SEARCH

SEARCH BY CITATION

Keywords:

  • transfusion;
  • tissue graft;
  • transplant;
  • infection;
  • validity;
  • administrative data

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

To systematically review algorithms to identify infections related to blood products, tissue grafts, or organ transplants in administrative and claims data, focusing on studies that have examined the validity of the algorithms.

Methods

A literature search was conducted using PubMed and the database of the Iowa Drug Information Service. 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

Searches identified one study that examined the validity of an algorithm to identify aspergillosis in transplant recipients and 16 studies that used nonvalidated algorithms to identify infections in recipients of blood products, tissue grafts, or organ transplants. Transfusion was studied as a risk factor for infection, but no studies attempted to identify infection transmitted by any of the exposures under review. Two studies reported sensitivity ranging from 21% to 83% and specificity of 100% of codes to identify allogeneic blood transfusion. No validation studies of algorithms to identify tissue grafts or organ transplant were identified.

Conclusions

There is little evidence to support the validity of algorithms to identify infections related to blood products, tissue grafts, or organ transplants in administrative data or algorithms to identify the exposures. Although it may be possible to validate algorithms to identify the exposures and infectious outcomes, the use of administrative data to identify infections transmitted by these exposures may be challenging. Codes indicating infections acquired through medical care may 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 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 pilot 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. Because of the interest of the US Food and Drug Administration in blood product safety, one HOI selected for a systematic review was infection related to blood products or tissue grafts. Infection related to organ transplants was also included in this systematic review.

Infections related to blood products, tissue grafts, or organ transplants could be broken down into several categories, such as donor-transmitted viral or other infections, bacterial contamination occurring after the product has been collected but before transplant, or infections resulting from a transfusion, graft, or transplant-related procedure. The category of greatest interest for surveillance is when the product is actually infected, such as with bacterial contamination of platelets or an organ infected with human immunodeficiency virus (HIV). Platelets with bacterial contamination have historically been problematic because of room temperature storage, although methods of screening for contamination have improved.[1] The transmission of viruses such as HIV or hepatitis has also become rare because of the universal screening of products.[2-5] However, it is still important to be able to track such infections for surveillance of the effectiveness of screening programs. It would also be useful to track newly emerging infection types potentially transmitted from these sources. Infections in people who have received allogeneic tissue grafts or organ transplants might be related to immunosuppressive medications used to prevent rejection and infected grafts or organs, or they might occur incidentally.[6] Surgical site infections occurring shortly after surgery are a final category that was not covered in the review.

This manuscript provides an overview of the infection related to blood products, tissue grafts, or organ transplant 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.[7] In brief, the base PubMed search was combined with the following terms to represent the HOI: ‘Infection’ and (‘transfusion’ [all fields] or ‘blood transfusion’ [medical subject heading] or ‘graft’ [all fields] or ‘transplant’ [all fields] or ‘transplants’ [medical subject heading]). Searches of Embase and the citation database of the Iowa Drug Information Service were also conducted. The details of these searches can be found in the full report on the Mini-Sentinel Web site. The PubMed search was conducted on 23 June 2010 and the Iowa Drug Information Service search on 10 May 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 were identified that performed validation of the algorithm, algorithms used without validation were reported. The data in the evidence table were extracted by one investigator and confirmed by a second.

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

Review of 402 abstracts led to 183 being selected for full-text review; 97 were excluded because they did not study the HOI, 111 were excluded because they were not administrative database studies, and 10 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 0.32. The major source of disagreement was whether or not a study was an administrative database study. This related to the decision to review single center studies that may have used administrative data-based algorithms using their own billing data, and the difficulty in discerning the likelihood that such an algorithm was used when reviewing abstracts of such studies.

