Advertisement
Mayo Clinic Proceedings Home

How Well Does the Patient's Burden of Illness Explain Differences in Outcome?

  • Katherine L. Kahn
    Correspondence
    Address reprint requests to Dr. K. L. Kahn, UCLA, Department of Medicine, Division of General Internal Medicine and Health Services Research, B542 Factor Building, Los Angeles, CA 90024
    Affiliations
    Department of Medicine, UCLA, Los Angeles, California
    Search for articles by this author
      As the costs of health care mount, clinicians, patients, and policy analysts increase their attempts to understand how hospitals compare with each other.
      • Hartz AJ
      • Krakauer H
      • Kuhn EM
      • Young M
      • Jacobsen SJ
      • Gay G
      • Muenz L
      • Katzoff M
      • Bailey RC
      • Rimm AA
      Hospital characteristics and mortality rates.
      • Keeler EB
      • Rubenstein LV
      • Kahn KL
      • Draper D
      • Harrison ER
      • McGinty MJ
      • Rogers WH
      • Brook RH
      Hospital characteristics and quality of care.
      Variations across hospitals and among types of hospitals in the characteristics of the patients receiving treatment, the use of services, and the style of medical care being delivered have prompted interest in comparisons of the outcomes experienced by patients after hospitalization.
      • Iezzoni LI
      • Shwartz M
      • Moskowitz MA
      • Ash AS
      • Sawitz E
      • Burnside S
      Illness severity and costs of admissions at teaching and nonteaching hospitals.
      • Park RE
      • Brook RH
      • Kosecoff J
      • Keesey J
      • Rubenstein L
      • Keeler E
      • Kahn KL
      • Rogers WH
      • Chassin MR
      Explaining variations in hospital death rates: randomness, severity of illness, quality of care.
      Although comparison of patients' quality of life after hospitalization would be ideal, easier to quantify measures, such as cost and rates of mortality and readmission, are more often used to compare individual and groups of hospitals and patients.
      In fact, data for each of these three outcome variables are available in electronic data bases in many hospitals in the United States.

      Factors That Affect Outcome.

      Outcomes reflect the burden of illness the patient has at the time of hospitalization, the quality of care the patient receives during the hospitalization, and the prevalence of in-hospital complications, which may result from both the severity of illness at admission and the quality of care.
      • Kahn KL
      • Brook RH
      • Draper D
      • Keeler EB
      • Rubenstein LV
      • Rogers WH
      • Kosecoff J
      Interpreting hospital mortality data: how can we proceed?.
      • Keeler EB
      • Kahn KL
      • Draper D
      • Sherwood MJ
      • Rubenstein LV
      • Reinisch EJ
      • Kosecoff J
      • Brook RH
      Changes in sickness at admission following the introduction of the prospective payment system.
      For comparisons of mortality rates, readmissions, and costs to be meaningful to clinicians, patients, and policymakers as a measure of the quality of care delivered to the patient during the hospital stay, such comparisons must account for the burden of illness the patient brings to the hospital.
      • Kahn KL
      • Rubenstein LV
      • Draper D
      • Kosecoff J
      • Rogers WH
      • Keeler EB
      • Brook RH
      The effects of the DRG-based prospective payment system on quality of care for hospitalized Medicare patients: an introduction to the series.
      The more valid the measure of sickness, the more meaningful the comparisons of outcomes become. The validity of systems for measuring illness is often determined on the basis of the amount of variance in outcomes that can be explained by the system.
      • Keeler EB
      • Kahn KL
      • Draper D
      • Sherwood MJ
      • Rubenstein LV
      • Reinisch EJ
      • Kosecoff J
      • Brook RH
      Changes in sickness at admission following the introduction of the prospective payment system.
      For example, the more that differences in mortality rates across hospitals can be explained by variations in the burden of sickness that the patients bring to the hospital, the more accurately the discrepancies in outcomes can be interpreted as reflecting differences in the care provided to the patients during their hospital stay.

      Assessment of Delivery of Care.

