Evelyn S. Mathis, National Center for Health Statistics
Genevieve Strahan, National Center for Health Statistics
Esther Hing, National Center for Health Statistics
Edward S. Sekscenski, National Center for Health Statistics
Jennifer Madans, Ph.D., National Center for Health Statistics
William Scanlon, Ph.D., Georgetown University
EVELYN MATHIS: The NNHS is conducted by the NCHS as part of its responsibility for the collection and dissemination of information on the health of the whole United States. The survey provides information on the size and composition of the population in nursing homes, the services they receive and the sources of funding for care. The development of the data set for the 1985 survey was initiated during 1982. The data set was developed from a review of the data items that were used in the 1977 NNHS, the recommended long term care minimum data set, and recommendations made by people working within the federal government in the field of long term care.
First we start off by sending letters to people in both public and private sectors of long term care to inform them of plans to conduct the NNHS and to solicit their recommendations as to the type of information that we should collect. Questionnaires were drafted and distributed to a wide variety of individuals representing organizations in the areas of policy making, research, evaluation and analysis, for review and comments. After several reviews and discussions, a data set to be tested was agreed upon. Five federal agencies entered into interagency agreements with NCHS.
These agencies either added new questionnaires or items to the existing questionnaires, or both. Office of Management and Budget (OMB) clearance was obtained. The pretest consisted of eight different components of questionnaires and was conducted among 150 nursing and related care facilities, in four metropolitan areas, each representing one of the four census regions, namely, the south, west, midwest and northeast.
All of the data collection instruments are color coordinated. Whether you are talking about a sample list or a questionnaire, they were all the same color.
The sponsoring agencies, in addition to the NCHS, were the National Institute of Mental Health (NIMH), the Bureau of Health Professions, the Health Care Financing Administration (HCFA), the Veterans Administration (VA) and the National Institute on Aging (NIA).
Training sessions were held for all aspects of the survey. After the training of the field supervisors and interviewers, the survey was initiated. Visits to the nursing homes took place from August 1985 to January 27, 1986 and the contacts of next-of-kin followed the field work and continued until October 1986.
The NNHS is a stratified two-stage probability sample of nursing and related care homes in the U.S. The first-stage consisted of the selection of facilities by NCHS, and providing them to the contractor. The second-stage consisted of the selection of samples of RN's, current residents and discharges, and was carried out by the field interviewers.
The universe from which the sample was selected was the National Master Facility Inventory (NMFI). The inventory is a comprehensive file of in-patient health facilities. The facilities in the inventory are categorized into three broad types; they are hospitals, nursing and related care homes, and then we have a catch-all category called other. Obviously, the category for the NNHS was the nursing and related care homes.
Two mechanisms are used to keep the NMFI as current as possible. One mechanism is through periodic surveys and the other is through our agency reporting system where state agencies and national organizations send NCHS their most recent directories and lists.
Another activity associated with the NMFI is the complement survey. The compliment survey is conducted periodically to assess the completeness of the NMFI.
The frame for the 1985 NNHS consisted of the results of the 1982 survey of the NMFI, homes that were identified in the 1982 complement survey, homes that opened for business after the 1982 survey, and hospital-based nursing homes identified in the records of HCFA. The resulting frame contained over 20,000 facilities.
These facilities were sorted into two strata, those certified by either Medicare or Medicaid, and those not certified by either Medicare or Medicaid. Facilities in each of these two strata were divided into two groups. One was the 1982 complement survey and then all of the other files. The facilities in the non-complement survey strata were further sorted into bed size groups, producing 20 primary strata.
The nursing homes in the universe were then ordered by ownership, geographic region, metropolitan status, state, county and Zip Code. The sample was then selected systematically after a random start within each primary strata.
The second-stage was done at the nursing home. The universes for the RN's, the residents and the discharges were developed as a part of the data gathering process. If the nursing home had a list that was already prepared, we allowed the interviewers to use the prepared list.
GENEVIEVE STRAHAN: The homes that participated in the 1985 NNHS were selected from a universe of over 20,000 nursing and related care homes. Of the 1,220 facilities selected, six were identified as having been included in the pretest phase of the survey. It was decided by the NCHS not to re-contact the same facilities, but instead to transcribe data from the pretest instrument to the national survey instruments. Of the remaining 1,214 facilities, 57 were identified as either out of scope or out of business. Of the remaining in-scope facilities, 1,079 participated in the survey for a response of 93 percent.
First contact with the facility was made in May 1985 prior to the beginning of the survey. A telephone pre-screening procedure was performed to verify contact information for the facility selected in the sample. This pre-screening was designed to update facility data concerning facility name, address, telephone number and the administrator's name. The next contact made with the sample facility was in the form of an introductory information package to the administrator. The packet contained a letter from the Director of NCHS explaining the importance of the survey and informing the administrator that an interviewer would be calling for an appointment. The packet also included letters of endorsement from professional health organizations.
After the packet should have been received, the interviewer contacted the administrator to set up an appointment and to conduct the survey. Depending on the size of the facility, one interviewer or a team of two or three interviewers visited the facility. A part of the facility visit included the administration of three questionnaires: (1) facility questionnaire, (2) expense questionnaire, and (3) nursing staff questionnaire.
