To the Editor—We thank Drs Manian and Karlapalem for their interest in our work and for raising several important points that promote discussion regarding monitoring hospital patients for acquisition of coronavirus disease 2019 (COVID-19).

Our work is part of a growing body of literature aiming to understand the risk of COVID-19 acquisition among hospital patients.^{Reference Rhee, Baker and Vaidya1–Reference Lewis, Ibukunoluwa, Seidelman, Anderson, Moehring and Smith3} Reliably attributing COVID-19 acquisition to the hospital setting is beset with challenges similar to other hospital-acquired infection events. Misclassification of putative hospital-acquired infections is inevitable due to uncertainty engendered during determinations, and awareness of this uncertainty creates unease among infection control, clinical, research, and public health communities. Although surveillance methods require evaluators to assign a value of “infection=yes” or “infection=no” to their reviews, in reality, there are probabilistic underpinnings to these determinations.^{Reference Trick4} For each potential infection event, evaluators’ estimations are influenced by many factors including presence of symptoms, temporal associations of specimen acquisition to symptoms and healthcare exposures, patient comorbidities, clinician documentation, and their own probability threshold for binary classification.

For any surveillance definition, the selection of a temporal cutoff point for categorizing an event as hospital-acquired should be examined. Although we selected a cutoff point later in the hospital stay relative to other infection-related patient-safety events, other investigators have specified even later cutoff points for COVID-19.^{Reference Habermann, Tande, Pollock, Neville, Ting and Sampathkumar2} Also, our study was conducted during the early phases of the pandemic and reflected the median incubation period (>4 days) for the ancestral strain of severe acute respiratory coronavirus virus 2 (SARS-CoV-2).^{Reference Nishiura, Linton and Akhmetzhanov5} During pandemics such as COVID-19, a stringent temporal criterion for manual case reviews likely is necessary to avoid further burdening stressed infection control departments. Individual hospitals can choose a more permissive cutoff point (evaluating events early during a hospitalization) for manual review of potential hospital-acquired cases, or perhaps clusters of cases. Such decisions need to consider infection control resources as well as the dynamic epidemiology of COVID-19—strains emerge that differ in transmissibility, incubation period, virulence, and symptom profile.

Permissive cutoff points increase sensitivity but at a cost of a reduced likelihood that an event actually represents hospital acquisition. Given challenges associated with attributing events to the community or hospital (eg, gaps in information due to incomplete serial testing to evaluate community acquisition, unavailability of whole-genome sequencing of SARS-CoV-2 strains for patients and community populations, or incomplete capture of exposure history), we opted for a temporal criterion that avoided inflating the risk of misclassifying events as hospital acquired. Importantly, to avoid calculating an artificially low incidence caused by denominator inflation, our patient-days denominator rejected convention and only counted at-risk patient days. We excluded all patient days through hospital day 5 and patient days for those who were SARS-CoV-2 positive. Our estimated incidence was, in our estimation, sufficiently low in both hospitals to claim it as an uncommon event (3.3 and 1.3 per 10,000 patient days), especially in the context of a potentially high-risk environment with a high volume of SARS-CoV-2 inpatients.

The points raised by Drs Manian and Karlapalem are relevant for other hospital-acquired conditions. Incubation periods are dynamic and clinical expression heterogeneous—influenced by inoculum, organism characteristics (species and strain), and endogenous or exogenous host characteristics (eg, vaccination status or pharmacologic immunosuppression). The possibility of colonization or asymptomatic infection poses challenges when making transmission determinations for other pathogens, notably influenza infection, for which we do not routinely test for asymptomatic transmission events.^{Reference Furuya-Kanamori, Cox, Milinovich, Magalhaes, Mackay and Yakob6} Fortunately, in our evaluation, a small minority (13%) of SARS-CoV-2 detection events that occurred after day 5 were asymptomatic.

We suspect that future, more comprehensive investigations (serial swab collection paired with exposure history and whole-genome sequencing) will better inform a temporal cutoff point for hospital-onset COVID-19 infection or possibly SARS-CoV-2 acquisition. A more permissive cutoff point for manual review of potential cases may be justifiable, especially since the volume of potential case reviews has dramatically decreased. Alternatively, the recognition that deterministic classifications of events as hospital-acquired can reduce reliability will allow for assignment of probability scores.^{Reference Hota, Malpiedi, Fridkin, Martin and Trick7} Probability estimates derived from intense data collection, expert review, and genomics could mitigate problems arising from the uncertainty in making determinations and variable application of surveillance definitions. Such probability estimates would account for a lower likelihood of hospital-acquisition among asymptomatic patients and early-onset episodes. Within the limitations of our study, our data and other publications suggest that hospital-acquired COVID-19 was uncommon during the early phases of the pandemic when hospitals enacted infection prevention control efforts to care for an unprecedented number of severely ill COVID-19 patients.