Northwest Renal Network

Mortality Report 1999-2001

December 25, 2002

Northwest Renal Network

4702 – 42^nd Ave SW

Seattle WA 98116

(206) 923-0714

fax (206) 923-0716

prepared by Jim Buss, MA, CDP

jbuss@juno.com

This report has been prepared under contract #500-00-NW16 with the

Centers for Medicare and Medicaid Services

Northwest Renal Network Mortality Report, 1999-2001

December 25, 2002

After a one-year hiatus, this is the sixth in our series of annual Network Mortality Reports.

A Standardized Mortality Ratio (“SMR”) compares mortality in a small population of patients, to mortality in a larger or reference population. An SMR of 1.00 means that mortality in the smaller population is the same as mortality in the reference population.

An SMR of 1.20 means that mortality in the smaller population is 20% higher than mortality in the reference population, and an SMR of 0.80 means that the smaller population has experienced 20% less mortality than that of the reference population.

In this report we compare mortality among chronic kidney dialysis patients in the Pacific Northwest (Washington, Oregon, Idaho, Montana, Alaska) during 1999-2001, to mortality among chronic kidney dialysis patients in the United States as a whole during the year 2000. That is, our “smaller population” is patients in the Northwest, and our “reference population” is patients in the US as a whole. For this comparison, patients are stratified by age, race, primary renal diagnosis, modality of treatment, and gender.

Our reference population rates are from the United States Renal Data System (“USRDS”). These rates can be found in tables H-4 and H-5 in section H on the website, http://www.usrds.org/reference.htm.

The method of comparison we use is a simple one. We look up an “expected mortality” for each patient in the national tables, based on their age, race, primary renal diagnosis, modality, and gender. We also compute 95% confidence intervals on each patient’s “expected mortality,” based on how much variation there is nationally among patients of this particular age-race-diagnosis-modality-gender group. An expected mortality is a fraction between zero and one which indicates a patient’s likelihood of dying within one year.

Then for any group of patients, we simply add up each patient’s expected mortality, and add up each patient’s “actual mortality.” “Actual mortality” is either zero if the patient is still alive, or one if not.

The SMR for the group is the total actual mortality divided by the total expected mortality.

We also compute an upper 95% confidence interval on the SMR by dividing total actual mortality by the sum of the lower confidence interval for each patient’s expected mortality, and compute a lower 95% confidence interval on the SMR by dividing total actual mortality by the sum of the upper confidence interval for each patient’s expected mortality. This method is explained in detail at

http://www.nwrenalnetwork.org/mortality/runsmrci.htm.

This method works very well for small groups of patients, and it’s a conservative method – that is, it’s unlikely to label a group as having high mortality when in fact it doesn’t.

A group is considered to have statistically significantly high mortality (at p < 0.05) if the lower confidence interval of its SMR is greater than 1.00.

Note: Since the method is conservative, however, it could label a group as not having low mortality when in fact it does.

A group is considered to have statistically significantly low mortality if the upper confidence interval of the SMR is less than 1.00.

If your group has an upper confidence interval greater than 1.00 and there are more than about 20 patients in the group, you can calculate an alternate upper confidence interval in the following way. Divide the total actual deaths by the square of the total expected deaths. Take the square root of the result, and multiply by 1.96. Add this to your SMR to obtain your adjusted upper confidence interval. The more patients there are in your group, the greater the difference between this confidence interval and the one calculated by the method used in our program. If this new upper confidence interval is less than 1.00, your mortality is statistically significantly low, at p < 0.05. This alternate calculation is the same one that was used in our previous Mortality Reports.

We eliminate patients during their first ninety days on dialysis ever, but we don’t exclude patients because of a failed prior transplant, a previous recovery episode, or a previous withdrawal episode. Except for excluding that first ninety days, we count patients whenever they’re on dialysis in this Network. Patients are counted at a facility whenever they are dialyzing there, and deaths are attributed to the facility at which the patient last dialyzed, even if the patient withdrew from dialysis prior to death. The SMR reports from the University of Michigan Kidney Epidemiological Center (and the Colorado Foundation for Health Care) use different methods, different assumptions, and a different dataset.

