PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of hsresearchLink to Publisher's site
 
Health Serv Res. 2011 February; 46(1 Pt 1): 120–137.
PMCID: PMC3015013
NIHMSID: NIHMS237690

The Residential History File: Studying Nursing Home Residents' Long-Term Care Histories*

Abstract

Objective

To construct a data tool, the Residential History File (RHF), that summarizes information from Medicare claims and nursing home (NH) Minimum Data Set (MDS) assessments to track people through health care locations, including non-Medicare-paid NH stays.

Data Sources

Online Survey of Certification and Reporting (OSCAR) data for 202 free-standing NHs, Medicare Denominator, claims (parts A and B), and MDS assessments for 60,984 people who were present in one of these NHs in 2006.

Methods

The algorithm creating the RHF is outlined and the RHF for the study data are used to describe place of death. The identification of residents in NHs is compared with the reports in OSCAR and part B claims.

Principal Findings

The RHF correctly identified 84.8 percent of part B claims with place-of-service in NH, and it identified 18.3 less residents on average than reported in the OSCAR on the day of the survey. The RHF indicated that 17.5 percent non-Medicare NH decedents were transferred to the hospital to die versus 45.6 percent skilled nursing facility decedents.

Conclusions

The population-based design of the RHF makes it possible to conduct policy-relevant research to examine the variation in the rate and type of health care transitions across the United States.

Keywords: Medicare, Minimum Data Set (MDS), transitions in care settings, linking administrative files, tracking health care utilization

The adoption of prospective payment, first for hospitals in the early 1980s and then postacute settings a decade ago, created conflicting payment silos, with hospitals reducing length of stay and postacute providers accepting complex patients likely to be rehospitalized. These conflicting Medicare reimbursement incentives have been associated with high rates of transitions between providers because there are no consequences for maximizing reimbursements in this way. In particular, patients who are disabled or chronically or terminally ill, who are often served in nursing homes (NH) as their main long-term care provider have been subject to the consequences of the conflicting reimbursement incentives and have thus suffered from multiple transitions. In spite of the increasing recognition of the importance of care transitions among long-term care residents, most of the literature has concentrated on reporting total utilization per service type (mostly either inpatient or Medicare-paid skilled nursing facility [SNF] care) (Coleman et al. 2004; Mor et al. 2010;). Even the recent focus on geographic variation in Medicare costs has emphasized regional and hospital differences in the average intensity of inpatient use rather than the extent of variation in patients' experiences across the continuum of care within and between geographic areas (http://www.dartmouthatlas.org).

Historically it has been difficult to assemble patient histories using existing claims data for those discharged from hospital to postacute settings because conflicting reimbursement incentives also translate to disparate reimbursement systems and therefore administrative data systems. Thus, composing transition histories using only Medicare claims does not provide information on long-term NH care (Burton et al. 1995; Brown et al. 1999; Cooper et al. 2000;). Alternatively, using only NH federally mandated regular assessment of all NH residents using the Minimum Data Set (MDS) resident assessment instrument data provides limited information about transitions outside of the NH (Coburn, Keith, and Bolda 2002).

The absence of data on NH use may lead to misleading conclusions. Some studies of Medicare expenditures at the end of life found those to be lower for older than younger persons (Gornick, McMillan, and Lubitz 1993; Levinsky et al. 2001;). However, Roos, Montgomery, and Roos (1987), using a Canadian longitudinal data set that included data on both acute and long-term care utilization, found end-of-life public health care utilization expenditures did not decrease with age. The discrepancy between the United States and Canadian findings may be explained by the absence of NH stays in the U.S. research and highlight the need for accurate information on all utilization when studying expenditures and budgeting for care.

