Prevalence of dyspnea among hospitalized patients at the time of admission

Abstract

Context:

Dyspnea is an uncomfortable and distressing sensation experienced by hospitalized patients.

Objectives:

There is no large-scale study of the prevalence and intensity of patient-reported dyspnea at the time of admission to the hospital.

Methods:

Between March 2014 and September 2016, we conducted a prospective cohort study among all consecutive hospitalized patients at a single tertiary care center in Boston, MA. During the first 12 hours of admission to medical-surgical and obstetric units, nurses at our institution routinely collect a patient’s (1) current level of dyspnea on a 0-10 scale with 10 anchored at “unbearable”, (2) worst dyspnea in the past 24 hours prior to arrival at the hospital on the same 0-10 scale, and (3) what activities were associated with dyspnea prior to admission. The prevalence of dyspnea was identified and tests of difference were performed across patient characteristics.

Results:

We analyzed 67,362 patients, 12% of whom were obstetric patients. Fifty percent of patients were admitted to a medical-surgical unit following treatment in the ED. Among all non-critically ill inpatients, 16% of patients experienced dyspnea in the 24 hours prior to the admission. Twenty-three percent of patients admitted through the emergency department reported any dyspnea in the past 24 hours. Eleven percent experienced some current dyspnea when interviewed within 12 hours of admission with 4% of patients experiencing dyspnea that was rated 4 or greater. Dyspnea of 4 or more was present in 43% of patients admitted with respiratory diagnoses and 25% of patients with cardiovascular diagnoses. After multivariable adjustment for severity of illness and patient comorbidities, patients admitted on the weekend or during the overnight nursing shift were more likely to report dyspnea on admission.

Conclusions:

Dyspnea is a common symptom among all hospitalized patients. Routine documentation of dyspnea is feasible in a large tertiary care center.

Introduction

Dyspnea, defined by the American Thoracic Society as, “is a symptom that leads to substantial patient distress” (1-4). Dyspnea is prevalent among multiple patient populations, including those with cardiovascular disease, primary pulmonary disease, cancer, metabolic disturbances and acidosis, and at the end of life, which suggests this powerfully troubling symptom is a common outcome for many disease processes (5-13). Further, in both the outpatient and inpatient setting, dyspnea has been identified as an alarm of possible patient harm and poor outcome (11, 12, 14).

Despite the potential significance of dyspnea to both patients and healthcare providers, uniform measurement and documentation in hospitalized patients is rare. Dyspnea is defined as the patient’s experience of his or her own breathlessness, therefore measurement of dyspnea requires asking the patient (1). Our hospital has implemented unidimensional assessments of patient dyspnea both on admission to all inpatient units in a single institution and at least once per nursing shift; we reported the prevalence of dyspnea in a small sample from four selected medical-surgical floors in a previous pilot study (15-17). In the present we report the magnitude of the burden of dyspnea among a very large sample of patients admitted to all medical and surgical units.

We describe the prevalence of dyspnea prior to hospitalization and on arrival to the hospital among inpatients at a single tertiary care facility in Boston, MA. Further, we describe how patient-reported dyspnea varied by demographic and clinical features.

Methods

Study Population and Data Source

We conducted a retrospective cohort study. We included all consecutive admissions to a single tertiary care facility with 651 inpatient beds between March 2014 and September 2016. We restricted our study sample to patients 18 or greater years of age and to patients who were admitted to non-intensive care unit floors. Patients were included if they were able to verbally participate in evaluating their on-going dyspnea; non-verbal patients were excluded.

Our study was approved by the institutional review board at the Beth Israel Deaconess Medical Center with a waiver of informed consent.

