Printer Friendly

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.

Variations in hospital admission rates have been shown to exist across areas (whether hospital service areas, states or countries) and appear unrelated to the health characteristics of area residents (Andersen and Mooney 1990; Wennberg and Fowler 1977; Roos and Roos 1982). Considerably more studies document these variations than explain why they come about. However, a consensus has been growing around the following three related beliefs:

1. Uncertainty gives rise to different practice styles. Variation in hospital

admission rates is possible because of the uncertainty

surrounding medical practice, that is, the lack of knowledge

of what produces effective care (Wennberg, Barnes, and

Zubkoff 1982).

2. The variety of physician practice styles produces high and low rates.

Physician practice styles are responsible for observed variations

in rates across small areas. Surgical admissions most

clearly demonstrate this phenomenon (Wennberg 1979;

Wolfe and Detmer 1984; Roos 1983).

3. High rates indicate discretionary practices. High rates indicate

discretionary decisions and overuse, possibly unnecessary

use, of the health care system Wennberg et al. 1989; Roos,

Wennberg, and McPherson 1988; Paul-Shaheen, Clark, and

Williams 1987).

This study examines whether physicians who are more prone to admit patients to hospitals use hospitals for discretionary purposes and admit patients who are less ill than those admitted by other physicians. Previously I developed an index of individual Physicians' Hospitalization Practice Style (PHPS) using administrative data from the early 1970s (Roos et al. 1986). After controlling for factors related to health status, access to care, and so forth, a patient's probability of being hospitalized was strongly related to his or her physician's score on the index (Roos 1989). Using data from the 1980s from Manitoba, Canada, I redeveloped this index for the current study.


The PHPS index was constructed by: (1) analyzing the risk characteristics of patients contacting a given physician; (2) estimating the predicted rate of these patients being hospitalized, given the experience of patients with these risk characteristics across the province; (3) observing the actual rate at which a given physician admitted patients to a hospital; and (4) calculating physicians' scores on the PHPS index by comparing the proportion of patients admitted to a hospital with the expected admission rate given the provincial experience. A physician's propensity to admit was graded as lowest, medium-low, medium-high, or highest according to his or her score on the index.


To aid in examination, admissions were characterized using several measures suggested by the literature (Figure 1).

The high-variation medical conditions were those noted by Wennberg et al. (1989) as being associated with markedly higher admission rates in Boston than in New Haven, Connecticut; the two sites had similar rates for low- and moderate-variation conditions (stroke, heart attack, and gastrointestinal bleeding). The diagnoses involving high physician discretion were those judged by two physicians as having room for discretion in the decision to hospitalize (Anderson et al. 1986). Diagnoses classified as involving high discretion included such conditions as migraine, hypertensive heart disease, varicose veins, and hemorrhoids. Surgical procedures often performed on an outpatient basis were also identified (Roos 1988).

Type of admission (emergency versus nonemergency) has been used in several studies as a proxy for severity of illness (Dubois, Rogers, Moxley, et al. 1987; Munoz 1989). In Manitoba, all scheduled admissions to hospitals are coded as "elective" to distinguish them from nonscheduled urgent or emergency admissions. The predicted hospitalization rate (as estimated from the logistic regression) for each physician's admitted patients provided another measure of illness level (Hartz, Krakauer, Kuhn, et al. 1989). The higher the predicted rate of hospitalization, the higher the level of illness among a physician's patients. High-risk discharge diagnoses were also examined (Charlson et al. 1987); these diagnoses, used to control for the potential con-founding effects of concurrent illnesses in retrospective, chart-based analyses of readmissions and mortality, have been adapted for use with insurance claims data (Roos et al. 1989). Finally, the average diagnosis-related group (DRG) weight, developed for the U.S. prospective payment system (Jencks and Dobson 1987), is a widely recognized method of describing hospital case mix. DRG weights are based on use of resources and can be used to describe the mix of cases admitted to a hospital by a given physician. The lower the mean weight associated with admissions, the less complex the cases.


