Report >> |
3. Data overview
| Introduction |
| Hospital participation |
| Clinical questionnaires |
| Organisational questionnaire |
| Age and sex |
| Admission method to hospital |
| Source of admission to the ICU |
| Severity of the patient’s condition |
Hospital participation
261 hospitals were identified as having a Level 3 adult, general, intensive care unit.
197 hospitals submitted at least one clinical questionnaire and we had an overall participation rate
of 88% (229/261).
The reason for the non-participation of the remaining 32 hospitals is shown in Figure 1.
Figure 1. Reason for non-participation
Initially the estimated sample size was 6,000 patients admitted to ICU with 20-30% of patients expected to die during the data collection period. As data collection began and the lists of suitable patients and questionnaires were returned to NCEPOD, it became apparent that the expected sample size would not be reached. To evaluate this further, hospitals were contacted and asked to provide additional information on the total number of admissions to their ICU during one year and the number of these admissions that were general medical admissions.
42 hospitals provided this extra information. Based on this, the average number of admissions to an ICU over one year was 468. Of these, only 40% were medical admissions. Therefore the average number of medical admissions per unit would have been approximately 15 in a one month period. This figure was more consistent with the numbers associated with the study period.
By taking the above findings into account, it was proposed that the original estimate had been overestimated by approximately 60% and a more realistic sample size to be expected was in the region of 2,400 cases. It was likely that the overestimation arose from the fact that the multiplication factor of 7 and the number of estimated beds were not for medical patients only.
These calculations did not take into account different sizes of units, bed numbers or the time of the year that the study was run, all of which may have impacted on the final sample size. However, it did provide a crude indication to the number of cases expected.