Although cardiovascular disease (CVD) mortality rates are declining for women, as they are for men, women’s CVD mortality rates continue to exceed those of men.1 Coronary heart disease (CHD), a subset of CVD, is responsible for more than 240 000 deaths among women in the United States each year.2 For many years, there has been a consensus that underrecognition of women’s symptoms and difficulty in diagnosing CHD in women contribute to their greater disability and mortality after a CHD event. In response to the challenges in diagnosing CHD in women, the Agency for Healthcare Research and Quality commissioned a review of the accuracy of noninvasive technologies for diagnosis of CHD in women.3 Although this is a seminal review, only studies of women with chest pain syndrome were included. However, numerous studies have reported that chest pain is not present in a significant number of women with CHD, and many women report little or no chest pain before or with myocardial infarction (MI).4–7
Thus, it is vital to identify symptoms other than chest pain that are associated with risk of progressing to a CHD event, such as MI, so that women experiencing non–chest pain symptoms may undergo appropriate diagnostic testing. Several risk score calculators, such as the Framingham and Reynolds scores, are useful in predicting long-term risk for CVD.8,9 However, these instruments do not assess the prodromal symptoms (PSs), other than angina, that women frequently experience before MI.10,11 The McSweeney Acute and Prodromal Myocardial Infarction Symptom Survey (MAPMISS) assesses the presence of a broad range of PSs among women. However, although the MAPMISS has been used in several studies,10,12,13 all have been retrospective. As a result, we know little about the predictive utility of symptoms in the PS section of the MAPMISS.
Therefore, the aim of this longitudinal observational study was to assess the utility of the MAPMISS PS scores and PS counts in predicting the occurrence of cardiac events in women during a 2-year period. Our goal was to identify the most parsimonious subset of PS data predictive of angioplasty, stent placement, coronary artery bypass, MI, or death attributed to CHD.
Setting and Sample
We recruited women who were either clinician referred (n = 903, 82.32%) or self-referred (n = 194, 17.68%) to cardiology practices for initial cardiology evaluation. We recruited women from 3 sites in Arkansas and 1 site in Kentucky. The sites were associated with either academic centers or large statewide private practices with multiple satellite clinics in urban, suburban, and rural areas. To be eligible for the study, women had to (a) have been referred or self-referred to a cardiologist for initial CHD evaluation and have no current or previous CHD diagnosis, (b) identify themselves as African American/black or Caucasian/white, (c) be at least 21 years old, (d) be cognitively intact, (e) have access to a telephone, and (f) be able to read or spell out medication information from their prescriptions or designate a family member to perform this task.
The MAPMISS has 3 sections: (1) the Acute Symptom section, which is administered after a cardiac event (eg, MI, angioplasty), (2) the 30-question PS section, administered in the absence of a known cardiac event, and (3) a Background section that addresses demographic characteristics, comorbid conditions, other risk factors, and current medications.10,14 The analyses reported here are based on data from the PS and Background sections. In the PS section, women are asked about the occurrence of each of 30 PSs in the preceding 3 months. For each of the symptoms reported, they are then asked to specify the intensity and frequency with which it occurred. Categorical response options for intensity are mild, medium, or severe, scored 1 to 3, respectively. Categorical response options for frequency are daily, several times per week, 1 or more times per week, 2 or more times per month, monthly, and less than monthly, scored 7, 3.5, 1, 0.5, 0.25, and 0.167 days per week, respectively. A PS not reported for the period is scored 0 for intensity and 0 for frequency. Individual PS scores are calculated as the product of intensity and frequency (range, 0–21, where 0 indicates the absence of the symptom).14 A cumulative PS score is calculated as the sum of individual symptom scores (range, 0–630). The MAPMISS has been shown to have content validity and solid test/retest reliability with both African American/black and white women (r = 0.91).14,15
The Short Blessed Test (SBT) assesses for cognitive impairment and was administered at baseline and again at 12 months to ensure that participants were sufficiently cognitively intact to provide meaningful information. The SBT includes 6 weighted items that evaluate orientation and memory and has established reliability and validity.16–18 Each item is scored 0 to 5, and lower scores indicate better cognitive function.17 Women who scored 16 or higher at baseline were not eligible to participate in the study.
