Infection Control & Hospital Epidemiology http://journals.cambridge.org/ICE Additional services for Infection
Control & Hospital Epidemiology:
Email alerts: Click here Subscriptions: Click here Commercial reprints: Click here Terms of use : Click here
Matching Bacteriological and Medico-Administrative Databases Is Efcient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital Brice Leclère, Camille Lasserre, Céline Bourigault, Marie-Emmanuelle Juvin, Marie-Pierre Chaillet, Nicolas Mauduit, Jocelyne Caillon and Matthieu Hanf Infection Control & Hospital Epidemiology / Volume 35 / Issue 11 / November 2014, pp 1330 - 1335 DOI: 10.1086/678422, Published online: 16 January 2015
Link to this article: http://journals.cambridge.org/abstract_S0195941700094558 How to cite this article: Brice Leclère, Camille Lasserre, Céline Bourigault, Marie-Emmanuelle Juvin, Marie-Pierre Chaillet, Nicolas Mauduit, Jocelyne Caillon and Matthieu Hanf (2014). Matching Bacteriological and Medico-Administrative Databases Is Efcient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital. Infection Control & Hospital Epidemiology, 35, pp 1330-1335 doi:10.1086/678422 Request Permissions : Click here
Downloaded from http://journals.cambridge.org/ICE, IP address: 132.174.255.116 on 05 Nov 2015
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
original article
Matching Bacteriological and Medico-Administrative Databases Is Efficient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital Brice Lecle`re, MD, MSc;1 Camille Lasserre, PharmD;1,2 Ce´line Bourigault, PharmD, MSc;1 Marie-Emmanuelle Juvin, PharmD;1 Marie-Pierre Chaillet, MD;3 Nicolas Mauduit, MD;3 Jocelyne Caillon, PharmD, PhD;1,2 Matthieu Hanf, PhD;4 Didier Lepelletier, MD, PhD1,2 for the SSI Study Groupa
objective. Our goal was to estimate the performance statistics of an electronic surveillance system for surgical site infections (SSIs), generally applicable in French hospitals. methods. Three detection algorithms using 2 different data sources were tested retrospectively on 9 types of surgical procedures performed between January 2010 and December 2011 in the University Hospital of Nantes. The first algorithm was based on administrative codes, the second was based on bacteriological data, and the third used both data sources. For each algorithm, sensitivity, specificity, and positive and negative predictive values (PPV and NPV) were calculated. The reference method was the hospital’s routine surveillance: a comprehensive review of the computerized medical charts of the patients who underwent one of the targeted procedures during the study period. setting.
A 3,000-bed teaching hospital in western France.
population.
We analyzed 4,400 targeted surgical procedures.
results. Sensitivity results varied significantly between the three algorithms, from 25% (95% confidence interval, 17–33) when using only administrative codes to 87% (80%–93%) with the bacteriological data and 90% (85%–96%) with the combined algorithm. Fewer variations were observed for specificity (91%–98%), PPV (21%–25%), and NPV (98% to nearly 100%). Overall, performance statistics were higher for deep SSIs than for superficial infections. conclusions. A reliable computer-enhanced SSI surveillance can easily be implemented in French hospitals using common data sources. This should allow infection control professionals to spend more time on prevention and education duties. However, a multicenter study should be conducted to assess the generalizability of this method. Infect Control Hosp Epidemiol 2014;35(11):1330-1335
Standard data from hospital information systems are easily available, and their management can be rapidly completed by the significant progress in computer technology in the past decades. This allows many aspects of routine surveillance of healthcare-associated infections (HCAIs) to be enhanced or automatized, including cluster detection, quality assessment, and risk factor identification.1 Infection control professionals may therefore spend their time more efficiently, focusing on prevention and education. However, even if these electronic surveillance systems (ESSs) can outperform routine manual surveillance, their use is not yet widespread.2,3 Most of the
published electronic HCAI identification methods differed in the targeted infections, databases, or identification algorithms. Their development according to each hospital information system is thus not transferable to other healthcare centers. Without a standardized and sustainable method, several useful functions of HCAI surveillance such as comparisons between institutions and evaluations of control measures cannot benefit from the genuine support offered by ESSs. In France, as in several other Western countries, monitoring surgical site infection (SSI) rates for selected proce-
Affiliations: 1. Department of Bacteriology and Infection Control, Nantes University Hospital, Nantes, France; 2. EA3826 The´rapeutiques Cliniques et Expe´rimentales des Infections, Medical School, University of Nantes, Nantes, France; 3. Department of Medical Information, Nantes University Hospital, Nantes, France; 4. Clinical Investigation Center, Nantes University Hospital, Nantes, France; a. Members of the SSI Study Group are listed at the end of the text. Received May 7, 2014; accepted July 15, 2014; electronically published September 30, 2014. 䉷 2014 by The Society for Healthcare Epidemiology of America. All rights reserved. 0899-823X/2014/3511-0002$15.00. DOI: 10.1086/678422
computer-enhanced ssi surveillance
dures is considered a key part of HCAI control programs and a useful benchmarking parameter.4 Moreover, SSIs are associated with substantial morbidity and mortality and have an important impact on patients and public health. Thus, they represent a pertinent target for a computer-assisted surveillance method. The aim of this study was to evaluate the performance of a generally applicable method for SSI detection using different databases available routinely in the medical information system of the Nantes University hospital during a 2-year period, 2010–2011.
