A Systematic Review and Meta-analysis of Upper Airway Swab Collection for Detection of Viral and Bacterial Pathogens by Individuals or Caregivers Compared to Health Care Workers

ABSTRACT

Self- or caregiver collection of upper airway swabs reduces infectious exposures of health care workers (HCWs) and the need to redeploy clinical staff to testing roles. We aimed to determine whether self- or caregiver collection has adequate diagnostic performance for detection of viral and bacterial upper airway pathogens. We did a systematic review and meta-analysis of studies comparing diagnostic accuracy of self- or caregiver-collected upper airway swabs collected by patients or caregivers compared to HCWs. All study types except case reports and series were included if sufficient data were presented to calculate sensitivity, specificity, and Cohen’s kappa. Studies published from 1946 to 17 August 2020 were included in the search. We did a meta-analysis to assess pooled sensitivity and specificity. Twenty studies were included in the systematic review and 15 in the meta-analysis. The overall sensitivity of swabs collected by patients or caregivers compared to HCWs was 91% (95% confidence interval [CI], 87 to 94), and specificity was 98% (95% CI, 96 to 99). Sensitivity ranged from 65% to 100% and specificity from 73% to 100% across the studies. All but one study concluded that self- or caregiver-collected swabs were acceptable for detection of upper airway pathogens. Self- and caregiver collection of upper airway swabs had reassuring diagnostic performance for multiple pathogens. There are numerous potential benefits of self- and caregiver-collected swabs for patients, families, researchers, and health systems. Further research to optimize implementation of sample collection by patients and caregivers is warranted.

INTRODUCTION

There are many clinical indications for upper airway sampling for pathogen detection. Samples are most commonly collected using a swab. Depending on the pathogen(s) being studied, these swabs are tested by culture and/or culture-independent (usually molecular) techniques (1). Swabs are typically collected by health care workers (HCWs), including doctors and nurses in the clinical setting, or by trained researchers and study nurses for research purposes (all considered HCWs here). Potential risks to HCWs collecting swabs include infectious exposure during the period of close contact with the patient, exacerbated by coughing and gagging (2). In the current coronavirus pandemic (COVID-19) caused by the SARS-CoV-2 virus, there have been unprecedented testing campaigns relying largely on collection of swabs by HCWs, with risk of infectious exposure to HCWs, potential for limiting the workforce available to deliver clinical care, and use of limited stocks of personal protective equipment (3, 4). Testing by HCWs in centralized settings may also hinder physical distancing efforts, with increased transmission risks while groups wait for testing (5). Collection of swabs by patients or caregivers could mitigate against many of these challenges.

The benefits of self-collection of diagnostic samples have been demonstrated previously. Influenza surveillance can be optimized using self-collection by mildly unwell patients. In 2007, self-collection by callers to a UK telephone health helpline identified some of the earliest reports of influenza in the community (6). Uptake and participation in testing initiatives may be improved with self- or caregiver swabs, as satisfaction is higher than collection by HCWs (7–10). An online survey of patients with SARS-CoV-2 symptoms found a much higher proportion willing to be tested via a self-collected saliva sample (92%) or nasal swab (88%) than by a drive-through HCW swab collection (71%) (11). A 2019 meta-analysis by Seaman et al. of self-swabbing for influenza detection found high sensitivity, specificity, and acceptability (12).

In this review, we aimed to determine whether self- or caregiver collection of swabs has sufficient diagnostic performance for detection of viral and bacterial upper airway pathogens broadly, which might enable its early routine implementation for surveillance and diagnosis of common and emerging upper airway infections.

METHODS

We undertook a systematic review and meta-analysis of studies comparing the diagnostic accuracy of HCW-collected swabs to self- or caregiver-collected swabs, following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (see the PRISMA checklist in the supplemental material) (13).

Search strategy.

