RUG III Case Mix Classification and Funding in Long-Term Care

by Dr. Gary F. Teare

In the last decade, provincial governments have increasingly sought to make the distribution of health care funding sensitive to differences in the amount of resources utilized to care for different kinds of patients or residents. Some of the earliest progress in that regard was seen in funding of acute care facilities. However, the funding of long term care has also been a focus of this policy direction, because without explicit – and effective - recognition of differing resource intensities among residents, a powerful disincentive exists to admitting or retaining residents with higher care needs.

Two basic elements are needed in order to successfully advance this movement towards funding equity: a valid and reliable method of identifying patients/residents with differing resource intensity of care, that is, a case-mix measurement system and a method of moving from case-mix measurement to a case-mix payment system.

Case-mix measurement is primarily a scientific issue of determining which characteristics relate to cost of care. Case-mix funding system design is mostly a political process, based on the goals and realities of a particular health care system’s funder(s) and providers. These elements are best considered at length separately to avoid confusing payment issues with measurement issues. Nevertheless, it is valuable to touch upon both elements to illustrate the importance of excellence in measurement to the quality of the payment system.

The science of developing a case-mix measurement system involves determining which characteristics of residents, and combinations thereof, are predictive of costs of care. However, simply applying statistical processes to identify those characteristics which best predict costs is inadequate. Careful attention must be paid to how the measurement system will ultimately operate in the real world and the implications of what gets measured.

Having most or all of the following characteristics characterizes an excellent casemix measurement/classification system for long term care:

The Resource Utilization Groups

The Resource Utilization Groups version III (RUG-III) case-mix classification system for institutional long term care meets each of the listed criteria. RUG-III is the case-mix system that drives government payment systems for nursing homes in the United States. It has also been adopted in Ontario for case-mix-based funding of hospital chronic care and its use is being explored by other provinces. The RUG case-mix classification system has been developed and refined through extensive research conducted over the last 20 years. RUGIII, most recently updated in 1997, is applied to patient/resident data collected using the second version of the Resident Assessment Instrument Minimum Data Set (MDS 2.0).

The MDS 2.0 is a comprehensive, standardized assessment of resident function, needs and strengths that was designed to assist the development of thorough plans of care. The validity and reliability of the MDS 2.0 assessment has been demonstrated both in research and routine clinical use (Hawes et al., 1995; Morris et al., 1997; Phillips and Morris, 1998; Teare et al., 2000). The MDS 2.0 has been translated into at least 14 languages and is in use, by government mandate or for research, in 20 countries worldwide. The development and multiple applications of the MDS 2.0 have been described elsewhere (Hirdes et al., 2000).

RUG-III was built and refined based on studies involving over 12,000 nursing home and skilled nursing facility residents in 9 states in the U.S between 1990 and 1997. These studies involved assessment of residents with the RAIMDS 2.0 (an earlier version of the MDS instrument was used in the 1990 study), and contemporaneous measurement of the staff time involved in care of individual residents on a daily basis. The MDS 2.0 assessment provided a data set of over 400 clinical characteristics for each resident, including measures of physical, mental and psychosocial functioning and service provision. Nursing staff time, resident-specific and otherwise, was measured separately for each kind of licensed and unlicensed nursing staff, over a 24 hour period. Staff time for non-nursing services (e. g., physical therapy, social work, etc.) which have less frequent contact with individual residents, was collected over a 7 day period then expressed on a per day basis.

All time attributable to the care of an individual resident (e.g., hands-on care, care planning conferences, resident-specific charting, discussions with family or physician) was assigned to that resident, while time not specific to particular residents (e.g., ward maintenance, breaks) was assigned across all residents. Daily minutes of care were weighted by the relative wage rates among nursing and non-nursing staff to derive nursing and auxiliary per diem costs of resident-specific time. Statistical classification methodologies were used to select the resident characteristics, and combinations of characteristics, which best predicted per diem costs of resident-specific time. Fixed costs such as utilities, laundry, capital expenses, were not included in the costs for deriving RUG-III as these costs are largely invariant among residents. These costs can be handled separately in a payment system.

RUG-III was found to explain over 50% of the variance in resident-specific care costs. The RUG performs better in this regard than other case-mix classification systems which have been designed for long term care, and compares favourably to the performance of other established case-mix systems such as the Diagnosis Related Groups (DRGs) or Case Mix Groups (CMGs) used in acute care in the U.S. and Canada, respectively. The variance of cost within groups was less than among individuals in the entire study sample, and the mean costs associated with the groups were statistically different from each other. Statistical criteria were not the only standards by which RUG was judged. Extensive clinical expert input was obtained for selection of a classification scheme that made clinical as well as statistical sense (Fries et al., 1994).