Institutional review board approval was obtained at each site before recruitment. At all sites, new female patients received a flyer describing the study as part of the packet of forms completed by new patients at the site. Interested women returned the flyer to office staff along with their names and telephone numbers and completed a Health Insurance Portability and Accountability Act authorization permitting the clinic to disclose their names and medical history (reason for clinic visit) to the research team. A research assistant (RA) collected the forms, telephoned interested women, explained the study, answered questions, and for interested women, obtained verbal consent and authorization to review their medical records. The RA then administered the SBT; if the woman scored lower than 16, the RA scheduled a telephone interview to administer the MAPMISS. Participants completed follow-up telephone interviews at 3-month intervals that asked about symptoms experienced during the previous 3-month period, for 2 years or until a CHD event occurred.
If a woman reported a CHD event (angioplasty, stent placement, coronary artery bypass surgery, or MI) during the previous 3 months, the woman completed the acute section of the MAPMISS questions, and an RA audited the woman’s medical records from the cardiology clinic, emergency department, and/or hospital where she received treatment to verify occurrence of the event. Once an event occurred, a woman’s participation in the study was complete and no further follow-up assessments were conducted. If a woman died, family members were asked about her hospitalization and cause of death. We also obtained a death certificate to confirm cause of death. The first author and other team members evaluated whether a CHD event had occurred and whether the death was cardiac related. A cardiologist adjudicated all disagreements and made the final determination.
All interviews were administered by RAs located at the site using a computer-assisted telephone interview system that allowed them to enter data directly into a computer program while conducting the interviews. The baseline assessment covered PSs in the previous 3 months, demographic characteristics, cardiovascular risk factors including comorbidities and CHD family history, and currently prescribed medications. It took approximately 60 minutes to complete. Follow-up assessments included only the MAPMISS PS section with respect to the previous 3-month period. Follow-up interviews took approximately 20 to 30 minutes. Women received a $40 gift certificate for completing the baseline interview, $15 for each subsequent 3-month follow-up interview, and $30 for the final interview.
Data from the MAPMISS were analyzed to generate PS symptom counts (present/absent; 0–30), individual PS symptom scores (0–21), and cumulative PS symptom scores (0–630). As described earlier, individual scores for each of the 30 PSs were calculated for each woman at each visit based on the frequency (0–7 days/week) and intensity (0–3) she reported for each symptom in the preceding 3 months. A cumulative PS score was also computed for each woman at each visit as the sum of her 30 individual PS scores.
We had 2 main analytic goals: (1) to assess the utility of PS scores in predicting study outcomes (angioplasty, stent placement, coronary artery bypass, MI, or death attributed to CHD) and (2) to identify the most parsimonious subset of PS data predictive of cardiac events. We used the cumulative PS score to address the first goal because it is the most information-rich MAPMISS measure, incorporating symptom count, intensity, and frequency. To this end, we estimated hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) for experiencing an adverse cardiac event using multivariable Cox proportional hazards regression modeling, including cumulative PS scores as a time-dependent covariate. The model was first adjusted for age and race only. To determine whether cumulative PS scores contributed independently to prediction after accounting for the variables included in existing risk measures, we repeated the analysis adjusting the model for body mass index, education, income, marital status, physical activity, family history of CHD, smoking status, diabetes, hypertension, and hypercholesterolemia, in addition to age and race.
