Using Real-Time Machine Learning to Prevent In-Hospital Severe Hypoglycemia: A prospective study
Bibliographic record
Abstract
Abstract Objective There are many examples of machine learning based algorithms with impressive diagnostic characteristics. However, a few published studies have evaluated how well they perform when deployed into clinical care. The objective of this study was to evaluate the performance of a recently validated machine-learned model to predict inpatient hypoglycemia following its implementation into clinical care on cardiovascular and vascular surgery ward. Methods We conducted a prospective analysis of a machine learning algorithm to predict hypoglycemia. The algorithm was trained, validated, and tested using data from 2013 to 2019. We employed multiple supervised machine learning techniques (e.g., extreme gradient boosting) to predict inpatient hypoglycemia and severe hypoglycemia using a wide-range of patient-level data (i.e., features) including medications, labs, nursing notes, comorbid conditions, among others. Results Our study included 3989 hospitalizations during the pre-implementation period and 1916 post-implementation. Approximately one-third of patients were women, the median age was 66 years, 23% received metformin in hospital, 7% received a sulfonylurea, and the median length of stay was 6 days. During the pre-implementation period, more than 5% of patients experienced hypoglycemia during 9.4% (N=12/127 weeks) of study weeks as compared to 0% (N=0/79 weeks) of weeks during the post-implementation period (p=0.012). The weekly variability in the rates of hypoglycemia decreased by approximately 50% from the pre-implementation (standard deviation 1.8, variance 3.4) to implementation phase (standard deviation 1.3, variance 1.6; p=0.03). There was a week-to-week decrease in hypoglycemia rates by 0.03 events per week [95% CI: -0.04, -0.01] (p = 0.004) but no significant change in weekly rates of hyperglycemia (−0.04 [95% CI: -0.10, 0.01]; p=0.102). The severe hypoglycemia events per 100 patients per year was 1.3 pre-implementation and 1.1 following implementation. Discussion and Conclusion Our prospective analysis of a recently validated machine learned model to prevent hypoglycemia demonstrated a reduction in the rates of inpatient hypoglycemia. Our study suggests that machine learning methods can be leveraged to prevent inpatient hypoglycemia.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.009 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".