Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2018 Mar 1.
Published in final edited form as: Infect Control Hosp Epidemiol. 2016 Dec 15;38(3):334–339. doi: 10.1017/ice.2016.271

Validation of ICD-9-CM Diagnosis Codes for Surgical Site Infection and Noninfectious Wound Complications after Mastectomy

Margaret A Olsen *,, Kelly E Ball *, Katelin B Nickel *, Anna E Wallace , Victoria J Fraser *
PMCID: PMC5490368  NIHMSID: NIHMS865132  PMID: 27974057

Abstract

Background

Few studies have validated ICD-9-CM diagnosis codes for surgical site infection (SSI), and none have validated coding for noninfectious wound complications after mastectomy.

Objectives

To determine the accuracy of ICD-9-CM diagnosis codes in administrative health insurer claims data to identify SSI and noninfectious wound complications, including hematoma, seroma, fat and tissue necrosis, and dehiscence, after mastectomy.

Methods

We reviewed medical records for 275 randomly selected women who were coded for mastectomy with or without immediate breast reconstruction and were coded with an ICD-9-CM diagnosis code for a wound complication within 180 days after surgery. We calculated the positive predictive value (PPV) to evaluate the accuracy of diagnosis codes to identify specific wound complications and the PPV to determine the accuracy of coding for the breast surgical procedure.

Results

The PPV for SSI was 57.5%, or 68.9% if cellulitis-alone was considered an SSI, while the PPV for coding of cellulitis was 82.2%. The PPVs of individual noninfectious wound complications ranged from 47.8% for fat necrosis to 94.9% for seroma and 96.6% for hematoma. The PPVs for mastectomy, implant, and autologous flap reconstruction were uniformly high (97.5%–99.2%).

Conclusions

Our results suggest that claims data can be used to compare rates of infectious and noninfectious wound complications after mastectomy across facilities, although the PPV varies by specific type of postoperative complication. The accuracy of coding was highest for cellulitis, hematoma, and seroma, and a composite group of noninfectious complications (fat necrosis, tissue necrosis, or dehiscence).

INTRODUCTION

Using health claims data for retrospective surveillance of postoperative complications is useful to track complications across institutions and the spectrum of care. The accuracy of identification depends on the validity of the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes chosen for surveillance. Results have been mixed regarding the accuracy of ICD-9-CM diagnosis codes for surgical site infection (SSI) after various procedures. Few studies have validated ICD-9-CM diagnosis codes for SSI after breast procedures,13 and to our knowledge, no group has validated coding of noninfectious wound complications (e.g., hematoma, fat necrosis) after mastectomy.

We previously reported on the incidence of SSI4 and noninfectious wound complications5 after mastectomy using private insurer claims data. We used this population to determine the positive predictive value (PPV) of ICD-9-CM diagnosis codes to identify infectious and noninfectious wound complications and ICD-9-CM procedure and CPT-4 codes to identify breast surgical procedures in claims data compared to medical record review.

METHODS

We utilized a retrospective cohort of women 18–64 years of age with an ICD-9-CM procedure or CPT-4 code for mastectomy from 1/1/2004–8/31/2009 in the HealthCore Integrated Research Database, a private insurer claims database, as described previously (see Supplemental Table 1).4 Newly coded SSIs, cellulitis, hematoma, seroma, dehiscence, fat necrosis, and tissue necrosis were identified by ICD-9-CM diagnosis codes on inpatient and outpatient facility and provider claims from 2–180 days after mastectomy (day 0 for hematoma), as described previously.4,5 The complication algorithm included diagnosis codes specific to breast (e.g., 611.3 for fat necrosis), and general postoperative complications (e.g., 998.59 for SSI, Supplemental Tables 2 and 3). We randomly selected a subset of women with an infectious or noninfectious complication for review. We prioritized medical records from hospitals for patients with an inpatient claim coded for SSI and/or noninfectious wound complication. If there were no hospitalizations coded with a complication, we prioritized records from outpatient facilities, and finally medical records from providers. The medical records were obtained by a third party vendor and redacted of identifying information before review.

