Association of extracerebral organ failure with 1-year survival and healthcare-associated costs after cardiac arrest: an observational database study

This is a retrospective study conducted at the five university hospitals of Finland. These five centres serve a population of 3.2 million covering 60% of the entire population of Finland. We included adult cardiac arrest patients treated in the participating ICUs between January 1, 2003, and December 31, 2013. The year 2003 was chosen as the starting point because it was the first year when therapeutic hypothermia became widely used in the treatment of cardiac arrest patients in the participating centres [15]. Only the first event of cardiac arrest was considered. Cardiac arrest was recognized as admission diagnosis of cardiac arrest according to Acute Physiology and Chronic Health Evaluation III (APACHE III) or positive value for Therapeutic Intervention Scoring System 76 (TISS-76) item of “cardiac arrest and/or countershock within past 48 h”. Complete data for admission date and day of death/vital status at the end of 1-year follow-up was required for inclusion. Patients with missing SOFA score were excluded.

The Finnish social security and healthcare system is government-based. Each Finnish resident is assigned a unique personal identification code which is used to identify the person in medical records, social security records, and national registers. We used this identification code to combine data from different sources.

Medical data were extracted from the Finnish Intensive Care Consortium (FICC) database [16]. These included the SOFA score and the premorbid physical status as a simplified World Health Organization-Eastern Cooperative Oncology Group (WHO/ECOG) classification [17], where “independent” was defined as the patient being independent in self-care and “dependent” was defined as the patient being partly or fully dependent on help in self-care prior to hospital admission. We linked these data with the confirmed date of death obtained from the Finnish Population Register Centre database, which records all deaths of Finnish residents.

Furthermore, we linked data on gross 1-year healthcare-associated costs from the hospital billing records, rehabilitation costs and social security costs obtained from the Social Insurance Institution. The hospital costs include all costs until hospital discharge (diagnostics, ICU stay, ward stay, operative treatment etc.). The rehabilitation costs were calculated by multiplying the number of days spent in the rehabilitation unit with the average price per day for units of the corresponding level. The Finnish Social Insurance Institution reimburses disability and sickness allowances, private physician and physiotherapist expenses, prescription drug expenses and medical transportation expenses. All social insurance reimbursements during the year following hospital admission were included. Effective cost per survivor (ECPS) was calculated by dividing the sum of total costs for all patients with the number of survivors [18]. We adjusted all costs to the value of euro in the year 2013 according to the consumer price index (CPI) in Finland.

We performed an additional analysis of a nested cohort consisting of patients treated in one (Helsinki University Hospital) of the five centres of the full data. For this cohort, additional data were available according to the Utstein criteria [19, 20] including the delay from collapse to return of spontaneous circulation (ROSC), type of initial rhythm during cardiac arrest (shockable or not), whether the cardiac arrest was witnessed or not, and 1-year cerebral performance category (CPC) [21, 22]. We defined good neurologic outcome as 1-year CPC 1 or 2 and poor neurologic outcome as 1-year CPC 3 to 5. Furthermore, the location where cardiac arrest had occurred was recorded, allowing stratification of the nested cohort data in out-of-hospital cardiac arrest (OHCA) and in-hospital cardiac arrest (IHCA). Cardiac arrests in the ICU (ICUCA) were included in the IHCA group in this study.

We performed the statistical analyses with the SPSS software (version 24.0, IBM, Armonk, NY, USA). We present continuous data as medians with interquartile ranges (IQR) and categorical data as percentages of the whole. We tested group differences with a chi-square test, Student’s t test, Mann-Whitney U test or Kruskal-Wallis test, as appropriate. We performed Kaplan-Meier analysis to evaluate the difference in 1-year survival across quartiles of the 24h-EC-SOFA score. We used logistic regression models to search for variables that were independently associated with outcome. For testing the association of the 24h-EC-SOFA score with healthcare-associated costs, we used linear regression models. To avoid the confounding effect of early deaths causing low costs, we calculated a “cost per day alive” variable by dividing the total costs of a given patient by the number of days the patient remained alive during the first year after the cardiac arrest. If the patient was still alive 1 year after the arrest, then the cost per day alive variable was calculated by dividing the total costs during the year after cardiac arrest by 365.

The manuscript was completed according to STROBE standards [23].

In Kaplan-Meier analysis, 1-year survival was significantly different between quartiles of 24h-EC-SOFA (logrank P < 0.001; Fig. 2). One-year survival had a rising trend from 38% in 2003 to 47% in 2013 (Additional file 3: Figure S2). In a logistic regression model adjusting for age and the premorbid physical status, the 24h-EC-SOFA score was associated with 1-year mortality [odds ratio (OR) 1.16 and 95% confidence interval (CI) 1.14–1.18; Table 2]. Similarly, the full 24h-SOFA score was associated with 1-year mortality in a logistic regression model adjusting for age and the premorbid physical status (OR 1.22, 95% CI 1.20–1.24; Additional file 4: Table S2). All sub-scores, except the cardiovascular sub-score, were independently associated with 1-year mortality when the individual EC-sub-scores were tested (respiration, OR 1.18, 95% CI 1.12–1.25; coagulation, OR 1.09, 95% CI 1.00–1.18; liver, OR 1.23, 95% CI 1.10–1.37; renal, OR 1.52, 95% CI 1.44–1.61; Additional file 5: Table S3). Univariate analysis of the variables used in the logistic regression models is presented in Additional file 6: Table S4.

The sum of costs recorded in all 5814 patients during the year following cardiac arrest was €230,000,000. The 1-year ECPS was €96,000. Median costs in the first year after cardiac arrest were €47,000 in 1-year survivors and €12,000 in non-survivors (Table 1). During the study years, ECPS remained between €85,000 and €110,000 (Additional file 7: Figure S3). Median costs recorded for 1-year survivors were lowest in 2003 (€38,000) and highest in 2004 (€52,000), plateauing thereafter (Additional file 8: Figure S4). Median costs recorded for 1-year non-survivors had a rising trend from €9900 in 2003 to €16,000 in 2013(Additional file 8: Figure S4). When the data were divided in quartiles of the 24h-EC-SOFA score, the total costs of 1-year survivors differed significantly between groups (P <  0.001); higher 24h-EC-SOFA score was associated with higher costs (Fig. 3). The same was true when the quartiles of the 24h-EC-SOFA score were tested for the association with cost per day alive during the first year after cardiac arrest (P < 0.001). The differences were similar for quartiles of the 24 h-SOFA score (P < 0.001 for the difference in total costs in 1-year survivors and P < 0.001 for the difference in cost per day alive; Fig. 3).

In a linear regression model adjusting for age and premorbid physical status, an increase of one point in the 24h-EC-SOFA score was associated with an increase of €170 (95% CI €150–190) in the cost per day alive in the first year after cardiac arrest (Table 3). When the EC-sub-scores were tested, renal, cardiovascular, respiratory and coagulation sub-scores were significantly associated with an increase in the cost per day alive (increase per one point rise: respiration, €120, 95% CI €61 - 170; coagulation, €300, 95% CI €230 - 370; cardiovascular, €90, 95% CI €41 - 140; renal, €400, 95% CI €350 – 440; Additional file 9: Table S5). In 1-year survivors, an increase of one point in the 24h-EC-SOFA score was associated with an increase of €4400 (95% CI €3300–5500) in total healthcare-associated costs recorded during the year following the cardiac arrest (Table 4). When the EC-sub-scores were tested, renal, respiratory and coagulation sub-scores were significantly associated with an increase of total healthcare-associated costs recorded during the year following the cardiac arrest (increase per one point rise: respiration, €7300, 95% CI €4800 - 9800; coagulation, €4600, 95% CI €1000 - 8300; renal, €7900, 95% CI €5000 - 11,000; Additional file 10: Table S6).

For the nested cohort, the logistic regression model was complemented with variables describing conditions during resuscitation: whether the initial cardiac rhythm during cardiac arrest was shockable, the delay from collapse to ROSC and whether the cardiac arrest was witnessed. In this model, the 24h-EC-SOFA score was associated with 1-year mortality (OR 1.19, 95% CI 1.12–1.25 per point of 24h-EC-SOFA) and poor neurologic outcome (OR 1.15, 95% CI 1.09–1.21 per point of 24h-EC-SOFA; Table 2). When the individual EC-sub-scores were tested, respiratory and renal sub-scores were independently associated with both 1-year mortality (respiration, OR 1.16, 95% CI 1.02–1.32; renal, OR 1.50, 95% CI 1.32–1.70) and poor neurologic outcome (respiration, OR 1.17, 95% CI 1.02–1.33; renal, OR 1.42, 95% CI 1.25–1.63; Additional file 11: Table S7). The association of the 24h-EC-SOFA score with 1-year mortality and poor neurologic outcome remained significant in subgroup analysis of both OHCA (1-year mortality OR 1.20, 95% CI 1.10–1.31 per point of 24h-EC-SOFA; poor neurologic outcome OR 1.15, 95% CI 1.05–1.25 per point of 24h-EC-SOFA) and IHCA patients (1-year mortality OR 1.15, 95% CI 1.06–1.24 per point of 24h-EC-SOFA; poor neurologic outcome OR 1.11, 95% CI 1.03–1.21 per point of 24h-EC-SOFA; Additional file 12: Table S8).