Because assessing intensity and frequency is clinically time-consuming and perception of symptom intensity may vary by individual, we next looked at the predictive utility of the PS count, following the same 2-step procedure used for cumulative PS scores. Lastly, we explored whether predictive utility might be retained if data (scores or counts) were collected on a specific subset of symptoms. In identifying the most parsimonious predictive models, we first fit separate score and count models for each PS individually. In each case (scores and counts), we then evaluated all PSs together and the simplest models with the smallest number of PS and the greatest explanatory power were selected using a strategy that combined stepwise regression, Akaike information criteria, and the best subset selection method, as suggested by Shtatland et al.19,20 This strategy was designed to retain the best qualities of each of these 3 model building methods.
The predictive ability of the final models was assessed by computing Harrell concordance or C statistic and the corresponding bootstrapped 95% CI.21,22 The Harrell C statistic is analogous to the area under the receiver operating characteristic curve for survival data and is an estimate of the probability that for 2 randomly selected individuals, the risk of failure predicted from the model will be higher for the individual that fails first. Thus, the C statistic is interpreted the same way that the area under the receiver operating characteristic is interpreted. Specifically, if C = 1, then the model has perfect prediction, whereas if C = 0.5, the model’s predictive ability is no better than a coin flip. Data analysis was performed using Stata version 1223 and SAS 9.3.24
A total of 1114 women were enrolled in the study who were either clinician referred or self-referred to cardiology practices for initial evaluation. Referrals occurred for a variety of reasons, including initial evaluation of symptom(s) (n = 469, 42.75%), to establish a relationship with a cardiologist (n = 286, 26.07%), or other reasons such as for clearance before surgery. Of the 1114 women, 3 were found ineligible and 14 did not complete the study (missed a total of 3 interviews or could not be located). Among the remaining 1097 (98.5%) women, a total of 77 (7%) experienced cardiac events during the 2 years of follow-up (Table 1). Most (57.1%) cardiac events were stent placements, either alone (38.9%) or in combination with angioplasty (18.2%). Ten of the 77 women had MIs, and 4 women died of cardiac-related causes.
The characteristics of the 1097 women who completed the study are summarized in Table 2, stratified by the occurrence of a cardiac event. Briefly, most of the women were white (86.5%) and older than 50 years (61.1%), had more than a high school education (64%), and were married (60.8%). Most of the women were overweight or obese (74.3%), and nearly all reported a family history of CHD (98.1%). The 77 women (7%) who experienced events during the 2-year follow-up were generally older and had less education, lower incomes, and more chronic conditions than did those who did not experience a cardiac event.
Predictive Utility of McSweeney Acute and Prodromal Myocardial Infarction Symptom Survey Cumulative Prodromal Symptom Scores and Prodromal Symptom Counts
To determine whether MAPMISS cumulative PS scores and total PS counts were significantly associated with the occurrence of cardiac events, controlling for known clinical and sociodemographic risk factors, we first entered the cumulative PS score, computed as the sum of individual PS scores (range, 0–630), as a predictor. Results were similar whether we adjusted for race and age only or for all the covariates listed in Table 2. After adjusting for all covariates, the HR for subsequent occurrence of a cardiac event increased by 10% for every 10-unit increase in the cumulative prodromal score (HR, 1.10; 95% CI, 1.06–1.13). The total PS count, that is, the number of PSs endorsed by each woman at each visit (range, 0–30), was also significantly associated with risk of a subsequent cardiac event. After adjusting for all covariates, the HR increased by approximately 17% for every additional PS reported (HR, 1.17; 95% CI, 1.13–1.21).
The Most Predictive Prodromal Symptom Subsets
Because both the MAPMISS cumulative PS scores and total PS counts were predictive of subsequent cardiac events, we explored individual PS data to identify the most parsimonious models for predicting cardiac events. We again began by examining PS scores incorporating symptom intensity and frequency and then looked at the simpler measures of symptom presence/absence.
Prodromal Symptom Scores
We looked at the relationships between each of the 30 individual PS scores and the risk of a cardiac event, first adjusted for age and race only and then for all covariates. The HRs and 95% CIs from these analyses are shown in Table 3. Estimates obtained after adjusting for all covariates were similar to those obtained after adjusting for age and race only. Figure 1 illustrates the adjusted HRs and CIs for the 10 PSs with the highest HRs based on individual PS scores.
