 15-44 was procured from the 2002 National Survey of Family Growth (NSFG), Cycle 6. Age at first sexual intercourse was used to define early onset of sexual debut(<18 years). Socio-demographic and behavioral characteristics of the respondents, demographic and selected reproductive characteristics of the respondent's parents were examined using multiple logistic regression modeling. Results: Non-fispanic black, being raised without both parents, having a mother less than 18 years old at the age of first birth and age difference between partners were significant predictors of early onset of sexual intercourse for both males and females. Maternal education less than high school was a significant protective factor for female respondents [OR=0.72 (95%CI=0.58- 0.90)] and paternal education completed high school only [OR=1.4 (95% CI=l. 1-1.7)] was a significant risk factor for male respondents. Conclusions: Racelethnicity, age difference between partners, not being raised by both parents, having a mother who had her first birth before the age of 18 and parental education are important predictor variables. Further study should be conducted to investigate the protective effect of lack of maternal education for female respondents. Intervention programs for teen pregnancy and sexually transmitted infection prevention should target these at risk groups.

Made available in DSpace on 2016-06-02T20:30:09Z (GMT). No. of bitstreams: 1 5826.pdf: 1477811 bytes, checksum: 7e98346cb582d2360d7e7ca4a0cb751a (MD5) Previous issue date: 2014-02-20 / Universidade Federal de Minas Gerais / School victimization may favor the occurrence of traumatic symptoms and Post Traumatic Stress Disorder (PTSD). This Doctoral Thesis had the following objectives: validate the American retrospective instrument Student Alienation and Trauma Survey - R (SATS-R), to Brazil, in terms of construct and content validity; characterize how violence is expressed at school, identifying the main types of violence, the worst school events experienced by students, the frequency and duration of these events, main perpetrators, as well as victims' characteristics (age, grade and type of school); investigate the occurrence of traumatic symptoms, especially PTSD, after the worst school experience; analyze the association between PTSD symptoms and variables associated with the worst school experience, and investigate the relationship between the explanatory variables (individual characteristics and aversive school experiences), and development of PTSD symptoms, using an ordinal logistic regression model. The study included 691 students (54.8% female and 45.2% male), of a public university in São Paulo State, Brazil, who responded to Portuguese versions of the Student Alienation and Trauma Survey-R (SATS-R) and the Post-Traumatic Stress Disorder Checklist - Civilian Version (PCL-C). In terms of content validity, the following procedures were conducted: translation, back-translation, semantic equivalence, instrument analysis by experts in the field, and a sample assessment of the target population; for construct validity, an exploratory factor analysis was conduct and Cronbach's alpha was calculated. The study results indicate the feasibility of using the instrument in the Brazilian context for research purposes. Frequency of victimization types reported by participants were: relational violence (at least one item reported by 85.2%); verbal violence (77.7%); physical violence (50.8%); unfair discipline (43.1%); property violence (33.4%); witnessing violence (27.9%); and sexual violence (21.4%). The most frequent types of worst school experiences described were: relational (35.7%), and verbal violence (27.4%). Girls experienced more episodes of verbal, relational and sexual violence, and boys experienced more physical violence and unfair discipline, and the aggressors were mostly male students. The mean age when these worst experiences occurred was 12.3 years, and although most events occurred at low frequency and with short duration, a considerable percentage of participants indicated a duration of "years", particularly in verbal and relational victimization cases. Most participants indicated that they were greatly bothered by their worst school experience, and 7.8% had PTSD symptons after experiencing this event. The percentage of participants with clinically significant scores on the subscales ranged from 4.7% (somatic symptoms) to 20% (hypervigilance), and described symptoms frequently in the literature, such as depression, hopelessness, cognitive difficulties and traumatic event recollection. Significant variables for the regression model were: age, duration and discomfort after the worst experience; relational violence; and verbal violence. In general, student who expressed the greatest discomfort, reported traumatic experiences that were longer in duration, occured when they were older, and the greater the number of verbal and relational victimization events experienced, the greater the possibility of presenting clinically significant symptoms of PTSD. Despite the limitations of the retrospective methodo, the study obtained interesting results which coincide with the literature, drawing attention to the long-term effects of school victimization. In addition, these results may contribute to the development of new research on the topic, as well as offering treatment parameters for victims who were traumatized in school, improving school violence prevention programs. / A vitimização escolar pode favorecer a ocorrência de sintomas traumáticos, como de Transtorno de Estresse Pós-Traumático (TEPT). A presente Tese de Doutorado teve os seguintes objetivos: buscar evidências de validade de conteúdo e de constructo do instrumento retrospectivo norte-americano Student Alienation and Trauma Survey R (SATS-R), para o Brasil; caracterizar como a violência se expressa na escola, identificando os principais tipos de violência, as piores experiências escolares vivenciadas por estudantes, a frequência e a