In this large nationwide population-based study, older patients admitted to the ICU and who survived the first year after admission had a slight increase in mortality rate in the following years compared with the general population. This difference was, however, strongly attenuated by adjustment for baseline comorbidity, leaving only a weak association that was least notable in the highest age groups. The reason for ICU admission had a variable impact on the long-term mortality rate, with the most apparent increase seen in patients admitted for acute-on-chronic respiratory insufficiency.
In ICU patients 75 years and older, survival within the first year after intensive care admission was slightly above 50%. Still, after the one-year landmark, survival was mainly comparable to the general population's age- and sex-matched controls. The difference compared to the general population was almost negligible in the oldest age group. Behind this age-related pattern, there are likely complex interactions between the type and severity of medical conditions leading to ICU admission, prevalent comorbidity, and other patient-related factors such as general frailty, but also external factors such as bed-space availability and regional clinical practice patterns influencing the decision to admit the patient to the ICU.
Patient frailty is a complex multisystemic syndrome that makes it difficult for a patient to recover from an acute insult. Several approaches have been applied in the ICU setting to measure the frailty syndrome, but they have had uncertain reliability. Patients with frailty admitted to the ICU have a more significant number of comorbidities and are generally older [18]. Our study's observed survival probabilities related to age and comorbidity appear reasonable and do not raise any substantial overall concerns regarding the selection of older patients with comorbidity for ICU admission.
The study population included patients admitted to general ICUs and cardiothoracic ICUs. This must be considered when interpreting the results. When the SAPS3 prognostic score was developed, many cardiac surgery patients were included in the development dataset [19]. SAPS3 has nevertheless been shown to perform less well in this patient category, and other risk adjustment instruments have been used instead [20]. The reason for this could be that the cardio-thoracic ICUs predominantly admit patients for postoperative care and that patient selection for cardiac surgery strongly impacts long-term survival probability [21]. Therefore, analyses restricted to admissions to general ICUs may be of particular relevance for the overall interpretation of our results.
Knowledge of how the individual patient’s age and history of comorbidity interact with the acute condition underlying the ICU admission and the long-term survival probability compared to the general population after recovery is essential background information for selecting appropriate treatment strategies. This understanding is also relevant for the communication with patient and family, and it would be of value if some reassurance could be provided that if the patient survives the ICU stay, life expectancy will return to what would be expected based on age and preexisting comorbidities. It should, however, be acknowledged that the results in this study are conditional on the selection of patients for ICU admission. We have no knowledge related to patients that were potentially eligible for intensive care but for some reason were not admitted to an ICU. It may also be relevant to consider the independent impact of specific comorbidities.
The strength and pattern of the association between the recency of a prior admission for a specific comorbid condition and mortality rate were variable, depending on the type of comorbidity. The more recent a previous hospitalization for chronic pulmonary disease was, the stronger the association with mortality rate was, both before and after the one-year landmark. Heart failure, another chronically progressive comorbidity, was also clearly associated with mortality rate after adjustment for other comorbidity, but without any apparent relation to recency for the prior hospital admission for heart failure.
The recency of a prior admission for ischemic heart disease is considered an established risk factor for mortality after non-cardiac surgery [22]. A similar pattern would be expected after ICU admission. No such association could, however, be observed in our study. After adjustment for other comorbidities, a prior hospitalization for ischemic heart disease was not associated with mortality rate, neither before nor after the one-year landmark. New treatment strategies for acute coronary syndrome have substantially improved the outcome of ischemic heart disease over the years, and this may partly explain this finding [23]. It should be recognized that the association between recency of prior hospital admission for a comorbidity and mortality rate following an ICU stay may also be variable depending on differences in healthcare utilization between countries and regions over time [24].
The observation that the strength of association with survival probability is variable depending on the specific comorbidity condition studied is expected. The mortality rate associated with an exacerbation of chronic obstructive pulmonary disease (COPD) is high, especially if requiring ICU admission [22]. Some associations observed are not as expected. A history of hypertension or a previous hospitalization for infection may not be expected to impact the survival probability associated with an ICU admission. Still, a recent prior hospital admission for any of these conditions appeared to confer an increased mortality rate also after adjustment for other comorbidities.
Some strengths of our study are the large population-based study population with almost complete coverage of all Swedish ICUs, comprehensive data collection, long-term follow-up, and individuals selected from the general population. The quantification of comorbidity is based on a model shown to outperform conventional comorbidity measures [11]. These measures may still be suboptimal and cause residual confounding. An example could be chronic obstructive pulmonary disease, where physiological measures of lung function or other functional measures could provide further risk stratification. Differences in coding practices and healthcare system characteristics between countries are also expected to limit generalizability and hamper international comparisons [25, 26]. Other factors related to patient frailty and survival probability, such as socioeconomic status and lifestyle, were not included in our study and could result in residual confounding. They may be essential determinants for long-term outcomes, but they were outside the scope of the present study [27–29].
The fact that follow-up is only available until December 2016 can be considered a limitation. It should, however, be acknowledged that inclusion of data from 2019 inevitably would have compromised the interpretation of results because of the COVID-19 pandemic. When we using SIR compare ICU casemix and mortality in 2008 with 2023, they are very similar [30]. The mean age on ICU admission was 54.5 years in 2008 and 57.5 years in 2023. The median max, Sequential Organ Failure Assessment (SOFA) score, was 9 both in 2011 (first year available) and in 2023. The median length of stay was 0.97 days in 2008 and 1.09 days in 2023. Mortality after 180 days was 22.0% in 2008 and 21.9% in 2023. During the pandemic years with COVID-19, however, the pattern was different due to the large number of people treated in intensive care for COVID-19 and the limited resources available. To include these years in the current study would notably complicate the interpretation and limit generalizability of the results. It should also be recognized that the risk adjustment instrument SAPS3 was developed on data collected from consecutively admitted patients between 14 October and 15 December 2002 [19].
One key limitation of the study is that admission to the ICU is preceded by an individual patient assessment, and this may involve selection mechanisms that are time dependent and context sensitive. Since we do not have any information on which patients that were potentially eligible but not actually admitted to the ICU, the consequences of such selection mechanisms remain elusive. The results should be considered with this limitation in mind.
The analyses stratified for admission diagnosis are hampered by the change from APACHE II to SAPS3 during the study period and the large proportion of patients without a registered score. Because the scoring is predominantly missing from cardio-thoracic ICU admissions and patients with higher survival probability and considering the difficulty in harmonizing APACHE II and SAPS3 admission diagnoses, the analyses restricted to when SAPS3 is available are considered to be the most relevant.
Another potential limitation is the selection of a one-year landmark in the survival analyses. It is unclear how long after an ICU stay the course of the acute illness can affect mortality. For randomized ICU intervention trials, long-term mortality is also considered an important outcome measure. The European Medicines Agency guideline on clinical investigation of medical products for treating sepsis recommends that longer-term mortality data should support the primary outcome measure with a minimum of 6 months follow-up [31]. Similar recommendations are given for treatments targeting patients with acute respiratory distress syndrome [32]. To isolate long-term effects from the acute phase, we therefore pre-specified the landmark to one year. This selection may have influenced the results but has a clinical rationale.