Description of data used
Unless otherwise specified, the analysis described in this report are based on all available data from 1 January 2011 until the launch of the OHCA Strategy for Scotland (28 March 2015).
This will enable future analysis to assess the impact of the OHCA Strategy.
Combining data from several consecutive years provides a more statistically robust baseline.
Calculations of survival rates - assumptions made about unlinked episodes
For the calculation of the survival percentages the assumption has been made that the OHCA cases which we were unable to link to their CHI number belonged to individuals pronounced dead at the scene of their cardiac arrest. It is relatively unusual for patients who are transported to hospital to have less than the minimum data required for linkage recorded in their SAS data. For this reason we have assumed that the majority of unlinked episodes represent patients who were not conveyed from the scene of their arrest by SAS.
Logistic regression analyses
The majority of the results presented in the section describing survival after OHCA are based on logistic regression analysis. Firstly the results from univariable logistic regression models are presented which represent the crude associations. In addition, the results from multivariable (adjusted) regression analysis is presented. The advantage of this approach is the opportunity to calculate estimates of the independent effect of variables while adjusting for other factors.
The choice of the confounding factors is based on existing clinical knowledge, literature review and availability of data. An example of this is the association between living in an urban or rural area and 30-day survival presented in Figure 11. First, the crude results show an odds ratio of 1.22 and an accompanying 95% confidence interval of 0.98-1.52. Based on previous knowledge and descriptive analysis it is known that the age distribution between patients living in urban and rural populations is different. Furthermore, earlier results in Figure 9 show that age and 30-day survival are associated (younger patients are more likely to survive an OHCA) and age is not part of the causal pathway in the association of urban/rural and survival. Therefore, age is a potential confounder in the association between urban/rural and 30-day survival. For this reason, taking age into account as a possible confounder will yield a better reflection of the true association between urban/rural residence and 30-day survival. The multivariable logistic regression model, adjusting for age, gender and SIMD quintiles results in an odds ratio of 1.32 and an accompanying 95% confidence interval of 1.04-1.66.
Adjustments for variables in the causal pathway
Where appropriate for analysis included in this report, adjustment for variables on the causal pathway has been made. For example, in Figure 8 the association between sex and survival is presented. First, the crude results are given, showing a difference in survival between males and females. Thereafter, the association is adjusted for the presenting heart rhythm (shockable or non-shockable). This variable is likely to be part of the causal pathway. The adjusted results show no survival difference between males and females. This demonstrates that the fact that more females present with a non-shockable rhythm makes an important contribution to the explanation of sex differences in survival.
Completeness and quality of the data
This report describes initial work using this Scottish OHCA data with only a part of the linked dataset analysed in detail. The completeness of the variables used in this report is shown in table 5.
Table 5: completeness of the variables
|Variable||Number of missing values|
|Age at time of OHCA||1|
|Urban rural category||207|
Table 5 shows that the majority of the data used for this report, with the exception of bystander CPR is reasonably complete. Completeness of the data varies between 91.1 and 99.9% for the variables in the dataset. As already mentioned the completeness and quality of bystander CPR data is poor and work is underway to improve this data.
Most of the datasets in these analysis are widely used in research (see Table 4) and the data quality is generally regarded as good. The SAS data used for this project is under-utilised and somewhat under-developed. One of the longer-term goals of this work is to identify mechanisms to improve the data quality of SAS data.
It is worth noting that in this report, location related characteristics of OHCA cases such as SIMD quintile and urban-rural classification are all based on the home address of the OHCA patient. For most analysis this will be valuable information, though some questions are better answered using the geographical data relating to the location where the OHCA occurred. This type of analysis will form part of future work.
There is a problem
Thanks for your feedback