Introduction

Heart failure (HF) is both a debilitating and costly clinical syndrome, with an expenditure of 346 billion US$ globally [1]. Despite medical advancement in the management acute coronary syndromes, the five-year survival rates for HF is poorer than most cancers [2]. Nevertheless, cardiac remodelling and progression to HF can be modified by appropriate preventive strategies and timely effective treatments are available. Estimates on the average prognosis of HF at the population level are important for monitoring changes in healthcare delivery. On the individual patient level, knowing the absolute risks of mortality and readmission enables shared decision-making between health providers and patients when making plans for disease management [3].

As it is in high-income countries, HF is already demonstrated to be a major burden in middle-income countries and this is expected to pose an equal, if not greater financial and morbidity impact as some of these countries are still grappling with a concurrent infectious disease burden [4]. The amount of data published on the epidemiology of HF from these countries is highly disproportionate compared to developed nations. Relying on data extrapolations from high-income countries is inadequate given the differences in healthcare infrastructure, health spending and demographic composition. The Asian HF registry, which included 11 countries in the region, found that one-year mortality rate was 9.6% among registry patients [5]. However, it was also noted that patients enrolled in this registry were mainly treated in academic hospitals with echocardiography expertise and more resources, and hence not necessarily representative of overall hospitalised patients [5]. Recent data from Malaysia, a middle-income country in Southeast Asia, has shown a 39% rise in the absolute number of incident HF hospitalisations from 2007 to 2016 (Su Miin Ong, MSc, unpublished data, 2020). This highlights an urgent need to fill the existing gaps on temporal trends for clinical outcomes, specifically for Asia [4]. Accordingly, our objective was to estimate the short- and medium-term mortality and readmission outcomes after an incident hospitalisation for HF. We then described the temporal variation in these outcomes by sex and ethnicity in this ethnically diverse country.

Methods

Results

Trends for readmission and mortality

A 17.8% increase in overall readmissions from 16.6% in 2007–2008 to 19.6% in 2015 was observed and this trend remained significant at +2% per calendar year after adjusting for age, sex and ethnicity (p trend < 0.001). Figure 1(a) shows that age-standardised trends for 30-day readmission in men were proportionally higher than the rate for women. Although Others had the lowest readmission rates in 2007, its rise with time was the largest compared to the other ethnic groups (Figure 2 (a)). The mean number of hospitalisations per patient within a year has risen slightly from 1.3 in 2007 to 1.6 in 2015 and a modest increase of 0.04 per calendar year was still evident after accounting for age, sex and ethnicity (p trend < 0.001).

Figure 1 

Trends for (a) 30-day readmission, (b) in-hospital, (c) 30-day and (d) one-year all-cause mortality rates by men and women.

Figure 2 

Trends for (a) 30-day readmission, (b) in-hospital, (c) 30-day and (d) one-year all-cause mortality rates by ethnicity.

Overall, in-hospital mortality rates nearly halved from 6.9% (95% CI 6.6–7.3) in 2007 to 3.7% (95% CI 3.5–3.9) in 2016 with similar trends for men and women (Figure 1(b)). This declining trend remained after adjustment for age, sex and ethnicity (average –7% per year; p trend < 0.001). Improvements in mortality were also evident within 30 days of hospitalisation, seen as a 26% decline from 13.1% (95% CI 12.6–13.6%) in 2007 to 9.7% (95% CI 9.3–10.1%) in 2016. Men had consistently higher 30-day mortality rates than women during the study period (Figure 1(c)). Upon full-model adjustment, the average improvement in 30-day mortality was 4% per calendar year (p trend < 0.001). By contrast, all-cause mortality in one year remained unchanged throughout the study period (p trend = 0.113) with men having almost uniformly higher rates than women (Figure 1(d)).

Despite poorer overall outcomes, Others showed the most pronounced improvements in short-term mortality over time compared to other groups (Figure 2(b) and (c)). The rates for 30-day mortality between Others and Malays narrowed from 6.2% in 2007 to 3.5% in absolute rate difference in 2016 but this difference was not significant after full model adjustment (p interaction = 0.091; Supplementary Table 2). With respect to 1-year mortality, only Others showed significant average decline by 1% per year (p interaction = 0.021) relative to Malays (Figure 2(d), Supplementary Table2).

Cause of readmission and death

Cardiovascular causes accounted for half of all 30-day readmissions (50.1%) (Figure 3) with HF specifically accounting for 27.8%. Analysis on all medically-certified deaths found 56% of patients died of cardiovascular causes within a year from index hospitalisation, with the leading cause being HF (21.1%) (Figure 4). Cardiovascular mortality rates have been decreasing by two percent per calendar year, with adjusted p trend = 0.001 (Supplementary table 3). Despite this decline, a corresponding increase in non-cardiovascular mortality with time resulted in the unchanged overall rates for one-year mortality and a look into the specific causes between two calendar year periods for 2007–2008 and 2012–2013 showed that the contribution of infections as a cause of non-cardiovascular deaths has been increasing (Supplementary Table 4).

Figure 3 

Causes of readmission in 30 days after index hospital admission from 2007–2015.

Figure 4 

Causes of death for patients who died within one year after index admission from 2007–2013.

When the definition of incident HF hospitalisation was increased from a two-year to three -year lookback period, the percentage of reduction in false positives were between 1.7 to 3% annually. Nevertheless, the magnitude of outcome rates, trends and statistical significance of regression estimates were similar to the main results, for both the one-year and three-year definitions of incident hospitalisations (Supplementary Tables 5 and 6).

Discussion

We present here contemporary crude and age-standardised estimates for mortality and readmission rates among hospitalised HF patients in a middle-income Asian country. Using linked population data, this study has revealed improving trends in short-term mortality following an incident hospitalisation for HF in Malaysia. However, mortality at one year has remained constant while readmissions within thirty days rose steadily during the observation period. We have noted distinct differences in patient outcomes by sex and ethnicity. First, readmission rates were consistently higher in men compared to women and in Chinese and Indians compared to Malay patients. Second, overall short-term mortality outcomes were poorest in Others compared to all other ethnicities while men had slightly worse outcomes than women for 30-day mortality. Third, improvements in survival after HF hospitalisation varied by ethnic groups with Others showing the steepest decline in one-year mortality.

In-hospital mortality after incident HF hospitalisation in Malaysia was higher compared to the 4.1% reported from a HF registry in China [17]. Nevertheless, it is necessary to take into consideration that being enrolled in a registry is associated with better outcomes [18]. Next, we observed a seven percent decline for in-hospital mortality for both men and women. This is in contrast to a rise in Brazil, from 8.3% to 10.8% between 2008 and 2017 [19]. Similar to inpatient mortality, the rates of mortality within 30 days was also decreasing, albeit to a smaller extent. No direct comparisons were available for middle-income countries. However, the decline observed here were consistent to those observed in several high-income settings including Western Australia, the Netherlands, and Sweden [20, 21, 22, 23]. Several explanations are possible. The observed decline in short-term mortality is partly a reflection of improving population health, as seen with life expectancy increases from 73.7 years in 2008 to 75 years in 2018 [24]. Other explanations include earlier identification of cases as a result of rising population health awareness, improved pre-hospital emergency services and an increase in the number of medical specialists in the past two decades [25, 26]. MOH hospitals have experienced almost doubling of the number of emergency medicine specialists from 93 to 167 from 2010 to 2013 and a modest increase in the number of cardiologists from 47 to 53 within the same period [27]. There are no known changes in reimbursement practices or implementation of nationwide quality improvement programs during the study period, therefore we expect the coding of ICD-10 to remain consistent over this duration.

In keeping with earlier and recent studies from Western populations [28, 29, 30], we have also showed that men had higher risks of mortality and readmission than women after accounting for age, ethnicity and time trends. Poorer survival and higher readmission rates in men after HF hospitalisation may be explained by a predominant heart failure with reduced ejection fraction subtype among men and higher prevalence of macrovascular disease, myocardial ischemia and infarction, which underlie the aetiology of HF in this subgroup [31]. Unfortunately, the type of HF and aetiology were not available in the present data.

Amidst the overall improvements in short-term mortality, it is necessary to note that striking ethnic differences exists. Others had poorer outcomes than the rest of the population. This difference was apparent even in the presence of under-reporting of deaths in East Malaysia, from which the majority of Others reside. Hence, we expect the true estimates to be even higher than what was observed. The health status of this subgroup is known to be poorer compared to the general population and is characterised by lower socioeconomic status, shorter life expectancy, undernutrition, insulin resistance and lack of trust in modern medicine [32]. Moreover, accessibility to hospitals remains a challenge for some residents of the interior and remote parts of Sabah and Sarawak and it is also likely that this region has a higher prevalence of rheumatic heart disease which may contribute to poorer outcomes among those hospitalised for HF [33, 34, 35]. These findings highlight a need to improve access to healthcare and focus resources to narrow the disparities in short-term mortality particularly in Others. For short-term mortality trends, it is reassuring to observe that the largest disparity in mortality outcomes between Others and Malays have been narrowing. The remaining differences in mortality outcomes between ethnicity represents opportunities for health interventions.