Medline (including Epub ahead of print, in-process, and other nonindexed citations) and Embase databases were searched using the Ovid interface on 17 August 2020. The search was restricted to English-language articles, and terms included nose, pharynx, nasopharynx, throat, oropharynx, specimen handling, swab, collected by parent, patient, self, carer, doctor, nurse or investigator, sensitivity, specificity, and observer variation (see the search strategy in the supplemental material). There were no further search restrictions. Articles published from 1946 to 2020 were included. Titles and abstracts of all results were screened for inclusion and exclusion criteria by a single reviewer (C.H.). Two reviewers (C.H., D.E.L.) completed a closer review of full-length articles and manual searching of reference lists in relevant articles to identify any papers missed by the initial search. Discrepancies were resolved through discussion between the two reviewers.

Inclusion and exclusion criteria.

Apart from case reports and series, all study types presenting original research comparing self- or caregiver-collected upper airway swabs to HCW-collected swabs were included if sufficient data from swabs were presented to calculate sensitivity, specificity, and Cohen’s kappa statistic. Meta-analyses and systematic reviews were excluded as per these criteria. Conference abstracts, letters, and commentaries were included if sufficient detail was presented. For inclusion, each participant must have had both self- or caregiver-collected and HCW-collected swabs from a site in the upper airways. Testing methodology was required to be the same for the paired samples. Studies investigating disease and/or carriage of any bacterial or viral upper airway pathogen were included. Studies including swabs from sites in addition to the upper airways were included, but only data relevant to the upper airways were extracted. There were no limitations on the age of study participants. For young children, studies that included swabs collected by caregivers were considered together with self-collected swabs. A parent or guardian who attended with the child could collect a swab instead of the child themselves.

Studies were excluded if samples were collected by a method other than swabbing or if the endpoint was not identification of a viral or bacterial pathogen. Studies were excluded if insufficient data were reported to calculate sensitivity, specificity, and Cohen’s kappa or did not include paired results for each participant. We excluded studies from the meta-analysis where initial screening was done (i.e., only participants with an initial positive result had a comparator swab collected) but included these in the systematic review.

Quality assessment.

Study quality and risk of bias were assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) checklist (14). Items 5 and 6 in the checklist relating to participants receiving both an index and reference test were removed, as these were key inclusion criteria.

Data collection.

The following variables were extracted: location, pathogen(s) of interest, site swabbed, setting of collection, job title of investigator collecting swabs, age range of participants, total paired swabs, total positives, number positive by both swabs, number positive by self-swab or caregiver-collected-swab only, number positive by HCW-collected swab only, and number negative by both. Data extraction was completed by two reviewers (C.H., D.L.). Where key missing data were unavailable, studies were excluded from the meta-analysis.

Data analysis.

We calculated sensitivity and specificity with 95% confidence intervals (CIs) as well as the Cohen’s kappa statistic to compare self-collected to HCW-collected swab results using Stata SE software (version 16) for each study. Where some studies collected multiple swabs from participants to investigate results from different swabbing sites, each site was considered separately in the meta-analysis.

Meta-analysis of sensitivity, specificity, and positive and negative likelihood ratios was done using a bivariate random-effects model designed for use in diagnostic accuracy studies (Stata SE commands “metandi” and “midas”). Subgroup analyses for pathogens other than SARS-CoV-2 were not pursued given the small number of studies for each. Publication bias was assessed by a Deeks’ funnel plot asymmetry test (Stata SE command “midas”).

RESULTS

Systematic review.

The search strategy returned 1,148 results with 947 unique results after duplicates were removed (Fig. 1). After abstract screening, full-text review, and manual reference list search, 20 studies were included in the systematic review (Table 1). Five were excluded from the meta-analysis based on our exclusion criteria.

FIG 1.

Search strategy flow diagram. This flow diagram follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (13).

TABLE 1.

