Comparison of Mortality Between Private For-Profit and Private Not-For-Profit Hemodialysis Centers:  A Systematic Review and Meta-analysis FREE

More than 20 years ago, Relman and Rennie¹ suggested that profit status may affect dialysis practices. Advocates of private for-profit health care delivery argue that for-profit providers can deliver high-quality care more efficiently than not-for-profit providers.² However, fears that for-profit facilities compromise quality of care to maintain shareholder returns^3- 4 have precipitated a heated debate about whether the profit status of hemodialysis facilities influences patient mortality.^5- 7

Separating issues of funding (ie, who will pay for the health care) and delivery (ie, who owns and administers the institutions providing care) helps to inform debates about health care systems. Both funding and delivery can occur through private or public means. Private funding or delivery can be for profit or not for profit. Private for-profit and private not-for-profit insurance companies use insurance premiums to pay for the health care of the patients they insure. Delivery of care can occur at private for-profit institutions that are owned by investors or at private not-for-profit institutions that are owned by communities, religious organizations, or philanthropic groups. Public funding indicates payment by the government, and public delivery indicates that a health care institution is owned and administered by the government. All public health care institutions are not for profit.

Since 1973, the US government has, through Medicare, funded the care of patients receiving dialysis.⁸ Currently in the United States, private for-profit (approximately 75%) and private not-for-profit (approximately 20%) facilities provide most hemodialysis care.⁹ Public institutions deliver the small remaining proportion of hemodialysis care.

Our study addresses issues of health care delivery, rather than health care funding. Accurate understanding of the impact of alternative health care delivery systems requires a systematic, comprehensive, and unbiased accumulation and summary of the available evidence. We therefore undertook a systematic review and meta-analysis to address the following question: what is the relative impact of private for-profit vs private not-for-profit delivery of hemodialysis care on patient mortality?

This systematic review is part of a series of reviews we are undertaking comparing health outcomes, appropriateness and quality of care, and costs in private for-profit and private not-for-profit health care institutions. We have previously reported on hospital mortality rates in private for-profit and private not-for-profit hospitals.¹⁰Figure 1 outlines the methodological steps for this systematic review. The Hamilton Health Sciences Research Ethics Board in Hamilton, Ontario, approved this study protocol.

We undertook 6 strategies to identify potentially eligible studies. Our search included 11 electronic bibliographic databases, our own files, and contact with experts. We further searched PubMed using the "related articles" feature, SciSearch, and the reference lists of all studies that fulfilled our eligibility criteria.

A medical librarian used all the studies of which we were initially aware to identify Medical Subject Heading terms and key words for the search. In each database, the librarian undertook an iterative process to refine the search strategy through testing of several search terms and incorporation of new search terms as new relevant citations were identified. The search included the following databases: EMBASE (1980-2001), MEDLINE (1966-2001), HealthSTAR (1975-2001), CINAHL (1982-2001), BIOETHICSLINE (1973-2000), Wilson Business Abstracts (1997-2001), EconLit (1969-2001), Cochrane Library (2001, issue 3), Dissertation Abstracts Ondisc (1861-2001), ABI (Abstracted Business Information)/Inform (1970-2001), and NTIS (National Technical Information Service) (1964-2002). Complete descriptions of the database search strategies are available from the authors.

Eligibility Criteria. We included published and unpublished observational studies and randomized controlled trials that directly compared the mortality rates of hemodialysis patients in private for-profit and private not-for-profit dialysis centers. Eligible observational studies had to provide adjusted mortality rates or mortality rates from matched cases.

Screening Process. Our 6 search strategies identified 7045 unique citations. Teams consisting of 2 individuals independently screened the titles and abstracts of each citation and identified all citations for full review where there was any possibility that the study contained a comparison in which we were interested. This screening process yielded 779 full-text publications identified by one or both screeners for full review (Figure 1).

Assessment of Study Eligibility. We masked the results (ie, we obscured them from the tables and text using a black marker) of all publications selected for full review. To determine eligibility, teams of 2 reviewers independently evaluated masked articles that they had not assessed during the screening process. When disagreements occurred, the reviewers discussed the reasoning for their decisions, and if one reviewer realized she or he had made an error then the process was complete. This occurred in all cases of disagreement, and, therefore, plans for an independent review of cases of substantive disagreement proved unnecessary. Our agreement on study eligibility was excellent (κ = 0.91; 95% confidence interval [CI], 0.85-0.97).