Full-text review of 183 articles led to one article being included in the final evidence table of validation studies; 15 articles were excluded because they did not study the HOI, 143 articles were excluded because they did not use an administrative database, 8 articles were excluded because the HOI identification algorithm was poorly defined, and 16 articles were initially excluded but ultimately included in the reported nonvalidated algorithms because they included no validation of the outcome definition or reporting of validity statistics. Reviewers identified two citations for review from full-text article references, both of which were excluded because the algorithms were poorly defined. Cohen's kappa for agreement between reviewers on inclusion versus exclusion of full-text articles reviewed was 0.66. The one article on which they disagreed included a validation component for hepatitis C diagnosis but excluded people with organ transplantation and did not focus on blood products as an exposure.[8] Thus, it was ultimately excluded.

All of the 17 studies included in the evidence tables were identified from the initial search strategy. None were identified through references of articles that underwent full-text review or by Mini-Sentinel collaborators. Because fewer than five validation studies were identified, 16 studies that did not include validation of the outcome or reporting of validity statistics were reviewed in the evidence table on nonvalidated definitions.

Summary of algorithms

Only one validation study was identified that examined a definition of infection in recipients of a blood product, tissue graft, or organ transplant (Table 1). Chang et al.[9] studied aspergillosis infections among solid organ or hematopoietic stem cell transplant recipients at a large academic medical center from April 2001 to September 2005. This hospital had a registry developed from an active, prospective surveillance system to examine the incidence of fungal infections in solid organ and hematopoietic stem cell transplant recipients. The medical records of registry-identified cases were reviewed to confirm that they met criteria for aspergillosis. The International Classification of Diseases, Ninth Revision (ICD-9), billing codes for aspergillosis identified 67 patients, and 38 patients identified by these codes or the registry ultimately were confirmed as having had probable or suspected aspergillosis. Validity statistics were calculated for a variety of ICD-9 codes and combinations of codes. ICD-9 code 117.3 (aspergillosis) seemed to have the best balance between sensitivity (63%) and positive predictive value (PPV) (71%). ICD-9 code 484.6 (pneumonia with aspergillosis) had a relatively good PPV (88%) but limited sensitivity (37%), which would be expected because it only describes one type of aspergillosis infection. Although the validity of several combinations of codes was examined, all of these combinations included codes less specific to aspergillosis. This resulted in low PPVs for these combinations, despite improved sensitivity. The combination of codes 117.3 and 484.6 might have been expected to achieve a good balance between sensitivity and PPV, but the validation statistics for that combination were not reported. Because this study only included patients from one center, generalizability is limited. In addition, the study methods assumed that the registry had complete ascertainment of aspergillosis cases. This may be a reasonable assumption given that transplant patients should be easily identified and prospective active surveillance for aspergillosis was conducted. ICD-9 codes were also integrated into the active surveillance system after January 2004, so case ascertainment in those patients with billing codes would be expected to be complete after that time. Regardless, no chart review was described for patients identified by billing data who were not in the registry. This is a limitation compared with many studies that perform chart review on all patients identified by billing codes. If the registry did not identify all aspergillosis cases, this might result in increased PPVs and decreased sensitivity for at least some codes.