We then constructed a parsimonious model taking the individual PS scores for all symptoms into consideration and retaining all covariates listed in Table 2. The adjusted HRs and the corresponding 95% CIs estimated from this Cox proportional hazards model are shown in Table 4. Three symptoms were significantly associated with increased risk of a cardiac event whether we adjusted for all covariates or for age and race only: general chest discomfort, shortness of breath, and unusual fatigue. The C statistics and, thus, the predictive ability of these models were high (0.78 when adjusted for age and race only and 0.82 when adjusted for all covariates).
Prodromal Symptom Presence/Absence
Again, we first modeled each of the 30 PSs individually, adjusting for age and race only and then for all covariates. The HRs and 95% CIs from these analyses are shown in Table 5. Estimates adjusted for all covariates were similar to estimates adjusted for age and race only. Figure 2 shows the HRs and CIs for the 10 PSs with the highest HRs based on individual symptom occurrence (presence/absence).
We then constructed a parsimonious model taking all symptoms into consideration and retaining all covariates listed in Table 2. The adjusted HRs and corresponding 95% CIs for the retained PSs estimated from this parsimonious Cox proportional hazards model are presented in Table 6. When we adjusted for race and age only, 4 symptoms were significantly associated with increased risk of a cardiac event: discomfort in jaws/teeth, unusual fatigue, discomfort in arms, and shortness of breath. Women reporting 1 or more of these symptoms were more than 4 times as likely to experience an adverse cardiac event as women with none of these 4 symptoms (HR, 4.19; 95% CI, 2.63–7.44). When we added the other Table 2 covariates to the model, general chest discomfort was significantly associated with increased risk, along with the other 4 symptoms. Women reporting 1 or more of these symptoms were almost 4 times as likely to experience an adverse cardiac event as women with none of these 5 symptoms (HR, 3.97; 95% CI, 2.32–6.78). The C statistics and, thus, the predictive ability for these models were also high (0.81 when adjusted for age and race only and 0.84 when adjusted for all covariates).
To better understand the nature of the chest discomfort reported by some of the women, we examined the sensations/descriptors associated with this discomfort (Table 7). Of the 1097 women in the study, 620 (57%) endorsed having chest discomfort in any 1 of the 4 chest symptom locations at any time during follow-up. The 4 chest locations were left side of the chest, general chest, right side of the chest, and centered high in the chest. Interestingly, although the difference was not statistically significant, women with subsequent events endorsed having chest discomfort less frequently (49%) than did women without subsequent events (57%) (P = .19). Among women who did not experience an event but who endorsed chest discomfort as a symptom, most described the discomfort as pressure (45%), tightness (29%), or sharpness (28%). In contrast, the women who experienced an event and reported chest discomfort described it primarily as either pressure (47%) or tightness (26%). We also examined discomfort in the back, between the shoulder blades. Back discomfort at any time during follow-up was endorsed by 526 of the 1097 women in the study (48%). Women without subsequent events endorsed back discomfort more often (49%) than did women with subsequent events (39%) (P = .10).
Models for prediction of heart disease have been in use for many decades, but they lack specificity and sensitivity for women, making it difficult for clinicians to know which risk model to use for routine screening. A recent article examined the accuracy of the most commonly used prediction models in women in the United States and concluded that many questions remain, and it is unclear how to use these models to make treatment decisions.25 Models such as Framingham9,26 and the Reynolds Risk score have been useful in calculating long-term risk, and current efforts are underway to add additional biomarkers to improve cardiovascular risk prediction.9,26–31 However, the 10 Q Report: Advancing Women’s Heart Health Through Improved Research, Diagnosis and Treatment and others concluded that lack of sufficient numbers of women and gender-specific analysis in clinical trials make it difficult for clinicians to know how to adequately assess a woman’s risk of CHD or to modify this risk.32,33 This report also emphasized the need to identify symptoms other than chest pain that may be suggestive of CHD in women. Our study identified specific symptoms that were predictive of progression to a cardiac event in a sample of women followed over a 2-year period who did not have a CHD diagnosis at the time of enrollment into the study.