The purpose of this paper is to describe the creation of a “residential history file” (RHF), using an algorithm that links Medicare claims and NH MDS assessments that results in a dataset (the RHF) which tracks the timing and location of health service use. Initially, we developed this method to track postacute care for patients hospitalized for hip fracture or stroke (Intrator et al. 2003). Subsequently we expanded the method to other applications ranging from studies of hospice use among Medicaid NH residents to tracking posthospitalization disposition of NH residents (Miller et al. 2004; Intrator et al. 2007;). In this paper we present the structure of the RHF algorithm and apply it to a cohort of all Medicare beneficiaries who were in one of 202 free-standing nonpediatric NHs at some time in 2006. We describe the resulting RHF, present an alternative RHF based only on MDS data, and conduct an illustrative analysis of a study of place of death. We then present results from comparisons identifying patients in NHs based on the RHF versus on Online Survey of Certification and Reporting (OSCAR) and on the place of service codes recorded on Medicare part B claims.

METHODS

Administrative data are a compelling source for the study of population-wide health care utilization, patient outcomes, and organization and system evaluation. In the United States several administrative data sources are available to researchers from the Center for Medicare and Medicaid Services (CMS). We used Medicare claims and NH resident assessments and a facility-level resource, the OSCAR, which reports results of the annual certification of NHs and contains information regarding NH deficiencies as well as information about the NH residents in aggregate. In particular, we used the reported total number of NH residents and the number of Medicare residents on the day of the survey.

MDS Data from NH Residents' Assessments

Filing of the MDS resident assessment instrument is mandated for every NH resident at admission, quarterly, annually, at discharge, and at any significant change of care. The instrument contains nearly 400 data elements which have been combined to create composite outcome scales, quality indicators, and various case mix and risk-adjustment measures (Fries et al. 1994; Hawes et al. 1995; Mor et al. 1995, 2003; Phillips and Morris 1997; Gambassi et al. 1998; Hirdes, Frijters, and Teare 2003; Grabowski, Angelelli, and Mor 2004; Mor et al. 2004; Wu et al. 2005). Since 1999 the MDS data have been used to determine level of payment of Medicare billable SNF services using 44–53 Resource Utilization Groups (Fries et al. 1994). Because the MDS is used to determine Medicare SNF payment levels, special Medicare MDS assessments are conducted more frequently during the first several weeks of a SNF stay to properly adjust payment levels to residents' care needs. Since 2002 the MDS data have been used to publically report quality of NH care (http://www.medicare.gov/nhcompare.). The use of MDS for both payment and quality monitoring requires that the MDS is done in a timely manner and able to withstand audit.

Medicare Eligibility and Claims Data

Medicare's routine administrative databases include detailed demographic, financial, and clinical data. All records in the claims and enrollment files include unique identifiers for Medicare enrollees that allow longitudinal linkage, enabling detailed descriptions of the specific clinical services provided to patients. Enrollment files include demographic data (birth date, age, gender, race, and place of residence), eligibility information (in particular, Medicare parts A and B eligibility and periods of HMO enrollment), and vital status (date of death [DOD]). Medicare part A, institutional claims files, have been used extensively in research. Generally, they contain information on dates of claims, diagnoses, services provided, charges, and reimbursements. Outpatient claims may be used by an NH to bill for skilled services for residents who had exhausted their part A SNF benefits or were otherwise ineligible for SNF care.

Part B claims have been used less frequently in research. Like part A claims, these claims contain information on the date of service, charges, services provided, and reimbursements. They also provide information on the place of service. Among the 50 or so place of service codes two indicate NH location (SNF and nursing facility). One study examined the utility of part B claims place of service code to identify NH utilization (Iwashyna 2003). Using self-reported utilization data from the 1993 Medicare Current Beneficiary Survey (MCBS) for validation, it reported that the place of service code was 89.7 percent sensitive and 97.7 percent specific in detecting any NH utilization within a year. Although an impressive rate over a 1-year period of time, that paper did not attempt to define the duration of an episode of NH utilization nor to validate the accuracy of the actual place of service. We examine the sensitivity and specificity of the RHF (as the test) in identifying part B claims in the NH (assumed as the gold standard).

