Determining the prevalence of palliative needs and exploring screening accuracy of depression and anxiety items of the integrated palliative care outcome scale – a multi-centre study

This was a multi-centred observational study. Data were collected in nine portuguese centres spread out from north to south and rural to urban locations to maximise generalizability using convenience sampling. There were seven hospital based palliative care services, one oncology service and one primary care facility (health centre). All patients attending the participant services were screened for eligibility by the participating healthcare professionals. Inclusion criteria were as follows: ≥18 years, mentally fit to give consent judged as such by the participating healthcare professional, diagnosed with an incurable, potentially life-threatening illness, read, write and understand Portuguese. Exclusion criteria were: patient in distress (unable to maintain a conversation during a period of time) with uncontrolled physical or emotional symptoms, and/or cognitively impaired, judged as such by the participating healthcare professional. A standard operating procedures manual was developed and distributed to all centres in the person of the facilitator/champion leading the study locally. The detailed protocol has been published elsewhere [18].

The Portuguese version of the IPOS reported by the patient, previously developed, was used. The original measure has been culturally adapted and validated to European Portuguese [20, 21]. The protocol of the latter study has been published elsewhere [18] which contains the full questionnaire in the appendix. Next, we present a summary of main procedures and results. Two native Portuguese speaking translators, one clinical and one non-clinical independently created two Portuguese versions. A consensus Portuguese version was developed by two native Portuguese speaking independent reviewers blind to the original IPOS. This consensus version was sent to two other independent native Portuguese speaking translators, also blind to the original English IPOS, who back translated it into English. A second Portuguese consensus version was developed by the same reviewers. Three clinical revisions were performed by one specialist palliative care doctor, one specialist palliative care nurse and one non-clinical researcher – all native Portuguese. A final Portuguese version was created. There were grammatical and content differences in the first translation stage, in the items/questions text as well as in the response categories. These were resolved by discussion by both reviewers. There were also differences in the backward translation, namely verb tenses and the use of synonyms rather than the direct translation of words. These were resolved by discussion by the same reviewers. The clinical revisions flagged differences in verb tenses in three items. Those were discussed, and changes were made to create the final version. A Portuguese version of the IPOS was developed. Next, psychometric characteristics were assessed, namely, Internal consistency (excluding open questions, which are free text data) Cronbach’s alpha varied between 0.68 and 0.72, reliability between patients and healthcare professionals scores was assessed by intraclass correlation which was higher for mobility (ICC = 0.726 and lowest for practical problems (ICC = 0.088). Regarding construct validity, Items with similar constructs showed convergent validity and items with different constructs showed divergent validity. Spearman’s Rho varied between .390 and .631 with p ≤ .000. The measure also displayed good sensitivity to change, as Wilcoxon ranked test showed significant statistically differences between T1 and T2 in three symptoms. IPOS It is a brief, 19-item, multidimensional scale that captures core concerns in palliative care. The first item is an open question on the three main problems or worries the respondent might have had in the past week (results of this item are not reported in the present study given that data are free text); items 2 to 9 are set on a 5 point Likert scale based on descriptors (zero – not at all, 1 – slightly, 2 – moderately, 3 – severely, 4 – overwhelmingly), item two is a list of 10 of the most common physical symptoms in a palliative population, with the possibility of adding up to three more symptoms which are not present in the list; item 3 pertains to anxiety, item 4 asks about family/friends worry, item 5 is on depression; item 6 is about being at peace; item 7 relates to sharing feelings with significant people; item 8 is about information needs and item 9 concerns practical problems related to their illness. In the patient version (as opposed to the healthcare professional version), the questionnaire has an extra item asking if the respondent filled the questionnaire alone or with help. At the very end, there is a footnote to trigger the patient to talk to their healthcare professional, if they feel they are worried about any of the issues raised by the items in the questionnaire. This feature allows for real time clinical utility of the measure.

The Portuguese Hospital Anxiety and Depression Scale (HADS) is a 14 item screening measure for anxiety and depression states. The two subscales are comprised of 7 items each, scored separately, with descriptive answers based in a 4 point Likert scale. The authors propose a cut-off threshold of 11 for depression and anxiety. The authors conclude that the Portuguese HADS is a reliable and valid measure for assessing anxiety and depression in different medical settings and disease populations [22].

After checking data quality and performing Little’s Missing Completely At Random (MCAR) test to check if data were missing at random, descriptive statistics were used to examine the distribution of demographic and clinical variables of interest. To determine accuracy of the items under study, we previously tested all possible cut-offs (results not presented) and decided to use the most appropriate, a cut-off score of 2/3. Then we compared all 5 psychological, emotional and spiritual needs items (IPOS items 3 to 7) against the Portuguese HADS. Receiver operating characteristic curves (ROC) were used to determine the 2 items displaying the highest AUC and assess discriminant properties, namely, sensitivity and specificity of the cut-off 2/3 for both items against cut-off 10/11 of the HADS Anxiety subscale and the HADS Depression subscale respectively. Positive (PPV) and negative predictive values (NPV), false negative (FNR) and false positive rates (FPR) and positive and negative likelihood ratios weighted by prevalence were also computed for both items. For sensitivity, specificity, PPV, and NPV we considered 70% or above values to be acceptable and 80% or above to be high. For FPR and FNR of 30% or less, we considered them to be low. 95% confidence intervals were used. No sample size calculation was performed given that the only study found in the literature used the Palliative care Outcome Scale questionnaire (not the IPOS) and the HADS to screen for depression and anxiety was a secondary analysis of several independent datasets.

Ethical approval was granted from all relevant Research Ethics Committees and was in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical Standard. All participants gave informed signed consent. SPSS, version 22 (SPSS/IBM Corp., Armonk, NY) software was used.

In terms of prevalence of needs, IPOS items scoring the highest (=4) were: family or friends anxious or worried (36.3%); feeling anxious or worried about illness (13.3%); feeling depressed (9.6%); feeling at peace (9.6%); share feelings (8.9%) and pain (7.4%). IPOS items scoring the lowest (=0) were: vomiting (77%); shortness of breath (67.4%); nausea (65%); information needs (60.7); practical problems (45.2%) and constipation (43%). (See Fig. 1).

Item 3 (anxiety) and item 5 (depression) presented the highest AUCs (see Fig. 2). The prevalence of depression was 24.4% (C.I. 17.6–32.7%). The AUC curve was 0.72 (C.I.:0.62–0.81), p < 0.001 (see Fig. 3). Sensitivity was 51.5% and specificity was 78.4%. Positive predictive value was 43.6% and negative predictive value was 83.3%. As for the anxiety item, the prevalence was 23.7% (C.I. 16.9%–.31.9%) and the AUC was 0.70 (C.I.:0.60–0.80), p < 0.001 (see Fig. 4). Sensitivity was 65.6% and specificity was 68.0%. Positive predictive value was 38.8% and negative predictive value was 86.4% (see Table 4).