Introduction
Dichotomous outcomes—such as mortality—rarely capture the diverse range of potential outcomes important to patients and clinicians. For infectious diseases, it is particularly important to assess a range of outcomes as antimicrobials have potential harms and benefits that need to be balanced. To address these needs, the Desirability of outcome ranking (DOOR) was created. DOOR is a method of analysis that assesses multiple potential outcomes simultaneously from least to most desirable.Reference Evans, Rubin and Follmann1
Currently, there is no standard method to create a DOOR end point with most developed based on clinical opinion. Furthermore, although DOOR was designed to capture outcomes important to the patient experience, patient feedback has not been consistently used to create new DOOR endpoints.Reference Ong, Petersiel, Loewenthal, Daneman, Tong and Davis2–Reference Doernberg, Tran and Tong4 This could impact the validity and patient-centeredness of trial results as patients and clinicians may differ in their outcome preferences.Reference Rubin, Buehler and Halpern5 Finally, there is no standardized DOOR to assess outcomes in adults with non-severe community-acquired pneumonia (CAP).Reference Ong, Petersiel, Loewenthal, Daneman, Tong and Davis2
We aimed to a) develop a novel DOOR for adults hospitalized with CAP and b) assess differences in ranking of DOOR components between patients and physicians.
Methods
To create the DOOR for adults admitted with CAP, a multidisciplinary group of physicians generated nine potential clinical cases describing the range of potential outcomes of a patient with CAP two weeks after their initial emergency department (ED) visit (e Table 1). Using prior DOOR endpoints as a model,Reference Ong, Petersiel, Loewenthal, Daneman, Tong and Davis2–Reference Doernberg, Tran and Tong4 we created 9 case vignettes describing variable clinical resolution of symptoms, adverse effects of antimicrobials (including healthcare-associated infections and side effects), and treatment failures such as readmission or death (see Table 1). Vignettes were entered into a REDCap survey so respondents could rank them from most to least desirable.Reference Harris, Taylor, Thielke, Gonzalez and Conde6 For the patient survey, we edited the vignettes to reduce jargon.
Desirability of outcome ranking combining physician and patient survey resultsa
aFinal DOOR component rankings ordered based on combined physician and patient median and IQR. Rankings that were not significantly different from each other are color coded to denote a collapsed final component. bComparisons between patient and clinician component rankings were made using Mann Whitney U test, those with P < .05 are considered statistically significant. Abbreviations: DOOR, desirability of outcome ranking; AE, adverse event; IQR, inter-quartile range
To ascertain physician DOOR preferences (ie, ordering of potential outcomes), we used a snowball samplingReference Johnson, Balakrishnan, Colton, Everitt, Piegorsch, Ruggeri and Teugels7 method to recruit clinicians in specialties that regularly treat CAP (ie, infectious disease, emergency medicine, hospital medicine, pulmonology) over a two-week period from 11/2023 – 12/2023 at one academic and one community hospital in Utah. To ascertain patient preferences, we recruited patients hospitalized with pneumonia on a general medicine unit at the academic hospital between 9/2024 and 10/2024. Patients were approached by a research coordinator to either answer the electronic survey or be assisted verbally with the survey (per their preference). Physicians were emailed an electronic survey to rank potential outcomes in order of preference. All survey data were entered using REDCap.
To create the final DOOR, we used a Friedman Rank Sum test on combined physician and patient rankings to collapse DOOR components that did not significantly differ from each other in their ranking (all p values > .2). We also compared DOOR component rankings between physicians and patients for the 9 vignettes using a Mann–Whitney U test. Results were presented to the Hospital Medicine Reengineering Network (HOMERuN) Patient and Family Advisory Council (PFAC)8 for their impressions on differences between patient and physician responses. The PFAC responses were transcribed using automated Zoom software and edited for accuracy by a research assistant; we selected exemplar quotes representing their perspectives (see Table 2). The University of Utah Institutional Review Board determined this project was exempt.
