The physician imperative: define, measure, and improve health care quality

The physician imperative: define, measure, and improve health care quality – Part 2: Value-Based Health Care

William C. Mohlenbrock

THE INITIAL SEGMENT OF THIS ARTICLE ASSERTED that physicians must support the movement to value-based health care as it is in the best interest of both patients and physicians. Increasing numbers of health care purchasers are eager to move to value-based contracting, which is quality for reasonable prices. Physicians are the only constituency capable of initiating a movement to replace price-based contracting with value-based health care delivery.

Several action steps were enumerated whereby physicians can regain clinical control of their practices through value-based health care delivery. The steps include forming a physician organization which often needs to be independent of the hospital for outpatient contracting. Independence may require a capital partner with an understanding of the primacy for physicians clinical autonomy. The third critical success factor involves the acquisition and effective use of clinical information which accurately risk-adjusts and integrates both inpatient and outpatient data for all episodes of care. These clinical data and how they are used to define, measure, and improve all aspects of health care quality is the physicians’ Imperative and the subject of this article.

The dimensions of health care quality

Patients, providers, and purchasers share common goals in their desire to receive, provide, and purchase the highest possible quality health care at cost efficient prices. Virtually all providers mission statements and marketing literature proclaim their organizations to be delivering the highest quality of care. However, they invariably fail to specify the measures of quality which purchasers and patients could interpret and use for provider selection.

Health care quality can be assessed from three perspectives: that of the provider, the individual patient, and the overall patient population. The accepted metrics of health quality fall into four general categories which are reflected in JCAHO’s performance measures: clinical, financial/administrative, health status, and patient satisfaction, But there is another equally important outcome which is the measure of process improvement. Health care outcomes are enhanced by the continuous improvement of care processes. The relationship of improved processes to quality outcomes is well documented in other industries. Process improvements are measured using Reductions In Variation (RIV). According to quality experts, RIV is the first and probably the most critical outcome measure. (1)

This article quantifies both process and outcomes values, but does not address providers’ administrative or structural measures of care delivery systems.

1. Quality definitions

No single definition of health care quality is perfect, but certainly no product or service can be improved if it cannot be defined and measured. The Office of Technology Assessment (OTA), JCAHO, and most other organizations use definitions that acknowledge the limits of medical knowledge and the necessity of measuring patients’ expectations of improved health status and satisfaction with their care. The most commonly agreed upon element of quality definitions Is the positive relationship between care processes and outcomes–” ..the degree to which the process of care increases the probability of desired patient outcomes… ” (2)

2. Process performance measure-Reduction In Variation (RIV)

At the heart of a Continuous (Clinical) Quality Improvement (CQI) strategy are clinically reliable data that demonstrate variations in observed outcomes within homogeneous patient cohorts. Reductions In Variation (RIV) over time are a direct result of providers reasoning together to improve their consistency and the quality of their patients’ outcomes. Variation is a quantifiable metric as to how consistent an individual or group utilizes treatments or resources for patients with similar illness characteristics. Health care systems and individual physicians should document their Reductions In Variation as a routine part of outcomes reporting.

3. Outcomes performance measures

Clinical Performance Measures: These are generally identified as risk-adjusted clinical indicators, such as mortality, morbidity, infection, readmission rates, etc. Population-based metrics, such as a pediatric population’s immunization rates, fall in this group. The National Committee for Quality Assurance’s Health Plan and Employer Data and Information Set (HEDIS) relies heavily on these types of measures.

Financial (Resource Consumption) Measures: Resource consumption is a sensitive quality measure. Quality and cost efficiency are synonymous because over-utilization of resources can lead to complications from tests or treatments with minimal or no indications. Medical and surgical complications ruin lives and often inflate the cost of care. “Doing it right the first time.” with the appropriate resources, in the appropriate care setting, should be the goal.

