Two-step screening for depressive symptoms in patients treated with kidney replacement therapies: a cross-sectional analysis

ABSTRACT

Background

Systematic screening for depressive symptoms may identify patients who may benefit from clinical assessment and psychosocial support. Here we assess a two-step screening using ultrabrief pre-screeners [Edmonton Symptom Assessment Survey–revised Depression item (ESASr-D) or Patient Health Questionnaire-2 (PHQ-2)] followed by the Patient-Reported Outcomes Measurement Information System Depression questionnaire (PROMIS-D) to identify depressive symptoms in patients on kidney replacement therapies.

Methods

We conducted a cross-sectional study of adults (kidney transplant recipients or treated with dialysis) in Toronto, ON, Canada. We simulated various two-step screening scenarios where only patients above a pre-screening cut-off score on the ESASr-D or PHQ-2 would move to step 2 (PROMIS-D). Screening performance was evaluated by sensitivity, specificity and positive and negative predictive values using the Patient Health Questionnaire-9 (PHQ-9) as the referent. The average number of items completed by patients in different scenarios was reported.

Results

Of 480 participants, 60% were male with a mean age of 55 years. Based on PHQ-9, 19% of patients had moderate or severe depressive symptoms. Pre-screening with a PHQ-2 score ≥1 combined with a PROMIS-D score of ≥53 provided the best two-step results (sensitivity 0.81, specificity 0.84, NPV 0.95). Two-step screening also reduces question burden.

Conclusions

A two-step screening using a PHQ-2 score ≥1 followed by a PROMIS-D score ≥53 has good sensitivity and specificity for identifying potentially significant depressive symptoms among patients on kidney replacement therapies. This approach has lower question burden. Screened-in patients will need further clinical assessment to establish a diagnosis.

Graphical Abstract

Graphical Abstract.

KEY LEARNING POINTS.

What is already known about this subject?

• Patients treated with kidney replacement therapies (KRTs) experience emotional distress that is currently not systematically monitored and managed.

• The high symptom burden and emotional distress experienced by these patients impact their quality of life and are associated with poor outcomes.

• Systematic screening for depressive symptoms may identify patients on KRTs who may benefit from clinical assessment and psychosocial support. However, there is no consensus on the optimal screening tool.

What this study adds?

• This study reports the sensitivity and specificity of a two-step screening approach to screen for depressive symptoms in patients on KRTs.

• We report that the two-step screening approach using a PHQ-2 score ≥1 followed by a PROMIS-D score ≥53 has good sensitivity and specificity for identifying potentially significant depressive symptoms among patients on KRTs.

• This approach has lower question burden compared with using the PROMIS-D or PHQ-9 alone.

What impact this may have on practice or policy?

• The two-step screening could deliver efficient and feasible screening with reduced question burden, especially among patients with no or a low level of symptoms.

• Clinical assessment following a positive screen will be necessary to establish the diagnosis and determine the appropriate support or treatment.

INTRODUCTION

Depressive symptoms are common among patients with kidney failure treated with kidney replacement therapies (KRTs; dialysis or kidney transplant) [1–3] and are associated with poor outcomes [4, 5]. However, depressive symptoms are not assessed systematically in these patients, therefore they often remain unrecognized and untreated [6, 7]. Screening for depressive symptoms is recommended by the US Preventive Services Task Force [8], although some professional organizations do not support that recommendation [9]. Nevertheless, for patients with cancer, such screening is mandated internationally [10, 11].

In the USA, the Centers for Medicare and Medicaid Services requires depression screening for patients with kidney failure [12]. However, there is no consensus about the optimal screening tools [13]. Defining the optimal cut-off score is also challenging. A sensitive lower cut-off may yield many false positive cases, which could overwhelm the assessment system. A higher cut-off is more specific, but may miss patients who would benefit from further assessment.

Ultrabrief screening tools, such as the single depression item from the Edmonton Symptom Assessment Survey-revised Depression item (ESASr-D) or the two-item Patient Health Questionnaire (PHQ-2), minimize question burden [14, 15]. However, they yield a crude score, where a 1-point change may result in large changes in sensitivity and specificity, making ‘fine-tuning’ of the cut-off difficult. Furthermore, their reliability may be suboptimal [16].

Longer questionnaires, such as the 9-item Patient Health Questionnaire (PHQ-9) and 21-item Beck Depression Inventory, are more precise [13], but include questions that are not relevant for patients with no or minimal depressive symptoms. This and the question burden may result in respondent fatigue and low completion rates and may not be feasible for regular clinical use.

The Patient-Reported Outcomes Measurement Information System Depression (PROMIS-D) item bank offers reliable and efficient measurement of depressive symptoms [17]. It has been validated in kidney transplant recipients and patients on dialysis and can be administered using fixed length short forms (four, six or eight items) or computerized adaptive testing (CAT) [18–20]. The raw scores obtained with the different versions can be transformed to a T-score metric, which ensures comparability. PROMIS tools have been increasingly integrated into electronic patient records, making them an appealing screening option [21]. CAT offers the advantage of administering only the most relevant and informative items to respondents based on their answers to previous items. CAT typically yields precise assessment (>90% reliability) with the administration of only four to six items [22]. However, due to the specifications of the current CAT stopping rule, some respondents without depressive symptoms may need to answer up to 12 items [23].

A two-step screening approach, which has been used in various patient populations, may optimize sensitivity and specificity while minimizing the question burden [24–26]. In this approach, an ultrabrief pre-screener (ESASr-D or PHQ-2) with a highly sensitive cut-off is followed by a more detailed tool administered only to patients who score above the pre-screener cut-off. Here we compare the screening performance and question burden between administering the PROMIS-D questionnaire alone and such two-step approaches to screen for moderate to severe depressive symptoms in patients treated with KRTs.

MATERIALS AND METHODS

Study design, participant selection and data sources

This was a secondary analysis of data stored in our research database—the COmprehensive Psychosocial REsearch Data System (Co-PreDS-CKD; REB 17-5916). The primary studies contributing data to the database included participants from multiple dialysis units in Toronto who have been on dialysis for ≥90 days. Other studies enrolled participants from the outpatient kidney transplant clinic of the Toronto General Hospital, who had their kidney transplant ≥30 days prior to enrolment. Exclusion criteria for those studies included insufficient English proficiency, diagnosis of dementia and/or severe acute medical conditions, all as established by the managing clinical team. Both self-reported data elements and data extracted from medical records were audited for completeness and accuracy before being entered into the database. Participants selected for this secondary analysis included patients on maintenance dialysis and kidney transplant recipients who completed the ESASr, PHQ-9, and PROMIS-29 or PROMIS-D CAT between September 2017 and April 2020. Participants were excluded if they had missing or ambiguous responses on any of the questionnaires.

For the primary studies, patients were approached during regular clinic visits or dialysis treatment. Participants completed questionnaires on iPads using an electronic data capture system [DAta Driven Outcomes System (DADOS), Techna Institute, University Health Network, Toronto, ON, Canada]. Self-reported sociodemographic characteristics were also collected. Clinical data were abstracted from medical records.

All participants provided written informed consent. The studies were approved by the research ethics boards of all participating sites (REB 15-9645, REB 2016-003, REB 17-0061 and REB 377-2017).

The ESASr assesses nine symptoms (pain, tiredness, drowsiness, nausea, lack of appetite, shortness of breath, depression, anxiety and well-being) using a 10-point Likert scale ranging from 0 (no symptom) to 10 (worst possible symptom) [27]. It has been validated in patients on dialysis and in kidney transplant recipients [14, 28]. An ESASr-D cut off of ≥2 has been recommended to identify clinically relevant depressive symptoms in oncology care and among patients on dialysis [29]. In this analysis, ESASr-D was tested as a pre-screener, with cut-off scores of ≥1 or ≥2 for two-step screening.

