Sex-based analysis of NSTEMI processes of care and outcomes by hospital: a nationwide cohort study Open Access

Abstract

Background

Contemporary studies demonstrate that non-ST-segment elevation myocardial infarction (NSTEMI) processes of care vary according to sex. Little is known regarding variation in practice between geographical areas and centres.

Methods

We identified 305 014 NSTEMI admissions in the United Kingdom (UK) Myocardial Ischaemia National Audit Project (MINAP), 2010–17, including female sex (110 209). Hierarchical, multivariate logistic regression models were fitted, assessing for differences in primary outcomes according to sex. Risk-standardized mortality rates (RSMR) were calculated for individual hospitals to illustrate the correlation with variables of interest. ‘Heat maps’ were plotted to show regional and sex-based variation in the opportunity-based quality indicator score (surrogate for optimal processes of care).

Results

Women presented older (77 years vs. 69 years, P < 0.001) and were more often Caucasian (93% vs. 91%, P < 0.001). Women were less frequently managed with an invasive coronary angiogram (58% vs. 75%, P < 0.001) or percutaneous coronary intervention (35% vs. 49%, P < 0.001). In our hospital-clustered analysis, we show a positive correlation between the RSMR and the increasing proportion of women treated for NSTEMI (R² = 0.17, P < 0.001). There was a clear negative correlation between the proportion of women who had an optimum OBQI score during their admission and RSMR (R² = 0.22, P < 0.001), with a weaker correlation in men (R² = 0.08, P < 0.001). Heat maps according to the Clinical Commissioning Group (CCG) demonstrate significant regional variation in the OBQI score, with women receiving poorer quality care throughout the UK.

Conclusion

There was a significant variation in the management of patients with NSTEMI according to sex, with widespread geographical variation. Structural changes are required to enable improved care for women.

Graphical Abstract

Open in new tabDownload slide

Introduction

Cardiovascular disease is the leading cause of death in women in Europe and worldwide,¹ accounting for greater than half of all female deaths in Europe, and has worse outcomes and higher mortality than in men.² Despite this, contemporary studies have demonstrated that in acute myocardial infarction (AMI), processes of care vary according to sex. Women have been demonstrated to be less frequently prescribed guideline-directed medical therapy (GDMT), less likely to be investigated by or receive timely invasive coronary angiography (ICA), and less likely to be treated with percutaneous coronary intervention (PCI) or coronary artery bypass graft surgery (CABG).^3–8

Whilst sex disparities in AMI outcomes have been studied extensively, there is limited information regarding whether there is significant variation in practice between geographical areas and individual centres and whether this impacts clinical outcomes. Thus, in this large national European study, we describe sex differences in the clinical characteristics, management strategies, and outcomes of patients presenting with non-ST-segment elevation myocardial infarction (NSTEMI) and explore regional/inter-hospital variations in the quality of care received by women. Finally, we aim to study whether there are sex differences in the association between the quality of NSTEMI care and clinical outcomes.

Methods

Study design

We used the Myocardial Ischaemia National Audit Project (MINAP), a prospective national clinical registry of patients admitted to hospitals in the UK with acute coronary syndrome.⁹ The MINAP dataset consists of 130 variables, including baseline demographics and clinical characteristics, comorbidities, management strategies, pharmacotherapy, in-hospital clinical outcomes, and discharge diagnosis,¹⁰ and there is a minimum standard for the data-completeness of submitted records, which is 95% completion of 43 key variables for NSTEMI patients. Data are submitted by hospital clinical and clerical staff, and approximately 90 000 pseudonymized records annually are uploaded to the National Institute for Cardiovascular Outcomes Research (NICOR). Patient data is collected up to the point of hospital discharge, without external adjudication of study endpoints. There is annual data validation at each participating hospital using randomly selected patient records and re-entry of data.¹¹

Study population

We included patients admitted with a diagnosis of NSTEMI in any of the 230 participating hospitals in England and Wales between 1 January 2010 and 31 March 2017. The discharge diagnosis of NSTEMI was determined by local clinicians according to presenting history, clinical examination, and the results of inpatient investigations in keeping with the consensus document of the Joint European Society of Cardiology (ESC) and American College of Cardiology.¹² Patients were excluded if they had missing data in our key variables for investigation: sex, in-hospital mortality, and major adverse cardiovascular events (MACE). Furthermore, any individual patient's subsequent admissions were excluded from analysis (Figure 1).

Outcomes

Primary

Primary outcomes of interest included in-hospital all-cause mortality and MACE (composite endpoints of in-hospital mortality and reinfarction).

Secondary

Secondary outcomes of interest included cardiac mortality [death attributable to myocardial ischaemia or infarction, heart failure, and cardiac arrest of unknown cause] and major bleeding.

Quality indicators

We assessed the (ESC) Association for Acute Cardiovascular Care (ACVC) quality indicators (QI's),¹³ including the use of ICA within 72 h of admission; the assessment of left ventricular (LV) function while an inpatient; the use of fondaparinux or low molecular weight heparin (LMWH); and the prescription of P2Y₁₂ inhibition, adequate dual antiplatelet therapy (DAPT), and statins on discharge. For patients with moderate and severe LV systolic dysfunction (LVSD), the use of angiotensin converting enzyme inhibitors (ACEi) or angiotensin receptor blockers (ARB) and beta blockers on discharge was also evaluated. The ESC QI for LVSD is defined as an ejection fraction (EF) less than or equal to 40%. The MINAP database does not have the same cut-off points for LVSD, thus moderate (EF <49%) and severe LVSD (EF <30%) were used as surrogates. Furthermore, MINAP does not record the specific type or dose of statin prescribed, so ‘statin prescription’ was used as a surrogate for high-intensity statin.

Statistical analysis

Demographics, clinical characteristics, and crude adverse outcomes of patients by sex were compared using the Pearson chi-squared test for categorical variables. Continuous variables were compared using the Student's t-test if normally distributed and using the Wilcoxon rank sum test if not. Normality of the distribution was assessed using the Shapiro-Wilks test. Continuous variables are presented as medians and interquartile ranges (IQR), and categorical variables by proportions. Multiple imputations with chained equations (MICE) were used to impute values for variables with missing data. MICE is considered to be the best practice when dealing with missing data and can provide unbiased estimates even when levels of missing data are significant, and also some protection when the pattern of ‘missingness’ is not random.¹⁴ For each binary outcome of interest, multivariable logistic regression analysis was applied to imputed datasets to estimate the risk of adverse outcomes between groups. Estimates were combined using Rubin's rules.¹⁵ Hierarchical logistic regression models with patients nested within hospitals using maximum likelihood estimation were adjusted for: age, ethnicity, creatinine, heart rate and systolic blood pressure on admission, history of angina, family history of coronary artery disease, co-morbid conditions (hypertension, hypercholesterolaemia, diabetes, smoking and history of asthma or COPD), pharmacotherapy [prescription of LMWH, unfractionated heparin (UFH), warfarin, GP 2b/3a inhibitor, IV nitrate, furosemide, aldosterone antagonist, fondaparinux, beta blockers, ACEi/ARBs, aspirin, P2Y12 inhibitor statins, thiazide diuretics], cardiac arrest, and procedures including ICA and revascularization (by PCI or CABG) during admission. A second model assessing the interaction between the Opportunity Based Quality Indicator (OBQI) score category and sex was performed, adjusting with the aforementioned variables. OBQI score was ordered into four categories, ‘Excellent’, ‘Good’, ‘Fair’, and ‘Poor’. Excellent refers to an OBQI score of ≥90 and ≤100, Good a score of ≥80 and <90, Fair ≥70 and <80, and Poor <70. The difference in adjusted mortality between categories of OBQI score was used to multiply the number of female patients per region to estimate the potential number of lives saved if women in that region received the same category of care as men.

Temporal and geographical changes

Risk-standardized mortality rates (RSMR), adjusted for patient demographics, were calculated for each hospital included in our study. The mortality assessed was in-hospital mortality only. We undertook logistic regression to ascertain whether there was a correlation with the adjusted mortality rates and proportion of patients that were female. The OBQI is a composite score comprising the prescription of aspirin, thienopyridine inhibitor, β-blocker, ACEi, HMG CoA reductase enzyme inhibitor (statin), and enrolment onto a cardiac rehabilitation programme at the time of discharge. The OBQI reflects the number of care opportunities fulfilled at each hospital (numerator) divided by the number of opportunities to provide care (denominator). Excluded from both the numerator and denominator were interventions that were contra-indicated, not applicable, not indicated in, or declined by, individual patients. We formed scatterplots to assess the correlation of RSMR for each hospital with the mean OBQI score for each sex. We created ‘heat maps’ according to mean the OBQI according to 2021 England and Wales Clinical Commissioning Group regions (CCGs), clustering hospitals together according to 2021 CCG geographical boundaries. ‘Heat maps’ were formed for the overall mean OBQI for each CCG and for women and men separately.

Statistical analysis was undertaken using Stata 14.2 (College Station, Texas, USA). All statistical analyses were two-tailed, with an alpha of 5% used.

Results

Demographics

During the study period, there were 369 435 patients admitted to England and Wales with a diagnosis of NSTEMI. Applying the exclusion criterion left an analytical cohort of 305 014 patients, of which 110 209 were women (36%) (Figure 1). The differences in clinical characteristics between the two groups are presented in Table 1. Women were significantly older at presentation [(77 years old (IQR 67–85) vs. 69 years old (IQR 59–79), P < 0.001], less frequently from an ethnic minority background (7% vs. 9%, P < 0.001), less frequently had a previous history of smoking (29% vs. 41%, P < 0.001) or current smoking (18% vs. 24%, P < 0.001), were less likely to be admitted under a consultant cardiologist (44% vs. 52%, P < 0.001) or admitted to a dedicated cardiology ward (48% vs. 56%, P < 0.001), and more frequently presented with a ‘High-risk’ GRACE score (>140) (82% vs. 71%, P < 0.001) than men.

Management and clinical outcomes

Pharmacotherapies, management strategies, and unadjusted crude clinical outcomes for both cohorts are presented in Table 2. Women were less frequently prescribed statins (80% vs. 84%, P < 0.001), ACEi or ARBs (70% vs. 74%, P < 0.001), beta-blockers (79% vs. 82%, P < 0.001), and P2Y12 inhibitors (90% vs. 92%, P < 0.001) (Table 2). They less frequently underwent ICA (58% vs. 75%, P < 0.001), PCI (35% vs. 49%, P < 0.001), or CABG surgery (5% vs. 9%, P < 0.001) during their index NSTEMI admission. Women had a higher frequency of in-hospital mortality (7% vs. 5%, P < 0.001), MACE (7% vs. 5%, P < 0.001), cardiac mortality (5% vs. 4%, P < 0.001), and major bleeding (2% vs. 1%, P < 0.001) when compared to men in our unadjusted data.

ESC ACVC quality indicators

Women less frequently received ICA within 73 hours’ time frame of admission (60% vs. 66%, P < 0.001). They were less frequently discharged on DAPT (82% vs. 84%, P < 0.001) and were less frequently discharged on a combination of DAPT and statin therapy (composite score) (68% vs. 73%, P < 0.001) (Table 3).

Multivariate analysis

Following adjustment for differences in age, ethnicity, serum creatinine, baseline comorbidities, LV systolic function, Killip classification, and cardiac arrest, women had significantly greater odds of in-hospital mortality (OR; 1.22, 95% CIs 1.17–1.28, P < 0.001) or MACE (OR; 1.18, 95% CIs 1.13–1.23, P < 0.001), which persisted when additionally adjusting for medication (in-hospital mortality OR; 1.12, 95% CIs 1.07–1.16, P < 0.001 and MACE OR; 1.10, 95% CIs 1.06–1.14, P < 0.001). When additionally adjusting for revascularization, there was no statistical significance in the odds of in-hospital mortality (OR; 0.99, 95% CIs 0.94–1.03, P = 0.52) or MACE (OR; 0.99, 95% CIs 0.94–1.05, P = 0.85) in women compared to men (Table 4).

Hospital-clustered analysis

There is a positive correlation between the increasing proportion of women treated for NSTEMI in each individual hospital and RSMR [coefficient 0.034 (95% CIs; 0.024−0.438), P < 0.001, (R² = 0.17)] (Supplementary material online, Figure S1). There is a stronger negative correlation between mean OBQI and RSMR in women [coefficient −0.22, (95% CI: −0.278 to −0.168), R² = 0.22, P < 0.001] (Figure 2), than is seen in men [coefficient −0.13, (95% CI: −0.190 to −0.076), R² = 0.08, P < 0.001] (Figure 3).

Heat maps

Figure 4 is a ‘heat map’ of the UK CCGs according to the mean OBQI score. Figure 5 is a collection of ‘heat maps’ of the mean OBQI score for each UK 2021 CCG according to sex. This shows a marked regional variation in the mean OBQI score throughout the UK, but notable are the lower mean scores in Wales, Shropshire, the Northwest, and the East of England. Figure 5(A) shows that for women only, the mean OBQI is lower throughout the UK but particularly poor in the same regions (Wales, Shropshire, Northwest, and East of England). The equivalent ‘heat map’ for men (Figure 5(B)) shows a higher mean OBQI score throughout the UK, with particularly high mean OBQI scores in the Northeast, East Midlands, and Southeast.

Supplementary analysis

Table 5 shows the adjusted mortality rates for the sex and OBQI care categories. A lower proportion of women received ‘good’ or ‘fair’ care compared to their male counterparts. ‘Poor’ and ‘excellent’ care was similar between the sexes. Table 6 shows the mean OBQI scores for men and women in all regions of the UK. There are no regions where care is deemed ‘excellent’ for women. Comparatively, care in south-east London is rated as ‘excellent’ for men. Furthermore, there are no geographical regions where care is superior for women compared to men.

Discussion

Our nationwide analysis of over 300 000 NSTEMI patients in the UK revealed a number of important findings. First, there are marked sex-based differences in the quality of care received by patients admitted with a NSTEMI, with women less frequently treated on a dedicated cardiology ward or admitted under a cardiologist, less frequently treated with GDMT, and less frequently in receipt of coronary revascularization. Second, we report marked regional variation in the quality of care received by women, with no geographical areas where the care for women was noted to be superior to that of men. Third, the difference in care between men and women varies significantly by both the individual region and hospital. Fourth, in our hospital-clustered analysis, we show a positive correlation between the RSMR and the increasing proportion of women treated for NSTEMI. Finally, there is a stronger negative correlation between the mean opportunity based quality indicator and RSMR in women than is seen in men, suggesting that there may be greater potential gains in improving the quality of care in women than in men.

Prior studies examining sex disparities in AMI have several important limitations and contextual differences. Several studies have been undertaken in the USA, a multipayer health care system in which the health insurance of patients may affect the quality of care they receive, both with respect to primary prevention and acute illnesses. Furthermore, previous studies have predominantly focused on AMI collectively, or just STEMI. Given the less algorithmic management of NSTEMI, larger inequalities in the quality of care are more likely. Our study is the first to assess sex disparities in NSTEMI on a nationwide scale with granular interhospital variability.

Our heat maps of the UK CCG's according to the mean OBQI score show that the overall care is inferior for women compared to men. The overwhelming majority of geographical areas for women are rated as ‘fair’ whereas for men, the overwhelming majority are rated as ‘good’. There are also more areas rated as ‘poor’ and fewer areas rated as ‘excellent’ for women in comparison to men. Geography has previously been shown to be a key determinant of health, and location has been shown to affect access to treatment as well as the baseline cardiometabolic risk factor profile of an individual.¹⁶ Baum et al. showed in their retrospective cohort study of over 5 million participants that in the United States (US), moving from a 10th to a 90th percentile zip code for a given outcome was associated with a significantly increased prevalence of uncontrolled blood pressure, diabetes, and obesity.¹⁶ Tu et al., as part of their retrospective cohort study involving 5.5 million patients in Ontario, Canada, showed that access to healthcare was better in geographic regions where mortality was lower. Specifically, residents in those areas received physician services more often, were more likely to hit targets for hypertension and hypercholesterolaemia, and were screened more often.¹⁷

Our scatter plots show a negative correlation of the RSMR and mean OBQI, with significantly lower RSMR associated with improved OBQI scores. This relationship is far stronger for women compared to men. Although the UK is a publicly funded healthcare system, where you would expect a degree of homogeneity in AMI management between hospitals, Chung et al. showed that there is greater variation in 30-day mortality post-AMI in the UK compared to Sweden, which they attribute to greater heterogeneity in the prescription of GDMT amongst UK hospitals.¹⁸ It is not necessarily clear whether this is reflective of poorer practice in the UK or whether this is reflective of excellence in the Swedish system. Gale et al. have shown this hospital-wide variation in practice across the UK using MINAP data, with numerous hospitals shown to be below control limits for ‘door to needle time’ and for aspirin and beta-blocker usage post AMI.¹⁹ Interestingly, the observed variation between hospitals may not all be attributable to in-hospital processes of care, with Asaria et al. recently suggesting that a significant proportion of inter-hospital variation in mortality is due to regional myocardial event rates,²⁰ influenced by risk factors such as smoking, hypertension, and hypercholesterolaemia. The strong relationship between the RSMR and OBQI scores in women compared to men suggests that there are significant gains to be made with respect to improving the mortality of women presenting with NSTEMI by improving the quality of care that they receive in the hospital.

It is important to note that the poorer quality of care for women is not just confined to the UK and US, with registry data from Portugal suggesting performance in AMI management with regard to time to ICA and GDMT prescription (using the ESC QIs) is poorer in women,²¹ and data from Australia suggesting that even after adjusting for comorbidities, women are less likely to undergo coronary intervention.²² Our results showed that when adjusting for baseline characteristics or when adjusting for baseline characteristics and management strategy (inclusive of medications), women had worse primary outcomes of in-hospital mortality and MACE and worse secondary outcomes of cardiac mortality. When additionally adjusting for invasive strategy and revascularization, there were no significant differences in primary or secondary outcomes between men and women, highlighting the importance of not just guideline-directed medical treatment but the importance of an invasive strategy in reducing the disparities in key outcomes such as mortality between men and women.

Our study has important clinical implications for practice. Following the passage of the 2022 Health and Care Act in the UK, significant reforms have shifted the way the English health and care system is organized.²³ The formation of integrated care systems, which are partnerships that bring together NHS organizations, local authorities, and others to take collective responsibility for planning services, improving health, and reducing inequalities across geographical areas, has resulted in 42 autonomous geographical entities in the UK. Whilst traditionally in many countries, financing and autonomy of healthcare systems have been a top-down approach, with regional variation less pronounced, advanced healthcare systems, including those in the US, differ significantly based on state, with funding and arrangements of healthcare services differing widely between the different states. With an increasing population and increasing geographical disparities in healthcare, it would be expected that over time, the regional autonomy of healthcare budgets will become more pronounced. Our study shows wide geographic variation in the quality of care received for women, with specific geographic areas having a worse mean OBQI (Wales, Shropshire, Northwest, and East of England), with women being more adversely affected in these regions. With the aim of these reforms to uniform the quality of care between geographic areas, our study highlights the significant geographic disparities in care for NSTEMI patients, particularly in areas where significant ‘levelling up’ is required for women.