methods Routine Surveillance Our study was conducted at the University Hospital of Nantes, a 3,000-bed hospital located in western France. Nine surgical procedures have been monitored to detect SSIs during a 2-year period from January 2010 to December 2011: coronary artery bypass grafting (CABG), valve replacement, elective (nonurgent) colectomy with immediate restoration of intestinal continuity, primary total hip arthroplasty (PTHA), primary total knee arthroplasty (PTKA), kidney transplantation, urinary sphincter implantation, neurostimulator implantation, and cesarean section (C-section). Surgical patients were identified by the French Common Classification of Medical Procedure (CCMP) codes that are recorded in the hospital’s medical activity database.5 A list of patients was monthly transmitted to the infection control team (ICT) by the medical information department. Then, infection control nurses or practitioners systematically reviewed the computerized medical records to detect SSIs (operative report, medical discharge letter, bacteriological results, follow-up appointment notes, etc). Each suspicion of SSI needed to be confirmed by the infection control practitioners (ICPs) or by the surgeon who operated on the patient. This surveillance was performed within 30 days, or 1 year for prosthetic surgery, using definitions from the National Healthcare Safety Network.6 We considered this routine surveillance the reference method (criterion standard) for performance assessment of 3 surveillance algorithms using 2 electronic data sources: bacteriological analysis results and medical activity. Data Sources In France, hospital medical activity is monitored using the Medical Information Program System (PMSI), a mandatory computerized medical database used for budget allocation. For each hospitalization, surgical procedures and relevant diagnoses are recorded using respectively the CCMP and the 10th revision of the International Classification of Diseases (IDC-10). At the University Hospital of Nantes, these medical codes were assigned by the surgeons, and quality control on the database was performed by medical coding specialists. The laboratory activity was also computerized, and the type,
1331
date, and culture results of every sample were recorded in the laboratory’s database. After selecting the targeted procedures with the appropriate CCMP codes and time frame (2010–2011), the two data sets were merged using a specific patient number as the key. This patient number is systematically and definitively assigned to each patient at the first admission in the hospital. Electronic Surveillance Algorithms Three algorithms were tested to retrospectively identify potential SSIs. Algorithm 1 was built on variables available in the PMSI database. According to the surgical procedures, relevant ICD-10 codes were selected to detect potential SSI cases (Table 1). Algorithm 2 was performed using the bacteriology laboratory database. Depending on the “sample type” variable (blood, stool, urine, biopsy, deep drainage, etc), samples were classified as “deep” or “superficial” for the sake of the analysis. The markers of potential SSIs were positive cultures from superficial samples (ie, cultures with abnormal bacterial growth) and requests for deep sample bacteriological examination, regardless of the results. For urinary sphincter implantation and kidney transplantation, simultaneously positive blood and urine cultures were seen as a sign of possible pyelonephritis and used as a marker as well. To be considered valid, the samples had to be taken within 40 days of the procedure, or within 400 days for prosthetic surgery procedures; these small margins were added to standard followup times (respectively 30 and 365 days) to take into consideration the delay between clinical signs and subsequent hospitalizations or revision procedures. Algorithm 3 was performed by combining algorithms 1 and 2. They were implemented by the ICT using STATA 10.0 software. Performance Analysis Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated with the corresponding 95% confidence interval for each of the three algorithms. Specific statistics were also estimated according to the SSI’s depth.
results Between January 2010 and December 2011, 4,400 targeted procedures were reviewed by the ICT using medical records and the CCMP codes. The number of surgical procedures varied according to the type of surgery, from 149 for neurostimulator implantation to 2,259 for CABG and valvular replacement in cardiac surgery (Table 2). The routine surveillance identified 112 SSIs, with an overall incidence rate of 2.5%. SSIs were defined as superficial SSIs (26%) or deep SSIs (74%; Table 2). The deep SSI incidence rates were 0% for C-section, 1.2% for cardiac surgeries, 1.4% for PTKA, 1.7% for urinary sphincter implantation, 2.0% for neurostimulator implantation, 2.2% for PTHA, 4.6% for elective colectomy, and 8.9% for kidney transplantation.
1332
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
table 1. International Classification of Diseases, Tenth Revision (ICD-10), Codes Used for Surgical Site Infection Detection according to the Targeted Procedures Surgery procedure and ICD-10 code Cesarean section T81.4 O86.0 Neurostimulator implantation T81.4 T85.7 Urinary sphincters implantation and kidney transplantation T81.4 T83.5 Coronary artery bypass grafting and valve replacement T81.4 T81.38 T82.7 T82.6 Colectomy T81.4 K65.0 K65.8 K65.9 Primary total knee or hip arthroplasty T81.4 T81.38 T84.5 T84.7
Description Infection following a procedure Infection of obstetric surgical wound Infection following a procedure Infection and inflammatory reaction due to other internal prosthetic devices, implants, and grafts Infection following a procedure Infection and inflammatory reaction due to prosthetic device, implant, and graft in urinary system Infection following a procedure Disruption of operation wound Infection and inflammatory reaction due to other cardiac and vascular devices, implants, and grafts Infection and inflammatory reaction due to cardiac valve prosthesis Infection following a procedure Acute peritonitis Other peritonitis Peritonitis, unspecified Infection following a procedure Disruption of operation wound Infection and inflammatory reaction due to internal joint prosthesis Infection and inflammatory reaction due to other internal orthopedic prosthetic devices, implants, and grafts
Sensitivity varied significantly between algorithms 1, 2, and 3 from 25% to 87% and 90%, respectively (P ! .001; Table 3). Variations for specificity (91%–98%), PPV (21%–25%), and NPV (98% to nearly 100%) were not significant. Overall, predictive values were higher for deep SSIs than for superficial SSIs. When bacteriological data were used (algorithms 2 and 3), higher sensitivities and specificities were also observed for deep SSIs, compared to superficial SSIs.
discussion An accurate and wholly automated surveillance system for HCAIs is an enticing prospect for ICPs but is not likely to be available soon.7 Indeed, the definition of most HCAIs, and especially SSIs, relies on subjective clinical observations that still require human validation.8-10 Depending on the objectives of the surveillance—cluster detection, incidence trends analysis, risk factor identification, and so on—different characteristics and performances might be relevant.1 The objective of this study was to develop an ESS that detects possible cases of SSI and reliably excludes noncases, in order to minimize the surveillance-related workload of infection control professionals. In this perspective, the number of false-negative cases should be kept as low as possible, and high sensitivity and NPV are thus the most relevant characteristics.11 The use
of bacteriological data with a nonrestrictive algorithm allowed such performance features: the overall sensitivity reached 87%, and the NPV was close to 100%. Among the 83 deep SSIs detected by the routine surveillance (criterion standard), 80 were also detected by using the bacteriological database (algorithm 2) with a very good sensitivity. The three falsenegative cases were patients with deep SSI (abscess) after colectomy. They were treated by intravenous antibiotherapy without bacteriological sample and no surgical reoperation. Overall, PPVs were inferior to 25%, mostly because of the large number of cases falsely detected using any of the algorithms. False-positive cases were due to diagnosis coding mistakes, infections occurring outside the scope of the surveillance (not related to surgery or not related to 1 of the targeted procedures), or unconfirmed SSI suspicions. It seemed clear that none of the tested algorithms could reliably be used for automatic detection of SSIs and that, consequently, ICPs needed to keep reviewing charts to validate SSI suspicions. The workload related to this task could, however, largely be reduced: using algorithm 2 allowed the number of medical charts to review to decrease by 90% (452 vs 4,400 files). So at the cost of a 4% loss of true SSIs due to false-negative cases (sensibility of 96%), this algorithm could allow ICPs to
computer-enhanced ssi surveillance
1333
table 2. Number of Surgical Site Infections (SSIs) Detected by Routine Surveillance and Number of Possible Infections Using 2 Different Databases at the Nantes University Hospital (2010–2011) Routine surveillance, gold standard Surgical procedure type CABG and valve replacement Colectomya PTHA PTKA Kidney transplantation Urinary sphincter implantation Neurostimulator implantation C-section Total
Algorithms
Surgical procedure no.