RHF Methodology

The goal of the RHF is to create a per-person chronological history of health service utilization and location of care within a prespecified calendar (e.g., throughout a calendar year). The first step of the algorithm assigns utilizations and associated locations to days in a calendar. Depending upon the type of claim, the basic information from a claim is the location of care (e.g., hospital, NH, emergency room, and home) type of provider (e.g., free-standing, hospital based, or swing bed SNF), and service type (e.g., hospice, SNF). The order of information entered into the daily calendar structure which controls the RHF (the data hierarchy) gives precedence to the records with dates that are most likely to be complete and accurate because Medicare payments depend upon them. Thus, inpatient claims are first filled into dates of the calendar followed by outpatient emergency department (ER) and observation days. Next SNF claims are entered onto days, followed by outpatient claims for skilled nursing service in a nursing home, and lastly home health claims are filled into days. The above claims are location specific. Hospice claims are not location specific, because hospice can be provided in community or institutional settings. Thus, the location of hospice care is defined using other data and includes hospice at home and hospice in NH. Consecutive days with the same location and provider form episodelets of care.

Once the calendar is populated by all information obtained from claims, remaining nonfilled days may be populated by projected NH days based on MDS assessment dates and type. For example, admission assessments are required within 2 weeks of admission; therefore, the RHF fills up to 14 days back during consecutive “gap” days to form an NH stay. Quarterly assessments are required every 3 months; thus, any gap days within 3 months preceding the quarterly assessment will be filled in the calendar as NH. Annual assessments are required each calendar year, around the time of the closest designated quarterly assessment. Discharge tracking assessments are required by CMS and are used to determine date of NH discharge, when present. A full list of MDS rules is available from the authors.

The DOD is determined using information from the Medicare Denominator file and claims. Death is recorded as occurring on the last RHF episodelet, thus enabling an easy identification of place of death.

We note that the RHF only summarizes the existing data and that several episodelets may need to be linked to define an “episode of care” which is relevant to the research question. For example, if the total number of NH days is of interest, several types of NH episodelets (e.g., SNF and MDS type) may need to be joined to span the complete time in NH.

Example 1: An RHF-Based Patient History

We present an example of a history constructed for a fictional Medicare beneficiary. A home health care claim was made on behalf of this individual spanning from January 1 through January 10. However, an ER outpatient claim was made on January 3, and an inpatient claim was made from January 8 through January 11. This was followed by a Medicare SNF claim January 11–15. The NH conducted an admission MDS assessment on January 12 and then a Medicare specified assessment on January 27. An MDS discharge tracking form was filed on January 31. Hospice filed a claim January 26–31, and another inpatient claim was made for January 30–31. The Denominator file reported a verified DOD on January 31.

Figure 1 is a diagram of the process of filling a calendar for the month of January for this patient. First circles are entered to mark inpatient care (January 8–10, 30–31), then triangles to mark SNF care (January 11–14). Next home health periods are entered on claimed days (January 1–10), although we note that they overlap ER on January 3 and inpatient claims on January 8–10. Following home health claims, MDS assessments are entered with their type (January 12 admission 1A, January 27 Medicare required 7O, and January 31, discharge tracking 8D). Rules based on regulations infer NH days during gap days (January 15–29). Hospice recorded on January 26–29 are hospice in NH, but on January 30–31 they are added as a secondary type of episodelet to the existing inpatient stay. Finally, death is noted on January 31st from Denominator eligibility records. The table under the monthly calendar presents the resulting RHF records.

Figure 1
Sequence of Filling Calendar Days with Claims and MDS Information for Fictional Patient

Cohort

The RHF algorithm was applied to a cohort of all Medicare beneficiaries identified in any one of 202 free-standing NHs during 2006 collected for another study (Katz et al. 2009). All Medicare eligibility records, parts A and B claims and MDS data, were obtained for these residents under a data use agreement (DUA) with CMS. The RHF algorithm was applied to these data to create an RHF that is described and used in the following sections.

Analyses

Site of death has been studied before using information from death certificates, or only to determine whether death occurred in the hospital. The RHF makes it possible to locate a person when s/he dies, allowing an examination of NH and home as places of death, and transfers in care sites at the end of life. Those are tabulated and the distributions described.

RHFs created using only MDS and Medicare Denominator data versus using all part A claims, Denominator and MDS data were compared in order to gain an understanding of the difference in identifying days in NH for residents who are not Medicare beneficiaries or who are receiving Medicare Managed Care. Total days in each episodelet type per person were compared, and the distribution of the difference described.

