California Cancer Registry 2013 Lung Recoding Audit Final Report The Production, Automation and Quality Control Unit (PAQC) of the California Cancer Registry (CCR) initiated a recoding audit of lung cancer cases in an effort to determine data quality among these cases. Audit Sampling and Performance Methodology The recoding audit consisted of admission level lung cancer cases diagnosed in 2012 that bypassed the visual editing process. There were 26 data items audited, 14 visually edited (VE) data items as well as 12 additional data items including behavior and treatment related data items. The results were calculated following the standard VE methodology, where the data items CS Site Specific Factors #1 and #2 were counted as one (1) discrepancy per case. Discrepancies in the data items Regional Lymph Nodes Positive and Examined were also calculated using the same methodology. Collaborative Stage coding system version 02.04 was used for this audit since the audit consisted of cases diagnosed in 2012. All regions were involved in this audit. Due to limited available resources, the sampling reflected the three California SEER registries. Therefore, the regions that make up the Cancer Registry of Greater California (CRGC) were audited as one region. Regions 2, 3, 4, 5, 6, and 10 are considered the CRGC. The Cancer Prevention Institute of California (CPIC), also known as Region 8, and the Cancer Surveillance Program (CSP), also known as Region 9, were the other two SEER registries involved in this audit. The CCR adopted the Proportional Stratified Sampling method which takes into account the population (number of cancer cases with this primary site), confidence level, and margin of error. By adopting this methodology, it is felt the audit results will yield a statistically reliable result. Adopting this methodology resulted in a larger number of cases audited for each region. This resulted in a total of 348 cases audited. The number of cases audited by region is as follows:    CRGC: 150 cases, CPIC: 100 cases, CSP: 98 cases. Each case was recoded based on the text documentation submitted on the abstract. The original code must match the recode by the auditor. A discrepancy resulted if these codes did not match. The audit was performed utilizing the peer review method. Each case was audited blindly and independently by two separate auditors. When both the primary and secondary auditors completed their assigned region, the primary auditor then compared the discrepancies between the auditors. The two auditors then reconciled their differences and discussed any issues discovered during the audit process. May 2014 2013 Lung Recoding Audit Final Report 1 California Cancer Registry 2013 Lung Recoding Audit Final Report Audit Findings There were 726 discrepancies noted on this audit. More than half of the discrepancies occurred in Collaborative Stage (CS) data items. There were 496 (68.3%) discrepancies identified in CS data items and 230 (31.7%) identified in Non-CS data Items. This result is demonstrated in Graph 1. A detailed distribution of all discrepancies by data item can be found in Table 1. Discrepancy Analysis - CS Related Data Items There were 496 (68.3%) discrepancies noted in CS data items. The distribution of the top five CS data item discrepancies is:      CS Extension CS Site Specific Factor 2 (Pleural/Elastic Layer Invasion) CS Mets at DX CS Site Specific Factor 1 (Separate Tumor Nodules – Ipsilateral Lung) CS Lymph Nodes 115 (23.2%) 74 (14.9%) 68 (13.7%) 57 (11.5%) 45 (9.1%) A detailed distribution of discrepancies in CS data items can be found in Table 2. CS Extension The data item CS Extension had the highest number of discrepancies identified on this audit. There were 115 (23.2%) discrepancies identified. A clear majority of the recodes in this data items involve the CS Extension codes 100, 300, and 999. There were 75 discrepancies (65.2%) that involved these codes. The breakdown of these discrepancies is: o 49 (42.6%) were recoded to extension code 100 (Tumor confined to one lung) o 18 (15.6%) were recoded to extension code 300 (Localized, NOS) o 8 (6.9%) were recoded to 999 (Unknown; extension not stated. Primary tumor cannot be assessed) Other issues identified are coding direct extension into the visceral pleura, correct use of TNM NOS codes (115, 120, 460, 465, 570 and 810) , coding atelectasis/obstructive pneumonitis when the entire lung is not involved, and properly coding the CS extension when the contralateral lung is not involved May 2014 2013 Lung Recoding Audit Final Report 2 California Cancer Registry 2013 Lung Recoding Audit Final Report CS Site Specific Factor #2 (Pleural/Elastic Layer Invasion (PL) by H and E or Elastic Stain) This data item had 74 (14.9%) discrepancies. Almost 90% of the discrepancies in this data item, 66 discrepancies (89.2%) occurred as a result of a misunderstanding of Note 2 in the coding instructions for this data item. Note 2 states “Code 998 if no histologic examination of pleura to assess pleural layer involvement.” There were only six (6) situations (8.1%) where the patient in fact had a histological examination performed and this data item could be coded. CS Mets at Diagnosis There were 68 (13.7%) discrepancies noted in the data field CS Mets at Diagnosis. This data item is very complex and easy to misunderstand the vast number of combination codes. Due to the enormous number of overly complex combination codes, it is nearly impossible to identify patterns and identify any one issue that is causing these discrepancies. The codes themselves play a role in causing these discrepancies. The only important discrepancies that can be noted are: o 15 (22.1%) were originally coded to 00 (No distant metastasis) when in fact the patient did have metastatic disease and resulted in a recode o 8 (11.8%) were originally coded as having metastatic disease when they did not have documentation that supported distant disease o 6 (8.8%) were originally coded as 99 (Unknown; distant metastasis not stated) when they did not have metastatic disease and recoded to 00 (No distant metastasis) Other issues included coding pleural effusions and cardiac effusions in the CS Mets at Diagnosis field, and capturing contralateral lung nodules. CS Site Specific Factor # 1 (Separate Tumor Nodules – Ipsilateral Lung) CS Site Specific Factor # 1 (Separate Tumor Nodules – Ipsilateral Lung) had 57 (11.5%) discrepancies on this audit. This data item captures multiple tumors within the ipsilateral lung. There were 25 (43.8%) discrepancies where this data item was recoded to 000 “No separate tumor nodules noted.” Of those, 16 (64%) were recoded from 999 “Unknown if separate tumor nodules.” This is due to the registrar not applying the rules. Note 3 in CS Site Specific Factor #1 which states “If separate tumor nodules are not mentioned in imaging and/or pathological reports, use code 000”. Nine (9) discrepancies (15.8%) were recodes to 000 “No separate tumor nodules noted” from another code indicating ipsilateral lung tumors (codes 010, 020, 030, or 040). Tumors in the contralateral lung (opposite lung from the primary tumor) are captured in the data item CS Mets at Diagnosis and are considered distant disease. May 2014 2013 Lung Recoding Audit Final Report 3 California Cancer Registry 2013 Lung Recoding Audit Final Report CS Lymph Nodes There were 45 (9.1%) discrepancies noted in CS Lymph Nodes. This field is to be used to code regional lymph node involvement only. Distant lymph nodes are captured in CS Mets at Diagnosis field. There seems to be considerable issues for abstractors to determine the correct lymph node code when coding this data item for lung cases. Among the various recodes in this data item were:  CS Lymph Node code 200 had the most recodes with 18 (40%) recodes. o 8 (44.4%) were recoded to code 600 o 4 (22.2%) were recoded to code 999 (Unknown lymph node involvement) o 3 were recoded to code 000 (no lymph node involvement) o 3 were recoded to code 100  CS Lymph Node code 100 had the second highest number of recodes with 10 (22.2%) recodes. o 6 (60%) were recoded to code 200 o 2 (20%) were recoded to 000 (no lymph node involvement) o 1 (10%) was recoded to code 500 o 1 (10%) was recoded to code 600 Since this data item is critical to ascertaining the correct AJCC summary stage, it is important to note those cases that involve the CS Lymph Node codes 000 (No lymph node involvement) and 999 (Unknown lymph node involvement).  There were 7 (15.5%) discrepancies noted where CS Lymph Node code 000 (No lymph node involvement) was recoded to another known code.  There were 7 (15.5%) discrepancies noted where CS Lymph Node code 999 (Unknown lymph node involvement) was recoded to another known code. Alternately, there were:  10 (22.2%) discrepancies that were recoded to CS Lymph Node code 000 (No lymph node involvement)  5 (11.1%) discrepancies that were recoded to CS Lymph Node code 999 (Unknown lymph node involvement) May 2014 2013 Lung Recoding Audit Final Report 4 California Cancer Registry 2013 Lung Recoding Audit Final Report There were other issue identified, such as: The correct application of Note 2 in CS Lymph Nodes which reads, “If at mediastinoscopy/x-ray, the description is ‘mass,’ ‘adenopathy,’ or ‘enlargement’ of any of the lymph nodes named as regional in codes 100 and 200, assume that at least regional lymph nodes are involved. If there is any mention of bilateral or contralateral mass, adenopathy or lymph node involvement, use code 500.” Other issues included applying the correct CS Lymph Node code that corresponds to the involved lymph node chain, such as: a hilar lymph node is coded to CS Lymph Node code 100 and the correct use of code 600 due to bilateral involvement Discrepancy Analysis - Non CS Data Items The remaining 230 (31.7%) of the discrepancies were among the non-CS data items. The distribution of discrepancies of the top five non-CS data items were:      Scope Regional Lymph Node Surgery Date of Diagnosis Regional Radiation Modality Primary Site Chemotherapy Summary 28 28 25 25 21 (12.2%) (12.2%) (10.9%) (10.9%) (9.1%) A detailed distribution of discrepancies in non-CS data items can be found in Table 3. Scope of Regional Lymph Node Summary There were 28 (12.2%) discrepancies noted in this data item. Twenty-three (23), or 82.1% of the discrepancies, were the result of not coding this field when the only procedure performed was an FNA or lymph node biopsy. In all of these cases, the original code was 0 “No regional lymph node surgery” and were all recoded to code 1 “Biopsy or aspiration of regional lymph nodes(s).” Regional Radiation Modality There were 25 (10.9%) discrepancies noted in this data item. Eight (8) discrepancies (32%) were the recoded to code 20 “External Beam, NOS.” The primary reason for these recodes was due to text documenting “photons” or “cGy” only with no other mention of the energy used. A common misconception is that “cGy” is the energy used; however, “cGy” is an abbreviation for the term “centigray”, which is defined as a unit of measurement of the radiation dose given to the patient and is equal to 1 rad (or radian/radiation). Therefore, the phrase “1800 cGy” means the patient received 1800 rads of therapy with no indication of the energy used to deliver the dose such as, photons, electrons, protons, etc. May 2014 2013 Lung Recoding Audit Final Report 5 California Cancer Registry 2013 Lung Recoding Audit Final Report Five (5) discrepancies (20%) were recoded to code 41 “Stereotactic radiosurgery” from various other codes; such as code 24 “Photons (6-10mv)” when there was clear documentation that the patient had stereotactic radiosurgery. There were three (3) recodes (12%) to code 42 “Linac radiosurgery” from radiation code 41 “Stereotactic radiosurgery.” It is understood why this type of discrepancy is made due to codes 41 and 42 residing near each other and having very similar definitions. A CAnswer Forum Thread describes CyberKnife, as a subtype of Linac radiosurgery and to code to 42. Primary Site There were 25 (10.9%) discrepancies for the data item primary site / subsite. These discrepancies were all identified in the subsite. Twelve (12) of these discrepancies (48%) were recoded to C34.1 (Lung, Upper Lobe), with six (50%) of those cases being recoded from C34.9 (Lung, NOS). Five (5) cases (20%) were recoded to C34.9 (Lung, NOS) from a specific subsite of the lung, such as C34.1 (Lung, Upper Lobe) or C34.2 (Lung, Lower Lobe). The anatomy of the thorax is can be very complex and difficult to determine the specific area where a tumor may arise. This is further complicated by physicians using various terms to describe to location of the original tumor. Another factor may be multiple tumors diagnosed simultaneously in both lungs. Paying close attention to the reports and documentation in the medical record is required to correctly code the primary site, as this data item is extremely important to research studies, such as outcomes and trends over time. Chemotherapy Summary There were 21 (9.1%) discrepancies noted in the data item chemotherapy. More than half of these discrepancies (52.4%) were due to not using the chemotherapy codes 82, 85, 87, and 88. These codes are defined as:  82: “Chemotherapy was not recommended / administered because it was contraindicated due to patient risk factors (i.e., comorbid conditions, advanced age).”  85: “Chemotherapy not administered because the patient died prior to planned or recommended therapy.”  87: “Chemotherapy was not administered. It was recommended by the patient's physician, but this treatment was refused by the patient, a patient's family member, or the patient's guardian. The refusal was noted in patient record.”  88: “Chemotherapy was recommended, but it is unknown if it was administered.” In all cases there was documentation that supported the use of these codes but instead, the abstractor used code 00 “No chemotherapy.” All remaining recodes in this data item were due to the lack of text documentation to support the code or not choosing the correct code that the text supports. May 2014 2013 Lung Recoding Audit Final Report 6 California Cancer Registry 2013 Lung Recoding Audit Final Report CS Discrepancy Impact on AJCC Stage All CS discrepancies were evaluated to determine the impact on the resulting AJCC stage. Using one of the production copies of the state data base (Eureka), each case where a CS discrepancy was noted, the CS changes were made and the TNM algorithm was rerun. The AJCC algorithm result of each CS change, whether there was a change in any of the T, N, M, or stage groups or not, was noted in a spreadsheet. The results were significant and intriguing. Of the 496 discrepancies in CS data items, the CS changes resulted in only 74 changes to T, N, or M codes. Furthermore, those 74 changes resulted in only 24 changes to the AJCC Stage grouping. Table 4 demonstrates these results. This result demonstrates the inefficiencies of the CS staging system. The Lung schema is particularly complex and any misunderstanding of the anatomy of the thorax, or terminology used by physicians will most likely result in the incorrect code being assigned. One example of the complexity of the CS system can be found in the data item CS Mets at Diagnosis. In the relating element in the AJCC manual, there are only four options for coding the M value of T-NM. The four options are: M0: No Distant metastasis M1: Distant metastasis M1a: Separate tumor nodules(s) in a contralateral lobe; tumor with pleural nodules or malignant pleural (or pericardial) effusion M1b: Distant metastasis (in extrathoracic organ) Comparatively, there are 27 options in the CS staging system, not including the four “obsolete” codes. Of the 27 CS Mets at Diagnosis code options, 10 map to M1a, 14 map to M1b and one maps to M1. Furthermore, 25of the options relate to a SEER Summary Stage of “Distant.” This complex coding structure has caused many hours of effort to understand and code with little to no impact on the AJCC stage grouping, which is the internationally recognized staging system for cancer. The CS coding system was designed to eliminate duplicate data collection, collect more relevant data, provide data reproducibility over time, and create compatibility between the two different staging systems. Instead, the CS system has successfully caused confusion among registrars as well as many hours of visually editing CS codes that have little to no impact to the resulting staging systems. May 2014 2013 Lung Recoding Audit Final Report 7 California Cancer Registry 2013 Lung Recoding Audit Final Report Resource Allocation, Benefit, and Core Team   Projected Hours to Complete Project: Estimated at 350 hours from project kick-off to Final Report submission. Actual Hours to Complete Project: Approximately 520 hours. This was the first recoding audit of lung cancer cases performed by the central registry in a number of years. Since the reduction of VE to 20% of the cases submitted to the CCR, these types of audits are increasingly more important and relevant. Lung cancer cases are particularly difficult to code and stage since the anatomy of the thorax, or chest cavity, is complex and contains many structures. Adding to this complexity is the fact that physicians refer to many of the same structures using different terms. The Collaborative Staging system does not help with its many codes and sometimes confusing instruction. Due to these and other factors, performing recoding audits of primary sites such as lung, give an insight as to the quality of the data on the Eureka data base and provides the central registry with topics for educational activities. The core team included the following: PAQC Unit  Kyle L. Ziegler, CTR,  Mary Brant, CTR,  Jenna Mazreku, CTR,  Marilyn Scocozza, CTR CPIC:  Chris Schwarz, CTR Conclusions It has been identified on this audit that lung cases have unique and complex coding issues. It was noted that abstractors are not properly following instruction given on how to code particular data items such as Note 2 in CS Lymph Nodes and Note 2 in CS Site Specific Factor #2 (Pleural/Elastic Layer Invasion (PL) by H and E or Elastic Stain). The complex anatomy combined with an even more complex and difficult Collaborative Stage coding structure makes discrepancies such as these common. Encouraging abstractors to approach coding CS in this site carefully and cautiously will eliminate a majority of these discrepancies. Further education on coding CS for lung cases is highly encouraged until the end of CS collection in order to reduce these errors and improve the quality of the data. May 2014 2013 Lung Recoding Audit Final Report 8 California Cancer Registry 2013 Lung Recoding Audit Final Report Recommendations It is recommended that specific actions be taken to improve the quality of the lung cancer cases on the CCR data base. Specifically:    Results to be published in the CCR bulletin “Innovations.” Create presentation or webinar highlighting at least the major findings, including but not limited to: o Coding Lung Cancer Cases with attention to the following data items:  CS Extension, CS Lymph Nodes, Scope of Regional Lymph Node Surgery, and Regional Radiation Modality. o Applying documented rules such as demonstrated in CS Site Specific Factor #1, CS Site Specific Factor #2 or Scope of Regional Lymph Node Summary. o Review of TNM staging for Lung Cancer Prepare a basic TNM presentation for Lung Cancer Educational Highlights to include the following: Scope of Regional Lymph Node Surgery o Remind registrars to code this field when a FNA or single lymph node biopsy was performed CS Extension o The correct use of CS extension codes 100 versus 300 o Properly coding the CS extension when the contralateral lung is not involved o The correct use of CS extension codes 100 versus 999 and 300. o Application of Ambiguous Terminology when coding involvement. o Coding atelectasis/obstructive pneumonitis when the entire lung is not involved. o Coding direct extension into the Visceral Pleura. o Capturing plural involvement in the CS extension codes o The correct use of TNM - NOS codes (115, 120, 460, 465, 570 and 810) o Coding 999 when a valid code is more appropriate CS Lymph Nodes o Applying the correct CS Lymph Node code that corresponds to the involved lymph node chain, such as: a hilar lymph node is coded to CS Lymph Node code 100 o Application of Note 2 in CS Lymph Nodes, “If at mediastinoscopy/x-ray, the description is ‘mass,’ ‘adenopathy,’ or ‘enlargement’ of any of the lymph nodes named as regional in codes 100 and 200, assume that at least regional lymph May 2014 2013 Lung Recoding Audit Final Report 9 California Cancer Registry 2013 Lung Recoding Audit Final Report nodes are involved. If there is any mention of bilateral or contralateral mass, adenopathy or lymph node involvement, use code 500.” o Correct use of the Inaccessible Lymph Node rule o Correct us of code 600 due to bilateral involvement SSF #2 (Pleural/Elastic Layer Invasion (PL) by H and E or Elastic Stain) o The misunderstanding of Note 2 in CS Site Specific Factor #2, “Code 998 if no histologic examination of pleura to assess pleural layer involvement” SSF #1 (Separate Tumor Nodules – Ipsilateral Lung) o Not applying Note 3 in CS Site Specific Factor #1 which states “If separate tumor nodules are not mentioned in imaging and/or pathological reports, use code 000”. o The coding of contralateral versus ipsilateral nodules in the coding of CS Site Specific Factor #1. o Coding of pleural implants in CS Site Specific Factor #1, instead of in CS Mets at Diagnosis. May 2014 2013 Lung Recoding Audit Final Report 10 California Cancer Registry 2013 Lung Recoding Audit Final Report Discrepancy Distribution between CS Data Items & Non-CS Data Items n=726 496 (68.3%) CS Data Items Non-CS Data Items 230 (31.7%) Graph 1 May 2014 2013 Lung Recoding Audit Final Report 11 California Cancer Registry 2013 Lung Recoding Audit Final Report Lung Recoding Audit Discrepancies by All Data Items N=720 Data Item CS Extension CS Mets at Diagnosis CS SSF #2 CS SSF #1 CS Lymph Nodes CS Mets Lung Diagnosis Date Regional Lymph Nodes Examined Scope Regional Lymph Node Primary Site/Subsite CS Tumor Size Regional Radiation Modality Regional Lymph Nodes Positive Chemotherapy Summary Laterality Grade CS Mets Bone CS Mets Liver Reason No Surgery Histology CS Mets Brain Radiation Sequence Diagnostic Confirmation Surgery Primary Site Regional Radiation Boost Behavior TOTAL TOTAL 115 68 74 57 45 30 28 29 28 25 24 25 22 21 19 17 11 12 12 12 9 11 9 10 7 0 720 Table 1 May 2014 2013 Lung Recoding Audit Final Report 12 California Cancer Registry 2013 Lung Recoding Audit Final Report Discrepancies in CS Data Items N=496 Data Item CS Extension CS SSF #2 CS Mets at Diagnosis CS SSF #1 CS Lymph Nodes CS Mets Lung Regional Lymph Nodes Examined CS Tumor Size Regional Lymph Nodes Positive CS Mets Liver CS Mets Bone CS Mets Brain TOTAL 115 74 68 57 45 30 29 24 22 12 11 9 TOTAL 496 Table 2 Discrepancies in Non-CS Data Items N=224 Data Item Diagnosis Date Scope Regional Lymph Node Primary Site/Subsite Regional Radiation Modality Chemotherapy Summary Laterality Grade Reason No Surgery Histology Radiation Sequence Surgery Primary Site Diagnostic Confirmation Regional Radiation Boost Behavior TOTAL TOTAL 28 28 25 25 21 19 17 12 12 11 10 9 7 0 224 Table 3 May 2014 2013 Lung Recoding Audit Final Report 13 California Cancer Registry 2013 Lung Recoding Audit Final Report CS Discrepancy on AJCC Stage N=496 CS Discrepancy Impact on AJCC Stage 496 500 450 400 350 300 Total CS Data Item Errors 250 Change in T - N - M 200 Change in AJCC Stage Grouping 150 74 100 24 50 0 Total CS Data Item Errors Change in T - N M Change in AJCC Stage Grouping Table 4 May 2014 2013 Lung Recoding Audit Final Report 14