Superficial SSIs
Deep SSIs
1
2
3
2,259 196 506 278 293 179 149 540 4,400
0 3 0 0 4 2 6 14 29
27 9 11 4 26 3 3 0 83
47 33 8 4 13 1 0 4 113
270 28 48 17 57 9 11 11 451
294 29 49 17 63 10 11 11 484
note. Routine surveillance was performed using all data available from medical and surgical records and the French Common Classification of Medical Procedure (CCMP) with a patient follow-up of 1 year when a prosthetic implantation was done. Algorithm 1 was performed by matching the data collected with the 10th revision of the International Classification of Diseases codes and the CCMP codes; algorithm 2 was performed by matching the data collected by the laboratory database and the CCMP codes; algorithm 3 was performed by combining algorithms 1 and 2. CABG, coronary artery bypass grafting; C-section, cesarean section; PTHA, primary total hip arthroplasty; PTKA, primary total knee arthroplasty. a Colectomy with immediate restoration of digestive continuity, excluding emergency situations.
spend their time more efficiently, focusing on prevention and education or expansion of the range of targeted procedures. We believe that this may be a valid trade-off. Previous studies assessing the relevance of microbiological data to identify SSIs showed lower sensitivity measures: 59% for CABGs12 and 84% for gastrointestinal surgery.13 The detection rules used in these studies were more specific, as only patients with positive bacteriological exams were selected, hence excluding possible culture-negative SSIs.14 Moreover, algorithm 2 is probably dependent on wound culturing routine practices, particularly for superficial SSIs, which may also explain these differences. If the bacteriological database is available and the strategy of wound culturing well defined, the surveillance method using algorithm 2 can be easily implemented in all the French hospitals with a monthly case review (surveillance currently conducted at the Nantes University Hospital, with additional surgical procedures such as plastic surgery, mammary surgery, thoracic surgery, and vascular prosthesis surgery). The method relying solely on diagnosis codes through PMSI data had less interesting performance features. Our results showed that sensitivity, especially, barely reached 25%, which means that 3 out of 4 SSIs were missed. Many studies have showed that discharge diagnosis codes alone are not a reliable data source for HCAI surveillance.14-18 However, as a complementary data source, it can help to identify more potential SSIs. In our study, the combined algorithm’s sensitivity and NPV were higher than those observed using only the bacteriological database. These differences were not statistically significant, but this can be explained by a lack of statistical power, the number of SSIs following the monitored procedures being rather low. These results highlight the need
for better coding of SSIs diagnosis in the French PMSI database. Indeed, Gerbier et al18 showed that using unspecific ICD-10 codes for infections and postoperative complications can increase the sensitivity of a PMSI-based SSI detection method for general surgery, from 26% to 79%. In orthopedic surgery, a recent study has showed that a specific algorithm including all possible codes defined with surgeons ensures optimal detection of SSIs with good performance.19 In France, diagnosis coding methods may differ between healthcare facilities, and it might be relevant and necessary to adapt the PMSI algorithm to each center, according to its coding habits. Studies on HCAI detection systems tend to emphasize the importance of using multiple data sources.3 Some of them, such as computerized antibiotic prescriptions and text mining in medical charts are promising20 and can partly offset the dearth of systematically computerized clinical data. In our study, for example, most of SSI cases that were not captured by the bacteriological data were superficial SSIs, diagnosed in outpatients and for which no microbiological diagnosis was required. In these situations, text mining in clinical notes and records might be relevant. However, these data sources are not yet widespread and thus cannot be included in a surveillance system designed to be applicable in every French hospital. SSIs diagnosed after discharge have always been a possible source of selection bias for SSI surveillance, especially since some of these cases are diagnosed and treated in a different hospital than the one in which the original surgical procedure was performed. This, however, is true for both manual and computerized surveillances, and up to the present, there is no valid solution to this problem,1,21 even though regional or
1334
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
table 3. Performance Analysis for Electronic Surveillance of Surgical Site Infections (SSIs) according to Data Sources and Infection Depth, Nantes University Hospital (2010–2011) Algorithm and SSI depth
Sensitivity
Specificity
Positive predictive value
Negative predictive value
Superficial Deep All
24.1 (8.6–39.7) 25.3 (15.9–34.7) 25 (17.0–33.0)
97.6 (97.1–98) 97.9 (97.4–98.3) 98 (97.6–98.4)
6.2 (1.8–10.6) 18.6 (11.4–25.8) 24.8 (16.8–32.7)
99.5 (99.3–99.7) 98.6 (98.2–98.9) 98 (97.6–98.5)
Superficial Deep All
58.6 (40.7–76.5) 96.4 (92.4–99.2) 86.6 (80.3–92.9)
90.1 (89.2–91.0) 91.4 (90.6–92.2) 91.7 (90.9–92.6)
3.8 (2.0–5.5) 17.7 (14.2–21.3) 21.5 (17.7–25.3)
99.7 (99.5–99.9) 99.9 (99.8–100) 99.6 (99.4–99.8)
Superficial Deep All
65.5 (48.2–82.8) 98.8 (96.4–100) 90.2 (84.7–95.7)
89.4 (88.4–90.3) 90.7 (89.8–91.6) 91.1 (90.2–91.9)
3.9 (2.2–5.7) 16.9 (13.6–20.3) 20.9 (17.2–24.5)
99.7 (99.6–99.9) 100 (99.9–100) 99.7 (99.6–99.9)
1
2
3
note. Surgical patients were first selected by using the French Common Classification of Medical Procedure (CCMP). Algorithm 1 was performed by matching the data collected with the 10th revision of the International Classification of Diseases codes and the CCMP codes; algorithm 2 was performed by matching the data collected by the laboratory database and the CCMP codes; algorithm 3 was performed by combining algorithms 1 and 2. 95% confidence intervals are presented in parentheses.
national health insurance data might be interesting sources of information.22 No superficial infections were reported for the targeted procedures of cardiac and orthopedic surgery. This was probably due to the fact that such infections are generally benign and sometimes difficult to diagnose with certainty, and, thus, they are probably rarely mentioned in medical records or coded in the PMSI database, and superficial samples might not usually seem relevant. Therefore, both routine and enhanced surveillances probably lack sensitivity for this kind of infection. Our study sample was determined using the surgical procedure codes registered in the hospital information system. A selection bias might therefore have occurred in case of incompleteness of these data, which could have skewed our estimations. However, unlike diagnosis coding, surgical procedure coding is straightforward, easily recordable, and verifiable by medical coders, and, most importantly, it is necessary for the right pricing of hospital stays in surgery. Therefore, a significant lack of exhaustiveness of surgical procedure coding seems unlikely. Regarding these limits, the reference method (routine surveillance) was probably not perfect (variability in diagnosing SSI between surgeons, infection control nurses or practitioners, no validation by an independent reviewer), but it represented common practice and therefore remains relevant for the purpose of our study. In conclusion, the surveillance system based on standard hospital data appeared to be sufficiently accurate for routine SSI detection, and its use allowed reducing by 90% the number of medical files to be revised by the ICPs. However, the results of our study have to be interpreted with caution, as some common practices such as wound culturing and diagnosis coding might differ between surgeons and hospitals. A larger multicenter study should therefore be undertaken
to assess the practicability of this system in other French hospitals and to study performance variations between surgical specialties and procedures.
