SEARCH

SEARCH BY CITATION

Keywords:

  • ABO incompatibility;
  • transfusion;
  • validity;
  • International Classification of Diseases;
  • administrative data;
  • sensitivity

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

Purpose

This paper aimed to systematically review algorithms to identify transfusion-related ABO incompatibility reactions in administrative data, with a focus on studies that have examined the validity of the algorithms.

Methods

A literature search was conducted using PubMed, Iowa Drug Information Service database, and Embase. A Google Scholar search was also conducted because of the difficulty identifying relevant studies. Reviews were conducted by two investigators to identify studies using data sources from the USA or Canada because these data sources were most likely to reflect the coding practices of Mini-Sentinel data sources.

Results

One study was found that validated International Classification of Diseases (ICD-9-CM) codes representing transfusion reactions. None of these cases were ABO incompatibility reactions. Several studies consistently used ICD-9-CM code 999.6, which represents ABO incompatibility reactions, and a technical report identified the ICD-10 code for these reactions. One study included the E-code E8760 for mismatched blood in transfusion in the algorithm. Another study reported finding no ABO incompatibility reaction codes in the Healthcare Cost and Utilization Project Nationwide Inpatient Sample database, which contains data of 2.23 million patients who received transfusions, raising questions about the sensitivity of administrative data for identifying such reactions. Two studies reported perfect specificity, with sensitivity ranging from 21% to 83%, for the code identifying allogeneic red blood cell transfusions in hospitalized patients.

Conclusions

There is no information to assess the validity of algorithms to identify transfusion-related ABO incompatibility reactions. Further information on the validity of algorithms to identify transfusions would also be useful. Copyright © 2012 John Wiley & Sons, Ltd.


INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

The US Food and Drug Administration (FDA) commissioned systematic reviews to identify validation studies of algorithms to identify 20 health outcomes of interest (HOIs) in administrative and claims data (hereafter “administrative data”), as part of its Mini-Sentinel program. These reviews provide the foundation for future studies of HOIs in Mini-Sentinel and other administrative data sources. In such studies, it is extremely important to understand the performance characteristics of the codes that might be used to identify an HOI, as the presence of a code is not always sufficient to determine that an HOI actually occurred. One HOI selected for a systematic review was transfusion-related ABO incompatibility reaction.

Transfusion-related ABO incompatibility reaction is a rare but sometimes catastrophic event. Passive reporting systems have approximated that mistransfusion of blood occurs once for every 13 000 to 19 000 units of blood transfused.[1-3] Rates have been declining with the implementation of processes to ensure that patients are correctly identified, as mistransfusion generally occurs because of a failure to properly identify the patient prior to giving a unit of blood.[3] ABO incompatibility reactions occur when type A or B blood is given to a person with another blood type. The consequence can be a rapid and severe hemolytic reaction that can lead to organ failure and death, or in other more fortunate cases, the reaction can be less severe. Because symptoms can be non-specific and similar to symptoms of other conditions that may be present in a person who receives a transfusion, ABO incompatibility reactions are not always identified.[3] However, if red blood cell hemolysis occurs and a thorough investigation discovers that the wrong blood type was given, the diagnosis can be confirmed with a high level of confidence.

This manuscript provides an overview of the transfusion-related ABO incompatibility algorithm review.The full report can be found at http://mini-sentinel.org/foundational_activities/related_projects/default.aspx.

METHODS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

Details of the methods for these systematic reviews can be found in the accompanying manuscript by Carnahan and Moores.[4] In brief, the base PubMed search was combined with the following terms to represent the HOI: “Blood Group Incompatibility” (Mesh), “ABO incompatible,” and “ABO incompatibility.” Searches of Embase and the citation database of the Iowa Drug Information Service (IDIS/Web) were also conducted. The details of these searches can be found in the full report on the Mini-Sentinel website. The PubMed search was conducted on 22 June 2010, the IDIS/Web search on 2 September 2010, and the Embase search on 24 June 2010. All searches were restricted to articles published in 1990 or after. Mini-Sentinel collaborators were also asked to help identify any relevant validation studies.

