Reliability and validity of the SF-12v2 in the medical expenditure panel survey

Objective  Evaluate the reliability and validity of the Medical Outcomes Study Short-Form version 2 (SF-12v2) in the 2003–2004 Medical Expenditure Panel Survey (MEPS). Research design  Data were collected in the self-administered mail-out questionnaire and face-to-face interviews of the MEPS (n = 20,661). Internal consistency and test–retest reliability and construct, discriminate, predictive and concurrent validity were tested. The EQ-5D, perceived health and mental health questions were used to test construct and discriminate validity. Self-reported work, physical and cognitive limits tested predictive validity and number of chronic conditions assessed concurrent validity. Results  Both Mental Component Summary Scores (MCS) and Physical Component Summary Scores (PCS) were shown to have high internal consistency reliability (α > .80). PCS showed high test–retest reliability (ICC = .78) while MCS demonstrated moderate reliability (ICC = .60). PCS had high convergent validity for EQ-5D items (except self-care) and physical health status (r > .56). MCS demonstrated moderate convergent validity on EQ-5D and mental health items (r > .38). PCS distinguish between groups with different physical and work limitations. Similarly, MCS distinguished between groups with and without cognitive limitations. The MCS and PCS showed perfect dose response when variations in scores were examined by participant’s chronic condition status. Conclusions  Both component scores showed adequate reliability and validity with the 2003–2004 MEPS and should be suitable for use in a variety of proposes within this database.     

The Use of Generic Patient-Reported Outcome Measures in Emergency Department Surveys: Discriminant Validity Evidence for the Veterans RAND 12-Item Health Survey and the EQ-5D

ObjectivesThis study aimed to compare discriminant validity evidence of 2 generic patient-reported outcome measures (PROMs), the Veterans RAND 12-Item Health Survey (VR-12) and level 5 of EQ-5D (EQ-5D-5L), for use in emergency departments (EDs).MethodsData were obtained via a cross-sectional survey of 5876 patients in British Columbia (Canada) who completed a questionnaire after visiting an ED in 2018. We compared the extent to which the VR-12 and the EQ-5D-5L distinguished among groups of ED patients with different levels of comorbidity burden and self-reported physical and mental or emotional health status. Multivariable logistic regression was used to evaluate the ability of the 2 PROMs to identify patients presenting with a mental health (MH) condition.ResultsAll the measures produced small effect sizes (ESs) for discriminating comorbidity levels (R² range: 0.00 [VR-12 mental component summary {MCS}] to 0.10 [VR-12 physical component summary score]). The EQ-5D visual analog scale offered the largest ES for discriminating self-reported physical health (R² = 0.48), whereas the MCS, the VR-12 MH domain, and the EQ-5D-5L anxiety/depression dimension had the largest ESs for discriminating self-reported mental or emotional health (R² = 0.42, 0.40, and 0.38, respectively). The MCS produced a medium ES (R² = 0.42) along with the VR-12 utility score (R² = 0.27) compared with the EQ-5D-5L index (R² = 0.19). Having a MH condition was predominantly identified by the MCS (Pratt index = 0.56).ConclusionsThe VR-12 PROM provides a more comprehensive measurement of MH than the EQ-5D-5L, which is important to inform healthcare service needs for patients who present in EDs with MH challenges.     

The SF-36 summary scales were valid, reliable, and equivalent in a Chinese population

OBJECTIVES: To find out whether the SF-36 physical and mental health summary (PCS and MCS) scales are valid and equivalent in the Chinese population in Hong Kong (HK). STUDY DESIGN AND SETTING: The SF-36 data of a cross-sectional study on 2,410 Chinese adults randomly selected from the general population in HK were analyzed. RESULTS: The hypothesized two-factor structure of the physical and mental health summary scales (PCS and MCS) was replicated and the expected differences in scores between known morbidity groups were shown. The internal reliability coefficients of the PCS and MCS scales ranged from 0.85 to 0.87. The effect size differences between the U.S. standard and HK-specific PCS and MCS scores were mostly <0.5. The effect size differences in the standard PCS and MCS scores of specific groups between the U.S. and H.K. populations were all <0.5. CONCLUSION: The PCS and MCS scales were applicable to the Chinese population in HK. The high level of measurement equivalence of the scales between the U.S. and H.K. populations suggests that data pooling between the two populations could be possible. To our knowledge, this is the first study to show that the SF-36 summary scales are valid and equivalent in an Asian population.     

Predictors of health-related quality of life and association with survival may identify colorectal cancer patients at high risk of poor prognosis

Purpose

To investigate the mediators of health-related quality of life (HR-QoL) in colorectal cancer (CRC) patients and effect on overall survival.

Methods

We analyzed baseline (within 1 year of diagnosis) SF-12v1 questionnaire data from 3734 CRC patients and assessed the differences in mental composite scores (MCS) and physical composite scores (PCS) by socio-demographics and risks of poor HR-QoL by these factors. Hazard ratios were generated using univariate Cox regression for MCS and PCS dichotomized using the normalized scoring-based mean of 50 and survival estimates generated using the Kaplan–Meier method.

Results

Differences in MCS and PCS were identified by sex, age, education level, alcohol use, tobacco use, and stage. Race, marital status, and cancer site differed only by PCS. Being female, never married, former alcohol user, or with stage IV disease significantly increased risk of a poor HR-QoL, with magnitudes of risk from 1.25- to 1.97-fold. Higher education level had a protective effect (MCS: P _trend = 2.32 × 10^?7; PCS: P _trend = 5.62 × 10^?14). Hispanics and African-Americans had a 1.35- and 1.57-fold risk of poor PCS, and increase in age had a protective effect for risk of poor MCS (P _trend = 1.84 × 10^?7). Poor MCS or PCS were associated with poor prognosis and decreased survival at 5 years (HR_MCS 1.57, 95 % CI 1.41–1.76 and HR_PCS 2.38, 95 % CI 2.08–2.72), and both remained significant when adjusting for age, gender, race, education level, tumor stage, and tumor site.

Conclusions

Our findings identify potential mediators for HR-QoL and suggest that baseline HR-QoL assessment may be prognostic for CRC.

     

Normative data of the SF-12 health survey with validation using postmyocardial infarction patients in the Dutch population

Objective  To gain Dutch population norms for the Short Form-12 (SF-12), a generic health status questionnaire, in a random sample of the general population and to validate these in postmyocardial infarction (MI) patients. Methods  2,301 respondents from the general population and 459 post-MI patients completed the Short Form-36 (SF-36), which was used to calculate SF-12 scores. Results  The SF-12 summary scores correlated highly with SF-36 summary scores, demonstrating that these scores explain the same amount of variance in health status. Significant sex differences (P < .001) existed for both the physical component summary (PCS) and the mental component summary (MCS). Multivariate analysis of variance showed a main effect of age in oblique (PCS-12: P < .001; MCS-12: P < .001) and orthogonally rotated PCS scores (PCS-12_uc: P < .001; MCS-12_uc: P = .07). As expected, post-MI patients reported statistically significant and clinically relevant poorer mental (P < .001) and physical functioning (P < .001). Differences were less pronounced for MCS and PCS derived from orthogonal rotation data. When controlling for covariates, MI did not significantly affect PCS-12_uc anymore in orthogonally rotated data, while PCS-12_uc was affected by fewer covariates compared with PCS-12. Conclusions  This study presents Dutch population norms for the SF-12 in a large random population sample obtained from both oblique and orthogonal PCA rotation methods, revealing systematic differences between the results based on these two methods. Furthermore, this study demonstrates the discriminative validity of the SF-12 by showing that post-MI patients differ significantly from the normative population on PCS-12 scores.     

Multidimensional item response theory models yielded good fit and reliable scores for the Short Form-12 questionnaire

ObjectivesTo propose a multidimensional item response theory (MIRT) scoring system for the Short Form 12 (SF-12) with good psychometric properties in terms of fit and reliability.Study Design and SettingsTwo models, indicating physical (PCS) and mental component summary (MCS) dimensions, were fitted to SF-12 data from the European Study of the Epidemiology of Mental Disorders, a representative sample from European adult general population (n = 21,425; response rate = 61.2%). Goodness of fit, information, reliability, and agreement of individual scores were compared with the classical SF-12 and RAND-12 algorithms.ResultsThe bidimensional response process (BRP) model, where all items are indicators of both dimensions, yielded the best fit (root mean square error of approximation = 0.057, comparative fit index = 0.95, and Tucker–Lewis index = 0.94), and highly agreed with PCS and MCS scores from the SF-12 (intraclass correlation coefficients of 0.92 and 0.88, respectively) and RAND-12 (0.88 and 0.95). Regarding reliability, the BRP yielded 0.75 and 0.77 (PCS and MCS, respectively), greater than SF-12 (0.65 and 0.66) and RAND-12 (0.65 and 0.67). As indicated by scale linking, MIRT scores can be interpreted similarly to the classical scores.ConclusionThe MIRT models showed a clear construct structure for the PCS and MCS dimensions, defined by functional and role limitation content. Results support the use of SF-12 MIRT-based scores as a valid and reliable option to assess health status.     

Single-change items did not measure change in quality of life

OBJECTIVES: The aim of this study was to demonstrate validity for the two items in the Veterans RAND 36 Item Health Survey (VR-36) that attempt to measure the change in health-related quality of life (HRQOL) over 1 year, using serial administrations of the full instrument as a gold standard. STUDY DESIGN AND SETTING: A total of 1,117 subjects in the Veterans Health Study completed the VR-36 instrument at study inception and 1 year later. Using nonparametric correlation and factor analyses, we compared the single-change items (SCIs) with 1-year changes in the physical component score (PCS) and mental component score (MCS), the summary scores of the VR-36. RESULTS: We found low correlations between the SCI and longitudinal changes in the PCS (0.21) and MCS (0.18) and moderate correlations between the SCI and the current PCS (0.50) and MCS (0.41). Factor analyses confirmed that the SCI loaded highly with current HRQOL. CONCLUSIONS: The two SCIs contained in the VR-36, which are intended to measure changes in health over the past year, are more highly correlated with current self-reported HRQOL. In this instrument, single-item measurements of health status change cannot be substituted for changes in serial measures of HRQOL.     

A preference-based measure of health: the VR-6D derived from the veterans RAND 12-Item Health Survey

Purpose

The Veterans RAND 12-Item Health Survey (VR-12) is currently the major endpoint used in the Medicare managed care outcomes measure in the Healthcare Effectiveness Data and Information Set (HEDIS^®), referred to as the Health Outcomes Survey (HOS). The purpose of this study is to adapt the Brazier SF-6D utility measure to the VR-12 to generate a single utility index.

Methods

We used the HOS cohorts 2 and 3 for SF-36 data and 9 for VR-12 data. We calculated SF-6D scores from the SF-36 using the algorithms developed by Brazier and colleagues. The values of the Brazier SF-6D were used to estimate utility scores from the VR-12 using a mapping approach based on a 2-stage mapping procedure, named as VR-6D.

Results

The VR-6D derived from the VR-12 has similar distributional properties as the SF-6D. The change in VR-6D showed significant variations across disease groups with different levels of morbidity and mortality.

Conclusions

This study produced a utility measure for the VR-12 that is comparable to the SF-6D and responsive to change. The VR-6D can be used in evaluations of health care plans and cost-effectiveness analysis to compare the health gains that health care interventions can achieve.

     

Assessment of Reliability and Validity of SF-12v2 among a Diabetic Population

Objectives

To validate the Medical Outcomes Study Short Form version 2 (SF-12v2) in diabetic patients.

Rethinking and rescoring the SF-12

SummaryObjectives To derive and assess the validity of an Australian version of the SF-12 quality-of-life questionnaire.Methods Using regression methods and structural equation modelling to obtain item weights, an Australian version of the SF-12 was derived from Australian population survey data and compared to the existing United States (US) SF-12 variable set.Results The Australian version of the SF-12 explained 94% of the variation for physical components summary (PCS) and the mental components summary (MCS) of the SF-36 questionnaire. There was high level of agreement on the MCS and PCS summary scores between both versions of the SF-12 and the SF-36.Conclusions Although it is possible to derive a valid Australian version of the SF-12 it is concluded the US version of the SF-12 be used for reasons of international comparability, but using item weights derived from structural equation modelling.