acknowledgments The SSI Study Group consisted of Olivier Baron (cardiac surgery), PaulAntoine Lehur (colorectal surgery), Eric Bord (neurosurgery), Franc¸ois Gouin and Sophie Touchais (orthopedic surgery), Olivier Bouchot and Loı¨c Lenormand (urologic surgery), and Henri-Jean Philippe (obstetrical surgery). Potential conflicts of interest. All authors report no conflicts of interest relevant to this article. All authors submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest, and the conflicts that the editors consider relevant to this article are disclosed here. Address correspondence to Didier Lepelletier, MD, PhD, Department of Bacteriology and Infection Control, Nantes University Hospital, Nantes, France ([email protected]). Presented in part: 24th European Congress of Clinical Microbiology and Infectious Diseases; Barcelona, Spain; May 10–13, 2014 (E-poster).
references 1. Platt R, Yokoe DS, Sands KE. Automated methods for surveillance of surgical site infections. Emerg Infect Dis 2001;7(2):212– 216. doi:10.3201/eid0702.700212. 2. Leal J, Laupland KB. Validity of electronic surveillance systems: a systematic review. J Hosp Infect 2008;69(3):220–229. doi: 10.1016/j.jhin.2008.04.030. 3. Freeman R, Moore LSP, Garcı´a A´lvarez L, Charlett A, Holmes A. Advances in electronic surveillance for healthcare-associated infections in the 21st century: a systematic review. J Hosp Infect 2013;84(2):106–119. doi:10.1016/j.jhin.2012.11.031. 4. Astagneau P, L’He´riteau F, Daniel F, et al. Reducing surgical site infection incidence through a network: results from the French ISO-RAISIN surveillance system. J Hosp Infect 2009;72(2):127– 134. doi:10.1016/j.jhin.2009.03.005.
computer-enhanced ssi surveillance
5. Agence Technique de l’Information sur l’Hospitalisation. CCAM Classification Commune des Actes Me´dicaux, V5. 2006. http:// www.atih.sante.fr. 6. Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification of CDC definitions of surgical wound infections. Infect Control Hosp Epidemiol 1992;13(10):606–608. 7. Woeltje KF, Lautenbach E. Informatics and epidemiology in infection control. Infect Dis Clin North Am 2011;25(1):261–270. doi:10.1016/j.idc.2010.11.013. 8. Wilson APR, Gibbons C, Reeves BC, et al. Surgical wound infection as a performance indicator: agreement of common definitions of wound infection in 4773 patients. BMJ 2004; 329(7468):720. doi:10.1136/bmj.38232.646227.DE. 9. Lepelletier D, Ravaud P, Baron G, Lucet J-C. Agreement among health care professionals in diagnosing case vignette-based surgical site infections. PLoS ONE 2012;7(4):e35131. doi:10.1371 /journal.pone.0035131. 10. Birgand G, Lepelletier D, Baron G, et al. Agreement among healthcare professionals in ten European countries in diagnosing case-vignettes of surgical-site infections. PLoS ONE 2013;8(7): e68618. doi:10.1371/journal.pone.0068618. 11. Landers T, Apte M, Hyman S, Furuya Y, Glied S, Larson E. A comparison of methods to detect urinary tract infections using electronic data. Jt Comm J Qual Patient Saf 2010;36(9):411–417. 12. Apte M, Landers T, Furuya Y, Hyman S, Larson E. Comparison of two computer algorithms to identify surgical site infections. Surg Infect (Larchmt) 2011;12(6):459–464. doi:10.1089 /sur.2010.109. 13. Chalfine A, Cauet D, Lin WC, et al. Highly sensitive and efficient computer-assisted system for routine surveillance for surgical site infection. Infect Control Hosp Epidemiol 2006;27(8):794–801. doi:10.1086/506393. 14. Moro ML, Morsillo F. Can hospital discharge diagnoses be used
15.
16.
17.
18.
19.
20.
21.
22.
1335
for surveillance of surgical-site infections? J Hosp Infect 2004; 56(3):239–241. doi:10.1016/j.jhin.2003.12.022. Song X, Cosgrove SE, Pass MA, Perl TM. Using hospital claim data to monitor surgical site infections for inpatient procedures. Am J Infect Control 2008;36(3):S32–S36. doi:10.1016 /j.ajic.2007.10.008. Bolon MK, Hooper D, Stevenson KB, et al. Improved surveillance for surgical site infections after orthopedic implantation procedures: extending applications for automated data. Clin Infect Dis 2009;48(9):1223–1229. doi:10.1086/597584. West J, Khan Y, Murray DM, Stevenson KB. Assessing specific secondary ICD-9-CM codes as potential predictors of surgical site infections. Am J Infect Control 2010;38(9):701–705. doi: 10.1016/j.ajic.2010.03.015. Gerbier S, Bouzbid S, Pradat E, et al. Use of the French medicoadministrative database (PMSI) to detect nosocomial infections in the University hospital of Lyon. Rev Epidemiol Sante Publique 2011;59(1):3–14. doi:10.1016/j.respe.2010.08.003. Grammatico-Guillon L, Baron S, Gaborit C, Rusch E, Astagneau P. Quality assessment of hospital discharge database for routine surveillance of hip and knee arthroplasty–related infections. Infect Control Hosp Epidemiol 2014;35:646–651. doi:10.1086 /676423. Bouzbid S, Gicquel Q, Gerbier S, et al. Automated detection of nosocomial infections: evaluation of different strategies in an intensive care unit 2000–2006. J Hosp Infect 2011;79(1):38–43. doi:10.1016/j.jhin.2011.05.006. Petherick ES, Dalton JE, Moore PJ, Cullum N. Methods for identifying surgical wound infection after discharge from hospital: a systematic review. BMC Infect Dis 2006;6:170. doi: 10.1186/1471-2334-6-170. Miner AL, Sands KE, Yokoe DS, et al. Enhanced identification of postoperative infections among outpatients. Emerg Infect Dis 2004;10(11):1931–1937. doi:10.3201/eid1011.040784.