The facility questionnaire, printed in canary yellow, was completed by the interviewer in a face to face interview with the administrator or his designee. Collected on the facility questionnaire was basic information about the facility: ownership, certification status, bed size, number of admissions and discharges, in-patient days of care, services provided to residents and nonresidents, and number of nonresidents served.
Staffing in several occupational categories was collected for full-time and part-time employees. Full-time equivalent employees for each category were tabulated utilizing the number of hours worked by part-time employees. Thirty-five hours of part-time work are taken to equal one full-time.
The 1985 survey collected for the first time per them rates for routine care set by nursing homes. These rates were collected separately for Medicare, Medicaid and private pay patients. Per them rates will be one of the key units of analysis from the facility file. By matching the unique facility ID number from all documents completed in the sample home, information collected in other components of the survey can be described by characteristics of the facility. For example, estimates of current residents can be tabulated by ownership of the facility.
The administrator did not always have all the data required for the facility questionnaire at hand, and needed to consult records and staff in other off ices. Questions that required specific numerical data were printed on a separate sheet, referred to as the facility questionnaire worksheet; it is also yellow in color, a single page printed on both sides. The questions were exact duplicates from the facility questionnaire.
The interviewer gave the work sheet to the administrator at the end of the interview to be completed later. She picked it up at the end of the day, or at some later date if that was required.
In 1985 the typical nursing home was independently and privately owned. It had about 85 beds, most of which had some form of certification. The estimated 19,100 nursing home sets average rates of $61 for skilled private pay daily care and $62 per day for Medicare skilled care. These data and more are included in Advance Data report number 131, nursing home characteristics, and preliminary data from the 1985 NNHS. Data from the facility file, along with data from four other components of the NNHS will be included in a special report to be published by the end of this year.
Upon completion of the facility questionnaire, the expense questionnaire and its accompanying definition booklet printed in green, like money, were presented to the administrator for completion. In many facilities the administrator completed the expense questionnaire. In others, he referred the interviewer to an accountant, a bookkeeper or a central off ice. This instrument was completed by a respondent at his or her convenience, and postage paid return envelopes were provided for the return of the expense questionnaire. The expense questionnaire collected data on two major topics; expenses and revenues. Expense data include payroll, health care services, insurance, taxes, food, utilities, maintenance and drug expenses. Revenue data included sources of income from patients and non-patient sources, such as contributions. In lieu of the completed expenses questionnaire, each facility was offered the option of providing the interviewer with a financial statement, and many homes did that.
After obtaining the financial statement or the name and address of the anticipated respondent for any necessary follow-up, the interviewer introduced the nursing staff component of the NNHS. These two documents, the nursing staff sampling list and the nursing staff questionnaire were used to collect data on RN's working in nursing homes.
The nursing staff sample list, printed in blue, was completed by the interviewer in collaboration with a staff member designated to help her. For the preparation of this list it was necessary to divide employment status of all facility RN's into three categories: those who are employed on the staff of the facility; those scheduled to work who were retained through a special contractual relationship; and those scheduled to work who were attained through a temporary service.
Three columns were provided in which to list separately persons in each of these categories. The sample list provided the universe of RN's separated into three groupings. With the introduction of the nursing staff sampling list came the first need to use sampling tables.
Each interviewer received a packet of sampling tables. The packet consisted of ten independent sets of three different kinds of tables which were numbered and color coded according to the component to which they were applied. Table 1 was blue and was used to select the nursing staff sample. In order to insure random in-facility samples, Table 1 had ten versions numbered 0 through 9. The fourth digit of the facility ID number determined which version of Table 1 was to be used to select the RN sample at that facility. For example, if the facility ID number was 123400-7, the interviewer would consult version 4 of Table 1-4. This method of assignment assured a fairly even distribution of facilities among all versions of the sampling tables.
After finding the total number of RN's recorded on the nursing staff sampling list, the interviewer referred to the version of Table 1 mandated by the fourth digit of the facility ID number and looked at column 1 which is total number listed, and went across. That interviewer could now decide which line numbers to choose for those nurses that would be in the sample. For instance, if there had been ten RN's in a facility that had a fourth digit of 7, then the line numbers would be 1, 4, 6, and 9. The RN's listed on those lines would be included in the survey for that facility.
The selection of up to four RN's from each sample nursing home yielded a sample of 3,439 RN's. The nursing staff questionnaire, also printed in blue, was personally distributed by the interviewer to those RN's selected in the sample. The nursing staff questionnaire was self-administered. When personal delivery was not possible, the questionnaire was either mailed to a home address or left at the facility. A postage paid business reply envelope was provided for return of the completed questionnaire. If a questionnaire was not received within 28 days of the facility visit, a reminder letter and a duplicate instrument were sent. The nursing staff questionnaire gave information on the work experience, hours, activities, education, training, salary, and opinion about recruitment and retention issues of RN's working in nursing homes. Basic demographics about each RN were also collected.
Data were collected for 2,763 of the sample RN's for an 80 percent response rate. The typical RN working in a nursing home was prepared to work as an RN in a diploma program and has been employed as an RN for more than 10 years. She, and I say she because 98 percent are female, worked full-time on a non-rotating day shift. She is white, married with either no children living at home or children of school age, that is, 6-18 years. She is scheduled to work an average of 32.5 hours per week and earns about $334 a week. An Advance Data report will be published this year reporting characteristics of RN's in nursing homes. Future reports will provide detailed information about RN's working in nursing homes, and will be published in both Series 13 and 14 reports.