In the nested cohort, the total costs documented for the 1324 patients was €65,000,000 and 1-year ECPS was €94,000. The cohort included 772 OHCA and 552 IHCA patients. Median (IQR) survival time for 1-year non-survivors was 5 (3–10) days in OHCA and 8 (3–20) days in IHCA. In the nested cohort, the linear regression model adjusting for age and premorbid physical status indicated an association of an increase of one point in the 24h-EC-SOFA score with a €280 (95% CI €230–330) increase in the cost per day alive. For 1-year survivors, the increase in the total healthcare-associated costs in the year after cardiac arrest was €7700 (95% CI €5700–9700) per one-point rise in the 24h-EC-SOFA score. When the model was complemented with Utstein variables, the increase in the cost per day alive was €210 (95% CI €170–260) per 24h-EC-SOFA point. In 1-year survivors, total healthcare-associated costs increased €7200 (95% CI €5000–9400) per each additional 24h-EC-SOFA point (Tables 3 and 4). When the data were stratified according to cardiac arrest location, the increase in the cost per day alive in OHCA was €210 (95% CI €150–280) and in IHCA €200 (95% CI €120–270) per 24h-EC-SOFA point. In OHCA patients, the total healthcare-associated costs in 1-year survivors increased €2500 (95% CI €200–4800) and in IHCA patients €11,000 (95% CI €5600–16,000) per each additional 24h-EC-SOFA point (Additional files 13 and 14: Tables S9 and S10).

Previous studies have shown an association of admission SOFA and EC-SOFA with 28-/30-day mortality in single-centre settings of OHCA patients [13, 24, 25]. Association with neurologic outcome has produced conflicting results [24, 25]. In a single-centre material including both OHCA and IHCA patients, the highest EC-SOFA score during the first 72 h of ICU stay was associated with in-hospital mortality and cardiovascular and respiratory sub-scores were independently predictive of in-hospital mortality [12]. In a multi-centre database study including both OHCA and IHCA patients, ICU non-survivors had higher cardiovascular, respiratory and renal SOFA sub-scores upon ICU admission but only renal sub-score was independently associated with ICU-mortality [14].

To the best of our knowledge, our current study is the largest study on the association of EC-SOFA with cardiac arrest patient outcome published to date. In our multi-centre setting with a mixed cohort of OHCA and ICHA patients, the association of EC-SOFA with 1-year outcome was clear. In contrast to other sub-scores, the cardiovascular sub-score was not independently associated with the outcome in our cohort, since almost three quarters of patients had 24-h cardiovascular sub-score of 3 or 4, regardless of the outcome group. In line with previous studies, respiratory and renal sub-scores had a clear association with the outcome in our study population.

It is important to notice that the large overlap between the outcome groups does not allow the use of 24h-EC-SOFA for predicting outcome on a single patient level. However, our results emphasize the importance of the evolving multiple organ failure for long-term morbidity, mortality and healthcare resource use in cardiac arrest patients on population level. This association could justify the use of SOFA or EC-SOFA score as an end-point in small randomized controlled trials on cardiac arrest in the future instead of mortality, for which the studies often prove underpowered.

Recently, there has been a growing interest in the obvious socioeconomic burden of cardiac arrest [26]. One of the first studies assessing the healthcare resource use in cardiac arrest reported combined pre- and in-hospital costs for OHCA 6-month survivors to be €36,000 (expressed as 2013 euros) [27]. More recent studies have reported in-hospital costs of €50,000–60,000 in 2013 euros for patients surviving to hospital discharge after OHCA or IHCA [28‐30]. One previous study reported OHCA survivors to have combined pre-, in- and post-hospital healthcare-associated costs of €37,000 in 2013 euros in a 6-month follow-up [31]. In a recent study of single-centre data published by our group, combined in- and post-hospital healthcare costs in 1-year follow-up were €59,000 for OHCA and €84,000 for IHCA hospital survivors in 2013 euros [18].

In our current study, the association of 24h-EC-SOFA with healthcare-associated costs was clear. Most of the sub-scores (renal and respiratory sub-scores in particular) had also independent association with costs. This suggests that early organ failure after cardiac arrest leads to both short- and long-term morbidity reflected as increased in- and post-hospital expenses.

The analysis of our nested cohort clearly shows that the association of organ failure with outcome and costs is not redundant to more detailed information considering the conditions during and after cardiac arrest and resuscitation. The similarity of the ECPS values of the nested cohort and the full data (€94,000 and €96,000, respectively) suggest that the case-mix in the nested cohort was not markedly different from that of the full data.

This study has a number of strengths. It is a large multi-centre study conducted in a country of government-funded social security and healthcare systems. The follow-up time of 1 year is long enough to reliably dissect long-term survivors from non-survivors. We have exploited the unique possibility provided by the nature of the government-funded system to combine data from different registries to acquire information on long-term survival and gross healthcare-associated costs from different healthcare providers. Naturally, there are also limitations. Our data are limited to one country—the generalizability to other healthcare systems is not certain. The study is retrospective, and the data were gathered during a period of 11 years. The treatment of cardiac arrest patients has evolved during the data gathering period and the 4-year delay from the end of follow-up to the publication of the study. Unfortunately, distinction between OHCA and IHCA, which have largely different aetiology, was not possible in the main cohort. Furthermore, the FICC database provided only a point measure of SOFA score as the worst values recorded during the first 24 h, not allowing for evaluation of the temporal change in SOFA score.