Study Variables

Assessment of dyspnea

Our institution implemented measures of dyspnea for all patients on admission to the hospital as part of the comprehensive assessment of the patient’s current and past symptoms and functional status obtained by nursing staff during the patient’s first 12 hours on the admitting unit, called the Initial Patient Assessment (IPA). (See Supplement Figure S1 for a screen shot of the dyspnea items on the electronic form). As part of a full assessment, nurses asked their patients to respond to the following items: 1) report their current breathing discomfort at rest on a 0 to 10 scale where 10 is “unbearable”, 2) report on the same 0 to 10 scale their worst level of breathing discomfort in the 24 hours prior to arrival at the hospital (0 to 10 scale), 2a) If the patient gave a non-zero rating, they were asked what they were doing when they experienced this dyspnea; nurses then categorized the activity as one of four levels: resting (e.g., sitting in a chair, lying in bed), light activity (e.g., eating, dressing), moderate activity (e.g,, walking, making the bed), or heavier activity (e.g.. mowing the lawn, raking leaves). Self-reported pain on a 0-10 scale was also documented during the initial patient assessment. Dyspnea and pain assessments were required fields on the electronic form.

Our primary outcome was the prevalence of dyspnea among all hospitalized patients on arrival to the hospital and in the past 24 hours. The outcomes were further dichotomized by patients with dyspnea 4 or greater; this value has been associated with substantial dyspnea in prior research conducted previously among 4 pilot medical-surgical units within our hospital (14). Admissions were excluded from analysis when the primary outcome was missing (see Results).

Patient Demographics and Clinical Characteristics

Demographic information, including age, race (black, white, or other, as identified by the patient on admission), and sex, was collected for all hospitalized patients.

We report patients’ clinical characteristics including the service of admission, the admission location, comorbidities (extracted using the Elixhauser method (18) which uses billing data to identify concomitant diagnoses and included as separate, binary variables), and severity of illness (using patient-level casemix, which estimates severity of illness as reflected by higher reimbursement to the hospital(19)). Admission diagnosis was further categorized using the multiple classification system as identified by Cowen and colleagues (https://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp) (20). The characteristics of patients’ hospital stays included total length of stay, weekend versus weekday admission, and admission during night shift (7p-7a) or day shift (7a-7p).

Statistical Analysis

All statistical tests were formed using SAS (v. 9.4, SAS Institute Inc., Cary, NC). The unit of analysis was admission with each admission treated as a unique value (prevalence of readmissions included in Results). We first estimated overall prevalence of current dyspnea and dyspnea during the prior 24 hours, both treating the values as ordinal values, and dichotomizing the 10-point scale into dyspnea < 3 vs 4 or greater in a second analysis. The values of dyspnea were treated as continuous values to identify estimates of central tendency by activity prior to admission. We then compared any dyspnea (dyspnea > 0) or substantial dyspnea (dyspnea greater than or equal to 4) by patient characteristics using chi-squared tests for categorical variables and Kruskal-Wallis test for continuous variables. A 2-sided type I error of 0.05 or less was used to indicate statistical significance for all comparisons.

We hypothesized that patients admitted overnight or on the weekend were likely to be sicker than patients admitted at other times. We also hypothesized that patients admitted through the emergency department would report dyspnea more frequently than patients admitted from other locations in the hospital (e.g., same day surgery, directly from primary care offices, other hospitals). We used multivariable logistic regression to risk adjust the odds of reporting any dyspnea or dyspnea greater than or equal to 4, using patient-level casemix and Elixhauser comorbidities, with each Elixhauser variable included as a separate, independent measure in the model as advocated by Elixhauser (18).

Results

86,638 admissions occurred between March 2014 and September 2014. Of these, 10,991 patients were excluded from the study as the patients were either considered under “observation” or were directly admitted to the intensive care unit without completing an initial patient assessment. An additional 7,579 admissions did not complete an initial patient assessment; 5,984 of these were admissions to the obstetric service. 706 patients were unable to respond to the nurse about their levels of dyspnea. In total, 19,276 subjects were excluded.

We studied a total of 67,362 admissions between March 25, 2014, and September 30, 2016, which represents 46,481 unique patients. The mean age of patients was 58 (SD 19), 57% were women, and 31% were nonwhite. The median length of stay was 3 days (IQR 4). Ten percent of patients were non-English speaking. Twelve percent of admissions were to the obstetric service, 51% to medicine or medicine specialties, and 24% to surgery or surgical specialist services. Five percent of all admissions were for respiratory diagnoses and 16% were for cardiac diagnoses. The most prevalent comorbidities among this cohort included hypertension (48%,32436/67362), chronic pulmonary disease (18%, 11893/67362), and depression (17%, 11359/67362). Of the total admissions during this time period, 20% (9286/46,481) of patients had more than one admission in the following 365 days and 9% (4069/46,481) had more than two admissions (Table 1).