Table 1. Algorithm Validation Studies
CitationStudy population and time periodDescription of outcome studiedAlgorithmValidation/adjudication procedure, operational definition, and validation statistics
Chang et al.[9]All patients who received solid organ or hematopoietic stem cell transplants (HSCT) at a large academic medical center between April 2001 and September 2005 were included in a cohort with prospective, active surveillance; n = 67 with an ICD-9 code for an aspergillosis infection in the registry were used to calculate validation statisticsInvasive aspergillosis among solid organ transplant and HSCT recipients117.3 (aspergillosis), 117.9 (other and unspecified mycoses), 348.8 (other conditions of the brain: cerebral calcification or fungus), 484.6 (pneumonia with aspergillosis), 484.7 (pneumonia in other systemic mycoses), and 495.4 (malt workers' lung alveolitis due to Aspergillus clavatus)Registry cases of aspergillosis were confirmed using standardized criteria. The medical records of each case in the hospital's registry were independently reviewed by a physician epidemiologist to ensure it met criteria.
Single codes:
117.3: sensitivity = 63% (95%CI = 38%–84%), PPV = 71% (95%CI = 44%–90%)
117.9: sensitivity = 32% (95%CI = 13%–57%), PPV = 15% (95%CI = 6%–31%)
484.6: sensitivity = 37% (95%CI = 16%–62%), PPV = 88% (95%CI = 47%–100%)
484.7: sensitivity = 32% (95%CI = 13%–57%), PPV = 24% (95%CI = 9%–45%)
Code combinations:
117.3 or 117.9: sensitivity 84% (95%CI = 60%–97%), PPV 30% (95%CI = 18–44)
117.3 or 117.9 or 484.6: sensitivity 84% (95%CI = 60%–97%), PPV 30% (95%CI = 18%–44%)
117.3 or 117.9 or 484.7: sensitivity 84% (95%CI = 60%–97%), PPV 28% (95%CI = 17%–41%)
117.3 or 117.9 or 484.6 or 484.7: sensitivity 84% (95%CI = 60%–97%), PPV 28% (95%CI = 17%–41%)

The 16 studies reviewed that provided algorithms but no validation of the HOI covered a broad range of infections (Table 2). Four studies examined any infections or infection-related hospitalizations.[6, 10-12] The remainder studied more specific types of infection. Two studies provided a set of codes to identify coronary artery bypass grafts and transfusions,[11, 12] and one provided a set of codes to identify solid organ transplant recipients.[13] The transfusion algorithm that was provided was validated in an earlier study, which found a sensitivity of 83% and specificity of 100% for ICD-9 procedure code 99.04 (allogeneic blood transfusion).[14] Another multicenter study, however, found poor sensitivity (21%–31%) despite excellent specificity (100%).[15] The reasons for the differences in sensitivity are unclear but may reflect differences in hospital billing practices at the time the studies were completed. Many of the studies with nonvalidated algorithms used the US Renal Data System (USRDS), which specifically prohibits chart review by investigators.[16] Because this is a registry, these studies also did not provide codes to identify renal transplant.

Table 2. Nonvalidated algorithms
CitationStudy population and time periodDescription of outcome studiedAlgorithm
Abbott et al.[16]33,479 primary renal transplant patients in the USRDS between July 1994 and June 1997Bacterial endocarditis within 3 years after first renal transplantationBacterial endocarditis was determined from the presence of ICD-9 codes 421.x as a primary hospital discharge diagnosis. Fungal endocarditis was excluded. First renal transplantation identified via USRDS registry Prior hospitalization for valvular heart disease was examined as a risk factor (ICD-9 codes 394.x-397.x and 424.0x-424.1x)
Abbott et al.[17]28,942 renal transplant patients in the USRDS transplanted between January 1996 and July 2000Urinary tract infections after renal transplantationUrinary tract infection was defined as at least two ICD-9 codes for the following:
590.x (kidney infection, including pyelonephritis both acute and chronic)
595.0 (acute cystitis)
599.0 (urinary tract infection, not otherwise specified)—(code listed incorrectly in manuscript as 599.x)
Cytomegalovirus disease (ICD-9 code 078.x) and sepsis (ICD-9 code 038.x) were examined as confounders
Chavers et al.[10]368,705 incident pediatric and adult dialysis and transplant patients from 1996 to 2001 in USRDSInfection-related hospitalizations in end-stage renal disease and renal transplant recipients within 3 years of initial presentationICD-9-CM codes were obtained from the author, as the appendix mentioned in the manuscript was not otherwise available
Urinary tract infection: 590.x, 595.0–595.4, 597.0–597.89, 599.0, 601.x, 604.x, 607.1, 614.0–616.1, 616.3, 616.4, 616.8
Infection and inflammatory reaction due to other vascular device implant and graft (for hemodialysis associated infection): 996.62
Infection and inflammatory reaction due to peritoneal dialysis catheter: 996.68
Any infection (in addition to previously mentioned codes): 001.x-139.x, 254.1, 320.x-326.x, 331.81, 372.0–372.39, 382.0–382.4, 383.0, 386.33, 386.35, 388.60, 390.x-393.x, 421.x, 422.x, 460.x-466.x, 472.x-474.0, 475.x-477.x, 478.22–478.24, 478.29, 480.x-491.x, 494.x, 510.x, 511.x, 513.x-518.6, 522.5, 522.7, 527.3, 528.3, 540.x-542.x, 566.x, 567.x, 569.5, 572.0–572.2, 573.1–573.3, 575.0–575.12, 611.0, 670.x, 680.x-686.x, 706.0, 711.x, 730.3, 730.8, 730.9, 790.7, 790.8, 730.0–730.2, 997.62, 998.5, 999.3, V01.x-V06.x, V09.x
Dharnidharka et al.[18]870 renal transplant patients from 1996 to 2000, ≤18 years of age, in the USRDS with Medicare as the primary payerUrinary tract infections after renal transplantationThe following ICD-9 codes on one institutional claim or at least two physician supplier claims:
590.x (kidney infection, including pyelonephritis both acute and chronic)
595.0x (acute cystitis)
599.x (urinary tract infection, not otherwise specified)—seems this should have been 599.0x because 599.x includes other conditions.
Dharnidharka et al.[19]28,924 USRDS Medicare primary renal transplant recipients from January 1996 to July 2000Bacterial or viral infections after renal transplantationPrimary discharge ICD-9 diagnoses as follows:
079.x (Unspecified viral infections)
078.5 (cytomegalovirus)
052.9 (varicella)
053.9 (herpes zoster)
038.x (septicemia)
481, 482.9, 486.x (pneumonia)
590.xx, 599.0 (acute pyelonephritis)
682.x (cellulitis)
Hurst et al.[20]23,622 men in USRDS with Medicare primary insurance who received transplants from January 2000 to July 2005 and without a diagnosis of benign prostatic hyperplasia before transplantUrinary tract infection (other noninfectious outcomes related to the prostate were also studied)Urinary tract infection was identified by one institutional claim or two or more physician supplier claims with ICD-9 codes of 590, 590.1, 590.2, 590.8, 590.9, 595, 595.89, 595.9
Johnston et al.[21]5,117 patients in USRDS with Medicare primary insurance who initiated dialysis for failure of a first renal transplant from 1995 to 2004SepsisSepsis was identified by a hospital discharge diagnosis of 038.xx (septicemia), which include the following:
038.0x (streptococcal)
038.1x (staphylococcal)
038.2 (pneumococcal)
038.3 (anaerobic)
038.4x (aerobic Gram negative)
038.8 (other specified septicemia)
038.9 (unspecified septicemia)
Klote et al.[22]15,870 patients in USRDS with Medicare primary insurance who received renal transplants from January 1998 to July 2000TuberculosisTuberculosis was identified by an ICD-9 code of 01x.x
The following were excluded:
647.x (tuberculosis in pregnancy)
V011.x (tuberculosis contact)
V032.x (vaccine for tuberculosis)
V12.01 (personal history of tuberculosis)
137.x (late effect of tuberculosis)
Kutinova et al.[23]44,916 patients in USRDS with Medicare primary insurance who received a first renal transplant from 1995 to 2001Sepsis and pneumonia before and after renal transplantationThe presence of sepsis or pneumonia was determined by at least one inpatient or two outpatient claims for the following ICD-9-CM codes:
038.x (sepsis); 480.x-487.x (pneumonia)
Menzin et al.[24]11,881 high-risk patients with invasive fungal infections and 11,881 matched high-risk controls in the 2004 Healthcare Cost and Utilization Project Nationwide Inpatient Sample. High-risk conditions included cancer, infection with HIV, chronic obstructive pulmonary disease, diabetes mellitus, and solid-organ, hematopoietic stem cell, or bone marrow transplant.Invasive fungal infectionHealthcare Cost and Utilization Project Clinical Classifications Software was used to identify high-risk conditions. The following ICD-9-CM procedure codes identify transplants:
Bone marrow transplant: 41.0x
