Abstract
Rossi KC, Kim AM, Jetté N, Yoo JY, Hung K, Dhamoon MS
Epilepsia 2018;59:1603–1611. To determine whether epilepsy admissions are associated with a higher
readmission risk for psychotic episodes compared to admissions for
other medical causes. The Nationwide Readmissions Database is a nationally representative
dataset from 2013. We used International Classification of Diseases,
Ninth Revision, Clinical Modification (ICD-9-CM) codes to identify
medical conditions. There were 58 278 index admissions for epilepsy,
and this group was compared against admissions for stroke (n = 215
821) and common medical causes (pneumonia, urinary tract infection
[UTI], congestive heart failure [CHF], and chronic obstructive
pulmonary disease [COPD], n = 973 078). Readmission rates for
psychotic episodes within 90 days from discharge for index
hospitalizations were calculated. Cox regression was used to test
for associations between admission type and readmission for
psychotic episodes up to 1 year after index admission, in univariate
models and adjusted for multiple medical, social, and psychiatric
variables. Up to 90 days from index admission, there were 683/100 000
readmissions for psychotic episodes in the epilepsy group, 92/100
000 in the stroke group, and 58-206/100 000 in the medical group.
The relative rate of readmission in the epilepsy group was highest
in the first 30 days following index admission (311/100 000).
Unadjusted hazard ratio (HR) for readmission for psychotic episodes
within 1 year in the epilepsy group compared to the stroke group was
6.58 (95% confidence interval [CI] 5.69–7.61,
p < 2 × 10–16), and 4.41
compared to the medical group (95% CI 4.00–4.85,
p < 2 × 10–16). The fully
adjusted HR for readmission in the epilepsy group remained elevated
at 3.63 compared to the stroke group (95% CI 3.08–4.28,
p < 2 × 10–16), and 1.95
compared to the medical group (95% CI 1.76–2.15,
p < 2 × 10–16).
Confounding factors most strongly associated with psychosis
readmission were documented psychosis history at the time of index
admission, younger age, and lower income quartile. An epilepsy admission was independently associated with subsequent
hospital readmission for psychotic episodes, even after adjustment
for confounding variables.
Kim AM, Rossi KC, Jette N, Yoo JY, Hung K, Dhamoon MS
Neurology 2018;91:e800–e810. To determine if epilepsy admissions, compared to admissions for other
medical causes, are associated with a higher readmission risk for
mood disorders. The Nationwide Readmissions Database is a nationally representative
dataset comprising 49% of US hospitalizations in 2013. In this
retrospective cohort study, we used ICD-9-CM codes to identify
medical conditions. Index admissions for epilepsy (n = 58,278) were
compared against index admissions for stroke (n = 215,821) and
common medical causes (n = 973,078). Readmission rates (per 100,000
index admissions) for depression or bipolar disorders within 90 days
from discharge for index hospitalization were calculated. Cox
regression was used to test for associations between admission type
(defined in 3 categories as above) and readmission for depression or
bipolar disorder up to 1 year after index admission, in univariate
models and adjusted for age, sex, psychiatric history, drug abuse,
income quartile of patient's zip code, and index
hospitalization characteristics. The adjusted hazard ratio (HR) for readmission for depression in the
epilepsy group was elevated at 2.80 compared to the stroke group
(95% confidence interval [CI] 2.39–3.27, p < 2
× 10–16), and 2.09 compared to the medical group (95%
CI 1.88–2.32, p < 2 × 10–16). The
adjusted HR for readmission for bipolar disorder in the epilepsy
group was elevated at 5.84 compared to the stroke group (95% CI
4.56–7.48, p < 2 × 10–16), and 2.46
compared to the medical group (95% CI 2.16–2.81, p < 2
× 10–16). Admission for epilepsy was independently associated with subsequent
hospital readmission for mood disorders. The magnitude of elevated
risk in this population suggests that patients admitted with
epilepsy may warrant targeted psychiatric screening during their
hospital admission.Objective
Methods
Results
Significance
Objective
Methods
Results
Conclusion
Commentary
We seem to be in an era of big data for psychiatric issues associated with epilepsy. We have already benefitted from recent large-scale studies in the United Kingdom and United States showing relationships between epilepsy and depression as well as other psychiatric conditions (1–4). Population studies offer real-life examples of illness progression and relationships and are not skewed by inevitable idiosyncrasies of recruiting and assembling study populations. The reports from Rossi et al. and Kim et al. offer more information for our consumption. Both papers reference the same dataset, though Rossi et al. focus on psychosis and Kim et al. on mood disorders.
The analyses for both papers stem from a nationwide anonymized tracking system from the federally funded Healthcare Utilization Project. The Nationwide Readmissions Database was created from this project and geared toward assessing factors leading to hospital readmission. Individual patients were tracked with unique identifiers, establishing which patients were readmitted to the hospital and for what reasons. The data set was massive, impressively capturing 15 million hospitalizations in the United States.
