The study protocol was approved by the University Hospital of Aachen (Aachen, Germany) research ethics committee. Formal consent prior to contact for patient follow-up was not required since all patients who were contacted had the chance to refuse completion of the questionnaire. All patients admitted to the ICU of Medical Clinic I from 1 January 1999 to 31 December 2001, who received CPR for out-of-hospital or in-hospital cardiac arrest for any cause, were eligible for study entry. For the purpose of this study, CPR was defined as at least one cycle of chest compression and ventilation in patients with signs and/or symptoms of cardiac arrest. Patient care was at the discretion of the intensivist in charge without any explicit standard of care beyond the normal institutional standards and guidelines. Neither care of the patients nor end-of-life decision making was influenced by the study protocol at any time. Demographic data, admission diagnoses, lengths of ICU and hospital stays, and ICU and hospital mortality rates were collected prospectively. Severity of illness was classified using the Simplified Acute Physiology Score II (SAPS II) 9 for the initial 24 hours after admission to the ICU. The simplified Therapeutic Intervention Scoring System (TISS-28) 10 and the Sequential Organ Failure Assessment (SOFA) 11 were collected daily, and the total maximum SOFA (TMS) was calculated at the end of the ICU stay 12.

Health-related quality of life (HRQL) was obtained 5 years after ICU discharge using a regular mail formal letter, including a return envelope, containing the validated German interview form 13 of the Medical Outcome Survey Short Form-36 questionnaire (SF-36) self-report form 14. In addition, the questionnaire assessed employment and marital status, dependency, re-hospitalizations, patients' recollection of their ICU stay, and their willingness to undergo critical care, if necessary, again. If patients did not respond to the questionnaire and telephone contact could not be established, the family doctor and/or relatives of the patients were contacted to provide the correct address of the patient or to confirm death after hospital discharge. Patients who could not be contacted, but were known to be alive, were considered lost to follow-up with regard to HRQL. Normative HRQL data, including apparently healthy controls 15 and patients with acute and chronic diseases 16, are available for different age groups of the German population. An HSI, which represents overall quality of life relative to an age-matched reference group, was calculated for each patient using the SF-36 results. The HSI is a weight ranging from 0 (indifference between life and death) to 1 (perfect health) and was calculated for each patient as the mean of the individual domain indices for the eight domains of the SF-36 (that is, by dividing the individual patient result for a particular domain by the domain mean of the normative data obtained from apparently healthy Germans 15, summing each domain index, and dividing by the number of domains). The HSI multiplied by life years gained results in QALYs 17. Patients discharged from the hospital with a Glasgow Coma Scale (GCS) score of below 6 points (that is, severely neurologically disabled patients) were analyzed separately since formal objective quality-of-life assessment was not possible.

The costing methodology in this study is a modification of the 'bottom-up' approach 1819. Costs were not limited to index hospital and ICU admission costs. From the perspective of the society's health care system, we included post-hospital discharge health care costs and used a 3% annual discount rate 20. Other costs to society, such as time lost from work, were not considered. To consider differences in the complexity of the individual patient, ICU costs were divided into patient-specific (variable) and non-patient-specific (fixed) costs. Non-patient-specific costs were calculated on a patient-day basis, whereas patient-specific costs were directly attributed to the individual patient. Labor costs were divided into patient-specific and non-patient-specific costs. Wages of nurses and physicians on duty were allocated to patient-specific costs and distributed according to the patient's TISS-28 score to account for patient differences in therapeutic activities. 'Backup staff' (that is, vacancies and off-shift) costs were calculated as day-related non-patient-specific costs. The costs for radiology, clinical chemistry, pathology, and microbiology were calculated according to the German regulation of charges for physicians 21. Non-patient-related administrative costs were calculated as a share of hospital costs in relation to the size of the unit (energy, heating, and maintenance) and the number of patients and staff (administrative costs).

This study presents both the outcomes and costs associated with this patient cohort. Cost-outcome descriptions are presented as costs per survivor, costs per long-term (5 years) survivor, costs per life year gained, and costs per QALY gained. To obtain costs per survivor and costs per long-term survivor, total ICU, hospital, and post-hospital discharge health care costs were divided by the number of patients who survived hospital discharge and the number of patients remaining alive at 5 years post-hospital discharge. Age- and gender-specific expected annual post-hospital discharge health care costs were obtained from the German Ministry of Health (that is, for males: 30 to 44 years, 1,270 € per year; 45 to 64 years, 2,760 € per year; 65 to 85 years, 5,830 € per year; for females: 30 to 44 years, 1,840 € per year; 45 to 64 years, 3,160 € per year; 65 to 85 years, 6,250 € per year as of the year 2004 22). Post-hospital nursing home costs were estimated for all patients with a GCS score below 6 points as a monthly average of 2,700 € (as of the year 2004 22). A yearly discount of 3% was subtracted or added for the years before or following 2004, respectively. To estimate the cost per life year gained, the total costs were divided by the total estimated life years gained. Life years gained was calculated as the total life years of follow-up time observed in all patients post-discharge plus the estimated remaining life span of the patients alive at 5 years. The estimated remaining life span was calculated conservatively based on an average age-adjusted life expectancy of 80.5 years for the male population and 84.3 years for the female population (for a 65-year-old person as of the year 2002, German Ministry of Health 22). Cost-utility descriptions were generated via an HSI adjustment of life years gained. The number of QALYs gained is the product of the number of life years gained multiplied by HSI. Note that the HSI could be obtained only for patients surviving at 5-year follow-up; thus, only their life years gained were adjusted. This is a conservative approach since patients who were discharged alive, but did not survive to year 5, may have gained quality survival time.

All variables were tested for the assumption of normality using the Kolmogorov-Smirnov test. Descriptive statistics are reported as mean and 95% confidence interval (CI), except when stated otherwise. The Student t test was used for comparisons of means of normally distributed data. A non-parametric rank test (Mann-Whitney U test) was applied in case of non-normally distributed data. Categorical data were tested using the χ² statistics with Yates correction when appropriate. Internal consistency of the various domains of the SF-36 was assessed using the Cronbach's alpha coefficient. A Cronbach's alpha exceeding 0.7 is considered to demonstrate acceptable agreement 23. In a two-way sensitivity analysis, both HSI and incremental life years saved for all 5-year survivors who completed the questionnaire were increased and decreased by 25% and 50%, respectively. Costs were adjusted considering the changes in post-hospital health care expenses owing to a longer or shorter remaining life span. All statistical tests were two-sided, and a significance level (P value) of less than 0.05 was applied, except when stated otherwise. Data were analyzed using SPSS 12.0 (SPSS Inc., Chicago, IL, USA).

Of 354 patients admitted to the ICU with cardiac arrest, 204 patients (58%) died prior to discharge from the hospital, either during their ICU stay (n = 171) or later on the ward (n = 26), not including 7 patients who were discharged with severe disabilities (that is, a GCS score below 6 points). Of the 150 patients (42%) remaining, 40 patients died before year 5, leaving 110 patients (31%) eligible to be surveyed at 5 years. Twenty patients declined to respond to the HRQL survey at 5-year follow-up, and 9 patients were known to be alive but were otherwise lost to follow-up. Eighty-one patients (74% of all 5-year survivors) completed the entire questionnaire. Complete demographic information is presented in Table 1. The 29 patients who were unavailable for final follow-up stayed significantly longer in the ICU and hospital compared with the cohort completing the questionnaire. Demographic data, severity of illness on admission (SAPS II), or morbidity (TMS and TISS-28) did not differ between the two groups. The final cost-utility description is based on the 81 complete data sets only (Figure 1). Prior to cardiac arrest, 60 patients (74%) lived self-supported, a status that was maintained by 56 patients (68%) 5 years later. The number of patients living in their own home with some level of support increased from 8 patients (10%) to 13 patients (16%). After hospital discharge, only a minority of patients relied on daily custodial support (3 patients, 4%) or lived in nursing homes (6 patients, 5%). Five years after hospital discharge, 13 (16%) survivors were employed, 13 (16%) were early retired, and 48 (59%) were regularly retired due to age. Forty-eight (59%) survivors were re-hospitalized during the 5-year follow-up (23 survivors once, 5 twice, and 5 three times). Twenty-two (27%) survivors recalled unpleasant or alarming memories with regard to their ICU stay. Sixty-nine (85%) reported that they would undergo intensive care again if necessary. HRQL is displayed in Figure 2. Except pain, emotional role function, and mental health, all other items were rated somewhat lower than in an age- and gender-matched population of apparently healthy Germans 15. The 81 long-term survivors reached a mean HSI of 0.77 (95% CI 0.70 to 0.85). Women rated their HRQL significantly better than men did (HSI 0.87 versus 0.74; P < 0.05). There were no differences in age, severity of illness, ICU and hospital lengths of stay, or admission diagnosis between men and women. The individual items of pain, emotional role function, and physical role function were rated superior by women (P < 0.05) after 5 years.

