Impact of to Coding Transition on Prevalence Trends in Neurology.

OBJECTIVE

To determine the effect of () to () coding transition on the point prevalence and longitudinal trends of 16 neurologic diagnoses.

METHODS

We used 2014-2017 data from the National Inpatient Sample to identify hospitalizations with one of 16 common neurologic diagnoses. We used published codes to identify hospitalizations from January 1, 2014, to September 30, 2015, and used the Agency for Healthcare Research and Quality's MapIt tool to convert them to equivalent codes for October 1, 2015-December 31, 2017. We compared the prevalence of each diagnosis before vs after the ICD coding transition using logistic regression and used interrupted time series regression to model the longitudinal change in disease prevalence across time.

RESULTS

The average monthly prevalence of subarachnoid hemorrhage was stable before the coding transition (average monthly increase of 4.32 admissions, 99.7% confidence interval [CI]: -8.38 to 17.01) but increased after the coding transition (average monthly increase of 24.32 admissions, 99.7% CI: 15.71-32.93). Otherwise, there were no significant differences in the longitudinal rate of change in disease prevalence over time between and . Six of 16 neurologic diagnoses (37.5%) experienced significant changes in cross-sectional prevalence during the coding transition, most notably for status epilepticus (odds ratio 0.30, 99.7% CI: 0.26-0.34).

CONCLUSIONS

The transition from to coding affects prevalence estimates for status epilepticus and other neurologic disorders, a potential source of bias for future longitudinal neurologic studies. Studies should limit to 1 coding system or use interrupted time series models to adjust for changes in coding patterns until new neurology-specific ICD-9 to ICD-10 conversion maps can be developed.