Summary of Results

In our 1997-1999 Mortality Report, we reported statistically significantly high mortality for Native American patients, high mortality for a number of Network facilities, and trends toward higher relative mortality (that is, SMRs) among female patients, patients with a primary renal diagnosis of diabetes, and in-center hemodialysis patients.

We’re happy to report that in the 1999-2001 period none of these subgroups has significantly high mortality, and no subgroup demonstrates a trend toward increasing mortality.

We’ve printed out two different sets of tables. The first set gives SMRs by year over the three-year period, 1999-2001. The second set shows the last six quarters, through October 2002. For the Network as a whole, these tables are reproduced on pages 7-8.

Native American Patients

For Northwest Native American patients, the actual mortality rate has statistically significantly decreased, from 24% in 1997 to 19% in 1999 to 16% in 2001 (p < 0.05), and the SMR is no longer higher than 1.00.

During this time span the average age of Network Native American patients rose from 56.4 to 56.8. The SMR for these patients in 2001 was 0.90, not yet

Figure 1. Comparing Actual Mortality Rates for Native American Patients and All Patients, US versus Network, 1997-2001

statistically significantly lower than US Native American patients, but if the current trend continues, it will take us there soon. In prior reports, mortality for Network Native American patients has been statistically significantly higher than mortality for all US Native American patients. Figure 1 shows the decline in actual mortality rates. SMRs for Native American patients are tabled on page 9.

Individual Facilities with High Overall SMRs

We asked thirteen facilities which had high SMRs in our earlier 1997-1999 Mortality Report to prepare QI plans to investigate possible root causes for these high SMRs, and to see if they could reduce their mortality – and they did.

At these thirteen facilities taken together, the actual mortality rate fell from 34% in 1999 to 24% in 2001, and the SMR from 1.61 to 1.02. This is a statistically significant reduction that extended more than a hundred lives.

Only three of the thirteen facilities remain statistically significantly high in the year 2001, and all three have substantially lower SMRs.

Figure 2. SMRs by Facility for All Patients

In the 1999-2001 period, fifteen currently operating facilities demonstrate statistically significantly high mortality, but many of these are trending down. Only seven statistically significantly high facilities show uptrends or consistently high SMRs during 1999-2002, and the Network will work with these facilities. We’re excluding facilities with less then ten patient-years during 1999-2001. Twenty-six facilities had statistically significantly low mortality. Individual facilities are plotted in Figure 2.

Patients with a Primary Renal Diagnosis of Diabetes

While patients with a primary renal diagnosis of diabetes have a 1999-2001 SMR of 1.0 in the Network as a whole, there remain seventeen currently operating facilities with statistically significantly high 3-year SMRs for these patients.

Figure 3. SMRs by Facility for All Patients with a Primary Renal Diagnosis of Diabetes

This compares to twenty-six facilities with statistically significantly low SMRs for patients with a primary renal diagnosis of diabetes. These are plotted on Figure 3. Again we’ve excluded facilities with less than ten patient-years. We will work with selected facilities that have high SMRs for diabetic patients, at the very least to make sure everyone is familiar with the new literature on treating diabetic dialysis patients (http://www.renalweb.com/topics/out_diabetes/diabetes.htm).

Patients with a Primary Renal Diagnosis of Hypertension

Network patients with a primary renal diagnosis of hypertension also have an overall 1999-2001 SMR of 1.0, but there are twenty currently operating facilities with statistically significantly high 3-year SMRs for these patients.

Figure 4. SMRs by Facility for All Patients with a Primary Renal Diagnosis of Hypertension

Twenty-one facilities have statistically significantly low 3-year SMRs for these patients. Figure 4 plots these facilities. Again, we’ve excluded facilities with less than ten patient-years.

