Association of social worker-assessed psychosocial factors with 30-day hospital readmissions among hemodialysis patients

In 2016, the Centers for Medicare & Medicaid Services (CMS) spent $28 billion on hemodialysis in the United States; about one-third of end-stage renal disease (ESRD) expenditures were for inpatient care [1]. As part of its ESRD Quality Incentive Program, CMS ties reimbursement of U.S. ESRD services to clinical performance [2]. In 2017, this pay-for-performance program, added the standardized readmission ratio (SRR), such that facilities’ Five-Star ratings may be reduced due to higher-than-expected hospital readmissions among their hemodialysis patient population. One of the major criticisms of the SRR is that, while it accounts for some demographic and clinical factors, it does not account for differences in psychosocial factors across facilities, which may disadvantage facilities with disproportionately vulnerable populations [3].

Despite this, relatively little is known about the effect of psychosocial factors on readmissions among hemodialysis patients. In a recent single-center study, El-Majzoub et al. [4] found that psychosocial distress was associated with shorter time to hospitalization, but did not examine hospital readmissions specifically. Flythe et al. [5] found that poor social support and depressive symptoms were associated with higher risk of hospital readmissions among dialysis patients in a prospective study. Both studies used validated tools that were administered in a study setting. However, it is possible that data collected routinely by social workers as part of usual hemodialysis care could capture a wide variety of psychosocial factors and potentially inform providers of patient risk of subsequent hospital readmission without the need for additional assessments. Thus, we aimed to use clinically available psychosocial information addressing a variety of domains, extracted from both structured and unstructured electronic medical record (EMR) data from three metropolitan Atlanta dialysis centers, to identify social worker-assessed psychosocial factors associated with risk of 30-day readmission among dialysis patients.

We found that 22.1% of index admissions were followed by a readmission, similar to recent national studies using similar methodology [9]. Importantly, in this study examining social worker assessment-derived psychosocial factors and 30-day readmission risk among prevalent hemodialysis patients, associations were not statistically significant. However, our results do suggest that history of substance abuse, being unmarried, and patient-reported difficulty adhering to dietary and fluid restrictions may be associated with higher readmission risk, independent of patient and index admission factors. In contrast, depression/anxiety, impaired memory, requiring ambulatory assistance, being work-disabled or employed vs. unemployed, and patient-reported difficulty coming to dialysis were associated with lower readmission risk in our analyses. The other factors we examined—living alone, less frequent social support, community-dwelling vs. assisted living/nursing home, reported difficulty with completing dialysis sessions and taking medications—were not associated with readmission risk overall. However, analyses with a stratified outcome (pulmonary edema-related and other readmissions vs. no readmissions) suggested that living in assisted living or a nursing home and difficulty adhering to fluid restrictions or completing dialysis were associated with increased risk of pulmonary edema-related readmissions, while living alone was associated with increased risk of other readmissions.

Unexpectedly, we found that signs and symptoms of depression/anxiety at start of treatment were associated with 20% lower readmission risk, in contrast to the > 2-fold higher risk of hospitalization [4] and readmission [5] associated with positive screening for depression reported among hemodialysis patients, and to the 1.7-fold higher risk of readmission seen in hospitalized patients generally [10]. It is possible that patients thought by social workers to have depression/anxiety are followed more closely, resulting in lower readmission risk, or that these symptomatic patients are less likely to seek treatment, delaying readmissions past the 30-day threshold. However, this result could be partially attributable to differences in the timing of measurement between our study and these prior studies, if symptoms at discharge have more effect on readmission than chronic depression/anxiety. Social workers may also underdiagnosing depression and anxiety—which seems likely, given that about half of prevalent hemodialysis patients show signs of depression [11] and we found that only 17% of our population were noted to have depression at baseline. Since depression is associated with lower treatment adherence among dialysis patients [12, 13] and transplant recipients [13, 14], it may be important to assess depressive symptoms in this population more accurately and more frequently over time to determine readmission risk. Such work could lead to clinical interventions to reduce depressive symptoms including psychotherapy [15] and mindfulness meditation [16], which may effective in this population.

Non-married status was associated with higher risk of readmission, with no associations seen for frequency of social support or living alone. These results suggest that marital status may provide some protection beyond the presence of social support. However, it is also possible our measure of social support, which did not distinguish types [4] or providers of social support received, may not fully capture social support. Interestingly, several factors that indicate potential lack of independence in these hemodialysis patients, including impaired memory, requiring ambulatory assistance, and being work-disabled vs. unemployed were all associated with lower risk of readmission in our study. It is possible that such patients are followed more closely, both by dialysis providers and by caregivers, thus reducing the risk of readmission. Living in a nursing home or in assisted living, where patients are theoretically followed more closely than community-dwelling patients, did not provide this protection, but this discrepancy may be partially due to increased medical complexity of patients who are no longer community-dwelling. Furthermore, stratified outcomes suggested that, while those in assisted living/nursing homes were at lower risk of other readmissions, they were at higher risk of pulmonary edema-related readmissions. This may reflect a general protective effect of continuing medical care post-discharge, but poor post-discharge dialysis management among these patients. Because nursing homes are also now held accountable for readmissions [17], it may be important to explore better continuity of dialysis care specifically between nursing homes and dialysis facilities.

Difficulty with adherence was associated with both higher (dietary and fluid restrictions) and lower (coming to dialysis) risk of readmission. The prior association may reflect patient’s assessment of self-management skills, which are needed to understand and execute discharge instructions, including medication management—although we found no association with reported difficulty taking medications and readmission risk. The latter may reflect provider knowledge of and attempts to circumvent barriers to coming to dialysis (e.g., lack of transportation) among these patients. However, reported difficulty completing dialysis was not associated with overall readmission risk. These inconsistencies may reflect different types of adherence in dialysis being associated with different factors: for example, depression has been associated with missed dialysis [18] and medication adherence [13], but financial difficulties and provider and health system factors are also associated with problems taking medications [19]. Of course, these somewhat unexpected patterns could also be partially attributable to the misclassification due to self-reported data: while patients may report adherence to be “easy,” they may not find it to be easy in practice; furthermore, patients may not adhere to treatment recommendations for many reasons, related to trust, health literacy, polypharmacy, side effects, and financial barriers [20]. Culturally sensitive interventions to increase self-efficacy, individualized to patients’ problems with adherence, may be needed and may help prevent some readmissions [18, 21, 22].

There are several possible explanations for the lack of statistical significance in our results, including that psychosocial risk factors may not be associated with readmissions in the hemodialysis population. It is also possible that psychosocial factors do not contribute substantially to the ability of other demographic and clinical factors to predict readmissions. Flythe et al. [5] found that the associations between poor social support and depressive symptoms and higher risk of hospital readmissions among dialysis patients were independent of patient and index admission factors, making the explanation of a true null association less likely. Furthermore, in the general population, both patients [23] and providers [24] identify psychosocial variables as the primary reasons for readmissions, and it seems unlikely that these factors have no effect in the hemodialysis population. A future prospective study using validated instruments that are administered by social workers but standardized across multiple centers, or a retrospective study of a much larger dialysis organization (assuming standardized data collection by social workers), is needed to confirm these associations.

Other potential explanations for lack of statistical significance include inadequate statistical power, particularly in smaller subgroups; lack of an effect among our relatively homogeneous patient population, which is entirely urban and predominantly poor; potential selection bias due to missing data; and potential misclassification of the psychosocial factors. The potential for social desirability bias may be even stronger at the baseline assessment, since trust and rapport between social worker and hemodialysis patient may take time to establish [25]. Related to the timing of psychosocial assessment, misclassification is also possible due to the variable time lag between social worker assessment and index admission, which, on average, was around 6 months but was much longer for some patients. However, our results adjusting for this lag gave similar results.

Despite such limitations, our results do provide hypotheses that could be tested in future studies that inform clinical care and policy. Prior studies in the dialysis population have primarily focused on clinical risk factors for readmission [26‐28], but the policy-driven nature of hemodialysis care has generated interest in other patient factors, including depression, social support, and health literacy [5], that could be modified to reduce readmission risk, or used for adjustment to make between-facility comparisons such as the dialysis SRR more fair. In our study, we leveraged existing social worker assessment data, which would be more readily available to policymakers, to examine the effects of multiple psychosocial factors addressing mental health, social support, independence, and ease of adherence on hospital readmission among hemodialysis patients.

There are additional limitations to this study that not already noted above. We excluded those patients with a recent previous hospitalization and those who died within 30 days from index discharge, which could lead to an underestimation of our outcome. Predictors of multiple readmissions or readmissions followed by death may differ from the predictors of the single readmission outcome examined here. While all social workers used the same form, there are no national standards for clinical collection of psychosocial information on patients and the assessment items on these clinical forms, including those for depression assessment, are not validated despite the availability of instruments for this population. Detailed information on medications used for depression/anxiety and on psychotherapy history was limited. Similarly, information on substance abuse did not include detailed information on duration or type of substance (e.g., intravenous vs. by mouth). However, strengths of the study include comprehensive evaluation of multiple psychosocial factors and linkage of data to administrative data with nearly complete capture of hospitalizations.

This study adds to the body of knowledge on how psychosocial factors may affect 30-day readmission in hemodialysis patients. Our results could generate several hypotheses regarding psychosocial factors as potential predictors of readmission, potentially leading to future studies that inform polices and strategies to reduce readmissions among patients on hemodialysis. More frequent and robust collection of data on psychosocial factors, whether by the dialysis social worker or others, is needed. Such data may help guide targeted interventions to reduce readmissions, ultimately reducing costs and improving quality of life among hemodialysis patients.