The PHQ-9 is a publicly available tool that asks about the extent to which respondents are bothered by depressive symptoms. It contains nine items rated on a scale of 0–3 (0, not at all bothered; 3, bothered nearly every day). The total score (0–27) is a sum of the item scores [15]. PHQ-9 has been recommended by the British Columbia Renal Agency for screening in patients with chronic kidney disease (CKD) [30]. A PHQ-9 cut-off ≥10 is used to identify moderate to severe depressive symptoms. The PHQ-2 questionnaire is the first two items of the PHQ-9. A score ≥2 is suggested to identify moderate to severe depressive symptoms. In our analysis, we used a PHQ-9 score ≥10 to classify patients with moderate to severe depressive symptoms. Moreover, we also tested PHQ-2 scores of ≥1 or ≥2 as an alternative pre-screener tool. Other studies have also assessed the performance of the PHQ-2 against the PHQ-9 [15, 26].

The PROMIS Depression item bank version 1.0 for adults in English includes 28 items that assess self-reported negative mood, views of oneself, social cognition, decreased positive affect and engagement with responses scored on a 5-point Likert scale (1–5) [31]. Depressive symptoms in this study were assessed with either the PROMIS-D 4-item short form (part of the PROMIS-29 profile) or PROMIS-D CAT through an application programming interface. The PROMIS-D CAT administers items based on responses to preceding items using a score estimation algorithm, achieving precise estimates using more relevant items compared with fixed-length questionnaires [32]. Based on the stopping rule of the algorithm, 4–12 items are administered until >0.90 reliability is achieved or a maximum of 12 items are asked [33]. Raw PROMIS scores are transformed to a T-score metric that yields a mean of 50 and a standard deviation (SD) of 10 [34], with the US general population as the reference. Higher PROMIS scores indicate more severe depressive symptoms.

The PROMIS-D T-scores obtained using the PROMIS-D or PROMIS-D CAT were used as the second tool in the two-step screening approaches. Cut-off scores of 0.5 or 1 SD above the mean on the PROMIS T-score metric have been suggested to identify individuals with moderate to severe symptoms [35]. Since a cut-off of 53 may offer higher sensitivity [23], which is potentially advantageous for screening, we used a primary cut-off of 53 in the two-step screening approach. Additionally, we also assessed the cut-off of 55, which has previously been used by others [23, 36].

All participants completed the ESASr, PHQ-9 and PROMIS-29 or PROMIS-D CAT. Based on this dataset, we simulated various two-step screening scenarios, where only patients above the specified cut-off for the ESASr-D or PHQ-2 pre-screener would move to step 2 (PROMIS-D).

Statistical analysis

Baseline characteristics are presented as frequency, mean, SD, median and interquartile range (IQR). Floor and ceiling effects were reported as the proportion of patients with minimum and maximum possible scores, respectively. For the PHQ-9 and PHQ-2, Cronbach's α was computed to assess internal consistency. We assessed group-level reliability for PROMIS-D scores as conceptualized within the item response theory [33]. The standard error of measure (SEM) was reported over the range of PROMIS-D scores. The SEMs were converted to reliability coefficients where reliability is equal to 1 − SEM². Group-level scale reliability was calculated as average reliability = 1 − [mean(SEM)]² [37]. Values >0.90 indicate excellent reliability [38].

Receiver operating characteristics (ROC) analyses were conducted to assess the discrimination of ESASr-D, PHQ-2 and PROMIS-D T-scores against moderate/severe depressive symptoms defined by a PHQ-9 cut-off score ≥10. Test discrimination was characterized by the area under the ROC curve (AUROC), with an AUROC ≥0.75 representing good discrimination [39].

