To identify potential cases of atypical cat-scratch disease, we conducted a retrospective analysis of persons enrolled in the Truven Health MarketScan Commercial Claims and Encounters Database (Truven Health Analytics, https://www.ibm.com) during 2005–2014. The MarketScan Commercial Claims and Encounters Database includes persons <65 years of age covered by select employer-sponsored health insurance plans in all 50 states and contains administrative claims data on outpatient visits, inpatient admissions, and emergency department visits. Demographically, the MarketScan population generally mirrors the US population, with a slight overrepresentation of persons 50–59 years of age and a slight underrepresentation of persons 20–29 years of age (28).

Billing codes from outpatient, inpatient, and emergency department visits are assigned by either a clinician or billing specialist according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), and procedures are captured as either ICD-9-CM codes, Current Procedure Terminology codes, or Healthcare Common Procedure Coding System codes. Because the International Classification of Diseases, 10th Revision, Clinical Modification, was not officially adopted in the United States until 2015, those codes were not included.

We identified cat-scratch disease cases by extracting all enrollee visit records during the study period with an ICD-9-CM code for cat-scratch disease (078.3). The first instance of a 078.3 diagnosis code in a patient record was considered the index event. Atypical manifestations of interest were selected for analysis if they had recorded precedent in the literature as a complication of cat-scratch disease and distinct, clearly discernable ICD-9-CM codes associated with the specific manifestation. Based on these criteria, the known complications of cat-scratch disease included for analysis were endocarditis, osteomyelitis, erythema nodosum, conjunctivitis, retinitis/neuroretinitis, encephalitis, neuritis, and hepatosplenic disease. We included ICD-9-CM codes associated with optic neuritis in the retinitis/neuroretinitis category. Encounters with an ICD-9-CM code for cat-scratch disease and an accompanying diagnostic code to indicate the anatomic location of a wound or inoculation site for B. henselae were also flagged for analysis and were categorized as either head or neck region, arm or shoulder region, leg or hip region, or torso region. We compiled a detailed list of all ICD-9-CM codes used to identify atypical manifestations of cat-scratch disease (Appendix Table).

We extracted insurance billing records of enrollees with ICD-9-CM codes for cat-scratch disease and selected manifestations at either the same encounter or within a 30-day window of one another. These records were evaluated along with previous and subsequent records by 2 independent reviewers (R.J.M. and C.A.N.) to ensure that the clinical picture was consistent with the coded atypical manifestation based on diagnosis codes, procedure codes, and provider types. If plausible alternative causes of the selected manifestation or likely coding errors were identified, we did not include the enrollee record as an atypical case. In cases of discordance, a third reviewer (Paul Mead) determined final categorization based on record review.

We included persons with an ICD-9-CM code for cat-scratch disease but without accompanying atypical manifestation as typical cases for comparison. Residence in a rural area was assigned if an enrollee did not reside in a metropolitan statistical area, as designated by the US Office of Management and Budget. Because previous research has identified increases in cat-scratch disease in late summer, fall, and January (13,14,29), we categorized month of onset as either late summer and fall, January, or all other months for analysis.

We performed descriptive and comparative statistical analyses by using JMP version 13.2.1 (https://www.jmp.com) and SAS version 9.3 (https://www.sas.com). We used Pearson χ² tests or Fisher exact tests for comparisons of categorical variables. To compare the conditional probability of having atypical cat-scratch disease across strata of potential variables of interest (e.g., sex, age category), we calculated risk ratios (RRs) and associated 95% CIs. Human subjects review at the Centers for Disease Control and Prevention determined that institutional review board approval was not required for this study.

Using US nationwide insurance claims data, we identified and characterized 224 atypical cases of cat-scratch disease during 2005–2014 and estimated an average annual incidence of 0.7 cases/100,000 population. Nearly half of all atypical cat-scratch disease cases had ocular manifestations, most of which were retinitis/neuroretinitis. Atypical cat-scratch disease was most prevalent among female patients 10–14 years of age, who most commonly had ocular manifestations.

Trends in hospitalizations of patients with cat-scratch disease highlight the severity of atypical cat-scratch disease compared with typical cat-scratch disease. Atypical cat-scratch disease appears to be particularly severe among children <14 years of age, who had an increased risk for hospitalization. Adults 50–64 years of age had the lowest risk for development of atypical cat-scratch disease and specifically ocular manifestations. Reasons that older adults might have complications associated with cat-scratch disease less often than other age groups are unclear and require further study.

Severity of cat-scratch disease in children has been previously documented. In a study conducted by Reynolds et al., ≈25% of hospitalizations of children for cat-scratch disease were caused by complications associated with atypical cat-scratch disease; neurologic and hepatosplenic complications were most common (30). Although children in our study were also particularly at risk for hepatosplenic disease, neurologic and hepatosplenic complications were associated with increased risk for hospitalization in our overall study population, indicating that these manifestations are particularly severe for all age groups. Encephalitis was the most common neurologic manifestation in our population, which was also consistent with findings of Reynolds et al., in which most hospitalizations of children for neurologic complications of cat-scratch disease were caused by encephalitis or encephalopathy (30). Thus, physicians should consider cat-scratch disease in patients who have encephalitis or new onset hepatosplenic abnormalities, especially children.

Previous studies have documented the highest rates of cat-scratch disease in late summer and fall and a separate peak often seen in January (13,14,29). One such study found that rates of B. henselae seropositivity among samples submitted to Mayo Clinic Laboratories over a 10-year period were highest during September–January, with the highest annual rates in January (29). Typical cat-scratch disease diagnoses in our study followed similar seasonal patterns to those previously reported. However, atypical cat-scratch disease appeared more concentrated during August–October and January–March. The reasons for this finding are unclear but might include delays in diagnosis of atypical cat-scratch disease. For example, patients who contract cat-scratch disease and had complications during January might not be given a diagnosis of atypical cat-scratch disease at that time because they do not show classic symptoms or their symptoms take time to develop and care-seeking is delayed.

Furthermore, a recent case series of ocular manifestations of cat-scratch disease reported that 9 of 10 patients had symptoms <3 months before showing development of ocular complications and that 3 patients had been originally given misdiagnoses of etiologies other than cat-scratch disease (31). Given that ocular manifestations of cat-scratch disease were most common in our study, increased diagnoses of atypical cat-scratch disease through March could be a sign of delayed diagnoses, particularly for manifestations that are less severe, such as those involving the eye.

Trends related to geographic distribution of cases did not differ between atypical and typical cat-scratch disease. Similar to findings from 3 previous studies that reported the highest incidences of cat-scratch disease in the southern United States (13,14,30), in our study, most typical (53.0%) and atypical (57.6%) cat-scratch disease cases occurred in patients residing in this region. In addition, a national survey of US healthcare providers found that those in the Pacific and southern regions of the United States were more likely to have been given a diagnosis of cat-scratch disease than in other regions (32). These findings are further supported by studies that have found higher average B. henselae seroprevalences and active bacteremia in pet cats from warmer, more humid climates, including the southern United States (33,34). Thus, healthcare providers in regions with climates that support flea abundance should be aware of the risk for cat-scratch disease and be able to recognize its atypical manifestations.

This study had several limitations. First, although MarketScan is a large database of insurance claims data from persons covered by employer-sponsored insurance, it is a convenience sample and may not accurately represent the characteristics of all persons in the United States. For example, trends we see in atypical cat-scratch disease by geographic region and rural residence might be biased by differences in coverage and access to care that are not accounted for here. Furthermore, MarketScan does not include data for adults >65 years of age, military personnel, uninsured persons, or Medicaid/Medicare enrollees. These specific populations might show varying degrees of cat-scratch disease severity or risk that are not captured in our results. In addition, because only persons <65 years of age are included in the database, the proportion of children who have cat-scratch disease might be artificially inflated. The number of patients who had atypical cat-scratch disease was small, especially when broken down by manifestation. Thus, it is difficult to draw conclusions regarding risk factors for specific manifestations of atypical cat-scratch disease and hospitalization within these groups.

Furthermore, misclassification could have occurred when ICD-9-CM codes were used to classify atypical cat-scratch disease for several reasons. ICD-9-CM codes are subject to error from the clinicians and billing specialists who enter them. In addition, we excluded records that fit our criteria for manifestations of atypical cat-scratch disease but lacked additional supporting information, which could have caused us to underestimate the true burden of atypical cat-scratch disease. Last, codes for some known atypical cat-scratch disease manifestations, such as pulmonary complications and thrombocytopenia, were excluded because of etiologic ambiguity in enrollee records.

In conclusion, our findings indicate that atypical cat-scratch disease in the United States follows trends similar to those for typical cat-scratch disease but is more prevalent and severe among children <14 years of age and is least likely to occur in older adults (50–64 years of age). In addition, differences in seasonality of diagnoses were seen, which might be an indication that diagnosis of atypical cat-scratch disease is often delayed. Ocular (retinitis/neuroretinitis and conjunctivitis) and hepatosplenic complications were the most common manifestations of atypical cat-scratch disease. Improved understanding of atypical cat-scratch disease might lead to better recognition of cases by clinicians, as well as inform efforts to clarify the pathophysiology of this disease.