---

ZusammenfassungFragestellung Eine australische Version des SF-12-Fragebogens zur Erfassung der Lebensqualität ableiten und dessen Validität beurteilen.Methoden Unter Verwendung von Regressions- und Strukturgleichungsmodellen zur Bestimmung von Gewichtungsfaktoren wurde eine australische Version des SF-12-Fragebogens abgeleitet. Hierzu wurden Daten einer Befragung der aus-tralischen Bevölkerung herangezogen. Der australische Fragebogen wurde anschliessend mit dem bereits bekannten U.S. SF-12-Variablensatz verglichen.Ergebnisse Die australische Version des SF-12 erklärte 94% der Variation für die physische (PCS) und psychische Summenskala (MCS) des SF-36-Fragebogens. Zwischen den beiden Versionen des SF-36 und SF-12 stimmtem die MCS- und PCS-Summenwerte sehr gut überein.Schlussfolgerungen Es ist möglich, eine valide australische Version des SF-12 zu erhalten. Aus Gründen der internationalen Vergleichbarkeit ist es jedoch besser, die U.S. Version des SF-12 einzusetzen, aber unter Verwendung von Gewichtungsfaktoren, die anhand struktureller Vergleichsmodelle abgeleitet wurden.

---

RésuméObjectifs Etablir une version australienne du questionnaire SF-12 sur la qualité de vie et évaluer sa validité.Méthode A partir de méthodes de régression et de pondération par des modèles d'équations structurelles, une version australienne du SF-12 a été établie à partir d'une enquête de population et comparée avec l'ensemble des variables composant la version nord-américaine existante du SF-12.Résultats La version australienne du SF-12 expliquait 94% de la variation du score synthétique physique (PCS) et mental (MCS) du questionnaire SF-36. Il y avait un haut niveau de concordance pour ces deux scores entre les deux versions du SF-12 et du SF-36.Conclusions Bien qu'il soit possible d'établir une version australienne valide du SF-12, l'utilisation de la version américaine du SF-12 peut être utilisée pour des raisons de comparabilité internationale, mais en utilisant un système de pondération spéifique basé sur des modèles d'équation structurelles.

     

Predicting an SF-6D Preference-Based Score Using MCS and PCS Scores from the SF-12 or SF-36

Background:  The SF-6D preference-based scoring system was developed several years after the SF-12 and SF-36 instruments. A method to predict SF-6D scores from information in previous reports would facilitate backwards comparisons and the use of these reports in cost-effectiveness analyses.
Methods:  This report uses data from the 2001–2003 Medical Expenditures Panel Survey (MEPS), the Beaver Dam Health Outcomes Survey, and the National Health Measurement Study. SF-6D scores were modeled using age, sex, mental component summary (MCS) score, and physical component summary (PCS) score from the 2002 MEPS. The resulting SF-6D prediction equation was tested with the other datasets for groups of different sizes and groups stratified by age, MCS score, PCS score, sum of MCS and PCS scores, and SF-6D score.
Results:  The equation can be used to predict an average SF-6D score using average age, proportion female, average MCS score, and average PCS score. Mean differences between actual and predicted average SF-6D scores in out-of-sample tests was −0.001 (SF-12 version 1), −0.013 (SF-12 version 2), −0.007 (SF-36 version 1), and −0.010 (SF-36 version 2). Ninety-five percent credible intervals around these point estimates range from ±0.045 for groups with 10 subjects to ±0.008 for groups with more than 300 subjects. These results were consistent for a wide range of ages, MCS scores, PCS scores, sum of MCS and PCS scores, and SF-6D scores. SF-6D scores from the SF-36 and SF-12 from the same data set were found to be substantially different.
Conclusions:  Simple equation predicts an average SF-6D preference-based score from widely published information.     

Prenatal Depression in Latinas in the U.S. and Mexico

Objectives The study aimed to investigate the prevalence of depressive symptoms and their associated risk factors during pregnancy in Latinas in the United States (U.S.) and Mexico. Method The sample included 108 women in the U.S. whose data were obtained from medical chart reviews in a community clinic in Washington, D.C., and 117 women in Mexico who participated in face-to-face interviews in the waiting rooms of primary care community centers in Mexico City. Variables, chosen to match in both countries for comparisons, were: socio-demographics, pregnancy gestation and order, social support, depressive symptoms, personal history of depression, family psychiatric history, and suicidal thoughts. Depressive symptoms were measured using the Center for Epidemiological Studies Depression Scale (CES-D). Results The prevalence of depressive symptoms was 32.4% for pregnant Latinas and 36.8% for Mexicans (CES-D ≥ 16), and 15.7% and 23.9% (CES-D ≥ 24), respectively, with no differences between groups. Separate multiple logistic regression analyses showed that for U.S. Latinas: (1) being more educated predicted depressive symptoms (CES-D ≥ 16), and (2) second trimester, as compared to first, also predicted symptoms (CES-D ≥ 24). (3) History of suicidal thoughts predicted symptoms in Latinas in the U.S. (CES-D ≥ 24) and in Mexico (using both definitions of high symptoms), and (4) living with a partner but not formally married and multi-parous condition predicted symptoms (CES-D ≥ 24) among pregnant Mexicans. Conclusions A high prevalence of depressive symptoms and significant risk factors during pregnancy were found in Latinas in U.S. and Mexico, suggesting increased risk for postpartum major depression. Implications for screening and interventions are discussed.

