Free
Correspondence  |   August 2009
Administrative Databases: Are They Useful for Clinical Analyses?
Author Affiliations & Notes
  • Eric B. Rosero, M.D.
    *
  • *University of Texas Southwestern Medical Center, Dallas, Texas.
Article Information
Correspondence
Correspondence   |   August 2009
Administrative Databases: Are They Useful for Clinical Analyses?
Anesthesiology 8 2009, Vol.111, 450-451. doi:10.1097/ALN.0b013e3181ae0742
Anesthesiology 8 2009, Vol.111, 450-451. doi:10.1097/ALN.0b013e3181ae0742
In Reply:—  We would like to thank Dr. Memtsoudis and Larach et al.  for their comments regarding our study. We welcome the opportunity to address the criticisms and concerns expressed by these authors.
First, Dr. Memtsoudis speculates that our study may have overestimated the number of malignant hyperthermia (MH) episodes, because some patients with the diagnosis may have been transferred between hospitals. However, this possibility is unlikely. The National Inpatient Sample is a stratified probability sample of hospitals in the United States, and the sampling strata are based on five hospital characteristics (geographic region, urban or rural location, teaching status, bed size, and hospital control). Within each stratum, a particular hospital has a 0.2 probability of being selected in the database, which decreases the chances of an MH case being captured multiple times in the database. We feel that this low probability decreases the chances of an MH case being captured multiple times in the database and, therefore, disagree that interhospital transfers may have impacted our estimates in any substantial way. Furthermore, information about vital status of patients at discharge (whether or not they died during hospitalization) is quite accurate in the National Inpatient Sample, and is not affected by transfers between facilities. Thus, if the incidence of MH was lower than that reported in our study, then the mortality rate from MH would be even higher than we found, leading to the same conclusion that current mortality from MH in the United States is higher than that previously reported.
We attempted to address the concern raised by Dr. Memtsoudis regarding the use of weighted or unweighted data in our manuscript, where we stated that our results were based on weighted analyses of the database, taking into account the sampling design of the National Inpatient Sample and using the statistical tools available for the analysis of stratified samples. Finally, we agree with the limitations raised by Dr. Memtsoudis related to the administrative nature of the NIS database, and discussed them in our manuscript. Nevertheless, the benefits of using such databases are well recognized and should not be overlooked.
The main concern of Larach et al.  is that our study may have included a significant number of cases erroneously coded as MH episodes. We acknowledge that case ascertainment is a major source of bias in studies using administrative databases. Accordingly, we excluded patients with diagnosis codes of other conditions associated with hyperthermia. While we may have missed some cases, we eliminated many that were most likely miscoded. However, given that information to confirm the diagnosis of MH and the type of anesthesia are not available in the database, our study has still a potential for overreporting. Although the median age of our sample was only 39 yr and more than 90% of the patients had a low comorbidity index, we found an overall mortality rate of 11.7% associated with a diagnosis of MH, which exceeds the in-hospital mortality reported for serious conditions affecting even older and sicker populations.1–3 This high mortality in a low-risk population supports the idea that our estimates about the incidence and fatality rate of MH are not distant from the true values in the general population. Furthermore, although the cases were selected based just on a diagnosis code, coding by medical records departments depends on information provided by clinicians. Therefore, as we stated in our discussion, our results underscore the magnitude of the clinical problem, given that patients with a diagnosis of suspected MH should be treated as MH-susceptible until proven otherwise.
Although there are limitations to our study, we disagree with the letter authors in that we believe these data support an increase in the incidence of MH. Although one of the causes of increasing incidence could be increased awareness of the MH code, we do not believe that this issue had an important impact on the trend of our observations. MH is so rare that during the five-year study period, each coding department of the more than 5,000 hospitals of the NIS universe was exposed on average to only 0.5 MH cases. Accordingly, the hypothesis that the coders became aware of the MH code seems to be baseless. Furthermore, to minimize this bias, we excluded from the study data on the first 3 yr (1997 to 1999) in which the diagnosis of MH was available in the International Classification of Diseases, 9th Revision, Clinical Modification. 
We agree with Larach et al.  that the reports to North American Malignant Hyperthermia Registry provide excellent information, and any cases of MH must be reported to the Registry. In addition, the efforts of American Society of Anesthesiologists and Malignant Hyperthermia Association of the United States in obtaining the approval for MH coding are well recognized and commendable. In fact, our manuscript does not suggest that readers stop reporting to the Malignant Hyperthermia Association of the United States registry. Despite their limitations, administrative databases provide valuable information, and it is our belief that information from the administrative databases and registries complements each other and neither should be excluded as we try to better understand MH. We acknowledge the error in the reference on the introduction of our paper, which should make reference to the 5% mortality rate cited by European reports, and not to the rate reported by the North American Malignant Hyperthermia Registry study. Nevertheless, the MH-associated mortality rate remains controversial.
*University of Texas Southwestern Medical Center, Dallas, Texas.
References
Jamieson GG, Mathew G, Ludemann R, Wayman J, Myers JC, Devitt PG: Postoperative mortality following oesophagectomy and problems in reporting its rate. Br J Surg 2004; 91:943–7Jamieson, GG Mathew, G Ludemann, R Wayman, J Myers, JC Devitt, PG
Chiappini B, Camurri N, Loforte A, Di Marco L, Di Bartolomeo R, Marinelli G: Outcome after aortic valve replacement in octogenarians. Ann Thorac Surg 2004; 78:85–9Chiappini, B Camurri, N Loforte, A Di Marco, L Di Bartolomeo, R Marinelli, G
Scarborough JE, Pietrobon R, Tuttle-Newhall JE, Marroquin CE, Collins BH, Desai DM, Kuo PC, Pappas TN: Relationship between provider volume and outcomes for orthotopic liver transplantation. J Gastrointest Surg 2008; 12:1527–33Scarborough, JE Pietrobon, R Tuttle-Newhall, JE Marroquin, CE Collins, BH Desai, DM Kuo, PC Pappas, TN