Table 1.

Participants’ Characteristics

	N	%
TOTAL	67362
Female	38248	56.8
Age, Year (mean, SD)	58.0 ±18.8
Race
Black	9666	14.4
White	46521	69.1
Other	11175	16.6
Ethnicity-Hispanic	4286	6.4
Day of admission
Weekday	55564	82.5
Weekend	11798	17.5
Time of admission
Day (7a-7p)	33975	50.4
Night (7p-7a)	33387	49.6
Service Type
Medicine	33988	50.5
Surgery	16421	24.4
Obstetrics	7911	11.7
Other	9042	13.4
Admission Source
Elective	14367	21.3
Emergency	44273	65.7
Urgent	8722	12.9
Admission from emergency department	34073	50.6
Length of Stay (days, median, ±IQR)	3 ±4
Admission Diagnoses *
Disease of the circulatory system	10910	16.2
Diseases of the digestive system	8371	12.4
Complications of pregnancy, childbirth	8006	11.9
Comorbidities**
Hypertension	32436	48.1
Depression	11359	16.8
Chronic pulmonary disease	11893	17.6

^*List includes 3 most common admit diagnoses in the study population

^**List includes 3 most common comorbidities in the study population

Dyspnea at the time of the IPA interview

Within 12 hours of admission, 11% of all hospitalized patients reported current dyspnea >0 on admission and 4% (2483/67,362) of patients reported a rating of 4 or greater (out of 10) on admission; among non-obstetric patients, the prevalence of dyspnea was 12% and of any dyspnea rated ≥4 was 4% (2476/59,451) (Figure 1, full distribution of patient ratings available in the Supplement Figure S2A).

Figure 1.

Prevalence of any dyspnea and dyspnea 4 among patients on admission (Figure 1A) and over the past 24 hours (Figure 1B). Patients (N=67,362) were approached during the first 12 hours after admission. Dyspnea was evaluated on a 0(none)-10(unbearable) point scale.

Patient characteristics of those reporting dyspnea ≥4, considered to have substantial dyspnea, are presented in Table 2. Nearly three-fifths of patients who reported substantial dyspnea were admitted for cardiovascular or respiratory diagnoses. Among all patients admitted with cardiovascular illnesses, 7% reported dyspnea ≥4 and of the patients admitted with respiratory disease, 21% reported dyspnea ≥4 (Figure 2).

Table 2.

Dyspnea by participant characteristics

	Overall	Dyspnea on admission≥ 4*		P-value	Overall	Dyspnea in past 24 hours ≥ 4*		P-value
	N	N	%		N	N	%
TOTAL	67362	2483	3.7		66809	6412	9.6
Gender
Female	38248	1289	3.3	<.0001	37739	3241	8.6	<.0001
Male	29114	1194	4.1		29070	3171	10.9
Age (Mean, SD)	67362	65.4 vs 57.7* (15.9) (18.9)		<.0001	66809	66.4 vs 57.3* (15.7) (18.8)		<.0001
Race
Black	9666	514	5.3	<.0001	9596	1206	12.6	<.0001
White	46521	1720	3.7		46255	4551	9.8
Other	11175	249	2.2		10958	655	6.0
Ethnicity
Hispanic	4286	162	3.8	0.73	4256	398	9.3	0.57
Non-Hispanic white	63076	2321	3.7		62553	6014	9.6
Day of admission
Weekday	55564	1934	3.5	<.0001	55127	5045	9.1	<.0001
Weekend	11798	549	4.6		11682	1367	11.7
Time of admission
Day (7a-7p)	33975	1101	3.2	<.0001	33729	2856	8.5	<.0001
Night(7p-7a)	33387	1382	4.1		33080	3556	10.7
Service Type
Medicine	33988	2166	6.4	<.0001	33913	5626	16.6	<.0001
Surgery	16421	267	1.6		16414	693	4.2
Other	16953	50	0.3		16482	93	0.6
Emergency Admission
No	33289	510	1.5	<.0001	32811	1453	4.4	<.0001
Yes	34073	1973	5.8		33998	4959	14.6
DRG Cost Weight (Mean, SD)	67362	1.7 vs 1.6* (1.8) (1.4)		.01	66809	1.8 vs 1.6* (1.8) (1.4)		<.0001
Pain Score
<4	43154	1394	3.2	<.0001	42686	4426	10.4	<.0001
≥4	24208	1089	4.5		24123	1986	8.2

^*Continuous dyspnea rating (0-10) dichotomized into dyspnea ≥4 and dyspnea <4. Chi-square difference between Dyspnea≥4 and Dyspnea<4

Figure 2.