We examined the completeness of the RHF in determining NH episodelets by comparing it with two other sources. The first is the number of residents in an NH on the day of certification reported on the OSCAR. The second is the place of service codes on Medicare part B claims.

Because the RHF can be based on all NH residents regardless of their payer source (all residents must be assessed), the number of residents identified in the RHF on the day of the survey should be comparable to that reported in the OSCAR. Moreover, the OSCAR also includes a count of Medicare residents which should correspond to the number of residents receiving either SNF or hospice care in an NH on the day of the survey. The distribution of the difference between the OSCAR and RHF reports is described.

NH place of service codes in part B claim, and the episodelete type reported in the RHF on the day of the part B claim, are crosstabulated. Results are presented with the claim/visit as the unit of analysis. Assuming that part B place of service indicating NH is the “truth,” we calculated the sensitivity and specificity of the RHF to identify days in NH. Because part B visits are not conducted every day a resident is in an NH, we do not expect the specificity go be good.

RESULTS

Description of Study Data and Cohort

We received 263,040 MDS assessments from 59,810 unique residents in any of the 202 free-standing nonpediatric facilities during 2006 and 71,473 SNF claims from 30,627 residents. A total of 61,479 residents had either SNF or MDS assessments, of whom 28,756 (46.8 percent) had both SNF claims and MDS assessments, another 1,774 (2.9 percent) had only SNF claims, and 30,949 (50.3 percent) residents had only MDS assessments. Of the 61,479 residents with either an MDS or an SNF claim during 2006, 132 were not Medicare eligible reducing the sample to 61,347 residents. The RHF also identified 363 residents who had claims after death and who were removed from the RHF. Thus, the final RHF dataset included 60,984 individuals.

The final RHF included 506,477 episodelets lasting an average of 38.7 days (median 11 days). Among the cohort residents, 456 (0.7 percent) had no part A eligibility throughout the year, 683 (1.1 percent) had no part B eligibility throughout the year, and 10,446 (17.1 percent) received Medicare benefits through a managed care organization some time throughout 2006, with 6,016 (9.9 percent) having MCO coverage throughout 2006.

The average length of stay in the hospital was 7.6 days, with a median of 5 days. The durations of SNF, MDS-based, and outpatient-based NH episodelets were 24.3 (median=18), 55.6 (median=7), and 13.9 (median=10) days, respectively.

Example 2: Site of Death

Overall, 15,341 (25.2 percent) residents died during the year. Death in the hospital occurred for 2,974 patients (19.4 percent). NH was the place of death for 9,134 patients (59.5 percent) with 2,423 (15.8 percent) dying while receiving SNF care and 3,450 (22.5 percent) while receiving hospice care. The remaining 3,233 decedents (21.1 percent) died at home. Residents who died in the ER or during an observation stay (n=480, 3.1 percent) were assigned a place of death based on the location they originated from. The majority 390 (2.5 percent) died after being transferred to the ER from NH and the rest (90) died after being transferred to the ER from home.

The RHF allows us to identify the location before the site of death (Table 1). Interestingly, 1,485 of 3,233 residents who died at home (45.1 percent) were in an NH before that. Among residents who died in the hospital 19.1 percent were transferred from home, 68.2 percent from an NH (36.1 percent from SNF and 32.1 percent from non-SNF), while 14.6 percent of residents who died in the NH were transferred from home.

Table 1
Transitions from the Location before the Location of Death to Location of Death for 15,341 Cohort Decedents

Example 3: Comparison of Full RHF and Resident History File Based Only on MDS Data

We compared the “full RHF” created from all Medicare parts A claims, Denominator, and MDS data with an “MDS only RHF” created only from MDS and Denominator data in terms of the total duration in MDS-identified NH stay, any NH stay, and gaps, per person (Table 2). On average, the MDS-only RHF listed more MDS days than the full-RHF (127 versus 101 MDS days, respectively). Correspondingly, there were many more gap days in the MDS-only RHF than in the full RHF (238 versus 181 days, respectively). However, when comparing the total number of NH days of any type in the full RHF versus those accounted for by MDS only, the difference was very small, on average 1 more NH day in the full RHF versus the MDS-only RHF (interquartile range [IQR] 0–2 days).