In Manitoba the cost of all medical and hospital care, with minor exceptions, is covered by a government health care plan. There is no limitation on use except for chiropractic care and optometrist visits. A complete history of physician visits, hospital admissions, and surgery can be reconstructed for each patient from health insurance (claims) data. Since out-of-province medical care is reimbursable by the Manitoba Health Services Commission and since physicians operate under a fee-for-service system, both patients and physicians have an incentive to document all use. The reliability and validity of the Manitoba claims data have been investigated extensively (Roos et al. 1982; Roos et al. 1989).


Figure 2 outlines the steps necessary to redevelop the PHPS index, following the method used earlier (Roos 1989). First, a representative sample of patients aged 65 years or older was drawn from the December 1982 registry of the provincial health insurance plan. For each patient in the sample with one or more physician visits between July 1, 1982 and June 30, 1984 (N = 32,904), the risk of admission to the hospital was estimated with logistic regression by modeling the patient's characteristics, including indicators of health status developed from claims. (For details on the variables used and the calculation of the index, see the Appendix.) Each patient was then classified as being at low, medium, or high risk of admission.

Second, for each patient in the sample, events surrounding the first hospital admission (if any) between July 1982 and the end of June 1984 were examined to identify the admitting physician. (Most physicians, including general practitioners, have admitting privileges for at least one hospital in Manitoba. Patients are admitted by their physicians or by appearing at the emergency room.) The hospital abstract identifies the attending physician as "that physician most responsible for the care of the patient," and this physician almost always (in 96 percent of cases) admitted the patient to the hospital. The principal physician is defined as the attending physician, if the patient contacted him or her before admission, or as the doctor contacted by the patient before admission who refers the patient to the attending physician. If the patient had no contact with the attending physician prior to admission, no identification of principal physician was made and the case was excluded. The "primary physician" measure appears to have been accurate in finding the physician who cared for the patient in the period before hospitalization and who therefore likely had the most influence on the decision to hospitalize. In 74 percent of cases, the principal physician was the doctor contacted most frequently by the patient over the two-year period 1982 to 1984. I also determined the number of patients in each risk category that each physician had seen over the period July 1982 through June 1984. If a patient had contacted two or more physicians, each physician was ascribed a visit for a patient of this risk status.

Third, I calculated each physician's score on the PHPS index by comparing the proportion of that physician's patients in each of the three risk categories actually admitted to the hospital with the proportion admitted provincewide. Scores were classified into four groups ranging from lowest to highest. Because the stability of this score is related to the size of a physician's practice, all analyses were weighted by practice size, as determined from the population sample.


Several conceptual and practical problems occur in developing an index of physicians' hospitalization practice style. It is appropriate to include only physicians who deliver primary care since it is difficult to adjust for differences in case mix across specialists and subspecialists. Some general practitioners lack hospital admitting privileges; others have relatively small practices, which makes it hard to accumulate enough patients, even in overall large samples, to draw conclusions.

To make this study as generalizable as possible, I included all physicians who deliver a significant amount of primary care in the province (internists, general surgeons, and general practitioners) (N = 995) in the analyses reported here. I Physicians were excluded from analysis if their practice included relatively few elderly patients (i.e., having fewer than 25 patients in the population sample) or if they lacked hospital admitting privileges. After exclusions, the practices of 571 physicians (27,483 patients) were included in the analyses (Figure 3).

To determine whether physicians with high scores on the PHPS index admitted different types of patients to the hospital compared with physicians with low scores, I used the first admission that occurred during the study period for patients from the population sample. Admissions were excluded if the physician who decided to admit could not be identified (often because the patient's attending physician was a specialist to whom no referral had been made by a primary care physician the patient had visited before), or if the physician making the decision to admit was excluded from analysis; 6,473 first admissions remained for analysis.

The comparison of admissions included and excluded showed few differences in the time of last contact with a physician before admission (after exclusion of patients with no contact), the type of admission (50.3 percent of the surgical admissions and 57.7 percent of the medical admissions were included), or the discharge diagnosis. However, a higher proportion of excluded admissions than included admissions were for cataract procedures (9.1 percent versus 1.0 percent).