Health Status and Patient Satisfaction Measures: The ultimate judge of any health care encounter should be the patient and his or her family, as measured by their perception of improved health status and satisfaction with the care rendered by the physicians, hospital, and health plan. Tools for such measurements are: SF [12/36.sup.i]. [TyPE.sup.ii] surveys, and patient satisfaction surveys like those used to satisfy HEDIS requirements.

4. Quality measured across the continuum

Most quality assessments have been confined to the inpatient setting because of data availability. These limited assessments are no longer sufficient and the limitation is no longer necessary. Providers are increasingly at risk for an entire patient population across the continuum of care. They must manage care in physicians’ offices, outpatient diagnostic and treatment facilities, long-term facilities, including home health, and all others. The following data will be identified: inpatient, physicians’ offices including emergency departments: day episodes, such as outpatient diagnostics and surgeries; therapeutic series, such as renal dialysis; home health and others. The physician specific data presented in the article are from office practices. though similar inpatient and other outpatient information is readily available as well.

Reliable clinical information

Clinically reliable information is fundamental to both quality improvement and value-based health care purchasing. The information must also be accessible to and understood by all parties.

1. Risk-adjustment

A patient’s overall risk is predicated primarily on his or her age, gender and the pathophysiologic processes associated with each disease state (illness). These illness interactions across multiple organ systems in an individual describe the “risk” or “complexity” of a patient. In the particular risk-adjustment methodology used in the examples herein, each patient’s diagnoses are collected from all episodes and assigned an intra-illness, “Clinical Complexity” level. This risk level Is based on the individual patient-level characteristics. The accumulation of all patients’ complexities and the number of patients describe the patient mix of an individual provider or care system and is termed “Aggregate Clinical Complexity”.

In order for physicians and hospitals to reduce their care variations, it is imperative to accurately quantify the risks that each patient brings to the patient-physician encounter. Software tools have been developed for estimating expected outcomes based upon readily available data that are fully adjusted for patient characteristics. Outcomes must be estimated for: risk of dying (mortality); risk of staying in the hospital (length of stay); risk of consuming resources (Relative Value Units, costs, charges); and the risk of having poor or improved health status and patient satisfaction.

A reliable risk-adjustment methodology must encompass all quality measures in every care setting and for all episodes. Equally important, the estimates must be based on patient level characteristics and not on parameters significantly influenced by physician discretion. Provider-centered and patient self-assessed outcomes need to be integrated in a decision support tool so each outcome can be correlated one with another for appropriate interpretation.

The particular risk-adjustment tool used in these examples uses linear modeling to account for each patient’s co-morbidities and complications without first assigning the case to one of an arbitrary number of severity levels, Linear modeling allows the system to distinguish cases with subtle, yet important, clinical differences and use this information to minimize unexplained variability. All of the following factors may be used by the algorithm to estimate outcomes: age, sex, coded diagnoses, and other diagnostic history (where available). Outcomes may be estimated in several ways: by the entire continuum of care settings, across the outpatient continuum of care, or by individual episode of care (such as inpatient stay or day encounter) within illnesses.

2. Creating manageable patient cohorts

Data which assesses physicians’ quality or efficiencies are only interpretable if they are risk-adjusted and, equally important, the patients’ illnesses can be disassociated for independent analysis. Figure 1 illustrates two years of an actual patient’s health care history across the continuum. The outcomes prediction algorithm assigned the patient the following illnesses based on claims data: asthma, prostate cancer, and two occurrences of fractured radius.

Asthma and prostate cancer are both chronic illnesses and, therefore, the patient is assumed to have these illnesses indefinitely, while a fracture of the radius is an acute Illness with a treatment duration of about six months. In Figure 1, illness occurrences are shown as horizontal boxes that extend across the full two-year health care period. Two smaller horizontal boxes represent the occurrences of the two radial fractures. Episodes have been assigned to the appropriate occurrence. For instance, the patient’s asthma required a total of nine office visits and an inpatient stay. For the patient’s first fractured radius, the figure shows one day encounter for X-rays and casting and three follow-up office visits,

This advanced modeling technique was developed to overcome the problems associated with outpatient grouper technologies. Such grouping would classify this patient into a single illness group, losing the ability to analyze each of the complicating conditions. Rather than calculating a single estimate based on a patient’s primary illness, unique estimates should also be made for each of the complicating illness occurrences.

