Discussion

In this cohort study emotional, social, functional and behavioural prognostic factors were tested for patients undergoing cardiac surgery. The principal findings were that 1) living alone predicted both prolonged ICU stay and death for the total cohort of 3217 patients, and 2) low HRQoL and loneliness (not having someone to talk to) predicted prolonged hospital stay for the nested cohort of 982 patients undergoing cardiac surgery. Thus, information on cohabitation status may potentially be added to existing risk evaluation models due to its predictive value.
The predictive value of living alone is supported by Murphy & colleagues who found patients undergoing CABG surgery and living alone, were more than three times more likely to be readmitted to hospital (OR, 3.42; 95% CI, 1.38– 8.48) than those living with others (32). Being married, especially being in a highly satisfying marriage, has been found to offer a significant benefit to long-term survival after CABG (OR, 2.49; 95% CI, 1.47–4.24) (33). The beneficial effect of cohabitation and relationship satisfaction on survival is likely multifactorial, which has been emphasised by earlier studies linking living alone with poor health outcomes. Patients who are socially isolated are more likely to smoke and have excessive alcohol intake (34,35), delay seeking treatment (36), and demonstrate non-compliance with medical regimens (37), which may be due to a lack of emotional or practical support gained through living with another person (32).
In earlier studies a feeling of loneliness has been linked to several adverse health outcomes. For example, endorsing “yes” to “I feel lonely” was associated with increased 30-day (Rate Ratio (RR), 2.61; 95% CI, 1.15-5.95) and 5-year (RR, 1.78; 95% CI, 1.17-2.71) mortality among patients undergoing CABG (38) , and a response of “often” to the question “do you feel lonely” was associated with increased cardiovascular mortality among elderly Danish men (Hazard Ratio, 1.70; 95% CI, 1.03-2.81) (39).
Several studies agree that HRQoL has become a necessary addition and key indicator of cardiac surgical outcomes (40–42). This study found that reduced health-related quality of life predicts prolonged LOS-HOSP. The predictive value of HRQoL has been confirmed in earlier studies that have found low HRQoL to be predictive of both mortality following CABG with a 10 point lower SF-36 Physical Component Summary score having an OR of 1.39; 95% CI, 1.11-1.77 (43) and of one year cardiac functional status (OR, 2.73; 95% CI, 1.43–5.23) (44).
For this study the intention was to investigate factors beyond the clinical indicators and physical health of the patients planned to undergo cardiac surgery. Traditional risk assessment in cardiac surgery has been a tool for patient selection and has been aimed at the perioperative patient pathways. With the proposed supplement the risk assessment can potentially be used to identify vulnerable groups of patients leading to improved patient management still with the overall aim to improve patient outcomes. Information on cohabitation status, loneliness and HRQoL could potentially be added to existing risk evaluation models in cardiac surgery. However, further research is warranted to validate the findings of the current study and to investigate interventions supporting the identified vulnerable groups of patients.

Strength and limitations

This study has several limitations. Firstly, we were restricted to the use of predictor variables based on existing data measured in previously collected data sets, which is a beneficial way to make full use of already collected data to address potentially important new research questions and avoid disturbing patients unnecessarily. However, we may not have included important prognostic variables (e.g. cognitive status and frailty), because they were not measured in the original studies. Secondly, the present study used corresponding datasets. When doing this there is a risk that the datasets differ in important aspects, such as baseline risk. However, in the current study a prediction model was developed for each dataset reducing bias due to this.
Non-response for the DenHeart study was high at 49% which might bias the results. Responders and non-responders of the DenHeart study has earlier been established to be similar regarding socio-demographics, however, the non-responders were more severely ill, had more comorbidity and thus a much higher mortality rate compared to responders (45), which could have resulted in an underestimation of the associations between the predictor variables and the outcomes .
Imputations were utilised in the present study to maintain the sample size, assuming the missing values were missing at random. The use of mean imputations does not affect the estimate of the mean for the variable; however, it reduces the variance of the imputed variables. Furthermore, it assumes that the mean value of the respondents was a good estimate of the missing values, which may have resulted in conservative bias.
We used an automated stepwise approach to specify the models, principally due to its objectivity and that it generally results in smaller, clinically applicable models (46), but stepwise methods have well-known limitations such as unstable variable selection (47) and biased coefficient estimation (46). It is therefore conceivable that our choice to use stepwise selection may have reduced the predictive performance of the models. The overall model fit statistics indicate that the variance explained by our prediction models is at best modest. Perhaps some factors that are yet to be tested thoroughly in cardiac surgery, for example, frailty and mental state, explain additional variance in cardiac surgery. Despite the limitations of the study the models made informative predictions that should be externally validated in a similar population of patients undergoing cardiac surgery.

Conclusion

We tested several emotional, social, functional and behavioural prognostic factors as a supplement to EuroSCORE and reported different aspects of model performance that can be interpreted for further research applications. Based on the cohorts included, living alone predicts death, prolonged hospital admission and prolonged ICU stay following cardiac surgery. Low educational level and impaired HRQoL were, furthermore found to be predictive of prolonged hospital admission.