OBJECTIVES: The New York University (NYU) emergency department (ED) algorithm is being increasingly used to categorize ED care as emergent and non-emergent. However, alternative algorithms are used with publicly available data sources where part of the diagnosis code is masked. The objective of this study was to compare the agreement between the ED visit classification based on all digits of ICD-9 versus using 3-digit ICD-9 code in combination with the clinical classification software (CCS) code.

METHODS: This study used a National Hospital Ambulatory Medical Care Survey (NHAMCS, 2013) and included all ED visits with a valid non-missing primary ICD-9 diagnosis code. The percent agreement (kappa) between the NYU ED and alternative algorithm classification was the primary objective. Furthermore, multivariable logistic regression was conducted to estimate the predictors of agreement between the original NYU ED & alternative algorithms.

RESULTS: Our study found 78.25% agreement between classification obtained using ICD-9 CM’s 3-digit+CCS algorithm and NYU ED visit algorithm. Further, the odds of classifying ED visits in concordance with NYU ED algorithm, and using the ICD9 3-digit+CCS algorithm was lower for age>= 18, non-Hispanic blacks, males, follow-up visits, and those with greater number of comorbidities.

CONCLUSIONS: This study finds that ICD9 3-digit+CCS algorithm provided good agreement between the ED visit classifications based on all digits of ICD-9 versus using ICD 3-digit+CCS algorithm, however, the agreement varies based on sociodemographic factors.