5 - Prevention | Understanding Thin Client/Server Computing (Strategic Technology Series)

Editors: Kane, Robert L.; Ouslander, Joseph G.; Abrass, Itamar B.

Title: Essentials of Clinical Geriatrics, 5th Edition

> Table of Contents > Part I - The Aging Patient and Geriatric Assessment > Chapter 4 - Developing Clinical Expectations

Chapter 4

Developing Clinical Expectations

Health professionals' and informal care providers' attitudes toward older patients can affect their care. As with most chronic disease, the effect of good care is generally slowing the rate of decline rather than a reversal of a clinical course. Thus, it may be harder to discern without information about what would happen in the absence of good care. Moreover, elderly patients are in danger of being dismissed as hopeless or not worth the effort on the basis of their age. Physicians faced with the question of how much time and resources to spend in searching for a diagnosis will want to consider the probability of benefit for the investment. In some cases, older patients are better investments than younger ones. This apparent paradox occurs in the case of some preventive strategies when high risk of susceptibility and the discounted benefits of future health favor older persons. But it also arises in situations where small increments of change can yield dramatic differences.

Perhaps the most striking example of the latter is found in the case of nursing home patients. Very modest changes in their routine such as introducing a pet, giving them a plant to tend, or increasing their sense of control over their environment can produce dramatic improvements in mood and morale.

At the same time, the risk:benefit ratio is different with older patients. Treatments that might be easily tolerated in younger patients may pose a much greater risk of producing harmful effects in older patients. As shown in Figure 4-1, the therapeutic window that separates benefit from harm is narrower. In effect, the dosage that will produce a positive effect more closely approaches one that can lead to a toxic effect. As noted earlier, one of the hallmarks of aging is a loss of responsiveness to stress. In this context, treatment may be viewed as a stress.

FIGURE 4-1 Narrowing of the therapeutic window. This diagram portrays in a conceptual manner how the space between a therapeutic dose and a toxic dose narrows with age.

Those who treat older patients must also consider the theory of competitive risks. Because older persons often suffer from multiple problems, treating one problem may provide an opportunity for more adverse effects from another. In essence, eliminating one cause of death increases the likelihood of death from other causes.

Physicians treating older patients may suffer simultaneously from too much information and too little. A patient with a substantial history of care may come

P.72

with a multivolume set of charts from which little usable data can be gleaned in the time allotted for an outpatient encounter. Modern laboratory reports often provide more data than is requested, distracting as well as overinforming.

A useful tool for creating a more proactive and focused attitude among those who care for older persons is the flowchart. Focusing on a few clinical parameters that are both significant and most likely to be affected by treatment helps the clinicians focus their attention and recognize changes early. Because the changes are likely to be subtle, it is often helpful to establish treatment goals with time frames for achieving them. Both the health care team and the patient can then agree on expectations and follow progress toward the goal.

It is important that the goals be achievable. Small successes are very important and reinforcing. Thus the units of measurement should be capable of detecting small but meaningful changes. In many instances, small gains can, in fact, make an enormous difference. The stroke patient, for example, who regains the use of hand muscles has a greatly improved ability to function. Being able to change position in bed may mean the difference between getting pressure sores and not. Regaining a method of communication, whether by speech or some other means, can restore social contact.

By introducing gradual, small steps, a functional task may appear more achievable. We have all had some experience in getting a bedridden patient to

P.73

resume a more active role. For an older person who has been at bed rest for a long period, this task requires overcoming both physiological and psychological problems. Small steps will often ease the transition and provide an opportunity to monitor the effects at each stage to minimize risk.

CLINICAL GLIDE PATHS

In one sense, geriatrics can be thought of as the intersection of chronic disease care and gerontology. The principles of chronic disease care apply in spades here. As noted earlier in this book, care for chronic disease is conceptually different from that for acute disease. To provide effective chronic care, one needs a longitudinally oriented information system that is sensitive to change. Each clinical encounter with a chronically ill patient is essentially a part of a continuing episode of care; it has a history and a future. Caring for a chronically ill patient, especially one with multiple problems, demands an enormous feat of memory as the patient's list of problems is unearthed and the history, treatments, and expectations associated with each are reviewed. Clinicians caring for such patients (often under enormous time pressures) may find themselves either overwhelmed with large volumes of data from which they must quickly extract the most salient facts or, alternatively, relying on inadequate data from which to reconstruct the patient's clinical course. Moreover, because patients live with their disease 24 hours a day, 7 days a week, they are best positioned to make regular observations about its progress. Such constructive patient involvement responds to another principle of chronic care. These goals can be achieved using a simple information system that can focus the clinician's attention on salient parameters.