The abstract of each citation identified was reviewed by two investigators. When either investigator selected an article for full-text review, the full text was reviewed by both investigators. Agreement on whether to review the full text or include the article in the evidence table was calculated using a Cohen's kappa statistic. If fewer than five studies that performed validation of the algorithm were identified, up to 10 algorithms used without validation were reported. The non-validated algorithms were reviewed to serve as a starting point for identifying the health outcome in administrative data in the absence of validated algorithms and possibly as the foundation of future algorithm validation studies. The data in the tables were extracted by one investigator and confirmed by a second.

Because of the limited number of citations identified for this HOI and the lack of validation studies identified, a Google Scholar search was also conducted on 28 June 2010, using the following search string: “ICD 996.6 transfusion.” This was the final search selected after attempting several searches, because it included all the articles thought to be relevant from other searches. The 996.6 term is the ICD-9 code that represents ABO incompatibility reactions.

RESULTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

Literature searches and reviews

The PubMed search identified 63 citations, the IDIS/Web search 0 citations, and the Embase search 7 citations. The total number of unique citations from the combined searches was 69. The Google Scholar search identified 20 citations.

Of the 69 abstracts reviewed systematically, 15 were selected for full-text review; 35 were excluded because they did not study the HOI, 17 were excluded because they were not administrative database studies, and two were excluded because the data source was not from the USA or Canada. Cohen's kappa for agreement between reviewers on inclusion versus exclusion of abstracts was 0.15. The primary reason for the low agreement was that one investigator was more liberal in including abstracts in hopes that something might be useful, because so few appeared relevant to the HOI.

Of the 15 full-text articles reviewed from the original three searches, none were included in the final evidence tables; two were excluded because they did not study the HOI, and 12 were excluded because they were not administrative database studies. One full-text article could not be obtained through interlibrary loan. However, this article had no abstract and appeared to be a news article as opposed to original research. Reviewers identified one citation for review in full-text article reviews for another HOI (infections related to blood products or tissue grafts).[5] Cohen's kappa for agreement between reviewers on inclusion versus exclusion of full-text articles reviewed was one (all studies from the original set were excluded).

A total of three articles were included in this report. One article included in this report was identified from the full-text reviews for another HOI.[5] The Google Scholar search identified 20 results, which captured the other two articles selected for inclusion in this report.[6, 7] A number of studies that used non-validated algorithms were identified in the Google Scholar search, all of which used ICD-9-CM code 999.6 to identify ABO incompatibility. Most did not focus on this event but included it in a list of quality indicators. Because of this, only one study that provided some uniquely useful information was included in this report.[7] A technical report on methodology for quality indicators was also included because it provided ICD-10 codes that can be used to identify ABO incompatibility.[6]

Summary of algorithms

Little information was found that was useful for determining the validity of algorithms to identify ABO incompatibility reaction from administrative data. This is likely due in part to the rarity of the event.

Scanlon et al.[5] examined the positive predictive value of patient safety indicators in 28 pediatric hospitals with the use of administrative data from 2003 to 2005 (Table 1). They found only seven patients with any one of three codes indicating a transfusion reaction had occurred, one of which was ICD-9-CM code 999.6 (ABO incompatibility reaction). The codes were accurate in all cases, for a positive predictive value of 100%. Two transfusion reactions were present on admission. They noted that it was unlikely that any of the reactions were preventable because they were reactions to antibodies or antigens that were not typed using the best pre-transfusion blood typing tests that were available. This infers that none of the reactions they validated was actually an ABO incompatibility reaction because this blood typing was part of standard pre-transfusion testing.

