Sex Differences in Cardiac Catheterization: Title and subTitle BreakThe Role of Physician Gender FREE

Many studies indicate that women are less likely than men to undergo cardiac catheterization after an acute myocardial infarction (AMI).^{1 - 12} Sex-based variations in cardiac procedure use after myocardial infarction (MI) have been documented in several settings and, in most studies, have persisted after adjustment for patient clinical characteristics,^{2 - 8 ,10 - 12} insurance and access to care,^{1 ,3 ,6 - 7 ,10 - 12} and hospital, physician, and health system effects.^{1 ,6 ,10 ,12} Studies that found no sex differences in cardiac procedure use^{13 - 18} have primarily incorporated patients drawn from selected centers^{13 - 18} and nonacute settings.^{13 - 14 ,18} Sex-based differences in cardiac catheterization use and similar variations in the management of other clinical conditions have raised the concern that sexual discrimination on the part of physicians may be responsible.¹⁹

A recent study²⁰ found differences in physicians' recommendation for cardiac catheterization by patient sex, suggesting that physician behavior may contribute to sex differences in cardiac procedure use. Although sex differences in recommendation for cardiac catheterization were modest, elucidated in a simulated clinical setting, and limited solely to black women, this study²¹ was widely interpreted in the popular press as evidence of physician sex bias in medical care. Limited evidence suggests that physician sex may influence patient satisfaction²² and treatment.^{23 - 25} However, no study, to our knowledge, has evaluated the relationship between patient sex, physician sex, and use of cardiac procedures. Thus, it is unclear whether sex disparities in cardiac catheterization use are attributable to sexual bias in clinical practice.

We examined the association of patient sex and physician sex with cardiac catheterization use after an MI by using data from the Cooperative Cardiovascular Project (CCP), a national initiative to assess the care and outcomes of Medicare beneficiaries. We hypothesized that if sex differences in procedure use were attributable to sexual discrimination on the part of physicians, such disparities would be greater when a patient and physician were of different sexes. In particular, we expected greater sex disparities in procedure use for patients treated by male physicians compared with patients treated by female physicians.

The CCP was a project sponsored by the Health Care Financing Administration (now the Centers for Medicare and Medicaid Services) and developed to improve the quality of care for Medicare beneficiaries hospitalized with AMI.²⁶ Patients in the CCP cohort (n = 234 769) included fee-for-service Medicare beneficiaries who had been discharged from a nongovernmental acute care hospital in the United States and had a primary discharge diagnosis of AMI²⁷ between January 1994 and February 1995, with the exception of AMI readmissions. Medical records for each hospitalization were forwarded to clinical data abstraction centers, and data were abstracted for each hospitalization. Data quality was ensured by trained technicians and software abstraction modules and was monitored by random record reabstraction.

We limited our analysis to patients 65 years and older who had been hospitalized with a confirmed AMI. Patients younger than 65 years (n = 17 593), those without clinically confirmed AMI (n = 31 186), or those who arrived by interhospital transfer (n = 34 409) were excluded from analysis. Patients with multiple admissions in the CCP sample were identified, and readmissions for AMI were excluded (n = 23 773). Patients whose records could not be linked to data obtained from the American Hospital Association (n = 2363) or from the 1990 US census (n = 10 810) or patients for whom Medicare Part A billing data were unavailable (n = 34 187) were also excluded. We excluded records for which physician data were unavailable or the physician's sex was unknown (n = 22 222). To restrict the influence of previous invasive cardiac care on diagnostic decisions made for patients in the CCP cohort, we excluded patients who had undergone coronary artery bypass graft surgery or percutaneous transluminal coronary angioplasty (n = 43 143). Similarly, we excluded patients whose attending physician was a cardiothoracic surgeon, since virtually all of these patients underwent cardiac catheterization as a precursor to coronary artery bypass graft surgery (n = 9577). Patients with metastatic cancer or a life expectancy of less than 6 months (n = 4617) were also excluded because their treatment decisions may not have been intended for mortality reduction. Patients hospitalized in Puerto Rico (n = 1519) were excluded as well to reduce sample heterogeneity. The number of patients meeting at least 1 of the exclusion criteria was 129 836, leaving 104 933 patients eligible for analysis. Of these, we excluded 702 patients treated at hospitals without at least 1 male patient and 1 female patient in the CCP sample; the remaining 104 231 patients constituted the study cohort.