One approach to help to focus clinicians' attention on the salient parameters of a chronic condition is the clinical glide path. This technique is modeled on the way one lands an airplane. The goal is to establish a flight path that will lead to a safe landing. Information about the plane's position is provided regularly to allow the pilot to make small corrections in order to avoid making drastic adjustments at the last minute or crashing. In a similar vein, one can manage chronic illnesses by establishing the salient parameters, the indications that the patient's clinical course is being achieved. For each chronic problem, the clinician should identify the one or two parameters (physiologic or functional) that will best indicate how things are going. Such information is often symptoms such as shortness of breath, pain, or fatigue; but they can also be simple signs such as weight change, blood pressure or blood sugar. This information is then systematically collected (by the clinician or more likely by the patient) and entered into a flow sheet, which can be converted into a graph. As long as the patient's course is as good as or better than predicted, no adjustments need be made; but indications that the course is beginning to slip below that expected

P.74

should trigger an early reassessment, in the same way that a pilot is warned when the plane begins to move out of the glide path box. Figure 4-2 shows an example of a glide path.

FIGURE 4-2 Clinical glide path models. In this model (A), the expected course (solid line) calls for gradual decline. The confidence intervals are shown as dotted lines. Actual measures that are within or better than the glide path are shown as o. When the patient's course is worse than expected, the o changes to an x. The design shown uses confidence intervals with upper and lower bounds, but actually only the lower bound is pertinent. Any performance above the upper confidence interval boundary is very acceptable. The design of the glide path can also take another form (B). It may be preferable to think in terms of reaching a threshold level within a given time window (e.g., in recuperating from an illness) and then maintaining that level.

The choice of the parameters to be tracked for each problem can either be left up to each practitioner or the practitioner can choose from a window of pertinent

P.75

options. In most cases, the latter makes the most sense. It provides some structure but permits an individualized approach.

The clinical glide path approach combines making a prognosis about a patient's expected course with the collection of systematic information on salient parameters. The prognosis can be generated intuitively or derived from statistical data. Unfortunately, we do not yet have systematic data collection from which to generate such statistics.

Implementing a glide path system has to rely on clinical judgment about prognosis. This judgment can be made by each clinician, or one can rely on the systematic judgment of experts. Initially the value of the glide path system will come less from assessing the rate of achieving the expected path than from the systematic attention to patients' courses. Rendering a great deal of complex information more coherently and setting expectations should have a positive effect on chronic care.

It is important to distinguish the clinical glide path approach from clinical pathways. The latter specify an expected course with specific milestones and dictates what care should be provided at specific junctures. This approach works well in very predictable situations such as postoperative recovery and even some instances of rehabilitation, but most of chronic care management is not as predictable. The glide path method specifies what data should be collected, not what actions should be taken. Its underlying premise holds that when clinicians can be aided in focusing their attention on a patient's salient parameters, they will be able to manage the chronic problems better.

DISEASE MANAGEMENT

Focusing attention on the management of specific problems has become a consistent theme in efforts to improve the management of chronic illness. Disease management is most commonly used by health plans, which use the available administrative data from encounters, drug records, and lab tests to identify all enrollees with a given condition. Protocols can then be applied to look for errors of both omission and commission. In some cases, potential complications can be flagged and checks built in to try to avoid untoward events such as drug interactions.

A more active approach to disease management uses case managers for patients who are determined to need special attention, either because they have a diagnosis that suggests high risk of subsequent use or their history indicates problems in controlling their disease(s). These case managers work with the patients to be sure that they understand their regimens. They encourage the patients to raise any questions early. They telephonically monitor the course of the illness using parameters like those described above. They may make home visits to ascertain how the patients are doing and to ensure that they can function effectively in their natural habitats. The positive reports from trials of this approach have encouraged many replications.

P.76

Another variation on disease management is being practiced in a few managed care organizations. Here patients with a given disease (sometimes a more heterogeneous cluster of patients is assembled) are brought together for periodic sessions that include health education and group support, as well as individual clinical attention. It has proven more efficient to use groups in this way. The same sessions can draw upon specialists to see problematic cases more efficiently.

Particularly in the context of managed care, there is a strong incentive to try to identify high-risk patients in order to attend to them before they develop into high-cost cases. Various predictive models have been developed to identify such cases. One widely used model is the Pra (probability of repeat admissions).

