Geographic variation in expenditures for physicians' services in the United States

BACKGROUND

The national volume-performance standard recently implemented by Medicare does not account for geographic variation in expenditures for physicians' services. To study this variation, we examined expenditures for physicians' services in all metropolitan areas in the United States.

METHODS

We used Medicare claims data for 1989 to measure rates of service use for beneficiaries living in the 317 U.S. metropolitan statistical areas (MSAs). The variables investigated were rates of admission to the hospital, payments to physicians for inpatient care per admission and per beneficiary, payments to physicians for outpatient care per beneficiary, and overall payments to physicians per beneficiary. Expenditures were measured in terms of allowed charges as adjusted to reflect prevailing charges in each MSA. Rates of use were adjusted for age and sex, with the exception of the variable for payments to physicians for inpatient care per admission, which was adjusted for case mix.

RESULTS

Expenditures for the delivery of physicians' services to Medicare beneficiaries varied markedly among MSAs, with those for the areas with the lowest and the highest rates differing at least twofold on each measure. The measures for specific areas varied in parallel: areas with high rates of admission tended to have high levels of payment to physicians for inpatient care per admission, and areas with high payments for inpatient services tended to have high payments for outpatient services. Expenditures were not related to the number of physicians per capita but were lower in MSAs with a high proportion of primary care practitioners. The variation persisted when the 25 largest MSAs were examined; for total payments to physicians per beneficiary, there was a twofold difference between the area with the lowest rate and that with the highest, San Francisco ($872) and Miami ($1,874). The states with the highest overall payments to physicians per beneficiary were Florida, Louisiana, and Michigan.

CONCLUSIONS

The marked variation among metropolitan areas in payments to physicians underscores the lack of consensus among physicians about which services are required. Moreover, the practice style in a given community appears to be influenced not by the aggregate supply of physicians but rather by the mixture of primary care physicians and specialists.

A method for screening the quality of hospital care using administrative data: preliminary validation results

Applying a computerized algorithm to administrative data to help assess the quality of hospital care is intriguing. As Iezzoni and colleagues point out, there are major differences of opinion as to the worth of such efforts. This article significantly advances the state of the art in using administrative data to screen for potential quality-of-care problems. In addition, this work on identifying complications of care goes well beyond the emphasis of many government organizations on hospital mortality rates. One question, however, not raised in the paper is: What is a practical upper limit to the sensitivity and specificity in comparing computerized screen results with the consensus judgments of a group of independent physicians? Advanced statistical techniques (such as bootstrapping) might be used to estimate the stability of consensus judgments by physician groups. When the judgments of two groups of physicians are compared with each other, the resulting sensitivity and specificity will not be .99! In addition, more training of members of the physician panels would probably have increased interrater reliability. While acknowledging this problem, the researchers' detailed analysis of the panel results is intriguing and represents a model for such studies. It is hoped that the authors will follow up on the avenues opened here. Furthermore, what degree of accuracy is necessary to identify facilities with higher-than-expected rates of complications? The authors discuss problems involved in using administrative data to target hospitals and departments for more costly in-depth reviews of quality. It is hoped that the promising findings that are reported here will be validated in other studies. Certainly their algorithms should find a ready audience in insurers and hospitals willing to try them out. Finally, should we expect additional research to lead to improvement in the authors' algorithms? I believe the algorithms will prove difficult to improve upon; but perhaps we should not worry about this. At some point, however, the cost of trying to identify and correct quality problems in "minimally outlier" hospitals will exceed the benefits, particularly given alternative uses for the funds. Might we now be close the the "flat of the curve" in the development of such systems for identification of quality problems? This issue should be discussed much further in future studies.

Studying outcomes and hospital utilization in the elderly. The advantages of a merged data base for Medicare and Veterans Affairs hospitals

That veterans aged 65 years and older are eligible to receive care either in the Veteran Affairs (VA) health care system or in the private sector under Medicare confounds the analysis of veterans' health services utilization and outcomes in two ways. First, changes in eligibility or financial barriers to access with regard to either system influence veterans' decisions about where to seek needed care. Second, analyses of VA care for elderly veterans that rely solely on VA data sources underestimate both overall utilization and treatment complications. Similarly, failure to consider the contribution of health care delivery in the VA system may confound analyses of health care utilization by the Medicare-eligible population. To study the magnitude of such confounding influences, we linked the Medicare and VA health care administrative databases for residents of New England and New York. Results indicated that, for ten surgical procedures commonly performed in the elderly, as well as for hospitalizations resulting from acute myocardial infarction and hip fracture, VA patients receive from 17.6% to 37.4% of hospital care outside the VA system. Private hospitalizations account for 5.5% to 19.5% of the care received by veterans within 6 months after an initial episode of care in a VA hospital. It was also found that initial hospitalizations for study conditions in the VA accounted for 3.6% of all such hospitalizations among elderly Medicare-eligible men. Although overall hospital utilization appears to be underestimated in VA data sources, it was found that ascertaining mortality from sources available within the VA produced excellent results when compared with deaths recorded in the Medicare enrollment files. A national, merged VA-Medicare data base is feasible and would enhance the validity of analyses of health care delivery both for elderly veterans and for the Medicare population.