RUG-III has seven major categories, representing broad clinical types, which form the first grouping split in the classification. These categories, Special Rehabilitation, Extensive Services, Special Care, Clinically Complex, Cognitive Impairment, Behaviour Problems, and Reduced Physical Function are arranged hierarchically (in the order listed) from greatest to least average case-mix associated cost per resident. Subsequent splits based on other clinical characteristics (e.g., activities of daily living) result in ultimate classification into one of 44 RUG-III groups (Table 1). Residents are classified into the highest category for which they qualify, which means that they will be classified into a group that exceeds the resource intensity of all groups for which they would qualify in lower categories. Each of the 44 groups has a casemix index (CMI) associated with it, which is a measure of the relative utilization of daily care resources compared to a standard. Generally, the standard (CMI value of 1.0) is set to the average per-diem cost in the population to which to case-mix based payment system applies. For example, the average per diem cost among chronic care hospital patients in Ontario was used to establish the Ontario CMIs shown in Table 1.

[Editor's Note: All tables, footnotes and references are available in the print edition of STRIDE.]

The Rehabilitation category is defined based on the provision of professional rehabilitation services (physical, occupational or speech/ language therapies). The individual groups within this category are defined based on the amount and number of kinds of therapy services provided with subsequent splits based on an activities of daily living index score. Using service intensity in the Rehabilitation groups is the major instance in RUG-III where service provision (rather than underlying physical/ medical characteristics) is used in the classification. However, the cost of providing these services offsets any increased payment associated with them and including them acts as an incentive to their provision within long term care.

The Extensive Services groups capture the relatively rare but expensive residents with high medical complexity requiring higher levels of registered nursing involvement (e.g., respirator/ ventilator, parenteral feeding, suctioning, tracheostomy). The Special Care and Clinically Complex groups represent residents with other, less extreme, skilled nursing needs. Notably, although urinary catheter use was associated with clinical complexity and increased resource utilization it is not used as a case-mix indicator in RUG-III because its inclusion could introduce a perverse incentive to catheterize residents in order to qualify for a higher classification group.

In the research to develop RUG-III, cognitive impairment was associated with slightly higher resource use only among residents that did not have major medical problems and functional impairments. Therefore, if a resident did not qualify for any of the first four categories, but had short-term memory, orientation and decision-making deficits and had, at most, moderate physical function deficits, they could qualify for the Cognitive Impairment category. The Behaviour Problems category is for residents who have neither major medical or physical function problems nor major cognitive impairment, but who manifest severe behaviour problems (e.g., wandering, physical or verbal abuse). Classification into the Cognitive Impairment and Behaviour Problems categories is relatively rare, considering the frequency of medical complexity and/or profound physical impairment in long term care residents. Residents/patients classify into the Reduced Physical Functions category if they do not qualify for any of the preceding categories. Figure 1 shows the distribution among RUG-III categories of patients at admission to chronic care hospitals in Ontario, during fiscal year 1997/98.

Although developed in the U.S., the validity of RUG-III for long term care case-mix classification has been validated by research in over 9 countries, including Canada. Although the actual costs associated with each of the RUG-III groups may vary, the relative costs are remarkably stable among jurisdictions (Carpenter et al., 1997). The implication of this finding is that it is not essential to perform staff time measurement studies, which are very expensive, to calibrate RUG-III to the local context. Rather, the U.S. average nursing and therapy discipline minutes per day for each RUG-III group, which are freely available on the World Wide Web (www.hcfa.gov/medicare/ snfpps.htm), can be weighted by jurisdictionspecific relative wage rates to derive local RUG group case-mix indices. This was the approach taken in Ontario for chronic care (Teare, 1999).

The U.S. Health Care Financing Administration (HCFA) has demonstrated a sustained commitment to ensuring that the MDS assessment and the RUG case-mix classification system keep current with changing realities in continuing care. Since it was first introduced in 1990, the MDS has been revised twice and the RUG-III system has been significantly revised once (in 1997). The revisions to the MDS delivered improvements to the quality and precision of the assessment data collected. The RUG revision improved the performance of the classification system in differentiating groups of the more medically complex, heavier care residents that are increasingly cared for in continuing care settings. In the past year, research was undertaken on behalf of HCFA to explore possible further refinement of RUG-III with respect to certain very rare patient types, such as ventilator dependent patients, where direct care costs not related to nursing and the major therapy disciplines (“auxiliary costs”) can be inordinately high. However, the research did not lead to changes to RUG-III, since good predictors of auxiliary costs were not found. In addition to the research and development work initiated by HCFA, an international consortium of researchers and clinicians called interRAI, which owns the copyright to the MDS outside of the U.S., is dedicated to research into the use and application of the assessment tool in support of care of the elderly and disabled in 20 countries worldwide (www.interrai.org).