The performance of the ESASr-D, PHQ-2 and PROMIS-D T-scores alone and the performance of the simulated two-step screening scenarios was assessed by calculating the sensitivity, specificity and positive and negative predictive value (PPV and NPV, respectively).

Currently there are no universally accepted screening performance criteria for instruments that screen for depressive symptoms. Based on reports about the sensitivity and specificity of commonly used questionnaires, and considering that high sensitivity and NPV are needed when screening for depressive symptoms, we considered a sensitivity and specificity of 80% as acceptable screening accuracy [13]. This is also consistent with the recommendation that the sum of sensitivity and specificity should be ≥1.5 for a test to be useful [40]. Moreover, we also considered a high NPV of >90% to be acceptable given the importance of reducing the number of false negative patients when screening for depressive symptoms [41].

All statistical analyses were conducted using Stata 14.0 (StataCorp, College Station, TX, USA).

RESULTS

A total of 283 kidney transplant recipients and 197 patients on maintenance dialysis were included in this analysis (Supplementary Fig. S1). The mean age of the entire cohort was 55 years (SD 17) and 60% of participants were male. Based on a PHQ-9 cut-off of ≥10, 89 participants (19%; 26% and 13% for patients on dialysis and kidney transplant recipients, respectively; P < .001) were classified as having moderate to severe depressive symptoms. The baseline characteristics of the study sample are shown in Table 1.

Table 1:

Baseline characteristics of the study sample.

Characteristics	Entire cohort (N = 480)	Dialysis (n = 197)	Transplant (n = 283)	P-value
Male, n (%)	289 (60)	115 (58)	174 (61)	.50
Age (years), mean (SD)	55 (17)	63 (15)	49 (17)	<.001
Ethnicity, n (%)				<.001
White	236 (50)	68 (35)	168 (61)
African, Caribbean and Black	101 (22)	75 (39)	26 (9)
Asian	98 (21)	37 (19)	61 (22)
Other	34 (7)	15 (8)	19 (7)
Education (>12 years), n (%)	252 (55)	81 (43)	171 (62)	<.001
Marital status, n (%)				<.001
Single	119 (26)	44 (23)	75 (27)
Married or common law	245 (53)	76 (40)	169 (61)
Divorced, widowed or separated	102 (22)	70 (37)	32 (12)
Self-reported annual income (CAD/year), n (%)
<$30 000	105 (32)	72 (62)	33 (16)	<.001
$30 000–$70 000	107 (33)	30 (26)	77 (36)
>$70 000	115 (35)	14 (12)	101 (48)
Charlson Comorbidity Index <4, n (%)	269 (59)	69 (39)	200 (71)	<.001
Diabetes (yes), n (%)	172 (36)	100 (51)	72 (25)	<.001
Serum albumin (g/dl), mean (SD)	4.0 (0.5)	3.6 (0.4)	4.2 (0.3)	<.001
Haemoglobin (g/dl), mean (SD)	12.0 (1.9)	11 (1.3)	12.8 (1.9)	<.001
eGFR (ml/min/1.73 m2), median (IQR)			55 (39–74)
Time on dialysis (months), median (IQR)		40 (17–76)
Time since transplant (months), median (IQR)			104 (35–188)
Depression (PHQ-9 ≥10), n (%)	89 (19)	52 (26)	37 (13)	<.001

CAD, Canadian dollars; eGFR, estimated glomerular filtration rate.

The distributions and descriptive statistics for PROMIS-D, ESASr-D, PHQ-2 and PHQ-9 are reported in Supplementary Table S1. The reliability of PHQ-9 and PHQ-2 in this sample was 0.87 and 0.76, respectively. The average reliability of the PROMIS T-scores was 0.86.

