Diplopia, commonly referred to as double vision, can be a disabling symptom, often leading to emergency department (ED) visits. Binocular diplopia can be a sign of neurological disease such as stroke, multiple sclerosis or brain tumours, necessitating urgent assessment and investigations. ^{Reference Nazerian, Vanni and Tarocchi1} A few single-centre studies have reported that among patients with diplopia seen in the ED, 16–33% could have a neurological disease; however, long-term outcomes of patients with diplopia in the ED are scarce due to the inability to identify these patients in population-based databases. ^{Reference Nazerian, Vanni and Tarocchi1,Reference Occelli, Coffin and Raynaud2} There is an International Classification of Diseases 10^th revision (ICD-10) code for diplopia, H53.2; however, the validity of this code for identifying people with true diplopia is not known. We evaluated the validity of the ICD-10 code (H53.2) for diplopia using a gold standard of the diagnosis of binocular diplopia based on chart abstraction by a neurology resident physician.

We performed this single-centre, retrospective cohort study at St Michael’s Hospital in Toronto, Canada. St Michael’s Hospital, a level 1 trauma centre, is an academic hospital in downtown Toronto that has ∼60,000 ED visits per year, of which 20% are from people experiencing homelessness. The hospital has ophthalmology, neurology and neurosurgery services. The study was approved by the St Michael’s Research Ethics Board (REB25-080). Patient consent was waived due to the limited use of the personal health information.

In Ontario, the National Ambulatory Care Reporting System (NACRS) records the main reason for visit for all patients seen in the ED, along with nine other fields that may include additional ED diagnoses (https://www.cihi.ca/en/national-ambulatory-care-reporting-system-nacrs-metadata). The main reason or “main problem” field is the most clinically significant diagnosis, condition, problem or circumstance for the client’s visit, as determined by the managing emergency physician. Based on what is listed on the chart by emergency physicians, certified health information professionals, with training in reviewing clinical records and using standardized methods for billing and recordkeeping of routine health data, will assign the ICD-10 code to this field.

We assessed the validity of the ICD-10 code for diplopia in a group of adult (age ≥ 18 years) patients who were seen in the ED at St Michael’s Hospital between 1 December 2023 and 30 November 2024. The study period was specifically chosen to avoid a change of electronic health record systems at the institution that occurred on 1 December 2024. Our cohort included any patients with ICD-10 H53.2 in the main field in NACRS and any patients who were seen in the ED who had a follow-up appointment with the hospital’s ambulatory ophthalmology clinic within 7 days of their ED visit.

Using a prespecified data collection form (Table 1), an undergraduate student (AT) abstracted the following information from the charts pertaining to the index ED visit chart, as well as any neurology or ophthalmology consults in the ED: age in years; sex; after-hours ED visit (defined as visits with registration time outside of regular business hours, Monday to Friday, 8–5 PM); ED arrival by ambulance (yes or no); history of diplopia, blurry vision or reduced vision; presence of eye misalignment on exam, which eye was misaligned; and associated eye symptoms such as ptosis. In case of a discrepancy in the information between ED physician records and those by the specialists, the information in the specialist’s note was abstracted. They were blinded to the diagnostic codes in NACRS.

Table 1.

Characteristics of the included sample of patients

Table 1. Long description

The table compares characteristics of patients with and without diplopia, including age, length of stay, sex, after-hours visits, ambulance arrival, history of vision issues, eye misalignment, and associated symptoms. It has 17 rows and 5 columns. Column headers are Variable, Diplopia as per the gold standard, No diplopia, and Standardized difference. Row labels include Mean age, Median length of stay, Female, Seen after hours, Arrived by ambulance, History of blurry vision, History of reduced vision, History of binocular diplopia, Presence of eye misalignment, Eyes that were misaligned, Ptosis, and Coded as diplopia in NACRS. Each row provides specific data points for patients with and without diplopia, along with the standardized difference between the two groups.

SD = standard deviation; NACRS = National Ambulatory Care Reporting System.

A neurology resident physician (AR) reviewed abstracted chart data to identify patients with diplopia. For this study, a patient was defined to have diplopia if a) diplopia was documented to be binocular (i.e., cases of monocular diplopia are not considered diplopia for this study) and b) had misalignment of one or more eyes on exam or reported diplopia while testing eye movements, that is, seeing two objects. In 30 patients, where the diagnosis was uncertain, a staff neurologist (MVV) determined the presence or absence of diplopia. The neurologist assessors were blinded to the diagnostic codes in NACRS.

The primary outcome was the sensitivity (true positive rate), specificity (true negative rate) and positive predictive value for the ICD-10 (H53.2) code using our gold standard. We also reported these metrics in the following subgroups: age (above or below 65 years), sex (female vs. male), after-hours visits (yes vs. no) and arrival by ambulance (yes vs. no). All analyses were conducted in SAS 9.4.

The study sample included 783 patients with either an ICD-10 code for diplopia in the ED (n = 27, 3.4%), an ophthalmology clinic visit within 7 days after an ED visit (n = 732, 93.5%) or both (n = 24, 3.1%) (e-Figure 1). The mean age of the cohort was 53.4 years (SD 17.7), and 372 (47.5%) were female (Table 1). The median length of ED stay was 188 minutes (Q1–Q3 112–307). Blurry vision was the most common complaint (n = 437, 55.8%), while double vision on history was reported only in 76 (9.7%) patients. On exam, misalignment of one or more eyes was only recorded in 15 (1.9%) patients (Table 1).

The ICD-10 code for diplopia was noted in 51 (6.5%) of all patients. 79 (10.1%) patients were ascertained to have diplopia by our gold-standard definition (Table 2). Using this definition, the specificity of the ICD-10 code H53.2 was 96.9% (95% confidence interval, 95.6–98.2), sensitivity was 36.7% (26.1–47.3) and positive predictive value was 93.2% (91.3–95.0) (e-Table 1). In sensitivity analyses, removing records of the 30 patients with inconclusive diagnosis by the trainee, the sensitivity (27.1%; 17.1–36.3) dropped, but not at a high cost of specificity (96.3%; 94.2–97.8%). These validity metrics were largely similar across the prespecified subgroups, although in female patients and those aged 64 years or older, both specificity and sensitivity improved compared with the overall cohort (e-Table 1).

Table 2.

Validity of the ICD-10 code for diplopia in the emergency department of a quaternary hospital in Ontario, Canada

ICD-10 = International Classification of Diseases 10^th revision; ED = emergency department; TP = true positive; FP = false positive; FN = false negative; TN = true negative.

In a single-centre retrospective observational study, we show that the ICD-10 code for diplopia in the ED has high specificity but low sensitivity. The findings support its use to identify true cases of diplopia in the ED with good confidence, but not to estimate the true burden of diplopia in the ED.

We are not aware of previous studies that have assessed the validity of the ICD-10 code for diplopia. In the USA, one study used the national Ambulatory Medical Care Survey between 2016 and 2017 and found 49,790 visits to the ED for diplopia-related concerns. They used the ICD-9 code (368.2) to identify the cohort. ^{Reference De Lott, Kerber, Lee, Brown and Burke3} However, the study did not report on the validity of the code in identifying patients with diplopia in the ED.

Our study was strengthened by blinding the chart abstractor and the neurology resident physician. There was no standard definition of diplopia in the literature. Thus, we defined diplopia based on history and features on exam; however, a standard definition of diplopia for future studies relying on chart abstraction may be helpful in determining all true positive cases. Further, we don’t know what ICD-10 codes were listed in the main problem list of those patients who were found to have diplopia by the gold standard but did not have an ICD-10 code for diplopia. The use of single-centre data meant a greater consistency in reporting practices, as medical documentation within an ED of one hospital would have a higher homogeneity than across hospitals. ^{Reference Cohen, Friedman, Ryan, Richardson and Adler-Milstein4} However, differences in reporting across sites may influence how the ICD-10 code for diplopia is coded in other EDs; thus, our findings need to be replicated in other cohorts. Because one individual was involved in defining the gold standard, we cannot report inter-rater reliability, which is a limitation of our work. ^{Reference McHugh5} Our sample may have missed patients with diplopia if they were not coded as such in the ED and were only seen by a neurologist in an outpatient setting or were seen by an ophthalmologist after the 7 days. Another important limitation is that people with diplopia who later get diagnosed with a neurological condition (stroke, multiple sclerosis or brain tumour) while in the ED may not receive an ICD-10 code for diplopia in the NACRS database, and since they are admitted or cared for by a neurologist, they may not end up in our cohort. Therefore, our validation study may be generalizable to patients with diplopia who are likely to be discharged from the ED, rather than those getting diagnosed in the ED.

Based on one study in Ontario that used ICD-10 code H53.2, there are an estimated 1225 patients with diplopia seen in the ED per year. ^{Reference Nanji, Gulamhusein, Jindani, Hamilton and Sabri6} The findings from our study would suggest that at least 1140 patients per year seen in the ED would need urgent attention and subsequent follow-up to identify potential disabling neurological disease. Thus, it is important to inform and educate healthcare providers on a systematic approach and adequate work-up of diplopia in the ED.

The ICD-10 code for diplopia in patients visiting the ED has high specificity and positive predictive value but low sensitivity, allowing future work to study outcomes among these patients using routinely collected health administrative data.