Patient and provider determinants of nephrology referral in older adults with severe chronic kidney disease: a survey of provider decision making

Chronic kidney disease (CKD) is a recognized public health concern that preferentially affects older adults, more than a third of whom over age 70 meet the formal definition of CKD [1, 2]. According to current Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines, patients with severe CKD (a glomerular filtration rate (GFR) of less than 30 ml/min/1.73 m²⁾ should be referred to a nephrologist to manage the potential complications of CKD, slow the progression to end stage kidney disease, prepare patients for renal replacement therapy, and reduce mortality [3–5]. However, older adults though more likely to have severe CKD, are less likely to be referred to a nephrologist [5–7]. Previous studies of both primary care (PCPs) and nephrology providers suggest that more severe kidney disease, increased complexity of medical co-morbidity, limited life expectancy and patient preferences against dialysis, in addition to older age, influence provider referral decisions [8–11]. Our own work suggests that providers also consider the cognitive and functional status of older patients when deciding whether or not to refer the patient to a nephrologist [12]. In addition to patient factors, PCP characteristics might explain some of the variability in referral decisions [11, 13]. However, precisely how PCPs prioritize and incorporate these patient factors into their decision-making strategy when addressing individual patients is not known. We previously outlined a conceptual framework that describes the complexity of PCP referral decisions for CKD and incorporates three main components relevant to the referral decision: patient characteristics, provider characteristics, and disease characteristics [14, 15]. This study was designed to systematically assess the role of specific patient characteristics while accounting for characteristics of practicing PCPs that might influence referral decisions for older adults with severe CKD.

Collectively, the respondent PCPs made 680 referral decisions. Despite all vignette patients being appropriate for referral based on KDOQI guideline recommendations, PCPs did not recommend referral in 257 (37.8%) of their decisions. PCPs most commonly cited multiple medical co-morbidities, no intervention warranted, and medical management that can be done in the PCP's office as reasons for non-referral.

The majority of PCPs reported that the severity of CKD was a "moderately" or "very important" factor in their decision to refer (69.5%). A much smaller group of providers noted that functional status (29.4%), cognitive status (32.9%) and the presence and severity of additional co-morbidities (45.9%) were at least "moderately" important in their referral decision. Only a few PCPs reported that patient age (16.5%) or race (3.5%) played an important role in their referral decision.

PCPs overwhelmingly expected nephrologists to evaluate the etiology of CKD (97%), discuss prognosis (98%), discuss the initiation of dialysis (94.1%), and discuss goals of care (96%) with referred patients. Most of them expected the nephrologist to manage CKD-associated anemia (64%), but not the blood pressure (40%).

CKD is an important health concern for older adults and yet they do not access nephrology services at the same rates as younger patients with equivalent severity of disease. In this study of PCP decision making about nephrology referral for hypothetical patients with severe CKD, we find that nearly 1/3 of older patients are not referred despite meeting guideline criteria for referral. We identified that older age and moderate dementia significantly reduce the likelihood that a patient will be referred while other patient factors including medical co-morbidity and functional disability do not impact such decisions. We also find that the likelihood of referral is impacted by provider characteristics and expectations independent of patient factors. These results suggest that referral decisions are complex in ways that current practice guidelines do not address.

Our finding that a large number of hypothetical patients with CKD are not referred to nephrology is consistent with previous studies of actual patient populations. Previous studies suggest that 33-85% of persons with moderate to severe CKD were unknown to nephrology services [5, 7]. In several retrospective studies of CKD populations, older age is associated with lower likelihood of referral [5, 7, 9]. Our study further confirms that patient age is a strong predictor of non-referral.

However, our study also evaluated several other factors commonly associated with age to determine whether PCPs use age alone or as a proxy for other age-associated conditions. In our study, moderate dementia is independently associated with a lower likelihood of referral. Previous studies have not identified dementia as an independent factor for non-referral. Perhaps this is because dementia is under-recognized and under-diagnosed even in the CKD population where persons are at higher risk [21, 22]. Contrary to other studies about CKD referral, medical co-morbidity is not associated with the referral decision in this study [9]. This finding may be a result of the study design in which we varied the presence or absence of a single co-morbidity (stage III CHF). The presence of stage III CHF alone may not influence the referral decision. Instead, provider referral decisions might be more strongly influenced by other significant co-morbidities [23]. Using this study design, we did not have the sample size to fully account for all the possible co-morbidities. Contrary to our previous findings, functional disability was not associated with the referral decision [12]. In these clinical scenarios, functional disability was represented as the use of an assistive device for ambulation. We hypothesized that PCPs would view the use of a walker as an indication of frailty and, therefore, they would be less likely to refer [12]. However, providers may not identify functional disability or frailty by the presence of impaired ambulation. Perhaps deficits in independent activities of daily living (IADLs), which we did not assess, are more important to providers. A fuller accounting of other relevant factors awaits future work.

PCP estimation of longer remaining life expectancy for a patient was significantly associated with the decision to refer in bivariate analyses although the difference was less than 1 year and of unclear clinical significance. However, estimated remaining life expectancy was not independently associated with likelihood of referral in multivariate analyses. Remaining life expectancy is a concept that incorporates age, gender, number/severity of medical co-morbidities, and the presence of functional disability into an estimation of physiologic age [24, 25]. Arguably, remaining life expectancy could be used to evaluate patients instead of age given the significant heterogeneity of older adults where chronologic age may not be as important for outcomes [25]. In this study, PCPs exhibited significant variability in estimating remaining life expectancy suggesting that across PCPs there are inconsistencies in how providers calculate or conceptualize life expectancy. These inconsistencies in life expectancy calculations have important implications for the possible use of life expectancy rather than age in practice guidelines for CKD. If provider life expectancy estimates are significantly inaccurate, then the decisions made based on such estimates might be inappropriate. Future decision making studies could address how providers estimate life expectancy and determine whether providing an estimate of life expectancy would alter the referral decision.

In comparison to the few patient factors which affected the referral decision, we found several PCP factors that were statistically associated with the decision. Although several of these PCP factors have wide confidence intervals suggesting imprecise estimates due to the sample size, the potential implications of their influence on provider decision making should be further examined in future studies. The greater the importance of CKD severity to the PCP, the more likely he or she is to refer the patient, suggesting that raising PCP awareness about the health consequences of CKD might impact referral decisions. Also, PCPs who report referring greater numbers of patients to nephrology in the preceding year were more likely to refer the hypothetical patients. PCP expectation that the nephrologist would discuss goals of care, but not dialysis initiation, was very strongly associated with a higher likelihood of referral. Perhaps PCPs are reluctant or unprepared to discuss care goals with older patients who have severe CKD. This finding suggests that PCPs expect nephrologists to engage patients in a broader discussion about goals and not simply discuss dialysis. Interestingly, we find that PCP gender and race are independently associated with referral consistent with previous studies of referral behavior in other contexts [26–28]. It is unclear why this gender and race difference occurs. Our analyses are adjusted for PCP age, awareness of guidelines, reaction to uncertainty, and risk aversion, which might be hypothesized to vary by gender and race. While some of these factors exerted their own independent effects on the referral decision, they did not attenuate the effect of provider gender and race. Specifically, we find that PCP anxiety about clinical uncertainty, concern about bad outcomes, and greater risk aversion have significant independent effects on the referral decision. These results do suggest a potential role for these provider characteristics in decision making. The full meaning of the associations of these general measures of clinical uncertainty and risk aversion on referral decisions management of CKD awaits future work.

We hypothesize that PCP factors including expectations, personality traits, and risk preferences are important in decision making when the outcome of the decision is uncertain. In the case of referral to the nephrologist for severe CKD, PCPs confront several uncertainties including: 1) how likely is this older patient with CKD to progress to end-stage renal disease; 2) will the patient have strong preferences about starting dialysis if it is indicated; or 3) will the referral to nephrology result in improved outcomes for this individual patient. Given the lack of evidence to answer these uncertainties, it is likely that PCP characteristics, expectations, and personality traits will be more important in the decision making and contribute to practice variation. As evidence accumulates to address these uncertainties and guidelines incorporate this evidence into algorithms for risk assessment of older adults, the effects of these PCP characteristics on the referral decision may diminish. It remains to be seen if CKD-focused educational efforts targeting PCPs would eliminate the gender and race discrepancies or reduce the effect of personality traits in referral decision making.

There are several limitations to this study. First, this is a study of decision making using hypothetical vignettes. Decision making studies that involve clinical vignettes are highly correlated with decisions made during actual patient encounters; and vignettes are a cost-effective means of evaluating decisions that may not be easily studied in routine practice [29, 30]. In this study, the careful control over variables of interest (e.g. GFR) provides insight into the specific role of the selected patient factors in the referral decision, but somewhat limits inferences for actual practice. Specifically, the effect of other medical co-morbidities associated with CKD, but not included in this study, on the referral decision was not assessed. Future studies involving actual patients would be needed to supplement this work and help to determine if these patient factors, or other disease-specific factors such as malnutrition or hypoalbuminemia, influence referral practices. Second, this is a study of a relatively small group of PCPs in practice-based research networks. As such, these results may not generalize to PCPs in other settings. However, this study did involve community-based PCPs from different types of practices and settings, in 2 different regions of the country. These findings are similar to results of our previous study of internists, geriatricians, and nephrologists at an academic institution [12, 23]. The lower response in this study is comparable to other physician surveys and is typical of internet-based studies [2, 9]. A larger, more representative sample of practicing providers would be needed to further test these findings. Third, through the use of the vignettes, we controlled for the degree of CKD and patient preferences, both of which may be important in the PCP's decision to refer the patient. Future studies could vary GFR to determine if providers use a threshold level of GFR to decide whether to refer regardless of other factors and whether this threshold differs from current guideline recommendations. We also recognize that patient (and family) preferences are important. The assumption that patients have no strong preferences about referral to a nephrologist may be incorrect. Perhaps patients (and families) equate a referral to nephrology with a decision about initiating dialysis. If this is true, then educating patients and PCPs about the important role of the nephrologist in managing severe CKD might improve referral rates. Additional studies are needed to understand patient expectations of a nephrology referral.

Our findings provide new insight into PCP referral decisions for older adults with severe CKD to the nephrologist. In this study of PCP decision making using hypothetical clinical vignettes to systematically vary specific patient factors, we identify previously unrecognized patient (e.g. cognitive impairment) and provider (e.g. risk preferences) factors which independently predict which older patients with severe CKD are more likely to be referred by PCPs to a nephrologist. Additional work is necessary to determine if geriatric conditions like cognitive impairment and provider characteristics are also important in the outcomes of older patients with advanced CKD. Still, the results of this study underscore the complexity of the referral decision where a seemingly simple clinical recommendation--to refer an older adult with severe CKD as guidelines recommend--may involve challenging tradeoffs among multiple factors.