Diagnosis Code for Family History of Blood Clots

• Research article
• Open Access
• Published: 08 June 2017

Limitations of pulmonary embolism ICD-10 codes in emergency department administrative data: allow the heir-apparent beware

BMC Medical Enquiry Methodology volume 17, Commodity number:89 (2017) Cite this commodity

• 3019 Accesses

• 34 Citations

• iii Altmetric

• Metrics details

Abstract

Background

Administrative data is a useful tool for research and quality improvement; still, validity of research findings based on these information depends on their reliability. Diagnoses assigned by physicians are subsequently converted past nosologists to ICD-10 codes (International Statistical Classification of Diseases and Related Health Problems, 10th Revision). Several groups take reported ICD-9 coding errors in inpatient data that take implications for research, quality improvement, and policymaking, but few have assessed ICD-10 code validity in ambulatory care databases. Our objective was to evaluate pulmonary embolism (PE) ICD-10 code accuracy in our large, integrated hospital organization, and the validity of using these codes for operational and wellness services enquiry using ED ambulatory care databases.

Methods

Ambulatory care information for patients (age ≥ xviii years) with a PE ICD-x code (I26.0 and I26.9) were obtained from the records of four urban EDs between July 2013 to Jan 2015. PE diagnoses were confirmed past reviewing medical records and imaging reports. In cases where chart diagnosis and ICD-10 code were discrepant, nautical chart review was considered correct. Physicians' written discharge diagnoses were besides searched using 'pulmonary embolism' and 'PE', and patients who were diagnosed with PE only not coded equally PE were identified. Coding discrepancies were quantified and described.

Results

One k, 4 hundred and fifty-three ED patients had a PE ICD-10 code. Of these, 257 (17.7%) were false positive, with an incorrectly assigned PE code. Among the 257 false positives, 193 cases had ambiguous ED diagnoses such as 'rule out PE' or 'query PE', while 64 cases should have had non-PE codes. An boosted 117 patients (eight.90%) with a PE belch diagnosis were incorrectly assigned a non-PE ICD-10 lawmaking (faux negative group). The sensitivity of PE ICD-10 codes in this dataset was 91.1% (95%CI, 89.4–92.half dozen) with a specificity of 99.9% (95%CI, 99.9–99.ix). The positive and negative predictive values were 82.3% (95%CI, eighty.3–84.two) and 99.9% (95%CI, 99.9–99.9), respectively.

Conclusions

Ambulatory care information, similar inpatient information, are field of study to coding errors. This confirms the importance of ICD-ten code validation prior to employ. The largest proportion of coding errors arises from ambiguous physician documentation; therefore, physicians and data custodians must ensure that quality comeback processes are in place to promote ICD-10 coding accurateness.

Peer Review reports

Background

The use of authoritative data for research provides multiple advantages: information technology is readily available, tin exist used to place big samples of patients over extended periods, and is relatively cheap to acquire. However, the utility of administrative data depends largely on its accuracy and reliability. Administrative database research often relies on diagnostic codes, now defined past the International Statistical Classification of Diseases and Related Health Bug, 10th Revision (ICD-10) [1].

The process for assigning diagnostic codes to patients visiting the emergency department (ED) is standardized in almost Canadian hospitals [2]. At the fourth dimension of patient discharge, ED physicians record a clinical description of the health problems (medico clinical notes) and write a discharge diagnosis or a provisional diagnosis pending further investigations. These belch diagnoses oftentimes neglect to adjust to ICD definitions, and in that location is potential for mistake when nosologists after translate them to ICD codes in infirmary administrative databases. Such coding errors may affect research validity, reported affliction trends, operational decisions, and wellness policies.

There is a growing body of research identifying errors associated with ICD-9 diagnostic code assignments in inpatient databases. Yet, few published studies accept assessed the accurateness of ICD-10 codes in ED authoritative information. Importantly, O'Malley et al. report that setting under which ICD codes are assigned is important; then although coding practices are similar for inpatient and ambulatory care settings, each has unique sources of error [three]. Thus, quantification of ICD-10 code accuracy in ED administrative data is important.

Pulmonary embolism (PE) is a potentially life-threatening disease that is diagnostically challenging. With symptoms including dyspnea, chest hurting, palpitations, hemoptysis, and or syncope, PE is considered in the differential diagnosis of many cardiopulmonary presentations, thus clinical inquiry to amend PE diagnosis and treatment remains of import [iv]. Administrative data is a powerful tool for studying PE. The accuracy of diagnostic codes has been well-defined in inpatient data; however, the undifferentiated emergency patient differs from a hospitalized patient, who is more probable to exist diagnosed with a PE equally a outcome of their well-recognized increased risk for developing PE. Thus, population differences, in combination with the inherent differences in coding errors inside convalescent and inpatient data, make it hard to directly compare the validity of ICD codes.

The objective of this study was to assess the validity of PE ICD-ten diagnostic codes every bit a sole means for identifying diagnoses in ED administrative data.

Methods

Data source

Authoritative data were obtained from the Convalescent Care Database of the Alberta Health Services Calgary Zone, located in the province of Alberta, Canada. Given that Alberta has a health care insurance plan that covers all healthcare costs, Alberta Health Services (AHS) is the single health authority in the province and its database includes over 99% of Alberta residents [v]. The Calgary Zone includes four adult acute care hospitals serving a similar demographic population of approximately one.two million people, and seeing approximately 325,000 ED visitors yearly. Data extraction included patient age and sex, engagement and location of hospital visit, presenting complaint, triage note, medico's written belch diagnosis, and the subsequently assigned ICD-x codes. Physician's clinical notes were likewise reviewed, providing a narrative and more complete clarification of the patient's disposition. Eligible patients were over 18 years of historic period with an ED visit between July 2013 and January 2015. In Calgary, and beyond Canada, a coordinator is responsible for establishing and maintaining consistent coding practices for authoritative data [2].

Identification of true positive and imitation positive PEs

We used ICD-10 codes (I26.9: pulmonary embolism without cor pulmonale, and I26.0: pulmonary embolism with cor pulmonale) to identify patients diagnosed with PE; we refer to these identified patients every bit the coded PE group (Fig. ane). Inside the coded PE group, nosotros identified true positives by comparing physicians' clinical notes and written discharge diagnoses to the assigned ICD-10 code. Patients with a congruent PE ICD-10 code were assigned to the truthful positive PE group (Fig. ane). Patients who did not have a PE according to the physicians' clinical notes and written discharge diagnoses but had a PE ICD-10 lawmaking were assigned to the false positive PE group. The false positive PE group was comprised of two populations: the commencement had written belch diagnoses that were obviously not-PE (miscoded PE group, Fig. 1). The 2d grouping was comprised of discrepant cases where the physician's written discharge diagnosis was unclear—for example, 'rule out PE' or 'query PE' (query PE group, Fig. 1). Patient medical records and imaging reports were reviewed by two trained investigators to confirm PE diagnosis. For patients whose records indicated a diagnosis other than PE, the most correct diagnosis was documented and the PE ICD-10 lawmaking was considered incorrect. An experienced ED physician adjudicated cases of disagreement and complex cases requiring additional expertise.

Fig. 1

Validation of PE ICD-10 codes. Coded PE: patients assigned I26.0 or I26.nine ICD-10 code; True Positive PE: patients assigned a PE ICD-10 lawmaking whose diagnosis was PE; False Positive PE: patients whose nautical chart revealed a diagnosis other than PE; Miscoded PE: patients who should have been assigned an alternative lawmaking; Query PE: patients assigned a PE diagnostic code workup revealed a probable absence of PE; No PE ICD-x code: patients not assigned I26.0 or I26.nine ICD-10 code; False Negative PE: patients with a PE who were not assigned a PE ICD-x code; True Negative PE: patients not diagnosed with PE and not assigned a PE ICD-x; Validation Negative PE: patients non diagnosed with PE (may or may non be reflected by their ICD-ten code consignment); Validation Positive PE: patients diagnosed with PE (may or may non exist reflected by their ICD-10 code consignment)

Full size image

Identification of true negative and false negative Human foot

Patients with ICD-x codes for diagnoses other than PE were identified; we refer to these identified patients equally the no PE ICD-10 code group (Fig. 1). To identify PE cases 'missed' past using ICD-10 codes, we performed a free-text search of the physician discharge diagnosis field, looking for the keywords 'PE', 'pulmonary embolism', 'pulmonary', and 'embolism'. Patients diagnosed with PE without an ICD-10 code for PE were moved to the simulated negative PE group (Fig. one). The remainder of patients in this written report made up the truthful negative PE group (Fig. 1). This group, together with the false positive PE group, comprised the validation negative PE grouping (Fig. 1). Similarly, the true positive PE and imitation negative PE groups comprised the validation positive PE group (Fig. one).

Statistical analysis

4 strategies can be used to identify patients diagnosed with PE in administrative data (Tabular array 3). For each strategy, sensitivity (SN), specificity(SP), and positive and negative predictive values (PPV and NPV) with 95% conviction intervals (CIs), were calculated using MedCalc, version 15.eleven.4 (MedCalc Software, Confirm, Belgium. Accessed November 29, 2016 at https://www.medcalc.org/calc/diagnostic_test.php). Confidence intervals for sensitivity and specificity are "exact" Clopper-Pearson confidence intervals. Conviction intervals for the predictive values are the standard logit confidence intervals as previously described [6].

Results

Between July 2013 and Jan 2015, 479,937 patients visited Calgary EDs, and 1453 (0.xxx%) received a PE ICD-10 diagnostic code (Fig. 1). However, assay of the raw data revealed that a subset of patients identified using PE codes were not diagnosed with PE (257 (17.7%)); these errors were preventable, since 64 patients who should have been assigned an culling lawmaking for diagnoses unrelated to PE; for instance chest pain, pleural effusion, anxiety, or substance corruption (Table one). Furthermore, 4 of the 64 miscoded patients were randomly assigned a PE ICD-10 code, only their written belch diagnosis was blank and triage and physician disposition notes indicated no suspicion of PE. The remaining 193 patients were misinterpreted equally being positive for PE during code abstraction; they had negative PE investigations and their actual diagnosis remained unspecified. A free text search of the database identified an additional 404 Query PE cases; however, these cases were more accordingly coded equally breast pain or dyspnea. Thus, the other 67.7% of Query PE patients received an advisable ICD-10 diagnostic code. Conversely, we identified 117 patients (8.9%) with PE who were non assigned a PE lawmaking during abstraction; only rather an unrelated ICD-10 lawmaking (Table 2). Furthermore, 33 patients diagnosed with PE that were assigned no diagnostic codes at all.

Table one Summary of patients assigned a PE ICD-ten code who should have been assigned an alternative diagnostic code (Miscoded Foot)

Full size table

Table two Summary of the Fake Negative PE group: patients with PE who were assigned ICD-x codes other than PE

Total size tabular array

We proposed iv strategies that could be used to place PE patients in administrative data. The showtime utilized PE ICD-10 codes with no farther verification (Table 3 – strategy A). Causeless accuracy refers to the perception of accuracy held past the investigator who assumes ICD-x codes are correct; in this state of affairs, SN, SP, PPV, and NPV appeared to be 100.00% (95%CI, 100.00–100.00) (Table three – strategy A, assumed accurateness). However, we demonstrated that SN and PPV following validation were actually 91.one% (95%CI, 89.4–92.6) and 82.3% (95%CI, 80.3–84.two), respectively (Table 3 – Strategy A, true accuracy). The assumed accuracy values for strategies B and C were also 100.00% (95%CI, 100.00–100.00) (Table iii – strategy B, C, assumed accuracy); reflecting that without validation, the investigator unknowingly studies PE using inaccurate data. Notably, the truthful PPV for strategy B, which included a step to place patients with missed PEs (i.e.: the false negative PE group (Fig. 1)), was was 83.vi% (95%CI, 81.9–85.two) (Table 3 – Strategy B, truthful accuracy). Conversely, the SN for strategy C, which instead included a step to remove incorrectly coded PEs (i.e.: the false positive PE group (Fig. 1)) was 91.1% (95%CI, 89.4–92.six). Finally, the SN, SP, PPV, and NPV for strategy D, which involved complete validation with missed and incorrectly PE patients beingness re-assigned to the appropriate grouping, were 100.00% (95%CI, 100.00–100.00) (Table iii – strategy D).

Table 3 Strategies to identify PE patients

Full size table

Discussion

ICD-10 codes are widely used in research involving administrative information and are assumed to be an accurate reflection of affliction incidence in the population studied, but several groups have identified ICD coding errors as a threat to inquiry validity. Our calculated SN of 91% is coinciding with reported sensitivities of PE ICD-9 codes in inpatient and postal service-operative patient databases (62–92%) [vii,8,9,10]. Our calculated SN is on the higher end, perhaps reflecting the the expanded listing of diagnostic codes in the ICD-10 schedule, meant to amend SN by allowing for more than specific diagnostic code assignment. Nosotros also identified a fake positive PE group, meaning that the PPV of PE ICD-10 codes was only 82.three%. Scarveli et al. assessed PE ICD-9 codes in inpatient information and similarly calculated a PPV of 80.5% [11]. They reported a false positive charge per unit of 18.5% [11]. Casez et al. determined that inpatient PE ICD-10 codes were 89% sensitive for PE, concluding that this is sufficient to use these codes to place PE patients, although they did not make up one's mind the rate of fake negatives in their data [12]. Our work builds on the work of Casez et al., though we are more than conservative in our conclusions, instead encouraging researchers to validate ED authoritative data prior to research use.

Our findings propose that strategies to prevent coding errors are necessary. Diligence during code abstraction and a requirement for imaging confirmation would reduce the number of false positive miscodes (64 in our written report). Most false positives in our study were in the query PE group — the result of ambiguous physician discharge diagnoses. Health information nosologists may add a "Q" prefix before ICD-10 codes to indicate "query" diagnoses or diagnostic doubtfulness. Increased use of this prefix may accost a big proportion of these errors, equally would clear management to nosologists as to the types of written diagnosis (e.g. "query PE" or "rule out PE") that would be appropriate to lawmaking with a Q prefix. Further utilization of ICD-x codes for vague or uncertain diagnoses may too aid these errors (see Chapter XVIII: Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified (R00-R99) [1]).

Previous piece of work has shown that code abstraction depends more on the quality of doctor documentation than on the judgment or experience of the nosologist [13, fourteen]. Considering the largest proportion of coding errors appear to consequence from cryptic documentation, physicians need to understand the importance of their written diagnosis. Specifically, in cases where a diagnosis is not yet confirmed, physicians should have a standard approach to documentation. For instance, a diagnosis of dyspnea NYD or Breast Pain NYD would exist preferable to 'rule out PE', and would betoken nosologists to look for additional confirmation. Periodic departmental audit and feedback of physician diagnostic coding could as well place situations more likely to contribute to ambiguous documentation. For instance, in situations where a physician hands over to a 2d md earlier the results of diagnostic tests are available, the second physician may non render to the nautical chart and modify the outset doc'south tentative discharge diagnosis ('?PE'). This problem could be largely eliminated by a policy that precludes writing in the discharge diagnosis field until investigations are complete.

If implemented, changes to ICD-10 code abstraction will take time; but there will exist no break in use of administrative data. Thus, we presented four strategies that can exist used for studying PE using authoritative data: A) utilize PE ICD-10 codes alone to identify patients diagnosed with PE, accepting that saving fourth dimension occurs at the expense of accuracy; B) employ strategies such as keyword identification to identify missed Foot, recognizing the existence of the faux positive grouping; C) instead remove false positive patients by reviewing patient charts and imaging studies; acknowledging that false negative patients might be missed; or D) utilise a consummate validation strategy equally described in this manuscript, which will be more than accurate, at the expensive of increased time and resources. Given the modest number of patients diagnosed with PE in a relatively large database, SP and NPV were of niggling utilize in our study. We propose that studies requiring accuracy, including those which assesses individual patient characteristics, might benefit from strategy D. On the other mitt, strategy A may be more suitable for studies concerned with the number of patients diagnoses rather than individual demographics, such as those monitoring interventions and disease trends.

This written report was limited because we did non seek imaging confirmation for all 1453 cases who had a PE ICD-10 codes. Rather, we only reviewed cases where the ICD-10 code was not coinciding with the physician'south discharge diagnosis, significant that we may have missed boosted fake positive patients. Also, our report reflects the work of physicians and nosologists in one Canadian region. However, understanding with findings of other groups suggests that our findings can likely be applied generally. Like other investigators, we were constrained by the quality of documentation within medical charts, which often lacked more detailed information regarding the physicians' diagnostic thought process. Though the power of our statistical analyses were limited by the low prevalence of PE in a very large database, our methodology could be practical during validation of other ICD-ten codes. Diagnoses with increased prevalence would benefit from using our strategies for validation.

Depending on the application, the false positive and negative rates seen in our data are a potential threat to validity of PE studies or initiatives that rely on administrative data. Nosotros suspect that not-PE ICD-10 codes are not allowed to coding errors, and propose that a validation strategy be employed when using authoritative data. Researchers and healthcare administrators should use caution in using ICD-10 codes from ED ambulatory care databases to place diagnostic groups without verifying the accuracy of ICD-10 coding.

Conclusions

Our written report shows that convalescent care data, like inpatient data, are subject to coding errors, and confirms the importance of validating ICD-10 diagnostic code accuracy prior to use for enquiry purposes. We demonstrate iv strategies for validating ICD-10 codes in administrative information, and these strategies can be applied broadly. The largest proportion of coding errors arises from ambiguous physician documentation; therefore, physicians and data custodians must ensure that quality improvement processes are in place to promote accurateness of ICD-10 diagnostic coding.

Abbreviations

CI:

Confidence interval

ED:

Emergency department

ICD-10 codes:

International Statistical Nomenclature of Diseases and Related Wellness Issues, tenth Revision

NPV:

Negative predictive value

PE:

Pulmonary embolism

PPV:

Positive predictive value

SN:

Sensitivity

SP:

Specificity

References

1. Earth Health System. International Statistical Classification of Diseases and Related Health Problems 10th revision teaching manual. Geneva; 2010.

2. Canadian Coding Standards for ICD-10-CA and CCI. [https://secure.cihi.ca/estore/productSeries.htm?pc=PCC189].

3. O'Malley KJ, Cook KF, Cost MD, Wildes KR, Hurdle JF, Ashton CM. Measuring diagnoses: ICD lawmaking accurateness. Health Serv Res. 2005;40(5 Pt 2):1620–39.

   Article  PubMed  PubMed Central  Google Scholar

4. Kline JA: Pulmonary embolism and deep vein thrombosis. In: Rosen's emergency medicine - concepts and clinical practice. Volume 1, viii edn. Edited by Marx JA: Elsevier, Saunders; 2014: 1157-1169.

5. Alberta Health and Wellness. Data disclosure handbook; 2003. p. one–15.

   Google Scholar

6. Mercaldo ND, Lau KF, Zhou XH. Confidence intervals for predictive values with an emphasis to case-control studies. Stat Med. 2007;26(10):2170–83.

   Article  PubMed  Google Scholar

7. Sadeghi B, White RH, Maynard G, Zrelak P, Strater A, Hensley 50, et al. Improved coding of postoperative deep vein thrombosis and pulmonary embolism in administrative data (AHRQ patient rubber indicator 12) afterward introduction of new ICD-ix-CM diagnosis codes. Med Intendance. 2015;53(v):e37–40.

8. Kniffin WD Jr, Baron JA, Barrett J, Birkmeyer JD, Anderson FA Jr. The epidemiology of diagnosed pulmonary embolism and deep venous thrombosis in the elderly. Arch Intern Med. 1994;154(8):861–vi.

   Article  PubMed  Google Scholar

9. Arnason T, Wells PS, van Walraven C, Forster AJ. Accuracy of coding for possible warfarin complications in hospital discharge abstracts. Thromb Res. 2006;118(2):253–62.

   CAS  Article  PubMed  Google Scholar

10. Tagalakis V, Kahn SR. Determining the test characteristics of claims-based diagnostic codes for the diagnosis of venous thromboembolism in a medical service claims database. Pharmacoepidemiol Drug Saf. 2011;20(3):304–7.

    Article  PubMed  Google Scholar

11. Scarvelis D, Anderson J, Davis 50, Forgie M, Lee J, Petersson L, et al. Hospital mortality due to pulmonary embolism and an evaluation of the usefulness of preventative interventions. Thromb Res. 2010;125(two):166–70.

    CAS  Commodity  PubMed  Google Scholar

12. Casez P, Labarere J, Sevestre MA, Haddouche M, Courtois X, Mercier S, et al. ICD-x infirmary belch diagnosis codes were sensitive for identifying pulmonary embolism but not deep vein thrombosis. J Clin Epidemiol. 2010;63(7):790–7.

    Article  PubMed  Google Scholar

13. Hennessy DA, Quan H, Faris PD, Beck CA. Do coder characteristics influence validity of ICD-10 hospital discharge data? BMC Health Serv Res. 2010;10:99.

    Article  PubMed  PubMed Central  Google Scholar

14. And so L, Beck CA, Brien S, Kennedy J, Feasby TE, Ghali WA, et al. Chart documentation quality and its relationship to the validity of administrative information belch records. Health Informatics J. 2010;sixteen(2):101–13.

Download references

Acknowledgements

We would like to thank Dongmei Wang for information extraction and Daniel Grigat and James Andruchow for feedback and guidance during the course of this projection. KB would similar to give thanks Mark Berkley for his assistance in managing large-scale databases in Excel, and his encouragement along the path of ICD-10 discovery.

Funding

This work was funded by Alberta Innovates - Wellness Solutions, through a Partnerships for Research and Innovation in Wellness Systems inquiry grant.

Availability of data and materials

The datasets during and/or analyzed during the current study cannot be released, as sharing of identifiable private information is not permitted per the terms of our REB approval and information sharing agreement with AHS.

Authors' contributions

KB conceived the projection and contributed to project pattern, reviewed medical records, analyzed and interpreted the data, and drafted and revised the manuscript. GI contributed to projection pattern, data interpretation, and made substantial contributions to manuscript revision. KS fabricated a substantial contribution to medical record review and data estimation. EL conceived the project contributed to project design and data estimation. AM contributed substantially to project design, data estimation, and manuscript revision. All authors read and approved the final manuscript.

Authors' data

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Consent for publication

Not applicable.

Ethics approval and consent to participate

Ideals approval for this study was obtained through the Academy of Calgary's Conjoint Wellness Enquiry Ethics Board (REB 14–0650). Patient consent was not required for this study every bit we performed a secondary assay of existing information.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Writer information

Affiliations

1. Cumming Schoolhouse of Medicine, University of Calgary, Calgary, Alberta, Canada

   Kristin Burles, Grant Innes, Boil Lang & Andrew McRae

2. Alberta Health Services, Section of Emergency Medicine, Calgary, Alberta, Canada

   Grant Innes, Kevin Senior, Eddy Lang & Andrew McRae

3. Emergency Strategic Clinical Network, Alberta Health Services, Calgary, Canada

   Grant Innes, Eddy Lang & Andrew McRae

4. Emergency Section, C231, Foothills Medical Centre, 1403 29 Street NW, Calgary, Alberta, T2N 2T9, Canada

   Andrew McRae

Corresponding author

Correspondence to Andrew McRae.

Rights and permissions

Open Access This article is distributed nether the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in whatever medium, provided you give advisable credit to the original author(southward) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/i.0/) applies to the data made available in this commodity, unless otherwise stated.

Reprints and Permissions

Well-nigh this article

Cite this article

Burles, K., Innes, G., Senior, 1000. et al. Limitations of pulmonary embolism ICD-x codes in emergency section administrative data: let the buyer beware. BMC Med Res Methodol 17, 89 (2017). https://doi.org/10.1186/s12874-017-0361-1

Download citation

• Received: 05 October 2016

• Accustomed: 23 May 2017

• Published: 08 June 2017

• DOI : https://doi.org/x.1186/s12874-017-0361-1

Keywords

• Pulmonary embolism
• PE
• ICD-10
• Miscoding

johnsonsount1937.blogspot.com

Source: https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-017-0361-1

Share this post