We received 132 records from women coded for SSI/cellulitis and 188 from women coded for one or more noninfectious wound complication that contained clinical information spanning the time period of the complication of interest. Signs and symptoms of postoperative complications documented by clinicians were abstracted from the medical records by a reviewer blinded to the codes in the claims data (KEB). The Centers for Disease Control and Prevention/National Healthcare Safety Network (NHSN) definition was used to define SSIs in the medical record; the pre-2010 (including cellulitis as SSI) and 2010 definitions (excluding cellulitis-only) were considered separately.6,7

PPV was calculated as the number of cases confirmed by medical record review divided by the number identified by the claims algorithm. The 95% confidence intervals were calculated with a finite population correction factor. Data management and analyses were performed using SAS v9.3 (SAS Institute Inc., Cary, NC) and SPSS version 21.0 (IBM SPSS Statistics, Armonk, NY). This study was approved with a waiver of informed consent by the Washington University Human Research Protection Office and Quorum Review.

RESULTS

We obtained medical records for 275 women coded for mastectomy; 260 (94.5%) had information regarding whether mastectomy was performed. Two women had undergone breast conserving surgery rather than mastectomy, for a PPV for mastectomy of 99.2%. The PPV for laterality of the mastectomy was 94.6%, with 244 of 258 coded correctly as unilateral or bilateral.

Two hundred fifty-nine records contained sufficient information to determine whether breast reconstruction was performed and the type of reconstruction. Of 122 women coded in the claims for breast expander or implant reconstruction, 119 were confirmed by medical record review (PPV 97.5%). Forty-eight women were coded for autologous flap reconstruction, and 47 were confirmed by medical record review (PPV 97.9%).

The PPVs for individual complications are shown in Table 1. The PPV was 57.5% for SSI using the current NHSN definition of SSI (excluding cellulitis-only), but increased to 68.9% if the pre-2010 NHSN definition including cellulitis-only as SSI was used. The PPVs of individual noninfectious wound complications ranged from 47.8% for fat necrosis to 96.6% for hematoma.

Table 1.

Positive Predictive Value of Coding in Administrative Claims Data Compared to Medical Record Review to Identify Surgical Site Infection and Noninfectious Wound Complications* after Mastectomy

Coded complication Number of women coded for complication Number of women with confirmed complication by medical record review Positive predictive value, % (95% confidence interval)
SSI and/or cellulitis 132 91 68.9 (61.4, 76.4)
SSI (not including cellulitis) 120 69 57.5 (49.1, 65.9)
SSI (not including cellulitis or 998.51) 107 67 62.6 (53.9, 71.3)
Cellulitis 45 37 82.2 (71.6, 92.9)
Hematoma 58 56 96.6 (92.2, 100.9)
Seroma 78 74 94.9 (90.2, 99.6)
Dehiscence 41 26 63.4 (49.2, 77.6)
Fat necrosis 23 11 47.8 (28.4, 67.2)
Tissue necrosis 23 15 65.2 (46.5, 84.0)
Dehiscence, fat necrosis, or tissue necrosis 76 66 86.8 (79.6, 94.1)
*

Women could have multiple different complications. Of the total 188 women coded for noninfectious wound complications, 157 were coded for one noninfectious wound complication, 27 were coded for two different noninfectious wound complications, and 4 women were coded for three different noninfectious wound complications (223 individual noninfectious complications). Forty-five women were coded for both SSI and a noninfectious wound complication.

Number of women coded for complication and with medical record received with clinical information spanning the time period of the complication.

Pre-2010 NHSN definition included cellulitis as SSI. The PPV for this definition, excluding 998.51, was 75.0 (67.6, 82.4).

The complications documented in the medical record at the time of false-positive coding for a wound complication are shown in Table 2. Of 100 discrepancies in 90 women, 4 (4.0%) had the correctly coded wound complication at another anatomic site (e.g., port infection, cellulitis), and 81 (81.0%) had a different breast wound complication recorded in the medical record. The most common complication noted in the medical records for women with false-positive coding for SSI was cellulitis, followed by seroma and noninfectious wound complications. Eleven of the 13 false-positive SSIs with medical record documentation of seroma had been coded with the ICD-9-CM diagnosis code 998.51 (infected postoperative seroma). The most common error in coding of noninfectious wound complications involved incorrect use of a code for a different wound complication. Twenty-four of these discrepancies involved women coded for fat or tissue necrosis or dehiscence, with medical record documentation of another of these noninfectious complications (Table 2). In the case of fat necrosis, 11 of the 12 complications miscoded as fat necrosis had medical record documentation of tissue necrosis.

Table 2.

Summary of Complications That Were Not Confirmed and Complications Documented in the Medical Records

False positive complication based on ICD-9-CM algorithm (Total false positive) Number not confirmed (n = 100)* Complication documented in the medical record at the time of false-positive coding
SSI (n = 51) 14 Cellulitis
13 Seroma
5 Hematoma
12 Noninfectious wound complication (dehiscence/fat or tissue necrosis)
3 Other infection (port, suture abscess)
4 Non-specific complication without specific indication of breast infection (chronic inflammation/mastitis, pain/swelling/fever, neutropenia)
3 No wound complication (breast revision procedure, drain removal, drainage with no complication)
Cellulitis (n = 8) 3 SSI
2 Other infection (cellulitis of leg/arm)
2 Seroma
1 Non-specific complication without specific indication of breast infection (pain/swelling)
1 Allergic reaction
Hematoma (n = 2) 2 SSI
1 Seroma
1 Noninfectious wound complication (dehiscence/fat or tissue necrosis)
Seroma (n = 4) 1 SSI
3 Hematoma
Dehiscence (n = 15) 4 SSI
2 Cellulitis
1 Seroma
1 Hematoma
10 Noninfectious wound complication (fat or tissue necrosis)
2 No wound complication (breast reconstruction/revision, no complication)
Fat necrosis (n = 12) 4 SSI
3 Cellulitis
2 Seroma
1 Hematoma
11 Noninfectious wound complication (dehiscence/tissue necrosis)
Tissue necrosis (n = 8) 1 SSI
2 Hematoma
3 Noninfectious wound complication (dehiscence/fat necrosis)
1 Non-specific complication (poorly healing wound)
2 No wound complication (drain removal, no complication)
*

An individual record could have multiple complications recorded

DISCUSSION

The PPVs of our ICD-9-CM diagnosis code algorithms to identify infectious and noninfectious complications after mastectomy were variable, but generally moderate to good. The PPVs of coding for SSI ranged from 58% to 69% depending on whether cellulitis-alone was considered to meet the SSI definition. Cellulitis-alone was excluded from the NHSN definition of SSI in 2010,6,7 and this change has had a big impact on reported infection rates since cellulitis is the most common postoperative breast infection.8 To address this, we calculated the PPV for SSI using the definition of SSI at the time of the study (2004–2009), as well as the current NHSN definition excluding cellulitis-only. Fourteen of the 51 miscoded SSIs using the strict definition of SSI were recorded as cellulitis in the medical record and 3 of the 8 women miscoded with cellulitis were recorded in the medical record as an SSI without clinician documentation of cellulitis.

The majority of non-confirmed individual complications coded in the claims were documented in the records as other noninfectious breast wound complications. Some of the miscodings could have been due to misinterpretation via term searches. For example, a common error involving false coding for SSI was due to the use of the diagnosis code for infected seroma (998.51), in women with a documented seroma (without evidence for infection). Similarly, difficulty appeared to exist in discrimination of fat versus tissue necrosis.

These results suggest that care should be used with respect to inclusion of certain codes in algorithms to identify specific complications (i.e., avoid 998.51 to identify SSIs). Since the majority of errors in coding of noninfectious wound complications involved misinterpretation of the specific complication, it may be better to focus on a composite group consisting of fat necrosis, tissue necrosis, or dehiscence, which had a higher PPV. In contrast, the PPVs for seroma and hematoma were very high, and therefore can be reliably identified as individual complications.

In the literature, validation of claims data for SSI surveillance has yielded mixed results. The sensitivities of algorithms using a comprehensive list of codes to identify SSI in several studies were high at 72–99%,911 while the PPVs were lower, ranging from 15–51%.3,1016 In contrast, in studies that used a small set of SSI-specific diagnosis codes,1,2,1722 sensitivities varied greatly depending on the surgery and data source, from 20% for 998.59 from a single academic medical center after general and vascular surgery18 to 100% after joint arthroplasty (using 998.5X and 996.66) or vascular surgery (using 998.5X and 996.62) using Medicare claims.17 The PPVs in these studies using SSI-specific codes were generally higher, with 4 of 6 studies reporting PPVs greater than 50% and up to 88%.1,2,1821

Three studies have assessed the accuracy of coding for SSI codes after breast surgery. We previously reported high sensitivity (88%), specificity (99%), and PPV (88%) for SSI using ICD-9-CM diagnosis codes 611.0, 682.2, 682.3, 996.69, and 998.5X from primarily inpatient billing data after breast surgery at an academic medical center.2 In a multicenter two-phase study, Yokoe and colleagues reported the sensitivity of SSI diagnosis codes 998.5X from inpatient billing data as 50% in phase 1 and 70% in phase 2, with PPVs of 58 and 79%.1 Miner and colleagues used claims data from a large health care system to evaluate a complex algorithm for SSI using a variety of ICD-9-CM diagnosis codes and procedure codes for wound care and culture. The PPV of the SSI algorithm using medical claims from inpatient, outpatient, or emergency department encounters within 60 days after breast surgery was only 18%.3

The choice of codes used to identify SSI and corresponding sensitivity and PPV depends on the intended goal of surveillance. Our intent was to develop an algorithm to compare complication rates among facilities and over time; thus our focus was on maximizing the PPV to identify SSI and noninfectious wound complications. Interestingly, Calderwood and colleagues reported low PPVs of a comprehensive list of diagnosis and procedure codes to identify SSI in Medicare claims after hip arthroplasty and vascular surgery, but found similar PPVs between hospitals in the best- and worst-performing deciles in terms of risk-adjusted infection rates.12,16 This highlights the potential benefit of administrative data for comparing SSI rates between hospitals, even when the PPV is low.

The PPVs for the procedure codes to identify mastectomy and breast implant and autologous flap reconstruction were all greater than 97%, consistent with previous reports.2,23,24 As described previously,4 before randomly selecting procedures for medical record review, we performed extensive filtering to exclude procedures that were unlikely mastectomy (e.g., mastectomy coded only by an assistant, brachytherapy catheters at time of procedure). Thus the true PPVs of the procedures codes for mastectomy and reconstruction procedures are likely lower than what we report. Provided that a comprehensive algorithm is used to identify the appropriate patient population, the high PPVs of procedure codes suggest that accurate denominators can be calculated from claims or billing data to compare SSI rates across different surgical procedures or institutions.

There are some limitations to our study. First, we only reviewed the record from one source (facility or provider) per patient, so it is possible that documentation of a complication may have been available in an alternative record. We could only determine the PPV of the ICD-9-CM diagnosis code algorithms and not sensitivity, specificity, or negative predictive values since we did not obtain medical records from women who were not coded for complications. A larger study would be helpful to assess whether our findings for individual noninfectious complications can be replicated.

Using health claims data, we captured complications coded at the time of an inpatient or outpatient hospital visit at the same or different institution than the index mastectomy, as well as complications treated in outpatient clinics. We found moderate PPVs for ICD-9-CM diagnosis codes for fat necrosis, tissue necrosis, and dehiscence, and high PPVs for coding of cellulitis, hematoma, and seroma. The PPV was modest for SSI without cellulitis, but improved if cellulitis-only was considered an infection and if the code for infected seroma was excluded from the algorithm. Given the relatively high PPVs for noninfectious complications, including noninfectious wound complications along with SSI may provide a more robust measure of quality of care to compare complication rates between facilities. New surveillance algorithms to identify infectious and noninfectious wound complications will have to be developed using ICD-10 codes in future, but this study can be used as the basis to develop ICD-10 algorithms. Finally, the high PPVs to identify breast procedures are encouraging, since this indicates high accuracy of the denominators used to calculate wound complication rates from claims data. Although imperfect, claims data can be used to screen for possible complications, with subsequent confirmation by medical record review, to improve the efficiency of routine SSI surveillance.

Supplementary Material

Supplemental material

Acknowledgments

We thank Cherie Hill for database and computer management support. Funding for this work was provided by the National Institutes of Health (NIH) (5R01CA149614 to MAO). Additional support was provided by the Centers for Disease Control and Prevention (CDC) Epicenters Program (U54CK000162 to VJF).

MAO reports consultant work with Pfizer, Merck, and Sanofi Pasteur and grant funding through Pfizer, Sanofi Pasteur, and Cubist Pharmaceuticals for work outside the submitted manuscript.

Footnotes

KEB reports no conflict of interest.

KBN reports no conflict of interest.

AEW is an employee of HealthCore, a wholly-owned subsidiary of Anthem, Inc., a health insurance company. She has received Anthem stock options and participates in an Anthem employee stock purchase plan.

VJF reports her spouse is Senior Vice President and Chief Medical Officer at Express Scripts.

References

  • 1.Yokoe DS, Noskin GA, Cunningham SM, et al. Enhanced identification of postoperative infections among inpatients. Emerg Inf Dis. 2004;10:1924–1930. doi: 10.3201/eid1011.040572. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Olsen MA, Fraser VJ. Use of diagnosis codes and/or wound culture results for surveillance of surgical site infection after mastectomy and breast reconstruction. Infect Control Hosp Epidemiol. 2010;31:544–547. doi: 10.1086/652155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Miner AL, Sands KE, Yokoe DS, et al. Enhanced identification of postoperative infections among outpatients. Emerg Inf Dis. 2004;10:1931–1937. doi: 10.3201/eid1011.040784. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Olsen MA, Nickel KB, Fox IK, et al. Incidence of surgical site infection following mastectomy with and without immediate reconstruction using private insurer claims data. Infect Control Hosp Epidemiol. 2015;36:907–914. doi: 10.1017/ice.2015.108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Nickel KB, Fox IK, Margenthaler JA, Wallace AE, Fraser VJ, Olsen MA. Effect of noninfectious wound complications after mastectomy on subsequent surgical procedures and early implant loss. J Am Coll Surg. 2016;222:844–852. doi: 10.1016/j.jamcollsurg.2016.01.050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.National Healthcare Safety Network (NHSN) Procedure-Associated (PA) module: surgical site infection (SSI) event. Centers for Disease Control and Prevention. Available at: http://www.cdc.gov/nhsn/PDFs/pscManual/9pscSSIcurrent.pdf. Accessed November 14, 2013.
  • 7.Mangram AJ, Horan TC, Pearson ML, Silver LC, Jarvis WR. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect Control Hosp Epidemiol. 1999;20:250–278. doi: 10.1086/501620. [DOI] [PubMed] [Google Scholar]
  • 8.Degnim AC, Throckmorton AD, Boostrom SY, et al. Surgical site infection after breast surgery: impact of 2010 CDC reporting guidelines. Ann Surg Oncol. 2012;19:4099–4103. doi: 10.1245/s10434-012-2448-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Sands KE, Yokoe DS, Hooper DC, et al. Detection of postoperative surgical-site infections: comparison of health plan-based surveillance with hospital-based programs. Infect Control Hosp Epidemiol. 2003;24:741–743. doi: 10.1086/502123. [DOI] [PubMed] [Google Scholar]
  • 10.Rhee C, Huang SS, Berrios-Torres SI, et al. Surgical site infection surveillance following ambulatory surgery. Infect Control Hosp Epidemiol. 2015;36:225–228. doi: 10.1017/ice.2014.23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Calderwood MS, Kleinman K, Murphy MV, Platt R, Huang SS. Improving public reporting and data validation for complex surgical site infections after coronary artery bypass graft surgery and hip arthroplasty. Open Forum Infect Dis. 2014;1 doi: 10.1093/ofid/ofu106. ofu106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Calderwood MS, Kleinman K, Bratzler DW, et al. Use of Medicare claims to identify US hospitals with a high rate of surgical site infection after hip arthroplasty. Infect Control Hosp Epidemiol. 2013;34:31–39. doi: 10.1086/668785. [DOI] [PubMed] [Google Scholar]
  • 13.Stevenson KB, Khan Y, Dickman J, et al. Administrative coding data, compared with CDC/NHSN criteria, are poor indicators of health care-associated infections. Am J Infect Control. 2008;36:155–164. doi: 10.1016/j.ajic.2008.01.004. [DOI] [PubMed] [Google Scholar]
  • 14.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;38:701–705. doi: 10.1016/j.ajic.2010.03.015. [DOI] [PubMed] [Google Scholar]
  • 15.Julian KG, Brumbach AM, Chicora MK, et al. First year of mandatory reporting of healthcare-associated infections, Pennsylvania: an infection control-chart abstractor collaboration. Infect Control Hosp Epidemiol. 2006;27:926–930. doi: 10.1086/507281. [DOI] [PubMed] [Google Scholar]
  • 16.Calderwood MS, Kleinman K, Bratzler DW, et al. Medicare claims can be used to identify US hospitals with higher rates of surgical site infection following vascular surgery. Med Care. 2014;52:918–925. doi: 10.1097/MLR.0000000000000212. [DOI] [PubMed] [Google Scholar]
  • 17.Calderwood MS, Ma A, Khan YM, et al. Use of Medicare diagnosis and procedure codes to improve detection of surgical site infections following hip arthroplasty, knee arthroplasty, and vascular surgery. Infect Control Hosp Epidemiol. 2012;33:40–49. doi: 10.1086/663207. [DOI] [PubMed] [Google Scholar]
  • 18.Hollenbeak CS, Boltz MM, Nikkel LE, Schaefer E, Ortenzi G, Dillon PW. Electronic measures of surgical site infection: implications for estimating risks and costs. Infect Control Hosp Epidemiol. 2011;32:784–790. doi: 10.1086/660870. [DOI] [PubMed] [Google Scholar]
  • 19.Lawson EH, Louie R, Zingmond DS, et al. A comparison of clinical registry versus administrative claims data for reporting of 30-day surgical complications. Ann Surg. 2012;256:973–981. doi: 10.1097/SLA.0b013e31826b4c4f. [DOI] [PubMed] [Google Scholar]
  • 20.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;48:1223–1229. doi: 10.1086/597584. [DOI] [PubMed] [Google Scholar]
  • 21.Yokoe DS, Khan Y, Olsen MA, et al. Enhanced surgical site infection surveillance following hysterectomy, vascular, and colorectal surgery. Infect Control Hosp Epidemiol. 2012;33:768–773. doi: 10.1086/666626. [DOI] [PubMed] [Google Scholar]
  • 22.Knepper BC, Young H, Jenkins TC, Price CS. Time-saving impact of an algorithm to identify potential surgical site infections. Infect Control Hosp Epidemiol. 2013;34:1094–1098. doi: 10.1086/673154. [DOI] [PubMed] [Google Scholar]
  • 23.Fisher ES, Whaley FS, Krushat WM, et al. The accuracy of Medicare’s hospital claims data: progress has been made, but problems remain. Am J Public Health. 1992;82:243–248. doi: 10.2105/ajph.82.2.243. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Talia J, Agarwal S, Momoh AO, Wilkins EG, Kozlow JH. The validity of hospital discharge data for autologous breast reconstruction research. Plast Reconstr Surg. 2015;135:368–374. doi: 10.1097/PRS.0000000000000894. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental material

RESOURCES