Variations in physicians' hospitalization practices: a population-based study in Manitoba, Canada

Analyzing the Correlation Between Surgeon Experience and Patient Length of Hospital Stay

Introduction  Many clinical, social, and even economic factors have been extensively analyzed in the literature and shown to influence the length of stay (LOS) after spinal procedures. However, surgeon's experience was mostly examined relative to a learning curve and not regarding the time in practice. The primary objective of this study was to determine the effect of one surgeon's experience on the LOS in patients undergoing one- to two-level transforaminal lumbar interbody fusions (TLIFs). Materials and Methods  The study design was a retrospective cohort study of hospital discharge data. The cohort was comprised of 240 consecutive patients who had undergone open one- or two-level elective TLIF procedures for lumbar degenerative disc disease. The primary predictor was the surgeon's experience based upon the years of practice. The primary outcome was LOS, which was controlled by the discharge criteria that remained consistent throughout the study. Results  Based on the Poisson regression model, it can be inferred that the LOS is not significantly associated with a surgeon's experience (Pr(>|t|) = 0.8985, CI: -0.5825 to 0.5114) while controlling for all other variables. Other independent factors did seem to significantly influence patients' LOS, including the admission type (Pr(>|t|) = 9.637^-08, CI: -0.8186 to -0.3786), the number of TLIF levels (Pr(>|t|) = 1.721^-06, CI: 0.0606 to 0.1446), the Clavien-Dindo ( Pr(>|t|) = 0, CI: 0.1489 to 0.1494), the American Society of Anesthesiologists (ASA) physical status classification scores (Pr(>|t|) = 4.878^-3, CI: 0.0336 to 0.1880), and being discharged to skilled nursing facility (Pr(>|t|) = 3.44^-2, CI: 0.0127 to 0.3339). Conclusions Based upon the years in practice, surgeon experience was not associated with length of hospitalization and estimated blood loss during surgery in patients undergoing one- and two-level TLIF surgeries. However, while controlling for all other variables, the surgeon's experience and surgical time had a highly significant correlation. The study results clearly demonstrated efficiency, but we did not identify a clear correlation between LOS and surgeon experience overtime suggesting that other factors are likely contributing to such outcome. The average LOS is a complex measure of healthcare resource use and hospital discharge policy or other variables are likely having more effect on LOS than individual surgeons' preferences.

The Effect of Medical Work Groups on Hospital Resource use

This study tests the ability of medical work groups to overcome coordination problems related to group decision-making in allocating clinical resources to inpatients. The study was conducted over a 32-month period in two medium-sized acute-care hospitals located in Montreal, Quebec, Canada. The data were collected by hand from the medical charts of 10456 patients in the surgical and medical departments. The Linear Structural Relations (LISREL) approach was employed to address the work-group issue using a task contingent model of work-group organization. In this model, the nature of the task is fundamental because its level of complexity determines both the organization of the work group and the use of resources. Medical work-group mechanisms should be efficient to the extent that resource utilization is explained solely by task characteristics rather than by work-group structure. In this study, the following two major organizational concepts were used as factors to explain resource use: task characteristics and work-group characteristics. Our analysis confirmed the main points of the task contingency theory as applied to the field of medicine. First, the results confirm that resource utilization is explained mainly by task complexity. Second, they confirm that medical work groups modulate their structures on the basis of task characteristics and do not explain resource use. The results also reveal a more complex model in which, for instance, the concepts of medical task and medical professional work are not easy to separate. The results highlight the interest in conceptualizing and analysing medical practice in work groups. It raises important issues that have seldom been taken into account in the study of medical practice variations, which has tended to focus on attending physicians.

In place of fear: aligning health care planning with system objectives to achieve financial sustainability

The financial sustainability of publicly funded health care systems is a challenge to policymakers in many countries as health care absorbs an ever increasing share of both national wealth and government spending. New technology, aging populations and increasing public expectations of the health care system are often cited as reasons why health care systems need ever increasing funding as well as reasons why universal and comprehensive public systems are unsustainable. However, increases in health care spending are not usually linked to corresponding increases in need for care within populations. Attempts to promote financial sustainability of systems such as limiting the range of services is covered or the groups of population covered may compromise their political sustainability as some groups are left to seek private cover for some or all services. In this paper, an alternative view of financial sustainability is presented which identifies the failure of planning and management of health care to reflect needs for care in populations and to integrate planning and management functions for health care expenditure, health care services and the health care workforce. We present a Health Care Sustainability Framework based on disaggregating the health care expenditure into separate planning components. Unlike other approaches to planning health care expenditure, this framework explicitly incorporates population health needs as a determinant of health care requirements, and provides a diagnostic tool for understanding the sources of expenditure increase.

Patient and physician decision styles and breast cancer chemotherapy use in older women: Cancer and Leukemia Group B protocol 369901

PURPOSE Physician and patient decision styles may influence breast cancer care for patients ≥ 65 years ("older") because there is uncertainty about chemotherapy benefits in this group. We evaluate associations between decision-making styles and actual treatment. METHODS Data were collected from women treated outside of clinical trials for newly diagnosed stage I to III breast cancer (83% response) from January 2004 through April 2011 in 75 cooperative group sites. Physicians completed a one-time mailed survey (91% response), and clinical data were abstracted from charts. Patient decision style was measured on a five-point scale. Oncologists' preference for prescribing chemotherapy was based on standardized vignettes. Regression and multiple imputation were used to assess associations between chemotherapy and other variables. Results There were 1,174 women seen by 212 oncologists; 43% of women received chemotherapy. One-third of women preferred to make their own treatment decision. Patient and physician decision styles were independently associated with chemotherapy. Women who preferred less physician input had lower odds of chemotherapy than women who preferred more input (odds ratio [OR] = 0.79 per 1-point change; 95% CI, 0.65 to 0.97; P = .02) after considering covariates. Patients whose oncologists had a high chemotherapy preference had higher odds of receiving chemotherapy (OR = 2.65; 95% CI, 1.80 to 3.89; P < .001) than those who saw oncologists with a low preference. CONCLUSION Physicians' and older patients' decision styles are each associated with breast cancer chemotherapy use. It will be important to re-evaluate the impact of decision styles when there is greater empirical evidence about the benefits and risks of chemotherapy in older patients.

