Case fatality ratios for serious emergency conditions in the Republic of Ireland: a longitudinal investigation of trends over the period 2002–2014 using joinpoint analysis

The Republic of Ireland is an island of 70.2 thousand km² on the west of Europe with a population of 4.8 million [21]. It is divided into 26 counties, which for the analysis of emergency care reconfiguration may be grouped into eight regions based on hospital networks identified by the organisation that delivers public healthcare services in Ireland (Fig. 1) [22].

Regional characteristics and reconfiguration of services are presented in online Additional file 1. Two regions (South and Mid-West) have implemented significant reconfiguration of urgent and emergency care over the period 2012–2014. Four regions (West, North-East, South-East and Dublin-South) have introduced some reconfiguration measures since 2006, but these do not cover all services. The two remaining regions (Dublin Midlands and Dublin North-East) have undertaken no major changes since 2006. Emergency departments were eliminated in four largely rural counties: Clare (2009), Tipperary North (2009), Roscommon (2011) and Monaghan (2009). Four other rural counties did not have an emergency department throughout the study period: Leitrim, Wicklow, Carlow and Longford.

We considered 16 serious emergency conditions, derived from consensus work carried out in the UK, for which the risk of death could be reduced by a well performing emergency care system [23]. The conditions were grouped into three categories; stroke, acute myocardial infarction/cardiac arrest, and ‘other’ (see Table 1 and Additional file 2).

The incidences of deaths from the selected emergency conditions are available from the Irish Central Statistics Office (see Additional file 3). The incidence of hospital admissions for these conditions is available from the Irish Hospital In-Patient Enquiry (HIPE) admissions system.

The regions and constituent counties analysed are outlined in Fig. 1. Due to concerns regarding the completeness of HIPE data for Roscommon over the period 2011–12, this county has been omitted from all analyses for these years.

Historically, Dublin County has been divided into three sections with respect to emergency care delivery, Dublin North-East, Dublin-South and Dublin Midlands. However, mortality and admissions data is not available at a sub-county level. Therefore, CFRs for Dublin were analysed as a whole and reported independently. As a consequence we only present data for the Midlands part of the Dublin-Midlands region (i.e. counties Kildare, Laois, Offaly, Westmeath and Longford) and for the Wicklow part of the Dublin-South region.

Results produced at a county level also allow for comparisons with routinely collected measures from other administrative data sources.

Case fatality ratios were constructed for each year from 2002 to 2014 inclusive. For 2002–2004 the national annual CFR was 187 per 1000, falling to 151 per 1000 over the period 2012–2014. The national annual percentage change in national total CFRs over the period 2002–2014 was a decrease of 2.12%.

Between 2002 and 2014 the national total deaths from the selected conditions fell from 7,978 to 5,205, decreasing across all groups (see Additional file 3). Total cases also decreased from 41,645 to 35,736, again decreasing in each group (see Additional file 4).

Joinpoint analysis found a statistically significant change in the CFR trend for total conditions in 2007 (Fig. 2). The APC for the period 2002–2007 was − 3.4 (95% CI: -4.4, − 2.4), with the APC decreasing to − 1.2 (95% CI: -1.9, − 0.5) from 2007 to 2014.

Analysis of the individual condition groups showed a significant change in trend for stroke; from 2002 to 2006 an APC of − 4.3 (95% CI: -6.2, − 2.4) was observed, which decreased to − 0.6 (95% CI: -1.4, 0.1) from 2006 to 2014. A significant change was also seen for the ‘other’ group; from 2002 to 2008 an APC of − 2.7 (95% CI: -4.6, − 0.7) was observed, reducing to 0.4 (95% CI: -1.7, 2.6) from 2008 to 2014.

The AMI and cardiac arrest group showed a consistent downward APC of − 3.2 over the full period (95% CI: -3.5, − 2.9) (Fig. 2).

Having identified 2007 as the relevant joinpoint in the national trend, analysis of the regional total CFR trend revealed a statistically significant downward trend from 2007 in three regions; Mid-West, West, Dublin (Table 2).

When compared to the national annual decrease (− 1.2%) for this period, two regions were found to have a statistically significant difference; the South-East had a lower rate of decrease, and the West had a higher rate of decrease (Fig. 3).

The county results identified six counties with significant decreases in their CFR trends (Donegal, Dublin, Kerry, Louth, Mayo, and Sligo) from 2007 to 2014 (Table 2). Carlow and Mayo were found to be outside of the 95% limit when compared to the national trend (Fig. 3).

A comparison of county total CFR results in 2002–2004 and 2012–2014 to the respective national CFRs can be seen in Fig. 4, represented by the horizontal and vertical red lines, and Table 1.

The Spearman rank correlation of the CFRs found a statistically significant relationship between county results over the 2 periods (ρ = 0.40, P = 0.04). Three counties (Monaghan, Roscommon and Wexford) had a decline in their position with regards to the national result between periods, and two counties (Galway and Sligo) improved (Fig. 4). The two counties identified as significantly different from the national trend, Carlow and Mayo, were above the national result in both periods.

