Although we have data at monthly intervals for nursing home transitions, data for community services in the NLTCCD were collected only at 6 month intervals (the survey points). For individuals in the community, we assume that the services they were found to be receiving at the 6 month survey are representative of those received in months 1-6 while those they reported receiving at the 12 month survey are imputed to months 7-12. Service variations within these intervals are thus not captured.
Individuals in a nursing home at a point in time are, of course, not actually consuming community services at that time. What is at issue in predicting their transition is what services are available to them in the community at that time. We do not have direct measures of such "potential" services. For individuals who exited a nursing home during the study period, we take the services they were found to be receiving at the next survey point after exit to be representative of what was available to them during the preceding period of institutionalization. A better measure would have been services received at the time of exit, but again we face the problem that services were measured only at regular 6 month intervals. On average, we are measuring services for these individuals about 2.5 months after discharge, and so are probably tending to underestimate services provided at time of discharge. On the other hand, insofar as discharge planning reflects not just services at point of discharge, but also a longer run consideration of service availability, our delayed data are capturing valid information. At a minimum, these data are a reasonable proxy measure.
For individuals in the sample who began the study period in a nursing home and did not exit alive, we have of course no data on community service use during the study period (these are about 25 percent of those initially in a nursing home--some 2 percent of the sample overall). For these persons, their baseline report of service use at the time of most recent admission was used as a proxy for services available. Individuals who had been in a nursing home more than two months at baseline and for whom recall error was likely to be excessive, were excluded.
Categorization of services used by provider type was done as follows. Respondents were asked to describe the services actually received on a given provider visit. Using uniform definitions developed from an extensive analysis of service definitions from a variety of official sources (Corson et al., 1986, App. B), researchers from Mathematica Policy Research (the principal contractor for the NLTCCD evaluation) coded each encounter as involving services from nursing, therapy, home health aide, personal care aide and housekeeper services. In effect, the NLTCCD evaluation identified each encounter and type of service provider by the highest level of service rendered in that encounter. Thus, if a function identified as "nursing" occurred during an encounter, the caregiver is taken to be a nurse and the length of the encounter (service hours) is attributed to "nursing hours." It should be noted that this does not assure that lower level services did not take up some portion of the encounter, and hence it seems likely that the measurement scheme will tend to overstate the intensity of higher level services being provided. Thus, a one hour nursing encounter might in principle consist of 30 minutes of attention to matters requiring professional nursing skills and 30 minutes of more routine matters that might have been within the skill range of a home health aide. The data do not permit us to distinguish these cases. Also, we cannot distinguish cases where an insufficiently credentialed provider may have performed higher level functions. While these measures are only proxies for the underlying structural service variables, they are the best longitudinal service use data currently available that are linked to a detailed nursing home use history.
For cost data, site-specific average costs experienced by the 20 percent subsample for whom provider record extracts were obtained were used to estimate the community service and nursing home prices facing individuals at that site (see Carcagno and Brown, 1986). The hourly cost of home nursing over the ten sites varied from $29 to $72 with a mean of $48. Home health aide hourly costs ranged from $7 to $33, with a mean of $16. Costs for personal care ranged from $4.50 to $11.50 per hour, with a mean of $7.20, and costs for housekeeper services ranged from $4.70 to $10 per hour, with a mean of $6.60. Bear in mind that sites varied widely in geographical region and degree of urbanization, with some being principally rural.