The extent to which the primary patients of physicians in each of the four quartiles of the PHPS index were similar across factors that should influence hospitalization rates was examined directly (Table 1) and was compared with that of an independent sample of 2,931 patients aged 65 years or more interviewed in 1983. Since the data for these patients were not used in developing the PHPS index, this latter comparison provided important evidence about whether or not physicians with different practice styles treat patients with similar needs characteristics. For this independent sample, the principal physician (the physician seen most frequently in the year of the interview) was identified and his or her PHPS index score determined. Patients of physicians most prone to admit (i.e., in the top quartile of the score distribution) were more likely to report being restricted in the basic activities of daily living (15.5 percent versus 11.4 percent), and were likely to be somewhat older, to be less educated, and to have lower incomes than were patients of physicians who were least prone to admit (i.e., in the bottom quartile) (see Appendix Table A). However, no systematic differences were found in five other measures of patient health status: the number of health problems reported, a diagnosis of diabetes mellitus or heart disease, the ability to go outdoors in good weather, or self-perceived health status. Patients of physicians in different quartiles also had similar scores on the mental health status items.



Physicians with widely divergent scores on the PHPS index treated patients with remarkably similar characteristics (Table 1). Small distributional differences were found to be statistically significant owing to the large sample. Since all of the variables listed in Table 1 were used in the construction of the PHPS index, these data are presented only to illustrate the similarity of patient characteristics across practice style groups. During the six months before the study period the proportion of patients admitted to a hospital with high-risk conditions ranged from 4.3 percent for physicians least prone to admit to 6.1 percent for physicians most prone to admit. This finding likely reflects physician practice style as much as patient illness, since physicians with a high score had higher than expected rates of admitting patients of both low and high risk. During the study period (as well as during the next two years), both the probability of admission to a nursing home and the death rates were indistinguishable among patients treated by physicians with very different scores. Patients' scores on the ambulatory illness scale (Mossey and Roos 1987) (see Appendix) were also very similar across physicians with different hospitalization styles. When analyses were restricted to patients of general practitioners, the findings were unchanged except that patients of physicians with high PHPS scores had somewhat higher scores on the ambulatory illness scale in three of the four years (p = .05).

Despite the general similarities in patient characteristics, physicians most prone to hospitalize admitted 30.3 percent of their patients at least once during the study period, compared with 10.2 percent for physicians least prone to hospitalize (Table 2). For substantial numbers of patients in each practice style group (ranging from 11.0 percent to 15.0 percent), no contact with the attending physician had been made before admission and, therefore, it was unclear who made the decision to hospitalize (many of these patients may have been admitted through the emergency room). Overall, 74.9 percent of the patients of physicians least prone to hospitalize were not admitted, compared with 58.6 percent of the patients of physicians most prone to hospitalize.


Analysis of physician characteristics (Table 3) revealed that internists and urban general practitioners were overrepresented among physicians least prone to admit (35.4 percent and 47.6 percent, respectively). Rural general practitioners, in contrast, accounted for 69.7 percent of the physicians most prone to admit. Younger physicians accounted for a larger proportion of physicians least prone to admit, whereas older physicians accounted for a large proportion of those most prone to admit. The score on the PHPS index bore little relationship to the site of a physician's medical training.


Physicians with lower scores on the PHPS index tended somewhat to have a lower proportion of their admissions for high-variation conditions and for discretionary diagnoses (Table 4). The physician groups did not differ in the proportion of their admissions for surgery that could have been performed on an outpatient basis (such surgery accounted for less than 2 percent of the admissions overall).


Only 41.9 percent of the admissions by physicians most prone to admit were for a high-risk diagnosis, compared with 55.3 percent of such admissions for physicians least prone to admit (Table 5). Physicians most prone to admit had the lowest predicted rate of hospitalization among their admitted patients, suggesting that these patients were somewhat less ill with, on average, a 44 percent probability of being admitted, given the characteristics adjusted for in the logistic regression (prior admissions, proximity to death, etc.). Patients of physicians who were prone to admit also had a somewhat lower mean DRG weight associated with admissions. Differences in these last two measures were evident only between those most prone to admit and all others. Finally, although the relationship is somewhat irregular, physicians with medium-low and medium-high scores on the index admitted a higher proportion of their patients on an urgent or emergency basis that did physicians most prone to admit. When restricted to patients of general practitioners, the analyses were unchanged, with one exception: differences in the proportion of admissions on an urgent or emergency basis were not significant, but the direction of differences was similar and more consistent (18.1 percent, 15.3 percent, 15.7 percent, and 14.8 percent, respectively, for the four practice style groups, from least to most prone to admit).


In Table 6, ratios are calculated for the three variables in which the greatest differences across practice style groups were found. Physicians most prone to admit were 6 percent more likely to admit patients with a high-variation condition and 34 percent more likely to admit patients with a discretionary diagnosis than were physicians least prone to admit. Moreover, patients of physicians least prone to admit were 32 percent more likely to be admitted with a high-risk diagnosis. While these differences are in the expected direction, they are miniscule compared to the 197 percent difference in admission rates between these two groups of physicians.



The results of this study suggest that physicians who are more prone to admit patients to the hospital do admit more patients with discretionary diagnoses and more patients who are less ill, thus supporting the contention that high rates of hospitalization indicate discretionary physician practices. This belief is intuitively obvious from a logical model of health care delivery, and this is the model that has led to assumptions that high users are over-users (de Lissovoy et al. 1987). However, the most compelling finding of the current study is the similarity of patients admitted by physicians with markedly different practice styles.

This study has several strengths. First the analysis focuses on physician practice style. This emphasis is appropriate, given the assumption that varied practice styles are responsible for variations in rates of use across small areas. Second, much more care was taken to adjust for differences in case mix across physicians' practices than was done in other studies on variations in rates across small areas (Blumberg 1987; McLaughlin, Normolle, Wolfe, et al. 1989). I have demonstrated that physicians with very different hospitalization styles treat very similar patients. If anything, the illness levels of the patients of physicians most prone to admit were overestimated: the patients' hospitalization history (1980-1982) was used to estimate illness levels for 1983-1984, and since most patients would have had the same principal physician during this entire period, these 1980-1982 admissions were likely also to have been influenced by practice style. Finally, the measures of discretion and patient illness levels have face validity and have been used in other studies.

This study also has limitations. It is a claims-based study, and the physician responsible for the admission is identified from patterns of contact with and referrals to the patient's attending physician. These criteria have resulted in the exclusion of a substantial proportion of admissions from the analysis. In part this is because I excluded surgical specialists; it would be difficult to adjust their case mix to make it similar to that of the primary care physicians who are the focus of this study. However, in certain key areas the admissions analyzed were similar to those excluded. When the analysis was restricted even further to the hospitalization patterns of general practitioners, the results were unchanged.

Seventy percent of the physicians most prone to admit were general practitioners in rural Manitoba. No doubt their decisions were influenced by important factors other than individual practice style, including the distances patients must travel, low occupancy rates, and community pressure to keep hospital occupancy up lest a major source of employment be lost. However, almost a quarter of the physicians most prone to admit were general practitioners in urban areas, where pressure on beds is considerable.

Medical practice has long been recognized to vary from physician to physician. One current explanation for these variations assumes that deviations from a logical model of medical practice are caused by uncertainty (Eddy 1984), and thus, that when uncertainty is reduced or eliminated, the forces producing this variation will disappear. The evidence supporting this model is limited and is derived primarily from the observation that the conditions with the least variation in hospital admission rates, acute myocardial infarction and hip fracture, are conditions relatively easy to diagnose and for which a consensus on the need for hospitalization exists.

Moreover, if medical practice were logical, in areas where resources are limited (for example, where per capita rates are low), patients with more appropriate indications would be treated first. This logical model of medical practice is described by McPherson (1989), who suggests that "if one country has a rate twice that of another for a procedure, and if this can be attributed to practice style, it is reasonable to assume that 50 percent of the most seriously ill patients in the latter country will be receiving the procedure." This logic suggests that no less seriously ill patients are admitted in low-rate areas. Such a model does not fit the results reported here. Instead, we find surprisingly high rates of discretionary admissions and of admissions of less ill patients not only by physicians most prone to admit but also by those least prone to admit. While the differences between the two groups are in the expected direction (i.e., physicians most prone to hospitalize admitted about one-third more patients with discretionary diagnoses than those least prone to admit), this is considerably less than the overall threefold difference in the proportion of patients admitted. These results are consistent with those of Chassin, Kosecoff, Park, et al. (1987), who found that although the differences were in the right direction (a larger number of inappropriate procedures were done in high-rate areas), much larger differences showed up in rates of use than in appropriateness of use.

Such findings suggest a very wide latitude in physicians' decisions to admit patients; this latitude is not well captured by our concepts of discretion or need, or by a model that posits a logical relationship between physician treatment patterns and patient need (Wennberg 1987). Other studies support this conclusion (Leape, Park, Solomon, et al. 1989; Jessee, Nickerson, and Grant 1982). For example, Pakistan is a country with a severe shortage of surgeons, and logic would suggest that only the neediest get treated and that doctors forgo the luxury of discretionary care. However, marked variations were noted in surgical rates, across Pakistani hospital regions, that could not be explained by variations in patient need (Blanchard, Blanchard, Toussignant, et al. 1987). Extremely low population rates for such urgent procedures as caesarean section and appendectomy were observed at the same time as high rates for hemorrhoidectomy in some district hospitals. Closer to home, Siu, Sonnenberg, Manning, et al. (1986) reported from the RAND Health Insurance Experiment that, although cost sharing reduced the rate of admissions to hospital, it did not reduce the proportion of inappropriate admissions.

The current study suggests a need to revise our assumption that a logical model underlies medical practice. The sum of decisions made by individual physicians exercising their mandate to do what they believe is best for the patient cannot be assumed to result in appropriate medical care for the population and effective allocation of scarce health care dollars. Physician uncertainty as an explanation for variations in medical practice style is clearly inadequate. Other, probably more realistic models that attempt to explain how physicians (and everyone else) operate in worlds of microcertainty using idiosyncratic rules for decision making are becoming available and should be examined closely (Baumann and Deber 1991; Evans 1990; Bloor 1976; Dixon 1990). If these models correctly describe how physicians make decisions, efforts to improve the health care system must become very explicit.

Where should these efforts be focused? Highest priority should be given to educating the public, patients, and physicians about the implications of variations in medical practice, the uncertainty surrounding medical care, and the dangers of the "more is better" philosophy. Since these efforts represent an attempt to change the ethos of the health care system, this will be a process fraught with difficulties.

Resources should also be allocated to explicit monitoring of admission patterns at hospitals in high-rate areas. In Maine, where consensus on appropriate treatment exists, physician panels have made outlier physicians aware of their practice patterns, and high rates have fallen (Keller 1987).

Given the great uncertainty in medical practice, research on outcome assessment and consensus conferences are clearly needed to develop guidelines for appropriate practice (Ellwood 1988). But the present study, as well as others (Lomas 1988), demonstrates that reducing uncertainty offers only a minimal first step. In many situations physicians are capable of assessing appropriate care, but they do not pass judgment on a colleague's decisions unless empowered to do so (i.e., feedback of outlier results, triage in wartime, second-opinion surgery). Even in these situations, the effectiveness of treatment is likely to be overestimated. Resources must be found in the system to encourage education, monitoring, feedback, and outcomes assessment.




Modeling an Individual Patient Risk of Hospitalization

An estimate of each patient's risk of admission to a hospital between July 1982 and the end of June 1984 was calculated from the parameters of a logistic regression model relating individual characteristics to whether or not the patient was admitted. Patient histories were constructed from each of the 32,904 elderly patients chosen in a provincewide stratified probability sample. An original sample of 40,000 was drawn, but the data for 7,096 people were not usable because they had no contact with a physician over the study period. Patient characteristics available for use in the regression model included age, sex, and individual scores on the ambulatory illness scale (Mossey and Roos 1987; Roos et al. 1988) in each of four one-year periods beginning June 1980. This scale included three variables: (1) number of different diagnoses made by physicians on ambulatory visits, (2) number of visits with a diagnosis judged by a physician to represent a serious or urgent condition (Roos 1980), and (3) number of categories in which the patient made visits for diagnoses defined as increasing his or her "risk of not recovering from an illness" (Jones 1974). In addition, the number of different physicians visited in each two-year period beginning June 1982 was used. Variables describing the characteristics of a patient's admissions to hospital, including the number of discharges and the score on the Charlson Comorbidity Index (Charlson et al. 1987; Roos et al. 1989), between July 1980 and the end of June 1982 were included in the regression.

