Browsing by Author "Holodinsky, Jessalyn K."

Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Comparing regression modeling strategies for predicting hometime
    (2021-07-07) Holodinsky, Jessalyn K.; Yu, Amy Y.; Kapral, Moira K.; Austin, Peter C.
    Abstract Background Hometime, the total number of days a person is living in the community (not in a healthcare institution) in a defined period of time after a hospitalization, is a patient-centred outcome metric increasingly used in healthcare research. Hometime exhibits several properties which make its statistical analysis difficult: it has a highly non-normal distribution, excess zeros, and is bounded by both a lower and upper limit. The optimal methodology for the analysis of hometime is currently unknown. Methods Using administrative data we identified adult patients diagnosed with stroke between April 1, 2010 and December 31, 2017 in Ontario, Canada. 90-day hometime and clinically relevant covariates were determined through administrative data linkage. Fifteen different statistical and machine learning models were fit to the data using a derivation sample. The models’ predictive accuracy and bias were assessed using an independent validation sample. Results Seventy-five thousand four hundred seventy-five patients were identified (divided into a derivation set of 49,402 and a test set of 26,073). In general, the machine learning models had lower root mean square error and mean absolute error than the statistical models. However, some statistical models resulted in lower (or equal) bias than the machine learning models. Most of the machine learning models constrained predicted values between the minimum and maximum observable hometime values but this was not the case for the statistical models. The machine learning models also allowed for the display of complex non-linear interactions between covariates and hometime. No model captured the non-normal bucket shaped hometime distribution. Conclusions Overall, no model clearly outperformed the others. However, it was evident that machine learning methods performed better than traditional statistical methods. Among the machine learning methods, generalized boosting machines using the Poisson distribution as well as random forests regression were the best performing. No model was able to capture the bucket shaped hometime distribution and future research on factors which are associated with extreme values of hometime that are not available in administrative data is warranted.
  • Thumbnail Image
    ItemOpen Access
    Impacts of the SARS-CoV-2 pandemic on the seasonal pattern of hospitalizations for acute respiratory diseases among children in Alberta, Canada
    (2024) Lukac, Christine D; Simms, Brett; Kwong, Grace P.S.; Holodinsky, Jessalyn K.; Johnson, David W.; Kellner, James D.
    Introduction: Acute infectious respiratory diseases (ARD) among children generally have a biennial pattern – peak incidence is highest every other winter. This seasonal pattern of ARD was interrupted in 2020 by SARS-CoV-2 and non-pharmaceutical interventions (NPI). We conducted a population based retrospective cohort study in Alberta, that measured the impact on (i) the weekly incidence of hospitalizations to quantify healthcare use, (ii) the weekly percent of PICU admissions to monitor clinical severity, and (iii) the weekly average age at discharge to characterise the affected population. Methods: From Apr 2003-Dec 2023, all hospital discharges and PICU admissions for ARD (i.e. bronchiolitis, pneumonia, influenza-like-illness, and croup) among children < 18 years old were identified in the provincial hospital Discharge Abstract Database. Weekly incidence of hospital discharge was calculated using population denominators. Weekly percent PICU admissions was calculated using all hospital discharges as the denominator. Weekly average age at discharge was calculated from birth to discharge in months. Seasonal autoregressive-integrated-moving-average (SARIMA) models predicted the expected weekly outcomes from Apr 2020 onward. Incidence ratios and percent change compared observed versus expected outcomes. Analyses were conducted in R version 4.2.2 (2022-10-31) and R studio build 2022.12.0+353. Results: There were 63,776 hospitalizations for ARD among children from Apr 2003-Dec 2023: 22,963 (36.01%) for bronchiolitis, 23,977 (37.44%) for pneumonia, 10,833 (16.97%) for influenza-like-illness, and 4,984 (7.81%) for croup. Of the hospitalizations, 4,167 (6.53%) included a PICU admission. The average weekly incidence of hospitalization for ARD per 100,000 children decreased 12.71-fold during Dec 2020-Feb 2021 (0.82 observed vs. 10.42 [95%CI 5.11, 15.73] expected) and increased 1.51-fold during Dec 2022-Feb 2023 (16.28 observed vs. 10.77 [95%CI 4.71, 16.83] expected). The average percentage of PICU admissions steadily increased from 4.07% (95%CI 1.22%, 6.91%) in Dec 2003-Feb 2004 to 10.48% (95%CI 8.36%, 12.60%) in Dec 2019-Feb 2020. There was no significant change in the percentage of PICU admissions in Dec 2020-Feb 2021 and Dec 2022-Feb 2023, 11.17% (95%CI 0.00%, 26.32%) and 11.86% (95%CI 9.33%, 14.39%) respectively. During each winter season, the average age at discharge decreased to 25 months (95%CI 17.85, 33.74) annually. Similar patterns for incidence of hospitalizations, percent PICU admissions, and average age at discharge were observed for bronchiolitis, pneumonia, influenza-like-illness, and croup. Discussion: SARS-CoV-2 and NPI had significant impacts on provincial hospitalization for ARD among children. Initially hospitalizations for ARD decreased 12.71-fold during Dec 2020-Feb 2021. With SARS-CoV-2 vaccine availability, increased population immunity, and relaxation of NPI, hospitalizations for ARD increased 1.51-fold during Dec 2022-Feb 2023. However, there was no change in clinical severity based on percent PICU admissions, and no change in affected population based on average age at discharge.