Background
Health-related quality of life (HRQOL) is a critically important outcome for patients with end-stage renal disease (ESRD). In recognition of this, in 2008, the Centers for Medicare and Medicaid Services (CMS) mandated annual assessment of HRQOL as part of its conditions for coverage for ESRD facilities [
1]. The National Quality Forum selected the Kidney Disease Quality of Life Short-Form survey (KDQOL™-36) as the tool of choice for assessing this outcome in adult patients with ESRD; assessment is required within 4 months of initiating dialysis, and annually thereafter [
2]. This 36-question survey instrument was published in 2000 [
3], based upon a longer KDQOL instrument first developed in 1994 [
4]. The KDQOL™-36 contains 5 subscales: the Physical Component Summary (PCS), Mental Component Summary (MCS), Burden of Kidney Disease (BKD), Symptoms and Problems of Kidney Disease (SPKD), and Effects of Kidney Disease (EKD). The first 2 subscales are a generic measure of HRQOL (and are identical to the SF-12), whereas the last 3 assess issues specific to patients with ESRD or earlier stages of chronic kidney disease.
In spite of the importance of HRQOL as an outcome, and the long-standing use of the KDQOL™-36 for its assessment among patients with ESRD, validation of the instrument has been fairly limited [
5]. Indeed, the renal community in the United States has expressed concern regarding the relevance of the KDQOL™-36 in the current environment [
6], given changes in ESRD care that have occurred since its development. Recent work has identified factors beyond those included in the KDQOL™-36 that patients have identified as important to their quality of life [
7,
8], raising questions as to whether a revision of the KDQOL™-36 may be necessary.
Here, we sought to understand the survey response rates, distribution of component scores, and distribution of responses to individual items among a large, nationally representative contemporary cohort of US patients with ESRD. We also sought to examine correlations between component scores, individual item responses, and fluid status. In particular, given the importance of disease symptoms as a driver of HRQOL [
9‐
12], we sought to determine the extent to which the items comprising the SPKD subscale have remained relevant since the instrument’s development.
Methods
Patients, timeframe, and data sources
This study considered KDQOL™-36 survey opportunities (ie, recorded instances at which the KDQOL-36™ survey was offered to a patient) that occurred between 01 January 2014 and 21 December 2016 at a large dialysis organization (LDO) operating in the United States. The LDO has clinics in 45 of the 50 states (the exceptions are Alaska, Delaware, Indiana, Vermont, and Wyoming). We considered patients who, as of the date of the survey opportunity, were ≥ 18 years of age, were not Veterans Affairs beneficiaries (contractual stipulation), and were receiving dialysis care at the LDO. All data, including KDQOL™-36 item responses, patient demographic information, and treatment information, were derived from electronic records maintained by the LDO.
Among patients treated with in-center hemodialysis (ICHD), information about treatment attendance and inter-dialytic weight gain (IDWG) in the 30 days prior to the survey opportunity was likewise derived from electronic health records. Treatment attendance records indicate whether a patient attended treatment, missed treatment due to hospitalization, or missed treatment due to other reasons (referred to as “absences.”). IDWG was considered as either the average percent of target weight gained prior to each attended treatment, or as frequent excessive IDWG based on the internal definition utilized by the LDO (> 5% of target weight in > 10% of treatment intervals).
This study was conducted using deidentified patient data; therefore, according to title 45, part 46 of the US Department of Health and Human Services’ Code of Federal Regulations, this study was deemed exempt from institutional review board (IRB) or ethics committee approval (Quorum IRB, Seattle, WA). We adhered to the Declaration of Helsinki and informed consent was not required.
