Thank you so much, Pavan and John, so far. Thank you very much for the opportunity to speak today. I am a medical intensivist and a clinical epidemiology researcher, and I study this phenomenon of healthcare capacity strain. And so I was brought into the SARI PREP group to look at the simultaneous measurements of hospital stress and strain that were going on during the pandemic alongside the rest of the SARI PREP study goals. So I'll be talking today about severe acute respiratory infection and hospital stress patterns and impact, and this is the results of that work. And here are my financial disclosures. I'm funded by federal and foundation sources, and I have a few financial relationships, none of which I think pose a conflict for today's talk. So as was mentioned, SARI PREP is a CDC Foundation-funded and SECM Discovery-housed multi-center prospective cohort study, and the primary goals, as were outlined, are looking at SARI patient care and clinical biologic characteristics and outcomes. But we were separately motivated by the idea that hospitals experienced substantial stress during the COVID-19 pandemic. We defined that as threats to standard operations, and it's not well known how this stress manifested at individual hospitals, aside what each hospital knows about their own experience, but not as kind of a collective population of hospitals measure, which would be very important for preparedness efforts and kind of future-looking preparedness. So we aim to understand four things, patterns of hospital stress over time, where stress was located within individual hospitals, within hospital correlations between individual stress measures, and then temporal relationships between stress measures and local SARS-CoV-2 case activity. So these data were collected with a weekly hospital stress survey that was part of the SARI PREP study. This was deployed from November 2020 through June 2022. It's been since modified, but the data that we'll present today is through June 2022. And it had rolling cohort entries, so SARI PREP study sites, as they came online in the study, could also enter the stress survey data collection. And the respondents were SARI PREP site leaders or their designees at each individual site. And the survey assessed a number of things. It assessed perceived stress, that is to say reported and perceived by the respondents at different levels, at the hospital level overall, in the emergency department, and in the ICU. It measured a host of deviations or potential deviations from standard operating procedures that we'll go through, and then staffing and equipment availability. This is a map of the SARI PREP study sites on the left, almost all of which, but not 100 percent, were involved with responding to the stress survey. And here's a list of the sites, both the health systems and individual hospitals, and their kind of rolling cohort entries. You can see how different sites ended up entering the study over the course of the last two years, three years. In addition to the data collected in the survey itself, we also extracted state and county SARS-CoV-2 case counts from public health departments and standardized those per capita as an additional data point to link with the stress survey data. We categorized those SARS-CoV-2 case counts as periods of surges or between surge periods, and the surges were defined by weeks that included a day with a 60,000 or greater incidence SARS-CoV-2 cases nationally. And our study period covered essentially five waves, the third of three ancestral waves, and then alpha, delta, Omicron BA1, and Omicron BA212 in the early parts of BA4 and 5. And then between surge periods, which were weeks that did not include any days with that greater than 60,000 incident cases, made up about 26% of study weeks, those between surge periods. We had two units of analysis. The first was hospital week. So this reflects the experiences of individual study hospitals and was used for within and among hospital analyses. And the second is pandemic or study week or calendar week, and that was useful for across pandemic analyses, including among variant analyses. And we did four modeling approaches to answer those four priority questions I put forward earlier. So the first is to look at the correlation between individual stress measures, and for that we used a Spearman's correlation coefficient analysis. The second was to look at the association between SARS-CoV-2 county cases and hospital stress. For that we used multivariable logistic regression adjusted for the variant surge period and the hospital on the level of the hospital week. And then for among hospital variation and stress, we looked at, again, a multivariable logistic regression models, this time stratified by hospital adjusted for SARS-CoV-2 case counts. And then finally, to look at the temporal relationship between SARS-CoV-2 cases and hospital stress, we measured different time intervals, the first from the start of stress measures to the peak of local cases, then from the peak of local cases to the end of stress, the total duration of stress, and then the level of case counts of SARS-CoV-2 case counts locally at the start and end of stress for each of those metrics. And then for each surge period and for each stress measure, we compared them using two-sample Wilcoxon-Rank sum testing. So here are the results. The data I'll present today is from 13 hospitals across seven health systems in six U.S. states that span 85 pandemic weeks, so 85 calendar weeks, and then 839 total hospital weeks of data and five variant surges. This is kind of an overview of how these data came together. So the x-axis is time, and the black line indicates the SARS-CoV-2 cases for all the contributing counties during that time period. The red, green, and blue lines represent hospital, ICU, and emergency department stress, perceived stress by week, cumulatively over all the study hospitals, a percentage of hospitals reporting for each of those. And then the shaded periods represent the five different surge periods, different variants, with the white being between surge periods. And you can certainly see correlations. I'll show you the specific analyses in a moment, but correlations between case counts and stress and potentially different patterns of stress between different stress measures. We collected a huge amount of data with this. I'll just highlight a few things. The first is that over the full study period, hospitals reported stress on the whole hospital level, around 43% of hospital weeks, in the intensive care unit, around 32% of hospital weeks, in the emergency department, around 14% of hospital weeks. And the most common deviations from standard operating procedures were increasing hospital staffing, denying inter-hospital transfers, and canceling elective surgeries. Those occurred in 19, 15, and 14% of weeks. So a substantial disruption for normal operating procedures. The Omicron BA1 subvariant surge in particular was very disruptive across reported stress and across reported care deviations. But even during between-surge periods, things did not return to normal. So 14% of hospitals were still reporting hospital stress on any given week. When we looked at correlation between different stress measures, ICU stress was highly correlated with overall hospital stress, with a row value of 0.82. Emergency department stress was much more moderately correlated with both overall hospital stress and ICU stress. And we can take a look at that in a moment. And then from a standpoint of association between local SARS-CoV-2 case activity and hospital stress, we found there was indeed an association. And so for every 10 new SARS-CoV-2 cases, or 10 additional SARS-CoV-2 cases per 100,000 residents, we would expect by this analysis to see an 8.7% increase odds of hospital reporting stress that week, a 6.5% increase odds of the ICU reporting stress, and a 3.8% odds of stress in the emergency department. And then when we looked at between or among hospital variation, this is data from the Omicron BA1 subvariant surge. We have our 12 study hospitals at this point displayed on the x-axis. And then on the y-axis is the percent of weeks during this period that they reported stress with the three different primary measures, hospital stress in red, ICU stress in green, and ED stress in blue. And then you can see, you know, going from hospital one all the way up to hospital 12, there was significant variation even within the same surge period with how often these hospitals were reporting stress. So different hospitals are experiencing these acute surge periods differently. And then finally, the results of our analysis looking at the temporal relationship between SARS-CoV-2 cases and hospital stress, to orient you, the left figure is the Delta variant surge, the right is the Omicron BA1 subvariant surge. The y-axis is county SARS-CoV-2 cases per 100,000 population on a county level. The x-axis is weeks with the zero line aligning with the peak of SARS-CoV-2 cases. And so you can see the black curve is SARS-CoV-2 cases that come up, peak, and then descend. And the indicators are for, again, red, green, and blue, overall hospital, ICU, and ED stress. And the first three are the start of stress, and the next three squares are the end of stress. And so I'll point out a couple things. One is that all the measures of stress seem to start around the same time, and they do so either at the peak or in the run-up to the peak. But on the contrary, or in contrast, the end of stress is much later. So in the Delta variant surge, it took six weeks for ED stress to abate, and it took 10, 11, 12 weeks for ICU and overall hospital stress to abate. And in the Omicron BA1 subvariant surge, it took, again, five weeks for ED stress, and kind of five and a half to six weeks for overall hospital and ICU stress. So in certain scenarios, ED stress may abate a bit earlier, but probably the most predominant finding here is that perceived stress, at least, persists well after SARS-CoV-2 cases peak and decline in a given area. And that has potential implications for hospital preparedness, for staffing, resiliency, and so forth. So I'll conclude by saying that hospital stress and care deviations during the COVID-19 pandemic were common. They varied by surge period and by hospital, and perceived potentially avoidable patient harm was present but rare, perceived. SARS-CoV-2 county cases was associated with overall hospital ICU and ED stress, and then overall stress during the pandemic for SARI patients was highly correlated with ICU stress but only moderately correlated with ED stress. And then finally, hospital stress measures persisted for weeks after a surge's peak, potentially with emergency department stress resolving the earliest. We have a lot of future work that's ongoing or planned I'll mention a few of these. One is to look at, we'll be linking this dataset with the primary, sorry, patient-level dataset to link hospital stress with patient-level process of care and outcomes. We'd like to link perceived stress, this is stress sensed by and reported by clinicians and leaders, with quantitative measures of hospital stress, things like occupancy and acuity of patients within a hospital. There's other nuances that we'd like to look at between individual strain metrics and individual strain metrics and their relationship to local cases. And with the overarching goal here, looking at developing a more sustainable longitudinal surveillance program that can pick up early changes for future threats. So I'll stop there. I just want to thank the many, many collaborators that contribute to this study, all the different study sites, the funder, the CDC Foundation, our home, and SCCM Discovery. And then this is just a partial list of all the collaborators that worked on this particular part of the project. So thank you so much for the opportunity.

healthcare capacity strain

SARI PREP group

hospital stress

inter-hospital transfers

hospital preparedness