Download presentation
Presentation is loading. Please wait.
Published byFlorence Stevenson Modified over 9 years ago
1
Dialysis Outcomes and Practice Patterns Study Impact of Dialysis Prescriptions and Practices on Outcomes Friedrich K. Port, MD, MS Arbor Research Collaborative for Health Ann Arbor, Michigan ESRD State of the Art, Boston, MA. April 23-26, 2009
2
DOPPS Overview Prospective observational study, 1997 – 2011 Representative HD samples in 12 countries Practice-patterns in dialysis facilities and outcomes 4 Phases: consistent data collection internationally DOPPS 2: Added focus on incident HD patients DOPPS 3: Added processes of care and nutrition DOPPS 4*: Added practice trends and MD opinions Goal: Improve Outcomes in Hemodialysis: - Mortality, Morbidity, and Quality of Life * 2009-2011 funded by Amgen, Kyowa Hakko Kirin, and Genzyme
3
Dallas Conference, 1989 Held PJ, Brunner F, Odaka M, Garcia J, Port FK, Gaylin DS: Five-year survival for ESRD patients in the U.S., Europe, and Japan 1982-87. Am J Kidney Dis 1990; 15: 451-457. US survival is lower than EDTA or Japan: Why –US captures all deaths, other registries don’t (?) –US case mix or practices explain the differences –Mortality differs in the general populations –Were authors simply wrong?
4
Outline International outcome comparisons –Are outcomes differences real? –Can we explain outcomes difference? Dialysis prescription over the last 20 years Opportunities to improve practices in dialysis –Treatment time, at same Kt/V –Blood pressure –Phosphorus Focusing analyses on Practice Patterns may improve evidence and have practical implications
5
5-Year Survival for ESRD Patients Based on Registries, Adjusted for Diabetes and Age USEDTAUSJapan 5-Year Survival (%) Held et al. AJKD 15: 451, 1990 U 277 99 Includes all dialysis and transplant patients
6
DOPPS I: Survival Among Hemodialysis Patients in Japan, Europe, and the United States: DA Goodkin et al. JASN 14: 3270-3277, 2003 Survival (%) Unadjusted Adjusted for demographics and comorbidities Years US/EU RR =1.33
7
Mortality in the General Population versus the Dialysis Patient Mortality Nathan Levin’s Hypothesis: In international comparisons, higher dialysis patient mortality is partly explained by higher mortality in the general population Methods: Correlate WHO data with Registry + DOPPS data Yoshino et al JASN 2006, 17:3510-3519
8
Relationship of All-Cause Mortality Rates Between Dialysis Patients (DP) and General Population (GP) Yoshino et al JASN (2006) Unadjusted:
9
Relationship of All-Cause Mortality Rates Between Dialysis Patients (DP) and General Population (GP) Yoshino et al JASN (2006) Adjusted for age in DP (overall median mean age [60.4 yr]) and GP (overall median percentage of population aged 65 yr [15.8%]). N=21 countries.
10
International Differences Differences confirmed for US versus EU using detailed adjustment for case mix and same data collection for death ascertainment Better outcomes in Japan may be exaggerated since selection to transplant of healthier patients is minimal in Japan Background mortality partially explains differences* Question: Do practice differences contribute?
11
Outcomes by Vascular Access Use: Problems with patient-based analyses Patients who use a catheter for dialysis tend to be sicker patients Patients using a catheter have higher mortality than patients using an AV fistula Is the higher mortality due to catheter use or due to the selection of sicker patients? The use of catheters varies widely from facility to facility even when adjusted for case mix. This may be a practice pattern: In DOPPS we studied overall mortality in facilities by level of catheter use Pisoni et al. AJKD 2009, 53: 475-491
12
RR of Death among Facility Patients per 20% more facility use of indicated access type *DOPPS I+II, 1996-2004; n=27,892; adjusted for age, gender, black race, yrs with ESRD, 14 comorbidity classes, weight, other practice indicators (median treatment time, % of pts with S. Ca >10 mg/dl or S. PO 4 >5.5 mg/dl) whether hosp unit, & accounted for facility clustering; stratified by study phase & region. Facility access use is adjusted for facility case-mix. Vascular Access Use and Mortality Risk Facility-Based Model p<0.0001 Ref. p=0.008p<0.0001Ref.
13
Fac. Catheter Use (R 2 =0.95) Fac. Graft Use (R 2 =0.966) RR of death % Adjusted Facility Access Use Mortality Risk in Facilities that have Greater Use of Catheters or AV Grafts versus low use Quintiles for Graft and Catheter Use
14
Mortality Risk for US versus European DOPPS is Largely Explained by Vascular Access Practice All models were adjusted for age, gender, race, time on dialysis, 14 summary comorbid conditions, weight, unit type, facility median treatment time, facility % pts with serum phos > 5.5 and serum Ca> 10 mg/dl, and stratified by study phase; accounted for facility clustering effects. DOPPS I + II; n=24,398; *EUR=France, Germany, Italy, Spain, and UK. 0.5 1 1.5 Adjusted for Case Mix 1.00 EUR US 1.36 + Adjusted for Facility Vascular Access Practice 1.06 US 1.00 EUR p<0.0001p=0.43 RR of Death
15
International Differences Differences confirmed for US versus EU using detailed adjustment for case mix and same data collection for death ascertainment Better outcomes in Japan may be exaggerated since selection of healthier patients to transplant is minimal on Japan Background mortality partially explains differences Differences in vascular access practice explain most of the mortality differences between Europe and US: This points to an opportunity to improve vascular access care and outcomes in the US
16
Dialysis Prescription Kt/V Trends Treatment time (TT) and mortality risk (independent of Kt/V)
17
sp Kt/V U S R D S S p e c i a l S t u d i e s CMMS CMAS DMMS U S D O P P S DOPPS 1 DOPPS 2 DOPPS 3 Mean Single-pool Kt/V in US HD Patients during the Past 20 Years Year Adapted from Port et al. CJASN 1:246-255, 2006 Cross-sections of patients by year
18
Mean and Median Patient Prescribed Treatment Time in the US, by DOPPS Phase* Minutes *Prevalent Cross-section of patients in each phase, weighted to represent total facility sample size. (n=3,856)(n=2,260)(n=1,814) Mean Median
19
Distribution of Facility Treatment Time by Country and Phase Mean Facility Treatment Time (min) Phase ANZBeCaFrGeItJaSpSwUKUS Box-plots show the 25 th to 75 th percentiles (box) with median (line) and 5 th and 95 th percentiles (whiskers)
20
Mortality Risk by Average Facility Treatment Time as a Practice Pattern RR=0.96 per 15 minutes, p=0.03 *Adjusted 2-stage model (instrumental variable) p=0.04 p=0.26 Ref. RR Mortality*
21
Treatment Time and Mortality: Summary Patients treated with longer dialysis sessions have lower mortality risk at the same Kt/V (Saran et al 2008) Patients treated in dialysis facilities that use on average longer treatment times have lower mortality (this analysis focuses on the practice and minimizes bias due to patient health status) The agreement of these results enhances the level of evidence
22
Predialysis Systolic Blood Pressure and Mortality Risk A New Analytical Approach Using Patient Exposure to Different Practices
23
N of facilities = 150 84 62 Median Facility Pre-dialysis SBP Japan North America EU & ANZ — Prevalent HD Patients — Facility Median Pre-dialysis SBP, by Region Substantial Variation Between Facilities and Regions *Based on initial prevalent cross section patients (n=8000) with ESRD >3 months in 296 facilities in DOPPS III (2005-2008). SBP=systolic blood pressure 25 th = 131 mmHg 75 th = 145 mmHg 25 th = 142 mmHg 75 th = 153 mmHg 25 th = 147 mmHg 75 th = 161 mmHg
24
<100 100-110 110-120 120-130 130-140 140-150 150-160 160-170 170-180 ≥180 % of Pts in Facilities Pre-Dialysis SBP (mmHg) * 22,559 initial prevalent cross section patients with ESRD duration > 180 days from 919 facilities in DOPPS I, II, III Within-Facility Distribution of Pre-Dialysis SBP* Substantial Variation Across a Wide BP Range
25
1.03 0.97 1.01 0.99 1.00 1.15 0.89 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 Pre-HD Systolic BP and All-Cause Mortality RR of death § Patient Level BP 110 120 130 140 150 160 170 180 Pt achieved pre-dialysis SBP (mmHg) Ref Facility Level BP RR for an additional 10% of patients compared to the ref category RR of death § 110 120 130 140 150 160 170 180 *P<0.05 * Pre-dialysis SBP Group (mmHg) Ref * * * § 21,388 prevalent HD, 919 facilities Excludes patients with SBP <110 mm Hg. Cox models adjusted for age, gender, black race, BMI, vintage, study phase, hemoglobin, s. albumin, phosphorus, creatinine, ferritin, PTH, intra-dialysis weight loss, treatment time, catheter use, 13 comorbidities, stratified by country and accounted for facility clustering. Facility level model also adjusted for facility mean levels of intra-dialysis weight loss, dialysate sodium, and treatment time (min), % of catheter use and % pts in albumin, Hgb, Kt/V, and phosphate guidelines. No meaningful change with the addition of anti- hypertensive medications to the models, or with the addition of pts with SBP<110. * *
26
Facility Predialysis Systolic BP and All-Cause Mortality Pre-dialysis SBP (mmHg) Facility Level Mortality RR per an additional 10% of patients by category compared to the reference Ref * * RR of death 110-129 130-160 >160 * p <0.05
27
Predialysis Blood Pressure Levels and Survival: Summary Optimal target BP has been difficult to identify, because BP is influenced by health status Facility-based analyses provide insights by minimizing bias due to patient health status, and by taking advantage of the large between-facility variation in BP as a likely reflection of practices or MD opinion Our data show that: –Patients treated at facilities where more patients have low pre-dialysis SBP (110-130 mmHg) have higher mortality risk –Patients treated at facilities where more patients have high pre-dialysis SBP (>160 mmHg) have higher mortality risk
28
Predialysis Blood Pressure Levels and Survival: Conclusion These facility-level findings suggest that both higher predialysis SBP (>160 mmHg) and lower SBP (<130 mmHg) are associated with elevated mortality risk The present results are not consistent with KDOQI Guidelines (SBP <140 mmHg) A clinical trial is needed to identify optimal predialysis SBP goals
29
Serum Phosphorus and Mortality Risk Patient-based analyses and Practice-based analyses
30
Practices of Better Control of High Phosphorus and Mortality Risk BACKGROUND: Patient level analyses showing higher mortality in patients with high P levels may be confounded, if sicker patients have higher P levels The new KDIGO Guidelines recommend control “toward normal P levels” since randomized trials are lacking Since randomization to poor P control is not feasible, can we make observational studies more informative?
31
Mortality Risk by Phosphorus Categories Patient-Level Analyses, Among Patients on HD > 180 days Cardiovascular All-cause Cox models used all DOPPS (n=25,529) and adjusted for age, sex, race, BMI, years on ESRD, 13 comorbid conditions, facility clustering. Hazard ratios and 95% confidence intervals (whiskers) for all- cause (events n=5,857) and cardiovascular mortality (n events=1,930) Tentori et al. AJKD 2008
32
Variation in Facility-Level Serum Phosphorus Facility % of Patients N=899 facilities Serum Phosphorus (mg/dl) The % of patients having a serum PO 4 of >7 mg/dl varies from 3% in some facilities to 40% of patients in other facilities.
33
Facility-Level Serum Phosphorus versus All-Cause and CV Mortality Risks Among Patients on HD > 180 days Hazard ratios and 95% confidence intervals (whiskers) for all-cause (events n=5,857) and cardiovascular mortality (events n=1,930). Models (n=20,561) were stratified by study phase and region and adjusted for facility clustering effect; baseline patient age, sex, race, BMI, time on ESRD, 13 comorbid conditions, hemoglobin, albumin, normalized protein catabolic rate, single-pool Kt/V, prior parathyroidectomy, and vitamin D prescription; the percentage of patients at a facility with serum calcium 10 mg/dL; and the percentage of patients at a facility with serum PTH 600 pg/mL. Tentori et al AJKD 2008 Ref +95% C.I.
34
Can we use Principles of Randomization in Observational Studies? If patients are assigned randomly to facilities: Yes Instrumental variables may reduce treatment by indication bias This is useful when large differences in practice are observed In DOPPS, we use facility-level treatment variables as instrumental variables
35
Facility-Level Treatment Variables Rationale in DOPPS Design Patients usually select dialysis facilities by factors independent of their own medical condition – e.g. by proximity to home Average treatment patterns differ substantially among facilities, in part due to provider opinion or preferences
36
Facility-Level Treatment Variables Since variations in treatment preferences are likely “random” with respect to medical condition, this provides a “natural experiment” with advantages similar to randomization in a clinical trial Randomization provides balance across both measured and unmeasured confoundersRandomization provides balance across both measured and unmeasured confounders
37
Facility-Level Treatment Variables as Instrumental Variables: Caveats Other treatment practices may vary together with the treatment of interest –Action: We adjust also for other treatment practices Unmeasured treatment practices may be confounders –Action: We measure many practices
38
Impact of Dialysis Prescriptions and Practices on Outcomes: Summary The DOPPS approach has allowed identification of opportunities to improve practices and outcomes, e.g. –Treatment time (>4 hours thrice weekly) –Systolic Blood pressure (130-160 mmHg pre-dialysis) –Phosphorus (Avoid PO 4 >6 mg/dl) International outcome differences are confirmed and the US-Euro difference is largely explained by case-mix and vascular access: Need to improve vascular access The instrumental variable approach is useful when based on large differences in actual clinical practice
39
Acknowledgements Thanks to participating DOPPS facilities for their data submission and dedication, and to patients for completing questionnaires DOPPS is supported by scientific research grants without restrictions on publications from –Amgen (1996-2011) –Kyowa Hakko Kirin (1999-2011 in Japan) –Genzyme (2009-2011)
41
Tightness of Hgb Control* and Mortality Risk Among Facility Patients * measured as facility standard deviation of Hgb levels
42
Facility with Std Dev = 1.0 g/dl Some facilities may have larger variation (standard deviation) in patient hemoglobin levels. This may be due to: (1) greater comorbidity and variation in ESA-responsiveness among patients in some facilities (2) differences in facility practices that impact anemia control Facility Hgb Standard Deviation Patients (%) Facility with Std Dev = 3.0 g/dl (measure of “tightness of Hgb control” among facility patients)
43
Facilities (%) Facility Standard Deviation (g/dl) for Hgb Levels among Facility Patients Variation in Facility Hemoglobin Standard Deviation n=921 facilities, DOPPS I, II, and III; Facility Hgb Std Dev based upon a prevalent cross-section of study patients in a facility at start of each DOPPS phase on dialysis > 180 days The mean facility Hgb level did not significantly correlate with the facility Std Deviation of Hgb
44
Facility Std Deviation in Hemoglobin Levels and Mortality Risk Relative Risk of death Ref Adjusted for age, gender, black race, years with ESRD, body mass index, 14 comorbidity classes and facility mean Hgb level; stratified by country and phase; accounted for facility clustering effects; n=23,245, DOPPS 1 to 3. Facility hgb std dev based upon facility prevalent cross-section, pts on dialysis > 180 days Facility Std Dev for Hgb Levels p=0.10 RR= 1.10 per 0.5 unit higher Std Dev (p=0.001) Adjusted for Facility Mean Hgb Levels p=0.0003p=0.003
45
Facility Std Dev in Hemoglobin Levels and Mortality Risk Relative Risk of death Ref Adjusted for age, gender, black race, years with ESRD, body mass index, 14 comorbidity classes and facility mean Hgb level and the facility practice of treatment time, Kt/V, catheter use, serum Ca and PO 4, and mean ESA dose; stratified by country and phase; accounted for facility clustering effects; n=23,245, DOPPS 1 to 3. Facility hgb std dev based upon facility prevalent cross-section, pts on dialysis > 180 days Facility Std Dev (g/dl) for Hgb Levels p=0.23 RR= 1.11 per 0.5 unit higher Std Dev (p=0.002) Adjusted for Facility Mean Hgb Level plus adjusted for 6 other facility practices p=0.002p=0.0008
46
Average Std Deviation in Hemoglobin by Country & DOPPS Phase* *Baseline prevalent cross-section of patients on dialysis > 180 days for each country & phase. Restricted to facility with at least 12 observations Japan ANZ Sweden US France Germany Italy UK Spain Belgium Canada UK Japan Ave of Facility Std Dev, g/dl
47
Practices Associated with Tighter Hgb Control at Facility Level Having a narrower Hgb target range Adjusting ESA dose more often (at least monthly) Checking Hgb levels more often (at least weekly) Prescribing ESAs for more patients (higher %) Giving ESA i.v. rather than subcutaneously
48
Tighter Hgb Control at Facility Level* Summary Tighter Control of Hemoglobin at the Facility Level is associated with lower adjusted mortality at the facility is associated with certain practice patterns appears to be feasible according to the observed improvements over time in most countries * i.e. smaller standard deviation of Hgb across patients
50
Adjusted Relative Risk of Mortality by Continent 1997-2001 Analyses performed on DOPPS I data, collected from 1997-2001. Adjusted for case-mix. a P-value for comparison with Japan. b Adjusted for demographics and 15 classes of comorbidity. US versus Europe: Crude RR=1.71, adjusted RR=1.33 (p<0.0001). NOTE: Japan data under represents hospital units. Including such patients in DOPPS II gives smaller RRs. Japan has very low transplantation rates Goodkin DA et al. JASN 14(12):3270-3277, 2003
51
Distribution of Facility Mean TT, by Region and Phase n = 546 facilities TT=Treatment Time * * * * # * p <0.05 vs US of same phase # p <0.05 vs US DOPPS I Saran et al. KI 69: 1222-8, 2006
52
Hospitalization Risk by Facility Catheter Use Median Facility = 18% of Patients Use Catheters Relative Risk (95% CI) Infection-Related Hospitalization Access-Related Hospitalization RR=1.79 (p<0.0001) per 20% Higher Facility Catheter Use RR=1.33 (p=0.008) per 20% Higher Facility Catheter Use 1.07 1.03 1.28 1.56 1.24 1.18 1.64 2.56 Facility (%) Reference = AVF; adjusted for case mix and AV Graft use
53
Hospitalization Risk by Facility Graft Use Median Facility = 32% of Patients Use Grafts Relative Risk (95% CI) Infection-Related Hospitalization Access-Related Hospitalization RR=1.29 (p<0.0001) per 20% Higher Facility Graft Use RR=1.11 (p=0.008) per 20% Higher Facility Graft Use 1.30 1.28 1.33 1.51 1.38 1.61 1.75 2.32 Facility (%) Reference = AVF; adjusted for case mix and Catheter use
54
Facility Target Pre-Dialysis Systolic BP by Medical Director Survey By Region * DOPPS III Medical Director Survey. N = 236 facilities SBP Target (mmHg) N of facilities 119 57 60 25 th = 140 mmHg 75 th = 160 mmHg Japan North America EU & ANZ
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.