The total ICU costs for all 354 patients with cardiac arrest amounted to 6,312,700 € (Table 2). The costs for the ward stay after ICU discharge accounted for 295 € per patient per day. The total hospital length of stay was 7,544 days. The total hospital costs, including the ICU stay, amounted to 7,492,771 €. The ICU stay accounted for 84% of total in-hospital costs.

Post-hospital costs of future health care utilization for all 150 patients discharged alive were estimated to be 16,856,851 €, based on the projected remaining life span of a total 2,409 person-years. For the 110 patients known to be alive at 5 years, costs of future health care utilization after hospital discharge would amount to 15,615,920 €, based on the projected remaining life span of 2,354 years. The estimated long-term survival of the seven patients with a GCS score below 6 points, including the incurred ICU, hospital, and post-hospital costs, is displayed in Table 3. The post-hospital costs of future health care utilization were estimated to be 1,179,329 €, based on the projected remaining life span of a total 141 person-years. Including average nursing home costs of 2,700 € per month, total post-hospital costs would amount to 5,747,729 €. For the 110 patients known to be alive at 5 years, costs of future health care utilization after hospital discharge would amount to 15,615,920 €, based on the projected remaining life span of 2,354 years.

The costs per hospital survivor were calculated to be 49,952 € (that is, 7,492,771 € total hospital costs divided by 150 hospital discharge survivors). For the 110 patients surviving 5 years, initial ICU and hospital costs per long-term survivor were 68,116 € (that is, 7,492,771 € total hospital costs divided by 110 5-year survivors).

The 150 hospital discharge survivors were calculated, using life table methods, to have an estimated total remaining life span of 16 years per patient (95% CI 14 to 18 years) at the time of hospital discharge, which provides an additional 2,409 person-years. The estimated age-adjusted post-hospital discharge health care costs for these patients were calculated to be 6,997 € per person-year. Considering all hospital discharge survivors, the costs per life year gained were 10,107 € (that is, 7,492,771 € total hospital costs plus 16,856,851 € post-hospital discharge health care costs divided by 2,409 person-years gained). Including the costs incurred for the seven severely disabled patients (that is, applying life table methods to 157 hospital survivors), these costs increased to 11.757 € per life year gained. Considering only long-term survivors, patients alive at 5 years were calculated to have an estimated total remaining life span of 15 years per patient (95% CI 13 to 18 years) which, including the 5 years already survived per patient, accounts for a total additional 2,354 person-years. The age-adjusted post-hospital health care costs for these patients were calculated to be 6,634 € per remaining life year for a total of 15,615,920 €. Ignoring life years gained from patients who died before 5 years post-discharge, the costs per life year gained for long-term survivors were calculated as 9,817 € (that is, 7,492,771 € total hospital costs plus 15,615,920 € post-hospital discharge health care costs divided by 2,354 person-years gained). Again, including the seven patients with a GCS score below 6 points, costs per life year gained account for 11,566 €. It is important to present these long-term survivors separately since they represent the total sampling frame from which HRQL information could be obtained at 5 years.

Information on HRQL, and thus HSI 5 years after hospital discharge, was available in 81 patients (74% of all 5-year survivors). These 81 patients had an estimated average remaining life span of 21 years per patient (95% CI 18 to 24 years) for a total of 1,709 person-years and a calculated average HSI of 0.77 (95% CI 0.70 to 0.85). In this group, total post-hospital health care costs were 11,572,491 €. Incremental costs per life year gained thus amounted to 11,156 €. The estimated remaining life span of 1,709 years multiplied by the HSI of 0.77 translates into 1,316 QALYs, averaging 14,487 € per QALY gained. A simulation, including the 20 patients who declined responses to the HRQL survey and the 9 patients who were lost during follow-up, revealed an estimated total remaining life span of up to 2,327 life years (average 21 years per patient, 95% CI 19 to 24 years). Assuming an HSI of 0.75, which is comparable to those patients who completed the HRQL questionnaire, an additional 1,766 QALYs would have been gained. Thus, incremental costs per life year gained for all 110 5-year survivors would have amounted to 9,931 €, with 13,085 € per QALY gained. Figure 3 illustrates the influence of changes in utility (life years gained and HSI) and costs for the 81 patients with completed 5-year HRQL follow-up.