Control & Hospital Epidemiology:
Email alerts: Click here Subscriptions: Click here Commercial reprints: Click here Terms of use : Click here
Matching Bacteriological and Medico-Administrative Databases Is Efcient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital Brice Leclère, Camille Lasserre, Céline Bourigault, Marie-Emmanuelle Juvin, Marie-Pierre Chaillet, Nicolas Mauduit, Jocelyne Caillon and Matthieu Hanf Infection Control & Hospital Epidemiology / Volume 35 / Issue 11 / November 2014, pp 1330 - 1335 DOI: 10.1086/678422, Published online: 16 January 2015
Link to this article: http://journals.cambridge.org/abstract_S0195941700094558 How to cite this article: Brice Leclère, Camille Lasserre, Céline Bourigault, Marie-Emmanuelle Juvin, Marie-Pierre Chaillet, Nicolas Mauduit, Jocelyne Caillon and Matthieu Hanf (2014). Matching Bacteriological and Medico-Administrative Databases Is Efcient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital. Infection Control & Hospital Epidemiology, 35, pp 1330-1335 doi:10.1086/678422 Request Permissions : Click here
Downloaded from http://journals.cambridge.org/ICE, IP address: 132.174.255.116 on 05 Nov 2015
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
original article
Matching Bacteriological and Medico-Administrative Databases Is Efficient for a Computer-Enhanced Surveillance of Surgical Site Infections: Retrospective Analysis of 4,400 Surgical Procedures in a French University Hospital Brice Lecle`re, MD, MSc;1 Camille Lasserre, PharmD;1,2 Ce´line Bourigault, PharmD, MSc;1 Marie-Emmanuelle Juvin, PharmD;1 Marie-Pierre Chaillet, MD;3 Nicolas Mauduit, MD;3 Jocelyne Caillon, PharmD, PhD;1,2 Matthieu Hanf, PhD;4 Didier Lepelletier, MD, PhD1,2 for the SSI Study Groupa
objective. Our goal was to estimate the performance statistics of an electronic surveillance system for surgical site infections (SSIs), generally applicable in French hospitals. methods. Three detection algorithms using 2 different data sources were tested retrospectively on 9 types of surgical procedures performed between January 2010 and December 2011 in the University Hospital of Nantes. The first algorithm was based on administrative codes, the second was based on bacteriological data, and the third used both data sources. For each algorithm, sensitivity, specificity, and positive and negative predictive values (PPV and NPV) were calculated. The reference method was the hospital’s routine surveillance: a comprehensive review of the computerized medical charts of the patients who underwent one of the targeted procedures during the study period. setting.
A 3,000-bed teaching hospital in western France.
population.
We analyzed 4,400 targeted surgical procedures.
results. Sensitivity results varied significantly between the three algorithms, from 25% (95% confidence interval, 17–33) when using only administrative codes to 87% (80%–93%) with the bacteriological data and 90% (85%–96%) with the combined algorithm. Fewer variations were observed for specificity (91%–98%), PPV (21%–25%), and NPV (98% to nearly 100%). Overall, performance statistics were higher for deep SSIs than for superficial infections. conclusions. A reliable computer-enhanced SSI surveillance can easily be implemented in French hospitals using common data sources. This should allow infection control professionals to spend more time on prevention and education duties. However, a multicenter study should be conducted to assess the generalizability of this method. Infect Control Hosp Epidemiol 2014;35(11):1330-1335
Standard data from hospital information systems are easily available, and their management can be rapidly completed by the significant progress in computer technology in the past decades. This allows many aspects of routine surveillance of healthcare-associated infections (HCAIs) to be enhanced or automatized, including cluster detection, quality assessment, and risk factor identification.1 Infection control professionals may therefore spend their time more efficiently, focusing on prevention and education. However, even if these electronic surveillance systems (ESSs) can outperform routine manual surveillance, their use is not yet widespread.2,3 Most of the
published electronic HCAI identification methods differed in the targeted infections, databases, or identification algorithms. Their development according to each hospital information system is thus not transferable to other healthcare centers. Without a standardized and sustainable method, several useful functions of HCAI surveillance such as comparisons between institutions and evaluations of control measures cannot benefit from the genuine support offered by ESSs. In France, as in several other Western countries, monitoring surgical site infection (SSI) rates for selected proce-
Affiliations: 1. Department of Bacteriology and Infection Control, Nantes University Hospital, Nantes, France; 2. EA3826 The´rapeutiques Cliniques et Expe´rimentales des Infections, Medical School, University of Nantes, Nantes, France; 3. Department of Medical Information, Nantes University Hospital, Nantes, France; 4. Clinical Investigation Center, Nantes University Hospital, Nantes, France; a. Members of the SSI Study Group are listed at the end of the text. Received May 7, 2014; accepted July 15, 2014; electronically published September 30, 2014. 䉷 2014 by The Society for Healthcare Epidemiology of America. All rights reserved. 0899-823X/2014/3511-0002$15.00. DOI: 10.1086/678422
computer-enhanced ssi surveillance
dures is considered a key part of HCAI control programs and a useful benchmarking parameter.4 Moreover, SSIs are associated with substantial morbidity and mortality and have an important impact on patients and public health. Thus, they represent a pertinent target for a computer-assisted surveillance method. The aim of this study was to evaluate the performance of a generally applicable method for SSI detection using different databases available routinely in the medical information system of the Nantes University hospital during a 2-year period, 2010–2011.
