Original ArticlesDuke Case-Mix System (DUMIX) for Ambulatory Health Care
Introduction
Can we measure how sick and how dysfunctional individual patients are at a current point in time and then use the results to predict the health services utilization of those patients during the coming year? This is a very important question for managed care, now that providers of health care must make a commitment in advance to care for certain groups of patients for a predetermined charge. Also, measurement of severity and dysfunction are important within provider groups to monitor and compare quality of patient care and use of resources among the different providers. Quality and efficiency are important to all parties concerned: the providers, the clinic administrators, the third-party payers, and the patients.
Most of the existing ambulatory case-mix measures combine variables such as age, gender, and diagnostic profile in a variety of scoring systems 1, 2, 3, 4, 5. The basic assumption is that knowledge of illness severity will give direction for the most rational and cost-efficient allocation of health care resources. Most of the case-mix measures depend upon indirect indicators of severity, such as diagnoses, rather than direct indicators, such as assessments of health and severity that are made by patients and their health care providers. Also, the existing measures classify patients according to their own diagnostic data that were generated during a previous time period.
We propose to improve ambulatory case-mix measurement with a system that incorporates direct input by both the patient and the provider, and classifies patients by using predictive coefficients derived from other patients in a peer group. The rationale for this approach is: (a) our intuitive belief that the two people who should know the most about how sick or how well a patient is, are that patient and that patient's health care provider, and (b) our empirical finding in the present study that peer group predictive data can be used to weight a given patient's current health and severity scores to estimate expected individual resource use.
For the new case-mix measure, we have built upon previous work that has demonstrated the accuracy of patient-reported functional health status and provider-reported severity of illness as predictors of the future cost of ambulatory health care [6]. We have combined the Duke Health Profile (DUKE) 7, 8, 9, 10, 11and the Duke Severity of Illness Checklist (DUSOI) 10, 12, 13, 14with patient's age and gender into a measure called the Duke Case-Mix System (DUMIX). The present study tests the ability of the DUMIX to classify primary care patients both prospectively and retrospectively for their expected utilization of ambulatory health services.
Section snippets
Methods
Data for developing the DUMIX were collected from 1990 to 1992 on a validation sample of ambulatory patients 18–65 years of age in the Caswell Family Medical Center (CFMC), a rural community health clinic in Caswell County, NC 6, 10, 13. At the time of the study, health care was provided for approximately 3000 patients by four primary care physicians and one physician assistant.
Consenting patients were selected for the study consecutively for several days a week during an 8-month period,
Results
Of the 561 patients in the Caswell Family Medical Center (CFMC) who were asked to participate, 534 consented (95.2%), and 414 (73.8%) were included in the original baseline visit analyses 6, 10, 13. One hundred twenty patients were excluded from analysis because of the following: literacy problems (n = 50), excessive omissions from the intake questionnaire (n = 42), or severe illness (n = 28). The present analyses include 413 patients because of the exclusion of one additional patient whose
Discussion
We have demonstrated in the DUMIX system that the combination of age, gender, current patient-reported perceived and physical health status, and current provider-reported severity of illness at one point in time can be used to predict future and past utilization of ambulatory primary care health services, based upon predictive coefficients from a peer reference group. This utilization forecast can be made for patients in the same clinic as the peer reference group, and for patients in a
Acknowledgements
The authors thank the following health care providers who completed the DUSOI severity assessments at the time of their patients' visits: W. E. Broadhead, M.D., Ph.D.; Janet Jezsik, P.A.-C.; Todd Shapley-Quinn, M.D.; and Bret C. Williams, M.D., M.P.H., in the Caswell Family Medical Center; and Joyce Copeland, M.D.; Clark Denniston, M.D.; Howard Eisenson, M.D.; Albert Meyer, M.D.; and Elisabeth Nadler, M.D., in the Duke Family Medical Center. Data were managed by Alverta Sigmon, and statistical
References (32)
- et al.
Quality of life and functional health of primary care patients
J Clin Epidemiol
(1992) - et al.
Associations among family support, family stress, and personal functional health status
J Clin Epidemiol
(1989) - et al.
The Duke Severity of Illness Checklist (DUSOI) for measurement of severity and comorbidity
J Clin Epidemiol
(1993) - et al.
Ambulatory Visit GroupsA framework for measuring productivity in ambulatory care
Health Services Res
(1984) - et al.
Ambulatory Severity IndexDevelopment of an ambulatory case mix system
J Ambulatory Care Manage
(1988) - et al.
PACsClassifying ambulatory care patients and services for clinical financial management
J Ambulatory Care Manage
(1988) - et al.
Design of a prospective payment patient classification system for ambulatory care
Health Care Financing Rev
(1993) - et al.
Ambulatory Care GroupsA categorization of diagnoses for research and management
Health Services Res
(1991) - et al.
Health status and severity of illness as predictors of outcomes in primary care
Med Care
(1995) - et al.
The Duke Health Profile, a 17-item measure of health and dysfunction
Med Care
(1990)
Development of the 17-item Duke Health Profile
Fam Pract
Comparison of the Duke Health Profile and MOS short-form in healthy young adults
Med Care
Disease-specific versus generic measurement of health-related quality of life in insulin-dependent diabetic patients
Med Care
Feasibility and potential clinical usefulness of a computerized severity of illness measure
Arch Fam Med
Computer-based medical recordsThe centerpiece of TMR
MD Comput
The Medical RecordA comprehensive computer system for the family physician
J Am Board Pract
Cited by (10)
Morbidity severity classifying routine consultations from English and Dutch general practice indicated physical health status
2008, Journal of Clinical EpidemiologyCitation Excerpt :Previous attempts to add a measure of morbidity severity to this basic picture have been based on additional assessment of severity in the consultation, focused on nonprimary care settings [7,8] or have used a restricted number of morbidities [9,10] that have not been validated fully. Such measurement of “severity” has taken two forms: (i) based on a priori classification for routinely collected morbidity data, that is, morbidity severity classified relative to one another [7,9,10] or (ii) based on classification of the severity of morbidity in each patient [8]. We have worked with a group of UK GPs and used their clinician constructs of morbidity severity to produce a classification of the first type.