Translating Case-Mix Classifications into Case-Mix Funding

By differentiating costs of care among different groups of patients, case-mix classification enables health care payers to arrange equitable distribution of health care funding. Relatively more funding can be directed towards providers caring for relatively more resource intensive residents, thus removing the incentive, present in a non- (or poorly-) case-mix based funding arrangement, to turn away heavy care residents. Case mix classification per se does not address the issue of the absolute amount of funding required to provide care. Deciding the size of the funding envelope for long term care is very much a political process. Furthermore, the method by which the case-mix classification is applied to funding decisions is also largely a function of policy. Different paths for translating classification into funding have been followed by the various jurisdictions using RUG-III.

With a case-mix measurement system in place, there are three essential elements to be considered in the design of a case-mix based funding system: (1) How payment is structured; (2) How case-mix is estimated at the facility level; (3) Auditing procedures to counteract “gaming” and ensure accuracy.

Structuring Payment and Estimating Facilities' Case-Mix

There are two basic approaches by which casemix adjusted funding can be structured: a pricing approach and a “budget adjustment” approach. Using the pricing approach, a payment amount is set for each case-mix classification level and payment is made on a per patient/resident basis or on the basis of the average case-mix of patients in the facility. With a budget adjustment approach, adjustments are made to the facility’s budget for the next fiscal period, based on its expenditure per case-mix adjusted unit of care (e.g., case-mix adjusted patient day) compared to a standard (e.g., the provincial average), in the previous fiscal period.

A measure of a facility’s case-mix can be derived by aggregating all the case-mix weighted units of care (e.g., days) for individuals during a fiscal period (longitudinal aggregation). Alternatively, facility case-mix can be derived by taking the average case-mix score of patients present on a particular day as an estimate of the average case-mix in the facility during the period (cross-sectional or ”snapshot” estimation).

In the U.S., nursing home case-mix funding has been designed with a pricing approach. Different jurisdictions have taken different approaches to applying this funding method. Medicare, the federally funded health insurance program for the elderly in the U.S., structured its payment system for skilled nursing facilities as a reimbursement for each day of care provided to a Medicare-qualified resident, at a preset per diem rate based on the resident’s most recent RUG group classification. The RUG classification is updated by a new MDS assessment at specified intervals during the resident’s stay. Medicare has a separate per diem reimbursement rate for nursing care associated with each of the 44 RUG groups and different per diem therapy rate for each of the 5 main levels of the Special Rehabilitation groups. The total per diem reimbursement for each resident is the sum of the applicable nursing and therapy rates for the RUG-III group into which the resident was classified.

The total reimbursement for care of a Medicare resident/patient is determined by summation of the days of care at the per diem rate(s) for the RUG group(s) the resident is classified into during the course of their stay. A disadvantage of this resident-level reimbursement funding approach has been that some nursing facilities have incorrectly interpreted the per diem therapy rates associated with the Special Rehabilitation RUG groups as prescriptive of the exact dollar amount available to individual residents for rehabilitation professional services. This misunderstanding led to situations where residents who did not classify into a Special Rehabilitation group did not have rehabilitation services made available to them.

In contrast to the longitudinal, resident-byresident reimbursement approach taken by Medicare, a cross-sectional (snapshot) approach was taken by many states for determining case-mix based nursing home reimbursement for care of recipients of Medicaid, the state-managed health insurance for people with low income. In this approach, the average of the RUG case-mix indices of the patients present on a certain day is taken as an estimate of the average case-mix index for the facility during a fiscal period. The facility is paid a single per diem reimbursement rate for all Medicaid residents, based on the facility’s average case-mix index at the time of the “snapshot”. A similar approach is used in Ontario for funding nursing homes based on an annual calculation of a facility case-mix index based on the Alberta Resident Classification System. A potential problem with this approach is that the facility may be over- or underfunded for the fiscal period if the group of residents present in the facility when the average case-mix index is calculated does not accurately represent the facility’s average casemix over the entire period to which the casemix “snapshot” is applied.

In designing a case-mix funding methodology for Ontario hospital chronic care, the preferred approach to estimating facility case-mix was longitudinal aggregation of case-mix weighted units of service for all patients within a fiscal period. Case-mix was to be applied to hospital budget adjustment to ensure funding equity across the province. A further consideration was that since chronic care (complex continuing care) in Ontario is provided by hospitals, most of which also provide acute care, hospital funding would be simpler and more transparent if case-mix based funding of chronic care could fit within a unified approach to funding all levels of hospital care.