Teams of 2 reviewers independently abstracted the following data from all studies meeting eligibility criteria: sampling method, source of data (eg, administrative database, patient chart), type of dialysis facilities evaluated (ie, freestanding, hospital-based, or both), dates when data collection was initiated and completed, duration of patient follow-up, number of private for-profit and private not-for-profit dialysis facilities and patients evaluated, mortality, potential confounders adjusted for in the analyses, and the primary intent of the study (ie, to compare outcomes in private for-profit vs private not-for-profit dialysis facilities or other). Reviewers resolved disagreements by consensus using the process described earlier. Our overall agreement was 98% for data abstraction. We successfully contacted all authors to obtain missing data, and in all instances authors were able to provide all the information we requested.

Prior to undertaking this study, we considered it appropriate to adjust for the following factors: age, sex, race, income, education, cause of end-stage renal disease (eg, diabetes, hypertension, chronic glomerulonephritis), comorbid conditions (eg, cancer, congestive heart failure, coronary artery disease), number of years receiving dialysis, market share of the dialysis facility (ie, competition), and whether the dialysis facility was part of a multinational chain corporation. We considered it ideal if measurement of these factors, particularly comorbid conditions that could result from suboptimal dialysis (eg, hypertension), was undertaken when patients initiated dialysis treatments. We considered analyses to be overadjusted if the investigators adjusted for variables that were under the control of the facility administrators, could be influenced by profit status, and could possibly affect mortality. These variables included hemoglobin or hematocrit levels after June 1989 (when Medicare authorized reimbursement for erythropoietin¹¹), duration of dialysis treatments, staffing levels (ie, number of staff per dialysis treatment), skill level of the dialysis staff (eg, aids/technicians vs registered nurses), quality of equipment, and reuse of dialyzers.

Our quality assessment of studies included whether the study appropriately adjusted for any of the factors listed herein and avoided overadjustment as described, as well as identification of the source of data (data from individual chart reviews was considered superior to data from administrative databases).

Prior to the analysis, we specified several hypotheses to explain variability (ie, heterogeneity) in the direction and magnitude of effect among studies. We hypothesized that the effect size may differ based on whether the primary intent of the study was to compare outcomes in private for-profit vs private not-for-profit dialysis facilities; whether the study was restricted to dialysis facilities that were freestanding, hospital-based, or included both types of institutions; whether there was adjustment for market competition; whether there was overadjustment for the variables described; and whether we had to compute a corrected relative risk (RR) using the methods proposed by Zhang and Yu (Box).¹²

Only 1 of the studies included in our systematic review directly provided an RR.²² For the other studies, we had to convert some other measure of effect size to RR.

For 3 studies,^16,20- 21 we generated estimates of 2 × 2 (profit status × mortality) tables and used these to estimate ln(RR) and its variance. Plough et al¹⁶ reported information that enabled us to generate five 2 × 2 tables, 1 for each severity group. We used these tables to calculate an RR of 1-year mortality (private for-profit [PFP] relative to private not-for-profit [PNFP]) for each severity group; we then calculated a pooled estimate for the study using a meta-analytic method.¹³ Irvin²⁰ reported a risk difference of 5.86%. McClellan et al²¹ provided a linear regression analysis from which we were able to infer a risk difference of 0.73%. For these studies, we generated patient-level 2 × 2 tables consistent with those risk differences.

Two of our RR estimates are derived from analyses that were not reported in the original studies. We report these estimates for McClellan et al²¹ and Port et al²² based on models with public hemodialysis facilities excluded. In the latter case,²² profit status also had to be added to the model.

For each study, we computed the RR of mortality in private for-profit dialysis centers relative to private not-for-profit dialysis centers. Two independent reviewers blinded to the study results selected studies to pool in our primary pooled analysis on the basis of the following criteria: studies in which patients were likely to be included twice had to be separated in their time of enrollment by at least 2 years. For example, if one study completed enrollment on January 1, 1990, the next study could not begin enrolling before 1992. The purpose of this criterion was to limit the extent to which 2 studies used results from the same patients. If studies did have overlapping enrollment as described, the reviewers chose the largest study that adjusted for age, sex, race, and comorbidity (including diabetes). The reviewers independently agreed on which studies to pool in the primary meta-analysis.