An estimate for all of the patient’s care may then be calculated by accumulating the estimates for each of the patient’s illness occurrences. These additional estimates are important because they allow an organization to drill down into the data to answer specific questions. such as the number of resources used for each illness.

3. Patient self-assessed outcomes

Patient self-assessed outcomes consist of health and functional status, as well as patient satisfaction. These are sensitive evaluations of a health care system’s long-term benefits. Patients’ satisfaction with their care givers, treatment facilities, and access to care must be reliably and repeatedly monitored to evaluate a community’s overall health care quality. Various assessment tools have been developed for health status and any may be used. This example uses the SF-12 which has mental and physical component scores.

The particular tool developed for this outcomes management strategy uses automated telephone surveys that last only a few minutes. Patients’ response rates average above 50 percent and are retrieved and automatically collated at a fraction of the cost typically incurred for mailed questionnaires. The results are loaded into a decision support software tool for correlation with all other outcome metrics.

A four quadrant graph is used to demonstrate the results from a Southern California clinic’s SF-12 (iii) survey at three months post-op post vaginal hysterectomies.

Each letter represents one physician with five or more cases. Scores above the horizontal line are better than the group’s norm for Physical Component Scores (PCS). To the right of the vertical axis are scores which are better than the group’s norm for Mental Component Scores (MCS).

Note that the majority of the physicians’ patients’ scores fall into the right lower quadrant indicating a better than average MCS but less than average PCS. These gynecologists concluded that these patients were secure with their decisions to have a hysterectomy, but at 90 days were not back to the norm for Physical Component Scores. The physicians thought this was reasonable but additional polls were scheduled for six and 12 months post-op and the doctors plan to make any necessary decision or treatment corrections if warranted by the data.

4. Provider specific outcomes

Process Performance Measures–Reduction In Variation (RIV): Clinically similar patients, managed in the same or even different health care systems, should theoretically be diagnosed and treated with similar quantities and types of resources. That is not the case, however, as Wennberg (3) and others have demonstrated. Clinical and financial data usually reveal that providers are inconsistent even within their own practices and Inter-physician variations are very large, even within the same hospital. (4) Figure 3 demonstrates resource consumption variations at the illness level.

In clinical terms, variation is a measure of providers’ consistency in the use of resources and other decisions, such as if and when to deploy surgery, admit, or discharge patients. The majority of clinical, rate-based indicators measure conditions that should not occur during the course of treatment, such as morbidity. Reduction In Variation is a different metric. It Is a positive quality indicator measured in statistical terms. Naturally, the goal is not to simply reduce variation but to continuously improve the targeted outcome over time. The key to process improvement is reliable, risk-adjusted information that identifies two specific attributes of the outcome that are selected for improvement. The first attribute is the observed variation of the outcome’s metric and the second is the patient cohort with the best-demonstrated outcomes. These superior outcomes serve as the target of the improvement process.

Variation within the targeted outcome is first quantified at the beginning of the clinical process improvement activity and then again at six and/or 12 months to document any changes. The patient group with superior outcomes is identified to serve as the model for replication. Using automated techniques, the best-demonstrated care processes from these patients’ care is identified, benchmarked, and formatted into a protocol or care path. Examples of such processes are the numbers and types of lab tests, X-rays, medications, techniques, or decisions used in the best-demonstrated patient cohort. In addition to these data, input from all clinicians and administrators involved in these patients’ care should be solicited to complete the care path before implementation.