Table 2
Comparison between MDS-Only and Full RHFs in Terms of Average Number of Days Per Resident

Comparison 1: RHF versus OSCAR Number of Residents on Day of Certification

Among the 202 NHs in the cohort 138 were surveyed in 2006. The OSCAR reported an average of 130.8 (IQR 85–159) residents in the NH on the day of the survey, and 18.6 (IQR 8–26) residents with Medicare as their primary payer. Using the RHF we identified an average of 112.5 (IQR 65–134) residents in the NH on the day of the survey, and an average of 21.0 (IQR 10–29) residents receiving Medicare paid care. On average, the RHF identified 18.3 (IQR 6–24) less residents than did OSCAR and the RHF identified 2.4 (IQR 0–6) more residents on Medicare than OSCAR.

Comparison 2: RHF NH Episodelets and Part B Claims Place of Service Code

We received a total of 5,365,457 part B claims for 53,527 beneficiaries from our sample (91.1 percent of residents in the cohort). Of these claims 917,059 (17.1 percent) reported place of service as an NH for beneficiaries either receiving SNF care or other non-SNF NH care. On the other hand, 1,920,728 (35.8 percent) part B claims had a date of service corresponding to when the RHF identified the beneficiary as residing in an NH. Table 3 shows the crosstabulation of place of service and location of resident on the day of the part B claim based on the RHF episodelets.

Table 3
RHF Location at Time of Part B Claims versus Part B Place of Service (row percents)

Among the 917,059 claims with place-of-service NH, 777,628 were observed in an NH by the RHF. Thus, the sensitivity of the RHF in identifying NH is 84.8 percent (777,628/917,059 part B claims). The specificity is 74.3 percent (1−[917,059–777,628]/[5,565,457–1,920,186] part B claims).

DISCUSSION

The Residential History methodology provides extremely useful information for many health care utilization research studies. By tracking peoples' health care utilization over time, the RHF provides a longitudinal picture of utilization, making possible the examination of access barriers, and discontinuity of care. There are many important research questions that require this technology in order to be conducted; for example, identifying a cross-section of NH residents at a particular day to describe their conditions (http://www.ltcfocus.org). Another example is identifying 30-day rehospitalizations from NH, currently hotly debated by policy makers, requires that NH admission be verified after an index hospitalization, and that rehospitalization only from NH be identified. While some of these differentiations can indeed be done without the benefit of the RHF, having it available makes creating such complex analyses files much more efficient. Even more important, the RHF summarizes information and knowledge about Medicare claims and MDS records that have been obtained and accumulated by experience, and thus provides a single resource for utilizing these components of data together. The RHF framework also provides a method to adjudicate overlapping claims. For example, determining DOD requires working with information from both Denominator and claims with knowledge-based algorithms designed to reconcile differences. Moreover, the comparison of the RHF based on MDS and claims data to the RHF based only on MDS and Denominator data reveals that the RHF is able to track NH location quite accurately for those residents who are not Medicare beneficiaries (approximately 5 percent of the NH population) or who have elected to receive Medicare Managed Care (approximately 17 percent).

The RHF algorithm resulted in an RHF file with much face validity. Compared with part B place of service, it was almost 85 percent sensitive in identifying NH location. Moreover, while it is expected that episodelets in MDS-only RHF would be longer than episodeletes in the full RHF given that part A claims provide more transfer information, it was reassuring that the total number of NH days identified by the two methods was similar.

Indeed, the study of site of death presented in this paper points to complexities in studying site of death. Using the RHF, a more nuanced understanding of end-of-life care is gained; for example, among decedents receiving SNF care before their site of death, 49.5 percent were transferred to the hospital to die. This compares to 18 percent of non-SNF NH decedents, raising the question, why that great a difference?

