Low Screening Rates for Diabetes Mellitus Among Family Members of Affected Relatives.

Fernanda Polubriaginof, Ning Shang, George Hripcsak, Nicholas P. Tatonetti, David K. VawdreyPubMed 2018

TL;DR

This paper tackles the problem of low diabetes screening rates among individuals with a family history of the disease, a key risk factor. Using a novel EHR-based method to identify family members of diabetic patients, they found that a significant portion (30.6%) were not screened according to guidelines, highlighting a critical gap in preventative care.

What Problem Does This Paper Address?

Cardiovascular disease is a leading cause of death, and abnormal blood glucose (diabetes) is a major risk factor. Delayed diagnosis of diabetes significantly increases morbidity. Current guidelines recommend screening for individuals with a family history of diabetes, among other risk factors. However, accurately capturing family history in Electronic Health Records (EHRs) has been a persistent challenge, potentially leading to missed screening opportunities. This paper aims to quantify the extent of this missed screening and identify factors associated with adherence to guidelines among a high-risk population.

Methods

The researchers employed a retrospective analysis of data from NewYork-Presbyterian Hospital/Columbia University Medical Center. They utilized a novel algorithm called RIFTEHR (Relationship Inference from the Electronic Health Record) to identify familial relationships within the EHR, leveraging patient emergency contact information (first name, last name, phone number, ZIP code). This method was previously validated and can infer relationships across multiple generations.

Once family members were identified, demographic information was extracted. Diabetes screening was measured by the presence of specific laboratory tests (fasting glucose, random glucose, or hemoglobin A1c) after the index case's diabetes diagnosis. A multivariate analysis was conducted to identify factors influencing the likelihood of receiving a screening test, using standardized coefficients (β) to quantify the impact of each parameter.

Datasets

The study used data from the EHR of NewYork-Presbyterian Hospital/Columbia University Medical Center, spanning from 2007 to 2017. This dataset included information on patients diagnosed with diabetes mellitus and their identified family members. The paper does not introduce a new benchmark dataset but rather leverages existing EHR data with a novel method for relationship inference.

Key Insights and Ideas

  • EHRs as a powerful tool for identifying at-risk populations: The study demonstrates that informatics methods applied to EHR data can effectively identify individuals at high risk for diseases like diabetes, even when traditional data capture methods are lacking.
  • The "family history gap" is real and impactful: Despite clear guidelines, a substantial portion of individuals with a known family history of diabetes are not being screened, indicating a significant gap in preventative care delivery.
  • RIFTEHR offers a promising solution for family history inference: The novel RIFTEHR method shows potential for overcoming the long-standing challenge of accurately capturing family history in EHRs, opening doors for more proactive risk identification.
  • Screening adherence is not uniform: Factors like having multiple affected family members and being female are associated with higher screening rates, suggesting potential areas for targeted interventions.

Key Results

  • The study identified 56,794 family members of diabetic patients, with 45,778 being over 18 years old.
  • Of the 27,757 family members deemed eligible for screening (over 18 with a clinical visit after the index case's diagnosis), 30.6% did not receive diabetes screening as recommended by guidelines.
  • Multivariate analysis revealed that age (β=0.67), having more than one family member diagnosed with diabetes (β=0.11), and being female (β=0.08) were the most significant contributors to being screened for diabetes mellitus.
  • Screening rates were higher for females (71.6%) compared to males (65.1%).
  • Individuals with more than one family member diagnosed had a significantly lower screening rate (28.4%) compared to those with only one affected relative (71.6% for first-degree, 63.1% for second-degree).

Limitations and Future Work

The authors acknowledge that while RIFTEHR has benefits, privacy concerns exist regarding its use in clinical practice. They also note that while cancer screening is often costly, diabetes screening is relatively simple and inexpensive, making the low adherence rate particularly concerning.

Future work could focus on understanding the specific challenges preventing recommended diabetes mellitus screening, especially given its low cost and complexity compared to other screenings. Further research is also suggested to explore the potential of RIFTEHR for identifying patients at high risk for disease development and supporting appropriate monitoring of prevention strategies. The authors also highlight the importance of family history data availability during clinical encounters.

Summary updated

Abstract

Cardiovascular disease is the leading cause of death in the United States, and abnormal blood glucose is an important risk factor. Delayed diagnosis of diabetes mellitus can increase patients' morbidity. In an urban academic medical center with a large clinical data warehouse, we used a novel algorithm to identify 56,794 family members of diabetic patients that were eligible for disease screening. We found that 30.6% of patients did not receive diabetes screening as recommended by current guidelines. Further, our analysis showed that having more than one family member affected and being a female were important contributors to being screened for diabetes mellitus. This study demonstrates that informatics methods applied to electronic health record data can be used to identify patients at risk for disease development, and therefore support clinical care.
BibTeX
@inproceedings{polubriaginof2018low,
  author = {Fernanda Polubriaginof and Ning Shang and George Hripcsak and Nicholas P. Tatonetti and David K. Vawdrey},
  title = {{Low Screening Rates for Diabetes Mellitus Among Family Members of Affected Relatives.}},
  booktitle = {PubMed},
  year = {2018},
}
Figure