We pooled these RRs using a random-effects model¹³ and tested for heterogeneity using a χ² test. Data management and analysis were performed using SPSS software, version 11 (SPSS Inc, Chicago, Ill), and a meta-analysis program written by one of the authors (B.W.). We conducted a visual examination of funnel plots for evidence of publication bias.^14- 15

We identified 7 publications reporting 8 observational studies that met our eligibility criteria.^16- 23 We also identified 12 publications that we believed might be eligible but required further information and/or data from the authors. After successfully contacting all authors, we confirmed that the majority of these studies had grouped private not-for-profit and public dialysis facilities together and that the authors either no longer had the data or could not rerun the analyses excluding the public institutions; these studies were excluded from our systematic review (Table 1).^24- 35

Table 2 presents the study characteristics and Table 3 presents the study methods of the 8 observational studies included in our systematic review. Table 3 lists the variables adjusted for in the analysis of each individual study. We successfully contacted all authors to clarify and obtain missing information. All studies were conducted in the United States and included data from January 1, 1973, through December 31, 1997. Three of the studies assessed only freestanding facilities and the remaining 5 evaluated both freestanding and hospital-based dialysis centers. The 8 studies included more than 500 000 patient-years of data and assessed a median of 1342 facilities per study. The most common duration of patient follow-up was 1 year.

Our quality assessment of studies revealed that all studies appropriately adjusted for many important determinants of mortality (including age, race, and cause of end-stage renal disease), 2 studies had overadjustments (including staffing levels and skill levels),^17,21 and only 1 study obtained data from patient charts.¹⁸

Six studies showed a statistically significant increase in adjusted mortality in for-profit facilities,^{17- 20,22- 23} 1 showed a nonsignificant trend toward increased mortality in for-profit facilities,²¹ and 1 showed a nonsignificant trend toward decreased mortality in for-profit facilities (Figure 2).¹⁶

Our primary meta-analysis pooled 4 studies and demonstrated that private for-profit dialysis centers were associated with an increased risk of death (RR, 1.08; 95% CI, 1.04-1.13; P<.001) that, with the exception of 1 study, was highly consistent across studies (P = .08 for heterogeneity) (Figure 2).^{16- 17,19,22} None of our predefined hypotheses to explain potential sources of heterogeneity demonstrated a statistically significant difference between the subgroup summary estimates. One study by Plough et al¹⁶ completely confounded profit status with whether facilities were freestanding or hospital-based (ie, the investigators compared private for-profit freestanding facilities with private not-for-profit hospital centers). For several reasons (see "Comment"), this very different study design could explain the difference in results across studies. The difference between the estimate of effect by Plough et al (RR, 0.71; 95% CI, 0.49-1.02)¹⁶ and those of the other 3 studies (RR, 1.09; 95% CI, 1.07-1.11)^17,19,22 was statistically significant (P = .02). In the 3 unconfounded studies, the increased RR of 1.09 for mortality associated with care in a for-profit facility was significant (P<.001) and consistent (P = .50 for heterogeneity).^17,19,22

We undertook 2 other sensitivity analyses that we planned a priori. Pooling the results of all 8 studies demonstrated that private for-profit dialysis facilities were associated with an increased risk of death (RR, 1.09; 95% CI, 1.05-1.12; P<.001; P = .004 for heterogeneity). Pooling the 3 studies that had restricted evaluation to freestanding dialysis units also demonstrated a statistically significant higher risk of death in private for-profit dialysis centers (RR, 1.11; 95% CI, 1.02-1.21; P = .02; P = .004 for heterogeneity).^18,20,23 Funnel plots did not suggest publication bias.

Our systematic review identified 8 studies that assessed adjusted mortality rates of hemodialysis patients in private for-profit and private not-for-profit dialysis centers. Six of the studies demonstrated a statistically significant increased risk of death in private for-profit dialysis facilities. Approximately 20% to 25% of US in-center hemodialysis patients die each year,³⁶ and our meta-analysis found a pooled RR of death of 8%, suggesting that private for-profit dialysis facilities may be responsible for a substantial number of excess deaths.

We are unaware of any prior systematic reviews or meta-analyses that have compared mortality rates of hemodialysis patients in private for-profit and private not-for-profit dialysis facilities. We undertook a very broad search to identify studies for our systematic review. We masked study results prior to determining study eligibility. We conducted all review procedures in duplicate and demonstrated a high degree of agreement in our eligibility decisions, data abstraction, and meta-analysis decisions. Our search identified 8 eligible observational studies with large sample sizes. We were successful in confirming and obtaining information from authors. In summary, our systematic review of these observational studies is methodologically rigorous³⁷ and transparently reported.³⁸

Our systematic review has several limitations. We did not identify any randomized controlled trials. However, since it is unlikely that patients will ever be randomized to private for-profit and private not-for-profit health care delivery systems, the large studies in our systematic review represent the strongest feasible study design (ie, observational studies with adjustment for potential confounders) for answering our question. All included studies adjusted for age, race, and cause of end-stage renal disease. The last variable also provided some adjustment for comorbidity (eg, diabetes and hypertension).