Prevalence of current dyspnea 4 stratified by admission diagnoses using single-level Clinical Classical Software (CSS). Patients (N=67,362) assessed for current dyspnea during the first 12 hours after admission to surgical and medical units.

Patients admitted during a night shift or on the weekend were significantly more likely to report any dyspnea, even after adjustment for severity of illness and comorbidities (night admission versus day admission: unadjusted relative risk (RR) 1.4, 95% CI 1.3-1.4, p<0.0001; adjusted RR 1.3, 95% CI 1.2-1.3, p<.0001; weekend admission versus weekday admission: unadjusted RR 1.3, 95% CI 1.2-1.4, p<0.0001; adjusted RR 1.3, 95% CI 1.2-1.3, p<0.0001).

Dyspnea in the past 24 hours prior to hospital arrival

16% of all patients reported any dyspnea in the 24 hours prior to admission and 10% of all patients reported dyspnea of ≥4 (Figure 1B, full distribution of patient ratings available in Supplement Figure S2B). These frequencies were slightly greater when obstetric patients were removed from the sample (18% and 11%, respectively). Although the majority of the patients who reported dyspnea prior to hospital arrival were engaged in at least moderate physical exertion at the time, about a quarter experienced their worst dyspnea at rest or during light activities of daily living. Notably, the patients who experienced dyspnea during minimal activity reported slightly more dyspnea (Figure 3); a possible explanation is put forth in the Discussion section.

Figure 3.

Dyspnea reported in the past 24 hours, stratified by associated activity level. Median and IQR presented. Patients (N=10,521) were asked what they were doing when experienced dyspnea in the past 24 hours; resting (e.g., sitting in a chair, lying in bed), light activity (e.g., eating, dressing), moderate activity (e.g,, walking, making the bed), heavier activity (e.g.. mowing the lawn, raking leaves).

Of the patients reporting any dyspnea prior to hospital arrival, 40% reported that their dyspnea had increased in the past week.

There were demographic differences in the frequency of reported dyspnea in the past 24 hours. Men were more likely than women to report recent dyspnea of 4 or greater (11% vs 9%, p<0.0001; African-Americans were more likely than white patients to report recent dyspnea ≥ 4 (13% vs 10%, p<0.0001).

Prevalence of dyspnea among admission diagnoses

Cardiovascular diagnoses were the most common admission diagnoses to our hospital for this cohort. 43% of patients with respiratory diagnoses and 25% of patients with cardiovascular diagnoses reported dyspnea ≥ 4 in the 24 hours prior to hospital arrival; these are significantly greater than reported for this time frame by patients with other diagnoses (5%, p<0.001). The prevalence of substantial dyspnea across all diagnoses and across specific respiratory and cardiac diagnoses is displayed in Supplement Figure S3.

Prevalence of dyspnea with planned vs emergency admission

Nearly 16% of patients admitted through the emergency department reported current dyspnea during the first shift and 23% of patients admitted through the emergency department described any dyspnea (>0) in the 24 hours prior to admission. Patients admitted through the emergency department were 3.7 times (p<0.001) more likely to report dyspnea of ≥ 4 at the time of admission than non-emergency admissions (5.8% versus 1.5%). Similarly, patients were 3.3 times (p<0.001) more likely to report moderate to severe dyspnea in the past 24 hours versus non-emergency admissions (14.6% versus 4.4%).

Prevalence of pain and overlap with dyspnea

Among non-obstetric patients, any current pain (> 0) on admission occurred at a frequency of 53% and substantial pain (≥ 4) at a frequency of 39% among all hospitalized patients. Among patients with substantial pain, 5% simultaneously reported substantial dyspnea. Among patients with respiratory diseases, 12% reported both substantial dyspnea and pain together.

