Symptom- and prevention based testing of COVID-19 in nursing home residents: Symptoms and mortality, a retrospective cohort study

Background: Nursing homes (NH) residents with COVID-19 can either be tested because of presence of core symptoms (fever, cough, dyspnea) (S-based) or because of transmission prevention (TP-based). We described the clinical presentation and course of COVID-19 in NH residents who were tested either because of presence of core symptoms (S-based) or because of transmission prevention (TP-based). Methods:. XXX (XXX), is a 1185-bed NH. All NH residents who underwent SARS-CoV-2 RT-PCR testing between March 16, 2020 and May 31, 2020 were included (n = 380) in this retrospective cohort study. Clinical symptoms, temperature and oxygen saturation were extracted from medical records, 7 days before testing up to 14 days after testing. Results: SARS-CoV-2 was conrmed in 81 (21%) residents. Of these 81, 36 (44%) residents were tested S-based and 45 (56%) residents were tested TP-based. Yet, CT-values did not differ between the groups. In the 7 days prior to the test the most common symptoms in both groups were: falling (32%), somnolence (25%) and fatigue (21%). Two days before the test, we observed a stronger decrease in oxygen saturation and an increase in temperature for the S-based group compared to the T-based group that remained up to 10 days after testing. Residents with in the S-based group were 2.5 times more likely to decease within 30 days than residents in the TP-based group (HR, 2.56; 95% 1.3 to 5.2). Even though, 73% of the T-based group did eventually developed core symptoms. Conclusions: Many NH residents with a positive PCR did not have core symptoms when tested but had other signs/symptoms in the week before the positive test. Testing policies should therefore be adjusted to prevent transmission. Daily measures of temperature and oxygen saturation can contribute to earlier detection.

develop fever, but still had temperature elevations 9, . If the focus for testing is only on residents with core symptoms, many residents with COVID-19 are missed, with consequences for transmission prevention.
Furthermore, as a large proportion of all NH residents has cognitive impairment and impaired ability to communicate it is often di cult to accurately observe and interpret symptoms. Therefore, although residents may only show a rise of temperature and not fever, measuring objective parameters like temperature and oxygen saturation may provide more reliable information in the NH population in addition to observing symptoms 10-12 . Objectives This study, in which NH residents were tested either because of presence of core symptoms (symptombased testing) or because of transmission prevention (transmission-prevention based testing), aims to describe the clinical presentation and course of COVID-19 (including 30-day mortality) in both groups and differences between both groups. A second aim is to compare the CT value of the RT-PCR test between the two groups.

Study population
This study was conducted in residents living in the NH XXX (XXX), a 1185-bed skilled nursing facility with 18 locations spread over [XXX].

Study design
In this retrospective cohort study, all residents who underwent a Reverse Transcription Polymerase Chain Reaction (RT-PCR) test from March 16, 2020 (the day we started testing), up to May 31, 2020 were included.
RT-PCR tests were performed with a throat and nasopharyngeal swab. Patients who had a positive RT-PCR test were considered to be COVID-19 con rmed (cycle threshold (CT) value was also reported). A test was performed a) when a resident had core symptoms according to Dutch guidelines (fever, cough or dyspnea) 5 (referred to as symptom based testing (S-based)) for transmission prevention (referred to as transmission-prevention based testing (TP-based)) if a resident had potentially contacted with individuals with con rmed COVID-19 (residents as well as healthcare workers), after hospital admission or outpatient clinic visit or when residents were newly admitted to the NH. Residents on some wards were tested several times because one of the residents turned out to be positive. It sometimes happened that residents with already con rmed COVID-19 were tested again on these wards.