Table 1. Algorithm validation studies
CitationStudy population and study periodDescription of outcome studiedAlgorithmValidation/adjudication procedure, operational definition, and validation statistics
Scanlon et al. 2008[5]Pediatric (age <18 years) hospital discharges from 28 children's hospitals in the Health Care Utilization Project State Inpatient databases.Transfusion reactionICD-9-CM codes:Medical records were reviewed by clinicians at each of the 28 participating hospitals. They assessed whether the outcome was present on admission by using a set of general and outcome-specific questions developed by experts.
999.6: ABO incompatibility reaction
Only seven patients in the validation study had a code for a transfusion reaction.999.7: Rh incompatibility reactionA transfusion reaction was present in seven of the seven patients with a code from the algorithm. Positive Predictive Value = 100%. However, none of the transfusion reactions identified was an ABO incompatibility reaction.
E8760: mismatched blood in transfusion
The transfusion reactions were present on admission in two of the seven patients, whereas the other five reactions occurred during the hospitalization in question.

A technical report by Drösler[6] was included because it provided some additional codes beyond ICD-9-CM code 999.6 for identifying transfusion reactions. An external cause of injury code, E8760 (mismatched blood in transfusion), was added to the ICD-9-CM code algorithm. Although this is not specific to ABO incompatibility reactions, it may be relevant for studying this HOI. This report also provided corresponding ICD-10-WHO codes for transfusion reactions, which are provided in Table 2.

Table 2. Non-validated algorithms
CitationStudy population and study periodDescription of outcome studiedAlgorithm
Drösler[6]No study population. Technical report on cross-national comparisons of patient safety outcomes.Transfusion reactionICD-9-CM codes
999.6 (ABO incompatibility reaction)
999.7 (Rh incompatibility reaction)
E8760 (mismatched blood in transfusion)
ICD-10-WHO codes
T80.3 (ABO incompatibility reaction)
T80.4 (Rh incompatibility reaction)
Y65.0 (mismatched blood in transfusion)
Morton et al. 2010[7]All discharges in the Healthcare Cost and Utilization Project 2004 Nationwide Inpatient Sample database.Transfusion-related complications. Transfusion codes and non-infectious complication codes are reported here.Transfusion was identified by ICD-9-CM procedure codes 99.0X (various transfusion types) or code V58.2 (blood transfusion without reported diagnosis) in the codes for supplementary classification of factors influencing health status and contact with health services.
Notably, the authors reported that no discharges with an ICD-9-CM code for a non-infectious complication of transfusion were identified, despite studying 2.23 million patients who received transfusions.Non-infectious transfusion-related complications were identified by the following ICD-9-CM codes:
518.7 (transfusion-related acute lung injury)
999.4 (anaphylactic shock caused by serum)
999.5 (other serum reaction)
999.6 (ABO incompatibility reaction)
999.7 (Rh incompatibility reaction)
Other complications described as possibly related to transfusion were identified by the following ICD-9-CM codes:
999.1 (air embolism)
999.2 (other vascular complications)
999.3 (other infection)
999.8 (transfusion reaction not otherwise specified)

Morton et al.[7] examined all discharges in the Healthcare Cost and Utilization Project Nationwide Inpatient Sample database from 2004, including data of 2.23 million patients who received a transfusion (Table 2). The mean age of the transfusion recipients was 66.9 years, and 41.1% were male. Transfusions were identified by ICD-9-CM procedure codes 99.0x or V58.2. It is notable that the 99.0x procedure codes represent transfusion with a number of different types of blood products. The investigators examined a range of transfusion-related outcomes. ABO incompatibility reaction was one of a number of non-infectious transfusion-related complications. Despite the large number of discharges studied, the authors reported that they did not identify a single code for a non-infectious transfusion-related complication, including ABO incompatibility reactions. This is in contrast to the rate of 0.01 transfusion reactions per 1000 hospitalizations identified in pediatric patients.[5] This finding raises doubt about the sensitivity of discharge abstract data in identifying ABO incompatibility reactions as it seems unlikely that not a single transfusion reaction would occur among 2.23 million transfusion recipients. In contrast to this finding, FDA staff report that they have identified ABO incompatibility reaction codes in Medicare data, although they remain rare (personal communication).

DISCUSSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

We were unable to identify any studies that validated algorithms for identifying ABO incompatibility reactions using administrative data. The one study that included ABO incompatibility in an algorithm and performed medical record review of the transfusion reactions identified from administrative data did not actually appear to identify any cases of ABO incompatibility reaction.[5] Another study that examined 2.23 million hospital discharge abstractions in which a transfusion was administered found no codes for any of the non-infectious complications of transfusion that they studied.[7] Even though ABO incompatibility reactions are rare, this raises strong concern about the sensitivity of hospital discharge codes for identifying ABO incompatibility reactions. In contrast, FDA staff noted that they have found cases in Medicare administrative data (personal communication). Future work might explore the rates of ABO incompatibility reactions identified in administrative data to the rates expected on the basis of other research.

International Classification of Diseases code 999.6 identifies ABO incompatibility reactions and was used for all studies identified that examined this outcome. An ICD-10-WHO code for this HOI is also available. A number of other codes that are less specific to this HOI but used to study transfusion reactions are provided in Tables 1 and 2. Transfusions themselves were identified using ICD-9-CM procedure codes 99.0x or V58.2. One study examining the validity of the code typically representing transfusion of allogeneic red blood cells (99.04) found a sensitivity of 83% and specificity of 100% for this code at a single center.[8] Another multi-center study using data from 1987 found a sensitivity of 21% if only three procedure codes were available (as in Medicare data) and 31% when up to 25 procedure codes were available, and a specificity of 100%.[9] Thus, it may be difficult to identify transfusions, despite what appears to be essentially perfect specificity of the code. Revenue codes are another potential source of transfusion data that might be used to augment procedure codes, although the validity of these codes is unknown.[10] Further research is necessary to characterize the performance of algorithms to identify transfusions in administrative data.

It should be noted that ABO incompatibility reactions can occur not only with transfused blood products, but also when mismatched organs are transplanted or when a mother develops antibodies to the fetus blood type. No administrative database studies were identified that studied these types of ABO incompatibility reactions using any alternative algorithms.

New ICD-9-CM codes have also been adopted as of 1 October 2010, which should increase the specificity of the information provided by code 999.6 and other transfusion reaction codes.[11] The new codes for ABO incompatibility reactions include 999.62 (ABO incompatibility with acute hemolytic transfusion reaction, i.e., <24 h post-transfusion) and 999.63 (ABO incompatibility with delayed hemolytic transfusion reaction, i.e., ≥24 h post-transfusion). These codes should be utilized in future research.

Because no studies were identified that validated an algorithm for identifying ABO incompatibility reactions, the possibilities for research on algorithm validity are wide open. One might speculate that the positive predictive value of this code would be high because there is likely little diagnostic ambiguity when a hemolytic reaction occurs after transfusion and it is determined that the wrong blood type was given. An exception might occur in the case of a transcription error in entering codes. Further validation studies, utilizing large population-based administrative databases and medical record review, are needed to establish positive predictive value of the ABO incompatibility codes.

It is not clear that administrative data are sensitive in identifying such reactions. Research to determine the sensitivity of coding algorithms will be difficult because of the rarity of the event. Table 3 reviews potential strategies for future validation studies. Surveying hospitals and blood banks to identify patients who have had ABO incompatibility reactions may be a feasible method for finding cases whose billing claims could then be examined. Review of billing claims of cases reported to the US Centers for Disease Control and Prevention Biovigilance Component, Hemovigilance Module,[12] may offer a good starting point for the process of establishing sensitivity of the codes. Reviewing random charts of patients who received transfusions would likely be too inefficient, even if they were restricted to patients with an intensive care unit stay or some other criterion that might increase the prevalence of these reactions. Another route might be to identify fatal transfusion reactions that have been reported to FDA's Center for Biologics Evaluation and Research, because fatal reactions are reportable, and to determine how these are coded in administrative data. It is likely, however, that the probabilities of submitting codes for fatal and non-fatal transfusion reactions may differ. Along similar lines, it is possible that less severe ABO incompatibility reactions might not be identified in the clinical setting, making them impossible to identify with administrative data. Signs and symptoms of these reactions, particularly if they are delayed or less severe, may be mistaken for other conditions because patients receiving transfusions often have very severe health conditions leading to the need for transfusion.

