Accuracy of hospital discharge data to identify pregnant women with disabilities

From the Mueller Group, the Division of Public Health Sciences

International Classification of Diseases (ICD) codes in hospital discharge records are an important tool for epidemiologic studies examining health conditions among women with deliveries. These records are a good resource for examining maternal and infant outcomes among pregnant women with conditions that may result in disability. This research is important because these women may have difficulty traveling to caregiver offices for prenatal care, and often the offices and facilities are inaccessible once they arrive.

Dr. Beth Mueller, from the Division of Public Health Sciences, listed examples of these difficulties, “Some offices are unable to adequately assist persons using manual or power wheelchairs for mobility. Routine written or oral instructions may not be ‘user-friendly’ for women with sensory disabilities (hearing loss/deafness or vision impairment/blindness) or delivered in a format or at a level of comprehension appropriate for women with intellectual or developmental disability. There are no population-based registries in the US for persons with these conditions, making it difficult to conduct studies to identify barriers to appropriate care, or to assess the occurrence of adverse outcomes relative to women without these conditions.” 

Few studies have assessed the accuracy of hospital discharge data to identify these conditions. Dr. Mueller stated, “Studies relying on hospital discharge data are limited by the accuracy with which health conditions and procedures are classified and coded. As part of a larger study on pregnancy outcomes in women with physical, sensory, or intellectual disabilities, we examined the validity of the ICD codes used by health registries. Based on database coding at the time of delivery, we identified women with selected conditions that may cause disability and checked their medical record to see if the coding correctly identified the relevant condition." The results of this study were published in the journal Epidemiology.

Hospital discharge records which were linked to live birth/fetal death records in Washington State were utilized for the analyses. The Mueller Group commented on the use of hospital discharge records in this study, “Hospital discharge databases are increasingly used for research. We often enrich cancer registry or other health registry data by linkage to these databases to gain insights into underlying comorbidities or conditions that may impact outcomes. This is an efficient and economical way to conduct studies and often provides a wider, population-based perspective and greater numbers than most single center studies or studies relying on patient interview.” After receiving human subject’s approval, the Mueller group identified 179 (during 1987 -2008) and 101 (during 2009 – 2012) pregnant women with deliveries at a large hospital and who had disabilities based on the ICD9 codes in hospital discharge records. Charts of 390 randomly selected pregnant women without disabilities who delivered at the same hospital were also reviewed. Doctoral student, Naomi Schwartz used the medical records as the gold standard to measure the sensitivity, specificity, and positive predicted values (PPV). 

Graphical Representation of the Classification of Diseases
Classification of Diseases Image provided by Dr. Beth Mueller.

The estimated PPV (percent of women with a code for a specific condition who truly have that condition) for identifying women with disabilities based on the ICD9 codes increased from 91% during 1987-2008 to 98% during 2009-2012.  Dr. Mueller summarized, “We found that, for the most part, conditions that may result in physical disability (e.g., multiple sclerosis, spinal cord injury) are more accurately recorded than sensory or intellectual disability-related conditions, at least as based on hospital discharge coding at the time of delivery.” 

This is one of the first studies to examine coding at the time of delivery. Many studies of pregnancy and infant outcomes are based on ICD codes collected at or around the time of the delivery hospitalization. Dr. Mueller discussed the importance of the study, “As many of these studies rely on data exclusively from this data resource, this knowledge helps us better interpret results related to subsequent morbidity, mortality, or other outcomes in this vulnerable population.”

This research was supported by from the Eunice Kennedy Shriver National Institute of Child Health & Human Development.

Fred Hutch/UW Cancer Consortium member Dr. Beth Mueller participated in this research.

Schwartz, N. R. M., Crane, D. A., Doody, D. R., Schiff, M. A., & Mueller, B. A. 2020. Assessment of the Accuracy of Identification of Selected Disabilities and Conditions in Hospital Discharge Data for Pregnant Women. Epidemiology, 1.