Surveys
At the LDO, KDQOL™-36 surveys are offered to patients within 4 months of their first treatment at the LDO, and a minimum of annually thereafter, in accordance with CMS policy [
2]. Surveys were considered as “completed” if, at minimum, questions 1–12 (comprising the PCS and MCS) were answered. Surveys were considered as “declined” if the patient declined to complete the survey when offered (patients were not required to provide a reason for declining the survey), or if questions 1–12 were not completed. No imputation methods were used to handle missing values. All survey scores (from completed surveys and surveys with missing values in items 13–36) were calculated according to the procedures set forth by the Rand Corporation, which developed the survey. Because CMS guidelines indicate that a survey with responses to items 1–12 be regarded as complete irrespective of the status of the other items, we applied the same criteria in our analysis. Survey opportunities at which the survey was not offered to a patient for reasons of dementia, active psychosis, or other barriers to survey completion, were excluded from the analysis. On occasion, social workers will administer an off-cycle KDQOL™-36 survey in response to changes in patient circumstance (eg, severe illness, death of a caregiver, etc), resulting in more than one recorded KDQOL™-36 survey opportunity during a survey period (ie at approximately 4, 16, or 28 months of dialysis initiation). In such instances, only the first opportunity in the survey period was considered in this analysis. Responses to individual items on the KDQOL™-36 survey were used to calculate scores on the 5 subscales according to formulae specified by the developer of the instrument [
13]. These formulae are specified such that a higher numeric value of score is always reflective of a better perceived quality of life. In the case of the BKD score, a higher score is reflective of a
lower perceived burden of kidney disease.
Statistical analyses
Data are presented as means and standard deviations, medians and interquartile ranges, or counts and proportions, as dictated by data type. Correlations between item responses, domain scores, and IDWG were calculated using Pearson correlations [
14] computed with pairwise complete observations.
Discussion
This study examined KDQOL-36™ survey response rates and scores from a large, nationally representative cohort of contemporary ESRD patients treated with four different dialysis modalities. The overall survey response rate was 78.2%, and mean component scores were similar to those reported for other US dialysis cohorts [
5,
16]. Stratification of the study cohort on the basis of demographic and clinical characteristics revealed specific characteristics that were associated with lower or higher participation rates in the survey. Similarly, this stratification enabled identification of characteristics associated with particularly high or low mean scores on specific KDQOL-36™ domains. Beyond specific clinical, social, or other factors that may influence quality of life, differences between patient subgroups may be driven by differences in patient expectations with regard to their quality of life. The same set of circumstances may thus correspond to a low or high perceived quality of life, depending on the particular patient.
Notably, despite differences in patient demographics, KDQOL-36™ scores did not vary substantially across the four dialysis modalities examined. Indeed, a majority of the differences observed across modalities were less than 3–5 points, which is generally considered the threshold for minimally important differences (MID) in score on the KDQOL-36™ instrument [
17,
18]. Several previous studies have examined correlations between dialysis modality and HRQOL, generally focusing on the comparison between ICHD and PD in cohorts substantially smaller than the one examined here [
19‐
23]. Although differing to some extent with respect to particular details, these earlier studies and the findings presented here support the conclusion that HRQOL tends to be quite similar across ICHD and PD, a finding that the present study now extends to NOC and HHD as well. The consistent observation that similar HRQOL is achieved across treatment modalities, despite the fact that patients on ICHD tend to be older and sicker than those treated with other modalities, suggests that appropriate matching of individual patients to the dialysis modality that best fits their needs, rather than the modality itself, may be a key determinant of HRQOL.
This study has identified some important limitations of the SPKD subscale of the KDQOL-36™. Mean scores on the SPKD were numerically higher than mean scores on the other 4 subscales. Although SPKD was moderately correlated with PCS, as might be expected for two scores that measure physical aspects of health, mean score on the SPKD exceeded mean score on the PCS by approximately 40 points. Thus, the two scores convey very different messages about patient health: a PCS score in the 30’s is suggestive of extremely poor overall health, whereas an SPKD score of 70 or higher suggests a relatively low symptom burden. This pattern is suggestive, although not proof positive, that the SPKD subscale may be topped out.