This tool uses an eight-item questionnaire to flag older patients who are most likely to have two or more hospital admissions in the next several years (Boult et al., 1993; Pacala et al., 1995). A modification of this method has been developed to use administrative data bases as well. A similar approach is being developed to identify those at high risk for needing long-term care. Once these patients have been targeted, an intervention is needed to change the predicted course. The Pra model does not specify what actions should be taken; it was initially developed as a method for identifying those in need of a comprehensive geriatric examination.

Other efforts have sought to target high-risk groups. An analysis of the Medicare Current Beneficiary Survey identified a model that could identify older persons at risk of death or functional decline (Saliba et al., 2001). Another index can identify older adults who have an increased risk of death 1 year after a hospitalization (Walter et al., 2001).

Interventions have also been developed to address those at highest risk. A meta-analysis of geriatric assessment declared it a substantial boon to care, because it was associated reduced mortality and improved function (Stuck et al., 1993). Another meta-analysis suggests that home visits can prevent nursing home admissions and functional decline (Stuck et al., 2002).

Function has proven to be an important predictive risk factor for both subsequent use of expensive services and or outcomes in general. Poor functional status in hospital patients predicts later mortality over and above the effects of burden-of-illness measures.

BENCHMARKS

The clinical glide path system is closely related to another approach that combines a set of prognoses and the criteria for predicted progress. The latter are termed benchmarks. Ideally, each physician would evaluate each patient as a unique combination of problems and potential.

However, the time involved may prove prohibitive. An alternative is to use a previously conceived set of basic benchmarks geared to the common problems of geriatrics. The generation of a problem list then leads to the implementation of the

P.77

appropriate benchmarks. These can be modified to fit the more specific characteristics of a given patient. A great advantage of this approach lies in its capacity to enhance delegation. The benchmarks provide a structure to guide ongoing information gathering. They target the pertinent information and can even offer general guidelines about what steps to take when the observed results deviate from those expected. (These actions are usually couched in terms of varying stages of urgency for seeking help.) The benchmark system does not preclude gathering additional data or recognizing the development of a new problem.

One setting ideally suited to such an approach is the nursing home. Because the system guides the collection of information and relates each item to one or more specific problems, it has an educational role as well as a clinical one. Nursing home staff who may feel isolated from the infrequently seen physician now have a more effective means of communication. They know what information is being sought and why. They have a format for recording those data, which offers desired information. The benchmark system was, in fact, developed for use in the nursing home and has been used effectively in that setting. Because the general level of literacy among nursing home aides is quite low, the system was specifically designed to require minimal narrative by using symbols in a flowchart format.

Table 4-1 presents a sample set of benchmark criteria for hip fracture. The two scales referred to independence and behavior were specially developed for the project but could be replaced by measures of activities of daily living and a standardized mood scale such as that for depression. Other clinicians may choose to add, delete, or modify items to fit their pattern of practice. Figure 4-3 offers a hypothetical chart for recording data pertinent to this problem. Specific narrative notes are made only when an event warrants them. The flowchart format readily shows when a change in pattern occurs and intervention is indicated. This basic approach has been taken a step farther. Extended care pathways have been developed, largely in the arena of nursing, whereby specific outcomes are defined and timelines laid out by which they should be met. Charting is done for exceptions. Treatments and next steps are usually indicated.

TABLE 4-1 EXAMPLE OF BENCHMARK INSTRUCTIONS: HIP FRACTURE

OBSERVATION	FREQUENCY	BENCHMARK	ACTION
1. Independence scale score	Every 2 weeks	Decreased score or no change within 6 weeks after therapy is begun	Inform physician
2. Behavioral scale score	Every 2 weeks	Decreased score or no change within 6 weeks after therapy is begun	Inform physician
3. Soreness in either calf (use: R5 right soreness, L5 left soreness)	Daily; check by pushing foot up with knee extended	Present Notify physician	Place patient on bed rest
4. Temperature ( F), pain, malaise	Increased hip pain, calf pain in afternoon	Above 100.8 F (38.2 C) orally for 8-16 h	Notify physician Take temperature every 4-8 h if elevated
5. Contracture or stiff joint, especially foot drop, hip flexion, knee flexion, hip abducted or turned out (+ or -)	Daily	Present	Continue to position properly and give range of motion to tolerance; inform physician; chart location of contracture
6. Distance walked at one time (in yards)	Daily	Decrease over 3 days or no improvement within 1 week after therapy is begun	Inform physician
7. Pain in hip (+ for pain)	Afternoon	No relief by medication after 24 h of administration	Notify physician
Source: From Pepper GA, Jorgonson LB, Kane RL, et al: Problem-Oriented Process: Nurse's Manual. Salt Lake City, University of Utah, 1972, with permission.

FIGURE 4-3 Hypothetical example of benchmark chart. See nurse's notes in chart, as some appropriate for charting this observation concerning this observation on this day.

The benchmark system was originally developed for use with a nurse practitioner as the person giving primary care under supervision of physicians. Nurse practitioners have proved themselves very effective in such roles for nursing home patients, and the care they provide is now covered by Medicare Part B.

Neither the benchmark system nor the extended care pathways fit easily into the current system of regulations for physician attendance to nursing home patients. The present rules are designed to achieve at least a minimal frequency of contact with these patients (usually once a month). Unfortunately, the result is too often superficial attention to the patient's chart. The doctor's signature is the ticket to continued coverage under Medicaid. Particularly when the same system discourages frequent visits, the care of nursing home patients is neglected.

Nursing home care has never attracted a great deal of physician enthusiasm, but this need not continue to be the case. If we can implement a new form of

P.78

P.79

P.80

record keeping that provides better information to staff and demands better performance from them, we should see an improvement in morale and hence a more attractive atmosphere in which to practice.

MINIMUM DATA SET FOR NURSING HOMES

The Omnibus Budget Reconciliation Act of 1987 (OBRA 1987) produced many changes in the way nursing homes were regulated. Perhaps none was as influential as the requirement that all nursing home residents covered by federal funds be assessed regularly using a standardized form, the Minimum Data Set for Nursing Home Resident Assessment and Care Screening (MDS). This information is designed to be completed by a nurse, but it draws on data from a number of disciplines.

The MDS summarizes a number of facets about each resident, including functional levels, cognitive and behavioral problems, special care needs, skin condition, nutritional status, and psychosocial well-being (the last not very well).

In addition to serving as a basic data set, problems identified trigger more detailed required documentation, called Resident Assessment Protocols (RAPs), in 18 areas. Table 4-2 lists the RAPs.

TABLE 4-2 RESIDENT ASSESSMENT PROTOCOL (RAP) TOPICS

Delirium	Cognitive loss/dementia
Visual function	Communication
ADL functional/rehabilitation potential	Urinary incontinence
Psychosocial well-being	Mood state
Behavior problem	Activities
Falls	Nutritional status
Feeding tubes	Dehydration/fluid maintenance
Dental care	Pressure ulcers
Psychotropic drug use	Physical restraints
ADL = activity of daily living.

The MDS is intended to provide a basis for developing better plans of care and for assessing the changes in functional levels over time. A copy of the MDS is shown in the Appendix.

The MDS can also prove a useful tool for physicians. It is a compact source of information about various aspects of each nursing home resident. If the pertinent parameters for goals determined to be achieved in the care plan were systematically charted in a flow sheet, it would be possible to see progress at a glance or to recognize the need for a change in the plan of care. Physicians can

P.81

play a key role in helping nursing home staff to see how such information can be used to improve care, not just to meet external mandates for better documentation.

However, several important shortcomings of the MDS must be acknowledged. The MDS was designed to be a means of recording judgments. These judgments inevitably pass through several hands. The persons with the most direct opportunity to observe behavior are the nurses aides, who then communicate their observations to the nurses completing the forms. The overall reliance on observations means that, in effect, all nursing home residents are being assessed as though they were cognitively impaired. This limitation is especially severe, because the MDS purports to measure critical elements of quality of life. Assuming that one can truly infer another person's emotional state, the degree to which they are engaged in meaningful activities or whether they have real social relationships seems like an act of hubris. Even using observations to determine a person's cognitive capacity seems to require heroic assumptions. It may be possible to detect extremes of behavior, but no one would want to argue that such an approach is the best way to assess many of these critical domains. Nonetheless, the MDS does not use specific questions put to those patients who can respond. Work is currently underway to test methods to assess quality of life among nursing home residents. Many, including those who are cognitively impaired can be interviewed directly. The challenge comes in how to gather information on those who cannot respond reliably. Proxy use works poorly at the individual level, although the mean values correspond well with those obtained from residents and thus can be used to assess the performance of nursing home as a whole.

The MDS has also been used as the basis for assessing the quality of nursing home care. A set of quality indicators has been developed on the basis of MDS information. These are now being nationally normed, although more work is needed on risk adjustment to allow for valid comparisons among nursing homes that may have quite different case mixes.

OASIS

The federal government has also prescribed a data system for home health care. OASIS (Outcome and Assessment Information Set) is intended to play much the same role in this venue that MDS does in the nursing home, providing both a consistent information base for quality assessment and serving as the basis for better care planning.

