An administrative claims study found that using International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification diagnosis codes to identify laboratory-confirmed chlamydial and gonococcal infections substantially underestimates the burden of these diseases and misclassifies infections.
Supplemental digital content is available in the text.
Background
Because most sources of administrative claims data do not contain laboratory result data, researchers rely on diagnosis codes to identify cases of disease. The validity of using diagnosis codes to identify chlamydial and gonococcal infections in administrative claims data remains largely uninvestigated.
Methods
We conducted a retrospective cohort analysis using OptumLabs Data Warehouse, which includes deidentified medical (inpatient and outpatient) claims and laboratory test results. Among males and females aged 15 to 39 years during the period 2003–2017, we identified chlamydia and gonorrhea test results and corresponding diagnosis codes. Using test results as the criterion standard, we calculated the sensitivity and specificity of chlamydia and gonorrhea diagnosis codes to identify laboratory-confirmed infections.
Results
We identified 9.7 million chlamydia and gonorrhea test results among 3.1 million enrollees. Of the 176,241 positive chlamydia test results, only 11,515 had a corresponding diagnosis code, for a sensitivity of 6.5 (95% confidence interval [CI], 6.4–6.7) and a specificity of 99.5 (95% CI, 99.5–99.5). Corresponding diagnosis codes were identified for 8056 of the 31,766 positive gonorrhea test results, for a sensitivity of 25.4 (95% CI, 24.9–25.8) and a specificity of 99.7 (95% CI, 99.7–99.7).
Conclusions
Our findings indicate that using only International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification diagnosis codes to identify chlamydial and gonococcal infections substantially underestimates the burden of these diseases and inaccurately classifies laboratory-confirmed infections.
Administrative health care claims are generated during routine health care encounters (including outpatient physician visits, hospital admissions, and pharmaceutical dispensing events) and collected for administrative billing and reimbursement purposes.1,2 Medical diagnoses and procedures are recorded in administrative claims data using standardized coding vocabularies, such as International Classification of Diseases, Ninth and Tenth Revisions, Clinical Modification (ICD-CM) codes. Administrative claims data are routinely collected and more readily available compared with other sources of data that rely on more labor-intensive practices such as medical chart reviews or patient surveys. In general, administrative claims databases are relatively affordable and provide large, population-based data sets. Because of the accessibility and availability, administrative claims data are being increasingly used for research and quality of care evaluation purposes. Because most sources of administrative claims data do not contain laboratory result data, researchers primarily rely on ICD-CM diagnosis codes to identify cases of various diseases and chronic conditions, estimate prevalence, and examine temporal trends.3–6 Researchers have also used identified cases to establish outcome-specific cohorts to examine potential risk factors,7–9 associated health care services,10–12 and the effect of pharmaceutical interventions.13,14
Administrative claims data have been used to investigate both Chlamydia trachomatis and Neisseria gonorrhoeae (CT/GC),15–20 2 common bacterial sexually transmitted infections (STIs) that are diagnosed by laboratory-based nucleic acid amplification tests. Because administrative claims data often lack laboratory test results, researchers have primarily relied on ICD-CM diagnosis codes to identify chlamydial and gonococcal infections. A recent study using 1 year of data from one US state found a low sensitivity of ICD-10-CM diagnosis codes for identifying laboratory-confirmed CT/GC infections.21 We expand on that work by using 15 years of data from a geographically diverse data source, investigating the validity of ICD-9-CM diagnosis codes to be used in temporal analyses using data before 2015, and identifying factors associated with the validity of these diagnosis codes.
To investigate the validity of CT/GC diagnosis codes, we used an administrative claims data set containing both ICD-CM diagnosis codes and laboratory test results. Using laboratory test results as the reference standard, we determined the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of corresponding CT/GC ICD-CM diagnosis codes for laboratory-confirmed chlamydial and gonococcal infections.
MATERIALS AND METHODS
Data Source
We conducted a retrospective cohort study using the OptumLabs Data Warehouse (OLDW), an administrative claims database containing deidentified, longitudinal health information on enrollees representing a diverse mixture of ages, ethnicities, and geographical regions across the United States.22 The OLDW contains medical (inpatient and outpatient) and pharmaceutical claims, laboratory test results, and enrollment records for commercial health insurance and Medicare Advantage enrollees.22 Although administrative claims are available for all enrollees, laboratory test results are only available for a subset of enrollees in the OLDW. Only laboratory test results associated with contracted vendors are included in the OLDW. Demographic factors (sex, year of birth, census region of residence, and enrollment dates) were obtained from enrollment records. Medical claims include ICD-CM diagnosis codes (ICD-9-CM for 2003–2014 and ICD-10-CM for 2015–2017), as well as procedure (ICD-CM, Current Procedural Terminology, Version 4, and Healthcare Common Procedure Coding System) codes. Laboratory test results contain Logical Observation Identifiers Names and Codes. All diagnosis and procedure codes used in this analysis are specified in the supplemental materials (Supplemental Tables S3–S5, http://links.lww.com/OLQ/A669).
Study Population
We identified all commercially insured males and females aged 15 to 39 years during the study period January 1, 2003, through December 31, 2017. Continuous enrollment was not required. We used their enrollee identification numbers to extract all laboratory test results during the study period and determine the availability of laboratory result data for our study population. To investigate if the limited availability of laboratory result data may have systematically biased our findings, we compared eligible enrollees with and without any available laboratory test results by sex, age, and census region of residence. We then identified CT/GC test results using Logical Observation Identifiers Names and Codes (Supplemental Tables S3, S4, http://links.lww.com/OLQ/A669). Age at the time of CT/GC testing was estimated using enrollee year of birth and test result service dates; only results that met our study age criterion for enrollees were included in our analytic data set. Thus, our analytic data set contained all CT/GC laboratory test results from males and females aged 15 to 39 years during the study period.
Data Analysis
C. trachomatis and N. gonorrhoeae test results were used as the unit of analysis. We categorized all CT/GC test results as positive, negative, or other (including missing result values). Infection status was determined using test result values (e.g., a positive test result was considered to indicate a laboratory-confirmed infection). To assess the validity of using ICD-CM diagnosis codes to identify laboratory-confirmed chlamydial and gonococcal infections, we first linked test results to all medical claims occurring within a 21-day period (7 days before through 14 days after the test result date). For each test result, we determined the presence or absence of at least one corresponding CT/GC ICD-CM diagnosis code within the linked medical claims. Using test results as the reference standard, we then determined the sensitivity, specificity, PPV, and NPV of CT/GC ICD-CM diagnosis codes. Test results categorized as other were excluded from the diagnosis code validity analysis, as infection status could not be determined. We conducted additional analyses stratifying diagnosis codes by ICD-9-CM and ICD-10-CM.
To explore factors associated with the validity of CT/GC diagnosis codes, we examined the prevalence of corresponding diagnosis codes among confirmed chlamydial and gonococcal infections by demographic factors and sex-specific concurrent diagnoses for pelvic inflammatory disease (PID) and infection-related symptoms (dysuria, vaginitis, vulvovaginitis, and urethritis). The diagnosis codes of the ICD-CM for these concurrent diagnoses were identified within the previously linked medical claims within the 21-day period of interest (Supplemental Table S5, http://links.lww.com/OLQ/A669). Among confirmed chlamydial and gonococcal infections, we also identified concurrent GC and CT ICD-CM codes, respectively. Beginning in 2012, the recommended treatment of uncomplicated GC includes an intramuscular injection of 250 mg of ceftriaxone23; for CT/GC testing occurring after December 31, 2011, we identified this treatment within the linked medical claims using Healthcare Common Procedure Coding System code J0696.
The frequency of laboratory result data availability, demographic factors, CT/GC test result positivity, and the presence or absence of a corresponding CT/GC ICD-CM diagnosis code were assessed using counts and percentages. We used log-binomial logistic regression to calculate the prevalence of corresponding diagnosis codes and prevalence ratios (PRs) with 95% confidence intervals (CIs) among confirmed infections by demographic and health-related factors. All analyses were conducted using SAS Version 9.4 (SAS Institute Inc, Cary, NC). Because the data were deidentified, the Centers for Disease Control and Prevention deemed this study to be research not involving human subjects; thus, institutional review board review was not required.
RESULTS
We identified 37.4 million commercially insured males and females aged 15 to 39 years during the period January 1, 2003, through December 31, 2017 (Table 1). Among these enrollees, 12.4 million (33.1%) had at least one laboratory test result for any health condition during the study period, and 3.1 million enrollees (8.3%) had at least one CT/GC test result. We compared eligible enrollees with and without any available laboratory result data by sex, age, and census region of residence and did not identify any meaningful differences (data not shown).
TABLE 1.
Demographic Characteristics of Study Population, OptumLabs Data Warehouse, 2003 to 2017
| Enrollees Aged 15–39 y | Enrollees Aged 15–39 y With >1 Test Result(s)* | Enrollees Aged 15–39 y With >1 CT/GC Test Result(s)† | ||||
|---|---|---|---|---|---|---|
| n | % | n | % | n | % | |
| Total | 37,409,990 | 12,383,586 | 3,117,701 | |||
| Sex | ||||||
| Female | 18,739,791 | 50.1 | 8,130,653 | 65.7 | 2,659,971 | 85.3 |
| Male | 18,670,199 | 49.9 | 4,252,933 | 34.3 | 457,730 | 14.7 |
| Census region of residence | ||||||
| Midwest | 8,771,934 | 23.4 | 2,003,875 | 16.2 | 426,968 | 13.7 |
| Northeast | 4,404,888 | 11.8 | 1,824,249 | 14.7 | 497,214 | 16.0 |
| South | 16,190,164 | 43.3 | 6,572,754 | 53.1 | 1,691,959 | 54.3 |
| West | 6,413,630 | 17.1 | 1,953,912 | 15.8 | 493,911 | 15.8 |
| Unknown | 1,629,374 | 4.4 | 28,796 | 0.2 | 7649 | 0.2 |
| Median | (SD) | Median | (SD) | Median | (SD) | |
| Age, y | ||||||
| Female | 27.5 | (7.6) | 29.5 | (7.6) | 29.5 | (7.6) |
| Male | 28.0 | (7.7) | 30.0 | (8.5) | 30.0 | (8.5) |
| Enrollment length, y | 1.4 | (3.0) | 2.7 | (3.6) | 2.7 | (3.6) |
| No. CT/GC laboratory test results | N/A | N/A | 2.0 | (2.6) | ||
*Any laboratory test result(s) for any health condition during the period 2003–2017.
†CT/GC laboratory test result(s) during the period 2003–2017.
CT/GC indicates chlamydia and/or gonorrhea; n, number; N/A, not applicable; No., number.
Among the 3.1 million enrollees with at least one CT/GC test result, we identified 5.0 million CT and 4.7 million GC laboratory test results during the study period (Tables 1, 2). C. trachomatis and N. gonorrhoeae test results were primarily among enrollees that were female (85.3%) and enrollees from the south (54.3%). Median enrollment length during the study period was 2.7 years. Overall test result positivity was 3.5% for CT and 0.7% for GC. A majority of laboratory-confirmed chlamydial (62.1%) and gonococcal (48.9%) infections were among enrollees aged 15 to 24 years (data not shown). Although a majority of chlamydial infections were among females (71.7%), a majority of gonococcal infections were among males (60.5%; data not shown).
TABLE 2.
Laboratory Test Results With Diagnosis Code Validity Measures, by Sexually Transmitted Infection, Test Result Value, and Presence or Absence of Corresponding Diagnosis Code, OptumLabs Data Warehouse, 2003 to 2017
| Corresponding Diagnosis Code(s)* | Measures of Validity of Diagnosis Codes† | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Test Results | Yes | No‡ | Sensitivity | Specificity | PPV | NPV | |||||||||
| n | Col. % | n | Row % | n | Row % | Value | 95% CI | Value | 95% CI | Value | 95% CI | Value | 95% CI | ||
| Chlamydia | |||||||||||||||
| Positive | 176,241 | 3.5 | 11,515 | 6.5 | 164,726 | 93.5 | 6.5 | 6.4–6.7 | 33.0 | 32.5–33.5 | |||||
| Negative | 4,450,150 | 88.8 | 23,366 | 0.5 | 4,426,784 | 99.5 | 99.5 | 99.5–99.5 | 96.4 | 96.4–96.4 | |||||
| Other§ | 384,089 | 7.7 | 4720 | 1.2 | 379,369 | 98.8 | |||||||||
| Total | 5,010,480 | 100 | |||||||||||||
| Gonorrhea | |||||||||||||||
| Positive | 31,766 | 0.7 | 8056 | 25.4 | 23,710 | 74.6 | 25.4 | 24.9–25.8 | 35.5 | 34.9–36.0 | |||||
| Negative | 4,360,138 | 92.7 | 14,671 | 0.3 | 4,345,467 | 99.7 | 99.7 | 99.7–99.7 | 99.5 | 99.5–99.5 | |||||
| Other§ | 311,958 | 6.6 | 2923 | 0.9 | 309,035 | 99.1 | |||||||||
| Total | 4,703,862 | 100 | |||||||||||||
*Corresponding ICD-CM diagnosis code(s) identified within a 21-day period of interest ranging from 7 days before to 14 days after the laboratory test result date.
†Test results categorized as “other” were excluded from the diagnostic code measures of validity as infection status could not be determined.
‡No ICD-CM diagnosis code identified within the 21-day period of interest, described above, because of either an absence of diagnosis code(s) or an absence of any medical claims within the 21-day period.
§All test result values that could not be classified as “positive” or “negative,” including missing (i.e., blank) values, were classified as “other.”
95% CI indicates 95% confidence interval; Col, column; n, number; PPV, positive predictive value; NPV, negative predictive value.
The sensitivity of the ICD-CM diagnosis codes was lower for CT (6.5; 95% CI, 6.4–6.7) compared with GC (25.4; 95% CI, 24.9–25.8), whereas the specificity was comparable (CT: 99.5 [95% CI, 99.5–99.5]; GC: 99.7 [95% CI, 99.7–99.7]; Table 2). The PPV of the ICD-CM diagnosis codes was lower for CT (33.0; 95% CI, 32.5–33.5) than GC (35.5; 95% CI, 34.9–36.0), as was the NPV (CT: 96.4 [95% CI, 96.4–96.4]; GC: 99.5 [95% CI, 99.5–99.5]). When stratified by ICD-9-CM versus ICD-10-CM, the results were similar; however, the sensitivity and PPV was slightly higher for both CT and GC ICD-10-CM diagnosis codes (Supplemental Tables S1, S2, http://links.lww.com/OLQ/A669).
Corresponding CT diagnosis codes among confirmed chlamydial infections were more prevalent among males than females (PR, 1.5; 95% CI, 1.4–1.5; Table 3). Concurrent GC (PR, 4.0; 95% CI, 3.8–4.3) and PID (PR, 5.2; 95% CI, 4.8–5.7) diagnoses were associated with the prevalence of a corresponding CT diagnosis code. Among confirmed chlamydial infections, concurrent GC treatment was associated with the prevalence of a corresponding CT diagnosis code (PR, 1.6; 95% CI, 1.5–1.7).
TABLE 3.
Prevalence of Corresponding Diagnosis Code by Sexually Transmitted Infection, Demographic Factors, and Concurrent Diagnoses Among Positive Laboratory Test Results, OptumLabs Data Warehouse, 2003 to 2017
| Positive Chlamydia Laboratory Test Results* | Positive Gonorrhea Laboratory Test Results* | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| No. Test Results | Corresponding Diagnosis Code† |
PR | No. Test Results | Corresponding Diagnosis Code† | PR | |||||
| n | % Prevalence | PR | 95% CI | n | % Prevalence | PR | 95% CI | |||
| Total | 171,212 | 11,515 | 6.7 | 30,828 | 8056 | 26.1 | ||||
| Sex | ||||||||||
| Female | 122,715 | 7248 | 5.9 | Ref. | 12,182 | 3727 | 30.6 | Ref. | ||
| Male | 48,497 | 4267 | 8.8 | 1.5 | 1.4–1.5 | 18,646 | 4329 | 23.2 | 0.8 | 0.7–0.8 |
| Age group, y | ||||||||||
| 15–24 | 106,261 | 6994 | 6.6 | Ref. | 15,089 | 3885 | 25.7 | Ref. | ||
| 25–29 | 37,838 | 2591 | 6.8 | 1.0 | 1.0–1.1 | 7476 | 1906 | 25.5 | 1.0 | 0.9–1.0 |
| 30–39 | 27,113 | 1930 | 7.1 | 1.1 | 1.0–1.1 | 8254 | 2265 | 27.4 | 1.1 | 1.0–1.1 |
| Census region | ||||||||||
| South | 105,258 | 6968 | 6.6 | Ref. | 20,058 | 4864 | 24.2 | Ref. | ||
| Midwest | 25,269 | 1368 | 5.4 | 0.8 | 0.8–0.9 | 3546 | 915 | 25.8 | 1.1 | 1.0–1.1 |
| Northeast | 17,021 | 1363 | 8.0 | 1.2 | 1.1–1.3 | 3503 | 1139 | 32.5 | 1.3 | 1.3–1.4 |
| West | 23,368 | 1783 | 7.6 | 1.2 | 1.1–1.2 | 3669 | 1118 | 30.5 | 1.3 | 1.2–1.3 |
| Unknown | 296 | 33 | 11.1 | 1.7 | 1.2–2.3 | 52 | 20 | 40.0 | 1.6 | 1.1–2.2 |
| Concurrent diagnosis† | ||||||||||
| Chlamydia | ||||||||||
| No | N/A | 29,526 | 7251 | 24.6 | Ref. | |||||
| Yes | N/A | 1302 | 805 | 61.8 | 2.5 | 2.4–2.6 | ||||
| Gonorrhea | ||||||||||
| No | 167,707 | 10,621 | 6.3 | Ref. | N/A | |||||
| Yes | 3505 | 894 | 25.6 | 4.0 | 3.8–4.3 | N/A | ||||
| STI-related symptoms‡ | ||||||||||
| No | 114,827 | 7678 | 6.7 | Ref. | 20,296 | 5307 | 26.2 | Ref. | ||
| Yes | 56,385 | 3837 | 6.8 | 1.0 | 1.0–1.1 | 10,532 | 2749 | 26.1 | 1.0 | 1.0–1.0 |
| PID§ | ||||||||||
| No | 121,350 | 6845 | 5.6 | Ref. | 11,871 | 3574 | 30.1 | Ref. | ||
| Yes | 1365 | 403 | 29.5 | 5.2 | 4.8–5.7 | 311 | 153 | 49.2 | 1.6 | 1.5–1.8 |
| GC treatment†,¶ | ||||||||||
| No | 82,049 | 7145 | 8.7 | Ref. | 8373 | 2318 | 27.7 | Ref. | ||
| Yes | 13,664 | 1923 | 14.1 | 1.6 | 1.5–1.7 | 10,946 | 3288 | 30.0 | 1.1 | 1.0–1.1 |
*Limited to positive laboratory test results with ≥1 medical claim(s) within a 21-day period of interest, ranging from 7 days before to 14 days after the laboratory test result date.
†Corresponding chlamydia/gonorrhea diagnosis codes, gonorrhea treatment procedure codes, and concurrent diagnosis codes were all identified within the 21-day period described above.
‡Sex-specific STI-related symptoms include dysuria, urethritis, vaginitis, and vulvovaginitis.
§Only identified among female enrollees.
¶Healthcare Common Procedure Coding System code J0696 (intramuscular injection of ceftriaxone, 250 mg) was identified within the 21-day period described above for all chlamydia and gonorrhea test results occurring after December 31, 2011.
95% CI, 95% confidence interval; GC, gonorrhea; n, number; N/A, not applicable; PID, pelvic inflammatory disease; PR, prevalence ratio; Ref., reference group; STI, sexually transmitted infection.
Compared with females, the prevalence of a corresponding GC diagnosis code among confirmed gonococcal infections was lower among males (PR, 0.8; 95% CI, 0.7–0.8; Table 3). Among confirmed gonococcal infections, the prevalence of a corresponding GC diagnosis code was associated with concurrent CT (PR, 2.5; 95% CI, 2.4–2.6) and PID (PR, 1.6; 95% CI, 1.5–1.8) diagnoses.
DISCUSSION
Currently, because of the limited availability of laboratory results in administrative claims data, researchers using these data rely solely on ICD-CM diagnosis codes to identify chlamydial and gonococcal infections. Using an administrative claims data set containing medical claims and laboratory test results, we evaluated the validity of using ICD-CM diagnosis codes to identify laboratory-confirmed chlamydial and gonococcal infections. Using laboratory test results as the reference standard, we found very high specificities for both CT (99.5%) and GC (99.7%) ICD-CM diagnosis codes, indicating that CT/GC diagnosis codes identified in administrative claims data likely represent laboratory-confirmed infections. However, we also observed extremely low sensitivities for both CT (6.5%) and GC (25.4%) ICD-CM diagnosis codes, indicating that ICD-CM diagnosis codes greatly underestimate the burden of these infections. Therefore, whenever available, we recommend researchers use laboratory data instead of ICD-CM diagnosis codes to identify chlamydial and gonococcal infections.
Given that administrative claims data are intended for billing and reimbursement purposes, rather than surveillance and research, the observed low CT/GC ICD-CM diagnosis code sensitivity is not surprising. Administrative claims likely reflect in-person health care encounters where a diagnosis was known at the time of service and when reimbursable services, such as pharmaceutical treatments, were provided. Although the current treatment of CT is an oral regimen that does not require in-office administration, the current treatment of GC includes an intramuscular injection that potentially requires a follow-up office visit that would generate subsequent medical claims.24 Thus, it is not surprising that the ICD-CM diagnosis code sensitivity was higher for GC than CT. We found that confirmed chlamydial infections were more likely to have a corresponding ICD-CM diagnosis code when GC treatment was provided (PR, 1.6). A CT diagnosis alone likely does not generate additional health care encounters or claims, which may explain the low usage of CT ICD-CM diagnosis codes by health care providers.
Because we detected differences among enrollees with laboratory-confirmed infections who are assigned a CT/GC diagnosis code (i.e., these codes are not missing at random) and most confirmed infections did not have a corresponding diagnosis code, using only ICD-CM diagnosis codes to establish outcome-specific cohorts will introduce misclassification bias. In this study, relying solely on ICD-CM diagnosis codes would have misclassified 94% and 75% of enrollees with laboratory-confirmed chlamydial and gonococcal infections, respectively, as not having an infection. Corresponding CT/GC diagnosis codes among laboratory-confirmed infections were more prevalent among enrollees with concurrent CT and GC diagnoses and among women with a concurrent PID diagnosis. Future research should continue to investigate how to improve the validity of using administrative claims data without associated laboratory result data to examine disease prevalence and establish outcome-specific cohorts.
Our work highlights the challenges of using administrative claims data to examine conditions requiring a laboratory diagnosis. This analysis expands on the previous work by Ho et al.21 that examined the validation of ICD-10-CM diagnosis codes for identifying chlamydial and gonococcal infections. Although not directly comparable because of different methodologies, their analysis also reported low sensitivities (CT: 10.6; GC: 9.7) and high specificities (CT: 99.9; GC: 99.9) for CT/GC diagnosis codes and emphasized the challenges of using administrative claims in the absence of laboratory test results.21 Compared with our findings, Ho et al.21 reported higher PPVs for CT/GC diagnosis codes (CT: 87.6; GC: 85.0), which they attributed to presumptive diagnoses based on the comparison of diagnosis and laboratory test result dates. Furthermore, our work demonstrates the potential for differences between ICD-9-CM and ICD-10-CM diagnosis codes, particularly for chlamydial infections (Supplemental Tables S1, S2, http://links.lww.com/OLQ/A669).
This analysis had several limitations. The OLDW does not represent the general US population or the commercially insured population, which limits the generalizability of our findings. In addition, laboratory test results were only included in the OLDW when associated with contracted vendors and may not be a representative sample of all enrollees. However, we compared eligible enrollees with and without available laboratory data by sex, age, and census region of residence and did not identify any meaningful differences. Other inherent limitations to administrative claims data include missing, incomplete, or inaccurate data. Because we did not require continuous enrollment or have data for encounters that occurred outside of the OLDW, it is possible that some enrollees with available test results received follow-up care outside of the OLDW. As a result, we may not have identified all CT/GC-related medical claims associated with laboratory test results among our study population. Because of these limitations, it is possible that some medical diagnoses and procedures informing CT/GC diagnoses were not available for analysis and may have contributed to our observed findings.
By using 15 years of data in a large, geographically diverse administrative claims data set, we expand on previous research to evaluate the ICD-CM diagnosis codes currently in use to identify chlamydial and gonococcal infections. This is the first study to investigate the validity of using ICD-9-CM and ICD-10-CM diagnosis codes to identify laboratory-confirmed chlamydial and gonococcal infections. The approach used in this analysis may help improve CT/GC research efforts (e.g., help establish outcome-specific cohorts).
Researchers should be cautious when relying on ICD-CM diagnosis codes to identify chlamydial and gonococcal infections in administrative claims data. Our findings indicate that using only ICD-CM diagnosis codes to identify chlamydial and gonococcal infections substantially underestimates the burden of these diseases and inaccurately classifies laboratory-confirmed infections. Whenever possible, we recommend researchers use laboratory data in place of diagnosis codes to identify chlamydial and gonococcal infections. As data systems continue to expand and improve, future research should focus on novel approaches to identifying chlamydial and gonococcal infections in administrative claims data to further enhance public health research and disease surveillance efforts.
Supplementary Material
Footnotes
Conflict of Interest and Sources of Funding: The authors have no conflicts of interest to disclose. The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (http://www.stdjournal.com).
Contributor Information
Elizabeth A. Torrone, Email: igf0@cdc.gov.
Elaine W. Flagg, Email: elaine.flagg@comcast.net.
REFERENCES
- 1.Cadarette SM, Wong L. An introduction to health care administrative data. Can J Hosp Pharm 2015; 68:232–237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dixon BE Rahurkar S Ho Y, et al. Reliability of administrative data to identify sexually transmitted infections for population health: A systematic review. BMJ Health Care Inform 2019; 26:e100074. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Nelson CA Saha S Kugeler KJ, et al. Incidence of clinician-diagnosed Lyme disease, United States, 2005–2010. Emerg Infect Dis 2015; 21:1625–1631. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Kong CL Thompson RR Porco TC, et al. Incidence rate of herpes zoster ophthalmicus: A retrospective cohort study from 1994 through 2018. Ophthalmology 2019; 127:324–330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tao G, Patel C, Hoover KW. Updated estimates of ectopic pregnancy among commercially and Medicaid-insured women in the United States, 2002–2013. South Med J 2017; 110:18–24. [DOI] [PubMed] [Google Scholar]
- 6.Berlinberg EJ Kim E Deiner MS, et al. Seasonality of herpes zoster and herpes zoster ophthalmicus. J Clin Virol 2020; 126:105406. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Abraham NS Noseworthy PA Inselman J, et al. Risk of gestational bleeding increases with combinations of antithrombotic agents and patient age. Clin Exp Gastroenterol Hepatol 2020; 18:337–346. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Radner H Lesperance T Accortt NA, et al. Incidence and prevalence of cardiovascular risk factors among patients with rheumatoid arthritis, psoriasis, or psoriatic arthritis. Arthritis Care Res 2017; 69:1510–1518. [DOI] [PubMed] [Google Scholar]
- 9.Falowski SM Provenzano DA Xia Y, et al. Spinal cord stimulation infection rate and risk factors: Results from a United States payer database. Neuromodulation 2019; 22:179–189. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Tao G, Irwin KL. Receipt of HIV and STD testing services during routine general medical or gynecological examinations: Variations by patient sexual risk behaviors. Sex Transm Dis 2008; 35:167–171. [DOI] [PubMed] [Google Scholar]
- 11.Song LD Newhouse JP Garcia-De-Albeniz X, et al. Changes in screening colonoscopy following Medicare reimbursement and cost-sharing changes. Health Serv Res 2019; 54:839–850. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ekperi LI Thomas E LeBlanc TT, et al. The impact of Hurricane Sandy on HIV testing rates: An interrupted time series analysis, January 1, 2011–December 31, 2013. PLoS Curr 2018; 10:ecurrents.dis.ea09f9573dc292951b7eb0cf9f395003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dawwas GK Brown J Dietrich E, et al. Effectiveness and safety of apixaban versus rivaroxaban for prevention of recurrent venous thromboembolism and adverse bleeding events in patients with venous thromboembolism: A retrospective population-based cohort analysis. Lancet Haematol 2019; 6:e20–e28. [DOI] [PubMed] [Google Scholar]
- 14.Palmsten K Chambers CD Wells A, et al. Patterns of prenatal antidepressant exposure and risk of preeclampsia and postpartum haemorrhage. Paediatr Perinat Epidemiol 2020; 34:597–606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Owusu-Edusei K Jr. Roby TM Chesson HW, et al. Productivity costs of nonviral sexually transmissible infections among patients who miss work to seek medical care: Evidence from claims data. Sex Health 2013; 10:434–437. [DOI] [PubMed] [Google Scholar]
- 16.Tao G Hoover KW Nye MB, et al. Infrequent testing of women for rectal chlamydia and gonorrhea in the United States. Clin Infect Dis 2018; 66:570–575. [DOI] [PubMed] [Google Scholar]
- 17.Heijne JC Herzog SA Althaus CL, et al. Insights into the timing of repeated testing after treatment for Chlamydia trachomatis: Data and modelling study. Sex Transm Infect 2013; 89:57–62. [DOI] [PubMed] [Google Scholar]
- 18.Bautista CT Wurapa EK Sateren WB, et al. Longitudinal association of gonorrhea and bacterial vaginosis with repeat chlamydia diagnoses among U.S. Army women: A retrospective cohort analysis. Mil Med Res 2018; 5:37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Christiansen-Lindquist L Tao G Hoover K, et al. Chlamydia screening of young sexually active, Medicaid-insured women by race and ethnicity, 2002–2005. Sex Transm Dis 2009; 36:642–646. [DOI] [PubMed] [Google Scholar]
- 20.Loosier PS Hsieh H Cramer R, et al. Young adults' access to insurance through parents: Relationship to receipt of reproductive health services and chlamydia testing, 2007–2014. J Adolesc Health 2018; 63:575–581. [DOI] [PubMed] [Google Scholar]
- 21.Ho YA Rahurkar S Tao G, et al. Validation of International Classification of Diseases, Tenth Revision, Clinical Modification codes for identifying cases of chlamydia and gonorrhea. Sex Transm Dis 2021; 48:335–340. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.OptumLabs . OptumLabs and OptumLabs Data Warehouse (OLDW) Descriptions and Citation. Eden Prairie, MN: n.p., July 2020. PDF. Reproduced with permission from OptumLabs. [Google Scholar]
- 23.Centers for Disease Control and Prevention (CDC) . Update to CDC's sexually transmitted diseases treatment guidelines, 2010: Oral cephalosporins no longer a recommended treatment for gonococcal infections. MMWR Morb Mortal Wkly Rep 2012; 61:590–594. [PubMed] [Google Scholar]
- 24.Workowski KA Bolan GA, Centers for Disease Control and Prevention . Sexually transmitted diseases treatment guidelines, 2015. MMWR Recomm Rep 2015; 64(RR–3):1–137. [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.