Discussion

Our study represents the first large-scale quantification of the prevalence of dyspnea among hospitalized patients. As many as 12% of patients admitted to general medical and surgical floors report dyspnea at rest, a third of whom report moderate to severe dyspnea. The prevalence of this symptom suggests a common and unrecognized symptom burden among patients.

Large-scale measurement of dyspnea on admission to the hospital is feasible. Nurses at our institution almost universally endorsed the ease of dyspnea assessment and documentation, cited its importance to patient-centered care, and found it did not impair workflow (21). Although our laboratory is a leading proponent of multidimensional dyspnea measurement, we recognized that a unidimensional scale is more appropriate for the purpose of universal screening (22, 23). The patients included in this population represent all patients directly admitted to non-intensive care units throughout our hospital over a two-year period. Baker et al described the process of implementing large-scale dyspnea measurement throughout an academic medical center; nurses enthusiastically endorsed the opportunity to further characterize patients’ symptoms (15).

Current dyspnea at the time of IPA interview

At the time of interview, patients were seated or in bed, thus these ratings represent dyspnea without exertion. We found that 4% of non-obstetric patients reported dyspnea ≥4/10. The prevalence of dyspnea ≥4 in this hospital-wide sample was substantially lower than the 13% prevalence reported in our pilot study of four medical-surgical units (14), which may be due in part to the fact that we used a written data collection strategy in the pilot phase but had incorporated dyspnea measurement into the electronic system used by all nurses during the main study.

In the present data set, patients who rated current pain ≥4/10 were about 10 times as prevalent as patients who rated dyspnea of similar magnitude; pain prevalence was about the same as seen in prior studies of pain in hospitalized patients (24, 25).

Dyspnea prior to hospital arrival

To estimate the level of dyspnea before treatment, we asked patients to recall their worst dyspnea in the 24 hours prior to arrival at the hospital. This recalled dyspnea rating cannot be interpreted without some knowledge about the level of physical activity at the time because dyspnea depends on cardiorespiratory demand in relation to cardiorespiratory delivery. That is, more exertion provokes more dyspnea. Most patients reported that they were engaged in moderate or heavier exertion at the time they experienced worst dyspnea; dyspnea ratings in these patients were generally modest (median 4/10, upper quartile 6/10). This should not be surprising, as people commonly modulate their physical activity in order to avoid dyspnea levels above midscale (5). In contrast, those patients who reported dyspnea at rest or with light activities of daily living reported more intense dyspnea (median 5/10, upper quartile 7-8/10), presumably because these patients were unable to reduce their activity and thus were unable to avoid dyspnea.

In the present study, 16% of patients reported any dyspnea (i.e., rated ≥1/10) in the 24 hours prior to hospitalization. This number is lower than we expected, as previous studies of dyspnea in the general public and in outpatients have reported somewhat higher rates of any dyspnea on exertion, including estimates close to 25% (26, 27). Among patients without heart or lung disease, Santos et al found about 3% of patients reported moderate dyspnea, similar to our findings of 4% of hospitalized patients (28). In part, the greater prevalence of any dyspnea may be due to differences in the way dyspnea was assessed. Most such studies have employed some version of the MRC dyspnea scale (29), which asks what level of exertion provokes dyspnea (dyspnea intensity unspecified). We asked what level of dyspnea had been experienced, without specifying a priori the activity – thus many patients who chose not to exert themselves probably avoided dyspnea altogether and reported experiencing no dyspnea.

Our present data are limited to ordinary medical surgical units, thus do not include many of the sickest patients who are admitted to intensive care. In high-risk groups such as heart failure and respiratory disease requiring ventilator support, other authors have found much higher rates of dyspnea, approaching 50-70% (30, 31).

We also found differences in the prevalence of dyspnea among patients of different gender, age, and race. Both patient and clinician gender and race appear to influence expectation of and treatment for pain (32-35). Similar to our findings in hospitalized patients, Santos et al noted a greater prevalence of dyspnea among both female and older community-dwelling people (28). Our data cannot address the question of whether this constitutes a difference in reporting, a difference in perceived experience, or an interaction between the patient reporting the measure of dyspnea and the nurse documenting it.