Hospital use is high in the period prior to admission to nursing home and prior to death. Separate dummy variables were therefore used to identify whether or not an individual died between July 1982 and June 30, 1984, or in the subsequent year, or whether he or she had entered a nursing home over either of these periods. To control for bed availability in a patient's area of residence, a measure of hospital beds per 1,000 residents was entered into the model. This was done both as an independent measure and as an interaction term combined with age of the patient, since we previously found that very elderly patients are more likely to be admitted to hospitals in areas that have more beds (Roos et al. 1986). Because specialists (particularly specialists frequently referred to by other physicians) would be expected to have a higher-risk case mix than other physicians, dummy variables were used to identify the specialty of the patient's primary physician (general surgeon or internist versus general practitioner) and whether the physician had received a large number of referrals over the study period.

Using the resulting parameters of the logistic regression model, I calculated a predicted risk of admission for each of the 32,904 patients. Patients were then classified as being at low, medium or high risk of admission. Every primary physician contacted by each of the patients between July 1982 and June 1984 was identified. This included those contacted over the entire period if the patient was never admitted or only those contacted before the first admission for those admitted. Each physician was then characterized by the number of patients at low, medium, and high risk for admission that he or she had seen during the study period.

Calculating the PHPS Index

The key to the development of the Index of Physicians' hospitalization Practice Style (PHPS) was identifying the physician who decided to admit the patient to the hospital. After reviewing many records, I defined this physician as that primary physician, general practitioner, internist, or general surgeon whom a patient contacted before entering the hospital and who then attended the patient in the hospital, or the physician who referred the patient to a physician who subsequently attended the patient in the hospital. A patient was excluded from the analysis when the decision to admit was made by a specialist with no clear link to the patient's primary physician or when a complex pattern of physician contact (or a complete lack of contact) before admission made it impossible to identify the physician responsible for the admission.

Each physician's score on the PHPS index was calculated by dividing the proportion of patients in each risk category actually admitted by the proportion of patients in each risk category actually admitted provincewide (i.e., the expected rate of admission). The three separate measures were then combined and weighted by the number of patients a physician had in each risk category. Most physicians delivering primary care in the province could then be characterized by a score indicating propensity to admit ranging from 0.05 to 4.9 (the latter score indicating that the physician was 4.9 times more likely to admit patients than the provincial average, after risk characteristics of the physician's patients were controlled for).

Table A shows the, characteristics of the 2,931 elderly patients in the independent sample by the PHPS Index score of their primary physician. These patients were first interviewed in 1971 as part of the Manitoba Longitudinal Study on Aging (Mossey et al. 1981) and were reinterviewed in 1983; only data from the second interview are reported here. The people interviewed in 1983 have been shown to be representative of the population of elderly in these age groups during 1983.



(1.) Given concerns that general surgeons and internal medicine specialists would care for very different types of patients than general practitioners, all analyses were repeated with these physicians and their patients excluded. The results were very similar, but where there were differences, they are so reported. The analyses are available from the author on request.