ROLE OF OUTCOMES IN ASSURING QUALITY OF LONG-TERM CARE

Quality of care remains a critical, if elusive, goal for long-term care. As we consider steps for resource allocation, we might first address the question of

P.82

whether we are spending our current funds most wisely. There is at once a growing demand for more creativity and more accountability in long-term care. It may be possible to reduce the regulatory burden, increase the meaningful accountability, and make the incentives within the system more rational. Progress in long-term care and chronic care will require more innovation and creativity, but it will also require accountability. Outcomes monitoring (and ultimately outcomes-based rewards) allows both to coexist.

Before we can talk about how to package care or how to buy it cheaper, we need a better understanding of what we are really buying. One hears more and more about the value of shifting attention from the process of care to the actual outcomes achieved in acute care. These arguments apply at least as strongly to long-term care.

Two basic concepts must be kept in mind when discussing outcomes.

The term outcomes is used to mean the relationship between achieved and expected.
Because outcomes relies on probabilities, it is inappropriate to base assessments of outcomes on an individual case. Outcomes are always judged on the basis of group data.

Table 4-3 summarizes the reasons for looking toward outcomes as the way to assess and assure quality.

TABLE 4-3 RATIONALE FOR USING OUTCOMES

Outcomes encourage creativity by avoiding domination by current
professional orthodoxies or powerful constituencies.
Outcomes permit flexibility in the modality of care.
Outcomes permit comparisons of efficacy across modalities of care.
Outcomes permit more flexible responses to different levels of
performance, and thus avoid the all-or-none difficulties of many
sanctions. At the same time, outcomes have some limitations.
Outcomes necessitate a single point of accountability; all the actors
facility operators, agencies, staff, physicians, patients, and family
contribute to them. Under this approach the role of the provider includes
motivating others.
Outcomes are largely influenced by the patient's status at the beginning
of treatment. The easiest and most direct way to address this issue is to
think of the relationship between achieved and expected outcomes as the
measure of success.
Outcomes must also take cognizance of case mix. Predicting outcomes
necessitates information about disease characteristics (e.g., diagnosis,
severity, and comorbidity) and patient characteristics (e.g., demographic
factors, prior history, and social support).

Clinicians frequently balk at being judged on the basis of outcomes. This discomfort can be traced to several issues.

Virtually all of clinical training addresses the process of care. Clinicians are schooled in what to do for whom. They reasonably believe, therefore, that if they do the right thing well, they have provided a quality service. They do not like to discuss clusters of patients, preferring to review their care one patient at a time.
Many factors can affect the outcomes of care that are out of the clinicians' control. They have difficulty with the concept of probability and prefer to either be responsible or not.
Outcomes are by their nature post hoc. Often, a long period can elapse between the time of an action and the report of its success. It is thus too late to intervene in that case.
Outcomes indicate a problem but offer no solution. Outcomes do not often point to specific actions that must be taken to correct the problems.

Hence, introducing outcomes, however rational, has not been easy. Making clinicians comfortable with an outcomes philosophy will require substantial training and new incentives. Physicians need to be trained to think in terms of both condition-specific and generic outcomes. They need access to data systems that can display the outcomes of their care for clinically relevant groups of patients under their care and compare them with what are reasonable outcomes for comparable patients receiving good care. Table 4-4 summarizes the key issues in outcomes measurement and its applications.

TABLE 4-4 OUTCOMES MEASUREMENT ISSUES

ISSUE	COMMENTS
Need outcome measures that are both clinically meaningful and psychometrically sound.	Use combination of condition-specific and generic measures. Usually better to adapt extant measures than to develop measures de novo.
Outcomes are always post hoc.	Expand outcomes information systems to include data on risk factors. These data should be useful in guiding clinicians to collect information that will identify potential problems. Use these data to create risk warnings to flag high-risk cases.
Every physician has all the tough cases.	Need to include a wide variety of case-mix adjusters for severity and comorbidity. Ask clinicians in advance to identify potential risk adjusters. Collect almost any item that a clinician might want to see. Test the ability of the potential risk factor to predict outcomes and discard if it has little predictive power.
Because no two clinicians see the same cases, comparisons are unfair.	Use risk adjustment. Create clinically homogeneous subgroups; use risk propensities (groups of patients with same a priori likelihood of developing the outcome).
Cannot control for selection bias; patients may receive different treatments because of subtle differences.	Adjust for all clinically identifiable differences. Use statistical methods (e.g., instrumental variables) to adjust for unmeasured differences.

P.83

Outcomes should be used as the basis for quality assurance in long-term care. The outcome approach can be used in several ways.

As reflected in the OBRA 1987 regulations (which, in turn, were stimulated by the Institute of Medicine's 1986 report), there is already growing national interest in increasing the emphasis on outcomes in regulatory activities. Outcome measures can be substituted for most of the current structure and process measures. It is appropriate to continue regulation in areas such as life safety. Concomitant with an outcomes emphasis would be the reduction of regulatory burden. It is important to recognize, however, that it is not appropriate to dictate structure, process, and outcome at the same time. Such a policy removes all degrees of freedom and stifles creativity at the point when we want to encourage it. Under an outcome-regulated approach, providers whose patients do better than expected are rewarded and are less worried about their style of caregiving, whereas those whose patients do relatively poorly are investigated more closely.
Outcomes can be incorporated into the payment structure to link payment with effects of care. Payments, either in the form of bonuses and penalties or
P.84

P.85

as a more fundamental part of the payment structure, can be used to reward and penalize good and bad outcomes, respectively. (For example, an outcome approach might use a factor reflecting the overall achieved/expected ratio for a patient as a multiplier against the costs of care to develop a total price paid for that period of time; or one might use a similar ratio to weigh the amount of money going to a given provider from a fixed pool of dollars committed to such care.) Such an approach must be viewed carefully within the context of our present case-mix reimbursement scheme for nursing homes, because the latter indirectly rewards deterioration in function. An outcome approach to payment is compatible with a case-mix approach that is used on admission only.
An outcomes approach can be incorporated into the basic caring process. Where the information base used in assessing patients and developing care plans is structured, the emphasis on outcomes can become a proactive force to guide care. Optimally, the information used to assess outcomes will come from the clinical records and will be the same information used to guide care. Using available computer technology, it is now feasible to collect such data, translate them into care plans, and aggregate these data for quality assurance at minimal additional cost. The great advantage of such a scheme is its potential both to provide a better information base with which to plan care and to reinforce the creative use of such information to achieve improvements in function. Much of the current efforts going into more traditional regulatory activities might be redirected to this effort, with assessors used to validate the assessment and to focus more intense efforts on the miscreants.

We have generally good consensus on the components of outcomes, which include elements of both quality of care and quality of life; but we are less clear about how to sum them to produce composite scores. The gerontologic literature is consistent in citing the following as outcomes:

Physiologic function (e.g., blood pressure control, lack of decubiti)
Functional status (usually a measure of activities of daily living [ADL])
Pain and discomfort
Cognition (intellectual activity)
Affect (emotional activity)
Social participation (based on preferences)
Social relations (at least one person who can act as a confidant)
Satisfaction (with care and living environment)

To these must be added more global outcomes, such as death and admission to hospital.

Work is already available with nursing home residents to show that these factors can be predicted with sufficient accuracy to be used in a regulatory model. There is similar work to show that there is reasonable consensus across a variety of

P.86

constituencies about the relative weights to be placed on them for different kinds of patients (e.g., different levels of physical and cognitive function at baseline).

The outcomes approach offers significant assistance with a recurrent problem in regulation the development of standards. This approach may avoid many of these difficulties by relying on empirical standards. Rather than arguing about what is a reasonable expectation, the standard can be empirically determined. Expectations can be derived from the actual outcomes associated with real care given by those felt to represent a reasonable level of practice. This could include the entire field or a designated subset. Under this arrangement, providers would be comparing their achievements to each other's past records, with the possibility that everyone can do better.

TECHNOLOGY FOR QUALITY IMPROVEMENT

Ideally, one would like to see a measurement approach that

Can cover the spectrum of performance
Is easy and rapid to administer
Is sensitive to meaningful change in performance
Is stable within the same patient over time
Performs consistently in different hands
Cannot be manipulated to meet the needs of either the provider or the patient

The solution to this challenge is to create an assessment approach that incorporates the features designed to maximize these elements.

To cover the broad spectrum sought and still be relatively quickly administered, an instrument should have multiple branch points. These permit the user to focus on the area along the continuum where the patient is most likely to function and to expand that part of the scale to measure meaningful levels of performance. Branching can also ensure that the assessment is comprehensive but not burdensome.

By using key questions to screen an area, interviewers can ascertain whether to obtain more detailed information in each relevant domain. Where the initial response is negative, they can go on to the next branch point. Reliability is more likely to be achieved when the items are expressed in a standardized fashion tied closely to explicit behaviors. Whenever possible, performance is preferred over reports of behavior.