The principal outcome was use of cardiac catheterization within 60 days of admission for AMI, as determined by evaluating Medicare Part A billing records for International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes associated with cardiac catheterization.

We compared rates of cardiac catheterization between the 4 groups of patient sex–physician sex combinations: male patients treated by male physicians, male patients treated by female physicians, female patients treated by male physicians, and female patients treated by female physicians. Patients' sex was abstracted from the medical record. Physician sex data were obtained for the attending physician, defined as "the clinician who [was] primarily and largely responsible for the care of the patient from the beginning of the hospital episode," as noted on the Medicare Part A claim.²⁸ The attending physician's identification number was linked with data from the American Medical Association's Physician Masterfile,²⁹ a validated source of physician demographic data,³⁰ to determine demographic characteristics.

Crude rates of cardiac catheterization were compared between each of the 4 patient sex–physician sex groups by means of a global χ² analysis. Comparisons were repeated pairwise to elucidate differences between particular patient sex–physician sex groups. Patient medical history, clinical presentation, and physician, hospital, and regional characteristics were also compared globally and pairwise across the 4 patient sex–physician sex groups by means of χ² tests and analysis of variance.

To evaluate the independent association of the patient sex–physician sex combination with use of cardiac catheterization, we performed multivariable logistic regression modeling. Models used sequentially cumulative adjustments for patient sociodemographic characteristics, medical history and clinical presentation, physician characteristics, hospital data, and other AMI therapies provided. Patient sociodemographic characteristics included race and residential ZIP code measures of median household income and percentage of population with at least a high school education, as reported in the 1990 US census.³¹ Patient medical history measures were drawn from previously identified predictors of procedure use and included age, left ventricular ejection fraction, prior AMI, congestive heart failure, smoking status, hypertension, diabetes, cerebrovascular disease, peripheral vascular disease, dementia, functional status, liver disease, immunocompromised status, and the following characteristics ascertained at admission: systolic blood pressure, heart rate, Killip class, AMI location, renal dysfunction, Q-wave infarction, ST-segment elevation, microalbuminuria, and anemia. Physician characteristics included age, race, years in practice, practice type, and specialty, as reported in the American Medical Association's Physician Masterfile. Hospital data were obtained from the 1994 American Hospital Association Survey of Hospitals³² and included level of cardiac care facilities (none, catheterization laboratory, coronary artery bypass graft surgery suite), ownership (public, not for profit, for profit), teaching status (nonteaching, affiliated with a graduate medical education program, member of the Association of American Medical Colleges' Council of Teaching Hospitals), AMI volume, rural location (location outside a metropolitan statistical area), and US census region. Finally, AMI therapies evaluated included use of reperfusion therapy (primary percutaneous coronary intervention or thrombolytic agents), aspirin, β-blockers, and angiotensin-converting enzyme inhibitors. Multivariable logistic regression models were used to estimate risk-adjusted predicted rates of cardiac catheterization for each of the 4 patient sex–physician sex groups. Risk-standardized rates of cardiac catheterization were derived by dividing crude rates of cardiac catheterization by risk-adjusted predicted rates of cardiac catheterization and multiplying this ratio by the full sample's crude cardiac catheterization rate (44.1%).

Analyses were repeated incorporating a patient sex–physician sex interaction term to empirically test the existence of a male physician sex bias. If a physician sex bias by male physicians were to influence cardiac catheterization use, an interaction effect for patient sex and physician sex should be observed in multivariable analysis. Analyses were also repeated evaluating the composite end point of cardiac catheterization refusal during hospitalization and cardiac catheterization within 60 days of admission to approximate the frequency of referral for cardiac catheterization and, among the cohort of patients not treated by cardiologists, to determine whether variations in referral were limited to cardiologists. Results were adjusted for clustering of patients by using Huber-White robust estimates of SE.³³ Odds ratios were converted to risk ratios according to the method outlined by Zhang and Yu.³⁴ All models demonstrated appropriate discrimination and calibration. Statistical analyses were conducted with SAS statistical software, Version 6.12 (SAS Institute Inc, Cary, NC) and STATA statistical software, Version 6.0 (Stata Corp, College Station, Tex).

