Northwest Renal Network

Mortality Report 2002-2005

March 15, 2006

Northwest Renal Network

4702 – 42nd 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 #500-03-NW16 with the

Centers for Medicare & Medicaid Services

A Standardized Mortality Ratio (“SMR”) compares mortality in some specified group to mortality in some appropriate reference population.  SMRs are expressed as ratios – an SMR of 1.00 indicates that mortality in the specified group is the same as mortality in the reference population, an SMR of 1.10 tells us that group mortality is 10% higher than the reference, and an SMR of 0.90 shows group mortality 10% lower than expected.  An SMR is basically actual mortality divided by expected mortality.

            An SMR is often accompanied by a “confidence interval” that identifies the degree of probable random error.  For example, an SMR of 1.07 with a confidence interval of 0.97 to 1.17 has more probable error than a similar SMR that has a confidence interval of 1.01 to 1.13.  A 95% confidence interval is often used.  This implies that there is a 95% probability that the real SMR falls within the range given by the specified confidence interval.  In general, when a group’s lower confidence interval is above 1.00, the group’s mortality would be considered “statistically significantly” higher than the reference standard.  When a group’s upper confidence interval is below 1.00, its mortality would be considered “statistically significantly” lower then the reference. 

            The University of Michigan Kidney Epidemiology Cost Center (“KECC,” http://www.sph.umich.edu/kecc) calculates a set of SMRs annually for each chronic kidney dialysis facility in the US, for inclusion in their Dialysis Facility Reports.  These use a different methodology than the SMRs described here, and both should be considered useful.  The KECC SMRs are the basis for mortality comparisons on the CMS Dialysis Facility Compare (“DFC”, http://www.medicare.gov/Dialysis/Home.asp) website.  On DFC a facility is considered to have mortality “higher than expected” if the lower confidence interval of its SMR is above 1.20. 

Every several years the Northwest Renal Network (“the Network,” http://www.nwrenalnetwork.org/) also calculates SMRS for its member facilities, for the Network as a whole, and for other subgroups of Network patients.  The Network uses a different method than KECC uses, different data, different assumptions, and more timely data.  Our method was chosen because it is accurate for small populations, and is easy enough to calculate that any facility could use it to track their mortality.  Instructions for using our method are given in http://www.nwrenalnetwork.org/mortality/runsmrci.htm. 

KECC uses billing data, while we use patient data.  The primary difference in our assumptions is that KECC attributes a patient to a facility for the whole year if they are being treated by that facility on January 1, while the Network attributes a patient to the facility at which they’re being dialyzed, no matter how often they transfer.  Accordingly, in our method patient deaths are attributed to the facility at which they last dialyzed, even if the patient voluntarily withdrew from dialysis prior to death.  Neither KECC nor the Network includes patients during their first ninety days on dialysis. 

SMRs are standardized by specific patient characteristics.  For instance, mortality tables for the reference population might be broken down by gender, race, and age.  Then the expected mortality for each patient would be read from the appropriate gender, race, and age categories in the reference mortality table.  The SMRs we report here are standardized by gender, race, age, primary renal diagnosis, and dialysis modality.  The reference table we use is the latest available complete national data, for 2002, taken from the USRDS 2004 Annual Data Report (“ADR,” http://www.usrds.org/adr.htm).  The mortality tables for 2003 in the 2005 ADR lack standard errors, which are necessary to calculate confidence intervals, so we did not use them. 

            In this report we calculate SMRs for the period between July 2002 and June 2005, and for each of the three twelve-month periods within that time period.  For July 2002 through June 2005, no major group of Network patients had statistically significantly high mortality, and no major group of Network patients had statistically significantly low mortality.  At the Network level, no major group of Network patients had mortality statistically significantly different from any other group.  These groups include: 

All patients (SMR 0.99, 95% confidence interval 0.83 to 1.24),

Each Network state (Alaska, Idaho, Montana, Oregon, and

Washington),

Each race (Native American, Asian, Black, White),

For-profit and not-for-profit,

Each 5-year age group,

Each major primary renal diagnosis group (hypertension, diabetes,

glomerulonephritis, and other),

Each gender,

Each major dialysis modality (in-center hemodialysis and home

peritoneal dialysis), and

Each provider group.

SMRs for the Network and subgroups are tabled at the end of this report. There are no published national rates for home hemodialysis.

            Not counting facilities with less than twenty patient-years over the three-year period, there were eighteen individual facilities with statistically significantly higher mortality than the US as a whole, and twenty facilities with statistically significantly lower mortality than the US as a whole, using p < 0.05 as the threshold of statistical significance.  With more than one hundred Network facilities, we would expect about five facilities to cross this threshold by chance, in each direction.   

        The eighteen individual facilities with lower 95% confidence intervals higher than 1.00 (that is, with statistically significantly higher mortality than the US as a whole at p < 0.05) were asked to bring this matter to the attention of their medical director and quality improvement committee, and report back to us a summary of

any plans they institute to study or correct this difficulty.  They were also invited to ask for Network assistance. 

         The twenty Network facilities which had upper 95% confidence intervals less than 1.00 (i.e., statistically significantly lower mortality than the US as a whole at p < 0.05) were sent a letter congratulating them, and inviting them to share any practices which they believe may have contributed to this outcome. 

            The remaining facilities were sent their SMR outcomes, asked to examine them to see if any subgroups were statistically significantly high or low, invited to share any information they thought might contribute to their low-mortality subgroups, and invited to ask for Network assistance with their high-mortality subgroups.  For instance, seven facilities have an SMR for patients with a primary renal diagnosis (“PRD”) of diabetes that is twice as high as their SMR for patients with PRD of hypertension, while at five facilities the SMR for patients with PRD of hypertension is statistically significantly high and the SMR for patients with PRD of diabetes statistically significantly low.  Five facilities have statistically significantly high mortality for home peritoneal dialysis patients, eighteen for female patients, nine for black patients, and six for Native American patients. 

            In our previous report (http://www.nwrenalnetwork.org/mort01a.htm) we discussed a steady improvement in Network Native American dialysis mortality rates, from 24% in 1997 to 16% in 2001.  Has this trend continued?  In 2002 yes, but in 2003 it jumped back to 1999 levels, before starting back down in 2004.

             The table at the end of this report is similar to the report received by each facility.  It includes sections for all patients, Native American patients, black patients, female patients, in-center hemodialysis patients, home peritoneal patients, patients with PRD of diabetes, and patients with PRD of hypertension.  For each subgroup, an SMR is given for each year, and, in bold, for the three years together.  Mortality for a single year can vary widely for small populations. The three-year SMR averages out those variations, while the three one-year SMRs can be used to identify trends.

            The columns in the table include: 

“Pat Count” – the number of individual patients counted during the period

“Pat Years” – the number of patient-years tallied

“Deaths” – the number of actual deaths

“Actual Rate” – the actual crude mortality rate

“Exptd Deaths” – the number of deaths that would have occurred had the subgroup experienced mortality identical to national rates

“SM Ratio” – the SMR

“Lower Conf” – the lower 95% confidence interval of the SMR; when this number is higher than 1.00, it indicates statistically significantly high mortality           

“Upper Conf” – the upper 95% confidence interval of the SMR; when this number is below 1.00, it indicates statistically significantly low mortality. 

All Patients

Standard mortality rate data source: USRDS national rates for 2001 (2004 ADR). Calculated figures exclude the First 90 Days of Dialysis ever

Date Range	Pat Count	Pat Years	Deaths	Actual Rate	Exptd Deaths	SM Ratio	Lower Conf	Upper Conf
All
7/1/2002 to 7/1/2003	9782	7163.9	1641	0.229	1674.4	0.98	0.82	1.23
7/1/2003 to 7/1/2004	10238	7530.5	1780	0.236	1724.0	1.03	0.86	1.29
7/1/2004 to 7/1/2005	10523	7855.8	1689	0.215	1752.9	0.96	0.80	1.20
7/1/2002 to 7/1/2005	15324	22550.2	5110	0.227	5151.3	0.99	0.83	1.24
Native American
7/1/2002 to 7/1/2003	423	347.6	51	0.147	60.1	0.85	0.63	1.28
7/1/2003 to 7/1/2004	437	341.3	69	0.202	56.7	1.22	0.92	1.81
7/1/2004 to 7/1/2005	457	355.4	65	0.183	57.8	1.13	0.86	1.64
7/1/2002 to 7/1/2005	622	1044.4	185	0.177	174.6	1.06	0.80	1.57
Black
7/1/2002 to 7/1/2003	911	698.2	122	0.175	120.0	1.02	0.85	1.27
7/1/2003 to 7/1/2004	926	732.3	109	0.149	122.9	0.89	0.74	1.11
7/1/2004 to 7/1/2005	958	764.7	115	0.150	125.1	0.92	0.77	1.15
7/1/2002 to 7/1/2005	1317	2195.2	346	0.158	368.1	0.94	0.78	1.17
Female
7/1/2002 to 7/1/2003	4371	3214.9	749	0.233	761.9	0.98	0.82	1.23
7/1/2003 to 7/1/2004	4570	3369.9	837	0.248	776.7	1.08	0.90	1.35
7/1/2004 to 7/1/2005	4651	3496.0	748	0.214	787.1	0.95	0.79	1.19
7/1/2002 to 7/1/2005	6806	10080.7	2334	0.232	2325.7	1.00	0.84	1.26
In-Center Hemodialysis Patients
7/1/2002 to 7/1/2003	8629	6155.8	1469	0.239	1464.8	1.00	0.84	1.25
7/1/2003 to 7/1/2004	9030	6473.6	1600	0.247	1510.8	1.06	0.88	1.32
7/1/2004 to 7/1/2005	9312	6774.3	1523	0.225	1542.0	0.99	0.82	1.23
7/1/2002 to 7/1/2005	13776	19403.7	4592	0.237	4517.6	1.02	0.85	1.27
Home Peritoneal Dialysis Patients
7/1/2002 to 7/1/2003	1388	869.5	145	0.167	179.8	0.81	0.67	1.02
7/1/2003 to 7/1/2004	1463	908.8	151	0.166	182.7	0.83	0.69	1.04
7/1/2004 to 7/1/2005	1506	932.4	147	0.158	181.7	0.81	0.67	1.01
7/1/2002 to 7/1/2005	2494	2710.7	443	0.163	544.3	0.81	0.68	1.02
Diabetes
7/1/2002 to 7/1/2003