In support of the idea that SPKD may be topped out, a majority (≥65%) of patients in the contemporary cohort studied here were “not at all” or only “somewhat” bothered by each of the symptoms queried by the instrument. Notably, this was true for 8 of these 12 items at the time that the KDQOL was first developed [
4], suggesting that even within the clinical context at that time, these symptoms may not have been particularly troubling to a majority of patients. Based on the findings presented here, it appears that the 4 symptoms that were more bothersome at the time that the instrument was developed (dry skin, itchy skin, washed out or drained, and muscle soreness) are less so in the current clinical context, given the substantial improvements in dialysis delivery and other aspects of ESRD management have occurred over the past two decades. This raises the possibility that the list of symptoms queried in the SPKD may need to be refreshed to account for corresponding changes in the experience of patients on dialysis over that time.
The degree to which symptoms queried on the SPKD correlate with objective measures of patient health during the 30 days prior to survey has not been extensively examined. Here, we found that there was no correlation between fluid status (ie IDWG) in the 30 days prior to survey and a symptom that is largely driven by fluid overload, namely shortness of breath. The absence of such a correlation raises questions as to how to interpret and address patient perceptions of the symptoms reported on the KDQOL-36™, and further underscores the need to reexamine this subscale.
Strikingly, the response to item 1 on the KDQOL-36™ (“In general, would you say your health is”) was not correlated with any of the 5 subscale scores, nor with the response to any individual item on SPKD. This is notable in that patient-reported general health is thought to reflect aspects of health that are difficult to capture via clinical measures, and is independently associated with mortality risk [
24]. This finding suggests that efforts to identify factors that influence perceptions of general health among dialysis patients, and the inclusion of such factors on survey instruments, may facilitate more nuanced understanding of HRQOL.
Although the survey completion rate in our study was high, the fact that approximately 20% of survey opportunities were declined indicates that there may yet be opportunities for improvement of the completion rate, particularly among the most vulnerable patients. Options such as electronic survey administration may increase participation while reducing burden on those administering the survey [
25]. Shortening the KDQOL instrument by eliminating less relevant items, and thereby reducing burden on patients and those administering the survey, could also be considered.
This study should be interpreted within the context of its limitations. A major limitation is that this study can, by definition, only report scores and correlations for patients who complete surveys. Patients who decline to complete the survey, or those who are not offered the survey due to clinical or other circumstances, cannot be considered. This may result in under-representation of some of the most vulnerable patients; indeed, previous work has shown that patients who decline to complete HRQOL surveys are at significantly greater risk of mortality [
16]. Furthermore, not all patients are eligible to complete the survey: for example, patients who are under the age of 18 or those who have a diagnosis of dementia are excluded. The degree to which the findings of this study extend to these under-represented groups is not known. A minority of patients may receive assistance from social workers or other staff members in completing their KDQOL-36™ surveys. Unfortunately our source data did not permit us to distinguish surveys that were completed with assistance, and thus a sub-group analysis of such surveys was not possible. Similarly, a minority of patients may have switched modalities during the study period, thus contributing surveys to multiple modality groups. However, because approximately 85% of patients are treated with ICHD at any given time, the number of patients who contributed surveys to multiple modalities will necessarily be quite small. Finally, because the intent of this study was to provide a descriptive overview of scores on the KDQOL-36™ among contemporary dialysis patients, no attempts were made to adjust scores for case-mix variables. Rather, the findings presented here are intended to serve as the basis for designing or interpreting future predictive or associative analyses.
Acknowledgements
A portion of the results were presented in poster form at the American Society for Nephrology Kidney Week meeting in New Orleans, LA in November, 2017. We thank all members of the Healthcare Analytics and Insights team at DaVita Clinical Research for helpful discussions during the execution of this project. We also gratefully acknowledge the contribution of the social workers at DaVita who administer the KDQOL-36™ survey, and the patients who complete it, without whom this work would not have been possible.