George M. Ghobrial, MD,
Michael J. Lang, MD
Department of Neurosurgery, Thomas Jefferson University Hospital, Philadelphia, PA, USA
Journal Club Article: Rincon F, Rosenwasser RH, Dumont A. The Epidemiology of Admissions of Nontraumatic Subarachnoid Hemorrhage in the United States. Neurosurgery
I. Significance/Context and Importance of the Study:
Aneurysmal subarachnoid hemorrhage (SAH) is the second most common cause of SAH after trauma and is the underlying cause of up to 5% of all new strokes.1,2
Epidemiological studies provide useful data granted they are applicable to the population of interest. In SAH, regional variations in outcomes and incidence have been demonstrated. More recently, the WHO MONICA subgroup study demonstrated significant variation in outcomes and mortality across regions in Europe and Asia.3
Limited epidemiologic data exist specifically for aneurysmal nontraumatic SAH in the US healthcare setting. One review published in 1996 by Lin et al.4
, a meta-analysis of 18 studies over 40 years, found no significant change in the incidence over time. This study is a key step in advancing the quality of care through national education and efficient allocation of healthcare resources. Changes in treatment outcomes over time have significant economic effect given the non-trivial lifetime cost of SAH survival. Finally, this work is significant in that it lends credibility to non-population-based studies providing complementary and valuable information to population-based studies. This manuscript does that through concordant findings with numerous other studies.
II. Originality of the work
Literature pertaining to the epidemiology of the US or North American SAH population is scarce, existing largely in the form of retrospective cohort studies as well as a few meta-analyses while being confined to a single city or single institution in many cases. There are several well-known international studies which have characterized international populations such as Finland and Japan, where SAH has a higher reported incidence, but the main drawback is the difficulty in interpreting outcomes in a setting with such different practice management. Given the vastly different style of healthcare delivery in the US, this epidemiological study is a valuable and relevant tool in that it is the first non-population epidemiologic study of SAH that is applicable to our own healthcare population.
III. Appropriateness of the study design or experimental approach
In an epidemiological study of new admissions and mortality, Rincon and colleagues analyze data from The National Hospital Discharge Survey (NHDS). This study is a 30-year analysis of the NHDS from 1979 to 2008, a function of the National Center for Health Statistics (NCHS). While the NIS is a 20% sample, this represents a 1% sample of patients discharged across 500 registered institutions. Using the ICD-9 code of 430 (SAH) the authors infer that 97% of patients can be ruled in using this previously reported method.5
One main limitation of following complex diseases within a generalized national system of classification is the tendency to oversimplify many complex diseases leading to an artificially lower disease incidence. Using this diagnostic code there is a vague delineation between aneurysmal and nonaneurysmal etiologies of subarachnoid hemorrhage as well as the ill-definition of cerebral vasospasm. Additionally, SAH may be captured by the less specific ICD-9 codes 432, 437, and 438. Isolated intra-parenchymal, or multi-compartmental aneurysmal hemorrhage may not be coded under ICD-9 code 430. It may well be the case that multi-compartmental hemorrhage carries worse prognosis, which may generate significant selection bias. Moreover, a patient admitted with a severe headache, without evidence of subarachnoid hemorrhage on CT scan, will undoubtedly be underreported as a delayed diagnosis of aneurysmal SAH may elude the system of reporting, as the sensitivity is approximately 90% of 3rd generation CT scanners.
The diagnostic algorithm for workup of SAH has changed considerably over thirty years primarily with the routine use of cerebral angiography at specialty referral centers. Routine use of angiography and noninvasive adjuncts to rule-in aneurysms as the cause of nontraumatic subarachnoid hemorrhage have become standard of care. One concerning finding of this and other prior studies is the stable incidence of SAH. With these new diagnostic techniques in greater use than ever, the lack of a rising incidence is puzzling.
While addressed briefly by the authors, care must be taken to avoid confusing trends in in-hospital mortality with overall case mortality, generally reported elsewhere in the literature as 30-day mortality rates. The NHDS sample indicates that significant reduction in hospital stay, associated with a trend toward higher rates of placement in short-term and long-term care facilities, occurred over the study period. Short-term “bounce-backs”, not coded as 430 at the time of readmission, may account for some mortalities that would have previously been captured as in-hospital mortality. As such, the stated reduction from 33.0% to 20.0% in-hospital mortality may not accurately reflect changes in 30-day case mortality over the same period.
There are many sources of bias introduced from sampling a heterogenous population that makes this estimate of SAH admission less representative of the true admission population. The authors go through great lengths to limit this bias. They found no statistically significant variation in representation of hospital admission by size, which would confound the results given that the survival benefit that has been implied by some studies due to aggressive mobilization of multiple specialty resources.5
However, the NHDS represents one consistent method for sampling SAH admissions until 2008, where the number of registered hospitals decreased from 500 to 239.