We will work with selected facilities that have high SMRs for hypertension patients, at the very least to make sure everyone is familiar with the new literature on treating hypertensive dialysis patients (http://www.renalweb.com/topics/out_bp/bp.htm).

Northwest Renal Network Mortality Report

All Patients 1/1/1999 to 1/1/2002

Standard mortality rate data source: USRDS national rates for 2000.

Calculated figures exclude the First 90 Days of Dialysis ever

Date Range Pat Pat Deaths Actual Exptd SM Lower Upper

Count Years Rate Deaths Ratio Conf Conf

All

1/1/1999 to 1/1/2000 8055 5876.6 1405 0.239 1395.0 1.01 0.86 1.21

1/1/2000 to 1/1/2001 8562 6203.2 1507 0.243 1442.8 1.04 0.90 1.25

1/1/2001 to 1/1/2002 9016 6626.0 1482 0.224 1516.6 0.98 0.84 1.17

1/1/1999 to 1/1/2002 13090 18705.8 4394 0.235 4354.4 1.01 0.87 1.21

Black

1/1/1999 to 1/1/2000 743 585.9 104 0.177 100.7 1.03 0.89 1.23

1/1/2000 to 1/1/2001 783 603.8 100 0.166 99.9 1.00 0.86 1.20

1/1/2001 to 1/1/2002 813 646.5 87 0.135 103.3 0.84 0.72 1.01

1/1/1999 to 1/1/2002 1118 1836.2 291 0.158 303.9 0.96 0.82 1.14

Female

1/1/1999 to 1/1/2000 3641 2673.0 663 0.248 645.2 1.03 0.88 1.23

1/1/2000 to 1/1/2001 3852 2799.6 682 0.244 656.2 1.04 0.89 1.25

1/1/2001 to 1/1/2002 4063 2981.8 671 0.225 684.0 0.98 0.84 1.18

1/1/1999 to 1/1/2002 5876 8454.4 2016 0.238 1985.5 1.02 0.87 1.22

In-Center Hemodialysis Patients

1/1/1999 to 1/1/2000 6881 4821.3 1196 0.248 1164.3 1.03 0.88 1.23

1/1/2000 to 1/1/2001 7419 5213.9 1323 0.254 1232.9 1.07 0.92 1.28

1/1/2001 to 1/1/2002 7894 5617.6 1317 0.234 1307.3 1.01 0.87 1.21

1/1/1999 to 1/1/2002 11582 15652.8 3836 0.245 3704.4 1.04 0.89 1.24

Home Peritoneal Dialysis Patients

1/1/1999 to 1/1/2000 1357 842.6 167 0.198 185.5 0.90 0.76 1.10

1/1/2000 to 1/1/2001 1311 809.1 145 0.179 172.2 0.84 0.71 1.03

1/1/2001 to 1/1/2002 1337 849.6 130 0.153 177.2 0.73 0.62 0.90

1/1/1999 to 1/1/2002 2316 2501.3 442 0.177 534.9 0.83 0.70 1.01

Diabetes

1/1/1999 to 1/1/2000 3193 2278.3 662 0.291 629.5 1.05 0.91 1.25

1/1/2000 to 1/1/2001 3457 2440.7 700 0.287 655.8 1.07 0.92 1.27

1/1/2001 to 1/1/2002 3683 2681.1 674 0.251 711.6 0.95 0.82 1.13

1/1/1999 to 1/1/2002 5409 7400.1 2036 0.275 1996.9 1.02 0.88 1.22

Hypertension

1/1/1999 to 1/1/2000 1402 1004.0 279 0.278 293.6 0.95 0.81 1.15

1/1/2000 to 1/1/2001 1494 1074.9 306 0.285 308.1 0.99 0.85 1.20

1/1/2001 to 1/1/2002 1615 1147.6 335 0.292 318.8 1.05 0.90 1.27

1/1/1999 to 1/1/2002 2344 3226.5 920 0.285 920.4 1.00 0.85 1.20