When comparing the RHF location with an NH place of service on the part B claims, we found the RHF correctly identified 84.8 percent of the NH part B claims. The majority of the part B dates not identified to be in the NH were identified in “gap” (presumably, community not receiving institutional care). This suggests that there is a potential to extend the RHF and increase the identification of NH stays using the part B claims. Indeed, if part B claims indicating the NH as the place of service infers that the existing RHF episodelet is in an NH, we find that 94.1 percent of NH episodelets had a part B stay with place of service NH.

The comparison of the number of residents observed on the day of the OSCAR survey with the OSCAR report indicates that, on average, the RHF reported 14 percent less total residents and 12.9 percent more residents receiving Medicare-paid (SNF or hospice) care than the OSCAR. However, based on the OSCAR, almost 10 percent of facilities a year have a discrepancy in the number of residents of at least 20 percent from the prior year, and almost 60 percent of facilities each year have at least 20 percent difference in total Medicare residents. Thus, it appears that the OSCAR report may be more erroneous than the report based on the RHF. Several other studies have questioned the validity of the OSCAR data (OIG 2003; Feng et al. 2005;).

Others versions of the RHF reported appear to be more limited and not well documented (Sood, Buntin, and Escarce 2008). One recent attempt to create a file that identifies periods of NH care has been conducted internally by CMS (L. “Spike” Duzor, personal communication). The “Stay File” identifies periods of NH stay and adjoining hospitalizations. It uses MDS data to determine periods of NH stays by using the assessment types recorded on the MDS and their corresponding dates. Unlike the RHF, this file was created in an attempt to examine NH utilization in isolation of other types of utilization and only uses the MedPAR data to determine hospitalizations surrounding or during NH stays. Based on our experience with the RHF algorithm, it would appear that the use of more limited Medicare data may create a file that would identify all NH days, but would incorrectly identify additional days as NH days. Therefore, it is not likely that this file will be as useful in examining transitions in care and discontinuity of care.

Several limitations are noted. The RHF methodology is usable only for researchers who have an approved DUA with CMS to use Medicare standard analytic files and NH MDS data. Although the information provided by the RHF is comprehensive, it pertains mainly to the Medicare fee-for-service population and does not include information of out-of-pocket expenses estimated to be about 17 percent of total expenses for Medicare eligible population (MEPS 2006). The program that implements the methodology is very computationally intensive and quite complex spanning over 10,000 lines of SAS code.

The RHF methodology can be useful for many studies that have linked survey data with Medicare claims and MDS data. For example, the MCBS, the SEER cancer registry, the Health and Retirement Survey, and the currently planned National Health and Aging Trends Study all routinely link to Medicare claims. MCBS adds MDS and OASIS data since 1999, building a crude timeline file of MDS and OASIS stays (F. Epig, personal communication). All these studies can benefit from knowledge on transition sequences and NH stays.

There are several ways to share this algorithm; one is to make it available as is on our website. However, the SAS program is very complex and would require a lot of support at our institution, which we cannot budget. Another method is to provide this code to CMS and to allow researchers to ask in a DUA that an RHF be created for them. CMS could also incorporate the RHF, or a reduced version of it, into the Chronic Condition Warehouse, which is intended to serve a somewhat similar, although less dynamic purpose.

Increasingly, the U.S. health care system will need to provide care for frail, older persons, many with a terminal disease trajectory of chronic, progressive illnesses with prolonged periods of functional dependency. Our current health care reimbursement system, for the most part, is based on fee-for-service payment to individual institutions, with recent policy providing incentives for cost containment. These incentives, in part, resulted in shorter hospital stays and a higher rate of transitions to SNFs. The population-based design of the RHF makes it possible to conduct policy-relevant research to examine the variation in the rate and type of health care transitions in the United States, examine the role of state policies and market characteristics on transition rates, and the impact on beneficiaries residing in geographic region with differing rates and types of transitions.

