Article Text

Download PDFPDF

mHOMR: the acceptability of an automated mortality prediction model for timely identification of patients for palliative care
  1. Stephanie Saunders1,
  2. James Downar2,3,4,5,
  3. Saranjah Subramaniam6,
  4. Gaya Embuldeniya7,8,
  5. Carl van Walraven3,5,9,
  6. Pete Wegier6,8,10
  1. 1 Department of Rehabilitation Sciences, McMaster University, Hamilton, Ontario, Canada
  2. 2 Division of Palliative Care, The Ottawa Hospital, Ottawa, Ontario, Canada
  3. 3 Department of Medicine, University of Ottawa, Ottawa, Ontario, Canada
  4. 4 Bruyère Research Institute, Ottawa, Ontario, Canada
  5. 5 Ottawa Hospital Research Institute, Ottawa, Ontario, Canada
  6. 6 Humber River Hospital, Toronto, Ontario, Canada
  7. 7 Toronto General Research Institute, University Health Network, Toronto, Ontario, Canada
  8. 8 Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada
  9. 9 Department of Epidemiology & Community Medicine, University of Ottawa, Ottawa, Ontario, Canada
  10. 10 Department of Family & Community Medicine, University of Toronto, Toronto, Ontario, Canada
  1. Correspondence to Dr Pete Wegier, Humber River Hospital, Toronto, ON M3M 0B2, Canada; pwegier{at}hrh.ca

Statistics from Altmetric.com

Introduction

Patients with non-cancer serious illnesses are under-recognised and receive palliative care only in the final weeks of life, if at all.1 The modified Hospitalised-patient One-year Mortality Risk (mHOMR) tool is a computer-based mortality prediction tool that accurately identifies patients at risk of 1-year mortality and is a feasible alternative to healthcare provider (HCP)-dependent models.2 Briefly, the tool uses data from the electronic health record to calculate an mHOMR score for each new hospital admission. The alert only notifies the lead physician, suggesting they refer the patient topalliative care and does not provide the actual score.2 In this study, we sought the perspectives of patients, family members, and HCPs to identify acceptability of mHOMR as a mortality risk tool. Together, these two studies represent the feasibility and acceptability components of the implementation outcomes (IO) framework.3

Methods

Previously we reported the development and feasibility of mHOMR (see Wegier et al 2 for more details). Alongside the feasibility study2 we collected qualitative data from November 2016 to May 2017 pre-implementation and from June to October 2017 post-implementation at two quaternary hospitals in Toronto, Canada. We used a postpositivist, qualitative content methodology4 and consecutively recruited: (1) English-speaking patients admitted to a medicosurgical ward with an mHOMR score >0.21 (ie, >21% risk of death in 12 months) and (2) HCPs who admitted patients with an mHOMR score >0.21 or were involved in advance care planning or goals of care (GOC) discussions with these patients. Substitute decision makers were recruited if a patient could not consent. In-person interviews with patients and caregivers and phone interviews with HCPs were conducted before and after implementation of mHOMR. We followed semistructured interview guides (Interview guides can be found here: https://osf.io/34dcm/?view_only=4eefb31c12404d55aec2ff697054f25d) asking about challenges to initiating a palliative care approach and both expectations and experiences with mHOMR. Interviews were conducted by an experienced qualitative researcher with a PhD in anthropology (GE) and were recorded, transcribed verbatim, and anonymised. Three coders (PW, SSa, SSu) analysed the data using MaxQDA,5 with at least two coders coding each transcript; SSa coded every transcript. Analysis was done using an iterative inductive and deductive qualitative content analysis.4 Findings from before and after implementation were compared and no noteworthy differences were found. In the event of disagreement, consensus was reached through discussion.

Results

Of 80 participants screened, patients (n=22), caregivers (n=15), residents (n=3), administrative staff (n=3) and physicians (n=21) participated (n=64, 80% participation rate). Median interview length was 12 min (IQR=13). Forty-nine participants replied to the question ‘Do you find this tool acceptable?’; answering yes (71%, n=35), no (12%, n=6) and unsure (16%, n=8). Those who found mHOMR unacceptable emphasised situational challenges, whereas acceptable responses emphasised the advantages of an automated approach. Facilitators and barriers for mHOMR uptake are reported with illustrative quotes in table 1.