Confirmatory factor analysis of the validity of the SF-12 for persons with and without a history of stroke

Purpose  

To assess the validity of the Physical and Mental Component Summary scores (PCS and MCS) of the 12-item Short-Form Health Survey (SF-12), a measure of health-related quality of life (HRQoL), among persons with a history of stroke.     

Why U.S. health care expenditure and ranking on health care indicators are so different from Canada’s

Compared to other industrialized countries, the U.S. spends most of all on health care. Nonetheless, the U.S. ranks relatively low on health care indicators. This paradox has been already known for decades. For example, the turning point comparing the U.S. and Canada was in 1972. Health expenditure as a percentage of GDP was higher in Canada than in the USA from 1960 until 1972. Since 1972 expenditure on health care has been higher in the U.S. than in Canada (OECD 2005a, Health data 2005, fourteenth OECD electronic database on health systems, date of release June 2005, last update 04/26/2005). The present study integrates the dispersed literature on spending and health care rankings and adds some statistical analysis to these studies. The evaluation of different factors influencing health care expenditure in the U.S. relative to other countries is restricted to a comparison with Canada. The U.S. and Canada are two countries that are sufficiently similar to make comparisons useful. The comparison of factors influencing health care expenditure in the U.S. and Canada in 2002 reveals that health care expenditure in the U.S. is higher than in Canada mainly due to administration costs, Baumol’s cost disease and pharmaceutical prices. It is not primarily inefficiency in health care production but the dominant prevalence for free choice and own responsibility that explains the paradox of high expenditure on health care and low ranking on health care indicators.

Interpreting SF&-36 summary health measures: A response

In response to questions raised about the “accuracy” of SF-36 physical (PCS) and mental (MCS) component summary scores, particularly extremely high and low scores, we briefly comment on: how they were developed, how they are scored, the factor content of the eight SF-36 subscales, cross-tabulations between item-level responses and extreme summary scores, and published and new tests of their empirical validity. Published cross-tabulations between SF-36 items and PCS and MCS scores, reanalyses of public datasets (N = 5919), and preliminary results from the Medicare Health Outcomes Survey (HOS) (N = 172,314) yielded little or no evidence in support of Taft's hypothesis that extreme scores are an invalid artifact of some negative scoring weights. For example, in the HOS, those (N = 432) with “unexpected” PCS scores worse than 20 (which, according to Taft, indicate better mental health rather than worse physical health) were about 25% more likely to die within two years, in comparison with those scoring in the next highest (21– 30)␣category. In this test and in all other empirical tests, results of predictions supported the validity of extreme PCS and MCS scores. We recommend against the interpretation of average differences smaller than one point in studies that seek to detect “false” measurement and we again repeat our 7-year-old recommendation that results based on summary measures should be thoroughly compared with the SF-36 profile before drawing conclusions. To facilitate such comparisons, scoring utilities and user-friendly graphs for SF-36 profiles and physical and mental summary scores (both orthogonal and oblique scoring algorithms) have been made available on the Internet at www.sf-36.com/test. This revised version was published online in June 2006 with corrections to the Cover Date.     

Results from several population studies show that recommended scoring methods of the SF-36 and the SF-12 may lead to incorrect conclusions and subsequent health decisions

Purpose

To compare the measurement properties of the physical component summary (PCS) and mental component summary (MCS) scores of the SF-36 and SF-12 based on the traditional orthogonal scoring algorithms with the performance of the PCS and MCS scored based on structural equation model coefficients from a correlated model.

Methods

This study used three large-scale representative population studies to compare the measurement properties of the PCS and MCS scores of the SF-36 and SF-12 with the performance of the PCS and MCS scores based on structural equation models producing coefficients from a correlated model. We assessed the relationships of these scores with selected important mental health measures and chronic conditions from three representative Australian population studies that address clinical conditions of high prevalence and health service importance.