Adjustment based on data from administrative databases is limited by the comprehensiveness and quality of the data. Careful data abstraction from individual patient charts represents a superior approach to account for important clinical confounders, particularly related to comorbidity. Only 1 study used data abstracted from individual patient charts; this study demonstrated the largest increase in mortality (RR increase of 18%).¹⁸

Another limitation of our systematic review was that in studies that combined private not-for-profit facilities with public facilities, many authors either did not have the data any longer or could not rerun analyses without data from the public dialysis facilities. However, 2 authors did rerun their analyses to remove the public institutions, and the results strengthened our systematic review.^21- 22

Pooling data from private not-for-profit and public dialysis centers is questionable. Theoretical reasons for potential differences in outcomes between public hospitals and private not-for-profit hospitals that provide dialysis include the public hospitals' funding disadvantage and the possibility that public hospitals use some of the funds they receive for dialysis to help subsidize the care of other patients within their institutions.

For those who consider it appropriate to pool private and public not-for-profit institutions, the results of the majority of studies that combined these 2 types of facilities support the results of our systematic review. Seven of these 13 studies (including 2 very large US Renal Data System [USRDS] studies) demonstrated a statistically significant lower risk of death in not-for-profit dialysis centers compared with private for-profit centers,^30- 33 and 4 demonstrated a trend favoring not-for-profit centers (Table 1).^24,27,32 In contrast, only 2 of these 13 studies demonstrated a statistically significant lower risk of death in patients receiving dialysis care in private for-profit hospitals compared with not-for-profit hospitals,³⁰ and 2 others demonstrated trends in favor of for-profit facilities (Table 1).³⁵ The 1 publication with 2 studies that showed a statistically significant result in favor of for-profit facilities was restricted to hospitals. Furthermore, they were the only 2 studies that restricted their samples to dialysis units with less than 25% of patients with high-efficiency/high-flux dialysis, and included an overadjustment for reuse of dialyzers.³⁰

Although formal tests of heterogeneity may lack power, aside from 1 study by Plough et al,¹⁶ the results of eligible studies are very consistent (Figure 2), strengthening the inferences from this systematic review. The results of Plough et al raise challenges in interpretation because profit status was completely confounded with hospital and freestanding facilities (the authors compared private for-profit freestanding facilities with private not-for-profit hospital-based centers). Hospital-based hemodialysis may be associated with higher mortality than dialysis in freestanding facilities, irrespective of profit status.²⁴

Our systematic review demonstrated an increased risk of death in private for-profit dialysis centers (RR, 1.08; 95% CI, 1.04-1.13). Several studies have demonstrated mechanisms through which this increased risk of death may be occurring. Studies (including 1 in our systematic review) have demonstrated that private for-profit dialysis centers employ fewer personnel per dialysis run and less-highly skilled personnel (ie, more licensed practical nurses, aids, and technicians compared with registered nurses).^17,27,39 Other studies have also demonstrated that patients at private for-profit dialysis facilities have shorter durations of dialysis treatment.^27- 28 Shorter durations of dialysis treatment are associated with higher mortality.^27- 28

In a health care system in which funding is relatively fixed, as with dialysis care in the United States, the private for-profit facilities face a difficult economic challenge. Shareholders expect 10% to 15% returns on their investments³ and taxes may account for 5% to 6% of total expenses.⁴⁰ Private for-profit facilities must generate these profits and pay taxes while endeavoring to provide the same quality care as private not-for-profit centers that are free of these expenses. Given that nurse and technician wages account for approximately 70% of total dialysis costs, it is understandable that private for-profit centers may try to minimize staff and the skill (and, thus, reimbursement) level of staff.^17,27,39

How many deaths might realistically be avoided if private not-for-profit facilities delivered all dialysis care in the United States? There are approximately 208 000 patients receiving in-center hemodialysis in the United States each year, of whom approximately 75% receive their dialysis in for-profit facilities.⁹ Making a conservative estimate of the proportion who die each year, 20%,³⁶ and using our pooled estimate of an 8% (95% CI, 4%-13%) relative increase in mortality in for-profit facilities, one can estimate that were these patients to receive their dialysis in private not-for-profit facilities, approximately 2500 deaths (range, 1200-4000) could be avoided each year.

Many countries face choices about the optimal methods of health care delivery. For instance, Canada is currently undergoing intense debate concerning the relative merits of private for-profit vs private not-for-profit health care delivery. Our results suggest the inadvisability of introducing private for-profit dialysis centers into the Canadian health care system.

A previous systematic review suggested that private for-profit hospitals increase mortality relative to private not-for-profit hospitals.¹⁰ Together with the results of this review, these data provide compelling evidence that profit status can have an important impact on the outcomes of medical care.