Andersen, T. F., and G. Mooney, eds. The Challenges of Medical Practice Variations. London: MacMillan Press, 1990. Anderson, G. F., J. C. Cantor, E. P. Steinberg, and J. Holloway. "Capitation Pricing: Adjusting for Prior Utilization and Physician Discretion." Health Care Financing Review 8, no. 2 (1986): 27-34. Baumann, A. O., and R. B. Deber. "Overconfidence among Physicians and Nurses: The |Micro-Certainty, Macro-Uncertainty' Phenomenon." Social Science Medicine 32, no. 2 (1991): 167-74. Blanchard, R. J. W., M. E. E. Blanchard, P. Toussignant, M. Ahmed, and C. M. Smythe. "The Epidemiology and Spectrum of Surgical Care in District Hospitals of Pakistan." American Journal of Public Health 77, no. 11 (1987):1439-45. Bloor, M. "Bishop Berkeley and the Adenotonsillectomy Enigma: An Exploration of Variation in the Social Construction of Medical Disposals." Sociology 10, no. 1 (1976): 43-61. Blumberg, M. S. "Inter-Area Variations in Age-Adjusted Health Status." Medical Care 25, no. 4 (April 1987): 340-53. Charlson, M. E., P. Pompei, K. L. Ales, and C. R. MacKenzie. "A New Method of Classifying Prognostic Comorbidity in Longitudinal Studies: Development and Validation." Journal of Chronic Disease 40, no. 5 (1987): 343-83. Chassin, M. R., J. Kosecoff, R. E. Park, C. M. Winslow, K. L. Kahn, N. J. Merrick, J. Keesey, A. Fink, D. H. Solomon, and R. H. Brook. "Does Inappropriate Use Explain Geographic Variations in the Use of Health Care Services?" Journal of the American Medical Association 258, no. 18 (13 November 1987): 2533-37. deLissovoy, G., T. Rice, J. Gabel, and H. J. Gelzer. "Preferred Provider Organizations One Year Later." Inquiry 24, no. 2 (1987): 127-35. Dixon, A. S. "The Evolution of Clinical Policies." Medical Care 28, no. 3 (1990):201-20. Dubois, R. W., W. H. Rogers, J. H. Moxley, D. Draper, and R. H. Brook. "Hospital Inpatient Mortality: Is It a Predictor of Quality?" New England Journal of Medicine 317, no. 26 (24 December 1987): 1674-80. Eddy, D. M. "Variations in Physician Practice: The Role of Uncertainty." Health Affairs 3, no. 2 (1984): 74-89. Ellwood, P. M. "Shattuck Lecture - Outcomes Management. A Technology of Patient Experience." New England Journal of Medicine 318, no. 23 (1988): 1549-56. Evans, R. G. "The Dog in the Night-Time: Medical Practice Variations and Health Policy." In The Challenges of Medical Practice Variations. Edited by T. F. Andersen and G. Mooney. London: MacMillan Press, 1990. Hartz, A. J., H. Krakauer, E. M. Kuhn, M. Young, S. J. Jacobsen, G. Gay, L. Muenz, M. Katzoff, R. C. Bailey, and A. A. Rimm. "Hospital Characteristics and Mortality Rates." New England Journal of Medicine 321, no. 25 (1989): 1720-25. Jencks, S. F., and A. Dobson. "Refining Case Mix Adjustment: The Research Evidence." New England Journal of Medicine 317, no. 11 (1987): 679-86. Jessee, W. F., C. W. Nickerson, and W. S. Grant. "Assessing Medical Practices through PSRO Cooperative Studies. An Evaluation of Caesarean Births in Nine PSRO Areas." Medical Care 20, no. 1 (1982): 75-84. Jones, E. W. "Patient Classification for Long-Term Care: User's Manual." DHEW Publication no. HRA 75-3107. Washington, DC: Department of Health, Education, and Welfare, 1974. Keller, R. "The Maine Experience: What Is Effective-Education, a Carrot or a Stick?" Proceedings of the Quality of Care Research Symposium. Working Paper Series 87-4. Baltimore, MD, Health Care Financing Administration, August 1987. Leape, L. L., R. E. Park, D. H. Solomon, M. R. Chassin, J. Kosecoff, and R. H. Brook. "Relation between Surgeons' Practice Volumes and Geographic Variation in the Rate of Carotid Endarterectomy." New England Journal of Medicine 321, no. 16 (1989): 653-57. Lomas, J. "Holding Back the Tide of Caesareans: Publishing Recommendations Is Not Enough to Stop the Rise." British Medical Journal 297, no. 6648 (3 September 1988): 569-70. McLaughlin, C. G., D. P. Normolle, R. A. Wolfe, L. F. McMahon, and J. R. Griffith. "Small-Area Variation in Hospital Discharge Rates: Do Socioeconomic Variables Matter?" Medical Care 27, no. 5 (1989): 507-21. McPherson, K. "International Differences in Medical Care