This study indicates that hyponatremia represents a substantial medical and economic burden in the U.S. There are approximately 1 million hospitalizations per year in the U.S. with a principal or secondary discharge diagnosis of hyponatremia, as well as an estimated 105,000 to 120,000 annual ER visits, and 1.4 million to 3.4 million annual office visits for hyponatremia. The cost of illness estimate of $1.6 billion to $3.6 billion for hyponatremia can be put into perspective by reviewing published direct cost estimates for other conditions (updated to Year 2004 US $), including $788 million for treating children with respiratory syncytial virus [29
], $1.5 billion for treating refractory epilepsy in adults [18
], $2.3 billion for treating hay fever [30
], $23.4 billion for treating urinary incontinence [31
], and $23.7 billion for congestive heart failure [32
There have been no previously published estimates of the total direct costs of treating hyponatremia, but several previous studies corroborate the conclusions of our analysis. Results from a prospective study of 435 patients admitted to a university hospital with evidence of congestive heart failure showed that hyponatremia (defined as serum [Na+
] less than or equal to 135 mEq/L) was significantly (P
≤ 0.01) and independently associated with an increased duration of hospital stay and higher hospital cost [10
]. The increased length of hospital stay in patients with hyponatremia was demonstrated in another retrospective analysis of 1,046 patients (58% older than 65 years) hospitalized for heart failure [13
]. In this study, 171 patients had hyponatremia (defined as serum [Na+
] less than 135 mEq/L) at admission and their mean length of stay was 5.78 days, versus 4.72 days among patients without hyponatremia (P
= 0.0001). The only variable other than hyponatremia that was associated with a longer duration of hospitalization in this study was admission from a skilled nursing facility (6.22 days). A multivariate linear regression analysis indicated that hyponatremia was a significant predictor of hospitalization duration in this cohort of patients.
The current study's cost of illness estimate for hyponatremia is most likely a conservative one. The prevalence estimate was based in large part on the number of hospitalizations for hyponatremia as recorded (by ICD-9-CM diagnosis code) in a national database, but there is evidence that the ICD-9-CM code for hyponatremia represents only one-third of the patients admitted to the hospital and experiencing hyponatremia, due to the low sensitivity (30%) of the diagnosis code [33
]. In addition, a high proportion of hyponatremia in the hospital setting is iatrogenic [12
] and hospitals may be reluctant to include the code in the discharge data.
More definitive resource use and cost data from longitudinal, patient-level databases would have been preferred. However, as noted above, existing databases have their own inherent weaknesses due to the lack of sensitivity with the ICD-9-CM diagnosis code for hyponatremia. Future studies should therefore consider a broader national survey of treatment patterns and resource use associated with hyponatremia.
There are additional limitations associated with this analysis. Although previous research has found that consensus panel decisions have a high degree of consistency and validity when compared with clinical practice [20
] the panel estimates in the current study are uncertain. A variety of formal and informal methods have been developed for use as consensus-building techniques in group decision-making [35
]. The consensus development process in this study was a variation of a modified Delphi panel. In the first stage of a two-stage process, participants privately completed a mailed questionnaire. In the second stage, their compiled responses were presented at a face-to-face meeting where the group engaged in open communication to discuss any variations in their responses. The panel members reached consensus as a group on an appropriate estimate for each question, often in the form of a range. Unlike a true Delphi panel where participants never meet directly, a noted strength of the interactive forum is the opportunity the participants have to provide information, insight, and rationales for their responses. However, a limitation of this approach is the potential for decisions to be reached by persuasion rather than consensus due to an influential member of the group. While no single member of the panel in this study appeared to dominate the consensus process, we recognize that social forces such as persuasion and conformity may have influenced panel members' final decisions.
Additional uncertainty in the panel's estimates lay in the subjective nature of their responses. Previous commentaries in the literature have suggested the potential for bias in prevalence estimates provided by practicing clinicians because their experience is based on the duration of illness, severity, and other clinical characteristics of patients who receive treatment [36
]. For example, the prevalence of severe and symptomatic hyponatremia may be easier to estimate than the number of patients who have undetectable symptoms. We believe we minimized this potential bias by having a cross-disciplinary panel familiar with the variety of ways hyponatremia can present itself (i.e. acute, chronic, symptomatic, asymptomatic).
Another study limitation is the lack of inclusion of costs associated with complications of hyponatremia, which although rare, can be substantial. The panel felt it would be difficult to quantify the complications for the extremely small percentage of patients who experience these events. Resource use and costs associated with complications vary depending on the nature and severity of the complication. Furthermore, many complications of hyponatremia are neurological with severe long-term sequelae. Therefore, an accurate assessment of the economic burden would have to include direct and indirect costs incurred over time, which would vary depending on several patient, clinical, and treatment factors. Given the high degree of uncertainty associated with estimating the economic impact of complications and the low percentage of patients involved, the panel deemed it most appropriate to exclude complications from the analysis.
The analysis also did not include the indirect costs associated with hyponatremia. The expert panel did not feel qualified to assign levels of work loss or caregiver burden based on the presence of hyponatremia; and there were no data sources available to directly link hyponatremia with work loss. The increased mortality risk that has been linked to hyponatremia [9
] was assumed to apply mostly to non-working elderly populations, and thus the productivity losses due to mortality were considered minimal.
This analysis of the economic impact of hyponatremia raises a number of clinical implications that have not been fully appreciated nor discussed regarding this disorder. The clinical importance of symptomatic hyponatremia has been well appreciated by clinicians over the past decade, both as a result of the morbidity and mortality associated with hyponatremic encephalopathy, as well as that associated from the production of pontine and extrapontine myelinolysis from overly rapid correction of severe hyponatremia [38
]. However, both of these situations are relatively rare in terms of overall incidence, likely representing 1% or less of all hyponatremic patients (Table ). While these dramatic cases have appropriately received much attention in the medical literature, they represent only a small fraction of the resource utilization and costs associated with hyponatremia. Rather, the bulk of the costs attributable to hyponatremia appear to result from a combination of inpatient hospitalization costs (70%) and subsequent follow-up evaluation and treatment (15%–20%), and 80% of these are attributable to those patients for whom hyponatremia was not the primary diagnosis. Thus, these relatively conservative estimates suggest that more than two-thirds of the cost of hyponatremia occurs from patients hospitalized for other conditions whose length of hospital stay is then extended due to coincident hyponatremia. Further analysis of the reasons underlying this observation is therefore indicated.
Several possibilities can potentially explain this association. First, hyponatremia may be a marker of the severity of the underlying disease, in which case hospitalizations are longer simply because the hyponatremic patients represent a sicker cohort of all those with the underlying disorder. Second, hyponatremia may add its own complications to those of the underlying disorder, thereby acting as an independent factor that extends the length of hospital stay due to the intrinsic complications of this disorder. Third, the presence of hyponatremia may limit or otherwise compromise optimal treatment of the underlying disorder. Finally, because newly-discovered hyponatremia represents a metabolic abnormality of uncertain etiology and significance, the medical evaluation required to ascertain the underlying cause of the hyponatremia will necessarily involve investment of additional time and resources. Each of these possible explanations will be considered in greater detail.
Hyponatremia has long been known to occur in association with a variety of underlying conditions, from tumors that synthesize and excrete arginine vasopressin ectopically [39
] to disorders such as congestive heart failure and cirrhosis where arginine vasopressin secretion from the posterior pituitary is stimulated by decreased effective circulating blood volume [40
]. It is striking that mortality rates have been found to be significantly higher in hyponatremic patients across a broad range of primary disorders, including congestive heart failure and acute myocardial infarctions [41
], pulmonary tuberculosis [42
], and childhood diarrhea [43
]. Perhaps the strongest data for hyponatremia as a marker of disease severity comes from multiple studies of patients with congestive heart failure, which have clearly shown that hyponatremia represents an independent risk factor in patients with heart failure [8
], nearly doubling the risk of mortality in this group [44
]. Most evidence suggests that this association reflects the underlying pathophysiology of the heart failure (i.e., that hyponatremia is a marker of severity of the underlying disease). This is partly based upon the findings that arginine vasopressin is one of the hormones stimulated during the activation of multiple neurohumoral systems that occurs in association with progression of the heart failure. In the SOLVD (Studies of Left Ventricular Dysfunction), subjects with left ventricular dysfunction had significantly higher plasma arginine vasopressin levels compared to controls, and arginine vasopressin levels were highest in the subjects with overt heart failure [46
]. While these data support the possibility that case hospitalizations are longer in hyponatremic patients because they represent a sicker cohort of all patients with the underlying disorder, there are a number of reasons to suggest that the elevated plasma arginine vasopressin levels associated with hyponatremia may in fact aggravate disease progression in patients with heart failure. Specifically, the excess water retention caused by arginine vasopressin may cause worsening of congestive heart failure due to diastolic wall stress from the intravascular volume expansion that is caused by the excess retained water; in addition, the elevated arginine vasopressin levels may lead to increased systolic wall stress as a result of arteriolar vasoconstriction produced by activation of vasopressin V1a receptors in the vasculature, and potential stimulation of myocardial hypertrophy because of growth-stimulating effects of vasopressin V1a receptors in the heart. Thus, the assumption that hyponatremia due to increased arginine vasopressin levels is simply a marker of the severity of the underlying left ventricular dysfunction in patients with congestive heart failure rather than a causal factor in the increased mortality of this subgroup has never been directly tested and remains a presumption.
Regardless of whether elevated arginine vasopressin levels and hyponatremia directly contribute to the morbidity and mortality of underlying primary diseases, there is little question that the presence of hyponatremia can and often does interfere with the treatment of underlying diseases through multiple mechanisms. Perhaps most importantly, standard therapy for euvolemic and hypervolemic patients with hyponatremia is fluid restriction in order to prevent further water retention and worsening of the hyponatremia. This necessity can limit therapies that involve concomitant fluid administration to patients, including antibiotic therapy, chemotherapy, and parenteral nutrition. Furthermore, hyponatremic patients with edema-forming diseases such as congestive heart failure and cirrhosis who require aggressive diuresis of retained water and sodium sometimes do not receive as large a dose of diuretics as otherwise might be given because of fears of worsening hyponatremia as a result of the natriuresis produced by conventional diuretic agents. In each case, this would result in prolonging the period to reach the medical endpoint of the hospitalization.
Finally, even if none of the above scenarios apply to a specific case, the current standard of care for newly diagnosed hyponatremia is to ascertain the etiology of the hyponatremia before ascribing it to the underlying disease [47
]. This requires a combination of both laboratory and radiological testing (Table ) that can add several days to hospitalization, or alternatively, the employment of these resources during follow-up visits. In many cases underlying etiologies are not found,[48
] raising questions about the efficacy of the minimum diagnostic evaluation that is appropriate for all cases of hyponatremia.
While no study to date has definitively ascertained among the various possible reasons that account for the increased length of stay in patients with coincident hyponatremia, it seems likely that all of the factors postulated as potential causes of increased resource utilization contribute to this occurrence to varying degrees in individual cases.