Kidney transplant: 55.61, 55.69
Other organ transplant: 33.5, 33.50, 33.51, 33.52, 33.6, 37.5, 37.51, 41.94, 46.97, 50.51, 50.59, 52.80, 52.81, 52.82, 52.83, 52.84, 52.85, 52.86
Invasive fungal infection was defined as any primary or secondary diagnosis of one of the following ICD-9-CM codes: 117.3 (aspergillosis), 116.x (blastomycosis), 112.4, 112.5, 112.81, 112.83, 112.85 (candidiasis), 114.0, 114.2, 114.3 (coccidioidomycosis), 117.5 (cryptococcosis), 115.01–115.05, 115.11–115.15, 115.91–115.95 (histoplasmosis), 117.6, 117.9 (other mycoses), 117.7 (zygomycosis)
Neff et al.[25]32,757 patients in USRDS with Medicare primary insurance who received renal transplants from January 2000 to July 2004Pneumocystis jiroveci pneumonia cytomegalovirus was examined as a risk factorThe following ICD-9-CM codes were used in this study: 136.3 (P. jiroveci pneumonia); 078.5 (cytomegalovirus)
Rogers et al.[12]24.789 fee-for-service Medicare beneficiaries who received coronary artery bypass graft (CABG) surgery from 2003 to 2006Infection during hospitalization CABG and transfusion codes were also providedCABG surgery was identified by ICD-9 procedure codes 36.1x
Blood transfusions were identified from procedure codes (99.0x) in addition to revenue codes for blood products and services (38x for purchased blood and 39x for donated blood)
The code for allogeneic red blood cell transfusion (99.04) was previously found to have a sensitivity of 83% and specificity of 100% in a single-center study.[12] Infection codes were not fully described
Rogers et al.[11]9218 Michigan Medicare beneficiaries ≥65 years of age with CABG surgery from July 1997 to 22 September 1998Infection during hospitalization CABG and transfusion codes were also providedCABG surgery was identified by ICD-9 procedure codes 36.1x
Blood transfusions were identified by the previously listed codes used in Rogers et al.[12], which were found to have a sensitivity of 83% and specificity of 100%.
The following ICD-9 primary and secondary codes from hospitalization data were used to identify infections: 001–136, 380.1, 382, 420–422, 440.24, 460–466, 473, 478.29, 480–487, 491.1, 511.1, 513.0, 513.1, 519.2, 528.2, 540, 550.0, 567.0–567.2, 595.5, 569.61, 575.0, 577.0, 590, 595.0, 599.0, 604.9, 616.1, 680–686, 711.0, 711.3–711.9, 728.0, 730.0, 790.7, 996.6, 998.50, 999.3, V08, V09.
Detailed code classifications and definitions can be found in the complete report
Shroff et al.[26]47,899 first time renal transplant recipients and 62,520 renal transplant waiting list patients with Medicare primary insurance in USRDS from 1995 to 2003Hospitalization for bacterial endocarditisBacterial endocarditis was identified by ICD-9-CM code 421.0
Infectious organisms were identified by ICD-9-CM codes 038.xx and 041.xx during the same hospitalization as the bacterial endocarditis claim
Snyder et al.[6]46,471 adult (≥18 years of age) first renal transplant patients in the USRDS from 1995 to 2003 with Medicare primary insurance coverage. Patients younger than 62 years of age were followed for up to 3 years because this is the duration of automatic Medicare coverage posttransplant. Patients 62 years and older were followed for up to 5 years because they would be Medicare eligible when the 3-year term expired.First infection following renal transplantation. The algorithm only includes infections that can be attributed to specific categories of pathogens. Hospitalizations due to urinary tract infections, pneumonia, and other infections that could not be attributed to specific causal agents were included in the study, but ICD-9-CM codes were not provided for these types of infections.ICD-9-CM codes by infection type:
Bacterial
Septicemia: 038.x
Tuberculosis: 010.x-018.x
Other bacterial: 001.x-004.x, 008.0x-008.5x, 020.x-027.x, 030.x-036.x, 039.x-041.x, 073.x, 076.x, 080.x-083.x, 087.x-088.x, 091.x-100.x, 102.x-104.x, 137.x
Viral:
Hepatitis B: 070.2x, 070.3x
Hepatitis C: 070.41, 070.44, 070.51, 070.54
Hepatitis (other): 070.0x-070.1x, 070.42, 070.43, 070.49, 070.52, 070.53, 070.59, 070.6x, 070.9x
Cytomegalovirus: 078.5x
Other viral: 008.6x, 008.8x, 045.x-051.x, 055.x-057.x, 060.x-066.x, 071.x, 072.x, 074.x, 075.x, 078.2x-078.4x, 078.6x, 078.7x, 079.51, 079.52, 079.81
Fungal: 112.0x, 112.4x, 112.5x, 112.81, 112.83, 112.84, 112.85, 112.89, 112.9x, 114.x-117.x
Parasitic:
Pneumocystis: 136.3x
Other parasitic: 006.x, 007.x, 084.x-086.x, 120.x-131.x, 136.2x, 136.4x, 136.5x
Other: 009.x, 101.x, 136.9x
Tong et al.[13]The analysis included everyone in the 2003 Nationwide Inpatient Sample, part of the Healthcare Cost and Utilization Project, which represents approximately 7.5 million hospital stays in a stratified sample of 20% of all US community hospitals. It also included 100% of Medicare Beneficiary hospitalizations for 2003 from the Medicare Provider Analysis and Review file, approximately 11.5 million records.Invasive aspergillosis. Solid organ transplantation codes were also described.Solid organ transplantation hospitalizations were identified by drug related groups 103, 302, 480, and 495
ICD-9-CM codes for complications of transplants were used to identify posttransplant hospitalizations because a drug related group is not available to describe posttransplant status. These included complications of transplanted kidney (996.81), liver (996.82), heart (996.83), lung (996.84), and bone marrow (996.85)
Aspergillosis was identified by a primary or secondary hospital discharge diagnosis ICD-9 code of 117.3. Pneumonia in aspergillosis was identified by ICD-9 code 117.3 with ICD-9 code 484.6 as a secondary diagnosis.