IV. Adequacy of experimental techniques
The NHDS is similar to the Nationwide Inpatient Sample (NIS), providing a representative database of hospital patients which is calculated annually and reported as a sample percentage. The Agency of Healthcare Research and Quality have compared the two databases finding them similar in most regards, with the exception to cardiovascular disease. Most of the statistics are consistent between the two groups. One notably difference between the NIS and NHDS is the higher overall in-hospital mortality for the NIS. This difference in mortality could be due to the differences in sampling of hospitals, differences in hospital coding schemes, and regional dissimilarities. Furthermore, to limit error, only unweighted patient populations with a desired trait of at least 60 are utilized to limit the standard error, which, given the variance inherent to a sampled population, would be much higher in a small population.
V. Soundness of Conclusions and Interpretation
The authors conclude that the incidence of SAH has remained stable over the past thirty years, given the evidence from the NHDS. They also cite a decrease in deaths and in-hospital mortality along the same time period. For the aforementioned reasons, there is a substantial amount of bias introduce by using this population sampling method. Again, the diagnostic classification system of ICD-9 is not only imperfect, but it’s implementation varies, both by user familiarity, as well as with the number of modifications to the classification system. This supports the need for an improved diagnostic system. Still, the authors validate their findings by citing agreement with previous studies. At the very least, this study does not provide surprising conclusions. Instead, it adds credibility to non-population based findings until a more comprehensive population survey becomes readily available.
VI. Relevance of discussion
The authors find in their database sample a moderately increasing age of admission with a concomitant deceasing in-hospital mortality and total deaths over the 30-year time period. If accurate, this could easily explained to the dramatic technological advancements in surgical methods of aneurysm treatment, as well as highly subspecialized management of this disease by a multidisciplinary approach with neurosurgeons, neurologists, neurointensivists, and highly trained support staff. The authors also find an incidence rate comparable to that of prior population-based cohorts, which has not significantly changed over the years.
The authors are realistic about their limitations of this study and go into detail in this regard, taking care to the conclusions that can be drawn from them. Lastly, they draw support of their work from the concordance of this study with the available literature.
VII. Clarity of writing, strength and organization of the paper
The strength of this manuscript is its novel application of US epidemiological data
to SAH as well as its validation of population based studies on SAH in the US with non-population based sampling data. This manuscript discussion is organized in an ordered, logical manner.
VIII. Economy of words
This manuscript is concise and well-organized. Given the exceedingly low
number of clinical series on the surgical treatment of SAH, a table listing the prior contributions would be helpful for reference. This could be organized by population and non-population based studies, or domestic and foreign studies, for example.
IX. Relevance, accuracy and completeness of bibliography
A MEDLINE search provides a limited number of prior epidemiological studies. The most recent comprehensive population study and meta-analysis was in 2010 by Lovelock et al., ‘Time Trends in Outcome of SAH.’ This, like most other studies, is of a non-US population. The authors quite comprehensively cite all relevant works on epidemiological studies on SAH.
X. Number and quality of figures, tables and illustrations
Again, a table with a summary of the literature would be helpful as this topic is underreported. In the remaining table and two accompanying figures the data is clearly and concisely illustrated for the reader.
XI. Future/next steps the paper logically leads to.
Cerebral Vasospasm is the leading cause of morbidity in SAH. The next interesting question would be to evaluate if hospitals are becoming more effective at limiting morbidity caused by vasospasm, and to determine if there are any changes in cost of treatment over time for treating vasospasm. While this would be interesting, major obstacles to using ICD-9 classification is the lack of diagnostic code for this. One frequent argument that is made with most epidemiologic studies is to provide improved and accurate data to help in allocate adequate resources and assist in future attempts at bundling costs.
1. Suarez JI, Tarr RW, Selman WR. Aneurysmal subarachnoid hemorrhage. The New England journal of medicine. Jan 26 2006;354(4):387-396.
2. Mayberg MR, Batjer HH, Dacey R, et al. Guidelines for the management of aneurysmal subarachnoid hemorrhage. A statement for healthcare professionals from a special writing group of the Stroke Council, American Heart Association. Stroke; a journal of cerebral circulation. Nov 1994;25(11):2315-2328.
3. Ingall T, Asplund K, Mahonen M, Bonita R. A multinational comparison of subarachnoid hemorrhage epidemiology in the WHO MONICA stroke study. Stroke; a journal of cerebral circulation. May 2000;31(5):1054-1061.
4. Linn FH, Rinkel GJ, Algra A, van Gijn J. Incidence of subarachnoid hemorrhage: role of region, year, and rate of computed tomography: a meta-analysis. Stroke; a journal of cerebral circulation. Apr 1996;27(4):625-629.
5. Cross DT, 3rd, Tirschwell DL, Clark MA, et al. Mortality rates after subarachnoid hemorrhage: variations according to hospital case volume in 18 states. Journal of neurosurgery. Nov 2003;99(5):810-817.