      Generally, when comparisons of outcomes adjusted for the patient's sickness are used as a method of evaluating differences in performance across hospitals, sickness is defined in terms of the burden of illness the patient brings to the hospital. If outcomes are adjusted by collectively accounting for the burden of illness the patient brings to the hospital and the burden of illness that develops as a result of in-hospital complications, the differences in outcomes that may be attributable to poor quality of care (as manifest by the occurrence of in-hospital complications) will be diluted.
      Imagine a previously healthy patient admitted to the hospital with an acute myocardial infarction who, because of poor management of fluids and arrhythmias, has decreased renal and cerebral perfusion that results in renal failure and a cerebrovascular accident. The presence of each of these unfortunate outcomes and their interaction substantially increase the probability of death for that patient. Obviously, including the diagnoses of in-hospital complications such as renal failure and stroke will improve precision in estimating the probability of death. This approach, however, will not reveal how well the hospital delivered care to the patients because those who were “well” at admission and subsequently had in-hospital complications are grouped with patients who had severe comorbid conditions at the time they were admitted to the hospital.

      Preterminal Conditions.

      Including in-hospital complications with high probabilities for death in a prediction of death will certainly improve the accuracy of the prediction. It does not, however, enhance our understanding of whether the quality of care or the burden of illness the patient brings to the hospital is the more important determinant of differences in outcomes.
      In fact, inclusion of in-hospital complications as a means to adjust outcomes can be misleading. For example, if a hospital assigns diagnostic codes for the in-hospital complications of cardiac arrest or stroke and they are used to adjust mortality rates for these diagnoses, the adjustment will show that hospital as having a lower mortality rate than if the patients did not have these complications. In addition, because cardiac arrest and stroke are such important predictors of death, the variance in mortality rate that is explained by including these preterminal events in the sickness measure will be high—when, in fact, all that is being reflected is that the patient is about to die.

      The COMPLEX System.

      In this issue of the Mayo Clinic Proceedings (pages 1140 to 1149), Naessens and associates describe the COMPLEX system as a means to adjust outcomes for the severity and complexity of the patient's illness at the time of admission. The COMPLEX system successfully increases the amount of variance in outcomes that can be explained in comparison with several other methods. Unfortunately, it is unable to distinguish sickness that is present at the time of admission from sickness that develops as an in-hospital complication. Moreover, the system is deficient in that the accuracy of discharge data—the basis for measuring sickness—varies widely across institutions. Even with the revolution in the accuracy of coding that accompanied the introduction of the Medicare prospective payment system based on diagnosis-related groups, wide variations in styles and accuracy of coding exist across hospitals. Thus, the probability of a diagnosis appearing in the discharge data set is influenced by many administrative and style factors beyond whether the patient has a particular condition.
      • Keeler EB
      • Kahn KL
      • Bentow SS
      Despite these limitations, the COMPLEX system for measuring the patient's illness described by Naessens and colleagues has many advantages. The authors cleverly approach the challenge of measuring the patient's burden of illness by taking advantage of the hospital's discharge data because the cost of using this already existent administrative data set is substantially lower than the alternative, which would involve primary data collection through interviews with the patient or clinician or abstraction of medical records. The discharge data set includes a listing of diagnoses assigned to the patient at any time during the hospital stay. Thus, it provides the opportunity of tracking the presence of comorbid diseases or clinical problems that are currently present in addition to the morbid disease or factor that prompted admission to the hospital.
      This new adaptation of the computerized Disease Staging system
      • Gonnella JS
      • Hornbrook MC
      • Louis DZ
      Staging of disease: a case-mix measurement.
      maintains clinical validity by assigning credit for involvement of a new body system only if the diagnosis does not also serve to define the stage of another coexisting disease category. Furthermore, it maintains clinical validity by minimizing the inclusion of relatively minor conditions as comorbidities. These characteristics plus ease of use and significant improvements in the amount of explained variance in outcomes demonstrate the strengths of the system.
      Naessens and colleagues have succeeded in developing a system to improve the amount of variance in outcomes that can be explained by the complexity of the patient's illness. Nevertheless, much of their progress in improving the explanation of outcomes may be attributed to their inclusion of in-hospital complications as a component of their measure of sickness.

      Suggested Enhancements.

      Because of the advantages of using existing secondary data sources to measure the patient's burden of illness, the efforts of Naessens and co-workers to improve the amount of variance in outcomes that can be explained with secondary data should not be abandoned. Shapiro and associates
      • Shapiro MF
      • Park RE
      • Keesey J
      • Brook RH
      Effect of casemix adjustment on estimated mortality differences between New York City municipal and voluntary hospitals (abstract).
      addressed the problem of the general inability to use existent secondary data for distinguishing diagnoses present at admission from those developing as in-hospital complications. They suggested categorizing each diagnosis in the administrative data on the basis of the probability that it represents either an in-hospital complication or a chronic preexisting condition. For example, they considered chronic obstructive pulmonary disease and cancer most likely to be preexisting conditions, whereas sepsis and acute tubular necrosis were thought most likely to be in-hospital complications. This approach could be applied as an amendment to the COMPLEX system described in this issue.
      Alternatively, a more ambitious program could be implemented in which the clinicians and hospital coders could collaborate to assign a “time” code to each diagnosis listed in the administrative data. For example, when renal failure or stroke is listed, it could be accompanied by a code that specifies whether the condition is chronic, recurrent, or new. This effort would enhance the utility of administrative data for meaningfully adjusting outcome data. As the era of the electronic medical record approaches, this option offers considerable promise.

      Conclusion.

      Further understanding of how patients' illnesses influence outcome is critical in improving the efficacy of the medical care delivered to patients. The current study by Naessens and colleagues is another step toward developing a reliable, valid, feasible, and relatively inexpensive system for such measurements.

      REFERENCES

        • Hartz AJ
        • Krakauer H
        • Kuhn EM
        • Young M
        • Jacobsen SJ
        • Gay G
        • Muenz L
        • Katzoff M
        • Bailey RC
        • Rimm AA
        Hospital characteristics and mortality rates.
        N Engl J Med. 1989; 321: 1720-1725
        • Keeler EB
        • Rubenstein LV
        • Kahn KL
        • Draper D
        • Harrison ER
        • McGinty MJ
        • Rogers WH
        • Brook RH
        Hospital characteristics and quality of care.
        JAMA. 1992; 268: 1709-1714
        • Iezzoni LI
        • Shwartz M
        • Moskowitz MA
        • Ash AS
        • Sawitz E
        • Burnside S
        Illness severity and costs of admissions at teaching and nonteaching hospitals.
        JAMA. 1990; 264: 1426-1431
        • Park RE
        • Brook RH
        • Kosecoff J
        • Keesey J
        • Rubenstein L
        • Keeler E
        • Kahn KL
        • Rogers WH
        • Chassin MR
        Explaining variations in hospital death rates: randomness, severity of illness, quality of care.
        JAMA. 1990; 264: 484-490
      1. Med Care. 1988; 26: S1-S141
      2. AL AL Ware Jr, JE Measuring Functioning and Well-Being: The Medical Outcomes Study Approach. Duke University Press, Durham, North Carolina1992
        • Kahn KL
        • Brook RH
        • Draper D
        • Keeler EB
        • Rubenstein LV
        • Rogers WH
        • Kosecoff J
        Interpreting hospital mortality data: how can we proceed?.
        JAMA. 1988; 260: 3625-3628
        • Keeler EB
        • Kahn KL
        • Draper D
        • Sherwood MJ
        • Rubenstein LV
        • Reinisch EJ
        • Kosecoff J
        • Brook RH
        Changes in sickness at admission following the introduction of the prospective payment system.
        JAMA. 1990; 264: 1962-1968
        • Kahn KL
        • Rubenstein LV
        • Draper D
        • Kosecoff J
        • Rogers WH
        • Keeler EB
        • Brook RH
        The effects of the DRG-based prospective payment system on quality of care for hospitalized Medicare patients: an introduction to the series.
        JAMA. 1990; 264: 1953-1955
        • Keeler EB
        • Kahn KL
        • Bentow SS
        Assessing Quality of Care for Hospitalized Medicare Patients With Hip Fracture Using Coded Diagnoses From the MedPAR Provider Analysis and Review Files. Publication No. N-3450-HCFA. RAND, Santa Monica, California1992
        • Gonnella JS
        • Hornbrook MC
        • Louis DZ
        Staging of disease: a case-mix measurement.
        JAMA. 1984; 251: 637-644
        • Shapiro MF
        • Park RE
        • Keesey J
        • Brook RH
        Effect of casemix adjustment on estimated mortality differences between New York City municipal and voluntary hospitals (abstract).
        Clin Res. 1991; 39: 189A