The coefficient of variation found that variability between county CFRs decreased in the two periods; 11.45 for 2002–2004 and 9.72 for 2012–2014.

There was a large decline nationally in case fatality from serious emergency conditions over the period 2002–2014. The continued decrease in case fatality, albeit at a slower rate from 2007, is a positive outcome at a national level. Pronounced changes for stroke and the ‘other’ group were seen, while the rate of decline for AMI and cardiac arrest did not slow. The number of events for acute myocardial infarction has consistently decreased during the period of study (see Additional file 4). This is despite changes in how the condition is defined, and continued improvements in its detection [13, 26].

Variability in outcomes persists at a regional level. Two of the eight regions showed a significant difference in total condition fatality when compared to the 2007–2014 national rate; the South-East improving slower, while the West improved faster. Variation at county level also demonstrates that within region results are not homogenous.

Cross-sectional analysis revealed many counties in the South-East, North-East and Dublin performed consistently better than the national CFR. These regions underwent limited attempts at reconfiguration. The South, Mid-West and Midlands regions were consistently worse than the national average CFR over the study period. Of these, the South and Mid-West regions underwent significant reconfiguration.

There is little evidence that identified changes in CFRs at a regional or county level were associated with the reconfiguration of services, such as the removal of emergency departments. For example, counties such as Monaghan and Roscommon, which experienced the closure of emergency departments, saw a decrease in their position relative to the national CFR over the two time periods (2002–2004 and 2012–2014) studied. However, the rate of CFR decrease in these counties was not statistically different from the national rate between 2007 and 2014, as per the constructed funnel plots.

The findings from this research reveal a complex picture. Undoubtedly outcomes have improved over the period in question; however the national rate of improvement slowed in the most recent years. An argument can be made that mortality may be the last thing affected by system change. Clinical professionalism may limit any potentially negative consequences of such changes. Aspects of quality, safety and morbidity, including a delay in care and unrelieved pain, may be more likely to experience adverse effects.

The concentration of emergency care to specialist centres is intended to improve outcomes [11‐14]. In the UK, for example, the reconfiguration of trauma care services led to a 60% improvement in the odds of surviving a major trauma over the period 2008–2014 [27]. The findings presented in this paper suggest that reconfiguration in Ireland, mainly implemented after 2006, has not resulted in improved outcomes, and has not altered long-term geographical differences between regions and counties. This may be due to poor resourcing and implementation of reconfiguration plans. It may also be due to long-term structural differences between geographical areas in social determinants of health such as rurality and deprivation. Any detailed cross sectional analysis of variations between counties would need to account for these characteristics. A study of mortality in England and Wales found that deprivation accounted for the majority of differences seen between urban and rural areas, with the exception of lung cancer, respiratory disease and accidents [19].

The period of study also saw improved clinical guidelines and documents of best practice, the establishment of offices of clinical audit, as well as the introduction of clinical care programmes for conditions such as stroke and AMI.

The National Stroke Programme launched in 2010 is considered to have substantially changed the level of specialised stroke care received by patients [28]. A 2015 national audit of stroke highlighted in-hospital improvements for stroke mortality, decreasing from 19 to 14% since 2008 [29].

One of the principal aims of this programme was the development of stroke units in all hospitals which accept stroke patients [28]. However, issues exist regarding the full implementation and staffing of these units. According to the audit, only 29% of patients were admitted directly to a stroke unit and almost 50% did not receive treatment in a unit during their stay in hospital [29]. Also, nearly a quarter of the hospitals providing acute stroke care did not meet the minimum standards of a stroke unit [29].

The goal of full national 24/7 thrombolysis has still not been achieved. It is currently supported through bypass protocols to larger tertiary hospitals when required, and the development of the Telemedicine Rapid Access for Stroke and Neurological Assessment (TRASNA). TRASNA allows doctors to provide consultations via video and supervise thrombolysis where necessary. Where implemented the rate of thrombolysed patients is 1 in every 3.5 patients, compared to 1 in 5 elsewhere [28]. However, delays have been experienced in the full roll out of this programme [28].

In terms of cardiac care, the Acute Coronary Syndromes (ACS) Programme was launched in 2012 [30]. This programme has supported the adoption of five 24/7 primary percutaneous coronary intervention (PCI) centres and one 9–5 Monday to Friday centre nationally [31]. Improvements have also been made to pre-hospital services for patients as a result of changes to pre-hospital emergency care council and ambulance protocols. It has subsequently been reported that the number of reperfused ST-Elevation Myocardial Infarction (STEMI) patients that receive PCI increased from 55% in 2011 to 94% in 2015 [31].