This technique of experience-based. clinical process Improvement maximizes physicians participation and outcomes improvement since the constructed care path is internally derived. Reductions in Variation (RIV) can be measured over time for all inpatient and outpatient treatment episodes and the results displayed for each illness, episode, physician, and clinical service.

In this example, the physicians’ resource utilization is both more efficient and consistent in “lower back pain” management than “asthma,” as demonstrated by the positions of the graph. These parameters will be recalculated at six or 12 months to demonstrate that variation has been reduced and the outcome (resource consumption) for each illness has improved to meet the stated goals of the clinicians.

Clinical outcomes performance: Rate-based indicators which identify the frequency of surgeries, mortalities, or potential morbidities have been used by generations of physicians and serve as important quality indicators. Clinical indicators to assess providers’ care and profile health plans have recently gained widespread use by the managed care industry and federal and state agencies through HEDIS measures. Additionally. JCAHO has launched its ORYX project for accrediting hospitals and other facilities and primarily uses these types of clinical outcomes.

Post operative pneumonia in orthopaedic cases is only one of the hundreds of possible indicators. Figure 4 compares the incidence of this complication at several hospitals in a community. These indicators should be accompanied by confidence intervals so physicians can be assured the indicator’s value is not a random event. Only the orthopaedic surgeons managing these cases can reliably determine if their own rates are too high or too low. The differences in this example are statistically significant and warrant study. Physicians and their hospitals must know more about their own processes, outcomes, and practices than does any managed care organization or governmental agency.

Financial (resource consumption) efficiencies: All health care systems are limited by the data which they gather. Most hospitals and virtually all physicians’ offices collect only charge data from UB-92 and HCFA 1500 forms. Though less than perfect, charge data are reasonable surrogates for resource consumption and have great utility, particularly when comparisons are made within a health system.

In order to equate efficiencies across all physicians of all specialties, It is necessary to create a means of normalizing the data as exhibited by the Resource Performance Score (RPS) in this example. Delta Resource Performance Scores (RPS) are Expected versus Actual values compared to risk-adjusted, clinically homogeneous, intra-iliness patient cohorts. Minus (-) numbers represent relative over-utilization and positive (+) numbers represent relative efficiencies compared to inter-facility, local, state, or national norms. A report of this nature gives the physician very specific information regarding his or her overall efficiencies, including a breakdown by resource consumption area, including lab tests, ER utilization, and X-rays.

Annual Total Charges are shown for the physician indicated. Average or delta charges are always followed by a “plus/minus” number representing the standard error.

Aggregate Clinical Complexity (ACC) is a percentile measure of patient risk (complexity) for all illnesses. In the example shown, the value is a measure of patient risk by physician. Patients with values less than 50 have less risk of resource utilization than the average patient in the normative data set. The ACC describes the risk of an individual clinician’s mix of patients based on an accumulation of the multiple clinical complexities of each patient.

Physician “J” noted the rather striking differences between her relatively efficient office practice (RPS) and inefficient emergency department outcomes (-RPS) in both patient groups (234 URI patients and 56 UTI patients). Several other physicians noted the same pattern which was the knowledge they needed to implement changes in call coverage and patient flow.

Reimbursing physicians on quality of care

Equitable distribution of pooled capital between primary care physicians and specialists, and surgical and medical specialties requires the ability to accurately account for physicians with high risk patients and have ready access to all metrics of quality. The physician group can then assign relative weights to quantify the measures they wish to differentially reimburse.

1. Identifying physicians with high risk patients

The data in Figure 6 clearly illustrates which physician had already been economically credentialed off the panels of three major HMOs and, therefore, lost many cases. Physician “C” was also on probation in his own physician group’s health plan because of the apparent inefficiency indicated by the high average charges per case.

After the underlying data was accurately adjusted to account for the each patient’s unique risks, the Aggregate Case Complexity (ACC) of physician G’s patients equaled 74, the highest value for any physician in the plan. Moreover, the analysis indicated that he actually had a more efficient than average Resource Performance Score (RPS) of +1. It was extremely insightful for him and his colleagues to have ready access to every patient, diagnosis, and co-morbidity, as well as having all types, numbers, and norms for all procedures, radiology tests, and medical services that were utilized.

Figure 7 details the care of a patient of Physician “G.” Note that the risk model estimated the patient’s charges to be $18,348 and the actual charges were $18,509. (The patient had been followed for 657 days in all treatment sites.) This evidence demonstrates the doctor was actually managing patients with exceptionally high risk and doing so very efficiently. Provider groups who are at financial risk cannot afford the loss of high quality, cost-efficient physicians such as Physician “G.”

2. Equitable reimbursement based on quality

When physicians share a risk pool, on what basis should the money be divided? Productivity, numbers of patients treated, and Involvement In quality assurance activities are a few criteria in use. Additionally, physicians want to be reimbursed on their ability to practice high quality, cost efficient medicine. Each physician group should collectively determine which quality metrics warrant greater reimbursement. If patients’ health status or satisfaction should be weighted higher than, for instance, low morbidity rates (clinical indicators) or efficiencies, then their individual reimbursements should reflect superior practice outcomes. A formula has been developed that is completely flexible and accomplishes these goals whether or not all outcomes are being measured. It integrates each physician’s outcomes scores and allows differential weighting of the measures, but does not penalize providers if any of the outcomes are not available.

Conclusion

The patient-physician relationship can and must be restored to preserve one of the essential features of America’s excellent health care system. This occurs when physicians use reliable information and take full responsibilities and risks for the clinical and financial outcomes of care delivery. Clinical and financial risks are predictable and manageable when the data is fully risk-adjusted. spans the entire continuum for all episodes of care, and is integrated into an outcomes management strategy.

In the past few years, managed care techniques have unquestionably controlled spiraling costs, but have impacted patients’ choice of providers and physician autonomy in ways that have caused concern to the public. To allay these concerns and produce even greater resource efficiencies, physicians must demonstrably improve health care quality and share the information. Physicians have a unique expertise, ethics, and incentives to affect the revolutionary changes necessary to move their communities to value-based health care. Purchasers and patients seek the highest quality, most cost efficient physicians who objectively demonstrate a commitment to their patients’ health and wellbeing. The task of defining, measuring, and improving all aspects of health care quality and efficiency is the physicians’ imperative, and the time is now.

[FIGURE 4 OMITTED]

FIGURE 5

PHYSICIAN “J” RESOURCE CONSUMPTION DETAIL

Total

Illness/Procedure Group # Patients Delta RPS Charges ACC

Upper Respiratory Infection 234 -11 (-12,-9) $49,352 53

Emergency Dept. Service -7 (-10,-5) $8,281

Blood Drawing -28 (-30,-26) $6,268

Microbiology Lab -26 (-28,-23) $4,658

Office Visit/Consult 28 (26,30) $4,326

Chest Radiology 8 (5,11) $3,116

Immunology Tests/Other -37 (-39,-36) $2,913

Immunization Injection -26 (-29,-24) $2,595

Hematology Lab Test -22 (-25,-20) $1,845

Urinalysis -18 (-21,-15) $1,440

Urinary Tract Infection 56 -7 (-10,-4) $26,350 47

Microbiology Lab -8 (-12,-4) $3,676

Urinalysis -9 (-13,-5) $3,140

Emergency Dept. Service -17 (-22,-13) $2,801

Urinary Surg & Procr 13 (9,18) $2,266

Office Visit/Consult 17 (12,21) $1,822

Blood Drawing -9 (-14,-4) $1,693

Pregnancy Test -27 (-31,-22) $1,019

–William C. Mohlenbrock, MD

FIGURE 6

PHYSICIAN “J” RESOURCE CONSUMPTION DETAIL

(1995-96 Data, Total Charge> $10K and 3 Patients)

Physician # Patients Total Charges Average Charge

A 34 $25,579 $752

B 31 $13,844 $447

C 16 $13,688 $856

D 30 $11,361 $379

E 32 $12,770 $399

F 21 $14,802 $705

G 4 $35,460 $8,865

H 31 $12,557 $405

I 13 $11,036 $849

William C. Mohlenbrock, MD

FIGURE 7

DIABETES MELLITUS CASE FOR PHYSICIAN “G”

14 FEB 1995 Sex: Male Age: 56

DX0 36201 DIABETIC RETINOPATHY NO cp0 67210.

DX1 36283 RETINAL EDEMA cpl 92235.

DX2 36202 PROLIF DIAB RETINOPATHY cp2 92250

DX3 25051 DMI OPHTH NT ST UNCNTRL cp3 99214.

DX4 25000 DMII WO CMP NT ST UNCNT cp4 90784.

DX5 25040 DMII RENL NT ST UNCNTRL cpS 67228.

DX6 25001 DMI WO CMP NT ST UNCNTR cp6 99245.

DX7 25041 DMI RENL NT ST UNCNTRLD cp7 99215.

DX8 25060 DMII NEURO NT ST UNCNTR cp7 99215.

DX9 3572 NEUROPATHY IN DIABETES cp7 99215.

IIO DM Diabetes Mellitus I & II IS1 SiDiabRet

II2 Retln Retinal Disorders Misc IS3 DMI

IX6 Endocal Endocardial Disease Misc IX5 ChrPhar

IX6 ExamE Eye & Vision Exam/Visit IX7 Epistaxis

IS8 SIDia Diabetic Nephropathy IS9 DMII

14 FEB 1995 Duration: 657 Days

DX0 36201 TREATMENT OF RETINAL LESION

DX1 36283 EYE EXAM WITH PHOTOS

DX2 36202 EYE EXAM WITH PHOTOS

DX3 25051 OFFICE/OUTPATIENT VISIT

DX4 25000 INJECTION (IV)

DX5 25040 TREATMENT OF RETINAL LESION

DX6 25001 OFFICE CONSULTATION

DX7 25041 OFFICE/OUTPATIENT VISIT

DX8 25060 OFFICE/OUTPATIENT VISIT

DX9 3572 OFFICE/OUTPATIENT VISIT

IIO DM Diabetic Retinopathy

II2 Retln Diabetes Mellitus Type I

IX6 Endocal Chr Pharyng/Nasopharygn

IX6 ExamE Epistaxis

IS8 SIDia Diabetes Mellitus Type II

Expected Charges: $18,348

Actual Charges: $18,509

William C. Mohlenbrock, MD

References

(1.) Juran, J. M. Juran On Planning For Quality. The Free Press, New York, NY, 1988.

(2.) Joint Commission on Accreditation of Healthcare Organizations: An Introduction to Quality Improvement in Health Care. Dept. of Publications, 1991.

(3.) Wennberg, J., Gittelsohn, A. Variation in medical care among small areas. Scientific American 1982;246:120-29.

(4.) Davis, K. Leader of the Pack: Physicians spearhead clinical efficiency efforts. Hospital Benchmarks, American Health Consultants, January 1995, Vol. 2, No. 1. 1-16.

(i.) SF-12 and SF-36, The Health Institute, New England Medical Center, 1995.

(ii.) Typology of Patient Experience, The Health Outcomes Institute.

(iii.) SF-12, The Health Institute, New England Medical Center, 1995.

William C. Mohlenbrock, MD, is the Cofounder and Vice Chairman of Jameter, Inc. in San Mateo, California. He has more than 25 years of clinical experience in the practice of orthopaedic surgery and is a member of the medical staff at Scripps Memorial Hospital in La Jolla, California. He also serves as an Assistant Clinical Professor at the University of California in San Diego. He can be reached by calling 650/349-9100 or via fax at 650/349-7839.

COPYRIGHT 1998 American College of Physician Executives

COPYRIGHT 2004 Gale Group