All three questionnaires demonstrated good discrimination for moderate to severe depression {AUROC 0.80 [95% confidence interval (CI) 0.75–0.86], 0.92 [95% CI 0.89–0.95] and 0.89 [95% CI 0.85–0.92] for ESASr-D, PHQ-2 and PROMIS-D, respectively} (Fig. 1). For the PHQ-2, a cut-off ≥1 had the highest sensitivity and NPV (sensitivity 0.98, specificity 0.67, PPV 0.40, NPV 0.99) (Table 2). Since the sensitivity of the ESASr-D at a cut off ≥1 (sensitivity 0.75, specificity 0.72, PPV 0.38, NPV 0.93) was below our pre-specified threshold of 0.8, it was not considered for the two-step approach (Table 2). A PROMIS-D score ≥53 (sensitivity 0.82, specificity 0.77, PPV 0.45, NPV 0.95) had an overall better screening performance compared with a cut-off ≥55 (sensitivity 0.73, specificity 0.85, PPV 0.53, NPV 0.93) (Table 2).

Figure 1:

ROC curve of (A) ESASr-D, (B) PHQ-2 and (C) PROMIS-D T-scores against moderate to severe depression defined by PHQ-9 scores in the entire cohort.

Table 2:

Screening characteristics of instruments assessing depressive symptoms using a PHQ-9 score ≥10 to identify potentially clinically significant (moderate to severe) depressive symptoms.

Tool	Cut-off scores	Sensitivity	Specificity	TP	TN	FP	FN	PPV	NPV
ESASr-D	≥1	0.75	0.72	67	283	108	22	0.38	0.93
	≥2	0.71	0.84	63	327	64	26	0.50	0.93
PHQ-2	≥1	0.98	0.67	87	261	130	2	0.40	0.99
	≥2	0.92	0.81	82	315	76	7	0.52	0.98
PROMIS-D	≥53	0.82	0.77	73	301	90	16	0.45	0.95
	≥55	0.73	0.85	65	333	58	24	0.53	0.93
PHQ-2 and PROMIS-D	≥1 and ≥53	0.81	0.84	72	327	64	17	0.53	0.95
	≥2 and ≥53	0.78	0.88	69	346	45	20	0.61	0.95
	≥1 and ≥55	0.72	0.87	64	342	49	25	0.57	0.93
	≥2 and ≥55	0.69	0.91	61	357	34	28	0.64	0.93

TP, true positive; TN, true negative; FP, false positive; FN, false negative; PPV: positive predictive value; NPV: negative predictive value.

Rows in bold indicate screening combinations with both sensitivity and specificity ≥0.80.

Table 2 presents the performance of the two-step screening using the PHQ-2 as a pre-screener followed by a PROMIS-D score ≥53. The highest sensitivity and NPV were seen with a PHQ-2 score ≥1 and a PROMIS-D score ≥53 (sensitivity 0.81, specificity 0.84, PPV 0.53, NPV0.95). Of the 217 patients that moved to step 2, 130 patients were false positives. In the second step, a PROMIS-D score ≥53 correctly reclassified 66 of the false positives. However, 15 true positive patients were incorrectly reclassified in the second step, yielding 17 false negatives overall.

The PROMIS-D cut-off of ≥55 yielded lower sensitivity when used in the two-step screening scenarios, which is not desirable in screening, and was therefore not considered (Table 2).

In this sample, the 480 participants completed a total of 2244 questionnaire items if using the PROMIS-D alone (5 items per person on average; Table 3). Table 3 presents the number of items patients would complete using the two-step screening combination that had the best performance in this study (PHQ-2 ≥1 followed by the PROMIS-D). With this combination, participants completed a total of 1857 items (4 items per person on average; Table 3). In this scenario, 263 patients with PHQ-2 scores <1 were screened out at the first step. They would have completed only 2 PHQ-2 items, 526 for the cohort, instead of 1347 PROMIS-D items (5 items per person on average). The 217 patients with a PHQ-2 score ≥1 would have completed 1331 items (6 items per person on average). This represents a reduction of ∼20% compared with the scenario with the PROMIS-D alone and a reduction of ∼25% compared with the 2479 items resulting from the PHQ-2 ≥1 followed by PHQ-9 two-step approach (i.e. patients scoring ≥1 on the PHQ-2 would go on to complete the full PHQ-9). It also represents a substantial reduction compared with using PHQ-9 for all participants (4320 items).

