Diagnostic accuracy and feasibility of depression screening in spinal cord injury: A systematic review

Abstract

Context: Individuals with spinal cord injury or disease (SCI/D) are at increased risk of depression, which is associated with poor short- and long-term outcomes. Accurate diagnosis is complicated by overlapping symptoms of both conditions, and a lack of consensus-derived guidelines specifying an appropriate depression screening tool.

Objective: To conduct a systematic review to: (1) identify the diagnostic accuracy of established depression screening tools compared to clinical assessment; and, (2) to summarize factors that influence feasibility of clinical implementation among adults with SCI/D.

Methods: A systematic search using MEDLINE, EMBASE, PsycINFO, CINAHL and the Cochrane databases using the terms spinal cord injury, depression or mood disorder, and screening or diagnosis identified 1254 initial results. Following duplicate screening, five articles assessing eight screening tools met the final inclusion and exclusion criteria. Measures of diagnostic accuracy and feasibility of implementation were extracted. The Quality Assessment Tool for Diagnostic Accuracy Studies 2 (QUADAS-2) was used to assess study quality.

Results: The Patient Health Questionnaire-9 (PHQ-9) had the highest sensitivity (100%), and specificity (84%). The 2-item version, the PHQ-2, comprised the fewest questions, and six of the eight tools were available without cost. Utilizing the QUADAS-2 tool, risk of bias was rated as low or unclear risk for all studies; applicability of the results was rated as low concern.

Conclusion: The PHQ-9 is an accurate and feasible tool for depression screening in the adult SCI/D population. Future studies should evaluate the implementation of screening tools and the impact of screening on access to mental health interventions.

List of Abbreviations

BDI

Beck Depression Inventory

BSI

Brief Symptom Inventory

CES-D

Center for Epidemiologic Studies Depression Scale

CI

Confidence interval

DASS-21

Depression Anxiety Stress Scales-21

MINI

Mini International Neuropsychiatric Interview

OAHMQ

Older Adult Health and Mood Questionnaire

PHQ-2

Patient Health Questionnaire-2

PHQ-9

Patient Health Questionnaire-9

QUADAS-2

Quality Assessment of Diagnostic Accuracy Studies-2

SCI/D

Spinal cord injury and disease

SCID

Structured Clinical Interview for DSM

SRS

Zung Self-Rating Depression Scale

TDI

Thai Depression Inventory

Introduction

Depression is a well-established psychological condition affecting adults with spinal cord injury and disease (SCI/D). A recent meta-analysis estimates that the prevalence rates for depression in this population is between 19% and 26%.¹ During the rehabilitation phase, depression is associated with longer lengths of stay and fewer functional improvements.^2,3 Once in the community, individuals with SCI/D and depression have less social integration, increased medical complications and greater medical expenses compared to their non-depressed counterparts.^4–6

Despite these negative consequences, individuals with SCI/D and depressive symptoms are thought to be undertreated. In a cross-sectional survey of community-residing individuals, only 29% of patients with probable depression were taking an antidepressant, with only 17% receiving guideline-level pharmacotherapy or psychotherapy.⁷ While individuals in this population often have multiple risk factors hindering their response to treatment, including increased medical co-morbidities and chronic pain, depressive symptoms can be improved. Studies focused on depression management in adults with SCI/D have reported benefits of Venlafaxine, cognitive behavioral therapy, and physical activity.^8–10 Clearly, identifying patients at risk for depression has the potential to guide appropriate diagnosis and ensure access to effective treatment in order to improve symptomatology.

The negative impacts of depression have led to interest in the use of established screening tools to streamline the diagnostic process. The challenge lies in making a diagnosis of depression in a population where symptoms of energy loss, anhedonia and sleep disruption are common.^11,12 There is conflicting literature surrounding whether attributing these symptoms to depression over-estimates the diagnosis.^13–15 However, ignoring somatic symptoms altogether potentially results in overlooking a serious condition.¹⁶ For this reason, it is critical that any screening tool intended for use in the SCI/D population be assessed for its diagnostic accuracy prior to deployment.