Results
In total, 25 physicians completed DOOR surveys: 4 infectious disease, 9 emergency medicine, 5 hospital medicine, 7 pulmonology/critical care (denominator/response rate unknown given snowball sampling). Of the 31 eligible patients contacted for inclusion, 24 agreed to participate (77.4% response rate); one patient did not respond due to altered mental status and six declined. Of patients agreeing to participate, four were omitted due to an incomplete survey or submitting improperly (eg, confusion evident after beginning survey). Patients surveyed (n = 20) had a median age of 56 years, 45% were male, and 75% self-identified as White (15% Hispanic, .05% Native Hawaiian, and .05% Native American).
When comparing component rankings between physicians and patients, ranks for 7 of the 9 initial case vignettes differed significantly (Table 1). Physicians ranked non-limiting dyspnea, multidrug-resistant urinary tract infection, hospitalization, and death as less preferable than patients. Patients ranked the antibiotic adverse effects of non-limiting nausea and vomiting and Clostridioides difficile infection as worse than physicians. After combining data from both patients and physicians, 5 components remained significantly different from each other, enabling us to collapse the 9 vignettes into a DOOR end point with 5 total components (see color coding in Table 1). PFAC quotes are included in Table 2.
Exemplar quotes from HOMERuN Patient and Family Advisory Council hypothesizing why physician patients might rank cases differently
Abbreviations: HOMERuN, Hospital Medicine Reengineering Network
Discussion
We found significant differences between patient and physician preferences in ordering of potential CAP outcomes. Trials should consider whose perspective they prioritize when deciding which order to include. The simplest DOOR—which collapsed outcomes across groups that received similar rankings—involves 5 final components.
While DOOR has been used as a primary outcome in multiple infectious diseases,Reference Ong, Petersiel, Loewenthal, Daneman, Tong and Davis2–Reference Doernberg, Tran and Tong4 there is a paucity of studies for adult CAP using DOOR as a primary outcome. Only one trial to date has targeted non-severe CAP and that trial only included pediatric patients.Reference Williams, Creech and Walter9 We found that patients ranked some antimicrobial side effects as worse than physicians including nausea and vomiting and Clostridioides difficile infection, while physicians ranked readmissions, death, and developing a multidrug-resistant UTI as worse than patients. Although both patients and physicians ranked mortality as the least desirable outcome, three patients ranked it more desirable than some of the other outcomes leading to statistically different distributions. Rubin et al noted similar patient sentiment with most hospitalized patients surveyed indicating death was preferable to several permanent debilities.Reference Rubin, Buehler and Halpern5
While not explored explicitly here, using the DOOR in studies may also allow individualized recommendations—for example, one of our PFAC members who had previously had Clostridioides difficile reported that, for them, avoiding a repeat case would be the most important goal. Regardless, the differences found in patient and physician ranking further supports that current outcome measures may miss the patient perspective (eg, side effects are an important outcome). One path forward is for trials to conduct sensitivity analyses of different DOOR rankings based on whose perspective (patient or clinician) they are prioritizing.
In terms of limitations, our sample size was small, hospitalized, predominantly over 50 years old and self identifying as white, potentially limiting generalizability of our DOOR more broadly and reducing power to detect smaller differences between physician and patient preferences. Ranks which were collapsed may have been significantly different with a larger sample size. Demographic data did not include education level or socioeconomic status which could impact survey responses. Not all variations in severity of potential outcomes were provided in initial cases given the wide range of potential outcomes. Surveys were slightly different for patients and clinicians to reduce jargon-though this could also bias responses. Furthermore, our DOOR does not account for potential magnitude differences in importance of outcomes between adjacent ranks. Methods of providing “partial credit” for some outcomes could be considered to help discriminate severity of outcomes.Reference Evans and Follmann10
Our study demonstrates the importance of including patient perspectives in the development of DOOR as the relative ranking of outcomes may differ. While we created a novel DOOR with 5 components for hospitalized adults with non-severe CAP that incorporates both physician and patient perspectives, comparative effectiveness research could consider sensitivity analyses of different component rankings based on which perspective is being prioritized.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ash.2025.10195
Data availability statement
All data available upon request.
Acknowledgments
We would like to thank all patients, physicians and members of the HOMERuN Patient and Family Advisory Council for participating in this study.
Financial support
No funding was used for this study.
Competing interests
All authors report no conflicts of interest relevant to this article.
Open access