methods Routine Surveillance Our study was conducted at the University Hospital of Nantes, a 3,000-bed hospital located in western France. Nine surgical procedures have been monitored to detect SSIs during a 2-year period from January 2010 to December 2011: coronary artery bypass grafting (CABG), valve replacement, elective (nonurgent) colectomy with immediate restoration of intestinal continuity, primary total hip arthroplasty (PTHA), primary total knee arthroplasty (PTKA), kidney transplantation, urinary sphincter implantation, neurostimulator implantation, and cesarean section (C-section). Surgical patients were identified by the French Common Classification of Medical Procedure (CCMP) codes that are recorded in the hospital’s medical activity database.5 A list of patients was monthly transmitted to the infection control team (ICT) by the medical information department. Then, infection control nurses or practitioners systematically reviewed the computerized medical records to detect SSIs (operative report, medical discharge letter, bacteriological results, follow-up appointment notes, etc). Each suspicion of SSI needed to be confirmed by the infection control practitioners (ICPs) or by the surgeon who operated on the patient. This surveillance was performed within 30 days, or 1 year for prosthetic surgery, using definitions from the National Healthcare Safety Network.6 We considered this routine surveillance the reference method (criterion standard) for performance assessment of 3 surveillance algorithms using 2 electronic data sources: bacteriological analysis results and medical activity. Data Sources In France, hospital medical activity is monitored using the Medical Information Program System (PMSI), a mandatory computerized medical database used for budget allocation. For each hospitalization, surgical procedures and relevant diagnoses are recorded using respectively the CCMP and the 10th revision of the International Classification of Diseases (IDC-10). At the University Hospital of Nantes, these medical codes were assigned by the surgeons, and quality control on the database was performed by medical coding specialists. The laboratory activity was also computerized, and the type,
1331
date, and culture results of every sample were recorded in the laboratory’s database. After selecting the targeted procedures with the appropriate CCMP codes and time frame (2010–2011), the two data sets were merged using a specific patient number as the key. This patient number is systematically and definitively assigned to each patient at the first admission in the hospital. Electronic Surveillance Algorithms Three algorithms were tested to retrospectively identify potential SSIs. Algorithm 1 was built on variables available in the PMSI database. According to the surgical procedures, relevant ICD-10 codes were selected to detect potential SSI cases (Table 1). Algorithm 2 was performed using the bacteriology laboratory database. Depending on the “sample type” variable (blood, stool, urine, biopsy, deep drainage, etc), samples were classified as “deep” or “superficial” for the sake of the analysis. The markers of potential SSIs were positive cultures from superficial samples (ie, cultures with abnormal bacterial growth) and requests for deep sample bacteriological examination, regardless of the results. For urinary sphincter implantation and kidney transplantation, simultaneously positive blood and urine cultures were seen as a sign of possible pyelonephritis and used as a marker as well. To be considered valid, the samples had to be taken within 40 days of the procedure, or within 400 days for prosthetic surgery procedures; these small margins were added to standard followup times (respectively 30 and 365 days) to take into consideration the delay between clinical signs and subsequent hospitalizations or revision procedures. Algorithm 3 was performed by combining algorithms 1 and 2. They were implemented by the ICT using STATA 10.0 software. Performance Analysis Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated with the corresponding 95% confidence interval for each of the three algorithms. Specific statistics were also estimated according to the SSI’s depth.
results Between January 2010 and December 2011, 4,400 targeted procedures were reviewed by the ICT using medical records and the CCMP codes. The number of surgical procedures varied according to the type of surgery, from 149 for neurostimulator implantation to 2,259 for CABG and valvular replacement in cardiac surgery (Table 2). The routine surveillance identified 112 SSIs, with an overall incidence rate of 2.5%. SSIs were defined as superficial SSIs (26%) or deep SSIs (74%; Table 2). The deep SSI incidence rates were 0% for C-section, 1.2% for cardiac surgeries, 1.4% for PTKA, 1.7% for urinary sphincter implantation, 2.0% for neurostimulator implantation, 2.2% for PTHA, 4.6% for elective colectomy, and 8.9% for kidney transplantation.
1332
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
table 1. International Classification of Diseases, Tenth Revision (ICD-10), Codes Used for Surgical Site Infection Detection according to the Targeted Procedures Surgery procedure and ICD-10 code Cesarean section T81.4 O86.0 Neurostimulator implantation T81.4 T85.7 Urinary sphincters implantation and kidney transplantation T81.4 T83.5 Coronary artery bypass grafting and valve replacement T81.4 T81.38 T82.7 T82.6 Colectomy T81.4 K65.0 K65.8 K65.9 Primary total knee or hip arthroplasty T81.4 T81.38 T84.5 T84.7
Description Infection following a procedure Infection of obstetric surgical wound Infection following a procedure Infection and inflammatory reaction due to other internal prosthetic devices, implants, and grafts Infection following a procedure Infection and inflammatory reaction due to prosthetic device, implant, and graft in urinary system Infection following a procedure Disruption of operation wound Infection and inflammatory reaction due to other cardiac and vascular devices, implants, and grafts Infection and inflammatory reaction due to cardiac valve prosthesis Infection following a procedure Acute peritonitis Other peritonitis Peritonitis, unspecified Infection following a procedure Disruption of operation wound Infection and inflammatory reaction due to internal joint prosthesis Infection and inflammatory reaction due to other internal orthopedic prosthetic devices, implants, and grafts
Sensitivity varied significantly between algorithms 1, 2, and 3 from 25% to 87% and 90%, respectively (P ! .001; Table 3). Variations for specificity (91%–98%), PPV (21%–25%), and NPV (98% to nearly 100%) were not significant. Overall, predictive values were higher for deep SSIs than for superficial SSIs. When bacteriological data were used (algorithms 2 and 3), higher sensitivities and specificities were also observed for deep SSIs, compared to superficial SSIs.
discussion An accurate and wholly automated surveillance system for HCAIs is an enticing prospect for ICPs but is not likely to be available soon.7 Indeed, the definition of most HCAIs, and especially SSIs, relies on subjective clinical observations that still require human validation.8-10 Depending on the objectives of the surveillance—cluster detection, incidence trends analysis, risk factor identification, and so on—different characteristics and performances might be relevant.1 The objective of this study was to develop an ESS that detects possible cases of SSI and reliably excludes noncases, in order to minimize the surveillance-related workload of infection control professionals. In this perspective, the number of false-negative cases should be kept as low as possible, and high sensitivity and NPV are thus the most relevant characteristics.11 The use
of bacteriological data with a nonrestrictive algorithm allowed such performance features: the overall sensitivity reached 87%, and the NPV was close to 100%. Among the 83 deep SSIs detected by the routine surveillance (criterion standard), 80 were also detected by using the bacteriological database (algorithm 2) with a very good sensitivity. The three falsenegative cases were patients with deep SSI (abscess) after colectomy. They were treated by intravenous antibiotherapy without bacteriological sample and no surgical reoperation. Overall, PPVs were inferior to 25%, mostly because of the large number of cases falsely detected using any of the algorithms. False-positive cases were due to diagnosis coding mistakes, infections occurring outside the scope of the surveillance (not related to surgery or not related to 1 of the targeted procedures), or unconfirmed SSI suspicions. It seemed clear that none of the tested algorithms could reliably be used for automatic detection of SSIs and that, consequently, ICPs needed to keep reviewing charts to validate SSI suspicions. The workload related to this task could, however, largely be reduced: using algorithm 2 allowed the number of medical charts to review to decrease by 90% (452 vs 4,400 files). So at the cost of a 4% loss of true SSIs due to false-negative cases (sensibility of 96%), this algorithm could allow ICPs to
computer-enhanced ssi surveillance
1333
table 2. Number of Surgical Site Infections (SSIs) Detected by Routine Surveillance and Number of Possible Infections Using 2 Different Databases at the Nantes University Hospital (2010–2011) Routine surveillance, gold standard Surgical procedure type CABG and valve replacement Colectomya PTHA PTKA Kidney transplantation Urinary sphincter implantation Neurostimulator implantation C-section Total
Algorithms
Surgical procedure no.
Superficial SSIs
Deep SSIs
1
2
3
2,259 196 506 278 293 179 149 540 4,400
0 3 0 0 4 2 6 14 29
27 9 11 4 26 3 3 0 83
47 33 8 4 13 1 0 4 113
270 28 48 17 57 9 11 11 451
294 29 49 17 63 10 11 11 484
note. Routine surveillance was performed using all data available from medical and surgical records and the French Common Classification of Medical Procedure (CCMP) with a patient follow-up of 1 year when a prosthetic implantation was done. Algorithm 1 was performed by matching the data collected with the 10th revision of the International Classification of Diseases codes and the CCMP codes; algorithm 2 was performed by matching the data collected by the laboratory database and the CCMP codes; algorithm 3 was performed by combining algorithms 1 and 2. CABG, coronary artery bypass grafting; C-section, cesarean section; PTHA, primary total hip arthroplasty; PTKA, primary total knee arthroplasty. a Colectomy with immediate restoration of digestive continuity, excluding emergency situations.
spend their time more efficiently, focusing on prevention and education or expansion of the range of targeted procedures. We believe that this may be a valid trade-off. Previous studies assessing the relevance of microbiological data to identify SSIs showed lower sensitivity measures: 59% for CABGs12 and 84% for gastrointestinal surgery.13 The detection rules used in these studies were more specific, as only patients with positive bacteriological exams were selected, hence excluding possible culture-negative SSIs.14 Moreover, algorithm 2 is probably dependent on wound culturing routine practices, particularly for superficial SSIs, which may also explain these differences. If the bacteriological database is available and the strategy of wound culturing well defined, the surveillance method using algorithm 2 can be easily implemented in all the French hospitals with a monthly case review (surveillance currently conducted at the Nantes University Hospital, with additional surgical procedures such as plastic surgery, mammary surgery, thoracic surgery, and vascular prosthesis surgery). The method relying solely on diagnosis codes through PMSI data had less interesting performance features. Our results showed that sensitivity, especially, barely reached 25%, which means that 3 out of 4 SSIs were missed. Many studies have showed that discharge diagnosis codes alone are not a reliable data source for HCAI surveillance.14-18 However, as a complementary data source, it can help to identify more potential SSIs. In our study, the combined algorithm’s sensitivity and NPV were higher than those observed using only the bacteriological database. These differences were not statistically significant, but this can be explained by a lack of statistical power, the number of SSIs following the monitored procedures being rather low. These results highlight the need
for better coding of SSIs diagnosis in the French PMSI database. Indeed, Gerbier et al18 showed that using unspecific ICD-10 codes for infections and postoperative complications can increase the sensitivity of a PMSI-based SSI detection method for general surgery, from 26% to 79%. In orthopedic surgery, a recent study has showed that a specific algorithm including all possible codes defined with surgeons ensures optimal detection of SSIs with good performance.19 In France, diagnosis coding methods may differ between healthcare facilities, and it might be relevant and necessary to adapt the PMSI algorithm to each center, according to its coding habits. Studies on HCAI detection systems tend to emphasize the importance of using multiple data sources.3 Some of them, such as computerized antibiotic prescriptions and text mining in medical charts are promising20 and can partly offset the dearth of systematically computerized clinical data. In our study, for example, most of SSI cases that were not captured by the bacteriological data were superficial SSIs, diagnosed in outpatients and for which no microbiological diagnosis was required. In these situations, text mining in clinical notes and records might be relevant. However, these data sources are not yet widespread and thus cannot be included in a surveillance system designed to be applicable in every French hospital. SSIs diagnosed after discharge have always been a possible source of selection bias for SSI surveillance, especially since some of these cases are diagnosed and treated in a different hospital than the one in which the original surgical procedure was performed. This, however, is true for both manual and computerized surveillances, and up to the present, there is no valid solution to this problem,1,21 even though regional or
1334
infection control and hospital epidemiology
november 2014, vol. 35, no. 11
table 3. Performance Analysis for Electronic Surveillance of Surgical Site Infections (SSIs) according to Data Sources and Infection Depth, Nantes University Hospital (2010–2011) Algorithm and SSI depth
Sensitivity
Specificity
Positive predictive value
Negative predictive value
Superficial Deep All
24.1 (8.6–39.7) 25.3 (15.9–34.7) 25 (17.0–33.0)
97.6 (97.1–98) 97.9 (97.4–98.3) 98 (97.6–98.4)
6.2 (1.8–10.6) 18.6 (11.4–25.8) 24.8 (16.8–32.7)
99.5 (99.3–99.7) 98.6 (98.2–98.9) 98 (97.6–98.5)
Superficial Deep All
58.6 (40.7–76.5) 96.4 (92.4–99.2) 86.6 (80.3–92.9)
90.1 (89.2–91.0) 91.4 (90.6–92.2) 91.7 (90.9–92.6)
3.8 (2.0–5.5) 17.7 (14.2–21.3) 21.5 (17.7–25.3)
99.7 (99.5–99.9) 99.9 (99.8–100) 99.6 (99.4–99.8)
Superficial Deep All
65.5 (48.2–82.8) 98.8 (96.4–100) 90.2 (84.7–95.7)
89.4 (88.4–90.3) 90.7 (89.8–91.6) 91.1 (90.2–91.9)
3.9 (2.2–5.7) 16.9 (13.6–20.3) 20.9 (17.2–24.5)
99.7 (99.6–99.9) 100 (99.9–100) 99.7 (99.6–99.9)
1
2
3
note. Surgical patients were first selected by using the French Common Classification of Medical Procedure (CCMP). Algorithm 1 was performed by matching the data collected with the 10th revision of the International Classification of Diseases codes and the CCMP codes; algorithm 2 was performed by matching the data collected by the laboratory database and the CCMP codes; algorithm 3 was performed by combining algorithms 1 and 2. 95% confidence intervals are presented in parentheses.
national health insurance data might be interesting sources of information.22 No superficial infections were reported for the targeted procedures of cardiac and orthopedic surgery. This was probably due to the fact that such infections are generally benign and sometimes difficult to diagnose with certainty, and, thus, they are probably rarely mentioned in medical records or coded in the PMSI database, and superficial samples might not usually seem relevant. Therefore, both routine and enhanced surveillances probably lack sensitivity for this kind of infection. Our study sample was determined using the surgical procedure codes registered in the hospital information system. A selection bias might therefore have occurred in case of incompleteness of these data, which could have skewed our estimations. However, unlike diagnosis coding, surgical procedure coding is straightforward, easily recordable, and verifiable by medical coders, and, most importantly, it is necessary for the right pricing of hospital stays in surgery. Therefore, a significant lack of exhaustiveness of surgical procedure coding seems unlikely. Regarding these limits, the reference method (routine surveillance) was probably not perfect (variability in diagnosing SSI between surgeons, infection control nurses or practitioners, no validation by an independent reviewer), but it represented common practice and therefore remains relevant for the purpose of our study. In conclusion, the surveillance system based on standard hospital data appeared to be sufficiently accurate for routine SSI detection, and its use allowed reducing by 90% the number of medical files to be revised by the ICPs. However, the results of our study have to be interpreted with caution, as some common practices such as wound culturing and diagnosis coding might differ between surgeons and hospitals. A larger multicenter study should therefore be undertaken
to assess the practicability of this system in other French hospitals and to study performance variations between surgical specialties and procedures.
acknowledgments The SSI Study Group consisted of Olivier Baron (cardiac surgery), PaulAntoine Lehur (colorectal surgery), Eric Bord (neurosurgery), Franc¸ois Gouin and Sophie Touchais (orthopedic surgery), Olivier Bouchot and Loı¨c Lenormand (urologic surgery), and Henri-Jean Philippe (obstetrical surgery). Potential conflicts of interest. All authors report no conflicts of interest relevant to this article. All authors submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest, and the conflicts that the editors consider relevant to this article are disclosed here. Address correspondence to Didier Lepelletier, MD, PhD, Department of Bacteriology and Infection Control, Nantes University Hospital, Nantes, France ([email protected]). Presented in part: 24th European Congress of Clinical Microbiology and Infectious Diseases; Barcelona, Spain; May 10–13, 2014 (E-poster).
references 1. Platt R, Yokoe DS, Sands KE. Automated methods for surveillance of surgical site infections. Emerg Infect Dis 2001;7(2):212– 216. doi:10.3201/eid0702.700212. 2. Leal J, Laupland KB. Validity of electronic surveillance systems: a systematic review. J Hosp Infect 2008;69(3):220–229. doi: 10.1016/j.jhin.2008.04.030. 3. Freeman R, Moore LSP, Garcı´a A´lvarez L, Charlett A, Holmes A. Advances in electronic surveillance for healthcare-associated infections in the 21st century: a systematic review. J Hosp Infect 2013;84(2):106–119. doi:10.1016/j.jhin.2012.11.031. 4. Astagneau P, L’He´riteau F, Daniel F, et al. Reducing surgical site infection incidence through a network: results from the French ISO-RAISIN surveillance system. J Hosp Infect 2009;72(2):127– 134. doi:10.1016/j.jhin.2009.03.005.
computer-enhanced ssi surveillance
5. Agence Technique de l’Information sur l’Hospitalisation. CCAM Classification Commune des Actes Me´dicaux, V5. 2006. http:// www.atih.sante.fr. 6. Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification of CDC definitions of surgical wound infections. Infect Control Hosp Epidemiol 1992;13(10):606–608. 7. Woeltje KF, Lautenbach E. Informatics and epidemiology in infection control. Infect Dis Clin North Am 2011;25(1):261–270. doi:10.1016/j.idc.2010.11.013. 8. Wilson APR, Gibbons C, Reeves BC, et al. Surgical wound infection as a performance indicator: agreement of common definitions of wound infection in 4773 patients. BMJ 2004; 329(7468):720. doi:10.1136/bmj.38232.646227.DE. 9. Lepelletier D, Ravaud P, Baron G, Lucet J-C. Agreement among health care professionals in diagnosing case vignette-based surgical site infections. PLoS ONE 2012;7(4):e35131. doi:10.1371 /journal.pone.0035131. 10. Birgand G, Lepelletier D, Baron G, et al. Agreement among healthcare professionals in ten European countries in diagnosing case-vignettes of surgical-site infections. PLoS ONE 2013;8(7): e68618. doi:10.1371/journal.pone.0068618. 11. Landers T, Apte M, Hyman S, Furuya Y, Glied S, Larson E. A comparison of methods to detect urinary tract infections using electronic data. Jt Comm J Qual Patient Saf 2010;36(9):411–417. 12. Apte M, Landers T, Furuya Y, Hyman S, Larson E. Comparison of two computer algorithms to identify surgical site infections. Surg Infect (Larchmt) 2011;12(6):459–464. doi:10.1089 /sur.2010.109. 13. Chalfine A, Cauet D, Lin WC, et al. Highly sensitive and efficient computer-assisted system for routine surveillance for surgical site infection. Infect Control Hosp Epidemiol 2006;27(8):794–801. doi:10.1086/506393. 14. Moro ML, Morsillo F. Can hospital discharge diagnoses be used
15.
16.
17.
18.
19.
20.
21.
22.
1335
for surveillance of surgical-site infections? J Hosp Infect 2004; 56(3):239–241. doi:10.1016/j.jhin.2003.12.022. Song X, Cosgrove SE, Pass MA, Perl TM. Using hospital claim data to monitor surgical site infections for inpatient procedures. Am J Infect Control 2008;36(3):S32–S36. doi:10.1016 /j.ajic.2007.10.008. Bolon MK, Hooper D, Stevenson KB, et al. Improved surveillance for surgical site infections after orthopedic implantation procedures: extending applications for automated data. Clin Infect Dis 2009;48(9):1223–1229. doi:10.1086/597584. West J, Khan Y, Murray DM, Stevenson KB. Assessing specific secondary ICD-9-CM codes as potential predictors of surgical site infections. Am J Infect Control 2010;38(9):701–705. doi: 10.1016/j.ajic.2010.03.015. Gerbier S, Bouzbid S, Pradat E, et al. Use of the French medicoadministrative database (PMSI) to detect nosocomial infections in the University hospital of Lyon. Rev Epidemiol Sante Publique 2011;59(1):3–14. doi:10.1016/j.respe.2010.08.003. Grammatico-Guillon L, Baron S, Gaborit C, Rusch E, Astagneau P. Quality assessment of hospital discharge database for routine surveillance of hip and knee arthroplasty–related infections. Infect Control Hosp Epidemiol 2014;35:646–651. doi:10.1086 /676423. Bouzbid S, Gicquel Q, Gerbier S, et al. Automated detection of nosocomial infections: evaluation of different strategies in an intensive care unit 2000–2006. J Hosp Infect 2011;79(1):38–43. doi:10.1016/j.jhin.2011.05.006. Petherick ES, Dalton JE, Moore PJ, Cullum N. Methods for identifying surgical wound infection after discharge from hospital: a systematic review. BMC Infect Dis 2006;6:170. doi: 10.1186/1471-2334-6-170. Miner AL, Sands KE, Yokoe DS, et al. Enhanced identification of postoperative infections among outpatients. Emerg Infect Dis 2004;10(11):1931–1937. doi:10.3201/eid1011.040784.
![[Surveillance of surgical site infections. What is the best way?].](/assets/img/hold.png)