Many factors play a role in deciding if a screening tool is appropriate for use in a specific population. Clinical utility is broadly defined as the usefulness of an intervention for clinical practice, which encompasses measures of effectiveness; however, it is often narrowed down to sensitivity and specificity in the literature.¹⁷ The majority of studies focused on depression screening tools assess measures of effectiveness, including validity and reliability, by psychometric analysis.^12,18 Validity refers to how a tool measures the domain of intent, which can be described in many ways including content, construct, and criterion validity.¹⁸ Criterion validity includes measures of diagnostic accuracy, mainly sensitivity (true positive rate) and specificity (true negative rate) of an index test when compared to a reference test. The reliability of a tool refers to its consistency; typically, assessed using parameters of internal consistency (Cronbach’s α) and test-retest reliability.¹⁸ With this array of psychometric variables, it can be challenging to determine which are most important when deciding on an appropriate tool. A literature synthesis suggests that the validity and reliability measures of depression screening tools in the general population are often similar.¹⁹ Beyond these measures, for a test to be clinically useful, it should also be brief, self-administered, multipurpose (used for screening, symptom severity and response to treatment), accessible, and easy to score.²⁰

Two reviews of depression screening tools in the SCI/D population have been published in the past decade. Kalpakjian et al.¹⁸ identified 24 studies that assessed the psychometric properties of seven depression screening tools. They concluded that there was insufficient evidence to recommend the use of one screening tool over others. Similarly, Sakakibara et al.¹² reviewed the psychometric properties of 13 depression and anxiety measurement tools. They were also unable to recommend one tool over another due to comparable performance and suggest that clinicians consider the specific purpose and setting when selecting a screening tool. To date, few studies have assessed diagnostic accuracy by comparing the results to a reference test, or “gold standard”, specifically clinical assessment by a mental health professional. Access to mental health professionals can be challenging in the community due to long wait lists and financial barriers are prevalent. An accurate and easily administered screening tool would therefore help triage which patients require more immediate assessment and follow up. The objective of this systematic review was to identify the diagnostic accuracy of depression screening tools compared to clinical assessment in the SCI/D population and to summarize features of these tools that impact the feasibility of implementation in a clinical setting.

Methods

The review protocol was registered in PROSERO (CRD42017069334, July 2017) and Covidence software was used for the paper selection process.²¹ The process has been reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.²²

Identifying studies/search strategy

Searches were conducted in Medline, EMBASE, PsycINFO, CINAHL and the Cochrane databases using the search terms “spinal cord injury” or “spinal cord disease”, “depression” or “mood disorders”, and “screening” or “diagnosis” with subject heading variations specific to each database. Limits were set to publications available in English and focused on adults. Studies were imported into the Covidence software where duplicates were removed. Two reviewers (J.L. and R.T.) independently screened titles and abstracts in the first phase of the reviews and any conflicts were automatically added to the full text review. Full text review and reasons for exclusion were also assessed independently. Conflicts were to be addressed with a third-party reviewer if necessary (B.C.C.). References for the included studies were additionally reviewed to ensure no other relevant studies had been missed in the search strategy.

Inclusion and exclusion criteria

Studies were included in the final phase of the review process if they were original publications available in English and assessed adult (over 18 years of age) SCI/D populations using a screening tool, which included depression as an outcome measure. Studies were excluded if the screening tool was not compared to the gold standard of clinical assessment for depression, although the type of clinical assessment was not specified.

Data collection

Data was extracted independently by each reviewer and compared. Primary outcome measures included the index screening tool, the reference test (clinical assessment), and the sensitivity and specificity of the index test compared to the reference test. If not provided in the original article, 95% Clopper–Pearson confidence intervals for sensitivity and specificity were calculated. Measures of implementation feasibility included number of survey items, time for administration, method of scoring, and cost. If not available from the original article, this information was accessed from the screening tool publisher’s summary. Both reviewers independently assessed study quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool.²³ Any discrepancies were discussed, and the definitions of quality criteria were clarified as suggested by the Cochrane Training protocol.²⁴ Should a consensus not be reached, a third party was available for involvement. A secondary search was completed to abstract the available psychometric properties of the final tools identified. Lastly, a repeat literature search was conducted within one month of manuscript submission to ensure no additional literature was available.