Individual and contextual determinants of regional variation in prescription drug use: an analysis of administrative data from British Columbia

Background

Increasing attention is being paid to variations in the use of prescription drugs because their role in health care has grown to the point where their use can be considered a proxy for health system performance. Studies have shown that prescription drug use varies across regions in the US, UK, and Canada by more than would be predicted based on age and health status alone. In this paper, we explore the determinants of variations in the use of prescription drugs, drawing on health services theories of access to care.

Methods

We conducted a cross-sectional analysis using population-based administrative health care data for British Columbia (BC), Canada. We used logistic and hierarchical regressions to analyze the effects of individual- and area-level determinants of use of prescriptions overall and rates of purchase of prescriptions from five therapeutic categories representing a range of indications: antihypertensives, statins, acid reducing drugs, opioid drugs, and antidepressants. To indicate the relative scale of regional variations and the importance of individual- and area-level variables in explaining them, we computed standardized rates of utilization for 49 local health areas in BC.

Results

We found that characteristics of individuals and the areas in which they live affect likelihood of prescription drug purchase. Individual-level factors influenced prescription drug purchases in ways generally consistent with behavioral models of health services use. Contextual variables exerted influences that differed by type of drug studied. Population health, education levels, and ethnic composition of local areas were associated with significant differences in the likelihood of purchasing medications. Relatively modest regional variations remained after both individual-level and area-level determinants were taken into account.

Conclusions

The results of this study suggest that individual- and area-level factors should be considered when studying variations in the use of prescription drugs. Some sources of such variations, including individual- and area-level socioeconomic status, warrant further investigation and possible intervention to address inequities.

The formation and evolution of physician treatment styles: an application to cesarean sections

Small-area-variation studies have shown that physician treatment styles differ substantially both between and within markets, controlling for patient characteristics. Using data on the universe of deliveries in Florida and New York over a 15-year period, we examine why treatment styles differ across obstetricians at a point in time and why styles change over time. We find that variation in c-section rates across physicians within a market is about twice as large as variation between markets. Surprisingly, residency programs explain no more than four percent of the variation in physicians' risk-adjusted c-section rates, even among newly trained physicians. Although we find evidence that physicians learn from their peers, they do not substantially revise their prior beliefs regarding treatment due to the local exchange of information. Our results indicate that physicians are not likely to converge over time to a community standard; thus, within-market variation in treatment styles is likely to persist.

Changes in and predictors of length of stay in hospital after surgery for breast cancer between 1997/98 and 2004/05 in two regions of England: a population-based study

Background

Decreases in length of stay (LOS) in hospital after breast cancer surgery can be partly attributed to the change to less radical surgery, but many other factors are operating at the patient, surgeon and hospital levels. This study aimed to describe the changes in and predictors of length of stay (LOS) in hospital after surgery for breast cancer between 1997/98 and 2004/05 in two regions of England.

Methods

Cases of female invasive breast cancer diagnosed in two English cancer registry regions were linked to Hospital Episode Statistics data for the period 1^st April 1997 to 31^st March 2005. A subset of records where women underwent mastectomy or breast conserving surgery (BCS) was extracted (n = 44,877). Variations in LOS over the study period were investigated. A multilevel model with patients clustered within surgical teams and NHS Trusts was used to examine associations between LOS and a range of factors.

Results

Over the study period the proportion of women having a mastectomy reduced from 58% to 52%. The proportion varied from 14% to 80% according to NHS Trust. LOS decreased by 21% from 1997/98 to 2004/05 (LOSratio = 0.79, 95%CI 0.77-0.80). BCS was associated with 33% shorter hospital stays compared to mastectomy (LOSratio = 0.67, 95%CI 0.66-0.68). Older age, advanced disease, presence of comorbidities, lymph node excision and reconstructive surgery were associated with increased LOS. Significant variation remained amongst Trusts and surgical teams.

Conclusion

The number of days spent in hospital after breast cancer surgery has continued to decline for several decades. The change from mastectomy to BCS accounts for only 9% of the overall decrease in LOS. Other explanations include the adoption of new techniques and practices, such as sentinel lymph node biopsy and early discharge. This study has identified wide variation in practice with substantial cost implications for the NHS. Further work is required to explain this variation.

There's No Reason for It, It's Just Our Policy

On June 1, 2009 the town of McAllen, Texas rose to brief prominence on the American political stage. With the highest (bar Miami) per-beneficiary costs in the entire US Medicare program, it was featured in an essay in The New Yorker by Atul Gawande, then seized upon by President Obama: "This is what we have to fix." Behind the headlines were decades of documentation of clinical practice and analysis of regional variations by John Wennberg, Elliott Fisher and their colleagues, and by Leslie and Noralou Roos and theirs. The implications for health systems were grasped over 30 years ago and have been confirmed by more recent work. Efforts to understand these variations within standard economic theory have, however, had limited success.

Does experience matter? A comparison of the practice of attendings and residents

OBJECTIVE

To compare the utilization of health care resources and patterns of chronic disease care by patients of medical residents and patients of their attending physicians.

MATERIALS AND METHODS

This study involved a longitudinal cohort of 14,554 patients seen over a 1-year period by 149 residents and 36 attendings located in an urban academic medical center. Data were acquired prospectively through a practice management system used to order tests, write prescriptions, and code ambulatory visits. We assessed resource utilization by measuring the total direct costs of care over a 1-year period, including ambulatory and inpatient costs, and the numbers and types of resources used.

RESULTS

Residents' patients were similar to attendings' patients in age and gender, but residents' patients were more likely to have Medicaid or Medicare and to have a higher burden of comorbidity. Total annual ambulatory care costs were almost 60% higher for residents' patients than for attendings' patients in unadjusted analyses, and 30% higher in analyses adjusted for differences in case mix (adjusted mean 888 dollars vs 750 dollars; P=.0001). The primary cost drivers on the outpatient side were consultations and radiological procedures. Total inpatient costs were almost twice as high for residents' patients compared to attendings' patients in unadjusted analyses, but virtually identical in analyses adjusted for case mix differences (adjusted mean of 849 dollars vs 860 dollars). Admission rates were almost double for residents' patients. Total adjusted costs for residents' patients were slightly, but not significantly, higher than for attendings' patients (adjusted mean 1,651 dollars vs 1,540 dollars; P>.05). Residents' and attendings' patients generally did not differ in the patterns of care for diabetes, asthma/chronic obstructive pulmonary disease (COPD), congestive heart failure, ischemic heart disease, and depression, except that residents' patients with asthma/COPD, ischemic heart disease, and diabetes were admitted more frequently than attendings' patients.

CONCLUSIONS

Our results indicate that residents' patients had higher costs than attendings' patients, but the differences would have been seriously overestimated without adjustment. We conclude that it costs about 7% more for residents to manage patients than for attendings. On the ambulatory side, the larger number of procedures and consults ordered for residents' patients appears to drive the higher costs.

Economics and the evaluation of health care programmes: generalisability of methods and implications for generalisability of results

Increasing attention is being given to identifying standardised methods of analysis for the economic evaluation of health care programmes and generating generalisable findings from these methods. In this paper, we show how these approaches fail to reflect the social science foundations of the economics discipline and the economic theory of individual behaviour. Using simple examples, we show that the technical efficiency of a particular programme differs between communities, even though the underlying technology is the same for the communities. Similarly, the subjective considerations represented by the utility function are not generally transferable between settings or between individuals within settings. As a result, the efficiency of an intervention will be influenced by the context in which the intervention is experienced, even in the presence of identical production and utility functions. The lack of generalisability includes the validity of the methods used to analyse the subjective component of the evaluation exercise. The adoption of standardised methods of measurement and analysis, together with the use of findings from the application of these methods in other settings, might ease the administrative burden presented in resource allocation exercises. However, these approaches do not accommodate the intellectual substance of the wide range of problems and circumstances that underlie these exercises.

Clinical practice guidelines: when the tool becomes the rule

This work provides a critical examination of the use of clinical practice guidelines to measure individual performance. The problems inherent in using a measure of central tendency derived from a distribution of individual performances are addressed, as is the translation of the collectively determined guidelines into a measurement instrument. It is suggested that every process on the distribution of processes used to determine the guideline must be considered equally legitimate representations of the process in question. It is further suggested that to accept as a standard of quality, a particular process simply because there is a minimum of variation between providers, is to ignore the importance of the linkage between process and outcome. The importance of an independent measure of quality based on outcomes is further emphasized by highlighting the tautological nature of analyses that include an input measure, such as nursing hours, in both the dependent variable and the list of independent variables. It is recommended that individual performance be evaluated within the tolerances of the distribution from which they were derived and not be held to some measure of central tendency of that distribution. The alternative is to use the measure of central tendency with plus or minus limits such as one, or more, standard deviations.

Differences between patients with heart failure treated by cardiologists, internists, family physicians, and other physicians: analysis of a large, statewide database

BACKGROUND

The management of heart failure (HF) by cardiologists may be better than that of other physicians in that cardiologists' treatment choices more frequently conform with published guidelines and the results of clinical trials. Whether cardiologists' management of HF is more or less cost-effective is up for debate.

METHODS

Information on all 1995 New York state hospital discharges assigned ICD-9-CM codes indicative of HF in the principal diagnosis position was obtained. Demographic and clinical characteristics, process of care, resource utilization, and short-term HF-related outcomes were compared between patients of cardiologists and patients of other physicians.

RESULTS

A total of 44,926 patients were identified, with 10,506 (23%) receiving care from cardiologists, 28,300 (63%) from internists, 4812 (11%) from family practitioners, and 1308 (3%) from other physicians. Patients of cardiologists were younger, more frequently male, and less frequently residents of nursing homes. They were more likely to have associated cardiovascular diagnoses but less likely to have comorbid general medical conditions. Patients of cardiologists were more likely to undergo cardiac catheterization (9%) than those of internists (3%) and family practice (2%) physicians but had similar adjusted hospital length of stay and charges. Mortality and hospital readmission rates for HF were similar among the groups. Patients in the "other" group (managed mostly by surgeons) were the youngest, underwent more invasive and cardiac surgical procedures, and had the longest length of stay and highest hospital charges.

CONCLUSIONS

Cardiologists' management of HF is not economically disadvantageous. The relations among physician specialty, process of care, resource utilization, and clinical outcomes require further study before rational and evidence-based health care staffing recommendations can be formulated.

Rural and urban physicians: does the content of their Medicare practices differ?

Rural and urban areas have significant differences in the availability of medical technology, medical practice structures and patient populations. This study uses 1994 Medicare claims data to examine whether these differences are associated with variation in the content of practice between physicians practicing in rural and urban areas. This study compared the number of patients, outpatient visits, and inpatient visits per physician in the different specialties, diagnosis clusters, patient age and sex, and procedure frequency and type for board-certified rural and urban physicians in 12 ambulatory medical specialties. Overall, 14.4 percent of physicians in the 12 specialties practiced exclusively in rural Washington, with great variation by specialty. Rural physicians were older and less likely to be female than urban physicians. Rural physicians saw larger numbers of elderly patients and had higher volumes of outpatient visits than their urban counterparts. For all specialty groups except general surgeons and obstetrician-gynecologists, the diagnostic scope of practice was specialty-specific and similar for rural and urban physicians. Rural general surgeons had more visits for gastrointestinal disorders, while rural obstetrician-gynecologists had more visits out of their specialty domain (e.g., hypertension, diabetes) than their urban counterparts. The scope of procedures for rural and urban physicians in most specialties showed more similarities than differences. While the fund of knowledge and outpatient procedural training needed by most rural and urban practitioners to care for the elderly is similar, rural general surgeons and obstetrician-gynecologists need training outside their traditional specialty areas to optimally care for their patients.

Using collectively-derived standards to evaluate individual performance: a cautionary note on clinical practice guidelines

The purpose of this work is to demonstrate the problem of evaluating an individual physician's performance relative to practice guidelines which have typically been derived from group consensus or some measure of central tendency. It is argued that when evaluated against a set of criteria derived at the macro-level, an individual physician's performance may justifiably vary due to the patient characteristics or the evolving process of care. It is also argued that it is not necessarily true that costs are reduced when practice variation is reduced. The results indicate that there are cost reduction in areas not targeted by the guidelines, suggesting a possible 'spillover effect' due to the increased vigilance in monitoring provider performance. The results also provide some evidence of increased costs following a reduction in variation. Caution should be exercised when evaluating individual physician performance relative to guidelines established at the aggregate level. Acceptable individual physician performance should be judged within the upper and lower boundaries of the implicit distribution of physicians' performances from which the established guidelines generated.

Cardiology or primary care for heart failure in the community setting: process of care and clinical outcomes

STUDY OBJECTIVES

Severity of illness, treatment choices, and clinical outcomes may vary with physician training. This study was performed to determine whether such differences exist among patients with congestive heart failure (CHF) treated by cardiologists and by noncardiologists in the community hospital setting.

DESIGN

Prospective cohort study.

SETTING

Ten acute-care community hospitals. PATIENTS, MEASUREMENTS, AND RESULTS: Two thousand four hundred fifty-four patients with CHF were identified and followed up for 6 months after hospital discharge. Patients who were not treated by a cardiologist (group I; n = 977) were compared with patients whose attending physician was a cardiologist (group II; n = 419) and patients who received consultative care from a cardiologist (group III; n = 1,058). When compared with group I patients, group II patients were more likely to receive the recommended diagnostic tests and treatment strategies, although some of these differences could be explained by variations in the case mix. Group II patients had higher hospital charges, but lower CHF readmission rates and better postdischarge quality-of-life measures. No differences in adjusted mortality rates were observed.

CONCLUSIONS

In the community-hospital setting, the clinical practices of cardiologists are more compatible with published treatment guidelines than the clinical practices of other physicians. The benefits of cardiology specialty care include lower CHF readmission rates and better postdischarge quality-of-life measures, rather than lower mortality rates, fewer hospital charges, or shorter length of stay.

Interpreting risk-adjusted length of stay patterns for VA hospitals

OBJECTIVES

The Veterans Health System must become more competitive with the private sector in terms of efficiency of care. Studies have shown significantly longer lengths-of-stay (LOS) in facilities operated by the Department of Veterans Affairs (VA) compared with private sector facilities. Most comparisons, however, have not controlled well for casemix differences or have involved small numbers of patients. The aims of this study were: (1) controlling for casemix, to accurately measure the degree by which average length of stay in Veterans Affairs facilities exceeds that of private sector hospitals and (2) to demonstrate a methodology with which individual VA facilities can identify clinical and demographic subgroups of patients associated with the higher length-of-stay averages.

METHODS

Subjects of the study were Veterans Health System patients hospitalized during 1991-1993 and veteran respondents to the 1991 National Hospital Discharge Survey. Hospitals' mean length of stay adjusted for patients' diagnosis related groups, severity, demographics, and travel distances were measured.

RESULTS

Veterans Affairs medical centers' average risk-adjusted length of stay was 36% higher (8.9 days compared with 6.5 days) than that of the private sector. For individual hospitals, relative length-of-stay efficiency typically varied by condition. Among 14 hospitals in the VA's midwest region, none were high risk-adjusted length-of-stay outliers in all conditions studied, and four were high outliers for some conditions and low outliers for others.

CONCLUSIONS

Controlling for differences in patient demographic and clinical factors, Veterans Affairs medical centers consumed significantly more days of care than private sector hospitals. Veterans Affairs medical centers will be able to improve efficiency by identifying specific subgroups of patients whose clinical treatment should be examined.

[Use pf patient management categories (PMC) in the study of the variability of mean hospital stay for 3 surgical procedures]

GOAL

To analyze variation in length of stay in acute hospitals by using a large administrative data base based on CMBD.

SUBJECTS AND METHODS

Surgical patients admitted from 93 to 95 in the seven hospital with the largest volume of admissions from the Servicio Vasco de Salud--Osakidetza with a discharge diagnosis of inguinal hernia (IH), benign prostatic hyperplasia (BPH) and total hip joint replacement (THJR). We used data generated by a Patient Management Categories (PMC) data base. We present median length of stay result adjusted by age, sex, type of admission and Risk Intensity Score (RIS). We employed Analysis of Covariance (ANCOVA) for the multivariate analysis including the relevant interaction terms.

RESULT

Adjusted median length of stay went from 2 to 6 days for IH patients, from 7 to 18 for BPH patients, from 5 to 18 for THJR patients; depending upon the different hospital and patient attributable considered confounders. By year, we saw a general decrease on length of stay.

CONCLUSIONS

We found important differences by hospital though there is a decrease in length of stay in the last years. The administrative data bases, in spite of presenting validity problems, are a cheap and quick way of analyzing certain indicators to help in care management; given their limitations, cautious interpretation of the results is necessary.

The impact of physician involvement in managed care on efficient use of hospital resources

OBJECTIVES

This research assesses the impact of managed care on the physician's efficient use of hospital resources. It examines three questions. (1) Does a higher percentage and volume of managed care patients in the physician's hospital practice lead to more efficient utilization? (2) Do physicians shift cost to nonmanaged care patients in an effort to compensate for lower reimbursement for managed care patients? (3) Are there threshold effects in the percentage and volume of managed care patients treated by physicians?

METHODS

The study combines patient discharge data from the state of Arizona with physician and hospital data for a 2-year period. Random effects maximum likelihood (REML) regressions were performed for four different diagnosis classifications to examine the effect of the physician's managed care caseload on mean-adjusted charges and length of stay.

RESULTS

The findings suggest that physicians with high percentages and volumes of managed care patients in their hospital practice are more efficient in using hospital resources. The findings also suggest that physicians may compensate for the lower reimbursement from managed care patients by increasing their resource use among non-health maintenance organization patients.

CONCLUSIONS

Finally, there appears to be a threshold effect of managed care activity on the physician's hospital utilization in one of the conditions studied.

Utilization of hospital services by the elderly: geriatric crisis in one Canadian single payer system

As the number and proportion of elderly persons in the Canadian population increase, utilization of health services by the elderly becomes a growing concern for health service insurers, financial managers and policy makers, as well as for care providers. The purpose of this paper is to present the results of a study to analyse the use of hospital services by the elderly in Alberta since the introduction of a universal single payer health care insurance system in 1970. The study period coincides with the implementation of publicly-financed comprehensive medical and hospital insurance programmes for all Alberta residents, making it possible to perform historical and population-based utilization analyses. Thus the data used for the study included all hospital discharge abstracts generated by all Alberta hospitals from 1971 to 1991. Trends in hospital service utilization by the elderly in terms of total number of separations, patient-days, and per case measures such as average length of stay as well as per capita utilization rates were reviewed to identify utilization patterns over the study period. Further, relative per capita utilization measures, in comparison with the base year (1971), age group 15-44, male, metropolitan residents, were derived and historical trends identified. A series of regression analyses were carried out to estimate the effects of age, sex and origin on utilization rates. In addition, for the period of 1984-1991, Diagnosis Related Groups (DRG) case weights were used to measure per capita and per case rates and to analyse historical relative utilization rates over the 8-year period. In general, there has been a significant decline in hospital utilization by Albertans under the publicly-financed single payer system, but utilization rates for elderly have remained high, resulting in high relative utilization rates in comparison with other age groups. It was also noted that per capita utilization rates for rural residents were substantially higher than urban residents. It appears that these higher utilization rates by the elderly and rural residents in combination with tight bed and financial control by the government have been causing significant bed shortage problems for non-elderly elective patients in urban areas.

Good technologies gone bad: how and why the cost-effectiveness of a medical intervention changes for different populations

Cost-effectiveness (CE) ratios vary considerably, not only across interventions, but within single interventions. Using a simple decision-tree model of the treat-vs no-treat decision to organize the analysis, four potential errors leading to these within-treatment differences in CE ratios are identified. These errors arise from estimates relating to 1) prior probabilities of disease; 2) treatment efficacies; 3) costs of treatment; and 4) patient preferences. Systematic biases, where present, suggest overuse of medical interventions. For diagnostic tests, two additional potential sources of error are considered (using a simple decision tree incorporating both test and treat decisions). These involve 5) sensitivity and specificity of the diagnostic test and 6) inappropriate choice of "cutoff" to determine abnormal patients, in part arising from errors in estimating prior probability of disease.

Classifying general medicine readmissions. Are they preventable? Veterans Affairs Cooperative Studies in Health Services Group on Primary Care and Hospital Readmissions

OBJECTIVES

To describe a new quality assessment method used to classify the preventability of hospitalization in terms of patient, clinician, or system factors.

DESIGN

The instrument was developed in two phases. Phase 1 was a prospective comparison of admitting residents' and their attending physicians' classifications of the perceived preventability of consecutive admissions to one Veterans Affairs Medical Center (VAMC) excluding admissions to the intensive care unit (ICU). In phase 2, a panel of 10 physicians rated 811 abstracted records of readmissions from nine VAMCs.

SETTING

Nine VAMCs across the United States with varying degrees of university hospital affiliation.

PATIENTS

Phase 1, 156 patients admitted to the general medicine service at the Durham VAMC. Phase 2, 514 patients accounting for 811 readmissions within 6 months of a general medicine service discharge at nine VAMCs.

MEASUREMENTS AND MAIN RESULTS

Physicians used a checklist to record the reason for hospitalization, the preventability of the hospitalization, and, if preventable, a reason defining preventability, which was classified in terms of system, clinician, and patient factors. In phase 2, two physician panelists assessed preventability for each chart. When two panelists disagreed on the preventability of hospitalization, a third panelist, blind to the original assessments, rated the chart. In phase 1, residents and attending physicians rated 33% and 34% of admissions as preventable (kappa = 0.41), respectively. In phase 2, 277 (34%) of 811 readmissions were deemed preventable. Intraobserver accuracy for the assessment of preventability was 96% (kappa = 0.89). interobserver accuracy was 73% (kappa = 0.43). Hospital system factors accounted for 37% of preventable readmissions, clinician factors for 38%, and patient factors for 21%. The nine hospitals differed markedly in their profile of reasons for preventable readmissions (p = .005).

CONCLUSIONS

Using a new method of determining the preventability of hospitalizations, we identified several factors that might avert hospitalizations. Focusing efforts to identify preventable hospitalizations may yield better methods for managing patients' total health care needs; however, the content of those efforts will vary by institution.

Intensive use of mental health care

OBJECTIVE

To describe the profile of the intensive use of mental health services over a 4-year period in a population of 1.1 million people.

METHODS

Data obtained from computerized hospital separation records and physician reimbursement claims were combined to form patient-based histories of mental health care utilization. Users of mental health services in a 24-month period were hierarchically classified as having a psychotic disorder (ICD-9-CM 295-299) or a nonpsychotic disorder (ICD-9-CM 300-301, 306-309, 311). Intensive use was defined as 12 or more contact months or a minimum of 2 episodes of therapy in the 24-month period. The cohort of intensive users were followed over the subsequent 24-month interval to describe the persistence of intensive use.

RESULTS

In the initial observation periods, intensive users constituted 27.4% of individuals in treatment for psychotic disorder and 4.4% of persons in treatment for nonpsychotic disorder. These 2 groups, which represent 7.4% of all users of mental health care, were responsible for 53% of physician services, 72.7% of contacts with psychiatrists, and 64.4% of acute psychiatric bed days in the initial period. In the follow-up period, intensive use status was replicated by 44.6% of the cohort.

CONCLUSIONS

The diagnostic and therapeutic characteristics of intensive users of mental health services are heterogeneous. There is substantial persistence of intensive mental health service use over time.

Linking cost control measures to health care services by using activity-based information

Canada's health care institutions are under pressure to limit expenditures, maintain or increase productivity, and assimilate new technology. Even though more than 75% of hospital operating expenditures are controllable, according to a study by the Economic Council of Canada, cost systems are needed to provided essential management information. The new Canadian Management Information System (MIS) Guidelines for health care are designed to provide accurate cost measurement of patient treatment and to help managers evaluate the impact of planned program changes on areas of operational responsibility. Other potential benefits of implementing the MIS guidelines include correcting dysfunctional funding of health care units with benchmarking and setting high reporting standards for resource use at the patient level (MIS, 1991). This paper focuses on one important aspect of bringing these costs under control by examining the relation between cost deviations (variances) and underlying cost drivers. Our discussion will lead to the conclusion that incompatibility of DRG methodology and traditional cost accounting models may be an important source of cost variability within diagnostically-related disease groupings.

The effect of physician practice organization on efficient utilization of hospital resources

OBJECTIVE

This study examines variations in the efficient use of hospital resources across individual physicians.

DATA SOURCES AND SETTING

The study is conducted over a two-year period (1989-1990) in all short-term general hospitals with 50 or more beds in Arizona. We examine hospital discharge data for 43,625 women undergoing cesarean sections and vaginal deliveries without complications. These data include physician identifiers that permit us to link patient information with information on physicians provided by the state medical association.

STUDY DESIGN

The study first measures the contribution of physician characteristics to the explanatory power of regression models that predict resource use. It then tests hypothesized effects on resource utilization exerted by two sets of physician level factors: physician background and physician practice organization. The latter includes effects of hospital practice volume, concentration of hospital practice, percent managed care patients in one's hospital practice, and diversity of patients treated. Efficiency (inefficiency) is measured as the degree of variation in patient charges and length of stay below (above) the average of treating all patients with the same condition in the same hospital in the same year with the same severity of illness, controlling for discharge status and the presence of complications.

PRINCIPAL FINDINGS

After controlling for patient factors, physician characteristics explain a significant amount of the variability in hospital charges and length of stay in the two maternity conditions. Results also support hypotheses that efficiency is influenced by practice organization factors such as patient volume and managed care load. Physicians with larger practices and a higher share of managed care patients appear to be more efficient.

CONCLUSIONS

The results suggest that health care reform efforts to develop physician-hospital networks and managed competition may promote greater parsimony in physicians' practice behavior.

Is the public being protected? Prevention of suboptimal medical practice through training programs and credentialing examinations

Governments have traditionally looked to the medical profession for leadership in health planning and have charged the profession with the responsibility of establishing and monitoring standards of medical practice. Training program accreditation and licensure/certification exams have been used as the primary methods of preventing unqualified individuals from entering medical practice. Despite the critical nature of the decision made at the time of licensure/certification, there is no information about the validity of these examinations for predicting subsequent practice and health outcome. In this article, the assumptions implicit in the current use of licensing/certifying examinations are identified, the relevant evidence is reviewed, and the implications of this evidence for current methods of measurement are discussed.

Influence of bed supply and health care organization on regional and local patterns of diabetes related hospitalization

This paper undertakes both a macro- and micro-scale analysis of the influences exerted by the health care system on patterns of hospitalization. The health disorder of diabetes mellitus is used as the case study and the analyses are based on New Zealand data sets. The article first examines the extent to which both the supply and organization of primary and secondary health care affect rates of hospitalization. The macro-scale analysis investigates the applicability of Roemer's Law to regional variations in diabetes hospitalization. The organizational control of hospital utilization via doctor gatekeeping functions and interaction between health services are then examined at the local level. This analysis assumes a population based approach using the Canterbury Register of Insulin-treated diabetic persons as the study population. Diabetes discharge rates were found to be most highly correlated with hospital bed supply in 5 of the 8 years studied (1979-1986). Stepwise regression analysis indicated area rates of diabetes hospitalization were significantly influenced by resource factors even after controlling for differences in the socio-demographic characteristics of the area populations. This confirmed the presence of Roemer's Law at the aggregate level with rates of diabetes hospitalization appearing to have more to do with the availability of medical resources than to population needs. At the local level, hospital admission patterns were found to vary by general practitioner age, practice type found to vary by general practitioner age, practice type and diabetic caseload. Overall, insulin-treated diabetic patients most likely to be hospitalized were those in the care of young doctors new to general practice, and those who attended doctors who had small diabetic caseloads. Solo practitioners had the lowest rates of patient hospitalization. There were marked disparities in patient access to specialist diabetes education and clinical outpatient services by patient age, duration of diabetes and attendance on primary care. Overall, no significant differences were found in the propensity for hospitalization between users and non-users of these specialist services. This does not imply however, service ineffectiveness but rather is indicative of the complexity of the local diabetes care organization and the differing needs of the insulin-treated diabetic population within the community as a whole.

Hospitalization style of physicians in Manitoba: the disturbing lack of logic in medical practice

Variations in hospital admission rates across small areas are ubiquitous, and it is increasingly assumed that high rates result from physicians' discretionary decisions. Data for elderly patients from the health insurance system of Manitoba were used to construct an index that divided physicians into four groups based on their propensity to admit patients to the hospital. I then determined whether physicians who are more prone to admit patients use hospitals for more discretionary purposes and admit patients who are less ill. Although the differences between physicians with different practice styles were in the expected direction, the most compelling finding was the similarity in characteristics of patients admitted by physicians with markedly different practice styles. Such findings suggest a very wide latitude in physicians' decisions to admit patients; this latitude is not well captured by a model that posits a logical relationship between physician treatment patterns and patient need.

Identification of factors associated with hospital readmission and development of a predictive model

Multiple hospital admissions, especially those related to the chronically ill, represent a particular challenge to both the acute and long-term care sectors to identify effective methods of resource management. This study analyzes the multiple admission patterns associated with a cohort of 4,219 adult medical-surgical patients discharged alive from a community teaching hospital in Michigan. The sample was divided into two groups: 3,818 patients who survived and 392 who expired during the one-year follow-up period. For the surviving subsample, the characteristics found to be directly associated with the likelihood of readmission included increased age, advanced stage of disease, greater index-episode length of stay, discharge by an internist rather than a surgeon, Medicare as expected source of payment, decreased physician age, discharge to a community setting, and increased number of prior hospital episodes. For the subsample who died, only one explanatory variable was significantly associated with an increased likelihood of readmission-discharge to a community setting (with or without home care) rather than a nursing home. The article includes illustrates of the importance of decisions regarding posthospital, long-term care services on the likelihood of rehospitalization.

The financial effect of physician practice style on hospital resource use

Several specifications of a statistical model were used to measure the effect that internal medicine attending physicians had on inpatient charges and length of stay at a large urban teaching hospital. The study was based on a sample of 1,458 patients discharged during 1985-1987 with 12 common principal diagnosis clusters. The relationship between 31 physicians' clinical decisions and hospital charges and length of stay was analyzed controlling for patients' health status, as measured by demographic characteristics, diagnostic group, and ratings for the Severity of Illness Index (SOII). Results indicated that attending physicians were statistically significant predictors of the log of total charges (p = .0030) and the log of length of stay (p less than .0001), and not as significant predictors of untransformed total charges (p = .1255). Equivalent results were obtained when overall SOII ratings were replaced by SOII subscale ratings for the presenting stage of the principal diagnosis on admission. Examination of individual physician regression coefficients revealed that physicians varied within a 40 percent range of generated per patient charges. No significant differences in mortality, early readmissions or residual impairment on discharge were found between the ten highest and ten lowest resource use physicians. The conservatively estimated range of attending physician practice variations observed in this study has serious financial implications for hospitals operating under incentives to minimize operating costs, particularly for teaching hospitals facing reductions in subsidies for graduate medical education.

Dental decisionmaking and variation in dentist service rates

Previous medical and dental studies report wide variations in service rates across small areas, regions and providers. A major source of this variation appears to be the pattern of clinical decisionmaking of the provider. This argument was tested using dentist service rates (the average number of services provided per patient) calculated from 1984-1985 claims data for a population of well-educated, middle-class patients in Washington state, U.S.A. (N = 23,153). Service rates were calculated for each dental procedure in the following pairs of alternative treatments: crown vs amalgam or crown build-up; root canal therapy vs extraction; and fixed bridge vs removable partial denture. For each pair, dentists identified patient (e.g. cost, patient preference) and technical (e.g. periodontal status, tooth damage) factors which they considered to be important in choosing therapy. Regression analysis revealed that the technical factors explained little variation in the rates, while at least one patient factor was significant across services, except prosthetics. Environmental characteristics, structural features of the practice, and the dentists' practice beliefs also explained variation in the rates.

The Pediatric Cardiac Care Consortium: a physician-managed clinical review program

A voluntary physician-initiated and physician-managed clinical review program for pediatric cardiology uses clinical practice data and expected outcome to influence physicians to change their clinical behavior. The program offers a model for clinical review in situations where disease incidence is relatively low and physician performance is difficult to assess and evaluate. The program's major accomplishment is a collective pooling of data across cardiac centers that allows for statistical analysis and comparison not routinely possible at a single center because of sample size; a forum for comparison of different practice styles; and feedback to participants of actual, in contrast to anecdotal, results. The principles, structure, and format of the program can be applied to other specialty areas.

Comparing the quality of referrals of general practitioners with high and average referral rates: an independent panel review

The quality of referrals of four general practitioners, two with high and two with average rates of referral to the department of internal medicine, was judged by an independent expert panel. The panel, consisting of two general practitioners and one specialist, reviewed a set of information about the referrals blindly and in random sequence. The same distribution of quality of referrals was found among the referrals of the two high referring general practitioners (n = 192) as among those of the general practitioners with average rates (n = 88); that is, 57% and 55% respectively, of the cases had clear medical indications for referral, while the data did not permit a conclusion in 15% and 10%, respectively, of the cases. Controlling for sex, age and status of the referral (first or repeat referral) did not alter the results. We conclude that using referral rates to judge referral quality is misleading. However, a blind and randomly performed panel review of referrals is a time consuming but feasible method of quality assessment.

Identifying factors associated with health care use: a hospital-based risk screening index

The purpose of the current study was to identify variables near hospital admission that could effectively discriminate patients at risk for nursing home placement, long hospital stay, or readmission. Risk factors reported in the literature were used to predict hospital outcome for 532 admissions. Factors that discriminated type of outcome included: two or more chronic medical conditions, living alone or being admitted from a nursing home, dependent ambulation, poor mental status, psychiatric comorbidity, prior admission, age over 75, and being unmarried. Using these criteria, an index was developed to determine risk for placement, readmission, or lengthy stay. Use of cumulative risk scores can result in accurate prediction of outcome and may be useful in targeting patients for intervention. Performance characteristics of the risk index are discussed.

Using diagnosis-related groups for studying variations in hospital admissions

The diagnosis-related groups (DRG's) have classically focused on resources consumed during a hospital stay. DRG's can also be considered categories for describing cases admitted to a hospital. In this article, we illustrate how consistent patterns of variations in admission rates can be used to classify DRG categories according to the Index of Discretionary Admissions. The consistency of variation in admission rates for modified DRG categories across hospital service areas in Iowa, California, Massachusetts, and Maine was high. The proportion of hospital admissions in the DRG's judged to be most discretionary ranged from 22 percent in Iowa to 14 percent in California.

Head injury with and without hospital admission: comparisons of incidence and short-term disability

All persons with head injuries (skull fracture or injury to the cranial contents resulting in a physician visit or at least one day of disability), regardless of treatment or hospital admission status, were identified from National Health Interview Survey data for the years 1977-81. Among those who reported such head injuries within the two weeks prior to interview, only 16 per cent were admitted to hospitals. Children, members of low-income families, and those injured at home, school, or in a recreational setting were less likely to be admitted to hospital than others. Among those who sustained a head injury in the previous three months and had some disability from that injury during the two weeks prior to interview, those not admitted to hospital included one-half of those with three to seven days of bed disability and one-third of those with more than seven days of bed disability; and they accounted for one-half of all disability days. These findings indicate that hospital-based head injury incidence data are incomplete and may contain substantial biases.

Geographic variations in elderly hospital and surgical discharge rates, New York State

Hospital and surgical discharges for 1981, as recorded by the uniform hospital discharge data of New York State, were aggregated by county of residence and converted into age-sex adjusted rates. Elderly hospital discharge and surgery rates in New York State, 1981, varied 2.4- and 2-fold, respectively. Discharge rates of elderly with specific surgical procedures showed even greater variation. However, proportions of highly complex and non-elective procedures performed on the elderly were similar in counties with high and low surgical rates. A multiple regression model consisting of independent dimensions of county demographic and medical resources characteristics plus a proxy variable for surgical practice styles was applied to hospital and surgery rates. Variations in elderly surgical discharge rates were found to be related to the supply of medical resources and to surgical practice styles.

Using administrative data for longitudinal research: comparisons with primary data collection

This paper discusses the advantages and disadvantages of using administrative data for longitudinal research, focusing on loss to follow-up. Comparisons between research relying on primary data collection and that using data bases are made. After development of a suitable framework, follow-up in several well-known projects based on primary data collection (the Seven Countries project on coronary heart disease, the Massachusetts research on long-term care and the Pittsburgh clinical trial of tonsillectomy) is compared with follow-up using the Health Services Commission data base in Manitoba, Canada. Overall follow-up in the Manitoba research compares favorably with participation and follow-up rates in other studies based on primary data collection. Initial nonresponse and nonlocation are major problems with studies using primary data; failure to locate earlier respondents in subsequent waves results in a wide range of overall response rates. Data bases do not require researchers to contact individuals and hence follow-up is simplified. Eight year follow-up rates in the Manitoba data base are almost always over 80% and often over 90%. Because records can be flexibly summarized for each individual over time, data bases facilitate certain types of longitudinal studies which would be difficult, if not impossible, to perform using other methodologies. If the desired data are available and recorded with acceptable accuracy, administrative data banks hold considerable promise for the health care researcher.

Using insurance claims to measure health status: the Illness Scale

Health insurance systems are generating large numbers of claims filed by physicians and hospitals for reimbursement and accounting purposes. This paper describes and evaluates a measure of health status derived from physician and hospital claims filed for a sample of older Canadians during 1970-1977. Information on the number, type, and seriousness of reported diagnoses and the number and duration of hospitalizations and surgeries during each year were combined to generate annual Illness Scales ranging from 0 to 24. Alpha coefficients, measures of internal consistency, were between 0.82 and 0.84. Consistent with high validity, Illness Scale scores increased with age, were significantly associated with other health measures, and were strongly predictive of death and hospitalization in the following year. The ability to develop valid and reliable health status measures from insurance claims substantially expands the potential use of these data for research and evaluation.

High Users of Hospital Days

Data from the Manitoba Longitudinal Study on Aging were used to focus on the elderly who used 31 or more hospital days in one year and to see if their high use was an isolated event or a persistent pattern by examining their hospital use over a 6-year period. The relationship of several service-provider characteristics to high use was also examined. The majority (58%) of high users were decedents or those about to enter a nursing home and another 22% used 31 or more days in only one of the six years. Indications are that bed/population ratios and physician practice styles are positively related to the hospitalization utilization patterns of high users. These and other findings raise issues requiring further research.