Statistical models for analyzing count data: predictors of length of stay among HIV patients in Portugal using a multilevel model

Enhancing detection of labor violations in the agricultural sector: A multilevel generalized linear regression model of H-2A violation counts

Agricultural workers are essential to the supply chain for our daily food, and yet, many face harmful work conditions, including garnished wages, and other labor violations. Workers on H-2A visas are particularly vulnerable due to the precarity of their immigration status being tied to their employer. Although worksite inspections are one mechanism to detect such violations, many labor violations affecting agricultural workers go undetected due to limited inspection resources. In this study, we identify multiple state and industry level factors that correlate with H-2A violations identified by the U.S. Department of Labor's Wage and Hour Division using a multilevel zero-inflated negative binomial model. We find that three state-level factors (average farm acreage size, the number of agricultural establishments with less than 20 employees, and higher poverty rates) are correlated with H-2A violations. These findings offer valuable insights into where H-2A violations are being detected at the state and industry levels.

Impact of COVID-19 epidemic on temporal pattern of new HIV diagnoses in Italy, 2021 database

BACKGROUND

New HIV diagnoses in Italy decreased drastically in 2020 due to COVID-19 related effects: 50% fewer diagnoses were reported by the National HIV Surveillance System. COVID-19 pandemic impact on HIV surveillance is unclear. We estimated the expected number of new HIV diagnoses in 2020 in order to isolate the impact of the COVID-19 pandemic.

METHODS

We analyzed 29 697 new HIV infections diagnosed from 2012 to 2020, reported to the National HIV Surveillance System. We assessed temporal trends of new HIV diagnoses applying negative binomial mixed effects models. We estimated the COVID-19 impact as the difference between the model-estimated slopes from 2012 to 2019 and the change reported in the diagnoses. The expected number of new HIV diagnoses in 2020 was also estimated and compared with the reported count.

RESULTS

Based on the historical trend, we expected a 15% (95% CI: 5-25%) decline of new HIV diagnoses in 2020. We reported, however, a 49% decrease, yielding to a 34% net decrease in the number of new diagnoses. The strongest impact was estimated in northern regions (-40%) and MSM (-38%). We estimated 761 (95% prediction interval: 350-1277) missed diagnoses during 2020, the majority of them occurring in the North (465 cases), among MSM (416) and heterosexual males (217).

CONCLUSIONS

In 2020, when excluding 15% decrease of new diagnoses attributable to the expected reduction, an additional 34% decrease was observed, representing a large decline in new HIV diagnoses associated with the COVID-19 pandemic.

Racial and Ethnic Differences in Length of Stay for US Children Hospitalized for Acute Osteomyelitis

OBJECTIVE

To examine the associations between race and ethnicity and length of stay (LOS) for US children with acute osteomyelitis.

STUDY DESIGN

Using the Kids' Inpatient Database, we conducted a cross-sectional study of children <21 years old hospitalized in 2016 or 2019 with acute osteomyelitis. Using survey-weighted negative binomial regression, we modeled LOS by race and ethnicity, adjusting for clinical and hospital characteristics and socioeconomic status. Secondary outcomes included prolonged LOS, defined as LOS of >7 days (equivalent to LOS in the highest quartile).

RESULTS

We identified 2388 children discharged with acute osteomyelitis. The median LOS was 5 days (IQR, 3-7). Compared with White children, children of Black race (adjusted incidence rate ratio [aIRR] 1.15; 95% CI, 1.05-1.27), Hispanic ethnicity (aIRR 1.11; 95% CI, 1.02-1.21), and other race and ethnicity (aIRR 1.12; 95% CI, 1.01-1.23) had a significantly longer LOS. The odds of Black children experiencing prolonged LOS was 46% higher compared with White children (aOR, 1.46; 95% CI, 1.01-2.11).

CONCLUSIONS

Children of Black race, Hispanic ethnicity, and other race and ethnicity with acute osteomyelitis experienced longer LOS than White children. Elucidating the mechanisms underlying these race- and ethnicity-based differences, including social drivers such as access to care, structural racism, and bias in provision of inpatient care, may improve management and outcomes for children with acute osteomyelitis.

Evaluating the performance of different Bayesian count models in modelling childhood vaccine uptake among children aged 12-23 months in Nigeria

BACKGROUND

Choosing appropriate models for count health outcomes remains a challenge to public health researchers and the validity of the findings thereof. For count data, the mean-variance relationship and proportion of zeros is a major determinant of model choice. This study aims to compare and identify the best Bayesian count modelling technique for the number of childhood vaccine uptake in Nigeria.

METHODS

We explored the performances of Poisson, negative binomial and their zero-inflated forms in the Bayesian framework using cross-sectional data pooled from the Nigeria Demographic and Health Survey conducted between 2003 and 2018. In multivariable analysis, these Bayesian models were used to identify factors associated with the number of vaccine uptake among children. Model selection was based on the -2 Log-Likelihood (-2 Log LL), Leave-One-Out Cross-Validation Information Criterion (LOOIC) and Watanabe-Akaike/Widely Applicable Information Criterion (WAIC).

RESULTS

Exploratory analysis showed the presence of excess zeros and overdispersion with a mean of 4.36 and a variance of 12.86. Observably, there was a significant increase in vaccine uptake over time. Significant factors included the mother's age, level of education, religion, occupation, desire for last-child, place of delivery, exposure to media, birth order of the child, wealth status, number of antenatal care visits, postnatal attendance, healthcare decision maker, community poverty, community illiteracy, community unemployment, rural proportion and number of health facilities per 100,000. The zero-inflated negative binomial model was best fit with -2Log LL of -27171.47, LOOIC of 54464.2, and WAIC of 54588.0.

CONCLUSION

The Bayesian zero-inflated negative binomial model was most appropriate to identify factors associated with the number of childhood vaccines received in Nigeria due to the presence of excess zeros and overdispersion. Improving vaccine uptake by addressing the associated risk factors should be promptly embraced.

Deep phenotyping of oxidative stress in emergency room patients reveals homoarginine as a novel predictor of sepsis severity, length of hospital stay, and length of intensive care unit stay

Background

We aimed to determine primary markers of oxidative stress (OS) in ED patients which predict hospital length of stay (LoS), intensive care unit (ICU) LoS, and sepsis severity.

Materials and methods

This prospective, single center observational study was conducted in adult patients recruited from the ED who were diagnosed with either sepsis, infection without sepsis, or non-infectious, age-matched controls. 290 patients were admitted to the hospital and 24 patients had direct admission to the ICU. A panel of 269 OS and related metabolic markers were profiled for each cohort. Clinical outcomes were direct ICU admission, hospital LoS, ICU LoS, and post-hoc, adjudicated sepsis severity scoring. Bonferroni correction was used for pairwise comparisons. Principal component regression was used for dimensionality reduction and selection of plasma metabolites associated with sepsis. Multivariable negative binomial regression was applied to predict admission, hospital, and ICU LoS.

Results

Homoarginine (hArg) was the top discriminator of sepsis severity [sepsis vs. control: ROC-AUC = 0.86 (95% CI 0.81-0.91)], [sepsis vs. infection: ROC-AUC = 0.73 (95% CI 0.68-0.78)]. The 25th percentile of hArg [odds ratio (OR) = 8.57 (95% CI 1.05-70.06)] was associated with hospital LoS [IRR = 2.54 (95% CI 1.83-3.52)] and ICU LOS [IRR = 18.73 (95% CI 4.32-81.27)]. In prediction of outcomes, hArg had superior performance compared to arginine (Arg) [hArg ROC-AUC = 0.77 (95% CI 0.67-0.88) vs. Arg ROC-AUC = 0.66 (95% CI 0.55-0.78)], and dimethylarginines [SDMA ROC-AUC 0.68 (95% CI 0.55-0.79) and ADMA ROC-AUC = 0.68 (95% CI 0.56-0.79)]. Ratio of hArg and Arg/NO metabolic markers and creatinine clearance provided modest improvements in clinical prediction.

Conclusion

Homoarginine is associated with sepsis severity and predicts hospital and ICU LoS, making it a useful biomarker in guiding treatment decisions for ED patients.

Influence of Sociodemographic and Psychosocial Factors on Length of Stay After Surgical Management of Traumatic Spine Fracture with Spinal Cord Injury

OBJECTIVE

Identifying patients at risk of increased health care resource utilization is a valuable opportunity to develop targeted preoperative and perioperative interventions. In the present investigation, we sought to examine patient sociodemographic factors that predict prolonged length of stay (LOS) after traumatic spine fracture.

METHODS

We performed a cohort analysis using the National Trauma Data Bank tabulated during 2012-2016. Eligible patients were those who were diagnosed with cervical or thoracic spine fracture with spinal cord injury and who were treated surgically. We evaluated the effects of sociodemographic as well as psychosocial variables on LOS by negative binomial regression and adjusted for injury severity, injury mechanism, and hospital characteristics.

RESULTS

We identified 3856 eligible patients with a median LOS of 9 days (interquartile range, 6-15 days). Patients in older age categories, who were male (incidence rate ratio (IRR), 1.05; 95% confidence interval [CI], 1.01-1.09), black (IRR, 1.12; CI, 1.05-1.19) or Hispanic (IRR, 1.09; CI, 1.03-1.16), insured by Medicaid (IRR, 1.24; CI, 1.17-1.31), or had a diagnosis of alcohol use disorder (IRR, 1.12; CI, 1.06-1.18) were significantly more likely to have a longer LOS. In addition, patients with severe injury on Injury Severity Score (IRR, 1.32; CI, 1.14-1.53) and lower Glasgow Coma Scale (GCS) scores (GCS score 3-8, IRR, 1.44; CI, 1.35-1.55; GCS score 9-11, IRR, 1.40; CI, 1.25-1.58) on admission had a significantly lengthier LOS. Patients admitted to a hospital in the Southern United States (IRR, 1.09; CI, 1.05-1.14) had longer LOS.

CONCLUSIONS

Socioeconomic factors such as race, insurance status, and alcohol use disorder were associated with a prolonged LOS after surgical management of traumatic spine fracture with spinal cord injury.

A comparison of statistical methods for modeling count data with an application to hospital length of stay

Background

Hospital length of stay (LOS) is a key indicator of hospital care management efficiency, cost of care, and hospital planning. Hospital LOS is often used as a measure of a post-medical procedure outcome, as a guide to the benefit of a treatment of interest, or as an important risk factor for adverse events. Therefore, understanding hospital LOS variability is always an important healthcare focus. Hospital LOS data can be treated as count data, with discrete and non-negative values, typically right skewed, and often exhibiting excessive zeros. In this study, we compared the performance of the Poisson, negative binomial (NB), zero-inflated Poisson (ZIP), and zero-inflated negative binomial (ZINB) regression models using simulated and empirical data.

Methods

Data were generated under different simulation scenarios with varying sample sizes, proportions of zeros, and levels of overdispersion. Analysis of hospital LOS was conducted using empirical data from the Medical Information Mart for Intensive Care database.

Results

Results showed that Poisson and ZIP models performed poorly in overdispersed data. ZIP outperformed the rest of the regression models when the overdispersion is due to zero-inflation only. NB and ZINB regression models faced substantial convergence issues when incorrectly used to model equidispersed data. NB model provided the best fit in overdispersed data and outperformed the ZINB model in many simulation scenarios with combinations of zero-inflation and overdispersion, regardless of the sample size. In the empirical data analysis, we demonstrated that fitting incorrect models to overdispersed data leaded to incorrect regression coefficients estimates and overstated significance of some of the predictors.

Conclusions

Based on this study, we recommend to the researchers that they consider the ZIP models for count data with zero-inflation only and NB models for overdispersed data or data with combinations of zero-inflation and overdispersion. If the researcher believes there are two different data generating mechanisms producing zeros, then the ZINB regression model may provide greater flexibility when modeling the zero-inflation and overdispersion.

Length of Stay Analysis of COVID-19 Hospitalizations Using a Count Regression Model and Quantile Regression: A Study in Bologna, Italy

This study aimed to identify and explore the hospital admission risk factors associated with the length of stay (LoS) by applying a relatively novel statistical method for count data using predictors among COVID-19 patients in Bologna, Italy. The second goal of this study was to model the LoS of COVID patients to understand which covariates significantly influenced it and identify the potential risk factors associated with LoS in Bolognese hospitals from 1 February 2020 to 10 May 2021. The clinical settings we focused on were the Intensive Care Unit (ICU) and ordinary hospitalization, including low-intensity stays. We used Poisson, negative binomial (NB), Hurdle–Poisson, and Hurdle–NB regression models to model the LoS. The fitted models were compared using the Akaike information criterion (AIC), Vuong’s test criteria, and Rootograms. We also used quantile regression to model the effects of covariates on the quantile values of the response variable (LoS) using a Poisson distribution, and to explore a range of conditional quantile functions, thereby exposing various forms of conditional heterogeneity and controlling for unobserved individual characteristics. Based on the chosen performance criteria, Hurdle–NB provided the best fit. As an output from the model, we found significant changes in average LoS for each predictor. Compared with ordinary hospitalization and low-intensity stays, the ICU setting increased the average LoS by 1.84-fold. Being hospitalized in long-term hospitals was another contributing factor for LoS, increasing the average LoS by 1.58 compared with regular hospitals. When compared with the age group [50, 60) chosen as the reference, the average LoS decreased in the age groups [0, 10), [30, 40), and [40, 50), and increased in the oldest age group [80, 102). Compared with the second wave, which was chosen as the reference, the third wave did not significantly affect the average LoS, whereas it increased by 1.11-fold during the first wave and decreased by 0.77-fold during out-wave periods. The results of the quantile regression showed that covariates related to the ICU setting, hospitals with longer hospitalization, the first wave, and the out-waves were statistically significant for all the modeled quantiles. The results obtained from our study can help us to focus on the risk factors that lead to an increased LoS among COVID-19 patients and benchmark different models that can be adopted for these analyses.