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

Algorithms for the detection of the HOI, infections related to blood products, tissue grafts, or organ transplant, using administrative data are currently of limited value. We identified no validation studies that have looked at administrative data and infections associated with receipt of blood products or tissue grafts. Only one study attempted to validate an algorithm to identify one type of infection, aspergillosis.[9] Most of the nonvalidated algorithms identified examined infections in recipients of organ transplants or coronary artery bypass grafts, and some examined transfusion as a risk factor for infection but not a source of infection. Even in the clinical setting, it may be difficult to confirm that an infection is transmitted by a blood product, tissue graft, or organ transplant. Recipients often have many risk factors for infection. Validation studies could be performed to examine the agreement of infection diagnoses in administrative data and clinical data, but the challenge of linking the infection to the transfusion or graft would still remain. The receipt of a blood product, tissue graft, or organ transplant and the concurrent presence of infection provide no clear link between the exposure and outcome. There are International Classification of Diseases, Ninth Revision, Clinical modification (ICD-9-CM) codes representing complications of medical care that might be considered to help increase the specificity of algorithms to identify infections transmitted by these sources. These include 996.6 (infection and inflammatory reaction due to internal prosthetic device implant and graft) and its subcodes, 999.3 (other infection due to medical care not elsewhere specified), and 999.39 (infection following other infusion, injection, transfusion, or vaccination). The code 999.32 has also been proposed to represent transfusion-transmitted infection, which will likely be very useful for increasing the specificity of future algorithms if it is adopted. We cannot comment on the performance of algorithms that would include these codes, however, given the lack of evidence.