Table 3:

Number of questionnaire items completed for patients completing PROMIS-D alone versus a two-step method, using a PHQ-2 score ≥1 as the pre-screening cut-off score.

	PROMIS-D alone		PHQ-2 followed by PHQ-9		PHQ-2 followed by PROMIS D
Results	Total number of PROMIS-D items completed	Average number of items completed (total number of items/n)	Number of PHQ-9 items completed	Average number of items completed (total number of items/n)	Number of items completed in step 1	Number of items completed in step 2	Total number of items completed in two-step procedure	Average number of items completed in two-step procedure (total number of items/n)
PHQ-2 cut off >=1
Screened out in step 1 of two-step process (n = 263)	1347	5	526	2	526		526	2
Screened in for second step of two-step process (n = 217)	897	4	1953	9	434	897	1331	6
Total sample (N = 480)	2244	5	2479	5	960	897 (n = 217)	1857	4

DISCUSSION

We evaluated the performance of two-step screening approaches for depressive symptoms in patients treated with KRTs. We simulated various scenarios where patients would have completed different combinations of questionnaires. Using a sensitivity and specificity of 80% as benchmarks for acceptable screening [13, 40], only the combination of a PHQ-2 score ≥1 followed by a PROMIS-D score ≥53 had acceptable performance for identifying patients with moderate to severe depressive symptoms using a PHQ-9 score ≥10 as the referent. This combination also had an excellent NPV of 95%. Similar results were seen for a PHQ-2 score ≥2 followed by a PROMIS-D score ≥53, with the sensitivity slightly below our pre-specified threshold of 80%.

Compared with the PHQ-2 at a cut-off of ≥1 alone or a PROMIS-D score ≥53 alone, a two-step screening reduced the number of false positives; however, the number of false negatives was higher with the two-step approaches. Compared with administering the PROMIS-D or PHQ-9 alone to all participants, the two-step approach reduced the question burden of the screening. This was especially remarkable for participants with no or a low level of depressive symptoms who were screened out with just two items using the PHQ-2.

The PHQ-2 had much higher sensitivity than the ESASr-D. Moreover, the PHQ-2 alone with a cut-off ≥2 had similar performance to the two-step screening of a PHQ-2 score ≥1 followed by a PROMIS-D CAT score ≥53. However, it should be noted that the overlap in PHQ-2 and PHQ-9 questions may have inflated screening performance, although a similar combination of these tools has been used by others [15, 26, 42]. The reliability of the PHQ-2 in our study was inferior to that of the PROMIS-D, as also reported by others [43].

The two-step screening had slightly lower sensitivity than the PROMIS-D alone. However, this may be because of their substantially improved efficiency with overall fewer questions completed by patients. This was especially prominent for patients with no or very low levels of depressive symptoms. As two-step screening would only administer additional questions to patients who have some degree of symptoms, it may improve the acceptability and feasibility of regular testing in clinical settings. Similar two-step screening approaches have been implemented in routine care for various patient populations [24, 44]. Given the results obtained in our simulated scenarios, future studies to confirm the feasibility and utility of two-step screening in patients with advanced kidney failure are warranted.

While the PHQ-9 is one of the most widely used instruments to screen for depressive symptoms [45, 46], a number of arguments support the consideration of using the PROMIS-D item bank for this purpose. Similar to the PHQ-9, the PROMIS short forms are in the public domain and are free. It has been repeatedly documented that measurement characteristics of PROMIS-D CAT and short forms are similar to those of the PHQ-9 [23, 36]. The PROMIS item banks can be administered as a CAT, which only administers relevant items for most patients. Compared with the two-step approach including the PHQ-2 and PROMIS D, the overall question burden would be higher if the PHQ-9 was used (either alone or as part of two-step screening with the PHQ-2). Many electronic patient record platforms include PROMIS CAT modules, integrating CATs into the routine clinical workflow. Moreover, the PROMIS item banks are available to assess additional symptoms that are frequently associated with depression (e.g. pain, anxiety, fatigue) [47]. These item banks are scored on a similar T-score metric as the depression item bank, providing a common metric for these frequently co-occurring symptoms. Finally, domain scores of the eight most frequently used PROMIS item banks (depression, anxiety, sleep disturbance, fatigue, pain interference, physical function, cognitive function and social participation) can yield summary scores to characterize mental and physical health-related quality of life [48] and a preference-based health utility score [49].

It is important to emphasize that screening for depressive symptoms will not improve outcomes if it is not linked to offering referral to appropriate psychosocial management and support systems for further assessment and management. Screening will only improve outcomes, and in fact only be ethical, if it is followed by connecting identified patients to diagnostic clinical assessment and to appropriate and effective interventions [24]. If the system is prepared, however, systematic screening will help with getting the right care at the right time for patients who need it.

The strengths of this study include a diverse sample of patients on dialysis and kidney transplant recipients. The sample was also ethnically diverse, with 50% of patients self-identifying as members of racialized communities. Furthermore, we had scores from multiple questionnaires to assess depressive symptoms.

We also acknowledge limitations of our study. While we report data from a diverse sample, most of the participants were KTRs and/or male. The results reported are also from a selected group of participants from an existing database. Patients with earlier stages of CKD were not included in this analysis; we are planning additional studies to assess in detail the utility of the PROMIS tools in patients with earlier stages of CKD. In the current study, we did not assess the utility of the two-step screening approach for mild depressive symptoms, to keep the focus on patients who will likely benefit from additional assessment for depression. In this study, only ∼10% of the patients who were screened out in the pre-screening phase had mild depressive symptoms based on the final PHQ-9 score. However, additional analysis will be needed to define the best screening threshold to identify patients with mild depressive symptoms.

In addition, we have not collected information on dialysis adequacy, which would have improved the granularity of describing the study sample. While dialysis dose may be associated with depressive symptoms, it is unlikely to be associated with the screening properties of the patient-reported outcome measures (PROMs) used in this study. Moreover, we used a PHQ-9 score ≥10 as opposed to clinical assessment to identify moderate to severe depressive symptoms. Nevertheless, it would have been challenging to obtain such clinical assessments for a large database of patients. Lastly, the two-step screening approach was assessed through simulations using data from patients who completed the PHQ-9, ESASr and PROMIS-D items rather than data from patients who completed the two-step screening. Future studies should assess two-step screening in the clinical setting. Moreover, to understand potential barriers to using any screening for depressive symptoms, staff and patient perspectives will need to be considered [50].

CONCLUSION

A two-step screening for depressive symptoms using a PHQ-2 score ≥1 as a pre-screener followed by a PROMIS-D score ≥53 has acceptable sensitivity and specificity in patients on KRTs. The two-step screening could deliver efficient and feasible screening with reduced question burden, especially among patients with no or a low level of symptoms. Nevertheless, clinical assessment following a positive screen will be necessary to establish the diagnosis and determine the appropriate support or treatment.

FUNDING

The study was supported in part by a grant from the Kidney Foundation of Canada (KFOC190008), the Canadian Institutes of Health Research (PJT 165915) and the Toronto General and Western Hospital Foundation.

AUTHORS’ CONTRIBUTIONS

All authors contributed intellectual content during manuscript drafting or revision and accept accountability for the overall work.

DATA AVAILABILITY STATEMENT

Deidentified data analysed during the current study are available from the corresponding author upon reasonable request.

CONFLICT OF INTEREST STATEMENT

The authors declare no potential conflicts of interest.