Caveats to the interpretation and understanding of these data

We used administrative data to identify diagnoses, which lack the clinical nuance of a manual record review. This method is imperfect in classifying diagnosis; other authors have demonstrated how billing patterns have changed over time even when disease incidences have not (36). We elected to use billing data as it was beyond our resources to extract admitting diagnoses from record review for a large cohort.

Dyspnea and pain data were collected by nurses as part of the usual clinical routine, not by dedicated research staff. The data therefore reflect the clinical setting, including practical realities that may be less than ideal:

1) Ideally, all patients would provide a numerical rating for dyspnea and for pain. However, to streamline the task, many nurses in our institution first ask patients a yes/no question about dyspnea and pain such as “do you have any breathing discomfort [or pain]?” and if the response is “no”, the nurse documents dyspnea (or pain) as zero on the rating scale (21). This may not capture full information, as there is a sensory threshold for a yes answer. Psychophysical studies have shown that a subject’s decision to say ‘yes’ to a binary question requires a finite magnitude of sensation: this “decision criterion” can vary with the situation and among individuals (37). Thus, it is likely that some subjects responding ‘no’ would rate 1, 2, or 3 on the scale (38). Our data are likely, therefore, to somewhat underestimate the prevalence of low levels of dyspnea (and pain). Substantial discomfort (>3/10) is much less likely to fall below the decision criterion, thus is less likely to be underestimated.

2) Ideally, dyspnea ratings should rest entirely on the patient’s report. Nurses, however, sometimes use observed signs of respiratory distress to supplement the patient’s report; the most common reasons for the use of observed signs were because the nurse judged that the patient was unable to sensibly use number scales to report pain and dyspnea or because the patient was unresponsive (21). In the future, addition of another scale for the nurse to report ‘observed respiratory distress’ would clarify the source of the information.

With regard to those alert patients who are judged unable to use a numerical scale: we recognize that 10-15 % of laboratory subjects are unable to sensibly use a number scale to rate sensations, and this is likely to be true in the patient population. Although these “poor raters” are usually excluded from sensory psychophysics experiments, we must accept their presence in the clinical population as another contribution to ‘measurement noise’. Surveyed nurses generally agreed that patients unable to rate dyspnea were the same patients who were unable to rate pain (21). Clinicians should be alert to the ‘poor rater’ problem when assessing individual patients.

3) Ideally we would have obtained a dyspnea rating at the moment the patient arrived at the hospital, but this was not possible. Current dyspnea as measured during the first inpatient nursing shift probably underestimates the prevalence of dyspnea on arrival at the hospital. A major factor is that about half of patients are admitted to our institution through the emergency department, with 10% of patients transferred from smaller surrounding hospitals. Our emergency department does not document patient-reported dyspnea ratings, nor do most surrounding hospitals. These patients are likely to have had treatment prior to their initial interviews with floor nurses. For logistical reasons, many nurses delay conducting the initial patient assessment until there is a quiet moment in the shift, again increasing the likelihood that patients already have had treatment. To help circumvent this gap in coverage, the IPA includes a question about the worst dyspnea experienced in the 24 hours before arrival at the hospital; recall for dyspnea in this time frame is reliable (39).

Conclusions

In our tertiary care hospital, nearly 1 in 20 patients are admitted with ongoing dyspnea of at least moderate severity. Dyspnea spanned multiple diagnoses, although it was significantly more common among patients admitted with respiratory and cardiovascular diseases. Puntillo et al reported that, although dyspnea was neither the most prevalent nor the most intense symptom in ICU patients, it did cause the greatest distress (40). Virtually all hospitals quantitatively assess and document pain routinely. Current dyspnea was a good deal less common than current pain, but nonetheless dyspnea affects a large number of patients, and the attendant suffering may be greater. Dyspnea and visceral pain activate the amygdala (41) and evoke emotions related to existential threat not inherent to peripheral pain, e.g., “when the shortness of breath was at its extreme, I thought I was going to die and saw a coffin beside me.” (4) Given the observation that assessment and documentation of current dyspnea requires less than a minute and is strongly supported by nurses (21), and given the potential for suffering, we suggest that dyspnea, like pain, should be routinely assessed and documented.