As mammography use increases, are some providers omitting clinical breast examination?

OBJECTIVE: To explore use of clinical breast examination (CBE) among women receiving mammography. METHODS: A retrospective cohort analysis of 100 women aged 50 years or older with at least one bilateral mammogram. Chart review documented demographic information, severity of illness, and performance of CBE (from 1 year prior to 18 months after the mammogram). RESULTS: The mean age of the 100 women was 63 years. They were predominantly unmarried (60%), nonwhite (58%), and not currently employed (57%). Three quarters (76%) had mammography and CBE (comprehensive screening), while the remaining 24% had mammography only. Sociodemographic factors did not differ for women with and without comprehensive screening (P>.1). However, patients of female providers were more likely to receive comprehensive screening than patients of male providers. Specifically, 95% of women seen by female attending physicians or fellows had comprehensive screening vs 67% for male attending physicians or fellows and 61% for residents (P=.008). CONCLUSIONS: Mammography may be replacing CBE especially among patients of male providers. Interventions targeted to these providers could help improve the use of CBE and mammography.

Does the survival advantage of nonwhite dialysis patients persist after case mix adjustment?

PURPOSE: Nonwhite dialysis patients survive longer than white patients; however, their clinical characteristics differ. We examined whether case mix differences explain the apparent survival advantage of nonwhite dialysis patients. SUBJECTS AND METHODS: We performed a prospective cohort study using data from the US Renal Data System Case Mix Severity Study that included 4,797 randomly selected dialysis patients 20 years of age and older who were followed up for up to 6 years. Demographic, comorbidity, laboratory, nutritional, and functional status data were obtained. Multivariable proportional hazards models adjusted for case mix differences between nonwhite and white dialysis patients. Additional analyses examined the effects of differences in transplantation rates, withdrawal from dialysis rates, and treatment modality selection. RESULTS: Unadjusted survival rates of black, Native American, and Asian or Pacific Islander dialysis patients were similar, and better than that for white dialysis patients. Relative to whites, the unadjusted relative risk (RR) for mortality among nonwhite patients was 0.64 (95% confidence interval [CI]: 0.58 to 0.70). Adjustment for case mix reduced, but did not eliminate, the survival advantage associated with nonwhite race (RR = 0.78, CI: 0.71 to 0.86). Adjustment for differences in transplantation rates (RR = 0.83, CI: 0.75 to 0.91), withdrawal from dialysis rates (RR = 0.81, CI: 0.73 to 0.90), and initial treatment modality (RR = 0.79, CI: 0.71 to 0.87) did not explain the lower mortality among nonwhites. CONCLUSIONS: A survival advantage for nonwhite dialysis patients persists after case mix adjustment. Future studies should explore additional physiologic and socioeconomic factors that might explain this difference.

Hospital cost of complications associated with coronary artery bypass graft surgery.

We identified 6,791 coronary artery bypass grafting (CABG) cases using the Massachusetts hospital discharge data to quantify the contribution of complications to the cost of a hospitalization for CABG. After adjusting for in-hospital mortality and baseline clinical severity as other contributors to cost, the additional costs associated with complications were substantial.

The importance of severity of illness adjustment in predicting adverse outcomes in the Medicare population.

The importance of using risk-adjusted mortality rates to measure quality of care is well-established. However, mortality rates may be an insensitive measure of quality for surgical patients since death is a relatively rare outcome. This study used Medicare files to identify, through chart abstraction, clinical postoperative complications of four surgical procedures (n = 8126) that could serve as measures of quality. Disease-specific severity of illness models using a moderate number of clinical variables and admission MedisGroups score models computed from approximately 250 clinical variables were compared in predicting postoperative adverse events. Initial differences between the two models disappeared upon cross-validation. Validated R-squareds and C statistics from models using half the data were generally positive, suggesting that these models had real, although modest, predictive power. We have shown that severity of illness on admission plays a role in predicting adverse events of surgery. Risk-adjusted outcomes may potentially be useful in screening for quality shortfalls.

Searching for an improved clinical comorbidity index for use with ICD-9-CM administrative data.

We studied approaches to comorbidity risk adjustment by comparing two ICD-9-CM adaptations (Deyo, Dartmouth-Manitoba) of the Charlson comorbidity index applied to Massachusetts coronary artery bypass surgery data. We also developed a new comorbidity index by assigning study-specific weights to the original Charlson comorbidity variables. The 2 ICD-9-CM coding adaptations assigned identical Charlson comorbidity scores to 90% of cases, and specific comorbidities were largely found in the same cases (kappa values of 0.72-1.0 for 15 of 16 comorbidities). Meanwhile, the study-specific comorbidity index identified a 10% subset of patients with 15% mortality, whereas the 5% highest-risk patients according to the Charlson index had only 8% mortality (p = 0.01). A model using the new index to predict mortality had better validated performance than a model based on the original Charlson index (c = 0.74 vs. 0.70). Thus, in our population, the ICD-9-CM adaptation used to create the Charlson score mattered little, but using study-specific weights with the Charlson variables substantially improved the power of these data to predict mortality.

Evaluating pressure ulcer occurrence in long-term care: pitfalls in interpreting administrative data.

Administrative databases for long-term care frequently collect information on fixed dates of the calendar year, rather than for entire episodes of care. Patients discharged or dying prior to an evaluation date are lost to follow-up. We used one such database, the VA Patient Assessment File, to examine pressure ulcer occurrence in long-term care. Clinical studies have established that most pressure ulcers develop during the first several weeks following admission. In these data, however, pressure ulcer development was less common in patients assessed within 2 months following admission, as compared to those examined at 3 to 6 months. This finding appears to be related to the selective discharge of patients, which makes these patient populations noncomparable. These results highlight that care must be exercised when interpreting results obtained from such administrative data.

Patterns of use of common major procedures in medical care of older adults.

OBJECTIVES: We analyzed Medicare data to determine the influence of age, mortality during the study year, and high individual Medicare costs on the frequency of hospitalization for the most common major procedures and for general medical diagnoses. METHODS: We used 1992 Medicare data to classify the most frequent diagnosis related groups (DRGs), representing one-half of all admissions, as procedural or general medical. We studied the frequency of hospitalization for the most common procedures (bowel surgery, cardiovascular procedures, hip and femur surgery) and for general medical care in the entire Medicare population, in persons aged 65 to 74, 75 to 84, and 85 or older, in those who died, and in the 5% of persons whose annual Medicare costs were the highest. RESULTS: In the entire population, common procedures accounted for 21% of the most frequent DRGs. Among those who died, the procedures were performed in only 8% of admissions. In the 5% of persons with the highest Medicare costs, 28% of admissions involved procedures. Admissions for cardiovascular procedures and for cancer chemotherapy decreased with age and were infrequent in persons 85 years and older. Hip and femur procedures increased with age and accounted for 70% of the common procedures in persons 85 years and older. CONCLUSIONS: The most common major procedures account for a minority of hospitalizations of persons more than age 65, of persons 85 and older, of those who died, and of persons with the highest Medicare expenditures. Most hospitalizations are for general medical care. Major procedures appear to be used with restraint in the very old and in persons in their last year of life.

Evaluation of a risk-adjustment model for pressure ulcer development using the Minimum Data Set.

OBJECTIVE: To validate a previously derived risk-adjustment model for pressure ulcer development in a separate sample of nursing home residents and to determine the extent to which use of this model affects judgments of nursing home performance. DESIGN: Retrospective observational study using Minimum Data Set (MDS) data from 1998. SETTING: A large, for-profit, nursing home chain. PARTICIPANTS: Twenty-nine thousand and forty observations were made on 13,457 nursing home residents who were without a pressure ulcer on an index assessment. MEASUREMENTS: We used logistic regression in our validation sample to calculate new coefficients for the 17 previously identified predictors of pressure ulcer development. Coefficients from this new sample were compared with those previously derived. Expected rates of pressure ulcer development were determined for 108 nursing homes. Unadjusted and risk-adjusted rates of pressure ulcer development from these homes were also calculated and outlier identification using these two approaches was compared. RESULTS: Predictors of pressure ulcer development in the derivation sample generally showed similar effects in the validation sample. The model c-statistic was also unchanged at 0.73, but it was not calibrated as well in the validation sample. On applying the model to the nursing homes, expected rates of ulcer development ranged from 1.1% to 3.2% (P <.001). The observed rates ranged from 0% to 12.1% (P <.001). There were 12 outliers using unadjusted rates and 15 using adjusted performance. Ten nursing homes were identified as outliers using both approaches. CONCLUSIONS: Our MDS risk-adjustment model for pressure ulcer development performed well in this new sample. Nursing homes differ significantly in their expected rates of pressure ulcer development. Outlier identification also differs depending on whether unadjusted or risk-adjusted performance is evaluated.

Deriving a risk-adjustment model for pressure ulcer development using the Minimum Data Set.

OBJECTIVE: To use the Minimum Data Set (MDS) to derive a risk-adjustment model for pressure ulcer development that may be used in assessing the quality of nursing home care. DESIGN: Perspective observational study using MDS data from 1997. SETTING: A large, for-profit, nursing home chain. PARTICIPANTS: Our unit of analysis was 39,649 observations made on 14,607 nursing home residents who were without a stage 2 or larger pressure ulcer on an index assessment. MEASUREMENTS: Pressure ulcer status was determined at an outcome assessment approximately 90 days after an index assessment. Potential predictors of pressure ulcer development were examined for bivariate associations, contributing to the development of a multivariate logistic regression model. RESULTS: A stage 2 or larger pressure ulcer developed in 2.3% of the observations. Seventeen resident characteristics were found to be associated with pressure ulcer development. These included dependence in mobility and transferring, diabetes mellitus, peripheral vascular disease, urinary incontinence, lower body mass index, and end-stage disease. A risk-adjustment model based on these characteristics was well calibrated and able to discriminate among residents with different levels of risk for ulcer development (model c-statistic = 0.73). CONCLUSION: A clinically credible risk-adjustment model with good performance properties can be developed using the MDS. This model may be useful in profiling nursing homes on their rate of pressure ulcer development.

Mammography use, breast cancer stage at diagnosis, and survival among older women.

BACKGROUND: Women age 65 years and older account for most newly diagnosed breast cancers and deaths from breast cancer. Yet, older women are least likely to undergo mammography, perhaps because mammography's value is less well demonstrated in older women. OBJECTIVE: To investigate the relationship between prior mammography use, cancer stage at diagnosis, and breast cancer mortality among older women with breast cancer. DESIGN: Retrospective cohort study using the Linked Medicare-Tumor Registry Database. SETTING: Population-based data from three geographic areas included in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program. PARTICIPANTS: Women aged 67 and older diagnosed with a first primary breast cancer, from 1987 to 1993, residing in Connecticut, metropolitan Atlanta, Georgia, or Seattle-Puget Sound, Washington. MEASUREMENTS: Medicare claims were reviewed and women were classified according to their mammography use during the 2 years before diagnosis: nonusers (no prior mammograms), regular users (at least two mammograms at least 10 months apart), or peri-diagnosis users (only mammogram(s) within 3 months before diagnosis). Mammography utilization was linked with SEER data to determine stage at diagnosis and cause of death. Our main outcome variables were (1) stage at diagnosis, classified as early (in situ/Stage I) or late (Stage II or greater), and (2) breast cancer mortality, measured from diagnosis until death from breast cancer or end of the follow-up period (December 31, 1994). RESULTS: Older women who were nonusers of mammography were diagnosed with breast cancer at Stage II or greater more often than regular users (adjusted odds ratio (OR), 3.12; 95% confidence interval (CI), 2.74-3.58). This association was present within each age group studied. Nonusers of mammography were at significantly greater risk of dying from their breast cancer than regular users for all women (adjusted hazard ratio (HR), 3.38; 95% CI, 2.65-4.32) and for women within each age group. Even assuming a lead time of 1.25 years, nonusers of mammography continued to be at increased risk of dying from breast cancer. Our findings remained significant for all women and for the two youngest age groups (67-74 years, 75-85 years), although the benefit was no longer statistically significant for the oldest women (85 years and older). CONCLUSIONS: Older women who undergo regular mammography are diagnosed with an earlier stage of disease and are less likely to die from their disease. These data support the use of regular mammography in older women and suggest that mammography can reduce breast cancer mortality in older women, even for women age 85 and older.

Variation in hospital rates of intraaortic balloon pump use in coronary artery bypass operations.

BACKGROUND: Little is known about regional patterns of intraaortic balloon pump (IABP) use in coronary artery bypass graft (CABG) operations. Our objectives were (1) to identify clinical variables associated with IABP use, and (2) to examine risk-adjusted rates of IABP use for 12 Massachusetts hospitals performing CABG operations. METHODS: We used hospital discharge data to identify 6944 CABG surgical cases. Logistic regression was used to identify clinical variables associated with IABP use, and the resulting multivariate model was then used to risk adjust hospital rates of IABP use. RESULTS: The IABP was used in 13.4% of the CABG surgical cases. The clinical variables independently associated with IABP use were cardiogenic shock, same admission angioplasty, prior CABG operation, cardiac arrest, congestive heart failure, recent myocardial infarction, and urgent admission status. Risk-adjusted rates of IABP use varied widely across hospitals from 7.8% to 20.8% (p < 0.0001). CONCLUSIONS: Hospital rates of IABP use vary considerably in Massachusetts. This practice variation may be related to the persistent uncertainty regarding the precise clinical indications for the IABP in this patient population.

Do severity measures explain differences in length of hospital stay? The case of hip fracture.

OBJECTIVE: To examine whether judgments about hospital length of stay (LOS) vary depending on the measure used to adjust for severity differences. DATA SOURCES/STUDY SETTING: Data on admissions to 80 hospitals nationwide in the 1992 MedisGroups Comparative Database. STUDY DESIGN: For each of 14 severity measures, LOS was regressed on patient age/sex, DRG, and severity score. Regressions were performed on trimmed and untrimmed data. R-squared was used to evaluate model performance. For each severity measure for each hospital, we calculated the expected LOS and the z-score, a measure of the deviation of observed from expected LOS. We ranked hospitals by z-scores. DATA EXTRACTION: All patients admitted for initial surgical repair of a hip fracture, defined by DRG, diagnosis, and procedure codes. PRINCIPAL FINDINGS: The 5,664 patients had a mean (s.d.) LOS of 11.9 (8.9) days. Cross-validated R-squared values from the multivariable regressions (trimmed data) ranged from 0.041 (Comorbidity Index) to 0.165 (APR-DRGs). Using untrimmed data, observed average LOS for hospitals ranged from 7.6 to 23.9 days. The 14 severity measures showed excellent agreement in ranking hospitals based on z-scores. No severity measure explained the differences between hospitals with the shortest and longest LOS. CONCLUSIONS: Hospitals differed widely in their mean LOS for hip fracture patients, and severity adjustment did little to explain these differences.

How much better can we predict dialysis patient survival using clinical data?

OBJECTIVE: To use three approaches to compare dialysis survival prediction based on variables included in the Standardized Mortality Ratio (SMR) with prediction based on a clinically enriched set of variables. DATA SOURCE: The United States Renal Data System Case Mix Severity data set containing demographic, clinical, functional, nutritional, and treatment details about a random sample of 4,797 adult dialysis patients from 291 treatment units, incident to dialysis in 1986 and 1987. STUDY DESIGN: This observational study uses baseline patient characteristics in two proportional hazards survival models: the BASE model incorporates age, race, sex, and cause of end-stage renal disease (ESRD); the FULL model includes these and additional clinical information. We compare each model's performance using (1) the c-index, (2) observed median survival in strata of predicted risk, and (3) predicted survival for patients with different characteristics. PRINCIPAL FINDINGS: The FULL model's c-index (0.709, 0.708-0.711) is significantly higher than that of the BASE model (0.675, 0.675-0.676), indicating better discrimination. Second, the sickest patients identified by the FULL model were in fact sicker than those identified as sickest by the BASE model, with observed median survival of 451 days versus 524. Third, survival predictions for sickest patients using the FULL model are one-third shorter than those based on the BASE model. CONCLUSIONS: The model with more detailed clinical information predicted survival better than the BASE model. Clinical characteristics enable more accurate predictions, particularly for the sickest patients. Thus, clinical characteristics should be considered when making quality assessments for dialysis patients.

An integer programming model to limit hospital selection in studies with repeated sampling.

OBJECTIVE: We describe an integer programming model that, for studies requiring repeated sampling from hospitals, can aid in selecting a limited set of hospitals from which medical records are reviewed. STUDY SETTING: The model is illustrated in the context of two studies: (1) an analysis of the relationship between variations in hospital admission rates across geographic areas and rates of inappropriate admissions; and (2) a validation of computerized algorithms that screen for complications of hospital care. STUDY DESIGN: Common characteristics of the two studies: (1) hospitals are classified into categories, e.g., high, medium, and low; (2) the classification process is repeated several times, e.g., for different medical conditions; (3) medical records are selected separately for each iteration of the classification; and (4) for budgetary and logistical reasons, reviews must be concentrated in a relatively small subset of hospitals. DATA COLLECTION/EXTRACTION METHODS. In each study, hospitals are ranked based on analysis of hospital discharge abstract data. CONCLUSIONS: The model is useful for identifying a subset of hospitals at which more intensive reviews will be conducted.

Specialty choices, compensation, and career satisfaction of underrepresented minority faculty in academic medicine.

PURPOSE: Despite efforts to increase the numbers of underrepresented minorities (URMs), only 3.9% of medical school faculty are URMs. The authors compared the specialty choices, compensation, and career satisfaction of minority faculty with those of their majority counterparts to determine whether there were differences that might affect the recruitment and retention of minority faculty. METHOD: In 1995, the authors mailed a self-administered survey to a stratified random sample of 3,013 eligible full-time salaried faculty in 24 randomly selected medical schools. Those schools, which had at least 200 faculty, did not include the Puerto Rican or historically black medical schools. RESULTS: Of the eligible faculty surveyed, 1,807 (60%) responded; 1,463 were majority faculty, 195 were URM faculty, and 149 were other-minority faculty. Similar proportions of the three groups were in the primary care specialties. Only 11% of the URM respondents were in basic science departments. There was no significant difference in adjusted mean compensation between majority, URM, and other-minority faculty. However, URM faculty were significantly less satisfied with their careers (adjusted scores: 60 versus > 65; p = .001) and more often considered leaving academic medicine within five years (58% versus < 45%). CONCLUSION: Given the demographic changes of the U.S. population, these issues should be addressed by deans and department heads in order to enhance recruitment and facilitate retention of URM faculty in academic medicine.

Fellowship training and career outcomes for primary care physician-faculty.

PURPOSE: To examine associations between fellowship training and career outcomes among primary care physician-faculty. METHOD: A total of 821 full-time primary care physician-faculty from 24 representative U.S. medical schools were surveyed using a self-administered questionnaire. Primary outcomes were recent grant submissions and funding, career referred publications, rank, and salary. Findings were adjusted for demographic and professional characteristics. RESULTS: Of the 500 respondents, 234 of the physician-faculty had completed a fellowship and 266 had not. Fellowship-trained physician-faculty were more than four times as likely to have submitted a grant proposal and to have had a grant funded (both p < 0.0001) than were physician-faculty without fellowship training. They were also more likely to have had any refereed publications (OR 3.8, p < 0.0001) and to have achieved senior academic rank (OR = 1.9, p = 0.02). Among those with fellowship experience, the amount of research training was important. Those with at least one year of research experience in their fellowship program had more grant proposal submissions (OR = 1.9, p = 0.02), more grants funded (OR = 2.9, p = 0.0003), more publications (OR = 2.4, p = 0.02), and higher academic ranks (OR 2.3, p = 0.03) than did those with less research training. Salaries were similar in every comparison. CONCLUSION: Fellowship-trained primary care physician-faculty were more productive researchers and were more likely to have achieved senior academic rank than were their no-fellowship-trained peers. Even among physician-faculty with fellowship experience, more research training was associated with higher productivity and rank. Salaries were not affected by training experience.

Junior faculty members' mentoring relationships and their professional development in U.S. medical schools.

PURPOSE: To determine (1) the prevalence of mentoring relationships for U.S. medical school junior faculty; (2) the quality of these mentoring relationships; (3) any variation by gender or race; and (4) the relationship between mentoring and junior faculty members' perceptions of institutional professional support; research-, teaching-, and clinical-skills development; allocation of time to professional activities; and career satisfaction. METHOD: In 1995 a 177-item survey was mailed to 3,013 full-time faculty at 24 randomly selected U.S. medical schools stratified on an area of medical specialization, graduation cohort, and gender. Mentoring was defined as "dynamic reciprocal relationship between an advanced career incumbent (the mentor) and a junior faculty member (the protégé) aimed at fostering the development of the junior person/protégé." Because mentoring is most crucial for junior faculty, the study focused on mentoring relationships within the previous three years ("recent mentoring") for faculty who were not full professors. Chisquare tests, analysis of variance, and principal-components analysis were used to analyze the data. RESULTS: In all, 1,808 (60%) of the 3,013 faculty surveyed, of whom 72% were junior faculty, returned completed questionaires. Fifty-four percent of the junior faculty had had a recent mentoring relationship. There was no significant difference between the men and the women faculty or between majority and minority faculty in the prevalence and quality of the mentoring relationships. The faculty with mentors rated their research preparation and research skills higher than did the faculty without mentors. Most of the women faculty (80%) and the minority faculty (86%) who had had mentors reported that it was not important to have a mentor of the same gender or minority group. CONCLUSION: Mentoring relationships are prevalent in academic medicine and should be promoted to support the career growth of junior faculty.

Using diagnoses to describe populations and predict costs.

The Diagnostic Cost Group Hierarchical Condition Category (DCG/HCC) payment models summarize the health care problems and predict the future health care costs of populations. These models use the diagnoses generated during patient encounters with the medical delivery system to infer which medical problems are present. Patient demographics and diagnostic profiles are, in turn, used to predict costs. We describe the logic, structure, coefficients and performance of DCG/HCC models, as developed and validated on three important data bases (privately insured, Medicaid, and Medicare) with more than 1 million people each.

Diagnosis-based risk adjustment for Medicare capitation payments.

Using 1991-92 data for a 5-percent Medicare sample, we develop, estimate, and evaluate risk-adjustment models that utilize diagnostic information from both inpatient and ambulatory claims to adjust payments for aged and disabled Medicare enrollees. Hierarchical coexisting conditions (HCC) models achieve greater explanatory power than diagnostic cost group (DCG) models by taking account of multiple coexisting medical conditions. Prospective models predict average costs of individuals with chronic conditions nearly as well as concurrent models. All models predict medical costs far more accurately than the current health maintenance organization (HMO) payment formula.

Principal inpatient diagnostic cost group model for Medicare risk adjustment.

The Balanced Budget Act (BBA) of 1997 required HCFA to implement health-status-based risk adjustment for Medicare capitation payments for managed care plans by January 1, 2000. In support of this mandate, HCFA has been collecting inpatient encounter data from health plans since 1997. These data include diagnoses and other information that can be used to identify chronic medical problems that contribute to higher costs, so that health plans can be paid more when they care for sicker patients. In this article, the authors describe the risk-adjustment model HCFA is implementing in the year 2000, known as the Principal Inpatient Diagnostic Cost Group (PIPDCG) model.