The impact of these condition specific service changes and other clinical programmes, together with higher level system changes, can be seen in the results of our analysis. The slowing of improvement, particularly for stroke, may now be a result of gains being harder to achieve as programmes start to focus on more complex changes. At a regional level, initial emergency care system resources and quality of care were not uniform and the implementation of reconfiguration differed widely across regions. Changes took place in the context of an initial period of national investment and growth, followed by an economic recession. Budgetary cuts were a contributing factor to the structural changes which resulted in the closure of emergency services. Restrictions on staff recruitment across emergency departments and ambulance services continue to be experienced to date. For instance, a review of the National Ambulance Service (NAS) in 2015 found that almost 300 additional staff would be required to cover best achievable performance [32], while the 2016 National Service Plan highlighted the continued gap between pre and post-recession employment in the acute hospital sector [33].

The restructuring of emergency services has been previously studied internationally, particularly with regard to the closure of rural emergency departments. Conflicting results have been found. Some studies [15, 34] have found a risk of higher mortality when distance to treatment is increased. In the UK, a study found that a 10-km increase in straight-line distance to treatment was associated with a 1% absolute increase in mortality [15]. Conversely, a study in the United States concluded that higher in-hospital mortality did not necessarily occur after the closure of a local emergency department [35]. It argued that where other appropriate services exist, the closure or reduction of certain services will not have a negative impact on in-hospital mortality outcomes [35]. However, remaining facilities must be adequately resourced and staffed to meet new demands [36].

A strength of this study is the shift from in-hospital mortality as the main measure of outcome. Using hospital mortality rates to predict the quality of hospital care can result in good or average hospitals being penalised [37]. Its continued use in outcome reporting [4‐6] over-emphasises the concerns of providers, rather than the needs of the population. Case fatality constructed by area of residence allows analysis of outcomes for those who need to engage with the system, rather than focusing on outcomes from a specific service [23].

This study is subject to a number of limitations. Emergency admissions to private hospitals were not included in this analysis; private hospitals are not required to submit data to the hospital inpatient enquiry system (HIPE). Consequentially, case fatality ratio results reported may represent a maximum level; results for counties with a high level of private hospital usage may be lower than stated here. However, we estimate the impact of private hospital admission on our results is low due to a number of factors. First, there were only five small private emergency departments open in Ireland over the study period and many of those were not open for the full study period. Second, these hospitals generally worked on a 8 am-5 pm schedule, Monday to Friday [38, 39] and during our study period would not have operated a weekend service. Third, private emergency departments generally did not accept the most serious emergency conditions, such as major trauma and acute stroke, over the study period [38‐40]. Fourth, serious emergency cases requiring ambulances were not taken to private emergency departments over the study period [38, 39].

Our analyses rely on the accuracy of the HIPE system for recording emergency admissions. A study by the Department of Health in 2013 has confirmed the robustness of the data available from HIPE, specifically as a tool for the development of indicators of quality of care in hospitals [41]. As a result, this data has formed the basis of the National Healthcare Quality Reporting System annual reports [4‐6] and such use is in line with the analysis produced within this study. However, within our analysis particular caution should be used when interpreting results for County Roscommon. Due to the absence of a HIPE coder for a period spanning part of 2011–2012 in Roscommon County Hospital, the accuracy of coding is limited for much of the county’s patient population.

The primary aim of this study is the evaluation of major system change. Reconfiguration of such a scale is likely to lead to improved results for certain conditions, but the deterioration of results for others. Therefore, to assess the overall impact on the system, the focus is necessarily on aggregated higher level data. The analysis of patients, or each condition, at an individual level is of limited benefit.

Cautions should be taken when using county level data in understanding change in complex, multi-factor situations. However, it is important to note that any lower level analysis is restricted in Ireland due to lack of access to more detailed data. Access to admissions data through the hospital admissions system is limited to county level. Similarly, personal individual level mortality data is unavailable from the Central Statistics Office due to concerns of identifiability.

Ireland also differs from many other European countries in that it does not have a unique patient identifier. This restricts the ability to link individuals to admissions and subsequent death for a specific condition. Therefore, analysis was limited to the ratio of deaths to cases in a year, as opposed to the rate of deaths per cases. There are measures underway as of 2014 to introduce a National Register of Individual Health Identifiers [42].

There is currently no independent routine health planning on behalf of populations in Ireland. The majority of planning is done by, or on the behalf of, the provider, the health service executive (HSE). Such planning is primarily based on once off national reports, as previously outlined [7‐10], which focus on the performance of the provider. This study provides a counterpoint to such reports, and aims to refocus attention to how well populations are served.

Our findings show that changes to the national CFR trend coincided with a period of recession in Ireland. With additional budget allocations as of 2015 [33], further monitoring will determine if there are future improvements to CFRs. Additionally, policies of reconfiguration do not appear to have significantly influenced CFRs. Continued observation will determine if on-going implementation of these policies also result in greater improvements. It may also be argued that much of the variance in case-fatality can be explained by non-health system factors such as deprivation and rurality [19, 43, 44], which have not been included in our model and merit further investigation.