Characteristics of studies included in the systematic review^a

Author (yr)	Location	Pathogen	No. of results	Setting	Age range	Self-swab site	HCW swab site	Test	QUADAS score
Altamirano (2020)	California, USA	SARS-CoV-2	30	Clinic	≥18 yrs	Nasal and pharyngeal	Nasal and pharyngeal	RT-PCR	12
Kojima (2020)	California, USA	SARS-CoV-2	45	Home	≥18 yrs	Nasal and oral fluid (supervised, unsupervised)	NP	RT-qPCR	12
Wehrhahn (2020)	Australia	SARS-CoV-2	236	Testing facility	7–81 yrs	Nasal and pharyngeal	Nasal, pharyngeal and NP	RT-PCR	12
Yuan-Po Tu (2020)	Washington, USA	SARS-CoV-2	530	Outpatient clinic	Not stated	Tongue, nasal, midturbinate	NP	RT-PCR	11
Fragoso (1989)	New Jersey, USA	Group A Streptococcus	98	Clinic	4–17 yrs	Pharyngeal	Pharyngeal	Culture	8
Katz (1974)	Maryland, USA	Group A Streptococcus	137	Clinic	“Children”	Pharyngeal	Pharyngeal	Culture	8
Murray (2015)	Minnesota, USA	Group A Streptococcus	402	Clinic	≥3 yrs	Pharyngeal	Pharyngeal	RT-PCR	12
Dhiman (2012)	Minnesota, USA	Influenza A and B	72	Emergency room	18–92 yrs	Nasal (midturbinate)	Nasal (midturbinate)	RT-PCR	12
Esposito (2010)	Italy	Influenza A and B	203	Emergency room	6 mo to 5 yrs	Nasal (midturbinate)	Nasal (midturbinate)	RT-PCR	11
Granados (2017)	Ontario, Canada	Influenza A and B	459	Home	Children and adults	Nasal (midturbinate)	Nasal (midturbinate)	RT-PCR	11
Goyal (2017)	Thailand	Influenza A and B	127	Clinic	≥65 yrs	Nasal	NP and nasal	RT-PCR	11
Ip (2012)	Hong Kong	Influenza A and B	196	Home and clinic	Children and adults	Nasal and pharyngeal	Nasal and pharyngeal	RT-PCR	10
Barbee (2020)	Washington, USA	C. trachomatis and N. gonorrhoeae	258	Clinic	Not stated	Pharyngeal	Pharyngeal	Culture	12
Dangerfield (2019)	Maryland, USA	C. trachomatis and N. gonorrhoeae	79	Clinic	≥18 yrs	Pharyngeal	Pharyngeal	NAAT	12
Freeman (2011)	California, USA	C. trachomatis and N. gonorrhoeae	480	Clinic	Adults	Pharyngeal	Pharyngeal	NAAT	10
Akmatov (2012)	Germany	Respiratory pathogens	82	Research institute	18–69 yrs	Nasal (anterior nares)	Nasal	Multiplex PCR	10
Larios (2011)	Ontario, Canada	Respiratory pathogens	60	Not stated	23–59 yrs	Nasal (midturbinate)	NP	RT-PCR	10
Vargas (2016)	New York, USA	Respiratory pathogens	30	Home	Children and adults	Nasal	Nasal	Multiplex RT-PCR	12
Lautenbach (2009)	Pennsylvania, USA	Staphylococcus aureus	56	Hospital (inpatients)	Children and adults	Nasal and pharyngeal	Nasal and pharyngeal	Culture	10
van Cleef (2012)	Netherlands	Staphylococcus aureus	105	Hospital (staff)	Adults	Nasal and pharyngeal	Nasal and pharyngeal	Culture	11

^aHCW, health care worker; NAAT, nucleic acid amplification test; NP, nasopharyngeal; RT-PCR, reverse transcription PCR; qPCR, quantitative PCR; SARS-CoV-2, severe acute respiratory coronavirus 2; QUADAS, evidence-based tool for Quality Assessment of Diagnostic Accuracy Studies.

Pathogens included in the review were influenza (n = 5 studies), SARS-CoV-2 (n = 4), group A Streptococcus (GAS; n = 3), Staphylococcus aureus (n = 2), pharyngeal Chlamydia trachomatis and Neisseria gonorrhoeae (n = 3), and a panel of respiratory viruses (n = 3). All studies were published after 2009 except for two studying GAS (1974 and 1989). Studies included swabs from a range of upper airway sites, with nasal swabs most common (n = 14). Pharyngeal swabs only were used for GAS, chlamydia, and gonorrhea. In five studies, there were HCW swabs collected from different anatomical site(s) than by self-collection. In each case, this was where there was an HCW-collected nasopharyngeal swab, a site generally considered unsuitable for self- or caregiver collection. A self-collected nasal swab was the usual comparator in these instances.

Study quality, rated by the QUADAS checklist, was high across all studies included in the systematic review, suggesting a limited risk of bias. All studies scored 10 to 12 out of 12 on the modified QUADAS checklist, except for the two older GAS studies (both 8 out of 12).

There was a total of 3,685 participants across all included studies, ranging from 30 to 530 in each study. Freeman et al. included the most stringent eligibility criteria (only men who have sex with men were eligible for enrollment) (15). In four pediatric studies, swabs were collected by caregivers. In the study by Murray et al., children aged 11 years and older could self-collect their pharyngeal swab (16). A further four studies had different age-related exclusion criteria for self-collection by children. The remainder of studies recruited only adults who self-collected swabs.

All but one study concluded that self-swabbing was acceptable and recommended its implementation for clinical and/or research settings. Fragoso et al. (1989) concluded that collection of pharyngeal swabs and preparation of culture plates for detection of GAS by parents was insufficiently sensitive compared to physician swab and culture preparation (17).

Meta-analysis.

Five articles included in the systematic review were excluded from the meta-analysis. Studies by Altamirano et al. (2020), Barbee (2020), Ip et al. (2012), and Lautenbach et al. (2009) were excluded because participants were prescreened for the disease being investigated and were only eligible for recruitment if positive (18–21). The study by Granados et al. (2017) was excluded because there were insufficient data reported (22). In the study by Wehrhahn et al. (2020), only SARS-CoV-2 data were included in the meta-analysis, as there were insufficient data reported for the other respiratory viruses (23).

In several studies, swabs by participants and HCWs were from different sites. In the study by Kojima et al. (2020), participants self-collected a nasal swab, a supervised oral fluid swab, and an unsupervised oral fluid swab, while a nasopharyngeal swab was collected by HCWs (24). In the study by Yuan-Po Tu et al. (2020), HCWs also collected nasopharyngeal swabs, and participants self-collected nasal, midturbinate, and tongue swabs (4). Results from each site were analyzed separately in the meta-analysis. The Wehrhahn et al. (2020) study included two separate study centers with different swabbing procedures (23). The two centers were considered separately in the meta-analysis.

The meta-analysis included results of 3,954 swabs. Overall sensitivity of self-collected upper airway swabs compared to HCW-collected swabs was 91% (95% CI, 87 to 94), and specificity was 98% (95% CI, 96 to 99; Fig. 2). Sensitivity ranged from 65% to 100% and specificity from 73% to 100% across the studies. The positive likelihood ratio was 51.5 (95% CI, 19.9 to 133.3), and negative likelihood ratio was 0.09 (95% CI, 0.06 to 0.14). The diagnostic odds ratio was 550.6 (95% CI, 180.1 to 1,683.2). In a meta-analysis of the 811 participants in the 3 studies looking at SARS-CoV-2 data, pooled sensitivity was 91% (95% CI, 82 to 95%), and specificity was 99% (95% CI, 93 to 100%), with a positive likelihood ratio of 95.4 (95% CI, 12.5 to 731.1) and negative likelihood ratio of 0.095 (95% CI, 0.048 to 0.189). Publication bias was assessed using Deeks’ funnel plot asymmetry (Fig. S1 in the supplemental material), with a P value of 0.17, suggesting low risk of bias.

FIG 2.

Forest plots of sensitivity and specificity of self-collected swabs compared to investigator-collected swabs with pooled estimates of sensitivity and specificity. Shown are sensitivity and specificity from each study (black dots in gray squares), corresponding 95% confidence intervals, and overall summary estimates for sensitivity and specificity (clear diamonds) calculated by bivariate random-effects meta-analysis. HCW, health care worker; SOF, supervised oral fluid; UOF, unsupervised oral fluid; NP, nasopharyngeal; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2.

For most studies, Cohen’s kappa was greater than 0.8, indicating almost perfect agreement (Table 2). Studies with low kappa values were Fragoso et al. (1989) comparing self-collected HCW-collected pharyngeal swabs for GAS detection, self-collected oral fluid compared to HCW-collected oral fluid for SARS-CoV-2 in the study by Kojima et al. (2020), and van Cleef et al. (2012) comparing nasal to nasopharyngeal swabs for Staphylococcus aureus (17, 24, 25).

TABLE 2.

Cohen’s kappa values for studies included in the meta-analysis^a

Author	Pathogen(s)	Self-swab site	HCW swab site	No. of self+/HCW− results	No. of self−/HCW+ results	Cohen's kappa value
Dhiman	Influenza A and B	Midturbinate	Midturbinate	1	2	0.881
Esposito	Influenza A and B	Midturbinate	Midturbinate	4	3	0.8571
Goyal	Influenza A and B	Anterior nares	Anterior nares	0	1	0.9292
Goyal	Influenza A and B	Anterior nares	NP	0	2	0.8667
Fragoso	Group A Streptococcus	Pharyngeal	Pharyngeal	0	25	0.4713
Katz	Group A Streptococcus	Pharyngeal	Pharyngeal	0	1	0.9802
Murray	Group A Streptococcus	Pharyngeal	Pharyngeal	14	10	0.8667
Kojima	SARS-CoV-2	Nasal	NP	4	2	0.7214
Kojima	SARS-CoV-2	UOF	NP	4	8	0.4579
Kojima	SARS-CoV-2	SOF	NP	6	3	0.5986
Wehrhahn	SARS-CoV-2 (center 1)	Nasal and throat	Nasal and throat	0	0	1
Wehrhahn	SARS-CoV-2 (center 2)	Nasal and throat	NP and throat	1	0	0.9565
Yuan-Po Tu	SARS-CoV-2	Tongue	NP	2	5	0.9186
Yuan-Po Tu	SARS-CoV-2	Nasal	NP	1	3	0.9547
Yuan-Po Tu	SARS-CoV-2	Midturbinate	NP	0	2	0.9782
Dangerfield	C. trachomatis and N. gonorrhoeae	Pharyngeal	Pharyngeal	0	0	1
Freeman	C. trachomatis and N. gonorrhoeae	Pharyngeal	Pharyngeal	15	4	0.7599
Akmatov	Respiratory pathogens	Nasal	Nasal	4	1	0.8487
Larios	Respiratory pathogens	Midturbinate	NP	4	3	0.7935
Vargas	Respiratory pathogens	Nasal	Nasal	0	3	0.7914
van Cleef	Staphylococcus aureus	Nasal	Nasal	6	1	0.847
van Cleef	Staphylococcus aureus	Pharyngeal	Pharyngeal	13	5	0.5982

^aSARS-CoV-2, severe acute respiratory syndrome coronavirus 2; NP, nasopharyngeal; self⁺/HCW⁻, positive test by self-swab with negative test by HCW swab; self⁻/HCW⁺, negative test by self-swab with positive test by HCW swab; UOF, unsupervised oral fluid; SOF, supervised oral fluid.

DISCUSSION

Our systematic review and meta-analysis of the diagnostic performance of self- or caregiver-collected swabs for detection of viral and bacterial pathogens across a range of upper airways sites support their use beyond influenza. We found a pooled sensitivity of 91% and specificity of 98% for self- or caregiver-collected upper airway swabs compared to HCW swabs, with high agreement despite swab collection from different anatomical sites. Findings were similar when only SARS-CoV-2 data were considered. This level of diagnostic performance should reassure clinicians, researchers, and public health officials that diagnostic performance is not necessarily compromised by self- or caregiver swabbing.

Three studies reported much lower sensitivity and/or specificity than the pooled results. In the studies by Fragoso et al. (1989), Kojima et al. (2020), and Goyal (2017), the sensitivity of self- or caregiver swabs was less than 80% (10, 17, 24). In the 1989 study, parents had to collect swabs and then inoculate an agar plate for culture so that deficient culture preparation may have explained the low sensitivity of 68% compared to HCW swabs (17). Other studies with parent-collected pharyngeal swabs reported much higher sensitivity. Unsupervised oral fluid swabs had low sensitivity (65%) in the study by Kojima et al., although supervised oral fluid swabs (87%) and nasal swabs (83%) had higher sensitivity, suggesting the unsupervised fluid swabs were poorly collected (24). In two other studies, including unsupervised swab collection at home, sensitivity was 88% and 96% (10, 26). The Goyal study had a low number of participants test positive for influenza with only 9 cases out of 127 participants; therefore, any discrepancy in results between the paired samples manifests as a large difference in sensitivity. Sensitivity of the self-collected swabs was 88% when comparing self-collected nasal swabs to HCW-collected nasal swabs, with only a single set of paired swabs discordant. There were two sets of discordant results when comparing self-collected nasal swabs to HCW-collected nasopharyngeal swabs, making the sensitivity 78%.

While our review and meta-analysis did not aim to consider acceptability of self- or caregiver-collected swabs, this was reported in many of the studies. Where participants were questioned regarding relative comfort of swabs, self-swabbing was consistently rated highly (8, 10, 15, 16, 23, 26, 27). In several studies, participants completed questionnaires and indicated a preference for self-collection of swabs (8, 15, 23). Parents also indicated that their children were less distressed when swabs were collected by parents (9).

Strengths of our review include the comprehensive search strategy, inclusion of databases listing very recent preprint papers with high contemporary relevance, and screening by two reviewers. There were also limitations to our study. Few studies were eligible for inclusion and covered a limited range of pathogens. Also, the pragmatic inclusion criteria meant that some of the comparisons between self- and caregiver-collected swabs were from different sites within the upper airways from the HCW swabs. Nonetheless, this reflects the reality that a deep nasopharyngeal swab is not a feasible sample type for self- or caregiver collection. Another limitation was the heterogeneity in testing methodology between the different studies, but as they were testing for different pathogens, this was unavoidable. Subgroup analyses for most pathogens or for adults compared to children were not possible. As for many studies comparing diagnostic tests, our assumption was that HCW swabs are the gold standard. For 10 of 22 comparisons between self- or caregiver and HCW swabs in the meta-analysis, there were more positive self-swabs with corresponding negative HCW swabs than negative self-swabs with positive HCW swabs. For the pathogens and testing methods studied, laboratory-level false positives are unlikely, so these discordant results were most probably due to deficient swabbing by HCWs than patients or caregivers.

The literature on self-collection of diagnostic samples is most robust in the field of sexual health. A 2015 systematic review and meta-analysis of self-collection of vaginal swabs for diagnosis of chlamydia and gonorrhea by Lunny et al. showed high sensitivity of self-collection of diagnostic swabs (92 to 99%) (28). In the context of the coronavirus pandemic, a recent interrupted time series study found acceptable performance characteristics for self-collected throat swabs after changing from clinician-collected swabs for oropharyngeal gonorrhea and chlamydia screening (29). While there are many benefits of self- or caregiver collection of swabs in clinical and research settings, there are relatively few publications describing their use outside sexual health. Our review shows that self- or caregiver collection has sufficient diagnostic performance to justify use when sampling from the upper airways, even when a nasopharyngeal swab is the standard HCW method. In mass testing campaigns, offering self-collection of swabs may increase uptake, minimize infectious exposures to the public and HCWs, and reduce the health care burden of testing for individuals attending for no other reason, including conservation of stocks of personal protective equipment. In the research setting, self-collection may improve recruitment and reduce burdens for researchers and participants, enabling participation at a distance without the need for repeated travel to attend the study center. This reduced commitment could allow for more intensive collection schedules, with more swabs across more time points for each participant. Self- or caregiver-collected swabs can contribute to surveillance efforts for conditions rarely requiring hospitalization.

In conclusion, self- and caregiver collection of upper airway swabs have reassuring diagnostic performance for multiple pathogens, with multiple potential advantages for individuals, families, researchers, and health systems over swabs collected by HCWs. Swabs by HCWs should not be automatically assumed to be superior to swabs collected by individuals or their caregivers. Future research efforts should focus on optimizing implementation of self- and caregiver collection of upper airway swabs.