Validation of the Supportive and Palliative Care Indicators Tool in a Geriatric Population

Offering timely palliative care to patients facing life-threatening illness has proven to be beneficial. It increases satisfaction and quality of life and diminishes depressive feelings, anxiety, and stress in both patients and families.^1–3 Unfortunately, this timely identification appears to be a challenge, especially for geriatric patients.^4–10 An important reason is that for the latter patients, the trajectory of disability in the last year of life is very heterogeneous, resulting in a nonpredictable course and prognostic uncertainty.¹⁰

To identify patients with unmet palliative care needs, several tools were developed over the years.^11–15 The Supportive and Palliative Care Indicators Tool (SPICT™) is one of those tools, combining 6 general and 21 clinical questions regarding deteriorating health.^11,13 Different versions of SPICT have been published and the April 2015 version is used in this study.¹⁶ In contrast to some other tools, such as the multidimensional prognostic index (MPI) or the study of osteoporotic fractures (SOF),^14,15 the prognostic value of SPICT in a geriatric population has never been validated before.

SPICT was revised in April 2016 with a decrease in cut-off values and thus a changing focus from prognostication to identifying care needs in a broad patient population.¹⁷ Nevertheless the authors believe that prognostication is an indispensable issue when assessing palliative care needs. Offering palliative care to patients who are not yet confronted with impending end of life may result in deprivation from curative care and healing opportunities, as well as the expression of desires that do not reflect true end-of-life wishes. The latter phenomenon is called pseudoparticipation.⁹ Only when the patient's prognosis is estimated can appropriate and timely medical decisions be delivered.^8,18 Although several studies state that the clinical survival predictions of physicians are superior to prognostic tools in the short term,^19,20 such prognostic tools may bring an added value for long-term predictions, wherein physicians often tend to be overoptimistic.⁸

This retrospective cohort study aims at validating the SPICT in a geriatric patient population admitted to the acute geriatric ward, by measuring the ability to predict one-year mortality. The association between the retrospectively calculated SPICT and the presence of a treatment limitation decision (TLD), as a measure for the clinical survival predictions of the geriatrician, is investigated and adds a second validation of SPICT.^21–23

All patients admitted to the acute geriatric ward of a Belgian university hospital for more than one day, between January 1 and June 30, 2014, were included. In an acute geriatric hospital ward, specialized acute care for geriatric patients is offered within a multidisciplinary approach.²³ Throughout the hospital, a standardized TLD coding system is used on which fixed categories of nontreatment decisions can be registered. A TLD code 0 means full therapy and a TLD code 1 stands for no cardiopulmonary resuscitation only. A TLD code 2 implies withholding of therapy (e.g., referral to the intensive care unit, upgrading of antibiotics, and dialysis), whereas a TLD code 3 stands for comfort care only with no life-sustaining therapies. As for all other medical interventions or decisions, the Belgian Law on Patients' Rights requires informed consent from the patient or his/her surrogate decision maker in case of incapacity.²²

Data regarding demographics, functional status, comorbidities, TLD on admission and discharge, as well as one-year mortality were collected from the electronic patient record. SPICT was measured retrospectively by an independent assessor, who holds medical skills and could rely on the supervision of two geriatricians throughout the entire process of data collection. The general and clinical indicators were mainly found in the comprehensive geriatric assessment (CGA), nursing records and report of multidisciplinary team meetings. Information concerning the one-year mortality was not recorded in the patient file for 158 patients. After telephone contact with patients' general practitioners, the number of cases with unknown one-year mortality was reduced to 25.

Statistical calculations were carried out using SPSS version 23. For continuous data, median and interquartile range were computed. The relationship between continuous data and the dichotomous outcome of SPICT was assessed using Mann–Whitney U tests. Categorical data were assessed using chi-square tests. The exact p values are reported, with statistical significance defined as p ≤ 0.05. To validate the SPICT, a receiver operating characteristic curve was built using SPSS, while the comparison of the areas under the curve (AUCs) was performed using MedCalc.

For this retrospective cohort study, acquisition of informed consent from studied patients was not obliged. The protocol of this study was approved by the ethics committee of the Ghent University Hospital.

To the best of the authors' knowledge, this study is the first to validate SPICT in predicting the one-year mortality and the association with TLDs in a geriatric population. The main finding of this study is that SPICT can predict one-year mortality in a geriatric population with a sensitivity of 0.841 and a specificity of 0.579.

Other prognostic tools, such as the MPI and SOF, show similar sensitivity and specificity when predicting, respectively, the one- and three-year mortalities (sensitivity: ±0.80 and specificity: ±0.40–0.50).^14,15 However, the use of SPICT is to be preferred because of several advantages. When there are no opportunities to carry out a CGA, the SPICT shows an equal capacity of prognostication as does the MPI, but using considerably fewer questions. All SPICT questions can be displayed synoptically on one page. In contrast to MPI and SOF, studies have shown that physicians assess the SPICT as convenient and feasible.^11,13,24,25

Although both parts of SPICT are equally strong at predicting the one-year mortality, a combination of both parts may bring an added value for physicians who are educated to mainly look for clinical symptoms (second part), rather than functional or general indicators of health (first part). Despite the major complexity of diseases in geriatric patients, 86.2% of patients in the studied acute geriatric unit had at least one positive clinical SPICT indicator. Furthermore, the addition of this second part of SPICT may have increased the specificity of this tool. The combination of cut-off value 2 for the first part and value 1 for the second part is preferred because of the balance between sensitivity and specificity. A rather low specificity, as in the 2015 version, results in identifying some patients who will not die within one year. Offering these patients advance care planning will not cause any harm, therefore, a rather low specificity can be tolerated.^1–3 In contrast, the specificity should not be extremely low. In the SPICT version of April 2016, where cut-off values 1 and 1 are used, the specificity is reduced to 0.245 in this patient population. Because of the risk of the so-called phenomenon of pseudoparticipation, the SPICT version of April 2015 is preferred by the authors.⁹

Given the fact that SPICT was measured retrospectively, it did not function as a guide to assign TLDs, but rather as a measure for the clinical survival predictions of experienced geriatricians. In 71.4% of cases, geriatricians assigned TLDs based on their own clinical knowledge and experience. The added value of SPICT in helping experienced physicians on the acute geriatric ward to identify patients in need of palliative care requires further investigation, as well as the possible added value for physicians less experienced in care of the older person.

A limitation of this study is the retrospective collection of data leading to the possibility of missing information. Future research should be multicentric, prospective, and should also include hospital wards where geriatric patients are not cared for by geriatric experts.

SPICT seems to be a valuable tool for identifying geriatric patients in need of palliative care as it demonstrates significant association with one-year mortality and with clinical survival predictions of experienced geriatricians, as reflected by TLDs given. SPICT predicted the one-year mortality of geriatric patients with a sensitivity of 0.841 and a specificity of 0.579 in this study. These values are comparable with those by prognostic tools commonly used in geriatric medicine, such as MPI and SOF. Because experienced geriatricians were able to identify most patients with limited prognosis based on their clinical survival predictions, the added value of SPICT on an acute geriatric ward can be questioned. The added value of SPICT on other hospital wards and in primary care needs to be further investigated.

We express our gratitude to all team members of the geriatric ward of the Ghent University Hospital for their support.