This study in more than 10,000 participants represents one of the largest prospective cost analyses of ACS and one of only few such analyses to provide cross-country comparisons. We found substantial variations across countries/regions and index diagnosis in healthcare costs incurred by patients during hospitalization for treatment of an ACS event. This is perhaps to be expected given the milieu of high, upper-middle and lower-middle income countries included in EPICOR Asia. What seems surprising is that the cost of treating ACS appeared relatively high in China across all three index-event types, exceeding those recorded for high-income countries/regions such as Singapore, the Republic of Korea, and Hong Kong. Interestingly, in contrast to Hong Kong and Singapore, costs in China were generally similar across all three index-event types. This suggests stratification of patients may not have been optimal, with patients at high- and lower-risk variably receiving high-level interventional therapy, and of variable cost. This may be compounded by inaccurate recording as to whether percutaneous transluminal coronary angioplasty was provided with or without stenting. Further detailed study is required to establish the multifarious factors underlying the apparent high costs of treatment in China and alleviate any concerns to decision makers given the increasing burden of ACS and growing proliferation of treatment. For example, between 2007 and 2011, there was a virtual doubling in the number of PCI procedures performed from 180,000 to 330,000 [19].
The finding that age is a positive predictor of high cost is consistent with potentially greater complexity and severity of illness, some of which may not have been captured and thus controlled for in the model. This is perhaps further evidenced by the positive association between hospitalization in the 3 months prior to the index event and high costs. Similar findings for male sex are consistent with a study in Italy where costs incurred by men were significantly greater than that for women, irrespective of index-event type [8]. Reports that women who present with ACS may be evaluated less intensively than men, may go some way towards explaining this [20].
The findings reported here also provide evidence of a potential income effect in that patients on relatively high income appear more likely to be categorized as highest cost. Odds ratios relative to income quintile 5 of 0.45 for quintile 1; 0.76 for quintile 2; 0.97 for quintile 3; 0.80 for quintile 4 (although only statistically significant in relation to quintile 2), ostensibly indicate a pattern in which the odds of incurring high costs increase with income. Such a finding, again, accords with expectations that wealthier patients will seek and have access to higher-cost treatments.
The findings that longer length of stay and having an invasive procedure (versus non-invasive medical management) were both positively associated with odds of incurring high costs is consistent with expectations and reflects, perhaps obviously, resource needs associated with longer treatment duration and need for an invasive procedure. Less intuitive is the finding that those patients who required no help with daily activities prior to hospitalization for their index condition (“no dependence”) had significantly higher odds of being in the highest-cost category. Here, it is possible that patients with dependency at baseline would have the ongoing support of a “carer” to rely on. The potential lack of such support for patients without dependency at baseline may have led to greater costs due to a greater need for in-hospital rehabilitation and extensive discharge planning.
Another ostensibly unexpected observation was that the odds of incurring high-cost treatment, relative to those encountered in patients admitted to a university general hospital, were higher for those patients admitted to all “other” categories of hospital, e.g. regional/community/rural hospitals, non-university general hospitals and other type of clinics. Here, it is possible that the multivariable analysis used in this study effectively controlled for factors implicated in higher costs seen in teaching (university) hospitals, such as size of facility, treatment mode, disease history and length of stay. Our findings suggest, therefore, that the independent effect of university status of a hospital was to lower costs, very likely associated with efficiency and an established degree of expertise in such centers.
The lack of significant association between insurance status and high-cost care may allay potential concerns about the inflationary effects of national programs to expand insurance coverage, e.g. due for instance to incentives created by a third-party payer for providers to overcharge/over-service (provider moral hazard) and patients to overuse (patient moral hazard) [21]. Although this study focuses only on ACS patients, the findings of this study found no evidence of an inflationary impact associated insurance coverage. Further country-specific research is needed to determine whether the roll out of social insurance programs will increase costs to any significant degree.
There were several limitations in the present study. First, inclusion only of patients alive and followed up at 6 weeks might suggest a possible survivor bias to the findings. As mentioned, earlier studies have reported in-hospital mortality to be associated with higher costs, suggesting our estimates of average costs may have been underestimated. In addition, the costs examined in this study reflect only health system cost whereas a broader societal perspective would have considered costs to households and the community associated with indirect loss of income and reduced productivity. Also, the costs included in this analysis were confined to hospitalization for the index condition and excluded costs of potential re-hospitalizations for ACS; in the US, such costs have been estimated at over 30% [22], suggesting there are significant costs associated with ACS outside of the scope of this analysis. Furthermore, the costs of sub-acute follow-up care were not included. These may vary across countries due to differences in treatment norms, funding models and other health system characteristics. Despite these potential limitations in capturing the high costs to health systems associated with ACS, the study highlights the major policy challenges associated with a high burden of illness in Asia. Some countries in this analysis were represented by a relatively small number of participants, thus precluding detailed country-specific analyses. Thus, the way in which the primary outcome for this study was specified (i.e. occurrence of cost in the highest quintile specific to each country and index condition) served as a standardized outcome that facilitated the pooling of data across all countries. An alternative approach would have been to adjust for differences in purchasing power by converting into international dollars; however, the problem with such a strategy is that costs reported in international dollars lack meaning for local policy makers since they do not reflect actual budgetary implications (nevertheless the conversions are provided in Additional file 1: Table S2 for reference). The inclusion of hospital length of stay as an explanatory variable and the likelihood of it being highly correlated with cost is a potential weakness in the modelling [23], as we may not be able to identify factors that affect the cost through the hospital length of stay. However, it is an important variable of interest and its inclusion is justified as it allows us to estimate the direct effect of other factors included in the model. Finally, without accounting for clustering in the analysis, variance and confidence intervals could be slightly underestimated. However, in international studies of this kind, it is conventional that such adjustments are not made.