About one in three patients hospitalised for HF in this study die within a year and this has remained fairly constant during the study period. Mortality at one year is almost 1.5-fold compared to the European Society of Cardiology Long Term Heart Failure registry, reflecting differences in income per capita, health systems and patient characteristics between Malaysia and the European and Mediterranean countries which participated in this registry [36]. Trend-wise, death due to cardiovascular disease has decreased and this could be the result of a higher use of disease-modifying therapy such as renin-aldosterone angiotensin system inhibitors and beta-blockers in recent years, as reported in several tertiary centres [37, 38, 39, 40]. Numerous reports have shown that people of South Asian descent are predisposed to higher risk of ischaemic heart disease compared to the rest of the population [41]. Accordingly, our previous study had also found the highest incidence of HF hospitalisations amongst Indians in the Malaysian population. Interestingly though, when it comes to survival, be it short-term or at one-year, Indians had significantly better survival compared to other ethnic groups. This suggests a stronger influence of environmental and behavioural determinants over genetic influences in HF outcomes. Further investigation into use of HF medications and lifestyle factors by ethnic subgroups would hence be warranted.

Preventing decompensation is an important therapeutic goal after a diagnosis of HF. To our knowledge, there is no published data on 30-day readmission trends after HF hospitalisation among middle-income countries. The steady annual rise in 30-day readmissions that we have found were comparable to those reported in Spain, but in contrast to a two percent reduction among Veteran’s Affairs hospital admissions in the United States [42, 43]. It is necessary though, to keep in mind that differences in healthcare financing and infrastructure exist between countries of middle- and high-income economies. Lowering admission threshold for HF is a potential reason for this observed rise in readmission rates in this study. Differential readmission rates by ethnicity were likely related to socioeconomic status and educational level. This is reflected as higher readmission rates among the Chinese and Indians who largely reside in urban locations, whereas greater access and logistical barriers to care exists among Others [44]. However, narrowing of the gap in readmission rates between Others and the population average suggests that physical access for Others to secondary care is improving over the last decade. The overall increasing trend in 30-day readmissions observed here deserve attention from researchers and policy makers alike because hospitalisations incur the greatest financial costs to HF health expenditures and about a quarter of these readmissions are preventable [45]. Standardised strategies to differentiate the severity of patients who present at the emergency department would be useful for risk stratifying them into those who require admission while the rest may be observed and treated on an outpatient basis. We know that half of these readmissions are due to non-cardiovascular causes; therefore, multi-faceted assessments which address all comorbidities can be incorporated into early care transition to outpatient clinics, nurse-led home visits and structured telephone monitoring, all of which have shown moderate effectiveness in reducing rehospitalisations [46, 47].

In this study, we estimated the average prognosis after an index hospitalisation for HF using representative data from a large national database. These findings are generalisable to other middle-income countries with similar government-funded health systems and diverse ethnic composition. While most HF hospitalisation data for middle income countries in literature come from urban tertiary centres [4], we have presented here data across a range of hospitals within the public health sector in Malaysia. Unlike patient selection in disease registries, the inclusion of unselected cases of HF hospitalisations here allowed us to make reliable comparisons between sex and ethnic groups.

There were several limitations in this study. Complete data were available for only primary discharge diagnoses; therefore, the absolute number of HF hospitalisations was likely underestimated. Nevertheless, the trend data is unlikely to be affected by this underestimation as the selection criteria used was uniform across time points. We explored the use of secondary discharge diagnoses as proxy for underlying disease severity but found that data completeness was not consistent across time. Therefore, it is difficult to draw conclusions on the severity of patients who were hospitalised for HF. Information on HF subtypes (reduced, preserved and mid-range ejection fraction), medical history, treatments and device therapy were not available in the discharge database and thus, does not allow for correlation of these factors with HF outcomes. While this analysis encompassed an average of 83% of annual HF hospitalisations in Malaysia, it is necessary to point out that we have included only HF patients who were hospitalised in MoH hospitals. Thus, these results are not generalisable to patients treated in the community setting or private hospitals. The 30-day and one-year mortality estimates were slightly underestimated due to incomplete death registration in East Malaysia. Lastly, there were some data losses in 2012 and 2013 and imputations were not feasible in this situation because the exact number of missing records was not known.

To the best of our knowledge, this is one of the first studies to report national, age-standardised estimates for HF prognosis and trends for hospitalised HF patients in the South East Asian region. The declining trend and narrowing of ethnic differences for short-term mortality showed that these outcomes are amenable to targeted interventions. Moreover, the differential HF outcomes by sex and ethnicity seen here highlights the importance of incorporating these determinants into risk predictions models or when calculating likely accrual endpoints in the design of therapeutic studies.

Conclusion

Gradual improvements in short-term mortality were seen across sex and ethnicities although relative differences between ethnic subgroups remain apparent. The steady rise in 30-day readmission post-discharge and stagnating 1-year mortality rate raises concern, signalling a need for pro-active efforts from policymakers, physicians and research in making HF as a priority disease area.

Additional File

The additional file for this article can be found as follows: