The American Journal of Managed Care August 2009
Cost Comparison of Peritoneal Dialysis Versus Hemodialysis in End-Stage Renal Disease
Patients with end-stage renal disease who began peritoneal dialysis had lower 1-year hospitalization rates and lower total healthcare costs than those who began therapy with hemodialysis.
Objective: To compare healthcare utilization and costs in patients with end-stage renal disease (ESRD) beginning peritoneal dialysis (PD) or hemodialysis (HD).
Study Design: Retrospective cohort study.
Methods: Using a US health insurance database, we identified all patients with ESRD who began dialysis between January 1, 2004, and December 31, 2006. Patients were designated as PD patients or as HD patients based on first-noted treatment. Patients with less than 6 months of pretreatment data and those with less than 12 months of data following initiation of dialysis (“pretreatment” and “follow-up,” respectively) were dropped from the study sample. The PD patients were matched to HD patients using propensity scoring to control for differences in pretreatment characteristics. Healthcare utilization and costs were then compared over 12 months between propensity-matched PD patients and HD patients using paired t tests and Wilcoxon signed rank tests for continuous variables and using Bowker and McNemar tests for categorical variables, as appropriate.
Results: A total of 463 patients met all study entrance criteria; 56 (12%) began treatment with PD, and 407 (88%) began treatment with HD. Fifty PD patients could be propensity matched to an equal number of HD patients. The HD patients were more than twice as likely as matched PD patients to be hospitalized over the subsequent 12 months (hazard ratio, 2.17; 95% confidence interval, 1.34-3.51; P <.01). Their median healthcare costs over the 12-month follow-up period were $43,510 higher ($173,507 vs $129,997 for PD patients, P = .03).
Conclusions: Among patients with ESRD, PD patients are less likely than HD patients to be hospitalized in the year following initiation of dialysis. They also have significantly lower total healthcare costs.
(Am J Manag Care. 2009;15(8):509-518)
Using a large US healthcare claims database, we compared healthcare utilization and costs in patients beginning renal replacement therapy with peritoneal dialysis (PD) versus hemodialysis (HD), matching PD patients to HD patients based on pretreatment characteristics.
- The PD patients were significantly less likely to be hospitalized during the year following initiation of dialysis.
- The PD patients also had significantly lower total healthcare costs over this period.
In the United States, the cost of dialysis is largely borne by the Medicare ESRD system, which accepts all patients previously enrolled in Medicare on initiation of dialysis (principally, persons ≥65 years) and those otherwise not eligible for Medicare benefits after they have received a minimum of 3 months of dialysis (for these latter patients, there is an additional 30-month “coordination of benefits” period during which Medicare remains the secondary payer, while the private insurer is the primary payer).4 Persons 65 years or older who are still working or who have a spouse who is still working also may have their costs borne (in part or in full) by private health insurers. It has been estimated that 25% of all patients with ESRD beginning HD and 37% of all such patients beginning PD are privately insured.5
There is a wealth of information about healthcare utilization and costs among patients with ESRD who are insured through the Medicare program. Comparatively little is known about the use and cost of healthcare services among patients with ESRD who are privately insured and, in particular, those beginning treatment with PD versus HD.
Data were obtained from the Phar-Metrics Patient-Centric Database, which is composed of facility, profes-sional service, and retail (ie, outpatient) pharmacy claims from more than 85 US health plans (PharMetrics, Watertown, MA). The plans provide healthcare coverage to approximately 14 million persons annually throughout the United States (35% in the Midwest, 21% in the Northeast, 31% in the South, and 13% in the West). All patient identifiers in the database are fully encrypted, and the database is fully compliant with the Health Insurance Portability and Accountability Act of 1996.
Information available for each facility and professional service claim includes the date and place of service, diagnoses (in International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] format), procedures (in ICD-9-CM [selected plans only] and Health Care Financing Administration Common Procedural Coding System formats), provider specialty, and charged and paid amounts. Data available for each retail pharmacy claim include the drug dispensed (in National Drug Code format), dispensing date, and quantity dispensed and number of days of therapy supplied (selected plans only). All claims include a charged amount; the database also provides the paid amounts (ie, total reimbursed, including patient deductible, copayment, and coinsurance).
Selected demographic and eligibility information is also available, including age, sex, geographic region, coverage type, and the dates of insurance coverage. All patient-level data can be arrayed chronologically to provide a detailed longitudinal profile of all medical and pharmacy services used by each insured person. Because this study was retrospective in nature, used completely anonymized data, and did not involve patient contact, institutional review board approval was neither required nor sought.
Using the PharMetrics database, we identified all patients with 1 or more medical encounters for PD or HD between January 1, 2004, and December 31, 2006 (“study period”), irrespective of whether they had any claims with a diagnosis of renal failure (ICD-9-CM diagnosis codes 403.X1, 404.X2, 404.X3, 585, 585.X, and 586) (additional criteria, listed herein, were used to exclude patients receiving dialysis for reasons other than ESRD). For each such patient, we then identified the first-noted claim for dialysis (either PD or HD) during the study period; the date of this claim was designated the “index date,” and patients were stratified into 2 groups (“PD patients” or “HD patients”) based on the treatment received on this date. All patients were required to be continuously enrolled for 6 months before their index date (“pretreatment”) and for 12 months after this
date (“follow-up”). Identification of patients who received PD versus HD was based on algorithms developed by us (eAppendix A and eAppendix B available at www.ajmc.com). Patients with claims for both PD and HD on their index date were excluded, as were patients enrolled in a Medicaid program and those 65 years or older who were enrolled in Medicare supplemental or capitated plans (because of incomplete claims histories). Additional exclusion criteria were taken from prior studies6-8 that identified patients receiving dialysis based on electronic claims databases and included the following: (1) any claims encounters with dialysis-related codes (ie, diagnostic, procedural, or equipment) during the pretreatment period, (2) less than 3 months of continuous enrollment following the index date, (3) evidence of initiation of dialysis for reasons other than ESRD (eg, because of trauma), and (4) patients who underwent renal transplantation during the first month of follow-up.
We examined the prevalence of several (medically attended) comorbidities among PD patients and HD patients, including the following: (1) diabetes mellitus (ICD-9-CM diagnosis code 250.XX or receipt of a-glucosidase inhibitors, insulin, metformin hydrochloride, nonsulfonylurea insulin secretagogues, sulfonylurea, or thiazolidinedione); (2) coronary artery disease (codes 410.XX-414.XX); (3) congestive heart failure (code 428.XX); (4) anemia (codes 280.XX-285. XX or receipt of darbepoetin alfa or epoetin alfa); (5) renal osteodystrophy (code 588.0); (6) sleep disorders (codes 307.4X, 780.5X, and V69.4); (7) amyloidosis (code 277.3); and (8) hypertension (codes 401.XX-405.XX, 459.10, 459.30, 459.31, 459.32, 459.33, and 459.39 or receipt of antihypertensives). Patients were deemed to have these conditions if they had 2 or more outpatient claims (medical or pharmacy) on different days or 1 or more inpatient claims (medical only) meeting the aforementioned criteria.
To control for potentially important differences in pretreatment characteristics between PD patients and HD patients, we matched the latter to the former using techniques of propensity scoring.9-11 Briefly, multivariate logistic regression analysis was used to generate a probability (“propensity score”) that each subject was a PD patient; covariates entered into the model included age, sex, comorbidities, and pretreatment healthcare utilization and costs. Once a propensity score was generated for each patient in the study sample, HD patients were matched to PD patients in a stepwise fashion that minimized the absolute difference in propensity scores for each match. Once a pair was matched, both subjects were removed from the pool of potential pairs, and the process was repeated until all possible pairs for which the absolute difference in propensity scores was less than 0.01 were matched (ie, caliper matching was used). All analyses were then undertaken on the propensitymatched sample.
Once the matched sample was created, we examined the use and cost of healthcare services during the 12-month follow-up period, including the following: (1) prescription medications, (2) physician office visits, (3) other outpatient visits, (4) emergency department (ED) visits, and (5) hospitalizations. The use of healthcare services was examined in terms of the percentage of patients receiving each service, as well as the number of times each service was rendered; for hospitalizations, length of stay also was examined. The total reimbursed amount (ie, the amount paid by the insurer plus the amount of patient liability [eg, copayment and deductible]) was used as a proxy for cost.
Kaplan-Meier methods were used to examine the incidence of and the time to hospitalization during follow-up among patients in the matched sample. Cox proportional hazards models were used to identify potential predictors of hospitalization; predictors in these models included age, sex, pretreatment comorbidities, pretreatment healthcare utilization and costs, and initial dialysis modality (ie, PD vs HD).
For variables that were not approximately normally distributed, 95% confidence intervals (CIs) were constructed by drawing 1000 samples (with replacement) from the source population, calculating the values for the relevant variables within each sample, and taking the 2.5 and 97.5 percentile values (ie, the bootstrap method).12 The statistical significance of differences between propensity-matched PD patients and HD patients was ascertained using paired t tests (age) and Wilcoxon signed rank tests (all others) for continuous variables and using Bowker and McNemar tests for categorical variables, as appropriate. All analyses were conducted using PC SAS version 9.1 (SAS Institute, Cary, NC).
We identified 56 PD patients and 407 HD patients who met all study entrance criteria. On average, PD patients were younger than those receiving HD, and fewer of them had a history of congestive heart failure (Table 1). The total healthcare costs during pretreatment were higher (albeit not significantly) among PD patients than among HD patients. Fifty PD patients were matched to an equal number of HD patients; 6 PD patients could not be matched. After matching, PD patients and HD patients were similar in terms of the pretreatment characteristics we considered (Table 2).
On a per-patient basis, those initiating dialysis with HD averaged 20 more outpatient visits over 12 months compared with matched patients in the PD group: the mean (95% CI) was 68.4 (57.3-82.1) versus 48.4 (41.0-57.1), and the corresponding median (interquartile range [IQR]) was 60 (38-90) versus 43 (29-70) (P = .01 for both) (Table 3). The HD patients also had nominally more ED visits (mean [95% CI], 3.3 [2.1-5.0] vs 2.3 [1.3-3.5] for PD; P = .28). Over the 12-month period of follow-up, more HD patients were hospitalized (80% vs 50% for PD, P <.01).
1. National Institute of Diabetes and Digestive and Kidney Diseases Web site. Kidney failure: choosing a treatment that’s right for you. http://kidney.niddk.nih.gov/kudiseases/pubs/pdf/choosingtreatment.pdf. Accessed September 2007.
2. Saxena R. Peritoneal dialysis: a viable renal replacement therapy option [published correction appears in Am J Med Sci. 2005;330(3):110]. Am J Med Sci. 2005;330(1):36-47.
3. National Institute of Diabetes and Digestive and Kidney Diseases Web site. Treatment methods for kidney failure: peritoneal dialysis. http://kidney.niddk.nih.gov/kudiseases/pubs/pdf/peritoneal.pdf. Accessed September 2007.
4. US Renal Data System. USRDS 2007 Annual Data Report: Atlas of End-Stage Renal Disease in the United States. Bethesda, MD: National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health; 2007.
5. Hirth RA. The organization and financing of kidney dialysis and transplant care in the United States of America. Int J Health Care Finance Econ. 2007;7(4):301-318.
6. Robbins JD, Kim JJ, Zdon G, Chan WW, Jones J. Resource use and patient care associated with chronic kidney disease in a managed care setting. J Manag Care Pharm. 2003;9(3):238-247.
7. Shih YC. Effect of insurance on prescription drug use by ESRD beneficiaries. Health Care Financ Rev. 1999;20(3):39-54.
8. Winkelmayer WC, Glynn RJ, Mittleman MA, Levin R, Pliskin JS, Avorn J. Comparing mortality of elderly patients on hemodialysis versus peritoneal dialysis: a propensity score approach. J Am Soc Nephrol. 2002;13(9):2353-2362.
9. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med. 1997;127(8, pt 2):757-763.
10. Rubin DB, Thomas N. Matching using propensity scores: relating theory to practice. Biometrics. 1996;52(1):249-264.
11. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516-524.
12. Efron B, Tibshirani R. Bootstrap methods for standard errors, confidence intervals, and other measures of statistical accuracy. Stat Sci. 1986;1:54-57.
13. Mehrotra R, Kermah D, Fried L, et al. Chronic peritoneal dialysis in the United States: declining utilization despite improving outcomes. J Am Soc Nephrol. 2007;18(10):2781-2788.