This study demonstrated the predictive utility of the PS section of the MAPMISS instrument. Both MAPMISS cumulative PS scores and total PS counts were significantly associated with subsequent development of a cardiac event among women, independent of traditional risk factors. After adjustment for clinical and sociodemographic characteristics, the risk of a cardiac event increased by approximately 10% for every 10-unit increase in the cumulative MAPMISS PS score and approximately 17% for every additional PS reported. The latter finding may be especially useful in the clinical setting because it is quicker and easier to count symptoms endorsed than to score and sum PS intensity and frequency. Furthermore, presence/absence reports are also likely to be less vulnerable than estimates of symptom intensity to individual differences, including symptom tolerance. This is important because it is unlikely that additional assessments will be incorporated into a clinical practice or prediction rule unless they minimally increase workload and do not bring additional expense.34 Assessment of individual symptoms included in the MAPMISS is easy to administer (yes/no) and adds minimal time and expense during a clinical encounter.
In the present study, we also looked at which symptoms and symptom scores were most predictive of a subsequent cardiac event. Even after adjusting for known clinical and sociodemographic risk factors, women who reported 1 or more of the following symptoms were approximately 4 times as likely to experience an adverse cardiac event as women reporting none of the symptoms (HR, 3.97; 95% CI, 2.32–6.78): discomfort in jaws/teeth, unusual fatigue, discomfort in arms, general chest discomfort, and shortness of breath. This suggests that a subset of MAPMISS PS items may provide a useful screen for primary care clinicians considering whether to refer women for additional cardiac work-up and provide key information to inform clinical management decisions. However, it is essential that this subset undergo further testing in primary care samples to confirm its usefulness in identifying women with early symptoms most indicative of CHD in need of further evaluation. The literature highlights the importance of identifying symptoms predictive of CHD in women.32,33 Only future research in primary care populations can determine whether the subset of MAPMISS symptoms identified in this study can meet this need.
Historically, chest pain has been considered the signature indicator of CHD. Interestingly, in our study, women who did not experience a cardiac event were more likely to report chest discomfort or pain in 1 or more sites (right, left, high, general) than were women who had such an event. This highlights the need for clinicians to consider a wider spectrum of symptoms than just chest pain or discomfort, especially unusual fatigue, discomfort in jaws/teeth, and shortness of breath, when evaluating women for risk of progressing to a CHD event. Although there is increasing awareness that women commonly experience symptoms other than or in addition to chest pain or discomfort, this study is the first to identify specific PSs that are predictive of subsequent cardiac events. The results indicate that these symptoms might precede an acute MI by weeks or months and should be useful to both clinicians and women. Earlier recognition of symptoms should contribute to decreasing MI-associated mortality rates in women.
In addition to its many strengths (sample size, sample diversity, 2-year longitudinal design), the study had several limitations. Data were based on self-report, making them vulnerable to recall bias. To minimize the risk of recall bias, we used a short (3-month) follow-up interval, worded questions simply, and, if the RA noted any hesitation in the participant’s response(s), rephrased the question. In addition, throughout the interview, the RAs frequently verified participants’ understanding of MAPMISS items. Moreover, the SBT cognitive screen was administered at baseline and again at 12 months to decrease the influence of change in cognitive status. None of the participants scored higher than 16 (ie, positive for cognitive impairment) on the SBT at 12 months. Individuals may also differ in their perceptions of symptom severity and, consciously or unconsciously, could exaggerate or minimize their reports of symptom severity or frequency. Therefore, with the participants’ informed consent, we audited their medical records over the 2-year period, to verify the information they provided. No discrepancies were noted. Finally, although we had sufficient overall power, subgroup analysis was limited by the relatively small number of events observed. Future studies with a greater number of events would enable additional subgroup analyses.
