Variations and characteristics of quality indicators for maintenance hemodialysis patients: A systematic review

Abstract Aims Several quality indicators (QIs) to improve the quality of practice for hemodialysis patients have been implemented. However, the variations and characteristics of these indicators in terms of their use and feasibility have not been investigated. We conducted a systematic review to evaluate the variations and characteristics of existing QIs for maintenance hemodialysis patients. Methods We conducted a systematic literature search of MEDLINE via PubMed, Scopus, the Cochrane Library, and CINAHL, without date limits, on February 26, 2016. We selected the English‐written articles regarding QIs for patients aged ≥18 years who were on maintenance hemodialysis therapy ≥3 months, and extracted the definition and development process of the reported QIs. We categorized each indicator into one of four types, namely, structure, process, surrogate outcome, and outcome, and assessed the data sources that were necessary to measure it. Results We included 70 articles and identified 101 indicators, and found that most of the consensus processes for selecting indicators were unclear. We also found that most indicators were not process indicators and that the measurement of some indicators required a chart review, which limits their use and feasibility. Conclusions Development of QIs for hemodialysis patients in the future should use a definitive consensus process and consider process‐centered indicators that can be measured automatically using claims data and test results contained in electronic medical records, to improve usability and feasibility.

hemodialysis patients, [10][11][12] the methods by which these QIs were established in these studies are not clear, and we cannot, therefore, be sure that their selection was based on scientifically valid methods.
In setting QIs, many have recognized the usefulness of Donabedian's framework, which defines quality measurement of health care in three parts: structure, process, and outcome. 13 This framework is sometimes expanded into four parts to include a surrogate outcome. 14 As each part is associated with its own advantages and disadvantages, quality can be precisely measured if the meaning of each part differs according to the aim of the measurement initiative. 14 In addition to variations in types of QIs, there are also differences in the process of developing QIs, for example, in the use of guideline-based versus Delphi methods. The Delphi method was originally developed to ensure an anonymous consensus to avoid domination by a few experts; however, even this method has some variations. 15 Moreover, although there are many variations in the components of QIs and in their development process, no systematic review of existing QIs for maintenance hemodialysis patients has yet been conducted, unlike the case of other areas such as palliative care, 16 trauma care, 17 and anesthesia. 18 Here, we conducted a systematic review of QIs for maintenance hemodialysis patients to construct item lists and to identify the pros and cons of existing QIs. Our findings will help improve the future development of QIs.

| MATERIALS AND METHODS
This current systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. 19 The study protocol was not registered in the PROSPERO because some standard methods of the systematic review process (eg, prespecification of the primary outcomes, risk of bias assessment, data synthesis including meta-analyses, or evidence synthesis using GRADE approach), which should be stated through the registration process in PROSPERO, were not required in this study.

| Literature search
We conducted a systematic literature search of MEDLINE via PubMed, Scopus, the Cochrane Library, and CINAHL, without date limits, on February 26, 2016. Our search strategy is shown in Text S1. We checked the references of all potential publications to extract the definition and development process of the reported QIs.

| Study selection
We included and excluded publications according to the following criteria: 1. Only English-written publications were included. 2. Those that described the development process or characteristics of QIs for patients aged ≥18 years who were on maintenance hemodialysis therapy ≥3 months were included. Publications examining patients using special modalities such as nocturnal hemodialysis, home dialysis, and combination therapy with peritoneal dialysis were excluded.
3. Those describing only the QIs that should be achieved on initiation of hemodialysis, such as arteriovenous fistula (AVF) creation during the initiation of hemodialysis, were excluded because these QIs could not be modified during the maintenance hemodialysis phase. These QIs were also excluded from the extracting items for each QI set, which we defined as a set of QIs examined in each included article. 4. Those not describing the rationale behind associations with the QIs were excluded. Those that discussed the rationale, such as that behind the association between the QI and clinical outcomes, but did not cite a reference(s), were included. 5. Those in which the numerators and denominators of the QIs were defined, or could be deduced from the description of the QIs, were included. 6. Those that described indicators with specific goals were included.
For example, the target hemoglobin (Hb) level, such as Hb ≥ 10 g/dL, had to be reported when Hb level was a QI. 7. Those describing QIs for primary care settings were excluded.
Four authors (I.T., S.S., N.K., and T.Y.) were divided into two teams, with I.T. and N.K. in one, and S.S. and Y.T. in the other. Each team examined half of the articles identified by the electronic search strategy described above and checked them according to the inclusion/exclusion criteria.
The two members of each team reviewed each article independently.
Articles that were considered to meet the inclusion criteria were obtained as full articles and independently reassessed for inclusion as described above. In the case of discordance in the selection of an article within one team, one author from the other team assessed its inclusion.

| Data extraction
We used a structured Excel data collection form designed by the authors to independently extract the required data from the included studies. Extracted data included the consensus process used to develop the indicator, references for the indicator, a general description of the items in each QI set, the types of indicators for each item, the data sources used to measure each item, and the clinical practice guidelines supporting each QI. We categorized each indicator into one of four types: structure, process, surrogate outcome, and outcome. We defined structure indicators as hospital or clinical resources such as the number of doctors and nurses. We defined process indicators as those that can only be modified by health care professionals and do not depend on the patient's condition, such as the frequency of blood tests and noninvasive procedures. We defined surrogate outcomes and outcomes as patient conditions, with surrogate outcomes represented by clinical signs such as test results that are associated with outcomes. We also categorized each indicator according to the data sources that were necessary to measure that indicator, such as claims data, test results, and medical chart review. We defined claims data as data such as information on a disease, procedure, or prescription.
As this is a systematic review of the literature, approval by the research ethics committee was not required.  Table 1.

| Variations in the development of QI sets
Most QI sets were developed by experts' consensus based on international guidelines such as the National Kidney Foundation-Dialysis Outcomes Quality Initiative (NKF-DOQI) Clinical Practice Guidelines 23 and the European Best Practice Guidelines (EBPG). 24 In terms of the CPM project, the first 16 QIs were developed based on the NKF-DOQI Clinical Practice Guidelines. These indicators have been updated in recent expert meetings, and the process is disclosed on their website. 20 Compared to QIs developed according to the CPM project, the selection process for most other QI sets that were developed according to the relevant guidelines and the consensus process of experts' meetings is unclear, because they have not been published or disclosed in English-language articles. Although the consensus processes for QIs for CPM have been disclosed, these were qualitative rather than quantitative processes using methods like the Delphi method.

| Characteristics of QI items
One hundred one QI items were identified among the included articles after identical indicators with different ideal values had been combined. Characteristics of the QI items referenced by more than one article are summarized in Table 2. The detailed characteristics of all QI items are shown in Table S1. The clinical practice guidelines or performance measures supporting the indicators, such as CPM, 20 EBPG, 24 and the clinical practice guideline endorsed by Japanese society of dialysis therapy, 25 are also shown in Table 2. These items were categorized into 10 areas: anemia, mineral and bone disorder (MBD), dialysis adequacy, vascular access, nutrition, fluid management, diabetes, dyslipidemia, infection, and others. The QIs for anemia, MBD, dialysis adequacy, vascular access, nutrition, and fluid management have been examined in several studies (Table S1). Most of these indicators were for surrogate outcomes, such as achievement of hemoglobin level, serum calcium level and Kt/V, and maximizing the use of AVF,

| Variations in data resources
The data sources used to measure each QI are summarized in Tables 2 and S1. Quality indicators for anemia and MBD can all be measured using information in claims data or blood tests. In contrast, several QIs for other areas need measurement using a chart review. For example, measurement of Kt/V or ultrafiltration rate requires detailed information such as dialysis time, postdialysis body weight, and ultrafiltrate volume, which can rarely be retrieved from test results or claims data.
Furthermore, a chart review is needed to measure QIs for infection due to the need for diagnostic information.

| DISCUSSION
We conducted a systematic review and generated an items list of existing QIs for adult maintenance hemodialysis patients to determine the pros and cons associated with their use and to discuss the requirements for the development of future QIs. We evaluated variations in the areas and types of indicators and the associated development processes. We also categorized the source data to measure each item in the QI sets. From the perspective of this information, we then assessed the pros and cons of the existing QIs for maintenance hemodialysis patients.
Most QI sets fell under the following areas: anemia, MBD, dialysis adequacy, vascular access, nutrition, fluid management, and infection.
Most of the QIs for these areas measured surrogate outcomes, and in particular, indicators for nutrition comprised only surrogate outcomes.
Although outcome indicators are, generally, intuitive and easy to understand, in practice, they often require long-term observation to detect changes. Furthermore, they often require case-mix adjustment, because they are easily influenced by a patient's condition. 14 In contrast, process indicators are so useful for detecting changes in practice within a short period of time, that those most associated with relevant outcomes have been recommended for the assessment of quality of care. 8,14 Additionally, the inclusion of more process indicators may benefit maintenance hemodialysis patients. In particular, as our findings show that indicators of nutrition include only surrogate outcomes, process indicators such as nutritional support should be included.
We Of the indicators developed using expert consensus, the development process of CPM developed by Medicare were disclosed in a peerreviewed article 27 and on their website. 20 The consensus was achieved by face-to-face discussion, and the evaluation was descriptive rather than using a quantitative approach such as the Delphi method. A previous report suggested that a consensus achieved using only a face-toface discussion could be biased toward the opinions of dominant persons or groups, owing to the difficulty in assuring anonymity in such processes. 15 To develop a validated set of QIs, it may be important to clarify the consensus process and to exclude such dominance.
The feasibility of measure indicators may be important in the selection of QI sets. Health care providers reportedly spend large sums of money to report their QIs. 8  QIs for maintenance hemodialysis patients should use definitive consensus processes and consider process-centered indicators, which can be measured automatically using claims data and test results contained in electronic medical records, to improve usability and feasibility.

CONFLICTS OF INTEREST
All authors have no conflicts of interest.