Can electronic health records be used to identify medical procedure‐related events indicative of difficult insertion of intrauterine devices?

Abstract

Background: Procedure‐related information is needed to contextualize use of a medical device or drug delivery system, but many procedure‐related events are poorly captured in claims data. One way to potentially obtain meaningful procedure‐related data is through electronic health record (EHR) review.

Objectives: We assessed the potential to collect procedure‐related information on difficult intrauterine device (IUD) insertions within 4 EHR systems, to satisfy a regulatory safety study.

Methods: A stratified random sample of 125 postpartum women receiving IUDs was assessed at each of 4 sites (3 Kaiser Permanente sites—Northern California, Southern California, and Washington—and Regenstrief Institute, Indiana). Structured (National Drug Codes, International Classification of Diseases 9‐Clinical Modification, Current Procedural Terminology) and unstructured (chart review of clinical notes) data were used to identify possible indicators of difficult IUD insertion. We assessed use of cervical dilation, ultrasound guidance, misoprostol, paracervical block, clinician note for “difficult insertion” or “complicated procedure,” and insertion of a second IUD within 30 days as possible indicators. Descriptive statistics for prevalence were calculated for each site.

Results: Prevalence of indicators of difficult IUD insertion across sites were cervical dilation, 0.0‐1.6%; ultrasound guidance, 0.8‐3.2%; misoprostol, 0.0‐1.6%; paracervical block, 0.0‐1.6%; provider note, 0.1‐1.6%; and insertion of a second IUD, 0.8‐1.6%. No site consistently reported the highest or lowest proportion across indicators. If all indicator categories were considered mutually exclusive, then difficult insertions would constitute 4.8‐8.8% of all insertions across sites.

Conclusions: The overall prevalence of indicators of difficult IUD insertion was slightly lower than has been published (8‐9%). The prevalence of some indicators (eg, dilation) was lower than previously published, while the prevalence of other indicators (eg, misoprostol) was similar to previous reports. Substantial effort was needed to capture specific, procedure‐related information for occasional events. Although we were able to identify indicators of difficult insertion using structured and unstructured EHR data, we cannot determine the proportion of the possible indictors that represent true difficult insertions.