Practices." Health Care Financing Review 11, no. 1 (1989): 9-20. Mossey, J. M., B. Havens, N. P. Roos, and E. Shapiro. "The Manitoba Longitudinal Study on Aging: Description and Methods." Gerontologist 21, no. 5 (1981): 551-58. Mossey, J. M., and L. L. Roos. "Using Insurance Claims to Measure Health Status: The Illness Scale." Journal of Chronic Disease 40, supplement 1 (1987): 41S-50S. Munoz, E. "The Impact of Diagnosis Related Groups and the Prospective Payment Assessment Commission." In Socioeconomics of Surgery. Edited by I. Rutkow. St. Louis, MO: CV Mosby Co., 1989. Paul-Shaheen, P., J. D. Clark, and D. Williams. "Small Area Analysis: A Review and Analysis of the North American Literature." Journal of Health Politics, Policy and Law 12, no. 4 (1987): 741-809. Roos, L. L. "Supply, Workload and Utilization: A Population-Based Analysis of Surgery." American Journal of Public Health 73, no. 4 (April 1983): 414-21. Roos, L. L., N. P. Roos, S. M. Cageorge, and J. P. Nicol. "How Good Are the Data? Reliability of One Health Care Data Bank." Medical Care 20, no. 3 (March 1982): 266-76. Roos, L. L., S. M. Sharpe, M. M. Cohen, and A. Wajda. "Risk Adjustment in Claims-Based Research. The Search for Efficient Approaches." Journal of Clinical Epidemiology 42, no. 12 (1989): 1193-1206. Roos, N. P. "Impact of the Organization of Practice on Quality of Care and Physician Productivity." Medical Care 18, no. 4 (April 1980): 347-59. _____. "Predicting Hospitalization Utilization by the Elderly. The Importance of Patient, Physician and Hospital Characteristics." Medical Care 27, no. 10 (1989): 905-19. _____. "What Is the Potential for Moving Adult Surgery to the Ambulatory Setting?" Canadian Medical Association Journal 138 (1 May 1988): 809-16. Roos, N. P., G. Flowerdew, A. Wajda, and R. B. Tate. "Variations in Physicians' Hospitalization Practices: A Population-based Study in Manitoba Canada." American Journal of Public Health 76, no. 1 (1986): 45-51. Roos, N. P., and L. L. Roos. "Surgical Rate Variations: Do They Reflect the Health or Socio-Economic Characteristics of the Population?" Medical Care 20, no. 9 (September 1982): 945-58. Roos, N. P., L. L. Roos, J. M. Mossey, and B. J. Havens. "Using Administrative Data to Predict Important Health Outcomes: Entry to Hospital, Nursing Home, and Death." Medical Care 26 (1988): 221-39. Roos, N. P., J. E. Wennberg, and K. McPherson. "Using Diagnosis-related Groups for Studying Variations in Hospital Admissions." Health Care Financing Review 9, no. 4 (1988): 53-62. Siu, A. L., F. A. Sonnenberg, W. G. Manning, G. A. Goldberg, E. S. Bloomfield, J. P. Newhouse, and R. H. Brook. "Inappropriate Use of Hospitals in a Randomized Trial of Health Insurance Plans." New England Journal of Medicine 315, no. 20 (1986): 1259-66. Wennberg, J. E. "Factors Governing Utilization of Hospital Services." Hospital Practice 14 (September 1979): 115-27. _____. The Paradox of Appropriate Care." Journal of the American Medical Association 258, no. 18 (1987): 2568-69. Wennberg, J. E., B. A. Barnes, and M. Zubkoff. "Professional Uncertainty and the Problem of Supplier Induced Demand." Social Science Medicine 16 (1982):811-24. Wennberg, J. E., and J. F. Fowler. "A Test of Consumer Contribution to Small Area Variations in Health Care Delivery." Journal of the Maine Medical Association 68, no. 8 (August 1977): 275-79. Wennberg, J. E., J. L. Freeman, R. M. Shelton, and T. A. Bubolz. "Hospital Use and Mortality among Medicare Beneficiaries in Boston and New Haven." New England Journal of Medicine 32 1, no. 17 (1989): 1168- 73. Wolfe, B. L., and D. E. Detmer. "The Economics of Surgical Signatures." Hospital Medical Staff 13, no. 10 (October 1984): 2-9.
COPYRIGHT 1992 Health Research and Educational Trust
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Roos, Noralou P.
Publication:Health Services Research
Date:Aug 1, 1992
Previous Article:Determinants of need and unmet need among cancer patients residing at home.
Next Article:Change in the Medicare case-mix index in the 1980s and the effect of the prospective payment system.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters