Targeting High Utilisers: Predictive Validity of a Screening Questionnaire
Julie A. Meek
Brenda L. Lyon
Frederick E. May
Wendy D. Lynch
Objective: To examine the utility of a self-reported health perception assessment as a screening tool to predict high near-term utilisation of healthcare services. Design and setting: Prospective cohort study in a Midwest US commercial managed-care population. Participants completed a 70-question/126-response item paper-based health perception assessment (including demographic items) in late August 1997. Healthcare claims data were subsequently obtained from the health plan for the next 6 months and converted to total number of encounters and total dollars for each respondent. The dependent variable was the total number of encounters re-coded to a dichotomous variable with the cut-off set at 6 or more encounters as a subsequent high care user. All health perception assessment variables were dichotomised as well and then evaluated as independent variables for their ability to predict the probability that a member would become a high care user over the next 6 months. A split-half technique was used to identify the predictive model from the first half of the sample using logistic regression analysis. A formula was subsequently developed from that defined logistic model and then tested on the first split-half for levels of sensitivity and specificity. The chosen predictive formula was then tested using data from the other half of the sample. Study participants: A sample of 4210 non-institutionalised enrollees of the health plan, ranging in age from 18 to 65 years, who responded to an initial health perception assessment and were continuously enrolled in the health plan for the next 6 months. Main outcome measures and results: Using logistic regression for the first split-half of the sample, the resulting predictive model included 39 health perception assessment variables, correctly predicting 68.1% of the high care users and 61.9% of the low care users. The final logistic model was converted to a formula resulting in a probability score for each member, which indicates the likelihood the person will become a high utiliser in the near term. This formula was tested on both split-halves of the population yielding 66.7% sensitivity and 63.4% specificity on the first split-half and 59.4% sensitivity and 53.3% specificity on the second split-half. The predictive model permitted the number of health perception assessment survey questions to be lowered to 48 with 74 responses. Conclusions: Easily ascertained self-reported factors predict an adult's probability of becoming a near term high care user. Utilising a powerful self-report survey overcomes many of the limitations of using less predictive traditional health risk/status models or cumbersome claims stratification methods.
In order to set up a list of libraries that you have access to,
you must first login
or sign up.
Then set up a personal list of libraries from your profile page by
clicking on your user name at the top right of any screen.