One cannot expect to totally avoid the gaming of an assessment. If the patient knows that poor performance is needed to ensure eligibility, he or she may be motivated to achieve the requisite low level. One can use some test of response bias, such as measures of social desirability, but they will not prevent gaming the system or detect all cheating.

P.87

COMPUTER TECHNOLOGY

Clinical medicine seems headed inevitably toward electronic medical records. This step could represent a major advance in the care of older people, if the opportunity is properly harnessed. Simply reproducing the current unstructured information set in a more legible and transmissible format will not suffice. Structured information provides the vehicle for assuring a more systematic evaluation and follow-up of cases. By distinguishing between missing and normal values, it can provide the structure to focus clinicians' attention on salient items.

Computer technology can dramatically reduce redundancy. Properly mobilized, computers can provide the structure needed to assure a comprehensive assessment with no duplication of effort. Because they are interactive, they can carry out much of the desired branching and can even use simple algorithms to clarify areas of ambiguity and retest areas where some unreliability is suspected. Similar algorithms can look for inconsistency to screen for cheating.

Data stored on computers can be aggregated to display performance across patients by provider (e.g., physician, nursing home, or agency). Data on a patient can be traced across time to look at changes in function and, in turn, can be aggregated.

The next important step in the progression is to move the focus from a single point of care to the linking of related elements of care. In an ideal system, patient information would be linked to permit tracing changes in status for that individual as he or she moves from one treatment modality to another. Thus, hospital admission and discharge information, long-term care information, and primary care information would be merged into a common, computer-linked record, which allows one to trace the patient's movements and status.

Finally, it would be desirable to have data on the process of care as well as the outcomes. This combination would permit analyses of what elements of care made a difference for which patients.

Such an approach to assuring quality is within our grasp if we are prepared to invest in data systems and to commit ourselves to collecting standardized information. It necessitates a shift in some of our fundamental paradigms from thinking about whether we did the right thing to deciding if it made any difference after all.

Two basic changes in thinking are necessary in order to establish an outcome-based philosophy, both of which are difficult for clinicians:

Thinking in the aggregate, using averages instead of examining each case; outcomes do not work well for individual cases because there is always a chance that something will go wrong, and life does not provide a control group.
Attributing responsibility to the whole enterprise rather than placing blame on an individual; a pattern of poor outcomes will mandate closer inspection of the process of care, but outcomes per se are a collective responsibility.

P.88

Computerized records greatly facilitate the task of monitoring the outcomes of care. Ideally, such a record system should be proactive, directing the collection of clinical information to encourage adequate coverage of relevant material. Long-term care is actually ahead of acute care in this regard, with the federal requirement for computerized versions of the MDS. Unfortunately, most of the systems in use are simply inputting mechanisms. They do not begin to tap the real potential of a computerized information system. Because long-term care depends heavily on poorly educated personnel for so much of its core services, the availability of an information support system, which can provide feedback and direction, is especially appropriate.

The computer can provide both the flexibility and the brevity sought by using branching logic to expand a category when there is reason to explore it more thoroughly. It can avoid duplication by displaying data already collected by others while still permitting the second observer to correct and challenge earlier entries. More important, it can display information to show change over time, thus permitting both the regulators and the caregivers to look at the effects of care.

Once the data are in electronic form, they are easily transmitted and manipulated. It is not hard to envision a large set of data derived from these systematic observations that would permit calculations of expected courses for different types of long-term-care patients. These could then be compared to individual patient's courses to assess the potential impact of care on outcomes.

The computer's ability to compare observed and expected outcomes extends beyond its role as a regulatory device. It could be a major source of assistance to caregivers. One of the great frustrations in long-term care, especially in the trenches, is the difficulty in sensing whether the caregiver is making a difference. Because so many patients enter care when they are already declining, the benefits of care are often best expressed as a slowing of that decline curve. Without some measure of expected course in the absence of good care, those who render care daily may not appreciate how much they are accomplishing and thereby may forgo one of the important rewards of their labors.

To display information about the change in patient condition over time, a simple task for a computer, will assist the long-term caregiver to think more in terms of the overall picture, rather than a series of separate snapshots in time. Given the computer's ability to translate data into graphics, it is a simple procedure to develop pictorial representations of the changes occurring for a given patient or group of patients and to contrast those with what might be reasonably expected.

Again the effort is directed toward changing perceptions about older persons, especially those in long-term care. For too long, long-term care has worked in a negative spiral a self-fulfilling prophecy that expected patients to deteriorate served to discourage both care providers and patients. Such an attitude is hardly likely to attract the best and the brightest in any of the health professions. As noted

P.89

earlier in this chapter, nursing home patients are among the most responsive to almost any form of intervention. Any information system that can reinforce a prospective view of long-term care, especially one that can display patient progress, represents an important adjunct to such care.

