Multilevel Modeling to Predict Factors Influencing the Length of Stay for An Inpatient Malignant Hematology Service
Bibliographic record
Abstract
Abstract Abstract 4733 Purpose: Inpatient care represents a significant cost and resource allocation issue for hospitals. The purpose of this study was to examine factors in an administrative dataset that might predict the length of stay (LOS) for inpatients with a variety of hematological malignancies and to create a simple predictive model. There is a lack of published models for this in malignant hematology patients. Methods: Data was obtained from the Discharge Abstract Database (DAD) maintained by Canadian Institute for Health Information (CIHI) for the years 2002–2005. The data abstracted included patient demographics, type and route of admission, length of stay, resource intensity weights (RIW) and a complexity code. The disease category/diagnosis is based on the case mix group (CMG) which is a simplified grouping of diseases based on the International Classification of Diseases (ICD) system. The Juravinski Hospital and Cancer Centre is a tertiary referral center dealing with all varieties of adult hematological malignancies including stem cell transplantation. The data for this study did not include patients who were treated with an allogeneic stem cell transplant. Potential factors collected on admission were modeled to be nested among individual patients using longitudinal multi-level modeling. Factors entered in the model included: age and age2, route of entry to the hospital (direct vs. emergency room vs. other), area of residence, gender, day of the admission (weekday vs. weekend) and diagnosis (CMG). We also added the number of previous admissions as a baseline variable. Institutional research ethics board approval was granted for this study. Results: Data was collected on 713 patients representing 1,739 admissions and 17,661 days of in-patient services. The number of admissions ranged from 1 to 15 with a mean of 2.5 and median 2.0 admissions per patient. The range for LOS was 1 to 155 days with mean of 10.6 days (SD 13.8) and median of 64 days. The median age of patients was 62.6 years with a range of 17–95 years, and 45% of patients were female. Thirty-nine percent of patients had a diagnosis related to a lymphoproliferative disorder or a chronic leukemia and 19 percent related to the administration of chemotherapy. Factors with non-significant B (slope) value were omitted from the model. Age is centered around the mean. Errors and residuals (μ , r) have a mean of zero and a normal distribution. Age and previous number of admissions were the most important factors in predicting LOS. The final model for estimated LOS is as follows: LOS = 3.9658+μ 0+(0.2068+μ 1)(age-65)+(4.508+μ 2)(number of previous admissions) +r The variances for the different residuals have been estimated from the model as: μ 0=73.82, μ 1=0.2436, μ 2=0.2436 and r=69.47. For each increment by 5 years above the average age, the LOS increases by 1 day and for each previous admission the LOS increases by 4.5 days. The Pseudo-R squared is approximately 0.50. Conclusion: Of the factors considered in this administrative database, age and the number of previous admissions were the only relevant factors at the time of admission which could predict LOS. This model represents a relatively simple method to estimate the LOS for patients with hematological malignancies admitted to a dedicated hematological unit. Such a model highlights patients at risk of prolonged LOS and may allow health care practitioners to focus on interventions geared towards shortening length of stay. Disclosures: No relevant conflicts of interest to declare.
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 | 0.005 | 0.014 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.003 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 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".