A methodology was devised to associate each patient’s days in chronic care with the RUG-III group into which they classified, based on their most recent MDS assessment. The case-mix index of that RUG group was used to weight the patient days, and these RUG-weighted patient days (RWPD) were summed for all patients in a facility during a fiscal year (Teare, 1999). Costs for chronic care were allocated from within costs reported by hospitals to the Ontario Ministry of Health and Long Term Care (MoHLTC) in the Ontario Hospital Reporting System, using a previously established methodology (the Ontario Cost Distribution Methodology [OCDM]). Dividing RWPDs, the measure of case-mix weighted chronic care activity (volume), by the OCDM measure of chronic care costs gave the cost per RWPD. Cost per RWPD can be calculated at the facility or provincial levels, depending on the level at which RWPDs and costs are aggregated.

Comparison of a facility’s cost per RWPD to the province-level cost per RWPD provides a measure of efficiency of chronic care facilities. As a measure of cost per case-mix adjusted unit of care, cost per RWPD can also be compared with cost per weighted acute care case (derived based on OCDM costs and CMGweighted care episodes) to derive costequivalent units of patient care activity for acute and chronic care.

The Joint Policy and Planning Committee, of the Ontario Hospital Association and the MoHLTC, has used this approach in developing methodology to support a unified hospital funding formula. The method compares hospitals’ actual costs per weighted activity (acute and chronic, and all levels of care in “small” hospitals) to the “expected” (provincial average) level, after adjusting for certain factors associated with increased costs (tertiary care, “teachingness” and geographic isolation). In a general sense, application of this method to hospital funding would see redistribution of funds from hospitals with weighted activity costs in excess of the expected level (inefficient, or relatively over-funded) to hospitals with lower than expected expenditure for their weighted activity (more efficient, or under-funded). Several options exist for how to apply this approach of comparing cost per weighted activity to an expected level in a funding formula and the choice among them rests with the MoHLTC.

Structuring Payment and Estimating Facilities' Case-Mix

It is important that payers establish, early in the design of a case-mix funding system, mechanisms to ensure the integrity of the casemix data upon which the system is founded. “Gaming” - the material misrepresentation of patient/resident characteristics in order to qualify for higher case-mix funding groups - is always a temptation to unscrupulous operators. Ideally, a case-mix funding system is built upon a case-mix measurement system designed to minimize gaming, and backed up by ongoing, targeted on-site and off-site data auditing to act as a deterrent to those who would “play” the system.

The MDS 2.0 assessment, which serves as the measurement system for RUG-III, has several important features built-in to counteract gaming. First, the MDS 2.0 assessment items have been framed to capture data in a manner that lends itself to verification. For example, patient/resident performance of functional tasks (such as activities of daily living), not their capacity to do them, are recorded.

Performance can be verified, whereas the judgment of capacity is much more subjective. Second, the MDS 2.0 was designed to have direct clinical application in support of care planning. The utility to the clinical team of the information gathered in the assessment creates conditions supportive of accurate assessment. Third, the MDS Quality Indicators, which serve as flags of potential quality of care problems, can be used to counterbalance some efforts to game the system. For example, presence of urinary tract infection (UTI) serves as a qualifier for the Clinically Complex RUG-III groups, but the prevalence of UTI is also a Quality Indicator. Thus, if a facility falsely inflated the number of residents with UTI, in order to have more residents classified into a higher case-mix group, the facility would also risk standing out from their peers with a higher prevalence of UTI, indicating potentially poorer care practices. In the U.S., the MDS QIs are already being used by state nursing home surveyors to identify facilities needing closer scrutiny for checks of quality of care and the fidelity of their MDS data.

Aside from having a case-mix data gathering system with built-in deterrents to gaming, it is important to have processes for assessing the quality and likely fidelity of case-mix assessment data submitted by facilities. There will always be a need for representatives of the payer to do “spot checks” at facilities to periodically verify the accuracy of a sample of assessments. However, the U.S. government has recently been doing a considerable amount of research on methods to assess MDS 2.0 data quality and the likelihood of data problems (including gaming) at particular facilities. Currently, both on-site and off-site MDS 2.0 data auditing methods are being developed and field-tested. The aim of their effort is to facilitate more efficient audits of MDS 2.0 data to maximize accuracy in case-mix classification and minimize gaming.

The RUG-III classification system has the qualities required for excellent case-mix measurement and lends itself to several approaches to the design of case-mix based funding systems for long term care. Therefore, it is perhaps not surprising that several Canadian provinces, and several countries around the world, are moving towards case-mix funding for long term care, based on RUG-III. Such widespread application of a single casemix system in long term care may present interesting opportunities to learn the impacts of different funding policies, as various jurisdictions apply RUG-III to long term care funding by a variety of methods.