Young Et Al. - 2017 - Discordance of Patient-Reported and Clinician-Orde

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Vol. 53 No.

4 April 2017 Journal of Pain and Symptom Management 745

Brief Report

Discordance of Patient-Reported and Clinician-Ordered


Resuscitation Status in Patients Hospitalized With Acute
Decompensated Heart Failure
_
Kathleen A. Young, MD, Sara E. Wordingham, MD, Jacob J. Strand, MD, Veronique L. Roger, MD, MPH, and
Shannon M. Dunlay, MD, MS
Department of Medicine (K.A.Y., J.J.S.), Mayo Clinic, Rochester, Minnesota; Department of Medicine (S.E.W.), Mayo Clinic, Scottsdale,
Arizona; Department of Cardiovascular Diseases (V.L.R., S.M.D.), Mayo Clinic, Rochester, Minnesota; and Department of Health Sciences
Research (V.L.R., S.M.D.), Mayo Clinic, Rochester, Minnesota, USA

Abstract
Context. Accurate documentation of preferences for cardiopulmonary resuscitation at hospital admission is critical to
ensure that patients receive resuscitation or not in accordance with their wishes.
Objectives. We sought to identify and characterize inconsistencies in patient-reported and clinician-ordered resuscitation
status in patients hospitalized with acute decompensated heart failure (ADHF).
Methods. Southeastern Minnesota residents hospitalized with ADHF were prospectively enrolled into a study that included
the administration of face-to-face questionnaires from January 2014 to February 2016. Patient-reported resuscitation status was
assessed at enrollment using a validated question. Clinician-ordered resuscitation preferences at hospital admission were
abstracted from the electronic medical record.
Results. Of the 400 patients administered the questionnaire; 213 (53.3%) stated their resuscitation preference as Full
Code, 166 (41.5%) do-not-resuscitate (DNR), and 21 (5.3%) were unsure. In comparison, clinician-ordered resuscitation
status was Full Code in 263 (65.8%) patients, DNR in 133 (33.3%), and not documented in four (1.0%). Patient-reported and
clinician-ordered resuscitation status was discordant in 20% of patients, of whom 5.6% elected Full Code by questionnaire and
had a DNR clinician order, and 14.4% elected DNR by questionnaire but had a Full Code clinician order. Differences in age,
comorbidities, health literacy, marital status, completion of advance directives, hospital length of stay, and discharge
destination in patients with discordant vs. concordant resuscitation preferences were observed.
Conclusions. Patient-reported and clinician-ordered resuscitation preferences were discordant in 20% of patients
hospitalized with ADHF. The underlying etiology of these inconsistencies may reflect factors such as patient indecisiveness or
patient-clinician miscommunication and requires further exploration. J Pain Symptom Manage 2017;53:745e750. Ó 2016
American Academy of Hospice and Palliative Medicine. Published by Elsevier Inc. All rights reserved.

Key Words
Heart failure, resuscitation, hospitalization, advance care planning

Introduction time of hospital admission for acute decompensated


heart failure (ADHF) has not been recently examined.
Accurate documentation of preferences for cardio-
In the Study to Understand Prognoses and Prefer-
pulmonary resuscitation in the hospital is critical to
ences for Outcome and Risks of Treatments that
ensure that patients receive resuscitative efforts in
enrolled patients hospitalized with serious illness in
accordance with their wishes. The accuracy of
the early 1990s, when physicians were asked if they
clinician-ordered resuscitation preferences at the
thought patients with heart failure (HF) would want

Address correspondence to: Shannon M. Dunlay, MD, MS, Mayo Accepted for publication: November 14, 2016.
Clinic, 200 First Street SW, Rochester, MN 55905, USA.
E-mail: dunlay.shannon@mayo.edu

Ó 2016 American Academy of Hospice and Palliative Medicine. 0885-3924/$ - see front matter
Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpainsymman.2016.11.010
746 Young et al. Vol. 53 No. 4 April 2017

to be resuscitated, they were wrong 24% of the time.1 know, there are a number of things that doctors can
Since that time, there has been increased recognition do to try to revive someone whose heart has stopped
of the importance of communication between clini- beating, which usually includes a machine to help
cians and patients with HF about their wishes for breathing. Thinking of your current condition, what
end-of-life (EOL) care, including preferences for would you want your doctors to do if your heart ever
resuscitation. National HF guidelines now recom- stops beating? Would you want your doctors to try to
mend that clinicians discuss preferences for EOL revive you, or would you want your doctors not to try
care and resuscitation with all patients with HF.2,3 to revive you?.’’5 Choices were ‘‘Yes, try to revive me’’
However, whether this enhanced recognition of the (Full Code), ‘‘No, do not try to revive me’’ (do-not-
importance of these discussions has translated to resuscitate [DNR]), and unsure/no answer.
improved clinician accuracy in documenting patient Clinician-ordered resuscitation data were collected at
resuscitation preferences remains to be determined. the time of admission as resuscitation status is a stan-
In this study, we sought to compare patient-reported dard part of all hospital admission order sets. Changes
and clinician-ordered resuscitation preferences in in orders for resuscitation during the hospitalization
contemporary patients hospitalized with ADHF. In were also collected. Clinician-ordered resuscitation
addition to delineating their frequency, we also aimed status was defined as either ‘‘Full Code’’ or ‘‘DNR’’
to characterize the nature and patient predictors of in- based on orders in the EMR. As defined at our institu-
consistencies when they occurred. tion, DNR status refers to withholding cardiopulmo-
nary resuscitation in the setting of unconsciousness
and pulselessness.
Methods
Study Design Additional Data
Patients hospitalized with ADHF at Mayo Clinic Hos- Patient activation level was assessed using the vali-
pitals in Rochester, MN, were approached for prospec- dated 13-item Patient Activation Measure6 with lower
tive cohort study participation between January 2014 levels reflecting worse activation. Patients who are acti-
and January 2016. The study included administration vated believe patients have important roles in main-
of a face-to-face questionnaire by a research assistant. taining their health and know how to manage their
All participants provided written authorization, and condition, prevent health decline, and collaborate
the study was approved by the Mayo Clinic Institu- with their clinicians. This instrument was included as
tional Review Board. to explore if there was an association of activation
with resuscitation preferences. Health literacy was as-
Patient Population sessed using the Health Literacy Screener7 and health
Patients with HF were identified using natural lan- status using the question, ‘‘In general, would you say
guage processing of the electronic medical record that your health is excellent, very good, good, fair or
(EMR) text.4 After a hospital encounter, documenta- poor?.’’8 Completion of an advance directive was as-
tion is transcribed and appears in the record within sessed using the question, ‘‘Do you currently have a
24 hours, making prompt identification of patients hos- signed medical power of attorney or living will?’’
pitalized with HF possible. The search was restricted to Marital status and education level were assessed by
patients at least 20 years old who were residents of a questionnaire. Demographics and comorbidities
seven-county area in Southeastern Minnesota, where were collected from the EMR. The Charlson comor-
the vast majority of residents have been shown to receive bidity index was used to assess comorbidity.9 Left ven-
health care at the Mayo Clinic. This approach was taken tricular ejection fraction (EF) at rest was collected
to enroll study participants from the local community, from transthoracic echocardiograms performed
rather than patients referred for specialty care from within six months before to one month after study
other parts of the state and country. Medical records enrollment. Preserved EF was defined as 50% or
were reviewed by experienced research nurses to deter- more. Vital status (alive or deceased) at 30 days was
mine whether patients were hospitalized with ADHF, available on all patients. The date of death was deter-
met Framingham criteria for HF, and had a previous mined using death certificates filed in local counties,
diagnosis of HF. Patients with a new diagnosis of HF obituary notices, and electronic files of death certifi-
were excluded to ensure that they had the opportunity cates obtained from the State of Minnesota Depart-
to thoroughly discuss their diagnosis with their clini- ment of Vital and Health Statistics.
cians before being approached for an HF study.
Statistical Analysis
Resuscitation Preferences Data collection was very complete; all variables had
Patient-reported resuscitation status was assessed at less than 1% missing. The level of agreement between
enrollment using the question ‘‘As you probably patient-reported and clinician-ordered resuscitation
Vol. 53 No. 4 April 2017 Resuscitation Preferences in Heart Failure 747

preferences were assessed using percent agreement Table 1


and the Cohen’s kappa coefficient. Baseline character- Patient-Reported and Clinician-Ordered Resuscitation
Preferences
istics are presented as number (%) for categorical var-
iables, mean (SD) for normally distributed Patient-Reported Clinician-Ordered Resuscitation Preference
Resuscitation
continuous variables, and median (interquartile Preference Full Code DNR Not Documented
range) for non-normally distributed variables. Differ-
Full code 190a 21b 2c
ences in baseline characteristics for those with discor- DNR 54b 110a 2c
dant and concordant resuscitation preferences were Unsure 19c 2c 0c
examined using the chi-square test, t tests, and Wil- a
Concordant patient-reported and clinician-ordered resuscitation status.
b
coxon rank-sum tests where appropriate. All analyses c
Discordant patient-reported and clinician-ordered resuscitation status.
Not able to determine concordance/discordance due to patient uncertainty
were performed using Stata, version 13 (College Sta- or undocumented resuscitation status.
tion, TX). A P < 0.05 was used as the level of
significance.
In patients with documented preferences both by
questionnaire and clinician order (n ¼ 375), resuscita-
Results tion preferences were concordant in 80% of patients.
A total of 498 patients were approached for inclu- The Cohen’s kappa coefficient was 0.67, indicating
sion in the study and 400 (80.3%) agreed to partici- good agreement. Of the 75 patients (20%) with discor-
pate (Fig. 1). The distribution of age (mean age 75.8 dant resuscitation preferences, 5.6% elected Full Code
vs. 77.7 years, respectively, P ¼ 0.22) and sex (49% by questionnaire and had a clinician order for DNR
vs. 46% women, P ¼ 0.56) were similar in nonpartici- status and 14.4% elected DNR by questionnaire but
pants and participants. The median (25th to 75th had a Full Code clinician order.
percentile) time from hospital admission to study Compared with those who were Full Code by both
enrollment and completion of questionnaires was questionnaire and clinician order (n ¼ 190), patients
two (1e3) days. Of those surveyed, 213 (53.3%) stated with discordant resuscitation status (n ¼ 75) were old-
their resuscitation preference as Full Code, 166 er (P ¼ 0.010), less often had peripheral vascular dis-
(41.5%) preferred no resuscitation (DNR), and 21 ease (PVD) (P ¼ 0.004), had a trend toward longer
(5.3%) were unsure. In comparison, clinician- hospital length of stay (P ¼ 0.068), and less often dis-
ordered resuscitation status at time of admission was charged to home (P ¼ 0.006, Table 2). Compared with
Full Code in 263 (65.8%) patients, DNR in 133 those who were DNR by questionnaire and clinician
(33.3%), and not documented in four (1.0%, order (n ¼ 110), patients with discordant resuscitation
Table 1). The median (25th to 75th percentile) hospi- status were younger (P < 0.001), more often married
tal length of stay was four (3e7) days. Before hospital (P ¼ 0.001), had better health literacy (P ¼ 0.012),
discharge, changes in clinician-ordered resuscitation lower EF (P ¼ 0.022), less PVD (P ¼ 0.011), longer
status occurred in 19 (4.8%) patients, including 13 length of stay (P ¼ 0.033), and less often had an
changes from Full Code to DNR and six changes advance directive (P ¼ 0.012). Overall, patients with
from DNR to Full Code. discordant resuscitation preferences (n ¼ 75) had a

Fig. 1. Patient enrollment.


748
Table 2
Patient Characteristics Stratified by Resuscitation Preferences
Full Code by Patient DNR by Patient Report P-value
Report and Clinician and Clinician Order Discordant Resuscitation P-value P-value Concordant (n ¼ 300) vs.
Order (n ¼ 190) (n ¼ 110) Status (n ¼ 75) Full Code vs. Discordant DNR vs. Discordant Discordant

Age, years, mean (SD) 73.4 (12.6) 85.0 (9.4) 77.8 (10.8) 0.010 <0.001 0.97
Male, n (%) 112 (58.9) 50 (45.5) 43 (57.3) 0.81 0.11 0.60
EF, mean (SD), % 45 (17) 48 (16) 42.6 (17.7) 0.28 0.022 0.089
Preserved EF $ 50%, 82 (43.6) 60 (55.6) 33 (45.2) 0.82 0.17 0.67
n (%)
Duration of 132 (69.8) 73 (67.0) 54 (72.0) 0.73 0.47 0.59
HF > 18 months, n (%)
NYHA class on admissiona 4 (3, 4) 4 (3, 4) 4 (3, 4) 0.28 0.70 0.55
Advance directive, n (%) 114 (62.0) 94 (87.9) 51 (68.9) 0.29 0.002 0.67
Lived at home before 171 (90.5) 78 (70.9) 62 (82.7) 0.076 0.067 0.90
hospitalization, n (%)
Comorbidities
Hypertension, n (%) 141 (82.9) 86 (88.7) 55 (82.1) 0.88 0.23 0.55
Diabetes mellitus, n (%) 90 (49.5) 42 (38.9) 32 (43.2) 0.37 0.56 0.73
Prior myocardial 58 (31.9) 30 (27.8) 15 (20.3) 0.062 0.25 0.086
infarction, n (%)

Young et al.
Cerebrovascular 37 (20.3) 23 (21.3) 16 (21.6) 0.82 0.96 0.86
disease, n (%)
Peripheral vascular 40 (22.0) 22 (20.4) 5 (6.8) 0.004 0.011 0.004
disease, n (%)
Dementia, n (%) 2 (1.1) 4 (3.7) 2 (2.7) 0.35 0.71 0.74
Creatinine, mean (SD) 1.6 (1.0) 1.5 (0.9) 1.7 (1.2) 0.50 0.22 0.31
Charlson comorbidity 3 (2, 6) 3 (1, 5) 2 (2, 5) 0.017 0.43 0.062
indexa
Psychosocial
characteristics
Married, n (%) 115 (60.5) 41 (37.3) 46 (61.3) 0.90 0.001 0.15
Post-high school 101 (53.2) 62 (36.3) 38 (50.7) 0.90 0.42 0.81
education, n (%)
Poor health literacy, 46 (24.3) 43 (39.1) 17 (22.7) 0.77 0.019 0.22
n (%)
Patient activation levela 3 (2, 3) 2 (2, 3) 2 (2, 3) 0.082 0.52 0.34
Health status fair or 140 (73.7) 80 (72.7) 51 (68.0) 0.35 0.49 0.36
poor, n (%)
Hospital and early post-

Vol. 53 No. 4 April 2017


discharge outcomes
Discharged to home, 149 (80.5) 67 (63.8) 47 (64.4) 0.006 0.94 0.084
n (%)
Hospital length of staya 4 (3, 7) 4 (3, 7) 5 (3, 9) 0.068 0.033 0.037
30-Day mortality, n (%) 10 (5.3) 11 (10.2) 6 (8.1) 0.38 0.46 0.75
EF ¼ ejection fraction; HF ¼ heart failure; NYHA¼ New York Heart Association.
a
Numbers shown are median (25th, 75th percentile).
Vol. 53 No. 4 April 2017 Resuscitation Preferences in Heart Failure 749

longer length of stay (P ¼ 0.037) and less PVD physicians, nurse practitioners, and physician assis-
(P ¼ 0.004) compared with those with concordant tants are often responsible for entering orders for
resuscitation preferences (n ¼ 300). resuscitation preferences on hospital admission and
may lack competence in facilitating these discus-
sions.11,13 Although a single standardized question
Discussion cannot replace a nuanced patient-clinician discussion
In this study of community patients hospitalized about resuscitation preferences, the fact that 20% of
with ADHF, patient-reported and clinician-ordered patients hospitalized with ADHF responded with a
resuscitation preferences were discordant in 20% of resuscitation preference that differed from what had
patients. A higher number of patients reported a pref- been ordered by clinicians in the same hospitalization
erence for no resuscitation than what was ordered by highlights the need for greater consistency in the
clinicians. Patients with discordant resuscitation pref- approach to resuscitation preference discussions.
erences were systematically different than patients Those with discordant resuscitation status were
with concordant Full Code and DNR resuscitation different from those that were Full Code or DNR by
preferences, with variances in age, EF, comorbidities, both patient report and clinician order. In some
completion of advance directives, psychosocial charac- ways, patients with discordant preferences had an in-
teristics, hospital length of stay, and hospital termediate phenotype compared with those with
disposition. concordant Full Code or DNR preferencesdfor
Despite increased awareness of the importance of example, they were younger than those who were
discussion of EOL preferences in patients with HF, DNR but older than those who elected to be Full
the frequency of discordance in patient-reported and Code. Other differences suggested that patients with
clinician-ordered resuscitation status we observed discordant preferences may experience more complex
is similar to older studies.1,10 There are multiple hospitalizations, as they had longer length of stay than
potential reasons why patient-reported and clinician- either concordant group.
ordered resuscitation preferences may be inconsis- This study has limitations that should be acknowl-
tent. First, patients admitted to the hospital are acutely edged to aid in the interpretation of the data. First,
ill, which may require clinicians to infer a patient’s the study was completed in a predominantly Cauca-
resuscitation status to be Full Code until stabilized sian population, which may limit the generalizability
and further discussion can take place. Our study to populations of varying racial and ethnic diversity.
design did not allow us to ascertain if resuscitation However, the characteristics of the study population
preferences were discussed with the patient or the are similar to the state of Minnesota and the Upper
family on admission, and this would be of interest in Midwest United States.14 Although we restricted our
future studies. Furthermore, discussion of resuscita- study to residents of our local community, where
tion preferences on admission to the hospital is most individuals are known to receive some medical
amongst a host of issues addressed in a short amount care at our institution, some residents may seek care
of time. This may be the first time that some patients elsewhere which could lead to selection bias. Patient-
have engaged in dialogue about their resuscitation reported resuscitation preference was assessed using
preference as EOL discussions have been shown to a single question. Although shown to have excellent
occur infrequently in HF.11 As such, their initial pref- test-retest reliability,5 the possibility of patient misin-
erence may change after further self-reflection and terpretation cannot be excluded and could contribute
discussion with loved ones. In this study, there were to the results observed. We did not perform multivari-
very few patients (n ¼ 19) whose ordered resuscitation able models to explore factors that are independently
preference changed during hospitalization. However, associated with discordant preferences because of the
the medical context and prognosis can change quickly relatively small number of observations, which would
in patients hospitalized with ADHF, and patients may be of interest in future studies. We made no mathe-
have changed their preference in response to our matical correction for multiple comparisons made in
question but not communicated that change in prefer- this study.15,16 We found that 13 patient-level factors
ence with their medical team. Finally, patients may were associated with discordant resuscitation prefer-
have chosen a different resuscitation preference as a ences. In comparison, if all the null hypotheses were
result of variation in the way the question was worded true, we would expect 5% (n ¼ 3) of the comparisons
compared with the way the choice was presented by to have P-values less than 0.05.
clinicians. The way in which resuscitation preferences
are discussed can vary12 and may affect the patient’s
decision. These conversations can be awkward and Conclusion
intimidating to clinicians, who often lack the skills to Patient-reported and clinician-ordered resuscitation
conduct effective discussions. In particular, resident preferences were discordant in 20% of patients
750 Young et al. Vol. 53 No. 4 April 2017

hospitalized with ADHF. Most inconsistencies understand prognoses and preferences for outcomes and
occurred in patients who stated a desire to be DNR risks of treatments. Am J Med 1996;100:128e137.
but had a Full Code clinician order in the EMR. The 6. Hibbard JH, Mahoney ER, Stockard J, Tusler M. Devel-
underlying etiology of these inconsistencies may opment and testing of a short form of the patient activation
reflect factors, such as patient indecisiveness or measure. Health Serv Res 2005;40:1918e1930.
patient-clinician miscommunication, and requires 7. Chew LD, Bradley KA, Boyko EJ. Brief questions to iden-
further exploration. tify patients with inadequate health literacy. Fam Med 2004;
36:588e594.
8. Chamberlain AM, Manemann SM, Dunlay SM, et al. Self-
Disclosures and Acknowledgements rated health predicts healthcare utilization in heart failure.
J Am Heart Assoc 2014;3:e000931.
This work was funded by the National Institutes of
9. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new
Health (K23 HL 116643, PI, Dunlay). The authors method of classifying prognostic comorbidity in longitudinal
declare no conflicts of interest. studies: development and validation. J Chronic Dis 1987;40:
373e383.
10. Brunner-La Rocca HP, Rickenbacher P, Muzzarelli S,
References et al. End-of-life preferences of elderly patients with chronic
1. Krumholz HM, Phillips RS, Hamel MB, et al. Resuscita- heart failure. Eur Heart J 2012;33:752e759.
tion preferences among patients with severe congestive 11. Dunlay SM, Foxen JL, Cole T, et al. A survey of clinician
heart failure: results from the SUPPORT project. Study to attitudes and self-reported practices regarding end-of-life
Understand Prognoses and Preferences for Outcomes and care in heart failure. Palliat Med 2015;29:260e267.
Risks of Treatments. Circulation 1998;98:648e655.
12. Deep KS, Griffith CH, Wilson JF. Discussing preferences
2. Lindenfeld J, Albert NM, Boehmer JP, et al. HFSA 2010 for cardiopulmonary resuscitation: what do resident physi-
comprehensive heart failure practice guideline. J Card Fail cians and their hospitalized patients think was decided? Pa-
2010;16:e1ee194. tient Educ Couns 2008;72:20e25.
3. Yancy CW, Jessup M, Bozkurt B, et al. 2013 ACCF/AHA 13. Sulmasy DP, Sood JR, Ury WA. Physicians’ confidence in
guideline for the management of heart failure: a report of discussing do not resuscitate orders with patients and surro-
the American College of Cardiology Foundation/American gates. J Med Ethics 2008;34:96e101.
Heart Association Task Force on Practice Guidelines. J Am
Coll Cardiol 2013;62:e147ee239. 14. St Sauver JL, Grossardt BR, Leibson CL, et al. Generaliz-
ability of epidemiological findings and public health deci-
4. Pakhomov S, Weston SA, Jacobsen SJ, et al. Electronic sions: an illustration from the Rochester Epidemiology
medical records for clinical research: application to the Project. Mayo Clin Proc 2012;87:151e160.
identification of heart failure. Am J Manag Care 2007;13:
281e288. 15. Bender R, Lange S. Multiple test procedures other than
Bonferroni’s deserve wider use. BMJ 1999;318:600e601.
5. Phillips RS, Wenger NS, Teno J, et al. Choices of seri-
ously ill patients about cardiopulmonary resuscitation: corre- 16. Rothman KJ. No adjustments are needed for multiple
lates and outcomes. SUPPORT Investigators. Study to comparisons. Epidemiology 1990;1:43e46.

You might also like