The data points included the International Classification of Diseases, Ninth Revision (ICD-9), diagnosis codes that were determined for the primary cause of admission. Although the rationale was not well described, the following three groups were formed based on reasons for the index hospitalization: epilepsy, stroke, and medical (e.g., cardiac events, pneumonia, or urinary tract infections). Readmissions within 30, 60, or 90 days of the index hospitalization were also correlated with ICD-9 diagnosis codes that were deemed the primary reason for re-hospitalization. Together, the papers assess whether hospitalization for epilepsy is associated with more frequent subsequent hospitalization for psychosis or mood disorder as compared with initial hospitalization for stroke or for other medical issues.
At first glance, the results are striking. The unadjusted hazard ratio for being readmitted with psychosis within a year of being hospitalized for epilepsy was markedly higher than that for stroke (6.58) or for other medical causes (4.41). Similar results were seen for mood disorders. Hazard ratios for the epilepsy group being readmitted for depression were increased as compared with the stroke group (2.80) and the medical group (2.09). Readmissions for bipolar disorder were also higher in the epilepsy group than the stroke group (5.84) or the medical group (2.46).
Perhaps the results are not so surprising. After all, mood and epilepsy have long been associated, so much so that many posit a bidirectional relationship (5). The idea that epilepsy and psychosis overlap has also long been considered. The notion of forced normalization has its roots in that overlap, though technically forced normalization involved seizures and psychosis that alternated rather than overlapped.
With admitted speculation, the findings could be viewed as affirming of the old ideas of forced normalization. Hospitalization for epilepsy could be presumed successful in treating the epilepsy, but then promote the subsequent development or exacerbation of psychotic symptoms, thus leading to a hospitalization for that reason. Such an alternating pattern would not occur with other illnesses, such as stroke or pneumonia. The fact that the most robust risk factor for readmission for psychosis is having psychotic illness identified upon the index hospitalization actually serves to reinforce this idea.
It should be noted that the authors did not make this conjecture for psychosis or for mood and did not even speculate as to causality given that the only data points available were admission statistics and diagnosis codes. However elegant a forced normalization/readmission angle may be imagined; the Rossi et al. study could not confirm it. Yet, in both studies, even the idea that psychosis or mood readmission happens at all may be suspect. The analyses only considered a calendar year of admissions. The readmission rates for a year were only valid for index hospitalizations that occurred relatively early in the year. The data for readmissions within 30, 60, or 90 days were eliminated if the index hospitalization occurred too late in the year to allow those time frames to occur.
Additionally, we have to take issue with some decision-making regarding categorization of epilepsy. The database was created with admission diagnoses. Upon admission, some patients may not have clear and convincing epilepsy that could be coded 345.xx. If the diagnosis code was 780.xx—unspecified convulsions—then they were eliminated from the database. The reasoning was that many of those would prove to not have epilepsy, which could be true, but then broader inclusion criteria were made for psychosis upon subsequent hospitalizations. Unspecified psychosis, as well as established schizophrenia, was included. Presumably, in the hospital course of some of these readmitted patients, psychosis was ruled out, but they would still be deemed readmissions for psychosis because of the initial coding.
Even beyond the diagnosis coding, the samples cannot be reasonably compared because of the simple fact that the epilepsy group is nearly 20 years younger than the stroke or other medical group. Readmissions for medical reasons (i.e., nonpsychosis or nonmood) may be more likely in these older, non-epilepsy groups purely because the older population has already presented with significant life-jeopardizing medical illness. Many of the patients with epilepsy could have been admitted for monitoring or for otherwise less acute causes than stroke or cardiopulmonary events.
Ultimately, the relationship between epilepsy and subsequent hospitalization for psychosis or mood in this dataset may not be particularly robust. Too many confounding variables are present that astute clinicians and researchers may be unable to ignore. Still, with such a massive database and full reporting as done by Rossi et al. and Kim et al., we may glean some important information.
Perhaps the most valuable information from these reports is in the initial baseline characteristics tables (Rossi et al. Table 1 and Kim et al. Table 2). Secondary diagnoses at the index hospitalization are fully presented, and it is clear that not only is psychosis more common in epilepsy than in stroke or other medical causes, but so are depression, bipolar disorder, anxiety, and substance abuse with the exception of tobacco, which is actually markedly less in epilepsy. These findings deserve some consideration. These data alone represent a valuable addition to the literature though they have little to do with reasons for readmission, which themselves may be speculative given the truncated sample and potential for miscoding. The real value of these reports may not be in the mood or the psychosis but may be in the reality that at baseline, epilepsy and psychiatric comorbidity is undeniable, and now is affirmed in one of the largest populations studied to date.