Comorbidities, complications, and coding bias. Does the number of diagnosis codes matter in predicting in-hospital mortality?

OBJECTIVE

Incomplete coding of secondary diagnoses may bias assessments of patient risks of poor outcomes using administrative health care databases, most of which allow only five diagnoses. The Medicare program is expanding the number of possible diagnoses from five to nine, aiming to improve coding completeness. We examined the impact of having more diagnosis codes available on assessments of risk of death.

DESIGN

We used 1988 computerized hospital discharge abstract data from California, which allow up to 25 diagnoses per discharge, to select a sample of hospitalized patients and assessed the relationship between the presence of 29 specific secondary diagnoses and the risk of in-hospital death.

SETTING

Nonfederal acute-care hospitals in California.

STUDY POPULATION

All patients at least 65 years of age who were hospitalized for stroke, pneumonia, acute myocardial infarction, or congestive heart failure in California in 1988 (N = 162,790).

MAIN OUTCOME MEASURES

Relative risk of death for each specific secondary diagnosis.

RESULTS

Many conditions that on a clinical basis would be expected to increase the risk of death, such as adult-onset diabetes mellitus, previous myocardial infarction, angina, and ventricular premature beats, were associated with a lower risk of in-hospital death.

CONCLUSIONS

Bias against coding of chronic or comorbid conditions on the computerized discharge abstracts of patients who die best explains these results. Efforts to improve diagnosis coding completeness solely by increasing the number of available coding spaces may not succeed.

Prioritizing Oregon's hospital resources. An example based on variations in discretionary medical utilization

OBJECTIVE

To provide an alternative to Oregon's treatment-specific approach to rationing, we propose a prioritization based on the local hospital resources invested in discretionary medical admissions.

DESIGN

We used 1988 Oregon hospital discharge data to determine age- and sex-adjusted per-capita rates of inpatient days for discretionary medical admissions (for high-variation medical conditions) in each of 33 hospital service areas. Potential ceiling rates were defined based on prevailing utilization rates for discretionary medical admissions in each hospital service area. Savings were calculated under the assumption that resources allocated for inpatient treatment of these conditions in areas that exceed the ceiling rates were reduced accordingly.

SETTING

Nonfederal, acute-care hospitals used by Oregon residents.

STUDY POPULATION

Oregon residents.

MAIN OUTCOME MEASURES

Savings were defined in terms of patient days, hospital beds, hospital charges, and average costs.

RESULTS

Among the 16 largest hospital service areas, patient-day rates for discretionary medical admissions ranged from 188 to 335 patient days per thousand. Potential savings from applying different ceiling rates ranged from $0.4 million to $94.7 million per year. If the rate in the state capital (Salem) were used as the ceiling (218 days per thousand), then 238 beds could be closed in 20 hospital service areas, for an estimated cost savings of $47.3 million.

CONCLUSIONS

Hospital resources invested in discretionary admissions in high-rate areas represent an important potential source of funds for reallocation to meet other defined health care needs. Setting limits based on units of health care supply (eg, beds, capital equipment, and physicians) should be considered as an option for resource reallocation within health care.

The accuracy of Medicare's hospital claims data: progress has been made, but problems remain

BACKGROUND

Health care databases provide a widely used source of data for health care research, but their accuracy remains uncertain. We analyzed data from the 1985 National DRG Validation Study, which carefully reabstracted and reassigned ICD-9-CM diagnosis and procedure codes from a national sample of 7050 medical records, to determine whether coding accuracy had improved since the Institute of Medicine studies of the 1970s and to assess the current coding accuracy of specific diagnoses and procedures.

METHODS

We defined agreement as the proportion of all reabstracted records that had the same principal diagnosis or procedure coded on both the original (hospital) record and on the reabstracted record. We also evaluated coding accuracy in 1985 using the concepts of diagnostic test evaluation.

RESULTS

Overall, the percentage of agreement between the principal diagnosis on the reabstracted record and the original hospital record, when analyzed at the third digit, improved from 73.2% in 1977 to 78.2% in 1985. However, analysis of the 1985 data demonstrated that the accuracy of diagnosis and procedure coding varies substantially across conditions.

CONCLUSIONS

Although some diagnoses and all major surgical procedures that we examined were accurately coded, the variability in the accuracy of diagnosis coding poses a problem that must be overcome if claims-based research is to achieve its full potential.

