Double-reading of mammograms does not necessarily improve diagnostic accuracy

Double-reading of the same screening mammogram by two radiologists has been proposed as a way to increase accuracy. By pairing with another radiologist, the average radiologist can expect an 8 percent to 14 percent gain in the true-positive rate (TPR), that is, diagnosis of cancer that is proven to be cancer. But this is at the expense of a 4 percent to 10 percent increase in the false-positive rate (FPR)—a diagnosis of cancer that is proven not to be cancer—according to a study supported by the Agency for Health Care Policy and Research (HS07845). For some radiologists, double reading increased the TPR with a small concomitant increase in FPR. Other radiologists, however, realized only small gains in TPR with large increases in FPR.

Not all radiologists complement each other diagnostically, and some add little, if anything, to the accuracy of the other. Even adding the reading from a more experienced radiologist does not necessarily improve the TPR of a less experienced radiologist. That is why decisions on double-reading of mammograms must be made on a practice-by-practice basis, explains Craig A. Beam, Ph.D., of the Medical College of Wisconsin. Dr. Beam and his colleagues based their findings on interpretation of the same set of 79 randomly selected screening mammograms by pairs of radiologists at 24 facilities.

The researchers compared the effect of double reading on the diagnostic accuracy of the readings. They found that the actual effect depended on the initial sensitivities, specificities, and degree of diagnostic concordance of the two radiologists whose interpretations were combined. They point out that U.S. radiologists vary widely in the accuracy with which they read mammograms, and that understandably, the effect of double-reading might vary widely. For this reason, they caution radiologists who may be considering the use of double-reading to learn the individual radiologists TPRs and FPRs and the degree of their diagnostic concordance.

See "Effect of human variability on independent double reading in screening mammography," by Dr. Beam, Daniel C. Sullivan, M.D., and Peter M. Layde, M.Sc., in Academic Radiology 3(11), pp. 891-897, 1996.

Public disclosure of hospital and physician CABG mortality rates has little effect on the types of patients they treat

Since 1990, the New York State Department of Health has made public the mortality rates of patients undergoing coronary artery bypass graft (CABG) surgery by individual hospital and surgeon. So far, making these records public has not prompted New York surgeons or hospitals to turn away the most risky CABG patients in order to improve their public performance ratings, according to a recent study. However, with the movement toward further dissemination of patient outcomes to health insurance companies, other payers, and the public, providers may become progressively more wary of high-risk cases, warns Edward L. Hannan, Ph.D., lead author of the study. The study was supported in part by the Agency for Health Care Policy and Research (HS06503).

The researchers analyzed New York State registry data on CABG patients of 31 hospitals and 87 surgeons from 1990 to 1992. They found that, after adjusting for patient risk status, the mortality rate following CABG for all high-risk patients (predicted death rate of at least 7.5 percent) was slightly lower than for other patients (2.94 percent vs. 3.02 percent). In general, hospitals and surgeons with the lowest risk-adjusted mortality rates for all cases also had the lowest mortality rates for high-risk cases. This indicates no systematic bias against operating on high-risk CABG patients in New York.

In fact, 16 of the 31 hospitals had risk-adjusted mortality rates for the high-risk cases that were lower than the overall State-wide mortality rate of 2.99 percent. And 8 of the 12 hospitals with more than 100 high-risk patients had a risk-adjusted mortality rate less than 2.99 percent. Of the 87 surgeons who performed at least 200 CABG surgeries from 1990 to 1992, 52 percent had risk-adjusted mortality rates for high-risk patients that were at least as low as the State-wide mortality rates, and 47 percent had rates for all patients (including those at high risk) that were as low as those of low-risk patients.

These findings suggest that there is no more reason for providers to avoid operating on high-risk patients than there is to avoid any other patients, conclude the authors. They suggest that future studies should include the development of new and improved methods for investigating bias against high-risk cases and that methods for disseminating such information to providers be improved.

See "Assessment of coronary artery bypass graft surgery performance in New York," by Dr. Hannan, Albert L. Siu, M.D., M.S.P.H., Dinesh Kumar, M.S., and others, in Medical Care 35(1), pp. 49-56, 1997.

Quality of physician/patient interaction may suffer when patients are anxious or very sick

A patient's health status influences the way physicians and patients interact with each other. Physicians find visits with sicker, frequently anxious patients, to be unpleasant and less satisfying than visits with healthier patients, and they often behave negatively toward sicker patients, according to a study supported by the Agency for Health Care Policy and Research (HS06999). During visits with these patients, doctors tend to get right down to the "business" of the visit. In turn, the sicker patient is often distressed or angry, and few social amenities are exchanged.

This scenario was evident in videotaped and audiotaped visits of 1,300 patients during routine outpatient visits to more than 250 physicians in four studies analyzed by Judith A. Hall, Ph.D., of Northeastern University, Boston, MA, and her colleagues. Generally, the less healthy the patient, the less the physician engaged in social conversation, agreed with the patient, or exhibited positive behavior. At the same time, the patients themselves conveyed more negativity and concern to their physicians than healthier patients through their verbal and nonverbal cues. Despite this seemingly negative interaction between physicians and their sicker patients, physicians nevertheless appeared intently engaged, both in terms of their own talk and in the way they encouraged the patient to talk, and showed concern for the patient during the visit.

The relative lack of social conversation could reflect a more serious, down-to-business attitude in keeping with the realities of sicker patients medical needs rather than psychological avoidance, explain the researchers. They suggest that increased attention by physicians to their own and their patients behavior may enhance diagnosis and prevent misunderstandings.

For more information, see "Patients' health as a predictor of physician and patient behavior in medical visits," by Dr. Hall, Debra L. Roter, Dr.P.H., Michael A. Milburn, Ph.D., and Lawren H. Daltroy, Dr.P.H., in Medical Care 34(12), pp. 1205-1218, 1996.

Caring for depressed patients in a changing health care system challenges providers and policymakers

One of the major concerns about the profound changes taking place in the U.S. healthcare system is whether they will improve or diminish the quality and cost-effectiveness of medical care. The shift from a fee-for-service to a prepaid method of reimbursement has greatly changed the incentives of patients to seek care as well as those of providers to supply it. This change poses a particular challenge for depressed patients, a vulnerable population that often does not advocate for its own care.

Although depression is a major illness affecting millions of people, it is seriously undertreated in the United States. Depression can be more devastating than other illnesses because it often begins at a younger age, when people are at their productive peak. It also differs from many medical conditions in that its indirect costs—i.e., lost wages and productivity—are usually higher than direct treatment costs.

A newly published book, Caring for Depression, documents what researchers learned about depression from the Medical Outcomes Study. The Medical Outcomes Study was a 4-year project that was conducted between 1986 and 1990 and involved more than 20,000 patients and 500 clinicians in alternative systems of care in three urban areas (Boston, Chicago, and Los Angeles).

Caring for Depression was written by Kenneth B. Wells, M.D., M.P.H., Professor of Psychiatry at the University of California, Los Angeles Neuropsychiatric Institute and School of Medicine, and a senior scientist at RAND; Roland Sturm, Ph.D., an economist at RAND; Cathy D. Sherbourne, Ph.D., a medical sociologist at RAND; and Lisa S. Meredith, Ph.D., a social psychologist at RAND. In the book, they discuss the inefficiencies of the current health care systems—both prepaid and fee-for-service—and explore how these systems can be improved. The research on which this book was based was supported in part by the Agency for Health Care Policy and Research (HS06802).

The book will be useful to a wide range of readers, including clinicians and educators; health services researchers; managers in practices, insurance plans, and the health care industry; policymakers; and researchers focusing on health policy development. The authors have included background information on the etiology, clinical characteristics, and treatment of depression, as well as chapters on social and clinical factors, health outcomes, differences in treatment by specialty and payment source, the cost-effectiveness of care, the effects of health care policy on the care of depression, methods for measuring quality of care and outcomes, and methods for evaluating health care systems.

Caring for Depression, a 260-page, hardcover book, is available from Harvard University Press (key code: 39-605/book code WELCAR), Cambridge, MA. Call 800-448-2242 for price and ordering information.

Return to Contents

Elderly/Long-Term Care

State programs for preadmission screening of nursing home applicants may not be cost-effective

State nursing home preadmission screening (PAS) programs are designed to screen out nursing home applicants who do not really need nursing home care, and by doing so, reduce unnecessary Medicaid costs by diverting these patients to community care settings. But, most State programs refuse very few nursing home applicants. In fact, a recent study supported in part by the Agency for Health Care Policy and Research (HS07574) found that four States denied admission to only 1 percent of applicants per year, and nine States had denial rates ranging from 2 to 6 percent. Other States were unable to report denial rates.

These low rates of denial suggest that the PAS programs may have little direct effect in diverting clients away from nursing homes and thus may not be cost-effective, conclude University of California, San Francisco researchers, Charlene Harrington, Ph.D., and Michael Curtis, Ph.D. They collected primary data on PAS programs from State officials in two separate surveys in 1989 (covering the 1978 to 1989 period) and 1992 (covering the 1990 to 1992 period).

The researchers found that States are becoming increasingly stringent in their screening of applicants, but that PAS programs vary widely in characteristics: geographic base (Statewide/local), population screened (Medicaid/private pay), client origin (hospital/community), whether decisions are binding, timing of screen (prior to/after admission), and who does the screening (State staff/contract agency/provider). The increase in spending on home and community-based services suggests that States may be substituting home and community long-term care services for nursing home care, even though the role that PAS programs may have played in this scenario is not clear.

In conclusion, the researchers note that although PAS program costs may increase with greater use of community care services, the benefits to the frail elderly population may also increase. Thus, any analysis of the costs and benefits of PAS programs should take into account clinical outcomes of care, quality of life, and quality of care for clients.

See "State variations and trends in preadmission screening programs," by Drs. Harrington and Curtis, in the December 1996 issue of the Journal of Applied Gerontology 15(4), pp. 414-432.

Return to Contents

Special Populations

Minority research center publishes reports on prenatal care and reproductive health

Studies continue to show ethnically-based differences in access to care, health status, and treatment outcomes for minority populations. To learn more about the effectiveness and outcomes of medical treatments in diverse populations, the Agency for Health Care Policy and Research supports a number of research centers on minority health, including the Medical Effectiveness Research Center on Diverse Populations at the University of California, San Francisco (HS07373). Led by A. Eugene Washington, M.D., M.Sc., the Center focuses on the clinical outcomes of blacks, Hispanics/Latinos, and Asian Americans. Areas of research include cancer, cardiovascular disease, diabetes, glaucoma, prenatal care, and reproductive health. Center researchers recently published articles on prenatal care and reproductive health which are summarized here.

Fuentes-Afflick, E., and Hessol, N.A. (1997). "Impact of Asian ethnicity and national origin on infant birth weight." American Journal of Epidemiology 145(2), pp. 148-155.

The birthweight of infants born to Asian women depends on both the mothers national origin and their use of prenatal care, according to the researchers. They analyzed the relationship between Asian ethnicity/national origin and low birthweight (LBW) among 50,044 Asian and 221,866 white women who delivered babies during 1992 in California (where 40 percent of all Asian births in the United States occur).

Results showed that women who made inadequate use of prenatal services were four to six times more likely to have LBW infants (less than 2,500 g or 4.5 pounds) than women with adequate use of these services.

Overall, rates of LBW among Chinese, Korean, and Vietnamese women ranged from 3.5 percent to 4.7 percent, lower than the U.S. national goal of 5 percent. On the other hand, the LBW rates among Cambodian, Filipino, Indian, Japanese, Laotian, and Thai women ranged from 5.3 percent to 8.0 percent. The relative risk of very LBW (500 to 1499 g, from 1 pound to less than 3 pounds) was significantly elevated among Filipino women relative to whites, although the absolute difference in incidence was only 0.3 percent. Asian Indian women were almost twice as likely to have moderately low birthweight infants (1,500 to 2,499 g, 3 to 4.5 pounds) than white women. The researchers conclude that these differences in LBW across different Asian groups highlight the need for health policies tailored to both population and outcome.

Grover, C.M., Kuppermann, M., Kahn, J.G., and Washington, A.E. (1996, December). "Concurrent hysterectomy at bilateral salpingo-oophorectomy: Benefits, risks, and costs." Obstetrics & Gynecology 88(6), pp. 907-913).

When a woman is undergoing removal of both ovaries (bilateral salpingo-oophorectomy, BSO) because of a suspicious mass, she is typically given a hysterectomy at the same time, even when the frozen tissue sections suggest that the mass is benign. Clinicians have recently begun to question whether the benefits of this approach outweigh the risks. Apparently they do, according to this study. The researchers found that concurrent hysterectomy causes short-term problems but appears to slightly increase average life expectancy among perimenopausal women and is cost-saving. But since medical outcomes and economic consequences only marginally favor the procedure, patient preference should play a key role in determining appropriateness of its use, recommend the researchers. They used decision analysis to compare the health outcomes and economic costs of performing BSO with concurrent hysterectomy versus BSO alone in theoretical groups of 10,000 45-year-old women undergoing surgery for benign masses.

The researchers calculated that performing concurrent hysterectomy would prevent about 71 future deaths by averting cancer and future surgery-related deaths at a cost of five immediate deaths from the hysterectomy. On average, hysterectomy at age 45 would add about 0.071 years of life expectancy and at age 55, 0.026 years. BSO with concurrent hysterectomy would result in cost savings of about $1,913 per patient at age 45 and $1,112 at age 55. Alternatively, the choice of uterine preservation (BSO alone) ultimately would result in 49 cervical cancer cases with 22 subsequent deaths and 228 uterine cancer cases with 44 deaths. An additional 795 women ultimately would undergo hysterectomy at a later date for benign gynecologic disease, with five suffering postoperative death because of increased age-related surgical mortality.

Return to Contents

New brochure helps consumers choose and use health plans

A new brochure, Choosing and Using a Health Plan, is now available from the Agency for Health Care Policy and Research to help consumers choose a health insurance plan that best meets their needs and to help them get the most from their plan. The brochure was developed through a partnership between AHCPR and the Health Insurance Association of America (HIAA), a national trade association whose more than 250 members are insurers and managed care companies.

According to AHCPR's Deputy Administrator, Lisa B. Simpson, M.B., B.Ch., people usually know that health insurance is important and that they need it, but all too often they may be confused by the wide variety of coverage that is offered and the options available from different plans. As HIAA's President Bill Gradison points out, this new guide provides clear, concise information that will enable consumers to choose the coverage that's right for them.

Choosing and Using a Health Plan was developed for consumers who have a choice of health plans and those who are already enrolled in a plan. Even when consumers are not able to choose their own health plan, it still is important for them to understand what kind of protection their plan provides. The guide stresses the importance of knowing the "basics" when choosing and using a health plan. When comparing plans, this includes considering the services offered, choice of providers, location, costs, and quality. For those already in a plan, the guide offers tips to help consumers learn how to obtain care, how to understand the plan's hospitalization rules, and steps to take if they are not satisfied with their care.

Print copies of Choosing and Using a Health Plan (AHCPR Publication No. 97-0011) are available from the AHCPR Publications Clearinghouse.

AHCPR publishes guide on the use of geographic methods in health services research

Medical geography uses the concepts and techniques of geography to investigate health-related issues. With the computer programs and applications now available, it is possible to create maps and spatial interpretations that are not complex and do not require special data or systems. Data that apply to standard geographic areas can be mapped quickly and included in almost any type of printed report or visual medium.

The Agency for Health Care Policy and Research has published a new report, Using Geographic Methods to Understand Health Issues, for use by health services researchers, policymakers, educators, health planners, and others interested in medical geography. It demonstrates the ways in which the methods and approaches offered by the discipline of geography can be applied to the field of health services research. The report was compiled by Thomas C. Ricketts, Ph.D., and his colleagues at the University of North Carolina, Chapel Hill, with support from the Agency (AHCPR grant HS06706).

This full-color, 24-page manual leads the reader through several commonly used methods for producing maps and visual displays that can be applied for policy analysis, research, and/or planning purposes. It uses real-world examples to demonstrate and interpret applications that reflect both the disease ecology and spatial analysis aspects of medical geography.

Print copies of Using Geographic Methods to Understand Health Issues (AHCPR Publication No. 97-N013), by Dr. Ricketts, Lucy A. Savitz, Ph.D., Wilbert M. Gesler, Ph.D., and Diana N. Osborne, B.A., are available from the AHCPR Publications Clearinghouse. The supply of free print copies is limited.

When AHCPR copies are exhausted, the publication may be obtained from the National Technical Information Service (NTIS accession no. PB97-137707, 24 pp., $19.50 paper, $10.00 microfiche).

Final reports now available from NTIS

The following grant final reports are now available from the National Technical Information Service. Each listing identifies the project, principal investigator and his or her affiliation, grant number, project period, project objective, and methods used. Findings and other information are detailed in the individual reports.

Hawaii/Asian Pacific Island MEDTEP Research Center. Robert M. Worth, M.D., Ph.D., Pacific Health Research Institute, Honolulu, HI. AHCPR grant HS07381, project period 9/9/92 to 8/31/96.

The Hawaii/Asian Pacific Island Medical Treatment Effectiveness Program (MEDTEP) Research Center was funded as a developmental center under a 3-year cooperative agreement with AHCPR. The Center conducted outcomes research, training, technical assistance, and dissemination of findings with the goal of improving the effectiveness and appropriateness of medical practice as measured by "premature" death, health status, functional capacity, patient satisfaction, and quality of life and focused on Native Hawaiians, Asians, and Pacific Islanders (75 percent of the States population). Two key projects were the evaluation of health outcomes in a patient and community participation model of chronic disease management and health outcomes. The Center conducted 31 projects, raising $1.5 million in external funds. Over 300 individuals were trained, and 54 investigators and 14 research assistants conducted research. Abstract, executive summary, and final report are available from the National Technical Information Service (NTIS accession no. PB97-136675, 31 pp, $21.50 paper, $10.00 microfiche).

Implementation and Evaluation of Practice Guidelines. Siew H. Lam, M.B.B.S., University of Utah, Salt Lake City. AHCPR grant HS07450, project period 9/1/92 to 8/31/94.

Practice guidelines for the management of deep vein thrombosis were implemented in a Problem-Oriented Patient Management System on the HELP Hospital Information System at the LDS Hospital in Salt Lake City, UT. The system was able to generate patient-specific guideline suggestions according to the clinical situation at the time of generation. A retrospective evaluation was used to compare the appropriateness of the generated guideline suggestions with the appropriateness of the attending physician's management decisions. A significantly higher proportion of the guideline suggestions were found to be appropriate compared with the case physicians management decisions.Abstract, executive summary, and final report are available from the National Technical Information Service (NTIS accession no. PB97-134365, 14 pp, $19.50 paper, $10.00 microfiche).

Lipid Lowering Medications and Risk of Injury. Viktor E. Bovbjerg, Ph.D., University of Washington, Seattle. AHCPR grant HS08469, project period 9/1/94 to 8/31/95.