Strengths of study

There are a number of strengths to this study. Our analysis represents the largest study to date that looks at differences in NSTEMI patients by sex in the UK and the first to look at interhospital variability in quality of care in such granular detail. The MINAP database encapsulates an almost complete record of NSTEMI patients admitted in the UK and represents one of the largest national real-world databases of this cohort of patients in the world, including those who are high-risk and have multiple comorbid illnesses, such that they are either not included or under-represented in clinical trials.

Limitations of study

There are important limitations to observational studies of this type. The MINAP data registry shares the weaknesses of other national registries, including self-reporting of adverse events with no external validation. Although the MINAP dataset included important clinical and demographic variables of interest, there are limitations to the data collected. For instance, the database does not capture frailty score, severity of CAD, psychosocial risk factors, access to use of healthcare, rationale for specific medications, or an exhaustive comorbidity list. Furthermore, the database does not capture markers of inflammation, biomarkers, LDL-c levels, or less common risk factors such as malignancy, lipoprotein(a) or clonal haematopoiesis of indeterminate potential. Furthermore, inherent to large registries such as MINAP is the issue of missing data, we have attempted to mitigate against this by using MICE. We acknowledge that there is no external adjudication of diagnosis and endpoints of cases in our registry and that there may be a range of practices across UK hospitals, including the usage of different troponin assays and changes in clinical practice regarding GDMT, cardiac rehabilitation referral, and invasive investigations over what is a relatively long study period; therefore, our results must be interpreted in the context of this. For example, in a small minority of multimorbid patients, there may be a misclassification of type-two myocardial infarction for type-one myocardial infarction, where our quality indicators may not be relevant to this population. For our OBQI score calculations, patients for which pharmacotherapy agents were contraindicated or not-applicable are excluded from the numerator and denominator; however, it is possible that in the data entry, there may be patients for which certain agents were legitimately contraindicated/not-applicable that may be recorded as being omitted if the reason for this was not clearly documented or communicated.

Our study used MINAP registry data from 2010 to 2017, whereas our heat maps are based upon 2021 CCG boundaries, due to the lack of availability of editable maps from that range of years. A small number of CCGs were combined into larger groups over this time, which required manual reclassification and recalculation. Finally, in our RSMR scatter plots, there are several outlier values, which reflect hospitals with small numbers of patients; thus, the individual rates in these hospitals should be interpreted with caution.

Conclusion

Our study showed that women presenting with NSTEMI in the UK are typically older and increasingly comorbid. Women received a poorer quality of care, with lower rates of ICA, PCI, and GDMT prescriptions when compared with men. There was wide geographical variation in the mean OBQI scores according to sex. There were no areas in the UK where women had superior care to men. The strong relationship between the RSMR and OBQI scores in women compared to men suggests that there are significant gains to be made with respect to improving the mortality of women presenting with NSTEMI by improving the quality of care that they receive in the hospital. Further work should focus on reducing the geographical disparities in access to high-quality NSTEMI care according to sex.

Acknowledgements

None.

Secondary use of anonymized MINAP dataset for research purposes is authorized under NHS research governance arrangements and further supported under section 251 of NHS act 2006 (NIGB: ECC1-06(d)/2011), which allows researchers to use patient information collected within the dataset for medical research without patient consent. Therefore, formal ethical approval was not sought for this study.

Funding

None.

Conflict of interest:

None declared.

Data availability

The data underlying this article were provided by the National Institute for Cardiovascular Outcomes Research (NICOR). Data will be shared on request to corresponding author with permission of NICOR.

References

Author notes