TERMINAL CARE

The physician's concern with the patient's functioning continues throughout the course of the chronic disease. Elderly patients will die. In many cases, death is not a reflection of medical failure. The approach to the dying patient will often raise difficult dilemmas. No simple answers suffice. Perhaps the best advice is not to take on the whole burden. Too often the dying patient is treated as an object. Ignored and isolated, the patient may be discussed in the third person.

Physicians must come to terms with death if they are to treat elderly patients. Often the patients are more comfortable with the subject than are their physicians. Fleeing from the dying patient is inexcusable. Dying patients need their doctors. At a very basic level, everything should be done to keep the patient as comfortable as possible. One simple step is to identify the pattern of discomforting symptoms and arrange the dosage schedule of palliatives to prevent rather than respond to the symptoms.

Patients need an opportunity to talk about their death. Not everyone will take advantage of that chance, but a surprising number will respond to a genuine offer made without time pressure. Such discussions are not conducted on the run. Often several invitations accompanied by appropriate behavior (e.g., sitting down at the bedside) are necessary.

Some physicians are unable to confront this aspect of practice. For them, the challenge is to recognize their own behavior and get appropriate help. Such help is available at various levels: help for the physician and for the patient. Groups and therapy are readily available to assist doctors to deal with their feelings. Patients of doctors who fear death need the help of other caregivers. Often other professionals nurses, social workers who are working with these patients already can play the lead role in helping them work through their feelings. But the active intervention of another caregiver is not justification to ignore the patient.

The rise of the hospice movement has created a growing cadre of persons and settings to help with the dying patient. The lessons coming from this experience suggest that much can be done to facilitate this stage of life, although the formal studies done to evaluate hospice care do not show dramatic benefits.

Patients should be encouraged to be as active as possible and as interactive as they wish. Even more than in other aspects of care, the unique condition of the dying patient necessitates that the physician be prepared to listen carefully to the patient and to share in decision making about how and when to do things.

P.90

SUMMARY

In many respects, geriatrics is the epitome of chronic disease care. New paradigms are needed, which recognize the changing role of patients their own care, the need to think differently about the pay-off horizons for investments in care, and how to track the course of disease to identify when intervention is needed. With geriatrics and chronic disease in general the benefits of good care may be hard to discern, because they represent a slowing of decline. This effect is invisible unless there is some basis for forming an expected clinical course against which to compare the actual course.

Physicians caring for older patients need to think in prospective terms. They will enjoy their practices more if they can learn to set reasonable goals for patients, to record progress toward these goals, and to use the failure to achieve progress as an important clinical sign of the need for reevaluation.

References

Beck A, Scott J, Williams P, et al: A randomized trial of group outpatient visits for chronically ill older HMO members: the cooperative health care clinic. J Am Geriatr Soc 45:543 549, 1997.

Boult C, Dowd B, McCaffrey D, et al: Screening elders for risk of hospital admission. J Am Geriatr Soc 41:811 817, 1993.

Clark F, Azen SP, Zemke R, et al: Occupational therapy for independent-living older adults: a randomized controlled trial. JAMA 278:1321 1326, 1997.

Fiatarone MA, O'Neil EF, Ryan ND, et al: Exercise training and nutritional supplementation for physical frailty in very elderly people. N Engl J Med 330:1769 1775, 1994.

Kane RL: The chronic care paradox. J Aging Soc Policy 11(2/3):107 114, 2000.

Pacala JT, Boult C, Boult L: Predictive validity of a questionnaire that identifies older persons at risk for hospital admission. J Am Geriatr Soc 43:374 377, 1995.

Saliba D, Elliott M, Rubenstein LZ, et al: The vulnerable elders survey: a tool for identifying vulnerable older people in the community. J Am Geriatr Soc 49:1691 1699, 2001.

Stuck AE, Egger M, Hammer A, Minder CE, Beck JC: Home visits to prevent nursing home admission and functional decline in elderly people: systematic review and meta-regression analysis. JAMA 287(8):1022 1028, 2002.

Stuck AE, Siu AL, Wieland GD, et al: Comprehensive geriatric assessment: a meta-analysis of controlled trials. Lancet 342:1032 1036, 1993.

Walter LC, Brand RJ, Counsell SR, et al: Development and validation of a prognostic index for 1-year mortality in older adults after hospitalization. JAMA 285:2987 2994, 2001.