Abstract
Objective. We examined hospitalizations for patients with known rheumatoid arthritis (RA) or giant cell arteritis (GCA) to evaluate whether hospitalization-related diagnoses accurately identified patients with rheumatologic diseases.
Methods. Diagnosis codes for hospitalizations in 1996–2012 among previously identified population-based cohorts of patients with RA or GCA were examined for RA or GCA mentions.
Results. RA or GCA mention occurred in only 55% of 2407 hospitalizations among patients with RA and 31% of 502 hospitalizations among patients with GCA. RA or GCA was mentioned more often in recent years, younger patients, and rheumatic medication users.
Conclusion. Coding for RA or GCA during hospitalizations was often missed. Research using hospital diagnoses alone could be biased.
- RHEUMATOID ARTHRITIS
- GIANT CELL ARTERITIS
- HOSPITALIZATION
- DIAGNOSIS
- INTERNATIONAL CLASSIFICATION OF DISEASE
- CODING
Electronic data identification is commonplace, but assessing the limitations of data obtained for 1 purpose (e.g., diagnosis coding) when using it for another purpose (e.g., research) can be challenging. Several studies have addressed the positive predictive value of the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM) codes and algorithms for the identification of patients with rheumatic diseases in electronic databases1,2,3. These studies showed high positive predictive value for patients with multiple codings of the same rheumatic disease over a period of time. However, the National Inpatient Sample, a free and widely accessible database from the Healthcare Cost and Utilization Project (HCUP), is limited to information regarding hospitalizations without the ability to link those hospitalizations to other healthcare for the same patient, so multiple codings are not available4. Another issue not well addressed by these studies is whether there are hospitalized patients with rheumatologic diseases who are not identified because of an absence of a rheumatologic code from an inpatient stay. We examined primary and secondary diagnosis codes from hospitalizations for patients with known rheumatoid arthritis (RA) or giant cell arteritis (GCA) to evaluate whether hospitalization-related diagnoses accurately identify patients with rheumatologic diseases. We also examined whether certain reasons for hospitalization were more or less likely to be associated with the mention of RA or GCA.
MATERIALS AND METHODS
Our retrospective, population-based cohort study used the resources of the Rochester Epidemiology Project, a medical record linkage system containing the complete inpatient and outpatient medical records from all healthcare providers in Olmsted County, Minnesota, USA, including the Mayo Clinic and its affiliated hospitals, Olmsted Medical Center, local nursing homes, and private practitioners5,6.
Two previously identified cohorts of Olmsted County residents were analyzed. The RA cohort consisted of patients aged ≥ 18 years who first fulfilled the 1987 American College of Rheumatology (ACR) classification criteria for RA in 1980–20077,8. The GCA cohort consisted of patients who fulfilled the 1990 ACR criteria for GCA in 1950–20099,10.
Hospitalization data were retrieved electronically from all Olmsted County medical providers. Followup began with the latter of RA or GCA diagnosis date or January 1, 1996, and ended at death, emigration from Olmsted County, or December 31, 2012. All primary and secondary diagnoses from admission to discharge were analyzed. RA mention was defined as ICD-9-CM 714.xx with expansion to include codes for closely related diseases (i.e, ICD-9-CM codes: 696.0, 715.9X, 716.5–716.59, 716.8X, 716.9X, 719.3X, 720.0, 720.2, 721.90). GCA mention was defined as ICD-9-CM 446.5 with expansion to include 446.7 and 447.6.
To examine whether certain reasons for hospitalization were associated with the mention of RA or GCA, the primary discharge diagnoses were grouped together using the Clinical Classifications Software (CCS) for the ICD-9-CM from the HCUP11. The CCS classifies diagnoses into 18 chapters: infections and parasitic diseases; neoplasms; endocrine, nutritional, and metabolic diseases and immunity disorders; diseases of the blood and blood-forming organs; mental illness; diseases of the nervous system and sense organs; diseases of the circulatory system; diseases of respiratory system; diseases of the digestive system; diseases of the genitourinary system; complications of pregnancy, childbirth, and puerperium; diseases of the skin and subcutaneous tissue; diseases of the musculoskeletal system and connective tissue; congenital anomalies; certain conditions originating in the perinatal period; injury and poisonings (which includes fractures); symptoms, signs, and ill-defined conditions; and residual codes, unclassified.
Information on RA disease severity was collected previously by review of each patient’s full medical record12. Characteristics included rheumatoid factor (RF) positivity and anticitrullinated protein antibody (ACPA) positivity at the time of RA diagnosis, as well as repeatedly high erythrocyte sedimentation rates (ESR; i.e., ≥ 3 ESR measures ≥ 50 mm/h with at least 30 days between 2 measurements), the presence of joint erosions/destructive changes, or rheumatoid nodules during the first year following diagnosis of RA. Information on the use of corticosteroids for both patients with RA and GCA and the use of disease-modifying antirheumatic medications and biologic response modifiers for patients with RA during each hospitalization was also collected.
Statistical methods
Descriptive statistics (means, percentages, etc.) were used to summarize the data. Within each cohort, comparisons of hospitalizations with and without mention of RA or GCA were performed using generalized linear models with random intercepts for each subject to account for multiple hospitalizations in the same subject. Analyses were performed using SAS version 9.4 (SAS Institute).
RESULTS
The RA cohort contained 499 patients (68.1% women) who had 2407 hospitalizations with a mean age at hospital admission of 68.8 years. RA mention by the ICD-9-CM 714.x in discharge diagnoses occurred in 1334 hospitalizations (55.4%). Using the expanded list of related codes, an additional 68 hospitalizations were identified for a total of 1402 hospitalizations with the mention of RA (58.2%).
Hospitalization with RA mention occurred more often in younger patients (mean 67.2 yrs vs 70.9 yrs, p < 0.001; Table 1), in patients with longstanding RA (mean 11.2 yrs vs 10.0 yrs, p < 0.001), and in more recent hospitalizations (p < 0.001).
The mention of RA differed significantly in primary discharge diagnoses for 3 CCS chapters. RA was more likely to be mentioned when the primary discharge diagnoses of the hospitalization concerned diseases of the musculoskeletal system and connective tissue diseases compared with all other primary discharge diagnosis categories (78% vs 55%, p < 0.001). RA was less likely to be mentioned when the primary discharge diagnosis of the hospitalization concerned neoplasms (49% vs 59%, p = 0.034) or symptoms, signs, and ill-defined conditions (47% vs 59%, p = 0.020).
RA was more likely to be mentioned if patients were RF- or ACPA-positive (Table 2). Other RA disease severity measures occurring in the first year of RA diagnosis, such as erosions, repeatedly elevated ESR levels, and rheumatoid nodules, were not associated with an increased likelihood of RA mention. Biologic use and corticosteroid use were both associated with higher likelihood of RA mention.
The GCA cohort consisted of 119 patients (79.0% women) who had 502 hospitalizations. GCA was mentioned in 155 hospitalizations (30.9%). Two additional hospitalizations (0.4%) were coded as unspecified arteritis (ICD-9-CM 447.6) and no hospitalizations were coded as Takayasu arteritis (ICD-9-CM 446.7). GCA was more likely to be mentioned in slightly younger patients (mean 82.7 yrs vs 84.8 yrs, p = 0.002; Table 3), those with shorter disease duration (mean 4.9 yrs vs 10.4 yrs, p < 0.001), and those taking corticosteroids (60.5% vs 13.8%, p < 0.001).
While no CCS chapter categorization of primary discharge diagnoses yielded statistically significant results, data from 2 chapters stood out. GCA was somewhat more likely to be mentioned when the primary discharge diagnosis concerning diseases of the circulatory system was compared (37% vs 29%, p = 0.07), and GCA was less likely to be mentioned when the primary discharge diagnosis concerned the diseases of the genitourinary system (14% vs 32%, p = 0.06).
DISCUSSION
Our study evaluated how often hospital diagnoses included the mention of 2 systemic rheumatologic diseases using cohorts of patients with known and well-characterized RA and GCA. Coding for RA in hospitalized patients with RA occurred only half of the time. Coding of RA during hospitalizations of patients with RA occurred more frequently in recent years, younger patients, and those taking rheumatic medications. In hospitalized patients with GCA, coding for GCA occurred only 30% of the time and generally in younger patients and those using corticosteroids.
These results raise the concern that a large number of patients with diagnosed rheumatic diseases may not be included in studies in which only hospital discharge diagnoses are used for the identification of patients with rheumatic diseases. This failure of identification could result in studies flawed by biases of selection and detection.
While use of hospital codes for identification of cases is a convenient and efficient strategy for epidemiologic and related healthcare use studies, this strategy has inherent potential shortcomings. The data are dependent on accurate diagnoses by physicians and complete coding by the hospital. The presence or absence of coding is influenced by a number of factors, including diagnoses or procedures considered primary for the hospitalization, and those which may increase reimbursement for services.
A study of accuracy of hospital diagnoses for RA in the Danish National Patient Register found that only 59% of RA mentions in the National Patient Register were confirmed in the medical records13. The investigators were not able to evaluate how many patients with RA were not coded in their register.
Strengths of our study include the use of well-characterized cohorts of patients who are known to have RA or GCA. It includes every hospitalization the patients underwent in Olmsted County over a 16-year period. To our knowledge, there are no other studies that have used cohorts of patients with known rheumatic diseases to evaluate the capacity of hospital discharge diagnoses to identify these patients.
Study limitations include lack of data for hospitalizations occurring outside Olmsted County and the uncertainty regarding the generalizability of the results from 1 county to the entirety of the United States; however, it is unlikely that hospital diagnoses differ significantly from region to region. Generalizability of these results outside the United States is difficult or impossible because of the differences in healthcare systems in other countries that may have different incentives for coding practices.
It is incumbent on researchers to be judicious in their use of hospital diagnoses and ICD-9-CM codes for the identification of patients with rheumatic diseases. It is equally important for reviewers and readers to be shrewd in their judgment of epidemiological studies based solely on these methods.
Footnotes
Supported by a grant from the National Institutes of Health, NIAMS (R01 AR46849), and made possible by the Rochester Epidemiology Project (R01 AG034676 from the National Institute on Aging).
- Accepted for publication June 5, 2015.
REFERENCES
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.