Joel P. Winter

Predicting different types of service use by the elderly : The strength of the behavioral model and the value of interaction terms

This study compared the ability of logistic regression models (with and without interaction terms) to predict the following four measures of service use: in-home, financial, community, and any service use. A representative sample of more than 4,000 older adults provided data for the study via telephone interviews. Policy and practice implications of the study results were (a) the behavioral model predicted any service use and in-home service use moderately well; need variables explained about half of the variance, with predisposing and enabling variables explaining the remainder; (b) the behavioral model did not explain much of the variance of the use of financial assistance or community services; and (c) although most of the interaction hypotheses in this study were not supported, interactions involving activities of daily living or instrumental activities of daily living needs with living situation and income helped explain which older adults were most likely to use services.

Who Attends Senior Centers

Abstract Logistic regression analysis was used to compare users and non-users of senior centers. Variables assessing linkage to the service system were more significant predictors of senior center utilization than predisposing, enabling, or need variables. More specifically, users of senior centers were older and more likely to live in rural areas. They also had more social contacts, better mental health, and fewer ADL problems. Senior center users were also more aware of specific service agencies, more likely to consult formal resources in making service decisions, and more likely to have used other services.

Understanding and Controlling Response Bias in Needs Assessment Studies

This study demonstrated that estimates of agency awareness in the typical needs assessment study are probably inflated by a response bias labeled “agency awareness overclaiming.” Overclaimers (respondents who reported being aware of fictitious agencies) reported being aware of more real agencies than other respondents. Estimates of agency awareness may also be biased, because certain segments of the population were more likely to exhibit agency awareness overclaiming. Age was positively correlated with overclaiming, and African Americans were more likely to exhibit agency awareness overclaiming than Caucasians. General overclaiming was correlated with agency awareness overclaiming, but social desirability and acquiescence were not.

Do consumers who have a choice of treatment have better outcomes

This study used a non-equivalent control group design to investigate the effect of consumer choice of treatment on both process and outcome variables. All study participants suffered from severe mental illness, were homeless at baseline, and were enrolled in a modified Assertive Community Treatment (ACT) program. Consumers in the choice condition had selected the ACT program from a menu of five treatment programs; clients in the no-choice condition were simply assigned to the ACT program by an intake worker. Results found that consumers in the choice condition visited the ACT staff at their offices more than consumers in the no-choice condition, but there were no significant differences between groups on the other treatment process variables. Although consumers in the choice condition increased their income more than consumers in the no-choice condition, there were no significant differences between groups on the other outcome variables (stable housing, psychotic symptoms, depression, and substance abuse).

The Dartmouth Assertive Community Treatment Scale (DACTS) A Generalizability Study

Generalizability theory was used to assess the reliability of the Dartmouth Assertive Community Treatment Scale (DACTS), which was developed to assess treatment reliability to assertive community treatment (ACT). Program staff and local evaluators who were participating in a national demonstration program to serve homeless mentally ill clients provided data. The total scale score for the DACTS demonstrated acceptable internal consistency and interrater reliability. Although the stability of the total DACTS score was quite low, many items on the DACTS were quite stable. The generalizability analyses provide additional detail on the effects of site, occasion, and site by occasion interactions on the reliability of the DACTS.

Predicting four types of service needs in older adults

Abstract Logistic regression analysis was used to predict four service need variables. A sample of nearly 5000 older Missourians were assessed on a comprehensive set of variables, representing all of the categories of the behavioral model. Variables in the behavioral model predicted perceived need for frail elderly services better than they predicted unmet need for frail elderly services, perceived need for community services, and unmet need for community services. Health need variables were better predictors of all of the service need variables than predisposing or enabling variables. Although the inclusion of interaction terms in the prediction models did not increase model fit, some of the interaction terms were significant and helped to clarify the relationship between certain predictor variables and the four service need variables.

Evaluation of Treatment Programs for Persons with Severe Mental Illness Moderator and Mediator Effects

This study evaluated several statistical models for estimating treatment effects in a randomized, longitudinal experiment comparing assertive community treatment (ACT) versus brokered case management (BCM). In addition, mediator and moderator analyses were conducted. The ACT clients had improved outcomes in terms of housing and psychiatric symptoms thanBCM clients. Case management housing assistance and financial assistance partially mediated housing outcomes. No reliable mediators were found for psychiatric symptoms, and no reliable moderators were found for either housing or psychiatric symptoms. The study also made several important methodological advances in the analysis of longitudinal data in randomized experiments.

Predicting Specific Service Awareness Dimensions

Nearly 5,000 older adults provided data for this study, which used factor analysis to identify the following four dimensions of agency awareness: in-home services, aging network services, financial services, and other services. Then, 18 variables assessing the various dimensions of the behavioral model were included in multiple regression equations predicting each of the awareness dimensions. The regression equations only accounted for 7% to 13% of the variance of the various service awareness dimensions.

Reducing Overclaiming in Needs Assessment Studies an Experimental Comparison

A randomized experiment investigated the effect of various instructional sets on reducing agency awareness overclaiming, that is, claiming knowledge of fictitious agencies. As predicted, respondents who were warned that the list contained fake agencies exhibited less agency awareness overclaiming than respondents who were not warned. However, providing respondents a memory retrieval strategy had no effect on agency awareness overclaiming. A multivariate model, which included demographic variables, response style variables, and knowledge variables, explained 40% of the variance of agency awareness overclaiming.