Table 1

Representative quotes

Perceived facilitators

Patients and caregivers perceived an advantage to their HCPs receiving a mortality prediction alert via mHOMR. Physicians felt similarly, stating the information provided context to the patient in front of them. Since mHOMR does not mandate any actions, HCPs valued receiving information while preserving judgement in care decisions. HCPs discussed the benefit of reminders or confirmations of their gestalt impression of patients’ potential palliative needs. Residents discussed the value of mHOMR as sometimes they lacked the clinical experience required to identify patients with an elevated risk of mortality.

Perceived barriers

When deploying the alerts, HCPs felt it was important to consider the situation and context. Some preferred alerts at specific times and directed to specific HCPs, such as the nurse leading rounds or residents on call. Some HCPs indicated the mHOMR alert itself did not include enough information about how the score was calculated. Alert fatigue was another common concern. Some physicians who felt they were already aware of the patient’s elevated mortality risk were concerned about redundancy of the alert. Physicians felt it was critical to address immediate and pressing issues (ie, the reason for the acute admission) over long-term care needs. Others felt mHOMR alerts added to their gestalt but felt unclear about appropriate next steps. Both physicians and patients voiced concerns over whether mHOMR would limit patients’ agency to make care decisions.

Discussion

This is the first qualitative study to demonstrate acceptability of using an automated mortality prediction tool to support care decisions in a hospital setting. Our findings are not surprising given that presumed acceptability rates, as evidenced by acceptance of a palliative care triggering mandate, among automated mortality prediction tools have been shown to be high.6–8 Previous research highlights the acceptability of patient and/or clinician-reported prognosis tools in both community and hospital settings.9–12 Reasons for this are similar to our findings, that it helps to provide context to patients9 13 and that it is individualised.13 Given the manpower required to implement self-report tools, the acceptability of automated tools is promising since clinicians have poor recognition of end of life (EOL)1 14 and report limited capacity,8 14 which often thwarts these conversations upstream. This is concerning since having conversations about EOL has been found to increase patient agency and satisfaction at EOL.15 Automated models, such as mHOMR, may contribute to increasing the number of these upstream conversations,16 thus improving quality of care at EOL.

Aligning the acceptability of this study with the commonly used Hexagon tool,17 which uses six criteria to assess acceptability within implementation sciences, we see that there is an obvious (1) Need and (2) Fit for mHOMR in the organisation. Regarding (3) Resources, (4) Capacity, and (5) Evidence, participants discussed few concerns. Primarily, the lack of ability to address the alerts as a result of capacity and concerns about needing to focus on acute needs over long-term concerns likely reflects a broader issue driving late adoption of a palliative approach to care, where more urgent issues justify delaying this discussion. With respect to (6) Readiness, participants reported tension between the desire for more information surrounding patients’ conditions and concern over agency in care decisions.

Regarding limitations, we were unable to collect demographic data or mHOMR scores of participants. However, given the consecutive enrolment and high degree of participation, our sample should be representative of patients who may be seen on a medicosurgical ward with an mHOMR score of >0.21. Second, some participants were unable to provide a large period of their time, resulting in a short average interview duration. While this work is still in the early phases, the feasibility study showed promise that the alert leads to changes in clinical practice and so future research will aim to scale up the use of this tool to better assess the remaining IOs.

This study, combined with Wegier et al’s2 study, represents two components of the IO framework proposed by Proctor et al.3 Taken together, the mHOMR tool is feasible and is acceptable, the results are promising to continue to assess implementation. Future research will continue to look at ideal implementation conditions, as dictated by the IO framework.

Ethics statements

Patient consent for publication

Acknowledgments

This work was conducted while Stephanie Saunders, Saranjah Subramaniam, and Pete Wegier were affiliated with the Temmy Latner Centre for Palliative Care, part of Sinai Health in Toronto, Ontario.

References

Supplementary materials

Related Data

Footnotes

  • Twitter @petewegier

  • Contributors JD conceived the study and developed the protocol. PW, SSa and SSu led the drafting of the manuscript. All authors contributed to data collection and/or analysis and interpretation, revising the manuscript, and approved the final version submitted for publication.

  • Funding This research was funded by Canadian Frailty Network (Technology Evaluation in the Elderly Network), which is supported by the Government of Canada through the Networks of Centres of Excellence (NCE) programme. This project was also supported financially by the Temmy Latner Centre for Palliative Care and the Toronto General/Toronto Western Foundation, and received in-kind support from the Ottawa Hospital Research Institute. JD received support for this project from the Associated Medical Services through a Phoenix Fellowship.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles