Back in January, there was a discussion of stroke certification at Alameda Hospital. CEO Stebbins was happy that the risk adjusted mortality rate for the Hospital for 2009 was not identified by the State as being statistically worse than the state average. I was disgusted with the whole discussion and the comfort level the people in the room had with outright ignorance, especially the medical people. For the record, Stebbins contended that these numbers showed Alameda Hospital doing a good job even though the statewide average for mortality was 10.4% and Alameda Hospital’s was 12.6.
I made a statement about the statistics and was roundly ignored or even outright criticized for trying to apply analysis to these numbers. Now that Alameda Hospital is authorized by the County EMS to receive stroke victims again, I want to revisit those numbers. Here is a simple chart of the 2009 (most recent) numbers for the hospitals that County EMS will allow stroke victims to be brought to if within 4 hours of onset of symptoms:
|Alta Bates Summit Medical Center – Alta Bates Campus||8.2|
|Alta Bates Summit Medical Center – Summit Campus – Hawthorne||7.8|
|Eden Medical Center||9.5|
|Kaiser Foundation Hospital – Hayward||7.3|
|Kaiser Foundation Hospital – Oakland Campus||7.1|
|Washington Hospital – Fremont||7.5|
|San Jose Regional Med. Ctr||8.3|
7 of thes hospitals are in Alameda County and 3 are outside but listed in the County EMS manual. If something doesn’t pop right out at you when looking at those numbers then I have to conclude that you are innumerate. Only one hosptial in that list has a double digit risk adjusted mortality. Only one hospital has a rate that is below the statewide average [“below” means inferior in this sentence] . But we, in Alameda, really do not care specifically about the statewide average, we care (or should unless you are trying to spin the numbers) about the alternatives where we may actually receive treatment. The average for these 9 hospitals (excluding Alameda Hopital) is 8.1%.
Now I can take this one step further and will, but didn’t earlier because I did not think it would convince anyone and because I am not a statistical expert (“a little knowledge is a dangerous thing”) even though I am fairly confident that I know more statistics than the rest of the Board and Management (“in the land of the blind, the one-eyed man is king”). I used an online calculator like this one. Given my hypothesis that Alameda Hospital is worse than average, I was comfortable using a one-sided test in which the difference is statistically significant p=0.04 (less than 5% chance that the difference is due to random variaton). A more conservative approach where the hypothesis is simply that the two proportions are not equal uses a two sided test where the difference, in this case, is not statistically significant p=0.14 (less than a total of 5% chance that the difference is due to random variation but split between both sides of the distribution – 2.5% chance on each side). Of course if you are a Bayesian and your prior is that Alameda Hospital is likely inferior in outcomes , then you don’t even have to resort to a test since the number for Alameda is the worst of the lot.
So you can decide whether I was doing a little spin of my own. I think whatever you decide about my statements, you can readily conclude that Stebbins’s contention that these numbers showed that these quality metrics are no cause for concern is a significant overstatement.