14. US Renal Data System. USRDS 2006 Annual Data Report: Atlas of End Stage Renal Disease in the United States. Bethesda, MD: National Institute of Diabetes and Digestive and Kidney Disease, National Institutes of Health; 2006.
15. Shih YC, Guo A, Just PM, Mujais S. Impact of initial dialysis modality and modality switches on Medicare expenditures on end-stage renal disease patients. Kidney Int. 2005;68(1):319-329.
16. Bruns FJ, Seddon P, Saul M, Zeidel ML. The cost of caring for endstage kidney disease patients: an analysis based on hospital financial transaction records. J Am Soc Nephrol. 1998;9(5):884-890.
17. Yu AW, Chau KF, Ho YW, Li PK. Development of the “peritoneal dialysis first” model in Hong Kong. Perit Dial Int. 2007;27(suppl 2):S53-S55.
18. Khawar O, Kalantar-Zadeh K, Lo WK, Johnson D, Mehrotra R. Is the declining use of long-term peritoneal dialysis justified by outcome data? Clin J Am Soc Nephrol. 2007;2(6):1317-1328.
19. Sennfält K, Magnusson M, Carlsson P. Comparison of hemodialysis and peritoneal dialysis: a cost-utility analysis. Perit Dial Int. 2002;22(1):39-47.
20. Lee H, Manns B, Taub K, et al. Cost analysis of ongoing care of patients with end-stage renal disease: the impact of dialysis modality and dialysis access. Am J Kidney Dis. 2002;40(3):611-622.
21. Erek E, Sever MS, Akoglu E, et al. Cost of renal replacement in Turkey. Nephrology (Carlton). 2004;9(1):33-38.
22. Hooi LS, Lim TO, Goh A, et al. Economic evaluation of centre haemodialysis and continuous ambulatory peritoneal dialysis in Ministry of Health hospitals, Malaysia. Nephrology (Carlton). 2005;10(1):25-32.
23. Miskulin DC, Meyer KB, Athienites NV, et al. Comorbidity and other factors associated with modality selection in incident dialysis patients: the CHOICE study: Choices for Healthy Outcomes in Caring for End- Stage Renal Disease. Am J Kidney Dis. 2002;39(2):324-336.
24. Mehrotra R. The John F. Maher Award Recipient Lecture 2006. The continuum of chronic kidney disease and end-stage renal disease: challenges and opportunities for chronic peritoneal dialysis in the United States. Perit Dial Int. 2007;27(2):125-130.
25. Steinbrook R. Medicare and erythropoietin. N Engl J Med. 2007;356(1):4-6.
26. Shahab I, Khanna R, Nolph KD. Peritoneal dialysis or hemodialysis? A dilemma for the nephrologist. Adv Perit Dial. 2006;22:180-185.
27. Blake PG. Peritoneal dialysis in the USA. Perit Dial Int. 2006;26(4):416-418.
28. Mehrotra R. Changing patterns of peritoneal dialysis utilization in the United States. Perit Dial Int. 2007;27(suppl 2):S51-S52.
29. Jaar BG, Coresh J, Plantinga LC, et al. Comparing the risk for death with peritoneal dialysis and hemodialysis in a national cohort of patients with chronic kidney disease. Ann Intern Med. 2005;143(3):174-183.
30. Rubin HR, Fink NE, Plantinga LC, Sadler JH, Kliger AS, Powe NR. Patient ratings of dialysis care with peritoneal dialysis vs hemodialysis. JAMA. 2004;291(6):697-703.
31. Kutner NG, Zhang R, Barnhart H, Collins AJ. Health status and quality of life reported by incident patients after 1 year on haemodialysis or peritoneal dialysis. Nephrol Dial Transplant. 2005;20(10):2159-2167.
32. Wasserfallen JB, Moinat M, Halabi G, et al. Satisfaction of patients on chronic haemodialysis and peritoneal dialysis. Swiss Med Wkly. 2006;136(13-14):210-217.
33. London R, Solis A, Goldberg GA, Wade S, Ryu S. Health care resource utilization and the impact of anemia management in patients with chronic kidney disease. Am J Kidney Dis. 2002;40(3):539-548.