Early primary prevention trials of lipid lowering therapies reduced cardiovascular disease but not total mortality. Meta-analyses showed increased injury deaths (chiefly motor vehicle collisions, suicide, and homicide) among persons treated to lower cholesterol. This study was a population-based case-control study to examine the association of lipid lowering medication use with death or hospitalization resulting from motor vehicle collisions or intentional injuries. Cases (n=298) were identified using diagnostic databases, and controls (n=332) were frequency matched to cases on age, sex, and year of presentation. The use of lipid lowering medication was assessed from pharmacy data. Other variables were assessed from databases and medical records. Current use of lipid lowering medication was not associated with injury in unadjusted analyses or in analyses adjusted for matching variables, behavioral disorders, clinical heart disease, and health status. Past use was not associated with injury in either unadjusted or adjusted analyses. Further analyses did not reveal any patient subgroups at increased risk of injury. No evidence was found of an association between use of lipid lowering medications and injury. This finding is consistent with more recent randomized clinical trials, recent meta-analyses, and current treatment guidelines. Abstract and executive summary of dissertation are available from the National Technical Information Service (NTIS accession no. PB97-134381, 16 p, $19.50 paper, $10.00 microfiche).

Medical Care and Risks of Dysfunctional Chronic Pain. Michael Von Korff, Sc.D., Group Health Care Cooperative of Puget Sound, Seattle, WA. AHCPR grant HS07759, project period 9/1/93 to 8/31/96.

The objectives of this study were two-fold: (1) assess the long-term outcomes of primary care back pain and headache patients, and (2) conduct an exploratory audiotape study of primary care pain visits to identify how behavioral approaches to pain management might be integrated into care. Important results of the long-term followup study included: (1) new methods of assessing chronic pain status in primary care were developed; and (2) primary care back pain and headache patients were found to have less favorable long-term outcomes than previously believed. The most significant results from the audiotape study were that: (1) back pain patients had many goals for their visit, but clinicians rarely elicited or clarified these goals; (2) clinicians were less likely to assess interference with activities than to ask diagnostic questions about back pain; (3) clinicians recommended an average of seven different interventions per visit to their patients; (4) physician recommendations were generally in accord with AHCPR's back pain guideline; (5) patients usually were not asked whether they agreed with the recommendations, and patients usually did not voice active acceptance; and (6) the importance of avoiding prolonged rest and resuming normal activities was not consistently discussed in the visit, even with patients who had significant activity limitations. Abstract, executive summary, final report, and appendixes A-C are available from the National Technical Information Service (NTIS accession no. PB97-136709, 89 pp, $25.00 paper, $10.00 microfiche).

Physician Interaction Style and Medication Management. Betsy Sleath, Ph.D., University of North Carolina, Chapel Hill. AHCPR grant HS08431, project period 12/1/95 to 9/30/96.

The objective of this project was to examine how physicians and patients discussed medications using a data set of 503 audiotapes and transcripts of outpatient visits and postvisit interviews with chronic disease patients and their primary care physicians. Ninety-eight percent of patients discussed medications with their physicians. Patients and physicians spent an average of nearly 4 minutes or 18 percent of the medical encounter discussing medications. Twenty-seven percent of patients expressed one or more complaints about their medications; the types of complaints most often expressed included dissatisfaction, drowsiness, upset stomach, and difficulty in purchasing. Nineteen percent of patients expressed one or more adherence problems, including running out of medication, not using the medication, and not following physician instructions. Fifty percent of the patients had their medication regimen modified in some way during their medical encounter. Patient consumerism, physician use of a participatory style, patient expression of one or more complaints, and patient expression of one or more adherence problems all significantly influenced whether the patients medication regimen was changed. Abstract, executive summary, and final report are available from the National Technical Information Service (NTIS accession no. PB97-136667, 97 pp, $25.00 paper, $14.00 microfiche).

Return to Contents
Proceed to Next Section