Computer simulations of cell-target encounter including biased cell motion toward targets: single and multiple cell-target simulations in two dimensions

In order for immune cells to carry out many of their functions, including clearance of infectious agents from tissue, they must first encounter their targets in the tissue. This encounter process is often the rate-limiting step in the overall function. Most immune cells exhibit chemotactic ability, and previous continuum models for encounter rates and dynamics have shown that chemotaxis can be a great advantage to cells by greatly increasing encounter rates relative to those for randomly moving cells. This paper describes computer simulations of discrete cell-target encounter events in two dimensions, for the two cases considered by the continuum models: where only a single cell and a single target are present, and where many cells and targets are present. The results of these simulations verify our previous model predictions that a small amount of chemotactic bias dramatically decreases the encounter time, while further increases in the amount of bias have a much smaller effect. Chemotactic ability is shown to be an important determinant of the kinetics of target clearance, and its effects depend on the initial cell-target ratio and the initial distributions of cells and targets. To the best of our knowledge, this work provides the first computer simulations of particle-target encounter in which there is biased motion of particles toward their targets, and is therefore of general interest beyond specific application to immune cell function.

Hip fracture incidence and mortality in New England

We used Medicare data to conduct a population-based study of osteoporotic hip fracture incidence and outcomes among New England residents. To reduce bias and improve data reliability, we combined data from multiple files; we found that 6% of cases would have been missed had we relied on hospital claims alone. Hip fracture incidence (per 1,000 person-years) increased for white females from 2.2 for ages 65-69 to 31.8 for ages 90-94 and for white males from 0.9 for ages 65-69 to 20.8 for ages 90-94. Incidence among blacks was lower in all age/sex groups. The female/male relative risk was greater among whites than among blacks. Case fatality following hip fracture was 12.5% at 90 days and 23.7% at 1 year and was higher among males, older patients, and those who had documented comorbidity or who were residents of nursing homes.

Overcoming potential pitfalls in the use of Medicare data for epidemiologic research

We used Medicare data bases and US Census data to address two questions critical to the use of Medicare files for epidemiologic research. First, we examined the degree to which the population enrolled in the Medicare program is similar to the elderly resident population of the United States, as estimated by the US Census. We found small differences in the total population estimates but substantial differences by age and race. Second, we found that among Medicare enrollees, physician claims identify a small proportion of hip fracture cases which are not documented in the hospital discharge files. This proportion varies by age, region, and state within the United States. Calculation of rates based on Medicare hospital discharge data, and probably other hospital discharge data sets as well, must take these limitations into account. Use of all available Medicare data files can overcome these limitations.

Further study of the increased mortality following transurethral prostatectomy: a chart-based analysis

Previous studies using large administrative databases found an elevated relative risk of reoperation and death after transurethral resection of the prostate compared to open prostatectomy. To investigate whether differences in case-mix unmeasured by administrative data explained this finding, we reviewed the charts of 485 patients who had undergone prostatectomy (236 open and 249 transurethral) at the Health Science Centre, Winnipeg, Manitoba, Canada between 1974 and 1980. Data from patient histories, physical examinations and laboratory evaluations were abstracted and used to control for case-mix in models comparing the rates of reoperation and mortality after transurethral versus open prostatectomy. Several models were specified. In all models the relative risk of dying after transurethral prostatectomy remained elevated (1.36 to 1.89), as did the risk for reoperation (3.62). A prospective trial is needed to establish the relative safety and effectiveness of transurethral and open prostatectomy.

Analysis of the effects of immune cell motility and chemotaxis on target elimination dynamics

White blood cells of the immune system must encounter specific targets such as bacteria, malignant cells, virus-infected cells or other cells of the immune response in order to carry out their function of protecting the host from infectious and malignant disease. To analyze the dynamics of this process, a mathematical model has been developed for elimination of proliferating targets by a constant population of motile immune system cells in two dimensions. Encounter is assumed to be the rate-limiting step for elimination. This model makes use of a previously derived analysis of single cell-target encounter times, which yields an encounter rate constant that is incorporated into a kinetic conservation equation for target number density. This paper focuses on the influence of directed cell movement, or chemotaxis, as well as other cell motility properties, such as cell speed and persistence, on target elimination dynamics. A particularly significant result is that a given relative decrease in chemotactic responsiveness leads to much more severe deficiencies in target clearance rates for low levels of baseline chemotactic responsiveness than for high levels of baseline responsiveness. The general model results are then applied to the particular example of bacterial clearance from the lung surface by alveolar macrophages. It is shown that moderate levels of macrophage chemotactic responsiveness, similar to those measured in vitro, can account for the experimentally observed rates of bacterial elimination from the lung for typical values of bacterial specific growth rate and alveolar macrophage number density.