Of the 104 231 patients in the study cohort, 52.0% were women. The largest group consisted of female patients treated by male physicians (50 408, or 48.4%), followed by male patients treated by male physicians (46 823, or 44.9%), female patients treated by female physicians (4240, or 4.1%), and male patients treated by female physicians (2760, or 2.6%). Female physicians treated 6.7% of all patients, including a greater proportion of female patients.

Clinical characteristics were generally comparable among the 4 patient sex–physician sex groups (Table 1a), although female patients treated by female physicians were more likely to be nonwhite; to have a higher Killip class at admission; to have undergone an AMI without ST-segment elevation; to have poorer functional status, including a higher frequency of dementia, limited mobility, and incontinence; and to be admitted from a nursing home. Male patients cared for by male physicians were most likely to be treated by a cardiologist, by a white physician, or at a hospital with invasive cardiac care facilities and higher volume and were most likely to receive reperfusion therapy during their infarction.

As shown in Table 2, female patients had lower crude rates of cardiac catheterization compared with male patients and were less likely to undergo cardiac catheterization whether treated by male physicians or female physicians. In addition, male physicians had higher crude rates of cardiac catheterization use than female physicians, whether treating male or female patients. Consequently, male patients treated by male physicians had the highest crude rates of cardiac catheterization use among the 4 patient sex–physician sex groups (50.8%), whereas female patients treated by female physicians had the lowest rates of procedure use (34.8%). Predicted rates of cardiac catheterization were similarly lower for female patients and patients treated by female physicians. Cardiac catheterization refusal rates were similar for patients treated by either male or female physicians and slightly higher among female patients (Table 2).

Female patients were less likely to undergo cardiac catheterization than male patients, regardless of their physician's sex, after adjustment for patient medical history and presentation, physician, hospital and geographic characteristics, and AMI therapies provided during hospitalization (Table 3). By conducting multivariable analysis, we also found that patients treated by male physicians were more likely to undergo cardiac catheterization than were patients treated by female physicians, regardless of the patient's sex. As a result, even after adjustments for baseline differences, male patients treated by male physicians were most likely to undergo cardiac catheterization and female patients treated by female physicians were least likely, compared with the referent group of female patients treated by male physicians (Table 3).

The absolute difference in risk-standardized rates of cardiac catheterization between male and female patients was 2.9% for those treated by male physicians (45.6% vs 42.7%) and 3.5% for patients treated by female physicians (44.4% vs 40.9%). This difference in risk-standardized rates (2.9% vs 3.5%) was not statistically significant, because we found no significant interaction between patient sex and physician sex when we tested this variable in the final multivariable model (P for interaction = .85). Findings were similar for the composite end point of catheterization refusal or undergoing cardiac catheterization and among the cohort of patients who were not treated by cardiologists.

We found no evidence that sex differences in the use of cardiac catheterization after an AMI were greater when patients and physicians were of different sexes. The similarity in sex-associated disparities in cardiac catheterization between male physicians and female physicians suggests that sexual discrimination by male physicians toward women (or vice versa) is likely not responsible for sex differences in the use of cardiac procedures. These data, drawn from a large, community-based cohort, contribute to the well-established body of evidence of patient sex differences in cardiac procedure use^{1 - 12} and augment a growing volume of literature that evaluates the role of physician sex and patient care.^{23 ,25 ,35 - 39}

The absence of greater disparities in procedure use when a patient and physician are of different sexes is reassuring and consistent with data indicating that physicians' interactions with patients of a different sex are similar in content and decision-making style to those with patients of the same sex.⁴⁰ Among studies that found differences in procedure use in sex-discordant pairings, these differences have been seen largely in sex-specific services such as mammography screening,^{25 ,35 ,39} prostate examination,³⁹ and pelvic examination or Papanicolaou smear testing,^{25 ,35 ,39} selected procedures that may have produced differences in use rates because of their sex-specific nature.³⁸ Although our data cannot preclude the possibility that physician attitudes and beliefs may still account for some of the sex-associated variation in cardiac procedure use,⁴¹ our findings would suggest that any such attitudes are present in both male and female physicians.

