840 likes | 994 Vues
Scottish Stroke Audit. 3rd National Meeting 7th Dec 04. Welcome. NHS QIS funded audit - Oct 02 - 05 Original plan - 6 to 10 hospitals Impact of CHD & Stroke strategy NHS QIS standards and visits. Program. Comparisons between hospitals Control charts
E N D
Scottish Stroke Audit 3rd National Meeting 7th Dec 04
Welcome • NHS QIS funded audit - Oct 02 - 05 • Original plan - 6 to 10 hospitals • Impact of CHD & Stroke strategy • NHS QIS standards and visits
Program • Comparisons between hospitals • Control charts • Demonstration of real time data capture system • Audit of swallow screening • Update on “MCNs on the Web”
How can these data help improve patient care? • Identify variation in “performance” and to raise questions about cause of variation • Identify methods which increase performance? • Highlight services requiring more investment or re design
Reasons for variation in “Performance” • Method of collection data • Definitions, case ascertainment and audit period • Method of analysing data • Which numerator and denominator? • Chance • Actual performance of service
Some statistical terms • Proportions (%) • 95% confidence intervals • Means and medians • Inter quartile range
Proportions • Numerator / Denominator = Proportion • 100 patients admitted • 60 enter stroke unit • Proportion is 60/100 = 0.6 or 60% • We have had problems with denominators • NHS QIS ask % admitted SU within 1 day • Is denominator 60 or 100?
Denominators • If admit 100 stroke patients • 60 enter the stroke unit • therefore 60% managed in a stroke unit • if half get into stroke unit within a day • % admitted to SU <1day = 50% or 30% • NHS QIS want 30% figure
95% confidence intervals • Measure the proportion entering your stroke unit once • Calculate the 95% Confidence intervals • Measure the proportion a further 100 times and one would expect 95 estimates to lie within the 95% confidence intervals.
A normal distribution No. of patients Mean = 10 Median = 10 Length of stay in Days
Length of stay Mean = 10 Median = 10 No of people. Length of stay (days) Mean = total no. of days / total no. of people Median = LOS where half the people have longer ones and half shorter ones
A skewed distributione.g. length of stay in acute stroke unit No. Mean = 7.3 Median = 6 Days
A very skewed distributione.g. delay to CT scan No. Mean = 4.9 Median = 3 Days
Quartiles (quarters) No. Mean = 10 Median = 10 Days Interquartile range (IQR) (half the patients are included)
Comparisons between hospitals A few hospitals which are currently collecting data are not included because too few data are available.
Data collection periods vary • Longer period will provide more patients and more precise estimates • Longer period will include older data • Recent short period will not include patients still in hospital - therefore may give biased estimates
No. of admissions available for analysis – Group 1 Ninewells estimates will have wide Confidence intervals so differences are more likely to be due to chance
No. of admissions per year Group 1 At Ninewells may be missing cases - not identified or simply not yet discharged
No. of admissions per year -Group 4 The estimate in your hand out for St Johns is incorrect
Length of Stay in HospitalMean MedianGroup 1 Patients with longer LOS in Ninewells not yet discharged Why is LOS shorter in ARI than Edinburgh??
Length of Stay in HospitalMean MedianGroup 2 Length of Stay in HospitalMean MedianGroup 2
Length of Stay in HospitalMean MedianGroup 3 Length of Stay in HospitalMean MedianGroup 3 Two fold difference Monklands & Falkirk - why?
Length of Stay in HospitalMean MedianGroup 4 Length of Stay in HospitalMean MedianGroup 4
Length of Stay in HospitalMean MedianGroup 5 Length of Stay in HospitalMean MedianGroup 5 Imprecise estimates because small numbers Shetland a different model of service?
Proportions admitted to Stroke Unit – Group 1 77 beds 42 beds 18 beds Note the 95% CI vary with amount of data collected
43 beds 35 beds 30 beds 16 beds 10 beds Proportions admitted to Stroke Unit – Group 2 Ayr & Crosshouse are doing well! - ? chance because only 3 month 7.6-8.3 pts/SU bed/yr cf 14 pts/SU bed/yr in Inverclyde
Proportions admitted to Stroke Unit – Group 3 24 bed 24 beds 25 beds 30 beds 15 male
Proportions admitted to Stroke Unit – Group 4 15 beds 14 beds 21 beds 17 beds 8 beds 0 beds
Proportions admitted to Stroke Unit – Group 5 6 beds Variable
Proportion of admission in Stroke Unit – Group 1 Reflects delay in admission % entering SU and exit from SU before discharge X
Proportions scanned – Group 1 ARI seem to be having problems getting scans
Proportions scanned – Group 2 Delays in Ayr and Crosshouse
Proportions scanned – Group 4 Raigmore and Victoria Hospital Kirkaldy having problems