Technology assessment using insurance claims. Example of prostatectomy

This article describes the findings of an ongoing assessment of prostatectomy that relied on the use of administrative data bases. Examples of the use of claims data for monitoring outcomes and treatment comparisons are provided, as well as a discussion of the strengths and limitations of administrative data for technology assessment.

Risk of carotid endarterectomy in the elderly

We used the Medicare claims files to describe operative mortality for 2,089 New England residents over the age of 65 who underwent carotid endarterectomy in 1984 and 1985. For patients ages 65 to 69, the risk of death within 30 days of surgery was 1.1 percent, (95% confidence interval = 0.5, 2.1), for those ages 70 to 74, 2.8 percent (1.7, 4.4), for those ages 75 to 79, 3.2 percent (1.8, 5.2), and for those over age 80, 4.7 percent (2.3, 8.5). Nearly 80 percent of patients underwent surgery at hospitals performing 40 or fewer carotid endarterectomies per year on the Medicare population. The adjusted odds ratio for 30 day mortality for patients undergoing surgery in these low-volume hospitals was 2.8 (95% CI = 1.1, 7.2) compared to higher volume hospitals. Although the Medicare claims data provided only limited data about post-operative strokes, analysis of post-operative stroke risk supported these findings.

Mortality and reoperation after open and transurethral resection of the prostate for benign prostatic hyperplasia

As part of an ongoing effort to evaluate alternative treatments for benign prostatic hyperplasia, we compared the outcomes of transurethral resection of the prostate with those of open prostatectomy. Men undergoing prostatectomy in Denmark (n = 36,703), Oxfordshire, England (n = 5284), and Manitoba, Canada (n = 12,090), were identified retrospectively through administrative data and followed for up to eight years. The cumulative percentage of patients undergoing a second prostatectomy was substantially higher after transurethral than after open prostatectomy (12.0 vs. 4.5 percent in Denmark, 12.0 vs. 1.8 percent in Oxfordshire, and 15.5 vs. 4.2 percent in Manitoba). The long-term age-specific mortality rates associated with transurethral prostatectomy as compared with open prostatectomy were also elevated in each country. The data on 1650 Canadian patients were used to investigate the contribution of coexisting morbid conditions to the elevated risk of death. The relative risk was 1.45 (95 percent confidence interval, 1.15 to 1.83) before risk adjustment and 1.45 (95 percent confidence interval, 1.15 to 1.84) after adjustment; the higher mortality was seen among low-risk as well as high-risk patients. These findings suggest that transurethral prostatectomy is less effective in overcoming urinary obstruction than the open operation. Our data also raise the possibility that transurethral prostatectomy may result in higher long-term mortality, although we cannot rule out potential confounding effects of unmeasured characteristics of patients.

The motile response of alveolar macrophages. An experimental study using single-cell and cell population approaches

In this report, we studied the applicability of a random walk model of individual cell motility in predicting the motile behavior of alveolar macrophage populations under agarose. The migration of a population of cells in the absence of a chemotactic or chemokinetic gradient can be characterized by the random motility coefficient, mu, which is analogous to a particle diffusion coefficient. Random walk theory relates this latter coefficient to particle speed and collision time (equivalent to the time between changes in particle direction). By analogy, according to a similar random walk theory for cell migration, mu for a cell population is a function of the speed and persistence time (with direction changes governed by cell behavioral processes rather than by collisions) of individual cells. To test the model, normal guinea pig alveolar macrophages were incubated in the presence or absence of uniform concentrations of the chemotactic tripeptide formyl-norleucyl leucyl phenylalanine (FNLLP) to elicit different levels of motile activity. Mu was calculated from cell population density profiles obtained by fixing and staining cultures after 2, 3, or 4 days. In parallel experiments, individual cell speeds and persistence times were measured from 1-h, time-lapse video microscopy recordings. The value of mu calculated from single-cell measurements was in good agreement with that from population studies for stimulated random migration (at 10(-7) to 10(-11) M FNLLP), but not in the absence of stimulant. Overall, these results support the applicability of the random walk model of individual cell migration to randomly migrating alveolar macrophage populations.

The effect of alveolar macrophage chemotaxis on bacterial clearance from the lung surface

In the distal airways, the alveolar macrophage plays a crucial role in defense of the lung against inhaled pathogens. These cells have been observed in vitro to move chemotactically in response to many types of attractants that may be present on the lung's surface during a bacterial or particulate challenge. This study investigated the hypothesis that chemotactic ability is an important part of the defensive action of these cells as they ingest bacteria on the lung surface. We compared our mathematical model for lung clearance to previously published bacterial clearance data and determined the amount of alveolar macrophage chemotactic ability required to account for observed clearance rates. The results showed that random motion is insufficient for clearance, and a moderate amount of chemotactic ability is necessary for our predicted clearance rates to agree with experimentally measured clearance rates.