Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 28
3 Currently Available Datasets This chapter describes ongoing surveys and other data collection efforts that include information about decedents. It describes (1) datasets traditionally used to describe dying; (2) Census Bureau surveys; (3) national health surveys; (4) surveys focused on the elderly; (5) disease registries and surveillance studies; (6) reimbursement-specific administrative databases; and (7) other data collected by and about health care organizations. This chapter describes the original purpose of each data source and a brief overview of the way in which the information gathered could contribute to an improved understanding of dying in America. The appendices provide details about the sample, information about the timeframe of data collection, and availability of the data, questionnaires and coding instructions for these data sources, as well as a bibliography of publications derived from the data. The chapter concludes with a discussion of the information most readily available from existing datasets and the limitations of these data in answering the important questions about the end of life that were posed at the end of Chapter 2. There are a number of ways to arrange the data sources described in this chapter. They are ordered with those most relevant to death and dying first, and the more general surveys, with some information of interest, later. All the entries are also listed in Table 3-1, which has a code for the nature of each data set—e.g., administrative, national health survey— in addition to noting the numbers of deaths covered and the types of information collected.
OCR for page 29
DATASETS TRADITIONALLY USED TO DESCRIBE DYING Death Certificates The most immediate information gathered about a death is found on the state-required death certificate. National vital statistics rely heavily on information from death certificates, although federal tracking of death registration was not complete across the entire United States until 1933 (Lilienfeld and Lilienfield, 1980). The U.S. Standard Certificate of Death (Figure 3-1) was developed through a collaborative effort between the National Center for Health Statistics and the states. Although there is no formal agreement to use this standard certificate, the National Center for Health Statistics has an agreement with the states through the Vital Statistics Cooperative Program to provide data in specified formats consistent with the standard certificate. The death certificate assigns physicians, medical examiners, and coroners the responsibility for documenting the cause of death through a system that acknowledges the possibility of multiple causes. The disease, injury or complication that caused death (not the mode of dying such as a cardiac arrest) is noted on line “a” as the immediate cause of death. Then diseases or injuries that initiated the events resulting in death are recorded, beginning with the condition that gave rise to the immediate cause of death (line b), which in turn resulted from a further condition (line c). Additional lines may be added as necessary such that the underlying cause of death is on the lowest line used in Part I. Part II of the death certificate is used to list other significant conditions contributing to death but not resulting in the underlying cause of death. Demographic data (such as age, race, place of residence and marital status) and crude measures of socioeconomic status (SES) (such as educational attainment, usual occupation, kind of business) are recorded by the funeral director on the death certificate. The National Center for Health Statistics will recommend a revised version of the standard death certificate in 2003, which will include the expansion of the question on Place of Death to include “Hospice Facility.” Although problems with improper completion of death certificates and inaccuracy in reporting the cause of death continue (Lloyd-Jones et al., 1998; Maudsley and Williams, 1996; Smith Sehdev and Hutchins, 2001), the National Center for Health Statistics continues to develop and advance standardized coding, training workshops and other quality improvement efforts. Part of the problem is the lack of formal training in many medical schools or residency programs. Further, there is considerable variation among states in the sophistication of record keeping, in
OCR for page 30
TABLE 3-1 Types of Data Available from Datasets That Contain Individual-Level Information Data type Number of deaths year[s] Cause of death Comorbidities Pain or other physical symptoms C2 AHEAD 1,900 (1993-98) ✓ ✓ C2 LSOA I & II I: 2,900 (1984-90) II: 1,900 (1994-2000) ✓ ✓ C1 MCBS 700/yr ✓ B MDS ✓ ✓ B Medicare claims 1.7 million (2000) ✓ C1 MEPS NA2 ✓ ✓ A Death Certificates (NDI) 2.4 M (2000) ✓ C1 NHANES 50/yr estimate ✓ ✓ C2 NHEFS 1,400 (1987-92)1 ✓ ✓ ✓ C2 NHHCS ~5,000 (2000) ✓ C1 NHIS 55,000 (1986-94) ✓ ✓ C2 NLTCS 17,000 (1982-86) ✓ C2 NMFS 23,000 (1993) ✓ ✓ C1 NNHS ~1,600 (1999) ✓ B OASIS ✓ ✓ C1 PSID 4,300 (thru 1999) ✓
OCR for page 31
Physical function Psychological symptoms QoL Cognition Site of death Lived with Social support ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
OCR for page 32
Data type Family care Out of pocket expenses Bereavement ADS C2 AHEAD ✓ ✓ ✓ C2 LSOA I & II ✓ ✓ C1 MCBS ✓ B MDS ✓ ✓ B Medicare claims C1 MEPS ✓ ✓ A Death Certificates (NDI) C1 NHANES C2 NHEFS C2 NHHCS ✓ ✓ C1 NHIS ✓ C2 NLTCS ✓ ✓ ✓ C2 NMFS ✓ ✓ ✓ C1 NNHS ✓ B OASIS ✓ C1 PSID ✓ KEY: First Column: A = Vital statistics; B = Administrative data; C = Research sample (C1 = continuous or regular sample; C2 = not continuous or regularly occurring sample); AHEAD—Asset and Health Dynamics Among the Oldest Old; LSOA—Longitudinal Survey of Aging; MCBS—Medicare Current Beneficiary Survey; MDS—nursing home Minimum Data Set; MEPS—Medicare Expenditure Panel Survey; NDI—National Death Index; NHANES—National Health and Nutrition Examination Survey; NHEFS—NHANES I Epidemiologic Followup Survey; NHHCS—National Home and Hospice Care Survey; NHIS—National Health Interview Survey; NLTCS—National Long-Term Care Survey; NMFS—National Mortality Followback Survey; NNHS— National Nursing Home Survey; OASIS—Outcome and Assessment Information Set; PSID—Panel Study of Income Dynamics. cluding the use of electronic records. The use of death certificate information will be greatly facilitated as the National Center for Health Statistics works in collaboration with the states to build a totally electronic infrastructure to collect and manage these data.
OCR for page 33
Decision making Satisfaction Support Use of services Quality Provider ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ 1Deaths during 5-year follow-up period total 20-year follow-up (1971-1992). 2MEPS data will be linked to the NDI when a large enough number of deaths have occurred. It is currently too early for this. National Death Index (NDI) The NDI is a central computerized index of death certificate information that was established as a resource for epidemiologists and other health researchers to determine if persons in their studies have died. The NDI provides researchers with the state where death occurred and the
OCR for page 34
FIGURE 3-1 Standard U.S. Death Certificate.
OCR for page 35
death certificate number, which researchers use to obtain death certificate information directly from the appropriate state. To use the system, investigators must first submit an application to the National Center for Health Statistics. Fees are based on a service charge plus a fee for each user record for each year of death searched. National Mortality Followback Surveys (NMFS) The NMFS supplement information on death certificates with information on important characteristics of decedents. Information is collected from proxy next-of-kin informants identified on the death certificate. First instituted in the early 1960s, the various NMFS have been collaborative projects between the National Center for Health Statistics and various federal, state, and local organizations. To date, there have been six surveys (1961, 1962-1963, 1964-1965, 1966-1968, 1986, 1993) designed to provide information about the use of health services prior to death, socioeconomic status, aspects of life style, and other factors that may affect when and how death occurs. The 1986 NMFS file consisted of four types of linked records for 18,733 decedents, including (1) data from death certificates; (2) proxy surveys; (3) the medical examiners file for deaths due to homicide, suicide, and accidents; and (4) information from health care facilities used in the last year of life. The 1993 NMFS was based on a sample of 22,957 death certificates and focused on socioeconomic differentials in mortality and the prevention of premature death. The 1993 NMFS excluded deaths of people under age 15. For this survey, additional information was collected about risk factors and disability. No future surveys are currently being planned. National Longitudinal Mortality Study (NLMS) The NLMS is a large study of the effects of occupation, industry, income, education, and other socioeconomic factors on mortality. Records from several Current Population Surveys of the Bureau of the Census (1979-1981) have been matched to the National Death Index to link individual social and economic data with mortality outcomes. While the focus of this is on mortality, rather than quality or care of dying, the published reports do enhance our understanding of the role of social and economic variables on the age at death (Backlund et al., 1999; Johnson et al., 2000; Kaufman and Kaufman, 2001). This effort also highlights the value of linking the National Death Index with large ongoing national data collection efforts. The Current Population Surveys include only individuals age 16 and older, so no children are represented in the dataset linked to the National Death Index.
OCR for page 36
Medicare Data Medicare Claims Data Medicare claims data also contribute to our understanding of dying. More than 80 percent of American decedents received health care under Medicare, either because they qualified by being at least 65 years old or because they were disabled or had end-stage renal disease (Hogan et al., 2001). In addition to the various Medicare claims files, the Medicare Continuous History File contains Medicare beneficiaries’ basic entitlement data as well as Part A and Part B detailed claims (which include CPT and procedure codes). It is available continuously for 5 percent of all Medicare beneficiaries, or more than 1.5 million individuals in any given year. The entitlement file is linked to decedent records derived from the Social Security Administration. Approximately 5 percent of Medicare beneficiaries die in any given year. While the range of beneficiary characteristics is limited to age, sex, race, and geographical region, a wide array of health care utilization variables can be derived from the records of hospitalizations; stays in skilled nursing facilities; hospice services; and outpatient clinic, physician, and home health visits. These data have been used extensively to describe the intensity of medical services in the year preceding death (Virnig et al., 1999a,b, 2000, 2001, 2002; Christakis and Escarce, 1996; Christakis and Iwashyna, 2000). However, it should be noted that there are important gaps because of the lack of specific data about care delivered to beneficiaries receiving capitated HMO or hospice services. The only administrative data available in those cases is the amount paid per contract (in the case of the HMO) or per day (for hospice). Approximately 20 percent of deaths among Medicare beneficiaries occur under the hospice benefit. The current lack of detail about the specific services or medications received during that care within Medicare claims data hampers the ability to evaluate the quality of that care. Minor increases in the data elements on these claims forms would provide a means to evaluate the content of care and begin to assess quality. A small number of children are eligible for Medicare because of disability, but this would be a negligible source of information about children who die. However, Medicaid data are available about deaths among children covered by that program (discussed later in this chapter). Medicare Current Beneficiary Survey (MCBS) Within the Medicare program, the Centers for Medicare and Medicaid Services (CMS)—formerly known as the Health Care Financing Administration—conducts the MCBS, a continuous, multipurpose survey of
OCR for page 37
a representative national sample of the Medicare population, including those living in institutions. The purpose of the MCBS is to determine health care expenditures and sources of payments for all services beneficiaries use (including non-Medicare covered services) and to trace changes in health status and spending over time. Participants are interviewed in person for about an hour every four months, for up to four years (at which time they are replaced by new interviewees). At each interview, they are asked about their use of health services, medical care expenditures, health insurance coverage, and sources of payment (public and private, including out-of-pocket payments). Other questions are asked once each year, including information of relevance to this report, on health status and functioning. If the person is unable to answer the survey questions, he or she is asked to designate a proxy respondent, usually a family member or close acquaintance who is familiar with his or her care. About 15 percent of the interviews of non-institutionalized individuals are with proxy respondents. The data are designed to support both cross-sectional and longitudinal analyses and can be linked to Medicare claims data. The first round of interviews was conducted in 1991, and the survey has been in the field continuously since then. The target sample for the MCBS is 12,000 persons with three years of cost and use information. The MCBS includes roughly 700 decedents each year (Hogan et al., 2000) and adds important information about health care utilization and functioning prior to death. CENSUS BUREAU SURVEYS Although the datasets described above have been the ones most traditionally used to describe dying in America, many others have the potential to add to our understanding of end-of-life issues. Data collected by the U.S. Bureau of the Census provide general statistical information about the condition of the country and its population. Census 2000 From the decennial Census 2000, information is available on about 115.9 million housing units and 281.4 million people. Although the primary purpose of the census is to identify demographic trends, two questions about disability were included in the Census 2000 long form that provide a general picture of physical impairment within the population. In approximately one out of six households, the respondent was asked about the presence of a long-lasting condition that limited basic physical activities such as walking, climbing stairs, reaching, lifting, or carrying. A
OCR for page 38
second question addressed difficulty learning, remembering, concentrating, dressing, bathing or getting around inside the home. Other Census Bureau Surveys The Census Bureau continually conducts surveys throughout the decade between censuses to track social and economic conditions. These include the following: Current Population Survey (CPS). The CPS is a monthly survey of 50,000 households, is primarily targeted to provide information on the labor force of individuals age 16 and older (i.e., no children). However, CPS questions also address disability, and recent supplements included a Health/Pension survey in 1994. A regular supplement deals with health insurance coverage. Survey of Income and Program Participation (SIPP). The SIPP is another source of information on disability from the Census Bureau. The SIPP is a longitudinal panel survey of non-institutionalized adults, in which approximately 14,000 households are entered at the beginning of each calendar year and interviewed at specific intervals over a 2- to 3-year period of time. An extensive set of disability questions was asked in the 1990, 1991, 1992, 1993, 1996, and 2000 panels. Although the data from these surveys conducted by the Census Bureau have limited direct value in understanding quality of life or care at the end of life, the sampling strategies established by the Census Bureau are the foundation for most federal surveys. The coordination of federal surveys is an efficient way of ensuring that samples will be representative of the national population and that individual households will not be burdened with multiple federal surveys. The Census Bureau applies oversampling techniques to improve precision for racial and ethnic minorities and other demographically defined domains. NATIONAL HEALTH SURVEYS Large federal health surveys are also useful to understanding what happens at the end of life. There is likely to be substantial variation in the health of persons in the months and years before death, but surprisingly little published data support this assumption. Examining the information provided by respondents to health surveys who later die will begin to fill this gap. For example, beginning with the survey year 1986, the National Center for Health Statistics collected linkage information on all respondents to the National Health Interview Survey (NHIS), described
OCR for page 42
Health and Retirement Study (HRS) and Asset and Health Dynamics Among the Oldest Old (AHEAD) Study The HRS, fielded first in 1992, focuses primarily on factors affecting retirement, health insurance, savings, and economic well-being. The target population first included household residents in the contiguous United States born during the years 1931-1941. The original HRS panel consisted of 12,600 persons in 7,600 households. AHEAD studies examine the implications of health dynamics in old age for transitions in economic well-being, changes in family and marital status, and for reliance on public and private support systems. The target population for the first AHEAD survey, fielded in 1993, consisted of U.S. household residents who were born in 1923 or before. The original AHEAD study sample consisted of 7,447 HRS respondents aged 70+. AHEAD data provide information on the costs of illness borne by the family and the effectiveness of varying care arrangements in preserving function and delaying institutionalization. In 1998, the HRS and AHEAD studies were combined to provide a rich longitudinal dataset concerned with the health, economics, and demography of aging. For the combined study, a sample of “War Babies” (born between 1942 and 1947) and a sample of “Children of the Depression” (born between 1924 and 1930) have been added. Respondents are followed longitudinally at 2-year intervals until they die. At the time of the 1998 interviews, about 1,900 of the original AHEAD participants had died. National Long-Term Care Surveys (NLTCS) The NLTCS are surveys of the entire aged population with a particular emphasis on the aged who are functionally impaired. The samples are drawn from Medicare beneficiary enrollment files and are nationally representative of both community and institutionalized residents. A screening questionnaire identifies elders with functional disabilities and those so identified are followed until death. Medicare records are continuously added to each person’s data file through an automatic linking process. To date, four surveys make up this collection: 1982, 1984, 1989, and 1994. Each contains a longitudinal follow-up of previously surveyed disabled elders as well as screening for additional Medicare recipients with disabilities. A supplemental survey of caregivers was fielded in 1989 and 1999, though comparisons over time are difficult because of changes in the survey. Approximately 17,000 deaths have been identified from Medicare administrative records for respondents interviewed between 1982 and 1996. A next-of-kin supplemental survey was conducted as part of the 1994 NLTCS, for the estimated 4,800 persons who died between 1989 and 1994.
OCR for page 43
Panel Study of Income Dynamics (PSID) The PSID is a recurring national survey focused on economic and demographic issues. The major source of funding is the National Science Foundation, with additional sponsorship by NIH and other government agencies. The survey is conducted from the Survey Research Center, Institute for Social Research, University of Michigan. A wide variety of information is collected at the family and individual level about family composition, childbirth, and residential location. Over the life of the survey, the National Institute on Aging has funded supplements on health, wealth, parental health and long term care, and the financial impact of illness. For example, in 1993 the PSID interview included detailed information concerning the health events of the elderly and their burden on immediate and extended families. As a general rule, prior to 1999 only global health questions had been asked; for 1999 and 2001, much greater detail on specific health conditions is included in the health module. The PSID has gathered data about the fact and date of death of many former PSID respondents during 1993 and 1994 efforts to recontact former respondents, and through use of the NDI. During the 1990 interviewing wave, in conjunction with an NIA-funded project, respondents age 55 or older who indicated they were Medicare beneficiaries were asked to sign permission forms for access to their 1984-1990 Medicare claim records. Those who agreed were asked to renew that permission verbally in 1991 and in following years for later Medicare claims. About 4,300 deaths had been recorded through 1999, and updates through 2001 will be made soon. The questionnaire information on out-of-pocket medical expenditures and the long time-series of core PSID information are being linked to Medicare claims data, and should result in a valuable data set. DISEASE REGISTRIES AND SURVEILLANCE STUDIES Disease registries provide targeted information about those with specific health conditions. Examples of these are the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute and the Framingham Heart Studies. Surveillance, Epidemiology, and End Results (SEER) Program The National Cancer Institute’s SEER Program has been collecting and publishing cancer incidence and survival data from 11 population-based cancer registries and three supplemental registries which together cover approximately 14 percent of the U.S. population. In 2001, the pro
OCR for page 44
gram expanded, adding several registries and expanding existing ones, which increased coverage to one-quarter of the U.S. population. The database currently contains information on more than 3 million cancer cases, and approximately 170,000 new cases are added each year. There were 5,689 cases recorded among children and young adults age 19 and under during the 5-year period 1993 through 1997, or roughly 1,000 cases per year (NCI, 2002). SEER registries routinely collect data about patient demographics, primary tumor site, morphology, stage at diagnosis, first course of treatment, and vital status follow-up. By itself, SEER registry data do not provide information on the overall quality of care or quality of life of individuals, but because it is population based, it does provide a solid sampling frame for further investigations, particularly when linked to other data sets. SEER data have been linked to the National Death Index and Medicare claims data, but the potential for other linkages exists. REIMBURSEMENT-SPECIFIC ADMINISTRATIVE DATABASES Although the use of data from the Medicare program was described earlier, there are other major sources of administrative data linked to reimbursement systems. All the systems are large, but none is representative of the population as a whole (as the Medicare system is for people over 65): State-specific Medicaid claims data are available from states and the Centers for Medicare and Medicaid Services (CMS) (formerly known as the Health Care Financing Administration). Although detail may be missing for some persons enrolled in managed care, the remaining data elements are detailed and quite accurate (Arnett, 1998). Unlike Medicare, however, longitudinal tracking of individuals is unreliable, because people often enroll and disenroll in Medicaid, and may move from state to state. Administrative data from private insurers are available. Large private insurers such as Blue Cross/Blue Shield and UnitedHealth Group maintain comprehensive claims databases, as do commercial vendors such as MEDSTAT/SysteMetrics, Inc. and Shared Medical Systems, Inc. (Arnett, 1998; Paul et al., 1993). Data from private insurers, because they are generally employer based, would be the best sources of information about health care for children, young and middle-aged adults who die. Data from private insurers vary in the level of detail regarding information on cost, diagnoses and procedures, and enrollment. Administrative data from the Defense Department and the De
OCR for page 45