Data on RN's will also appear in a special report that will also include data from four of the other components.
ESTHER HING: Data from the current resident component of the NNHS are cross-sectional and are representative of the nursing home population in the U.S. as of the night of the survey. To draw a sample of residents, lists of residents in the facility were constructed at the time of the survey. Nearly half of the homes that we surveyed, the lists were provided by the facility and were computer generated lists. In the remaining homes, lists had to be constructed by the interviewers, and in 3 percent of the remaining homes the lists had to be constructed by copying the names of residents from ledgers or other patient lists. A sample of five or fewer residents were selected per sample home, resulting in an overall sample of 5,395 current residents.
The current resident questionnaire, which was Orange, was used to collect data on the sample of current residents. This questionnaire was administered by personal interview with a knowledgeable staff member who referred to the resident's medical record when necessary. The most frequent respondent was a nurse, that was 55 percent of the time, followed by the administrator or owner of the nursing homes. In about 3 percent of the cases, no staff was available and the interviewer had to extract the data from the medical records herself.
Participation for this questionnaire was very high. The response rate was 97 percent. Item response rates for this questionnaire were also generally high as a result of our pretest during which we eliminated most of the items which had low responses. In addition to the data collected from the nursing home staff on this sample of current residents, we also had a telephone follow-up of these current residents in a component called the next-of-kin component. In this component, the resident's next-of-kin, guardians or friends might have been contacted. Only residents who had a next-of-kin or other known contacts were eligible for this telephone follow-up. In general, the types of data items collected in this telephone follow-up were data items that were not available to the nursing home staff.
The current resident questionnaire collected information on the demographic, medical and utilization characteristics of the nursing home population. Demographic variables include age, sex, race, Hispanic origin, and martial status. Medical data include diagnoses at admission and currently. Up to eight diagnoses were listed for each time period. The data were coded according to the Clinical Modification of the Ninth Revision of the International Classification of Diseases. Other medical data collected include vision and hearing status, and prevalence of mental disorders. Utilization data collected include the length of stay since admission and the total monthly charge last month.
Items collected for the first time in the NNHS include marital status at the time of admission, presence of living children, diagnosis related group (DRG) data for hospital transfers, hospital stays while a resident, history of other nursing home stays, instrumental activities of daily living (IADL), disorientation or memory impairment, and sources of payment at the time of admission.
The tape for the current resident questionnaire will include the facility number and the resident weight. The facility number uniquely identifies each facility in the survey. By matching the facility number on the current resident questionnaire with the facility number on the facility tape, information about the facility, such as bed size or ownership, can be analyzed with the current resident data.
The resident weight is used to inflate the sample data to national estimates.
One of the principal strengths of the current resident data is that it provides national estimates of the population in nursing homes. This is useful to help planners and policy-makers who need descriptive data on the utilization of nursing homes. In addition, the 1985 NNHS also includes some items of particular interest to policy-makers. The item on sources of payment at admission and last month, for example, will provide estimates of nursing home residents who had to spend-down, in other words, exhaust their own income sources before they could become eligible for medical assistance from Medicaid.
A question was also added to the current resident questionnaire on the DRG for all persons transferred to the nursing home from short stay hospitals. This data, along with other variables from the survey, may be used to assess the impact of the Medicare Prospective Payment System (PPS) on nursing home care since its implementation in 1983.
Because the residents selected are patients currently in the facility as of last night, the length of stay for the residents is incomplete and underestimates the true length of stay that would be achieved at some point in the future. Residents with long lengths of stay, however, are overrepresented in the current resident sample because of the fact that only persons who are in the facility as of last night were included in the sample. Thus, a person admitted to a nursing home for one day has fewer chances to be included in the sample than a person who had been in the nursing home 1 year. An example of the skewness of the data from the current residents is the average length of stay of about 3 years, whereas the median length of stay from the current resident data is only 1.7 years. Because of these limitations, the current resident data is inappropriate to examine the flow of patients in and out of nursing homes. The best data for investigating this issue would be the longitudinal study of persons admitted to nursing homes. Longitudinal surveys, however, are expensive to conduct. We plan to construct a cohort of nursing home admissions using data from both the current resident and discharged resident questionnaire as well as the next-of-kin follow-up.
To date, one report on the use of nursing homes by the elderly has been published using the current resident data. This report showed that about 5 percent of the elderly resided in nursing homes on any given day during the survey period of the 1985 survey. Use of nursing homes increases with age for both sexes, but was greater for females than males. Use of nursing homes was lower for elderly persons who are Black or of other races than for White. For the most part, these trends have not changed since 1973 when the first NNHS was conducted. However, there are some exceptions. There has been an increase in the use of nursing homes by elderly Black persons and a decrease in the use by those 85+ since 1973.
The next report that will present current resident data will be the summary report that Genevieve has already mentioned. This report will probably be released at the end of the year. The next report after that using current resident data will be the study of impact of the Medicare PPS on nursing home care.
QUESTION: Who administered the interviews to the residents in the facility?
ESTHER HING: No resident was ever directly interviewed. We interviewed the staff of the nursing home.
QUESTION: Could you give me some examples of behavior problems and disturbance in general?
ESTHER HING: The disturbance of mood question is basically a question with a checklist that asks whether the resident displayed depression, anxiety, fearfulness or worry to such a degree that they were distressed in functioning nearly every day. A general question asked whether the resident had displayed any behaviors that were considered dependent or disruptive. The disruptive behaviors included disrobing, exposing oneself, screaming, being physically abusive to self or others, stealing, getting lost, wandering into unacceptable places and the inability to avoid simple dangers.
EDWARD SEKSCENSKI: Although continuities do exist with many of the data items in the 1973-1974 and 1977 discharged resident segments of the NNHS, some of the items have not been repeated and a number of other items have been added to the 1985 survey. I will briefly outline some of the similarities and differences between the 1973-1974, 1977 and 1985 surveys, and hope to cover all the items in the 1985 survey of the discharged resident component.
Data in the discharged resident file of the 1985 survey were obtained from personal interviews conducted in the sample nursing homes with employees deemed most knowledgeable of the discharged resident's health status and condition during their stay at the sample home. In most cases the interviewee was either a nurse or medical records person who consulted with the available medical records of the discharged resident during the course of the interview. As was true in both previous NNHS's and in the current resident segment of the 1985 survey, no resident was consulted personally in the discharge component of the 1985 survey. Unlike the 1973-1974 and 1977 surveys, the 12 month reference period from which the discharged resident sample was drawn for the 1985 survey ended on the date immediately preceding the survey date. The survey was conducted between August 1985 and January 1986. Therefore, the 12 month reference periods could range from August 1984 through January 1986.
Previous survey reference periods for the discharges were calendar years 1972 and 1976. The survey's reference period was changed for the 1985 survey in an attempt to obtain both more current and readily available data and to provide information on the utilization of nursing homes by both residents and discharges over a more closely related period of time. However, data from the 1985 NNHS for discharged resident population and the current resident population continue to differ in several major areas.
Briefly, while the discharged resident estimates represent all discharges over a 12 month period, the current resident population is estimated for a single night, that immediately prior to the survey dates. The discharge sample, therefore, may underestimate those nursing home residents who tend to stay for very lengthy durations and the current resident population may underestimate those persons with very short durations of stay. While the current resident file provides for what may be considered a snapshot of nursing home residents on any given day, the discharged resident file provides for some indication of the over the year changes in the nursing home population, at least, that is, in terms of who is being discharged from the nation's approximately 20,000 nursing and related care homes.
A randomized sample of six or fewer discharged residents was selected per sample home. This resulted in an overall sample of 6,023 discharged residents. The discharged resident questionnaire collected data on the discharged residents' demographic characteristics, including age, sex, race, Hispanic origin, marital status, their discharge diagnoses, the discharge destinations of live discharges, whether or not the resident had difficulty controlling his bowel or urine in the last 7 days before discharge, and whether or not he or she was bedfast or chairfast during those last 7 days.
Also obtained was information on all sources of payment for the month of discharge from the nursing home. Unlike the current resident questionnaire, however, no charge data were obtained on the discharged residents in this discharge questionnaire. All of these above data items provided continuity with similar data items in the 1977 NNHS.
New to the 1985 discharged resident component of the survey are data items on primary diagnoses of discharged residents at admission, categorical information on prior living arrangements immediately preceding the admission, and the sources of payment for the month of admission. All of these data items focus on the characteristics of discharged residents prior to or at the time of their admissions to the nursing home. The admission that is relevant to the discharge which fell in the purview of the survey.
Also new were questions on the discharged residents' history of other stays in the sample home, in other nursing homes and the total number of homes which the discharged resident had ever been a resident in. The data from these latter questions will begin to provide some evidence of patterns of nursing home utilization over a longer period of time than a single stay in a single nursing home.
We received information on the marital status of the discharged resident both at the time of admission and the time of discharge. We received for the first time information on the discharged residents' living arrangement immediately prior to the admission to the nursing home.
We also asked information on the living arrangements of live discharges subsequent to their discharge from the nursing home. We asked a question as to whether they went to another health facility and whether they died in that other health facility.
Other questions dealt with whether the resident had been a resident of any other nursing home besides the nursing home that was in the sample, and we obtained information on the next-of-kin, a friend, or a person who would know about the condition of the discharged resident after leaving the nursing home and immediately preceding admission to the nursing home. This respondent was utilized in the next-of-kin survey.
We also received admission and discharge diagnoses data, the primary diagnosis and all listed diagnoses on the medical records, and they were subsequently coded into ICD9/CM classifications.
In the 1985 survey, we did not receive detailed activities of daily living (ADL) information on discharges. The functional status information on discharges was limited to these categories: whether they were chair fast, bedfast, and whether they had difficulties in controlling bowel or bladder during the last 7 days of their stay in the nursing home.
QUESTION: Why did not you do all ADL?
EDWARD SEKSCENSKI: In the pretest that information was not deemed as being as reliable for discharges as for the current residents.
The final questions that we asked were on sources of payment. The new question here compared to the 1977 survey was sources of payment, both all sources and primary sources of payment, in the month of admission. This was repeated for the month of discharge.
An Advance Data report on discharges will be coming out in about 2-3 months in the publications of NCHS and a Series report on discharges will be coming out next year.
QUESTION: I have a concern about sources of payment. I am concerned that if someone uses that for public policy decisions that it might be misleading. I am not aware that sources of payment is a part of a medical record that the nursing staff and the medical records technician would know. How did you test the reliability of sources of payment?
EVELYN MATHIS: The sources of payment did not come from the medical record. For the current resident and for the discharges, when we got into certain kinds of information, we had to go to the financial people who kept the bills, paid the bills or did the billing. It did not come from the medical record.
JENNIFER MADANS: The next-of-kin component is unique in that it has not been a part of the NNHS previously. The fact that it has been included in 1985 survey reflects the more complicated questions that are being asked of long term care data bases.
First, some historic background. In the pretest, there was interest in looking at an admission cohort, not only in sampling current residents and discharged residents. In the pretest we actually took a sample of admissions. It proved to be very expensive and very time-consuming. As an alternative, we decided to change the sampling list for the discharges. The sampling frame was discharges in the past 12 months. An admission component could then be recreated from the other two resident components.
In the pretest, a nursing home history questionnaire was part of that admission cohort design. A family questionnaire was used in conjunction with the current resident component. Those two were combined into the next-of-kin component.
The next-of-kin component was designed to collect information that would not be available on the record. There are two reasons for doing this. One, to increase the analytic capability of the current resident and discharged resident components if you are dealing with them by themselves. The other, when you put together an admission cohort, was to look at a longer time-frame and to look at transition probabilities. We were interested in the subject's health and functioning status prior to the admission; for example, why were they admitted, what were they like before they were admitted to the nursing home, and did they have any history of previous nursing home utilization. These are things that might be absent from the record, especially in some smaller homes where they do not do a very extensive preadmission screening. We looked at ADL's at the time of the admission, so you could look at change over time in this factor and also in change in living arrangements that might have taken place since the facility-based component.
The next-of-kin component has three general objectives. We wanted to look at patterns of nursing home utilization and also patterns of hospital care that often accompanied the use of nursing homes. We did some analysis of the admission data from the pretest to show just how important it is to look at the combination of these two. We also wanted to look at predictors of nursing home care and predictors of different patterns of care. For example, what would predict a long term user, a short term user or someone who had multiple hospital stays in between nursing home stays. Finally, to the extent possible, we were interested in describing the natural history of functional dependence over time.
The original admission cohort included four re-contacts. When we went to the national design there was only one re-contact which took place relatively soon after the facility contact. However, we are now trying to follow-up on those people and we have about 2 years time lapse between the follow-up and the original data collection.
The people who are eligible for the next-of-kin are those 5,243 current residents and 6,023 discharged residents who had the resident questionnaires filled out. All of those people were eligible for the next-of-kin component. lf you add those up you do not get 11,196. That is because the discharge component is an event-based sample. Someone could have multiple discharges. We were only interested in following people so you have to subtract out all of the overlap cases.
We got the name of the next-of-kin; in some cases it was the name of the resident if that person had been discharged from the questionnaires themselves from the nursing home. Only 90 percent of the eligible residents were included in the data collection. There were many reasons for this. Primarily it was because the facility refused to release the name of the resident or the name of the next-of-kin. In some cases there just was no next-of-kin and either the resident was still in the home or had been discharged dead. Of the 10,123 where we attempted to get an interview, there was a 90 percent completion rate. The interviewing of the next-of-kin component was done through a computer assisted telephone interview. We felt that we had to go with something like that because there are so many possibilities, so many different scenarios of utilization patterns depending on the administrative definition of the stay or where the person was at the time of the facility contact. Any hard copy version would just be too complicated for an interviewer to administer. We developed this computer assisted telephone interview which worked quite well. It also allowed us to do a lot of data editing and consistency checks while collecting the data, so when we got the file out we had less cleaning to do.
In the proposed follow-up, we hope to re-contact someone for each of the residents who were included in the next-of-kin component who was alive at the time that the next-of-kin interview was done. One alternative that is being considered is just to go back to the person who gave us the information at the time of the next-of-kin. Recall that when we did the next-of-kin, we got someone who knew the person prior to the admission because we wanted some information about their history. However, in the case where someone was still in a home, either in the sample facility or some other facility, it is likely that the facility itself could give us more information about their current status and also their use of hospitals. There is a proposal to go back to the home that the person was in if they were still a nursing home resident.
We are also going to try to pull in people who got dropped at various stages, either because there was no name or because they could not contact the person at the time we were in the field. Six months after we do the first phase of this follow-up, we hope to re-contact using the same design and the same questionnaire to update the information. Here we are basically going to collect utilization information, such as where the person is currently living, their vital status, and if they have had any nursing home or hospital stays in the intervening follow-up time.
We hope to go into the field with this follow-up sometime this summer. We have sufficient funding for a core interview and we are soliciting additional funding to expand the scope. If there are any funders; out there, contact me.
NIA is interested in funding some work on analyzing these data tapes, particularly when the follow-up data tape comes out. They are a major funder of the follow-up. You should contact Richard Suzman on that.
ESTHER HING: The design of the 1985 NNHS is a complex multi-stage probability sample. For the data to reflect national estimates, the data needs to be inflated by a weighting factor. The weights for the 1985 NNHS estimators included three basic components: (1) inflation by the reciprocal of the probability of selection, (2) adjustment for non-response, and (3) a first-stage ratio adjustment to total beds in the sampling frame. For facility level estimates, such as the number of homes, beds or total costs, the probability of selection is a product of the facility's probability of being included in the sampling frame times the probability of its being selected from the frame. Only homes from the complement survey had a probability of being included in the sampling frame of less than one. For second-stage estimates of current residents, discharged residents, and RN's, the probability of selection is the product of the probability of the facility's selection times the second-stage probability of selection for these sampling units.
The non-response adjustment factor brings estimates based on the responding cases up to the level that would have been achieved if all eligible cases had responded. The effect of the first-stage bed ratio adjustment is to bring the sample in closer agreement with the known universe of beds.
All three components were used to estimate the facility characteristics correlated with bed size and estimates of current residents, discharged residents and RN's. The first-stage bed size radio adjustment, however, was not included in the estimates of nursing homes and facility characteristics unrelated to bed size.
Weighting factors used to estimate the number of residents and discharges with next-of-kin are similar to the weights for the current resident and discharged residents with the exception of an additional non-response adjustment factor for whether they responded to the question requesting the names of next-of-kin and an adjustment factor for the existence of next-of-kin. As a result, estimates of the residents and discharges with next-of-kin will be less than the overall estimates of residents and discharges.
QUESTION: What are some findings from the survey?
ESTHER HING: There are two Advance Data reports with data from the 1985 NNHS. One of them on the use of nursing homes by the elderly shows that the use rate of nursing homes, or residents per 1,000 population, has not really changed since 1977, or even from 1973. It is still about 5 percent on any given day. Basically, females are still using nursing homes at a higher rate than males. Whites are using nursing homes at a higher rate than Blacks and others.
These trends have not changed also since the first NNHS was conducted, but there was an increase in the percent of Black elderly in nursing homes. It is increased from 2 percent to almost 4 percent in 1985, and there has been a decline in the use rate by elderly 85 years and over since 1973.
QUESTION: There seems to be a new found interest in board and care facilities. To what extent can you use this data to take a look at these facilities?
EVELYN MATHIS: Board and care and almost all of your in-patient facilities are defined by different states or by different counties, and so the definition is going to vary.
What we have defined for the NNHS, is defined by our NMFl. If a facility meets the requirements or the criteria to be in the NMFI, then it is in the NNHS. For a facility to be in the NMFI, it must have at least three beds, it must provide care on an overnight basis, and it must provide something other than just room and board. If a facility is called room and board in a given area, but they provide medical, nursing, or personal care services, then it would be in.
QUESTION: Based on those criteria, what numbers are we talking about?
EVELYN MATHIS: You can look at our 1982 NMFI report. In 1986, the NMFI was surveyed again. That particular survey was called the Institutional Population of the Inventory of Long Term Care Places (ILTCP). It included not only what we call nursing and related care homes, but also some residential places.
If you look in one of the reports on current residents, it will tell you how the universe was compiled for this particular survey. Again, if something is called room and board but it provided the services to make it meet the qualification for the NNHS then it is in there.
QUESTION: On nursing home characteristics, you do have level of care of nursing homes, but, is it possible to differentiate?
EVELYN MATHIS: You can only use these data if the respondent answered questions about certification status and whether or not they had beds that were just residential care.
QUESTION: Are you seeing an average level of impairment higher than in 1977?
ESTHER HING: One Advance Data report shows that, in general, there are more elderly who have higher levels of ADL impairment in 1985 than in 1977.
QUESTION: The number of blacks per 1,000 nursing home residents has risen rather dramatically since 1973. Do you have any information to explain this? The two possible explanations that come to my mind are, one, the effect of anti-discrimination laws making it possible for people to go into nursing homes, and the second would be some family support.
ESTHER HING: In this report I could not explain it. Basically, a variable like antidiscrimination laws is something that is beyond the scope of our survey. We have no information on that. Insofar as what is happening out in the community, we do know that the Black elderly do tend to stay out of nursing homes as long as possible, and rely on their relatives and friends. That is about the only thing we could say about this trend.
QUESTION: Is there anything different about the factors that lead to utilization among black residents as opposed to others?
ESTHER HING: We have limited information in the NNHS. We have few demographic variables and basically what we do have on the demographics is in this report. There are other variables that we just do not have information on, and we could not say any more than that because it would be just speculation.
JENNIFER MADANS: One of the objectives of the next-of-kin component and the planned nursing home follow-up is to look at patterns of utilization and also factors that are associated with admission. We could look at differences in the distribution of the White/Black population at admission and also look at differences in the utilization pattern, the family characteristics and the living arrangements prior to admission. I do not know if there are enough Blacks in the sample to really do a very detailed analysis, but we could definitely start looking at the problem.
QUESTION: In your Advance Data report on use of nursing homes by the elderly, there is a category age 85+ which represents 25 percent of residents of national nursing homes. Merely by numbers in that category, is it not time to stop giving that as an open-ended category?
ESTHER HING: The next report that we publish will have information from all the questionnaires and we will be presenting more detailed age breaks in that report.
QUESTION: I recall in the survey you were going to be doing some tracking to try to establish some longitudinal categories and some longitudinal information.
EVELYN MATHIS: One aim of the entire NNHS is to look at utilization patterns and also to look at the interrelationships of hospital and nursing home utilization. The original design of the pretest had built into it a longitudinal component where we would go back and interview the next-of-kin or the resident if they had been discharged, for a sample of people who had been admitted in the prior 12 months. That component was dropped.
It will be possible to construct an admission cohort from the current resident and discharged resident samples. We have gone back and contacted the next-of-kin for anyone who had a completed current resident or discharged resident questionnaire. Next-of-kin was contacted possibly 3 months after the facility contact and information was collected on that person's history of use. We have information from the next-of-kin and from the facility.
We hope to go back into the field sometime this summer and update that information. We have approximately 2 years more of follow-up through which to track their utilization patterns. We would like to match this data to the HCFA files to get a more complete reporting of all hospitalizations.
From the analysis of the pretest I can tell you that there is a huge interrelationship. If you look at the difference in facilities, how they determine whether someone who is being readmitted if it is a new admission or an old admission, a fair amount of hospital utilization that is surrounding many of these stays would have been internal to them.
QUESTION: Are you linking the different data on the survey?
JENNIFER MADANS: We collect the Social Security number. That is not perfect, because it is not the claim number. On the follow-up we will get Medicare numbers and just do a computer match. We also hope to use the Social Security number to match to the National Death Index (NDI). A fair number of these discharged residents are deceased already, but we could follow the current residents. This becomes, I think, increasingly important if you want to look at an admission cohort since most of those people were alive at the time of the facility contact.
QUESTION: You just might be interested in the experience in tracking by Medicaid number in New York State, to take one state. We were amazed by the turnover of Medicaid numbers there was.
JENNIFER MADANS: We originally were going to match to Medicaid records. That proved to be infeasible. We will only match to Medicare records to get the hospital utilization. We hope to be able to get the nursing home utilization from a respondent whether it is the facility itself or from the next-of-kin. Where there is not next-of-kin it will be impossible, of course, to get the information.
It will be necessary to take the raw data as it appears on the file from the resident samples and from the next-of-kin sample and massage it, because you are going to get conflicting reports about when the stays started and stopped. You are using multiple sources of information with problem of the administrative versus "real" stays. What does that stay really look like? Since our interest was in not that administrative definition, we let the respondent on the next-of-kin questionnaire define the admission date. We got their characteristics prior to that admission, but it is going to be inconsistent with what the home told us. It becomes extremely complicated to use this kind of data. You have to make an attempt or start by trying to reconstruct these histories. As I said the data is not crystal clear; it is not very clean.
QUESTION: Given at the beginning of the year a person is at a certain age, a certain sex, obviously, and a certain nursing status, what is the probability within the following year that he is going to die, be transferred to a higher level of nursing care or be discharged? You were saying it is even difficult to establish the date of admissions. What reliability do you place on the data on the level of nursing care the person is receiving?
WILLIAM SCANLON: The level of care is not necessarily a variable which would be reliable, because level of care definitions are so different across states. What I think you are talking about is trying to create a cohort for a particular point in time and then try to look at that from another point in time. Some of that is feasible.
JENNIFER MADANS: You may also be aware that they have been discharged, if it was to another facility or hospital. If we can add the ADL's on to the next wave of contacts, you will have some change in functional status, but that is one of the variables that is now unfunded.
QUESTION: Have you thought about linking data with NDI?
EVELYN MATHIS: We plan to link to the NDI.
QUESTION: Is there source of payment information?
EDWARD SEKSCENSKI: It will be available in the discharged resident Advance Data report. It also will be available in the Series report on the current residents. There is information on all sources and primary source of payment for the admission month and discharge month, with no charge data for the discharge population. You have some charge data for the current resident population, from the current resident questionnaire.
ESTHER HING: In the next report, the summary report from the 1985 NNHS, one table will show the payment source at admission crossed by the payment source last month.
EDWARD SEKSCENSKI: That is for the current resident component. For the discharge component, hold your calls for awhile.
QUESTION: In what respect does the 1985 survey differ from the 1977 survey?
ESTHER HING: The 1985 survey was designed to be comparable with the 1977, but there are differences. In 1977 we had a number of questionnaires. There was a facility questionnaire which got information about the nursing home and its utilization characteristics, such as bed size, the number of resident days, and other certification. There was a questionnaire on expense which collected information on the costs of providing care. We also had a questionnaire on staff, in which we collected information on the demographic characteristics of all staff who provided direct services to the patients. We also had two patient samples. One of them was on current residents, which included all persons who were in the home as of the night before the survey, and there was a sample of persons discharged during calendar year 1976. The reason why we selected that time period is this sample was for people who had left nursing homes. They had completed length of stay, we could also find out their outcome, or if they went to another health facility. We actually wanted calendar year data for the previous year to be comparable with the NNHS previous to that.
If you can remember all of those components, basically the same components were fielded in 1985, but there were differences. In the 1985 NNHS the staff questionnaire, rather than collecting information on all staff who provided direct patient care, only was administered to RN's. There was a particular interest in finding out the factors that affect the recruitment and retention of RN's in nursing homes because, apparently, there is a shortage of RN's in nursing homes.
Another big difference is that in 1985, rather than having a sample of discharges for the previous calendar year, the discharges were for a calendar year as of one year preceding the day of the survey. It is basically the last 12 months of data. That was done in conjunction with the new component that we added to the NNHS which we called the next-of-kin. We are planning to try to have an admissions cohort using both data from the current resident sample and the discharge sample.
We also conducted a follow-up of those samples with the next-of-kin or any other contacts that we could identify in our nursing home visit. That next-of-kin component was conducted by telephone. There are a lot of changes, but there were still a lot that is comparable to the last survey.
QUESTION: Can an admissions cohort be constructed?
JENNIFER MADANS: We were interested in getting an admissions cohort. One way to do it is to get a list of all the people who were admitted that year, but it would be expensive.
Using both resident files to construct the cohort is just a more cost-efficient way of collecting that data. The questionnaire was modified so that some analysis of utilization numbers could be reflected. This is in conjunction with the interests of getting the admissions cohort.
The other reason to modify the questionnaire relates to the problem of hospital stays that occur on either end of the nursing home stay or within a stay. Different facilities use different definitions of admissions. What we want to do is redefine these stays so that you have a consistent definition across facilities.
EVELYN MATHIS: If you deal with the admission event, as it is defined by the home, the admission took place in the last 365 days. If you go in on June 1, the last 365 days, would be June 1 to May 31. Somebody who was admitted in January, has either been discharged between January and June, so they could fall in the discharge sample, or they were still in the home, so they could fall in the current resident sample. If you sample current residents and discharged residents you get all the people who were admitted in the last 365 days, plus a whole bunch of other people. Those other people you throw out in the analysis when you look at the admission cohort. You keep them in the analysis if you look at current residents and discharged residents.
EDWARD SEKSCENSKI: One of the things we already have looked at was the completed length of stay of discharged persons. There was an observed increase in the median length of stay, but this increase was not statistically significant.
JENNIFER MADANS: If the nature of homes are changing or reimbursement rules are changing, that may be affecting how a home defines a readmission or a new admission. In other words, do they or do they not hold the bed. That seems to be the difference. If they hold the bed, it is not a readmission. If they do not hold the bed it is a new admission. That tends to be related to funding or payment. If there has been a change in the kinds of facilities that do one thing or the other, you may see something that looks like an increased length of stay, but it is purely an administrative artifact. When we did the pretest for the 1985 survey, and we had the admission cohort, we redefined the stay. We said, all right, if we want to look at an admission date, where were these people before they were admitted. They were in a hospital. Where were they before that? They were in that home. That is not a real admission. We backed up the date until we came to what we called the real admission date.
We recalculated the length of stay and we compared that to what Manton and Liu calculated, where they made that kind of synthetic cohort for 1977, and you get an increase in length of stay. What you do not know is that is it a real change or is it just a change in how you define the stay. When you look at the difference between length of stay calculated on the administrative definition and length of stay calculated on this "analytic" definition, you get much longer lengths of stay.
EDWARD SEKSCENSKI: The median length of stay that I was talking about was calculated from the difference between the admission and discharge dates that were given in the answers to the question in the discharge questionnaire. That median changed from 75 to 82 days. That difference was not significant. The length of stay distributions for less than 1 month, 1-2 months, 3-6 months, 6 months-1 year, 1-3 years, 3-5 years, and 5+ years also showed observed increases. Again, none of these was significant either.
JENNIFER MADANS: There was some idea with DRG's there would be more short stayers and a lot of little stays, but most of the long stayers are in for cognitive impairment and they have very long stays. To the extent that their use increases, the proportion of the long stayers is going to increase. I do not think we will ever be able to really figure out a real difference.
QUESTION: Can you tell if people were readmitted?
ESTHER HING: We added a question this time to capture those people that were readmitted to the same nursing home.
EDWARD SEKSCENSKI: We have any other nursing home stays, including the nursing home that is in the sample.
QUESTION: Does the discharge sample follow people or events?
ESTHER HING: The discharge is an event sample. You come in as many times as you have been discharged. The next-of-kin follows people.
QUESTION: Is Medicaid spend-down information available?
JENNIFER MADANS: You can get spend-down information from the record and from the next-of-kin. We ask questions to the next-of-kin about a primary payment source for each stay and when it changed. We would like to ask much more direct questions in the follow-up: when was the person first admitted; were they Medicaid eligible; if not, when did they become eligible. You can get at it the way it is now, it is just a little bit harder.
QUESTION: Can you get payment source?
JENNIFER MADANS: As part of the current resident and discharged resident questionnaires you get payment source. Therefore, you have it for the admission component. I think you get It from both of them in admission and for the current residents, currently; and for the discharged residents, at discharge. You really have to combine the sources of data and then you have all the problems that we had that people talked about earlier. All the identifiers will match. Sometimes all the data does not.
QUESTION: What reports will be issued on the survey?
ESTHER HING: There will be a summary volume similar to what we had in 1977. The next individual report of current resident data will be the impact of the hospital payments system on the nursing home care. There are a number of other reports. There is going to be another utilization report that I am trying to work on. I think there is going to an Advance Data report on the RN's.
QUESTION: Can you compare full-time with part-time employees?
EVELYN MATHIS: In the facility questionnaire itself, there is a list of various kinds of employees. The instruction is to list the number of part-time and full-time employees. We did not ask them to give us the ratios, but one would be able to determine that.