The devastating impact of CHD is largely preventable if the condition is diagnosed early in the disease process. Instruments to facilitate early identification of high-risk individuals are vitally important. This need is especially great for women who often present with a variety of symptoms. This study demonstrated that a valid and reliable symptom instrument, the MAPMISS, can be used by clinicians in assessing women’s CHD risk. Using the MAPMISS, this study identified specific PSs that are predictive of future cardiac events. We observed that women who reported 1 or more of 5 of the MAPMISS PS items, (1) discomfort in jaws/teeth, (2) unusual fatigue, (3) discomfort in arms, (4) general chest discomfort, or (5) shortness of breath, were 4 times as likely to experience an adverse cardiac event as women reporting none of those symptoms. Informing clinicians and women of the implications of these symptoms could contribute to early detection and initiation of treatment to prevent or delay progression to a cardiac event. This in turn could assist in reducing the high disability and mortality rates in women associated with acute MI.
What’s New and Important
- The MAPMISS may be useful in predicting coronary events in women and may be useful in clinical screening.
The authors wish to acknowledge the assistance of David M. Mego, MD and Beth Crowder, PhD, APRN in recruitment of subjects.
1. Wenger NK. Women
and coronary heart disease: a century after Herrick: understudied, underdiagnosed, and undertreated. Circulation. 2012; 126 (5): 604–611.
2. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics—2012 update. Circulation. 2012; 125: e12–e230.
4. Romeo K. The female heart: physiological aspects of cardiovascular disease
. Dimens Crit Care Nurs. 1995; 14 (4): 170–177.
5. Canto JG, Shlipak MG, Rogers WJ, et al. Prevalence, clinical characteristics, and mortality among patients with myocardial infarction
presenting without chest pain. JAMA. 2000; 283 (24): 3223–3229.
6. Canto JG, Rogers WJ, Chandra NC. Differences in men and women
in heart attack treatments and outcomes are not explained by insurance status [abstract]. AHRQ Research Activities. 2002; 263: 3–4. http://www.ahrq.gov
. Accessed July 7, 2003.
7. Canto JG, Goldberg RJ, Hand MM, et al. Symptom presentation of women
with acute coronary syndromes: myth vs reality. Arch Intern Med. 2007; 167 (22): 2405–2413.
8. D’Agostino RB Sr, Pencina MJ. Invited commentary: clinical usefulness of the Framingham cardiovascular risk profile beyond its statistical performance. Am J Epidemiol. 2012; 176 (3): 187–189.
9. Cook NR, Paynter NP, Eaton CB, et al. Comparison of the Framingham and Reynolds Risk Scores for Global Cardiovascular Risk Prediction in the Multiethnic Women
’s Health Initiative. Circulation. 2012; 125 (14): 1748–1756.
10. McSweeney JC, Cody M, O’Sullivan P, Elberson K, Moser DK, Garvin BJ. Women
’s early warning symptoms of acute myocardial infarction
. Circulation. 2003; 108 (21): 2619–2623.
11. Graham MM, Westerhout CM, Kaul P, Norris CM, Armstrong PW. Sex differences in patients seeking medical attention for prodromal symptoms before an acute coronary event. Am Heart J. 2008; 156 (6): 1210–1216.
12. McSweeney JC, Cleves MA, Zhao W, Lefler LL, Yang S. Cluster analysis of women
’s prodromal and acute myocardial infarction
symptoms by race and other characteristics. J Cardiovasc Nurs. 2010; 25 (4): 311–322.
13. McSweeney JC, O’Sullivan P, Cleves MA, et al. Racial differences in women
’s prodromal and acute symptoms of myocardial infarction
. Am J Crit Care. 2010; 19 (1): 63–73.
