Reducing Nursing Home Use Through Community Long-Term Care: An Optimization Analysis Using Data from the National Channeling Demonstration. The Data Set


Data for the analysis are from the public use data files for the NLTCCD. The overall NLTCCD study sample was a targeted group from 10 sites representing considerable geographic and socioeconomic diversity (the sites being Baltimore; Houston; Cleveland; Miami; Philadelphia; an eight county region in Eastern Kentucky; a two county region in southern Maine; Middlesex County, New Jersey; Rensselaer County, New York; and Greater Lynn, Massachusetts). The NLTCCD enrolled persons who were relatively old and frail. Persons who applied for its programs were screened to be of age 65 or older, substantially impaired in functional capacity and to have some degree of unmet need in important functional areas. Those enrolled while in nursing home care were further required to be deemed a good prospect for discharge within 90 days. Details of eligibility criteria, and a comparison of sample characteristics with those from a simulated national probability sample, can be found in Applebaum (1988). Because the observed sample is self-selected by application to the NLTCCD, as well as screened on the targeting criteria, the reference population for the data analysis to follow should be confined to the universe of eligible persons who would apply and qualify for such programs. Insofar as factors that influence the likelihood of application are correlated with factors unmeasured in our statistical models, inferences to the otherwise eligible population which did not apply will to an unknown extent be subject to sample selection bias.

The data sets contain a 12-month nursing home use history for 4,593 individuals (see Wooldridge and Schore, 1986, for a complete description of this data, which is a subsample of the 6,326 individuals for whom survey data were obtained). From this group we eliminated all individuals for whom there was missing data for any variable in our models, or who did not survive long enough after baseline to contribute at least one person-month of nursing home use history, which reduced the effective sample size to 3,446. The descriptive statistics for this subgroup (see Table 1) are virtually identical to those for the full NLTCCD data set (Wooldridge and Schore, 1986; Applebaum, 1988), suggesting that the subsample is accurately representative of the entire study sample. This subsample was used to statistically estimate the transition logits. For the optimization analysis, the analysis sample was reduced further to 2,406 individuals because of missing data on variables used in the tobit models in estimating the proportion of time in the community not spent in hospital. Again, descriptive statistics for this subsample differ very little from those for the full sample, indicating that the subsample remains representative.

View full report


"rednh.pdf" (pdf, 1.92Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®