Results

Structural equation model scoring methods produced summary scores with higher correlations than the recommended orthogonal methods across a range of disease and health conditions. The problem experienced in using the orthogonal methods is that negative scoring coefficients are applied to negative z-scores for sub-scales, inflating the resulting summary scores. Effect sizes over a half of a standard deviation were common.

Conclusions

If health policy or investment decisions are made based on the results of studies employing the recommended orthogonal scoring methods then the expected outcome of such decisions or investments may not be achieved.     

A new comorbidity index: the health-related quality of life comorbidity index

Objective

To derive and validate the health-related quality of life comorbidity index (HRQL-CI).

Study Design and Setting

Of 261 clinical classification codes (CCCs) in the 2003 Medical Expenditure Panel Survey (MEPS), 44 were identified as adult, gender-neutral, chronic conditions. The least absolute shrinkage and selection operator (LASSO) procedure identified CCCs significantly associated with the Short Form-12 physical component summary (PCS) and mental component summary (MCS) scores. Regression models were fitted with the selected CCCs, resulting in two subsets corresponding to PCS and MCS, collectively called the HRQL-CI. Internal validation was assessed using 10-fold cross-validation, whereas external validation in terms of prediction accuracy was assessed in the 2005 MEPS database. Prediction errors and model R² were compared between HRQL-CI models and models using the Charlson-CI.

Results

LASSO identified 20 CCCs significantly associated with PCS and 15 with MCS. The R² for the models, including the HRQL-CI (0.28 for PCS and 0.16 for MCS) were greater than those using the Charlson-CI (0.13 for PCS and 0.01 for MCS). The same pattern of higher R² for models using the HRQL-CI was observed in the validation tests.

Conclusion

The HRQL-CI is a valid risk adjustment index, outperforming the Charlson-CI. Further work is needed to test its performance in other patient populations and measures of HRQL.     

Correlation of Risk Adjustment Measures Based on Diagnoses and Patient Self-Reported Health Status

Case-mix adjustments have traditionally used diagnosis-based models such as Diagnostic Cost Groups (DCGs). The recent development and availability of reliable and valid patient self-reported health status measures such as the Veterans SF-36 (Short Form Health Survey) may be useful in complementing existing diagnostic information in describing patients' health status for purposes of risk adjustment. However, the correlation between these two approaches has not been explored. We collected SF-36 data from 31,419 veterans nationwide based on a national probability sample of veterans receiving ambulatory care to assess the physical (PCS) and mental (MCS) component of patient self-reported health status. In addition, we used inpatient and outpatient diagnoses from one year (1/1/97 to 1/1/98) to calculate DCG relative risk scores, with the 1991 Medicare beneficiary population as the benchmark. We found that higher DCG relative risk scores were associated with worse PCS (r=–0.223, p<0.05) and MCS (r=–0.174, p<0.05) scores. Further examination of the distribution of MCS categories (MCS40) across the five psychiatric hierarchical condition categories (HCCs) in the DCG/HCC model showed a small association between MCS category and disease severity level. These results suggest that risk adjustment approaches based on patient self-reported health status and diagnoses convey different case-mix information, specifically for patients with psychiatric conditions. These two approaches can be used as the basis for the development of a more comprehensive risk adjustment model which incorporates both the providers' and the patients' perspectives in predicting resource utilization.     

Predictors of Secondary Abstinence in U.S. College Undergraduates

This study examined (1) the percentage of participants who practiced secondary sexual abstinence and (2) factors associated with its practice among a sample of U.S. college students. College undergraduate men and women (n = 1,133) in Texas completed a web-based survey assessing abstinence status and predictors of abstinent behavior. Results revealed that 12.5% of participants practiced secondary abstinence. Of eight variables, five significantly predicted secondary abstinence (following sexual initiation). Predictors were positive attitude toward abstinence, subjective norm supporting abstinence, greater religious ties, and previous negative sexual experiences. The fifth variable, participation in abstinence education, however, was associated with a significantly reduced likelihood of secondary abstinence. Fewer perceived barriers, less environmental manipulation (efforts to make physical and social environments supportive of abstinence), and greater religious ties significantly predicted self-efficacy for secondary abstinence. Findings provide an estimate of the percentage of participants who practiced secondary abstinence and suggest focal points for future research.