Acknowledgments

Joint Acknowledgment/Disclosure Statement : Supported in part by National Institute on Aging grants R01 AG-14427, R01 AG 020557, R21 AG 030191, and P01 AG027296, and Agency of Healthcare Research and Quality R01 HS10549. Special thanks to Julie Lima, Ph.D., who helped create Figure 1 and who has dealt with all the DUA issues over the past many years, to Linda Laliberte-Cote for her confidence in this concept and in continuing to support this effort throughout the years, and to Christian Brostrup-Jensen for his programming support and his comments and feedback in testing the residential history file program. A previous version of this paper was presented at the annual meeting of the Gerontological Society of America in 2003.

Disclosures: None.

Disclaimers: None.

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article:

Appendix SA1: Author Matrix.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

REFERENCES

  • Brown ML, Riley GF, Potosky AL, Etzioni RD. Obtaining Long-Term Disease Specific Costs of Care: Application to Medicare Enrollees Diagnosed with Colorectal Cancer. Medical Care. 1999;37(12):1249–59. [PubMed]
  • Burton LC, Steinwachs DM, German PS, Shapiro S, Brant LJ, Richards TM, Clark RD. Preventive Services for the Elderly: Would Coverage Affect Utilization and Costs under Medicare? American Journal of Public Health. 1995;85(3):387–91. [PubMed]
  • Coburn AF, Keith RG, Bolda EJ. The Impact of Rural Residence on Multiple Hospitalizations in Nursing Facility Residents. Gerontologist. 2002;42(5):661–6. [PubMed]
  • Coleman EA, Min SJ, Chomiak A, Kramer AM. Posthospital Care Transitions: Patterns, Complications, and Risk Identification. Health Service Research. 2004;39(5):1449–65. [PMC free article] [PubMed]
  • Cooper GS, Yuan Z, Chak A, Rimm AA. Patterns of Endoscopic Follow-Up after Surgery for Nonmetastatic Colorectal Cancer. Gastrointestinal Endoscopy. 2000;52(1):33–8. [PubMed]
  • Feng Z, Katz PR, Intrator O, Karuza J, Mor V. Physician and Nurse Staffing in Nursing Homes: The Role and Limitations of the Online Survey Certification and Reporting (OSCAR) System. Journal of American Medical Director Association. 2005;6(1):27–33. [PubMed]
  • Fries BE, Schneider DP, Foley WJ, Gavazzi M, Burke R, Cornelius E. Refining a Case-Mix Measure for Nursing Homes: Resource Utilization Groups (RUG-III) Medical Care. 1994;32(7):668–85. [PubMed]
  • Gambassi G, Landi F, Peng L, Brostrup-Jensen C, Calore K, Hiris J, Lipsitz L, Mor V, Bernabei R. Validity of Diagnostic and Drug Data in Standardized Nursing Home Resident Assessments: Potential for Geriatric Pharmacoepidemiology. SAGE Study Group. Systematic Assessment of Geriatric Drug Use via Epidemiology. Medical Care. 1998;36(2):167–79. [PubMed]
  • Gornick M, McMillan A, Lubitz J. A Longitudinal Perspective on Patterns of Medicare Payments. Health Affairs (Millwood) 1993;12(2):140–50. [PubMed]
  • Grabowski DC, Angelelli JJ, Mor V. Medicaid Payment and Risk-Adjusted Nursing Home Quality Measures. Health Affairs (Millwood) 2004;23(5):243–52. [PubMed]
  • Hawes C, Morris JN, Phillips CD, Mor V, Fries BE, Nonemaker S. Reliability Estimates for the Minimum Data Set for Nursing Home Resident Assessment and Care Screening (MDS) Gerontologist. 1995;35(2):172–8. [PubMed]
  • Hirdes JP, Frijters DH, Teare GF. The MDS-CHESS Scale: A New Measure to Predict Mortality in Institutionalized Older People. Journal of American Geriatrics Society. 2003;51(1):96–100. [PubMed]
  • Intrator O, Berg K, Hiris V, Mor V, Miller STG. Development and Validation of the Medicare-MDS Residential History File. Gerontologist. 2003;43:30–1.
  • Intrator O, Grabowski DC, Zinn J, Schleinitz M, Feng Z, Miller S, Mor V. Hospitalization of Nursing Home Residents: The Effects of States' Medicaid Payment and Bed-Hold Policies. Health Services Research. 2007;42(4):1651–71. [PMC free article] [PubMed]
  • Iwashyna TJ. On the Detection of Nursing Home Use in Medicare Claims. Health Services and Outcomes Research Methodology. 2003;4:187–96.
  • Katz PR, Karuza J, Intrator O, Mor V. Nursing Home Physician Specialists: A Response to the Workforce Crisis in Long-Term Care. Annals of Internal Medicine. 2009;150(6):411–3. [PubMed]
  • Levinsky NG, Yu W, Ash A, Moskowitz M, Gazelle G, Saynina O, Emanuel EJ. Influence of Age on Medicare Expenditures and Medical Care in the Last Year of Life. Journal of American Medical Association. 2001;286(11):1349–55. [PubMed]
  • MEPS. 2006. “Household Expenditures” [accessed on October 4, 2010]. Available at: http://meps.ahrq.gov/mepsweb/data_stats/tables_compendia_hh_interactive.jsp?_SERVICE=MEPSSocket0&_PROGRAM=MEPSPGM.TC.SAS&File=HCFY2006&Table=HCFY2006%5FPLEXP%5F%40&VAR1=AGE&VAR2=SEX&VAR3=RACETH5C&VAR4=INSURCOV&VAR5=POVCAT06&VAR6=MSA&VAR7=REGION&VAR8=HEALTH&VARO1=4+17+44+64&VARO2=1&VARO3=1&VARO4=1&VARO5=1&VARO6=1&VARO7=1&VARO8=1&_Debug=
  • Miller SC, Intrator O, Laliberte L, Cang Y. End-of-Life Care for Medicare Skilled Nursing Facility (SNF) Residents: Patterns of Care and Expenditures by Hospice Enrollment Status. Journal of the American Geriatrics Society. 2004;52(4):S193–S93.
  • Mor V, Angelelli J, Jones R, Roy J, Moore T, Morris J. Inter-Rater Reliability of Nursing Home Quality Indicators in the U.S. BMC Health Service Research. 2003;3(1):20. [PMC free article] [PubMed]
  • Mor V, Branco K, Fleishman J, Hawes C, Phillips C, Morris J, Fries B. The Structure of Social Engagement among Nursing Home Residents. Journal of Gerontology B Psycholical Science Society Science. 1995;50(1):P1–8. [PubMed]
  • Mor V, Grabowski D, Intrator O, Feng Z. The Revolving Door of Rehospitalizations from Skilled Nursing Facilities. Health Affairs. 2010;29(1):57–64. [PMC free article] [PubMed]
  • Mor V, Zinn J, Angelelli J, Teno JM, Miller SC. Driven to Tiers: Socioeconomic and Racial Disparities in the Quality of Nursing Home Care. Milbank Quarterly. 2004;82(2):227–56. [PubMed]
  • OIG. 2003. “Quality Assurance Committees in Nursing Homes.” OEI-01-01-00090 [accessed on October 4, 2010]. Available at: http://www.oig.hhs.gov/oei/reports/oei-01-01-00090.pdf.
  • Phillips CD, Morris JN. The Potential for Using Administrative and Clinical Data to Analyze Outcomes for the Cognitively Impaired: An Assessment of the Minimum Data Set for Nursing Homes. Alzheimer Disease Association Disorders. 1997;11(Suppl 6):162–7. [PubMed]
  • Roos NP, Montgomery P, Roos LL. Health Care Utilization in the Years Prior to Death. Milbank Quarterly. 1987;65(2):231–54. [PubMed]
  • Sood N, Buntin MB, Escarce JJ. Does How Much You Pay Matter? Evidence from the Inpatient Rehabilitation Care Prospective Payment System. Journal of Health Economics. 2008;27(4):1046–59. [PubMed]
  • Wu N, Miller SC, Lapane K, Roy J, Mor V. The Quality of the Quality Indicator of Pain Derived from the Minimum Data Set. Health Service Research. 2005;40(4):1197–216. [PMC free article] [PubMed]

Articles from Health Services Research are provided here courtesy of Health Research & Educational Trust