Results

From the initial search 1,254 articles were identified, with 1,024 remaining for the initial screen after removing duplicates (Fig. 1). The review of titles and abstracts revealed 28 articles for full text review. A total of 23 studies were excluded, leaving 5 studies to be included for data extraction. Of the studies excluded in this phase, 12 did not compare the index test to clinical assessment. The majority of these studies assessed psychometric properties of the screening tool, comparing outcomes to either depression, quality of life or functional questionnaires. Ten studies were excluded for incorrect study design, including either systematic reviews or editorials. One study was excluded as it focused on depression in musculoskeletal spine disorders. Two additional studies were found during reference review; however, after independent full text screening neither met the inclusion and exclusion criteria. There were no conflicts identified between the reviewers regarding final study selection.

Figure 1.

PRISMA flow diagram of study selection process.

Screening tools

From the five studies identified, eight depression screening tools were assessed as index tests (Table 1). This included the Beck Depression Inventory (BDI),^25,26 the Brief Symptom Inventory (BSI),^27–29 Center for Epidemiologic Studies Depression Scale (CES-D)³⁰, Depression Anxiety Stress Scales-21 (DASS-21),²⁷ Patient Health Questionnaire-2 and -9 (PHQ-2, PHQ-9),³¹ Zung Self-Rating Depression Scale (SRS)²⁸ and Thai Depression Inventory (TDI).³⁰ The reported psychometric properties for each tool is shown in Table 2; data for the PHQ-2 or TDI were not available for review.

Table 1. Summary of identified depression screening tools and their characteristics.

	Length	Time to administer	Scoring	Cost (USD)
BDI	21 items	5 min	Summed score	$142.50*
BSI	53 items	8–10 min	Summed score	$136.75*
CES-D	20 items	5–10 min	Summed score	Free
DASS-21	21 items	<10 min	Summed score	Free
PHQ-2	2 items	Not available	Summed score	Free
PHQ-9	9 items	<5 min	Summed score	Free
SDS	20 items	10 min	Score key required	Free
TDI	20 items	Not available	Summed score	Free

*For manual scoring starter kit.^25,28

Note: BDI, Beck Depression Inventory; BSI, Brief Symptom Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; DASS-21, Depression Anxiety Stress Scales; PHQ, Patient Health Questionnaire; SDS, Zung Self-Rating Depression Scale; TDI, Thai Depression Inventory.

Table 2. Summary of psychometric properties of identified screening tools.

Tool	Internal consistency (Cronbach’s α)	Concurrent validity (r)	Construct validity (r)
BDI	+++25,32	—	Pessimism (0.76)33 State Anxiety Inventory (0.82)34 Suicidal ideation (0.69)33
BSI	+++28	DASS-21 Depression Subscale (0.70)27 SDS (0.52)28	—
CES-D	++ to +++10,30,35–44	—	Perceived control (−0.56)44 Perceived health (−.49)44,45 Perceived Stress Scale (0.645-0.713)43 Satisfaction with Life Scale (−0.65)43 Medical Outcomes Study Short Form-36 Mental Health Subscale (0.75)36 Short-Form State-Trait Anxiety Inventory (0.80)43
DASS-21 Depression	+++46,47	BSI Depression Subscale (0.70)27	—
PHQ-9	+++48–50	Hamilton Depression Rating Scale (0.68)50 Hopkins Symptom Checklist-20 (0.78)50 OAHMQ (0.78)51 SCID (0.79)50	Satisfaction with Life Scale (−0.51–−0.35)48,49,51 Subjective Health (−0.50)48 Greater Difficulty with Daily Functioning (0.37-0.62)31,48 Medical Outcomes Study Short Form-1 (0.37)31 European Quality of Life Current Health State Thermometer (−0.36)31 Life-1(−0.38)31 Patient Global Impression – Improvement Scale (0.58)50 Clinical Global Impression – Improvement Scale (0.66)50 Maier Subscale (0.67)50 Bech Subscale (0.69)50
SDS	+++28	BSI Depression Subscale (0.52)28 Ilfeld Psychiatric Symptom Index (0.72)52	—

+ poor ++ adequate +++ excellent.⁵³

— No data available.

Note: BDI, Beck Depression Inventory; BSI, Brief Symptom Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; DASS-21, Depression Anxiety Stress Scales; OAHMQ, Older Adult Health and Mood Questionnaire; PHQ, Patient Health Questionnaire; SCID, Structured Clinical Interview for DSM; SDS, Zung Self-Rating Depression Scale.

Clinical assessments

Reference tests were comprised of a clinical interview, with either with a physician, psychologist, social worker or trained research associate. The studies by Bombardier et al.,³¹ Mitchell et al.,²⁷ and Radnitz et al.²⁵ used a more structured clinical assessment, either the Structured Clinical Interview for DSM (SCID) or the Mini International Neuropsychiatric Interview (MINI). Radnitz et al.²⁵ did not use a mental health professional for administration, but a trained research associate. They outlined the training program for the research associate, which included observed interviews and ensured good inter-rater reliability. Tate et al.²⁸ and Kuptniratsaikul et al.³⁰ did not specify the type of clinical assessment beyond a “clinical interview”.

Diagnostic accuracy

The diagnostic accuracy data from the included studies are summarized in Table 3. The PHQ-2 and PHQ-9 had the highest sensitivity of 1.00 (95% CI 0.78–1.00 for both), while the BSI total score had the lowest at 0.29 (95% CI 0.04–0.71). The BDI, scored with a cut-off of 27, had the highest specificity at 1.00 (95% CI 0.97–1.00), while the PHQ-2 had the lowest specificity at 0.48 (95% CI 0.39–0.56). Assessing the tests individually, the PHQ-9 had the highest combined sensitivity and specificity (1.00 (95% CI 0.78–1.00) and 0.84 (95% CI 0.76–0.89) respectively). Predictive values were only reported in the studies by Bombardier et al.³¹ and Kuptniratsaikul et al.³⁰ Notably, the negative predictive value of both the PHQ-9 and PHQ-2 were 1. Additionally, the PHQ-9 had the highest area under the curve of 0.92, when compared to the CES-D and TDI.

Table 3. Study parameters and diagnostic accuracy indices of the included studies.

Study	Participants (n)	Setting	Reference Test	Index Test	Sensitivity (95% CI)	Specificity (95% CI)
Bombardier et al. 201231	142	Inpatient rehabilitation	SCID, DSM-IV	PHQ-9	1.00 (0.78–1.00)	0.84 (0.76–0.89)
				PHQ-2	1.00 (0.78–1.00)	0.48 (0.39–0.56)
Kuptniratsaikul et al. 200230	83	Not specified	Psychiatric interview, DSM-IV	CES-D	0.80 (0.56–0.93)	0.69 (0.57–0.80)
				TDI	0.70 (0.56–0.93)	0.79 (0.67–0.88)
Mitchell et al. 200827	40	Inpatient and outpatient	MINI, DSM-IV	DASS-21	0.57 (0.18–0.90)	0.76 (0.58–0.89)
				BSI	0.57 (0.18–0.90)	0.82 (0.65–0.93)
Radnitz et al.199725	124	Inpatient and outpatient	SCID, DSM-III-R	BDI – Cut off 27	0.50 (0.21–0.79)	1.00 (0.97–1.00)
				BDI – Cut off 18	0.83 (0.52–0.98)	0.91 (0.85–0.96)
Tate et al. 199328	30	Inpatient	Clinical interview, DSM-III-R	BSI total score	0.29 (0.04–0.71)	0.73 (0.52–0.90)
				BSI depression subscale	0.57 (0.18–0.90)	0.87 (0.66–0.97)
				SDS	0.86 (0.42–1.00)	0.61 (0.39–0.80)

Note: BDI, Beck Depression Inventory; BSI, Brief Symptom Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; CI, Confidence interval; DASS-21, Depression Anxiety Stress Scales; DSM, Diagnostic and Statistical Manual of Mental Disorders; MINI, Mini International Neuropsychiatric Interview; PHQ, Patient Health Questionnaire; SCID, Structured Clinical Interview for DSM; SDS, Zung Self-Rating Depression Scale; TDI, Thai Depression Inventory

Feasibility of implementation

These tools ranged in length from a 2-item questionnaire (PHQ-2) to a battery of 53 questions (BSI), although all were reported to take less than 10 min to complete. Scoring for all tests was done by summation of a Likert scale score, with exception for the SDS, which used a reverse scoring system for positively associated question items, requiring the use of a scoring key. All tools, excluding the BDI and BSI, were available free of charge on public domain websites.

Risk of bias assessment

Results from the risk assessment using the QUADAS-2 tool are summarized in Fig. 2. Overall, risk of bias in all studies was either low or unclear. Only the study by Bombardier et al.³¹ specified the timeframe within which the index test and reference test were administered. This was unclear in the other four studies resulting in an unclear risk of bias in the flow and timing category. The study by Kuptniratsaikul et al.³⁰ did not specify the criteria for patient selection, resulting in an unclear risk of bias and unclear applicability of the results for the patient selection category. Two studies did not specify whether the clinicians completing the reference test were blinded to the screening tool results, earning an unclear risk of bias in the reference standard category.^25,30 Lastly, the study by Tate et al.²⁸ used a clinical interview by a clinician in the patient’s care team, resulting in a high risk of bias for the reference standard category.

Figure 2.

Summary of risk of bias assessment using the QUADAS-2 tool.

Discussion

The present review was conducted to assess the diagnostic accuracy of established screening tools for depression in SCI/D when compared to clinical assessment, in addition to measures of feasibility of implementation. Although abundant literature exists on depression measurement tools in SCI/D, few studies actually assessed diagnostic accuracy.^12,18 From the articles identified in the systematic review, the PHQ-9 had the highest combined sensitivity and specificity, with a low burden of administration. This tool has the added benefit of a 2-step process utilizing the PHQ-2 as an initial screen with excellent sensitivity. Bombardier et al. propose that screening begins with the PHQ-2 items, and if scored one or greater then proceeding to the remaining PHQ-9 items.³¹ A recent study used archival data of veterans with SCI/D who had completed both the PHQ-2 and PHQ-9 to determine the accuracy of different cut-off scores.⁵⁴ In this sample, a cut-off score of 3 or greater performed best, correctly classifying 94.8% of cases, although with a lower sensitivity of 0.83. Multiple cut-off values may be reasonable depending on the goal of screening. If the clinical priority were to minimize false negatives, a lower cut-off score would be utilized. If efficiency were a major concern, a higher cut-off score would reduce the number of false positives while saving time on completing the entire questionnaire.

Direct comparisons of screening tool psychometric properties often show comparable outcomes, moving the focus of tool selection to implementation feasibility.^12,50 In addition to containing few items, and having a simple scoring method, the PHQ-9 is available on a public domain website in multiple languages reducing the need for administrator interpretation. PHQ-9 has also been validated for telephone administration in the general population.⁵⁵ Lastly, the PHQ-9 has been accepted into practice in multiple rehabilitation settings, including stroke and brain injury populations.^56,57 These factors add to the ease of implementation in neurological rehabilitation centers, further enhancing its clinical utility.

Limitations

A major limitation of the current review lies in the study design utilized to address diagnostic accuracy. Articles that did not directly compare the screening tool to a clinical assessment were excluded. Other screening tools have been reported to have good measures of reliability and validity in the SCI/D population, including the General Heath Questionnaire,⁵⁸ the Hospital Anxiety and Depression Scale,⁵⁹ the Inventory to Diagnose Depression¹³ and the Older Adult Health and Mood Questionnaire (OAHMQ).^60,61 Many of these tools have been used in studies of disease prevalence and treatment efficacy.^8,9,62 While these tools may have their role in certain study designs, this review demonstrates that the diagnostic accuracy has not been established, and their use in the setting of depression screening should be cautioned. Another consideration is the clinical gold standard to confirm the presence of depression to which each tool was compared. Tate et al.²⁸ and Kuptniratsaikul et al.³⁰ did not specify the type of clinical assessment used. Radnitz et al.²⁵ used a trained research associate as opposed to a mental health professional. They outlined their method of standardization, including ensuring test re-test reliability, however one could question if this rigorous enough to qualify as a gold standard.

Another limitation of this review is the focus solely on depression. It is well known that other psychiatric conditions are prevalent alone or in combination among individuals with SCI/D including anxiety disorders, post-traumatic stress disorder (PTSD) and substance abuse.⁷ A recent meta-analysis found a prevalence of anxiety diagnoses ranging from 15% to 32%, which is also significantly higher when compared to the general population.⁶³ Additionally, anxiety and depression are often co-morbid conditions. A study of veterans with SCI/D and depression, found 70% had co-morbid psychiatric diagnoses, most commonly PTSD (21%) or another anxiety disorder (13%).⁶⁴ To truly address the complex psychiatric needs of this population, appropriate ways to screen for all common concurrent diagnoses should be investigated.

The patient demographics in the studies identified may limit the generalizability of the findings, as they consist predominantly of Caucasian men with traumatic spinal cord injuries. Studies frequently show that women and minorities, are at increased risk of depression.^46,65 Some studies suggest that individuals with a non-traumatic etiology of SCI/D are at increased risk of depression, while others do not.^46,66,67 It is even less clear how gender, ethnicity and etiology of injury impacts screening tool selection. In a direct comparison of participants with depressive symptoms utilizing the PHQ-9 and the OAHMQ, White and Non-White population symptom scores were consistent using the PHQ-9.⁵¹ However, significantly more Non-White participants reported depressive symptoms when utilizing the OAHMQ. In primary care populations, PHQ-9 scores are not dependent on gender or ethnicity, although individuals over 55 years of age were more likely to endorse feelings of anhedonia and low mood.^68,69 This would suggest that while these demographic factors may impact disease prevalence, they are less likely to influence the diagnostic accuracy of screening tools.

Risk of bias was assessed using the QUADAS-2 tool and was overall low to unclear. One particular source of bias specific to studies of diagnostic accuracy is spectrum bias. This is a form of sampling bias where performance of a diagnostic test may vary in clinical settings due to a different distribution of patients.⁷⁰ Thombs et al.⁷¹ suggest that including individuals with diagnosed depression or who are receiving depression treatment artificially inflates the diagnostic accuracy of screening tools. Only the studies by Tate et al.²⁸ and Radnitz et al.²⁵ commented on exclusion of those with psychiatric diagnoses. Bombardier et al.³¹ excluded patients with psychosis only if their ability to complete assessments was impaired. None of the studies commented on whether patients involved were currently receiving treatment for depressive symptoms. It is therefore possible that these measures of diagnostic accuracy are inflated, which should be taken into consideration during clinical implementation.

Lastly, while there is extensive research into the clinical utility of screening tools in SCI/D, it is unclear whether the implementation of routine screening improves identification, treatment or the overall emotional wellbeing of individuals. The recommendations for depression screening in the general population are mixed. The Canadian Task Force of Preventative Health Care advises against routine screening for both adults at average risk, or subgroups with increased risk of depression.⁷² This recommendation is based on the lack of evidence of benefit versus harm. They note the need for further research on the benefits of implementing screening, particularly in high risk groups like the SCI/D population. Conversely, a more recent publication from the US Preventative Services Task Force Recommendation Statement is supportive for screening for depression in adults with “high certainty that the net benefit is moderate or there is moderate certainty that the net benefit is moderate to substantial”.⁷³ It is qualified that screening should be implemented with “adequate systems in place to ensure accurate diagnosis, effective treatment, and appropriate follow-up”.⁷³ Further research should be focused on better understanding of how depression screening may impacts a patients’ mental health service provision and mental health outcomes.

Conclusions

This systematic review highlights that few screening tools have formally been validated for diagnostic accuracy compared to clinical assessment in SCI/D. Of the available screening tools, the PHQ-9 has high clinical utility in both its diagnostic accuracy and feasibility for implementation. Clinicians should note that anxiety disorders are also common and co-morbid in this population and anxiety symptoms would not be assessed using this tool. Further research into accurate and feasible screening processes for co-morbid conditions such as PTSD and generalized anxiety disorder would be valuable. Most importantly, further investigation into the implementation and benefits of routine depression screening in SCI/D is critical to advancing emotional wellbeing following SCI/D and preventing the adverse consequences of untreated depression on health and community participation.

Acknowledgements

The authors wish to acknowledge Dr. Mohammad Alavinia for his guidance on methodology and manuscript preparation and Ms. Maureen Pakosh for her assistance with the literature review.

Disclaimer statements

Contributors: All authors participated in the development, writing and review of the manuscript.

Conflicts of interest: There are no disclosures or conflicts of interest to report.

Funding: None.

ORCID

Rebecca Titman http://orcid.org/0000-0001-9950-744X

Jason Liang http://orcid.org/0000-0001-8648-7785

B Catharine Craven http://orcid.org/0000-0001-8234-6803