PREDICTING UTILIZATION OF HOME HEALTH RESOURCES - IMPORTANT DATA FROM ROUTINELY COLLECTED INFORMATION

This study examined the feasibility of using routinely collected information on patients enrolled in home health care to predict their subsequent use of services. Data were gathered from 1,984 episodes of care randomly sampled from home health care agencies of the Virginia Health Department. Age, sex, Medicare and Medicaid enrollment, referral source, medical diagnosis, and prognosis were used to predict the total number of visits, the duration of enrollment, and the intensity of service. Since the data were originally gath-ered to study the effects of the implementation of diagnosis-related groups (DRGs) on home health services, half of the patients were enrolled before and half after the implementation of DRGs. Using multiple linear regression analy-sis, significant amounts of variance in each measure of home health care utili-zation were explained by the predictor variables (R2 = 0.04 to 0.10). For exam-ple, after controlling for other predictor variables, age 75 years or older pre-dicted longer durations of enrollment and lower intensities of service as compared with other age groups (P < 0.05), and four of 14 diagnosis categories predicted at least one measure of utilization (P < 0.05). Medicaid enrollment predicted longer durations of enrollment and lower intensities of service in home health care (P < 0.05) in the post-DRG but not the pre-DRG period. These results demonstrate the value of routinely collected information in predicting the use of home health services. To develop more accurate estimates of needs for home health services for particular groups of patients, additional informa-tion on chronic functional impairments, informal caregiving, and the chronicity of needs may be useful. © Lippincott-Raven Publishers.