duração desses eventos, os agressores principais, bem como as características das vítimas (idade, série e tipo de escola); investigar a ocorrência de sintomas traumáticos, principalmente TEPT, nos estudantes, após a vivência da pior experiência escolar; analisar a associação dos sintomas de TEPT a variáveis relacionadas à pior experiência escolar; e investigar o relacionamento de variáveis explicativas (características do indivíduo e das experiências escolares aversivas vivenciadas) e o desenvolvimento de sintomas de TEPT, por meio de um modelo de regressão logística ordinal. Participaram do estudo 691 estudantes (54,8% do sexo feminino e 45,2% do masculino), de uma universidade pública do interior do estado de São Paulo, que responderam a versões brasileiras dos instrumentos Student Alienation and Trauma Survey e Post-Traumatic Stress Disorder Checklist Civilian Version (PCL-C). Para a validação de conteúdo, foram feitas: tradução, retrotradução, equivalência semântica, análise do instrumento por profissionais da área e avaliação por amostra da população alvo; para a validação de constructo foi realizada análise fatorial exploratória e cálculo do alfa de Cronbach do instrumento. Os resultados do estudo apontaram para a viabilidade da utilização do instrumento no contexto brasileiro para fins de pesquisa. A frequência dos tipos de vitimização relatados pelos participantes foi: violência relacional (ao menos um item relatado por 85,2%), violência verbal (77,7%) violência física (50,8%), disciplina injusta (43,1%), violência contra o patrimônio (33,4%), presenciar violência (27,9%) e violência sexual (21,4%). Os tipos de piores experiências mais frequentes descritos foram violência relacional (35,7%) e verbal (27,4%). As meninas sofreram mais episódios de violência verbal, relacional e sexual e os meninos violência física e disciplina injusta, sendo que os agressores foram, em sua maioria, estudantes e do sexo masculino. A idade média de ocorrência das piores experiências foi 12,3 anos e, embora a maior parte dos eventos tenha ocorrido em baixa frequência e com curta duração, porcentagem considerável dos participantes apontou duração de anos nos casos de vitimização verbal e relacional, principalmente. A maior parte dos participantes apontou ter se incomodado muito com a pior experiência escolar, sendo que 7,8% apresentaram indicação de TEPT após a vivência dessa experiência. A porcentagem de participantes com escores clinicamente significativos nas subescalas variou de 4,7% (sintomas somáticos) a 20% (hipervigilância), sendo frequentes sintomas comumente descritos na literatura como depressão, desesperança, dificuldades cognitivas e rememoração do evento traumático. As variáveis significativas para o modelo de regressão realizado foram: idade, duração e incômodo após a pior experiência, violência relacional e violência verbal. De forma geral, quanto maior o incômodo do estudante, maior a duração da experiência, maior a idade e quanto mais eventos vivenciados de vitimização relacional e verbal, maior a possibilidade de apresentação de sintomas clinicamente significativos de TEPT. Apesar das limitações da metodologia retrospectiva, foram obtidos resultados interessantes que coincidem com a literatura, chamando a atenção para os efeitos a longo prazo da vitimização escolar. Além disso, o estudo pode contribuir para o desenvolvimento de novas pesquisas sobre o tema, bem como oferecer parâmetros de tratamento às vítimas que apresentem sintomas decorrentes de experiências traumáticas na escola, podendo aprimorar, também, programas de prevenção à violência escolar.

This was a retrospective study of a community-based methadone maintenance treatment (MMT) program in Toronto. Participants (N = 170) were federally sentenced adult male offenders admitted to this voluntary program between 1997 and 2009 while subject to community supervision following incarceration. The primary investigation examined correlates of treatment responsivity, with principal outcome measures including MMT clients’ rates of: (i) illicit drug use; and (ii) completion of conditional (parole) or statutory release (SR). For a subset (n = 74), recidivism rates were examined after a 9-year interval. Findings included strong convergent evidence from logistic regression and ROC analyses that an empirically and theoretically derived set of five variables was a stable and highly significant (p <.001) predictor of release outcome. Using five factors related to risk (work/school status, security level of releasing institution, total PCL-R score, history of institutional drug use, and days at risk), release outcome was predicted with an overall classification accuracy of 88%, with high specificity (86%) and sensitivity (89%). The logistic regression model generated an R2 of .55 and the accompanying AUC was .89, both substantial. Work/school status had an extremely large positive association with successful completion of community supervision, accounting for > half of the total variance explained by the five-factor model and increasing the estimated odds of successful release outcome by > 15-fold. Also, when in the MMT program, clients' risk taking behaviour was significantly moderated, with low overall base rates of illicit drug use, yet the rate of parole/SR revocation (71%) was high. The 9-year follow-up showed a high mortality rate (15%) overall. Revocation of release while in the MMT program was associated with a significantly higher rate and more violent recidivism at follow-up. Results are discussed within the context of: (a) Andrews' and Bonta's psychology of criminal conduct; (b) the incompatibility of a harm reduction treatment model with an abstinence-based parole decision-making model; (c) changing drug use profiles among MMT clients; (d) a strength-based approach to correctional intervention focusing on educational and vocational retraining initiatives; and (e) creation of a user friendly case-based screening algorithm for prediction of release outcome for new releases.