Data collection
Data about age, gender, BMI, type of ward, comorbidities and renal function of all residents that were tested on COVID-19 were extracted from the electronic health record (EHR) PinkRoccade Healthcare myCaress (myCaress).
For all residents with a con rmed COVID-19, we searched for clinical symptoms and for temperature and oxygen saturation data in the EHR from 7 days before testing up to 14 days after testing. The oxygen saturation was measured with a pulse oximeter and the body temperature was measured differently per ward, rectally or tympanum. Clinical symptoms were extracted by reviewing the EHR, using registrations of physicians as well as of nursing staff. We searched for clinical symptoms that were new or signi cantly changed (i.e. a change from incidental to more frequent falling) for each individual resident.
A symptom was regarded as new if it had not been reported in the previous 4 weeks, and changed if it signi cantly changed compared to the previous 4 weeks.
From the day that the rst patient tested positive, it was agreed for the whole NH XXX to measure the temperature and the oxygen saturation daily in all residents and to register it in the EMR (speci c sections). Data about temperature and oxygen saturation were extracted from this speci c sections. Data about 30-day mortality were also collected from the EHR.

Statistical analysis
Baseline characteristics of all tested residents were analysed descriptively. Clinical symptoms, temperature and oxygen saturation of residents with con rmed COVID-19 were also analysed descriptively. To indicate the number of missing values, the N on which percentages were calculated, are reported. All continuous data were presented as means with standard deviation and 95% con dence intervals (CI). To evaluate the differences in clinical symptoms, temperature and oxygen saturation between the two groups of residents with con rmed COVID-19 (symptom based testing vs transmissionprevention based testing) group ANOVA's were applied. We considered a P-value of < .05 to be statistically signi cant. Survival curves on 30-day mortality were estimated based on the days between the positive RT-PCR test and the date of death using Kaplan Meier curves in residents with con rmed COVID-19 (symptom based testing and transmission-prevention based testing). The mortality rate of residents with con rmed COVID-19 (symptom based testing vs transmission-prevention based testing) was compared using a Cox proportional hazard model, using three models. Model 1 unadjusted, model 2 included gender and age, model 3 included gender, age and reduced renal function. Reduced renal function was signi cant in the univariate analysis and was much more common in the S-based group. Results are presented with 95% con dence intervals and all reported P values are two-sided.
Finally, using ANOVA, we compared the CT-value of both groups (symptom based testing vs transmissionprevention based testing).
All analyses were performed with the use of the SPSS statistical package, version 26.0 (SPSS, Armonk, NY: IBM Corp).

Ethics
The Medical Ethics Committee of the Academic Medical Centre in Amsterdam, The Netherlands, approved the study protocol.

Results
On March 16, 2020, the date the rst RT-PCR SARS-CoV-2 test took place, XXX had a bed capacity for 1185 residents. Between March 16th and May 31th, 2020, 380 residents underwent a RT-PCR test for COVID-19 (601 tests). COVID-19 was ruled out in 299 (79%) residents and con rmed in 81 (21%) of residents. Of these 81 residents, 36 (44%) residents were tested because of presence of core symptoms (S based), compared to 45 (56%) residents who tested positively but were tested because of transmission prevention (TP based). (Fig. 1)

Patient characteristics
Mean age for the COVID-19 residents was 80 years (SD ± 11.2) in the S-based group and 82 years (SD ± 8.7) in the TP-based group. There were more women in both groups, S-based (53%) and TP-based (60%).
In residents who were tested but for whom COVID-19 was ruled out the mean age was 79 years (SD ± 11.1), and the majority were woman 62%. Almost all residents in the TP-based group had dementia (96%) whereas in the S-based group this was 83%. In residents tested with COVID-19 ruled out, just over half of these residents had dementia (59%). After dementia, cardiac problems, hypertension and a reduced kidney function were the most common comorbidities in residents who tested positive or negative for COVID. (Table 1.)

Symptoms 7 days before RT-PCR-con rmed SARS-CoV-2 infection
In the 7 days before a positively con rmed COVID-19 test, only fever was reported of the core symptoms and only in the S-based group. No core symptoms were present om the TP based group. The most common other symptoms were: falling, somnolence, fatigue and asthenia ( Table 2). In the TP based group of patients falling occurred more often than in the S-based group (42% vs. 19%). Prior to the moment of testing other reported symptoms were: restlessness, dysphagia, diarrhea, sore throat, drowsiness and headache. Signs/Symptoms on the day of the RT-PCR-test On the day of COVID-19 testing, cough and fever were more frequent than dyspnea in the S-based group, 75%, 74% and 28% respectively. Other symptoms in the S-based group that were reported were somnolence (33%), rhinorrhea (33%), dysphagia (25%), body weakness (19%) and diarrhea (19%).

Signs/Symptoms up to 14 days after RT-PCR-con rmed SARS-CoV-2 infection
Fever is the most common symptom in the 14 days after the test, the majority of residents with a fever were in the S-based group (76% vs. 53% TP-based). The core symptoms cough and dyspnea were also most common in the S-based group compared to the TP-based group(cough 42% vs. 13% ; dyspnea 50% vs. 13%). Of all residents with con rmed COVID-19 tested based on transmission prevention, 73% eventually developed core symptoms after an average of 4.2 days (SD ± 2.3).
Decreased appetite was almost twice as common in the S-based group as compared to the TP-based group (64% vs. 38%). Agitation and confusion were more often reported in the TP-based group compared to the S-based group (agitated 31% vs. 6% and confusion 36% vs. 11%). Body weakness and anxiety were reported more often in the S-based group (body weakness 47% vs. 24% and anxiety 28% vs. 9%). Furthermore, the following symptoms were reported: dysphagia and somnolence, both more often in the S-based group, 25% and 31% respectively. Diarrhea, fatigue and nausea were more frequently present in the TP-based group, 24%, 27% and 18% respectively.
Day-to-day uctuations in temperature, 7 days prior and 14 days after testing.
The mean temperature of the residents with con rmed COVID-19 was lower in the 7 days prior to the positive RT-PCR test compared to the day of the RT-PCR test or up to ~ 10 days after positive RT-PCR testing ( Fig. 2A). About up to two days before the positive con rmed COVID-19 test an increase in temperature was observed, and on the day of the RT-PCR-test the average temperature was highest. About seven days after the positive RT-PCR test, the temperature slowly decreased again. The temperature in the S-based group was higher before testing, on the day of the RT-PCR test and after the RT-PCR test compared to residents in the TP-based group (Fig. 2B).
Day-to day uctuations in oxygen saturations, 7 days prior and 14 days after testing by residents with COVID-19 ± and COVID-19+.
The mean oxygen saturation by the COVID-19 con rmed group was higher in the 7 days prior to the positive RT-PCR test compared to the day of the RT-PCR test or up to ~ 10 days after positive RT-PCR testing. The oxygen saturation dropped about two days before the positive RT-PCR test. The oxygen saturation uctuated until approximately 11 days after the positive con rmed COVID-19 test, after which it slowly increased again (Fig. 2C). Overall oxygen saturation was lower in the S-based group prior to testing, on the day of the RT-PCR test and after testing in comparison to the TP-based group (Fig. 2D).

Discussion
We were able to gather information about signs and symptoms about NH residents with con rmed COVID-19 in the week prior to the RT-PCR SARS-CoV-2 test up to two weeks after testing; we also collected data about 30 day mortality. Of the residents with con rmed COVID-19, 44% of the residents were tested based on the presence of core symptoms (S-based) and 56% of the residents were tested based on transmission prevention (TP-based). CT-values did not differ signi cantly between both groups. In the 7 days prior to the test, only fever was observed as core symptom in the S-based group. We found that falling and somnolence were the most common reported other symptoms prior to the test in the TP-based group. Thus, extra attention should be paid to these symptoms since they might indicate a possible infection. This is in line with previous studies, in which symptoms such as falling and delirium were reported, but in which debuts with core symptoms were more common [13][14][15] . In the 14 days following the test, apart from the core symptoms, other symptoms that were common were decreased appetite, body weakness, dysphagia, somnolence, confusion and being anxious. Residents with con rmed COVID-19 who were tested based on transmissions prevention (TP-based) and not because of the presence of core symptoms on day of testing did develop core symptoms (fever, cough, dyspnea) in 73% of the cases, but 27% did not. To prevent transmission, we cannot base the testing policy on solely testing residents with core symptoms. The results of this study ask for repeated testing of all residents that have been in close contact with a resident with con rmed COVID-19. Furthermore all residents living together with a resident with con rmed COVID should be isolated for a quarantine period and personal protective equipment should be used in the care of these residents. Repeated testing of all care personnel of these wards is also appropriate 16 .
Our results thus support the approach for repeated testing, irrespective of symptoms, in skilled nursing facilities that has been advocated since May 2020 17 .
Our study showed no difference between CT-values of S-based residents and TP based residents, similar to previous studies. 18 -19, 16,20 In our study, 40% of the residents with con rmed COVID-19, died within 30-days. This percentage is higher than the 34% reported by McMichael et al., 21 and the 26% reported by Arons et al 2 . This differences could be due to a difference in the prevalence of dementia, in our study the prevalence is much higher compared to the other studies. Moreover, we observed that residents with core symptoms on day of testing (S-based testing) were 2.5 times more likely to decease within 30 days than residents without these core symptoms on day of testing (TP-based). Thus, even though for transmission prevention all residents should be tested irrespective of symptoms, staff should still be more alert when the core symptoms are present since the mortality risk is higher for these residents as compared to residents without these symptoms. However, please note that mortality rate was still high 27% in the TP-based group and that a large part of the TP-based group did develop core symptoms after ~ 4 days.
We are the rst to show day-to-day uctuations of the oxygen saturation in COVID-19 positive NH residents; oxygen saturation decreases approximately two days before testing in all residents who tested positive. We observed that oxygen saturation was lower for the S-based group prior to test than for the TP-based group. In addition, we observed an increase in temperature ~ 3 days prior to test. Day-to-day uctuations in temperature in residents with COVID-19 were described previously 9 . Yet, our day-to-day temperature measurements add to this by making a distinction between NH residents with and without core symptoms on day of testing. We observed that temperature for the NH residents without core symptoms at day of test (TP-based) also increased 2 days prior to test. These results of the daily uctuations in temperature and oxygen saturation (shifts 1 to 2 days before the positive test) can contribute to earlier testing and earlier detection of the 'invisible' SARS-CoV-2 virus. If testing is done sooner, the residents can also be placed in a separated cohort earlier and the chance of possible spreading of the SARS-CoV-2 virus is smaller.
Furthermore, measuring the temperature and oxygen saturation on a daily basis is non-invasive and is independent from the ability to express symptoms. Many NH residents frequently have di culty putting their symptoms into words and therefore run the risk of being missed. For example, our results demonstrated that almost all residents who were tested based on transmission prevention were living in a psychogeriatric ward and had dementia. By measuring daily changes in their temperature and oxygen saturation we may reduce the chance of transmission in this vulnerable group of NH residents.
This study had some limitations that must be acknowledged when interpreting results. The study was carried out within one organization and based on EHR data. Because of the use of EHR data we may have missed symptoms and also workload during the COVID-19 pandemic may have in uenced symptom registration in the EHR negatively. As we used both care staff and physicians registration, this might have been somewhat overcome. Measuring the oxygen saturation and especially the temperature (tympanic and rectal) will not have been the same in every ward, but it is close to practice. A RT-PCR test has relatively low sensitivity (63-78%) 22 . Consequently we will have missed cases of COVID-19.

Conclusions
Many NH residents with positive PCR did not have core symptoms when tested (fever, cough and dyspnea) but had other signs/symptoms at the day of testing and in the week before the positive test. Yet, a large part of this group, did develop these core symptoms after the testing day. Daily uctuations in temperature and oxygen saturation can contribute to earlier detection. The results of this study underscore the importance of current testing policies that advice ample and repeated testing of all residents and personnel that are in close contact with a resident with con rmed COVID-19. This study shows that course pro les may be present. However, in order to substantiate this with more certainty, more research is needed that is prospective and longitudinal. Consent to participate was not required in this study. The work was conducted on retrospective deidenti ed nursing home administrative data. The Medical Ethics Committee of the Academic Medical Centre in Amsterdam, The Netherlands waived the need for informed consent.

Abbreviations
All methods were performed in accordance with the relevant guidelines and regulations.