The only study that validated an infection outcome was one that looked at aspergillosis in hematopoietic stem cell transplants and solid organ transplants.[9] The sensitivity of ICD-9 code 117.3 was described as modest (63%) and the code 117.9 was described as poor (32%). If either code was present, the sensitivity was 84%, but the positive predictive value was 30%, which would likely be considered unacceptable without subsequent medical record review to identify true cases. Even this study was limited by the fact that it was a single center study that is difficult to generalize and it was unlikely that these infections were transmitted by the transplanted tissue or cells. These infections would have been secondary to the necessary immune suppression that these transplant recipients must receive. The other nonvalidated studies had similar limitations in that they detected infections because of immune suppression and not receipt of the medical product of interest.

One general concern of the administrative data approach to surveillance is the rarity with which the infections likely to be studied are actually transmitted by a blood product, tissue, graft, or organ transplant. Many infections historically linked to these sources are now regularly screened for at donation. Thus, the likelihood of transmission of viral hepatitis or HIV from these sources is very rare and declining.[2-5] In addition, the time from transplant to infection recognition is long, and thus it would be difficult to link the two. For some blood products such as platelets, there are specific organisms for which codes might be included in algorithms because they are known to be common contaminants. The most common organisms have been reviewed elsewhere.[27] Other infections linked to receipt of blood products or tissue grafts are often of microbiologically very rare organisms. One example was the recent report of transplant-associated encephalitis from the donor to the two kidney recipients of Balamuthia granulomatous amebic encephalitis, a rare disease caused by Balamuthia mandrillaris, a free-living ameba found in soil.[28] This was identified by physician self-report to the Centers for Disease Control and Prevention. Recent cases of rabies transmission via solid organ transplantation are another example of an extremely rare event that would not be detected via administrative data but by other reporting mechanisms.[29] Recent efforts at mandatory reporting have identified many of these cases for investigation by public health experts, but this can only occur after astute recognition by involved clinicians.

We also identified little information on the validity of codes for the exposures examined in this review. The code for allogeneic blood transfusion was found to have an acceptable sensitivity (83%) perfect specificity in one single center study[14] but poor sensitivity (21%–31%) despite excellent specificity in another multicenter study from the late 1980s.[15] No validation studies of codes for tissue grafts or organ transplantation were referenced by the studies reviewed in this report. Many used the USRDS, which includes registry data on renal transplant recipients. Review of medical records of patients identified in USRDS is prohibited, so diagnoses cannot be validated.

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

There is little evidence to support the validity of algorithms to identify infections in recipients of blood products, tissue grafts, or organ transplants using administrative data. No studies were identified that attempted to validate algorithms to identify infections transmitted by these sources. Two studies examined the performance of algorithms to identify allogeneic red blood cell transfusions, but their results were inconsistent. No information was found to validate algorithms to identify recipients of other types of transfusions, grafts, or organ transplants. Future research might particularly attempt to construct algorithms that would identify infections transmitted from these sources. However, given the challenge in linking an infection with one of these sources in the clinical setting, it may prove very difficult to link them in administrative data. Several codes indicating infections related to medical care are available, and another has been proposed, which may improve the specificity of such an algorithm.

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 authors report no conflicts of interest related to this work.

KEY POINTS

  • The only validation study identified found variable validity among a set of codes to identify aspergillosis in transplant recipients.
  • No study, even those with nonvalidated algorithms, attempted to use administrative data to identify infections specifically transmitted by a blood product, tissue graft, or organ transplant.
  • Confirming that the source of infection was a blood product, tissue graft, or organ transplant is likely to be difficult using administrative data, although codes are available and have been proposed that may increase the specificity.
  • This review identified only two studies that validated algorithms to identify allogeneic blood transfusion and no studies that validated algorithms to identify receipt of other blood products, tissue grafts, or organ transplants.

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

Mini-Sentinel was funded by the US Food and Drug Administration through the Department of Health and Human Services Contract Number HHSF223200910006I. 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.

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