14. McSweeney JC, O’Sullivan P, Cody M, Crane PB. Development of the McSweeney Acute and Prodromal Myocardial Infarction
Symptom Survey. J Cardiovasc Nurs. 2004; 19 (1): 58–67.
15. McSweeney JCM, Fischer E, Rojo MAN, Moser D. Reliability of the McSweeney Acute and Prodromal Myocardial Infarction
Symptom Survey among black and white women
[published online ahead of print October 8, 2012]. Eur J Cardiovasc Nurs. doi:10.1177/1474515112459989.
16. Ball LJ, Bisher GB, Birge SJ. A simple test of central processing speed: an extension of the Short Blessed Test. J Am Geriatr Soc. 1999; 47 (11): 1359–1363.
17. Katzman R, Brown T, Fuld P, Peck A, Schechter R, Schimmel H. Validation of a short orientation-memory-concentration test of cognitive impairment. Am J Psychiatry. 1983; 140 (6): 734–739.
18. Wilkins CH, Wilkins KL, Meisel M, Depke M, Williams J, Edwards DF. Dementia undiagnosed in poor older adults with functional impairment. J Am Geriatr Soc. 2007; 55 (11): 1771–1776.
19. Shtatland ES, Kleinman K, Cain EM. A new strategy of model building in PROC LOGISTIC with automatic variable selection, validation, shrinkage, and model averaging. Cary, NC: SAS Institute, Inc.; 2008:191–229.
20. Shtatland ES, Kleinman K, Cain EM. Model building in proc phreg with automatic variable selection and information criteria. Cary, NC: SAS Institute, Inc.; 2005:1–10.
21. Harrell FE, Califf RM, Pryor DB, Lee KL, Rosati RA. Evaluation the yield of medical tests. JAMA. 1982; 247 (18): 2543–2546.
22. Pencina MJ, D’Agostino RB. Overall C as a measure of discrimination in survival analysis: model specific population value and confidence interval estimation. Stat Med. 2004; 23 (13): 2109–2123.
23. STATA [computer program]. Version 12.0. College Station, TX: StataCorp; 2012.
24. SAS [computer program]. Version 9.3. Cary, NC: SAS Institute Inc; 2012.
25. Michos ED, Blumenthal RS. How accurate are 3 risk prediction models in US women
? Circulation. 2012; 125 (14): 1723–1726.
26. D’Agostino RB Sr, Ramachandran S, Vasan RS, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008; 117 (6): 743–753.
27. Tzoulaki I. Assessment of claims of improved prediction beyond the Framingham risk score. JAMA. 2009; 302 (21): 2345–2352.
28. Ridker PM, Buring JE, Rifai N, Cook NR. Development and validation of improved algorithms for the assessment of global cardiovascular risk in women
: the Reynolds Risk Score. JAMA. 2007; 297 (6): 611–619.
29. Arant CB, Wessel TR, Ridker PM, et al. Multimarker approach predicts adverse cardiovascular events in women
evaluated for suspected ischemia: results from the National Heart, Lung, and Blood Institute-sponsored Women
’s Ischemia Syndrome Evaluation. Clin Cardiol. 2009; 32 (5): 244–250.
30. Pelter MM, Riegel B, McKinley S, et al. Are there symptom differences in patients with coronary artery disease presenting to the emergency department ultimately diagnosed with or without acute coronary syndrome? Am J Emerg Med. 2012; 30 (9): 1822–1828.
31. Meisel ZF, Armstrong K, Mechem CC, et al. Influence of sex on the out-of-hospital management of chest pain. Acad Emerg Med. 2010; 17 (1): 80–87.
32. Wenger NK. Cardiovascular disease
: the female heart is vulnerable. Clin Cardiol. 2012; 35 (3): 134–135.
34. Grady D. Why is a good clinical prediction rule so hard to find? Arch Intern Med. 2011; 171 (19): 1701–1702.
Keywords:© 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
cardiovascular disease; coronary disease; myocardial infarction; women