The modest differences in cardiac procedure use associated with physician sex are difficult to explain. These variations remained after adjustment for patient, clinical, physician, and hospital characteristics and were not an artifact of subspecialty training, because they were also observed among noncardiologists. Although data from the ambulatory care setting have suggested that differences exist between patients who seek care from male physicians and those who seek care from female physicians,³⁵ it is unlikely that patients hospitalized with an acute event such as an MI would select a physician based on sex. Studies^{35 ,37} have identified significant physician sex differences in medical care attitudes, beliefs, and subsequent recommendations for preventive care services. Similar differences in referral for cardiac catheterization may exist between male and female physicians. Although modest in magnitude, small variations may result in significant differences in procedure use when applied to the large population of patients who experience AMIs.

Several factors may combine to explain sex differences in referral for cardiac catheterization. First, women with cardiac disease are less likely to report typical angina symptoms or report them consistently, or to have their ischemia perceived as cardiac in origin.^{42 - 46} Differences in coronary disease presentation have been shown to explain sex differences in referral for cardiac catheterization in the outpatient setting.¹³ However, it is unclear what effect such variation would have among patients hospitalized with a documented MI and in whom coronary disease is clearly presumed. Second, sex differences in the use of cardiac procedures may be attributable to lower income or wealth and the concomitant lack of "Medigap," or supplemental Medicare insurance among female Medicare beneficiaries.⁴⁷ Patients with fewer economic resources are less likely to undergo cardiac procedures,⁴⁸ and without supplemental coverage, women may be unable to incur the marked out-of-pocket costs associated with cardiac catheterization. Alternatively, treatment differences may be attributable to sex-associated variations in patients' willingness to undergo cardiac procedures. Although treatment differences between men and women in our analysis were unchanged when patients who refused cardiac catheterization were accounted for, more subtle sex-associated variations in preferences may still be plausible. Last, it is possible that unmeasured differences in comorbid conditions or cardiac risk factors explain the lower use of cardiac catheterization for women. However, the effect of residual confounding on cardiac catheterization rates should apply to all women, regardless of the sex of their attending physician.

Several issues should be considered when this study is interpreted. Without a definitive standard of appropriateness for cardiac catheterization, it is unclear whether variations in treatment reflect undertreatment, appropriate use, or overtreatment. However, our primary objective was to ascertain whether physician gender influenced sex-associated variations in cardiac catheterization use and to use this information as a means of empirically testing concerns of sexual bias by male physicians in cardiac procedure use, not necessarily the appropriateness of such variations. Also, our data were drawn from a cohort of elderly Medicare beneficiaries who had experienced an AMI and may not be applicable to younger patients, patients who undergo catheterization before an AMI, or patients treated at independent cardiac catheterization laboratories. However, most patients who undergo catheterization each year are older than 65 years, insured by Medicare, and treated in hospital-based laboratories; thus, our data are representative of most cardiac care nationwide. Although we evaluated the sex, race, training, and practice location of the attending physicians in the cohort, we were unable to account for the influence of physician consultations, the effect of other physicians involved in each patient's care, or physician subspecialty training in interventional or invasive cardiology, all of which may be associated with procedure use.⁴⁹ The absence of data related to the indication for cardiac catheterization precluded evaluation of the severity of coronary artery disease, although it is unlikely that differences in patient coronary artery disease prevalence would explain differences in procedure use associated with physician sex. Finally, our exclusion criteria resulted in a population that included a greater proportion of female patients than the original CCP baseline, although this result was expected, given our exclusion of patients younger than 65 years and those who had undergone revascularization. Nevertheless, we believe that these exclusions were appropriate, given our study's purpose.

Our results suggest that sexual discrimination, principally by male physicians toward women, does not explain sex-associated disparities in cardiac catheterization use after an AMI. However, other attitudes common to both male and female physicians may contribute to lower rates of cardiac procedure use in women.