partment of Veterans Affairs include the following (these include virtually no children and very few women in the VA system): The Defense Medical Information System (DMIS) contains patient data with data elements comparable to those found on the National Hospital Discharge Survey (reported in the next section). Health-related data files maintained by the Veterans Health Administration include the Patient Treatment File (PTF)—patient-specific claims type data for care received at VA facilities, including admission date, diagnosis, and procedures; the Outpatient Care File; and the Long-Term Care Patient Assessment Instrument File. The VA system tracks drug prescriptions in a centralized system, so it can be very helpful in studying the use of drugs by people with particular conditions (or, e.g., in the period before death). OTHER DATA COLLECTED BY AND ABOUT HEALTH CARE ORGANIZATIONS Data collected on a national level by health care organizations are another potentially important source of information about dying. Health care agencies collect these data to both understand the characteristics of the clients served and to monitor the quality of care provided. Such data allow us to use aggregated characteristics of those who are cared for and who die in various types of health care agencies to better understand patterns of end-of-life care. National Health Care Survey (NHCS) The National Center for Health Statistics merged and expanded its surveys of health care centers into one integrated survey of health care providers called the NHCS. The NHCS builds upon and now includes the National Hospital Discharge Survey, the National Ambulatory Medical Care Survey, the National Nursing Home Survey, and the National Health Provider Inventory. National Home and Hospice Care Survey (NHHCS) The new package also includes the NHHCS, which is of particular relevance to this report. The NHHCS is a continuing series of surveys of home and hospice care agencies in the United States. Data are collected through personal interviews with administrators and staff and include information that profiles the clients served at the time of the survey. Client data include referral and length of service, diagnoses, number of vis
OCR for page 46
its, patient charges, health status, reason for discharge, and types of services provided. The 1998 NHHCS sample consisted of 1,088 agencies. National Hospital Discharge Survey (NHDS) The NHDS, conducted annually since 1965, collects data from a sample of approximately 270,000 inpatient records acquired from a national sample of about 500 hospitals. Data about patients include demographic information, length of stay, and medical information about diagnoses and procedures. In 1996, deaths accounted for 3 percent of all discharges (which would include only a few children). Every state has hospital discharge datasets and these are useful in examining deaths occurring within hospitals. Healthcare Cost and Utilization Project (HCUP) The HCUP is a federal-state-industry partnership created to build a standardized, multi-state health data system. It comprises a family of administrative longitudinal databases that contain patient-level information in a uniform format with privacy protections in place: The Nationwide Inpatient Sample (NIS) includes inpatient data from a national sample of over 1,000 hospitals. The State Inpatient Databases (SID) cover inpatient care in community hospitals in 22 states that represent more than half of all U.S. hospital discharges. The State Ambulatory Surgery Databases (SASD), the project’s newest phase, currently provides data from ambulatory care encounters in nine states. Variables include primary and secondary diagnoses and procedures, admission and discharge status, demographics, charges, and length of stay. The uniform data in the HCUP are designed to facilitate comparative studies of health care services and the use and cost of these services. The HCUP also has been a powerful resource for the development of tools such as the Clinical Classifications Software, Comorbidity Software and the HCUP Quality Indicators. Children are included in these databases, but they would probably yield relatively little about children who die. Nursing Home Minimum Data Set (MDS) The nursing home MDS resident assessment and care screening system has evolved from a long-standing recognition that comprehensive
OCR for page 47
functional assessment of nursing home residents is central to maximizing their physical functioning and quality of life. The development of a uniform, comprehensive resident assessment system was one of the key recommendations of the 1986 report of the Institute of Medicine’s Committee on Nursing Home Regulation. In 1990, the Health Care Financing Administration (now known as the Centers for Medicare and Medicaid Services, or CMS) published a Resident Assessment Instrument (RAI) and specified that assessments would be conducted on each Medicaid certified nursing home resident upon admission, upon “significant change” in status, and at least annually after admission or significant change. Beginning in 1998, all nursing home facilities were required to automate and transmit MDS information to the states, which were required to forward it to the Health Care Financing Administration (HCFA, 1998). The MDS potentially offers a very rich resource of clinical information regarding nursing home residents’ functional status, health conditions, services received, demographic characteristics, and other items such as the presence of advanced directives and family participation. As reliability trials have demonstrated, however, nursing facility staff can produce research-quality data, but, in practice, facilities differ in their commitment to ensuring that staff are trained and that they adhere to assessment protocols (Hawes et al., 1995). Furthermore, in spite of the mandate to perform an assessment with each change in status, there is wide variation in the timing of assessments prior to the death of a nursing home resident. Outcome and Assessment Information Set (OASIS) OASIS was developed for the purposes of measuring patient outcomes in home health care and is now an integral part of the revised Conditions of Participation for Medicare-certified home health agencies. OASIS is the result of many years of research, development and demonstration programs funded by CMS and the Robert Wood Johnson Foundation. The current release (OASIS-B1) is a set of 79 items refined by expert panels and demonstration programs. Routine patient assessment approaches are the intended source of information. Among the data items is the fact of death at home. Data Related to Initiatives for Assessing and Improving the Quality of Care Two large initiatives aimed at assessing and improving the quality of care provided by health care organizations also warrant consideration: (1)
OCR for page 48
the ORYX initiative of the Joint Commission on the Accreditation of Healthcare Organizations (JCAHO); and (2) the Health Plan Employer Data and Information Set (HEDIS) of the National Committee for Quality Assurance (NCQA). JCAHO’s ORYX Initiative JCAHO evaluates and accredits nearly 19,000 health care organizations and programs in the United States. Accreditation by JCAHO is recognized nationwide as an indication that an organization has met certain performance standards. To earn and maintain accreditation, an organization must undergo an on-site survey by a JCAHO survey team at least every three years. In 1997, JCAHO announced the ORYX initiative, which integrates outcomes and other performance measurement data into the accreditation process. During the current phase of ORYX, accredited agencies collect and submit data on a minimum of six performance measures available through enrollment in any of over 250 listed performance measurements systems. In May 2001, JCAHO announced four initial core measurement areas for hospitals: acute myocardial infarction, heart failure, community-acquired pneumonia, and pregnancy and related conditions. Hospitals will begin collecting these data for patient discharges beginning July 1, 2002. JCAHO is striving to make core measures for home care and long-term care organizations as consistent as possible with the CMS requirements for OASIS and the nursing home MDS. The ORYX initiative should prove to be a useful source of detailed information about care provided for select conditions relevant to the end of life, but until a significant amount of data become available, its value will not be known. NCQA’s Health Plan Employer Data and Information Set (HEDIS) NCQA is an independent, non-profit organization whose mission is to evaluate and report on the quality of the nation’s managed care organizations. NCQA evaluates health plans against 60 different standards that fall into the following broad categories: access and service, qualified providers, health promotion activities, and quality of care of acute and chronically ill. NCQA’s HEDIS is a tool used by more than 90 percent of health plans to measure performance on important dimensions of care and service. Among the more than 60 measures in HEDIS are measures related to patient satisfaction, inpatient utilization, frequency of selected procedures, and disease-specific treatments such as the use of beta-blockers after a
OCR for page 49
heart attack. The purpose of this data collection is to assist employers and consumers in selecting the best health plan for their needs. The Quality Compass, a national database of HEDIS and NCQA accreditation information from hundreds of health plans, makes it possible to look at health plans side by side to see how they compare. TOOLS FOR ACCESSING DATA Many of the agencies sponsoring large data collection efforts have made useful tools publicly available to facilitate access to the aggregate data. For example, the Federal Electronic Research and Review Extraction Tool (FERRET) is a collaborative effort between the National Center for Health Statistics and the Bureau of the Census to provide full access to complex federally sponsored datasets through the Internet. Currently, the 1994 Underlying Cause-of-Death File, the 1993 NHIS, and NHANES III are available via FERRET, which provides the capabilities to create cross tabulations, frequencies, and SAS or ASCII output files. CDC Wonder is an easy-to-use system that provides a single point of access to a wide variety of reports, guidelines, and data from the Centers for Disease Control and Prevention. Data can be readily summarized and analyzed from public-use datasets about mortality, cancer incidence, hospital discharges, AIDS, behavioral risk factors, diabetes and many other topics. The National Center for Health Statistics also maintains the Statistical Export and Tabulation System (SETS), which gives data users the tools to access and manipulate large data files such as the NHIS, LSOA, NHHCS, and NNHS. An example of a more specialized data tool is SEER Stat, which provides cancer investigators with an easy-to-use desktop system for the production of statistics useful in studying the impact of cancer on a population. In addition to these aggregate-level data tools, data warehouses, such as the National Archive of Computerized Data on Aging (http://www.icpsr.umich.edu/NACDA/index.html) and the National Center for Health Statistics Data Warehouse (http://www.cdc.gov/nchs/datawh.htm), provide means to access the public use data files for many federally sponsored surveys. INFORMATION MOST READILY AVAILABLE FROM EXISTING DATA SOURCES Table 3-1 summarizes the types of data available from those datasets presented in this chapter that contain individual-level information. Addi
OCR for page 50
tional information about how each dataset captures information in selected categories is found in Appendix B. LIMITATIONS OF EXISTING DATA SOURCES IN ANSWERING KEY QUESTIONS Currently available datasets often provide only sketchy answers to pressing questions about the end of life among adults. For children, no reliable answers to these questions emerge from any dataset. The recent IOM report, When Children Die (IOM, 2002), found a similar situation, and recommends better data collection. (The reader is referred to that report for more detail.) Examples of the limitations of existing data in answering the questions listed at the end of Chapter 2 are noted below and comprehensively in Table 3-1. Question #1: Where are people dying and how much of the end of their lives is spent in those settings? The site of death is recorded on a death certificate, but the period of time spent in that final setting is not noted. Question #2: Who is providing care for them as they die? Inferences about the use of formal care can be derived from administrative reimbursement-related data. The National Hospital Discharge Survey provides an opportunity to look at the subset of discharged deceased to learn more about patterns among final hospital stays (length of stay, cost, procedures performed). Little information is available about informal caregiving. Some health surveys ask community-based respondents questions about assistance with activities of daily living. However, there are few surveys seeking detailed information directly from care-givers. We know little about who was available to provide informal care at the end of life, who actually provided care, and who assisted in the coordination of care. Question #3: Are physical and psychological symptoms being identified and treated? Only 6 of the 16 datasets analyzed included questions about physical symptoms. All 6 asked about pain; 3 also included questions about shortness of breath, and 2 collected data about fatigue. Information was obtained about anxiety, depression, or “emotional problems” in 6 datasets. In addition to the presence of symptoms, however, treatments and procedures directed at ameliorating symptoms are important. The recent increase in inclusion of information about pain in record keeping by hospitals, home health agencies, hospices and nursing homes data systems will provide an opportunity to better monitor progress towards reducing pain at the end of life. By allowing us to track the aggregate responses to these questions over time, future data from sources such
OCR for page 51
as the MDS, OASIS, and JCAHO-mandated records will play a key role in determining if persistent pain is a problem and, if so, whether it is the result of inattention, inadequate treatment, or is simply unavoidable. Question #4: How many persons experience impaired cognitive function before death, to what extent, and for what period of time? Information about cognitive function is available in nine datasets, though these data are generally cross-sectional. Important information could be gained from the analyses of this information if data points are close in time to the date of death. We know very little about the degree to which cognitive impairment affects communication about dying. Question #5: How many persons experience physical disability or social isolation before death, to what extent, and for what period of time? Physical disability is captured in 14 of the analyzed datasets, though the structure and wording of questions vary considerably, making comparisons and assessments of change very difficult. Social networks may be inferred from information about living arrangements collected in many datasets, but the perception of social isolation is not measured. Question #6: How do patients and loved ones perceive their qual ity of life at various time points prior to death? Information about quality of life is not readily available. Questions were included in the first wave of the AHEAD study, but other surveys and forms of ongoing record keeping do not generally seek this information. Question #7: Are patients and loved ones achieving a sense of life closure in their relationships and spiritually? Although the importance of closure at the end of life is being increasingly acknowledged, empirical inquiry is at an early stage of development. No approach to asking about life closure or spirituality has been accepted for widespread use at this time. Question #8: Are patients and loved ones involved in decision making about treatment and care options? Very limited information is available about the existence of advance directive documents in patient records. Essentially no information is widely available that describes the degree to which these are implemented, how often they are overridden, or if providers and patients have had discussions about goals of treatment. Question #9: Are patients and loved ones receiving timely and adequate information on which to base informed decisions? No questions surveying patients or loved ones about the degree to which they felt adequately informed were identified. Question #10: Are patients and loved ones receiving supportive services from chaplains, therapists, health aides, and other providers? None of the datasets analyzed provide a framework to create a comprehensive picture of the availability or utilization of support services.
OCR for page 52
Question #11: Are family physical, emotional and financial re sources being heavily depleted by the care of a dying family member? Surveys such as the AHEAD study and the National Long-Term Care Survey have the potential to provide information about the depletion of financial resources at the end of life, but little information is being collected about the physical and emotional burdens of caregiving. Question #12: Are loved ones supported through the grieving pro cess? No population-based information appears to be available about the availability or utilization of bereavement services. The recommendations in this report are intended to provide initial direction concerning how we might improve the availability and quality of data to better answer these questions.
Representative terms from entire chapter: