We aimed to provide evidence that the ICD-10-CM phecode map resulted in phenotypes like those sourced from the ICD-9-CM phecode map. First, we selected 357,728 patients in the VUMC EHR who had ≥1 ICD-9-CM and ≥1 ICD-10-CM codes in two 18-month windows. We selected windows to occur prior to and after VUMC’s transition to ICD-10-CM. To reduce potential confounders, we left a 6-month buffer after ICD-9-CM was replaced with ICD-10-CM. Further, the ICD-10-CM observation window ended before VUMC switched from its locally developed EHR [27 (link)] to the Epic system. This created two windows ranging from January 1, 2014 to June 30, 2015 for ICD-9-CM, and January 1, 2016 to June 30, 2017 for ICD-10-CM (Figure 3 ). The final cohort consisted of 55.10% (197,109/357,728) females with mean age of 45 (SD 25) years old. From the two observation periods, we extracted all ICD-9-CM and ICD-10-CM codes for each patient. We then mapped these codes to phecodes using the ICD-9-CM phecode [14 ] and ICD-10-CM phecode maps.
We used the patient cohort to test our hypothesis that the ICD-10-CM phecode map created phenotype definitions that were comparable to those generated using the gold-standard ICD-9-CM phecode map. For this analysis, we used four common chronic diseases (Hypertension, Hyperlipidemia, Type 1 Diabetes, and Type 2 Diabetes) and chose one acute disease (Intestinal infection) as a negative control. We expected that a large majority of the chronic disease patients and a small minority of the acute disease patients from the ICD-9-CM era would reproduce the same phenotypes during the ICD-10-CM era. We defined the phenotype cases as follows: Hypertension with phecodes 401.* (* means one or more digits or a period); Hyperlipidemia, phecodes 272.*; Type 1 diabetes, phecodes 250.1*; Type 2 diabetes, phecodes 250.2*; Intestinal infection, phecodes 008.*.
For each phenotype, we reported the number of ICD-9-CM cases and the number of those individuals who were also ICD-10-CM cases. To identify the possible reasons for individuals who were not identified as phenotype cases in the ICD-10-CM period, two authors with clinical training (PW, WQW) manually reviewed the EHRs of ten randomly selected patients from each chronic disease group, except Type 1 diabetes, for a total of thirty patients.
We used the patient cohort to test our hypothesis that the ICD-10-CM phecode map created phenotype definitions that were comparable to those generated using the gold-standard ICD-9-CM phecode map. For this analysis, we used four common chronic diseases (Hypertension, Hyperlipidemia, Type 1 Diabetes, and Type 2 Diabetes) and chose one acute disease (Intestinal infection) as a negative control. We expected that a large majority of the chronic disease patients and a small minority of the acute disease patients from the ICD-9-CM era would reproduce the same phenotypes during the ICD-10-CM era. We defined the phenotype cases as follows: Hypertension with phecodes 401.* (* means one or more digits or a period); Hyperlipidemia, phecodes 272.*; Type 1 diabetes, phecodes 250.1*; Type 2 diabetes, phecodes 250.2*; Intestinal infection, phecodes 008.*.
For each phenotype, we reported the number of ICD-9-CM cases and the number of those individuals who were also ICD-10-CM cases. To identify the possible reasons for individuals who were not identified as phenotype cases in the ICD-10-CM period, two authors with clinical training (PW, WQW) manually reviewed the EHRs of ten randomly selected patients from each chronic disease group, except Type 1 diabetes, for a total of thirty patients.
Free full text: Click here