Table 3. Potential strategies for future ABO incompatibility reaction validation studies
Data sourceMethodValidation statistic obtainedComments
US Centers for Disease Control and Prevention National Healthcare Safety Network Biovigilance Component: Hemovigilance ModuleReview of billing claims of patients with ABO incompatibility reactions reported to the Network.SensitivityDepends on clinical case identification and subsequent reporting to the Network. The Hemovigilance Module to track adverse events associated with blood transfusion was launched recently, in 2010. Participation is voluntary. Unknown whether the number of cases reported to date would be adequate for a precise estimate of sensitivity. The same approach could be applied with any hemovigilance networks or studies large enough to capture a sizable number of ABO incompatibility reactions, as long as billing claims can be accessed.
Survey of hospitals and blood banksSurvey to identify patients who have experienced ABO incompatibility reactions. Review of billing claims to determine whether they include a code for the reaction.SensitivityGiven the rarity of the event, the survey process is likely to be resource intensive. It would require the participants to review their hospitals' billing data for any cases that they report. Hospitals may by default use billing data to identify cases if not clearly directed, falsely inflating the calculated sensitivity, so blood bank records would need to be specified as the case ascertainment source.
Transfusion reactions reported to FDAReview billing claims of patients with ABO incompatibility reactions reported to FDA.SensitivityOnly fatal transfusion reactions are required to be reported to FDA. Less severe cases may not be well-represented. If more severe or fatal cases are more likely to be submitted as claims and to the FDA, this could falsely inflate calculated sensitivity.
Large administrative database with medical record reviewReview medical records of patients with ABO incompatibility reaction diagnosis codes in claims.Positive predictive value (PPV)Could also determine whether the presence of external cause of injury codes changes the PPV. A very large database is likely necessary to identify enough cases for a precise estimate of PPV. Requires health system cooperation in providing potentially legally sensitive medical records.

CONCLUSION

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

This systematic review identified no studies that validated algorithms to identify transfusion-related ABO incompatibility reactions in administrative data. This is a very rare event, so any such study that might be conducted in the future will need a large population from which to draw cases. Newly adopted ICD-9-CM codes that delineate whether an ABO incompatibility reaction is acute or delayed should be incorporated into research that utilizes data from after 1 October 2010. Although the data on algorithm validity are sparse, the algorithms themselves would be quite simple, using a small number of ICD-9-CM codes and possibly determining utility of adding an E-code that represents mismatched blood in transfusion. The sensitivity of any algorithm is unclear. Determining the sensitivity, at least to diagnosed cases, may require starting with a set of known cases identified through surveillance and examining how they were coded in billing data. A small number of studies have attempted to determine the performance characteristics of algorithms to identify allogeneic red blood cell transfusion, but those studies had highly disparate results suggesting a need for more research on this topic using recent data. Overall, there is much potential work to be performed to fully characterize the utility of administrative data for conducting transfusion safety research.

CONFLICT OF INTEREST

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

The views expressed in this document do not necessarily reflect the official policies of the Department of Health and Human Services, nor does mention of trade names, commercial practices, or organizations imply endorsement by the US government.

KEY POINTS

  • No studies provided information on the validity of codes for identifying ABO-incompatibility reactions in administrative and claims data.
  • Several studies used codes specifically representing ABO incompatibility reactions to identify these reactions. One included an external cause of injury code representing mismatched blood in transfusion.
  • There is evidence raising doubt about the sensitivity of administrative data for identifying ABO incompatibility reactions.
  • Two studies reporting on the validity of transfusion codes reported widely varying sensitivity.

ACKNOWLEDGEMENTS

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES

The authors would like to thank Mikhail Menis, Pharm.D., and colleagues at the US Food and Drug Administration's Center for Biologics Evaluation and Research for their input on this report. Mini-Sentinel is funded by the Food and Drug Administration through Department of Health and Human Services (HHS) contract number HHSF223200910006I.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. CONFLICT OF INTEREST
  9. ACKNOWLEDGEMENTS
  10. REFERENCES