PREDICTION OF RECOVERY FROM ACUTE ASTHMA .... subjects with data for the key variables were in- ..... and 4) the description of a possible tool for facili-.
28
ASTHMA RECOVERY
McCarren et al. • PREDICTION OF RECOVERY FROM ACUTE ASTHMA
Predicting Recovery from Acute Asthma in an Emergency Diagnostic and Treatment Unit MADELINE MCCARREN, PHD, MPH, ROBERT J. ZALENSKI, MD, MA, MICHAEL MCDERMOTT, MD, KULVINDER KAUR, MPH
Abstract. Objective: Optimal use of emergency diagnostic and treatment unit (EDTU) resources for treatment of acute asthma should be facilitated by the selection of patients with a high probability of discharge from the EDTU. The study goal was to identify characteristics of the patient or exacerbation that could be used to predict recovery of pulmonary function within 12 hours. Methods: Comprehensive cohort design in an urban public hospital. The subjects were 269 patients with moderately severe asthma exacerbations. Data were collected for historical and presenting features and response to treatment over 12 hours. Two outcomes were examined: 1) discharge from the EDTU and 2) achieving 50% predicted peak expiratory flow rate (PEFR) within 12 hours. Results: The two outcomes showed good concordance. The third-treatment PEFR was found to be predictive of both discharge and reaching 50% predicted PEFR within 12 hours. Since the objective
M
OST patients with acute asthma need neither hospital admission for days nor prolonged treatment (12 hours or more) in an emergency diagnostic and treatment unit (EDTU).1 In the more severe patients who do need more than just a few hours of initial treatment, discriminating between those who need only EDTU treatment and those requiring inpatient admission is a chal-
From the Center for Health Services Research (MMcC, RJZ, MMcD, KK), and Department of Epidemiology, School of Public Health (MMcC), University of Illinois at Chicago, Chicago, IL; Departments of Emergency Medicine (RJZ, MMcD) and Internal Medicine (MMcD), Cook County Hospital, Chicago, IL; Cooperative Studies Program Coordinating Center, Hines Veterans Affairs Hospital, Hines, IL (MMcC); and Department of Emergency Medicine, Wayne State University, School of Medicine, Detroit, MI (RJZ). Received February 26, 1999; revision received August 16, 1999; accepted August 30, 1999. Presented at the SAEM annual meeting, Chicago, IL, May 1998. Supported by grants from the Agency for Health Care Policy and Research (HHS HS07103 and HHS HS07969), the Center for Health Services Research of the University of Illinois at Chicago, and Cook County Hospital. Address for correspondence and reprints: Madeline McCarren, PhD, MPH, Hines Veterans Affairs Hospital, P.O. Box 5000, Cooperative Studies Program Coordinating Center 151K, Hines, IL 60141-5151. Fax: 708-202-2324; e-mail: mccarren@ research.hines.med.va.gov
measure of reaching 50% predicted PEFR is more readily defined and thus more generalizable, the authors focused on this outcome when describing prediction zones. Patients with 40% or higher PEFR after third treatment had an 89% probability of reaching 50% predicted in 12 hours, while those with a third-treatment PEFR lower than 32% predicted had only a 22% probability. Conclusions: A simple objective measure of pulmonary function early in treatment discriminated among those with high, low, and intermediate probabilities of achieving a specified level of PEFR within 12 hours. Awareness of this probability could assist clinicians attempting to predict discharge from the EDTU and facilitate decision making regarding utilization of EDTU resources. Key words: prognosis; probability; asthma; peak expiratory flow rate; emergency medicine; models, statistical. ACADEMIC EMERGENCY MEDICINE 2000; 7: 28–35
lenging issue.2 Many studies have attempted to specify an emergency department (ED) treatment response that results in a successful discharge from the ED within a few hours.3–6 However, these findings have questionable relevance to the type of patient treated in an EDTU, i.e., those who are still not dischargeable after the first few hours of treatment. Extrapolation of findings from ED studies to the EDTU is difficult also because of the outcomes considered relevant in the ED, as well as the spectrum of patients studied. For instance, short-term relapse is often included in the definition of poor response to ED care.6–8 However, a mixed outcome that includes both response during care and relapse following discharge may be problematic, as it has since been shown that different factors are associated with discharge3,4,9,10 and relapse.11–15 Furthermore, the early relapse rate under the modern EDTU protocol studied is very low, and it is comparable to the rate seen in patients who are admitted.1 This indicates that the early relapse rate cannot be further lowered by admitting patients rather than treating in the EDTU, and therefore early relapse is not a relevant outcome for the purpose of a rule predicting appropriate utilization of the EDTU. In addition, many earlier
ACADEMIC EMERGENCY MEDICINE • January 2000, Volume 7, Number 1
studies have included patients for whom a rule is not needed, i.e., those patients at the extreme ends of the exacerbation spectrum.3,4,7,10 Including such patients tends to exaggerate a rule’s performance. For these reasons, it is necessary to examine predictors of response to emergency care specifically in those asthmatic individuals who are candidates for EDTU care. The goal of this study was to identify characteristics of the patient or the asthma exacerbation that could be used to predict recovery of pulmonary function within 12 hours.
METHODS Study Design. This was an analysis of asthma patients treated in the EDTU using a comprehensive cohort design including both subjects enrolled in a randomized controlled trial and subjects who were eligible for but did not consent to randomization, yet were followed concurrently.12,16,17 The primary findings in the randomized arms of the main source study1 and secondary analyses using the comprehensive cohort have been published previously.12 Consent to randomization was not associated with the rate of recovery. The trial was approved by the institutional review board, and all patients gave written informed consent to be randomized or simply followed, if the patient did not choose randomization, under the treatment protocol. Study Setting and Population. Subjects were asthma patients enrolled in a comprehensive cohort study conducted from December 1992 through April 1995. In the current investigation, only those subjects with data for the key variables were included (n = 269, 88% of the full cohort). Patients were included only once. We selected for study only those patients who failed to reach discharge criteria within three hours in the EDTU but were not in imminent danger of ventilatory failure. Eligible patients presented to the EDTU for asthma exacerbation, had a history of asthma,18 and were aged 18–55 years. Patients were excluded if they had any of the following: a pCO2 ⱖ 45 mm torr or pO2 ⱕ 55 torr; a peak expiratory flow rate (PEFR) ⱕ 80 L/min after the first beta-adrenergic treatment; asthma onset after age 45 and a ⱖ20-pack-year history of smoking; an EDTU-documented best PEFR less than their discharge criterion (since such subjects are unlikely to ever meet targeted PEFR); pregnancy; or a diagnosis of pneumonia, congestive heart failure, or restrictive lung disease prior to eligibility assessment. Study Protocol. Treatment of the index exacerbation was based on study protocol.1 In the first
29
three hours, all patients were treated in the EDTU and received beta-adrenergic agonists via a handheld nebulizer (HHN) at 0, 30, 60, and 120 minutes after presentation and systemic corticosteroids within the first hour. At three hours, consenting patients were randomized to EDTU or admission in the source study.1 A subset of additional patients specifically desiring EDTU care were not randomized but simply followed. The EDTU patients were then scheduled to receive further HHN treatments every two hours and a repeat steroid dose (half the previous dose) at hour six, but could be discharged whenever discharge criterion was achieved. Admission was advised for EDTU patients who failed to reach discharge criterion within nine additional hours (12 hours total). The inpatient protocol1,19 specified three HHN treatments every two hours after admission then four times daily, target serum theophylline levels of 8–12 g/mL for all patients, and corticosteroids upon arrival to the ward and every six hours thereafter. Other medications were discretionary. The PEFRs (best of three attempts) were collected before and after HHN therapy and are expressed as percent of predicted.20 Data were collected on historical and presenting features and response to treatment over 12 hours. Two outcomes were determined: discharge from the EDTU and achieving a 50% predicted PEFR within 12 hours. Data Analysis. In 18 cases (7%), a missing value for the third-treatment PEFR was estimated as the mean of the PEFRs before and after (i.e., secondand fourth-treatment PEFRs). This procedure was validated by inspection of cases with nonmissing data. Survival analysis was used to estimate the time to reach outcome criterion and the log-rank test (PROC LIFETEST, SAS Institute Inc., V6.07, Cary, NC) to test differences between groups. Observations were censored at 12 hours or when subjects were lost to follow-up (e.g., those leaving against medical advice). Identification of individual characteristics that might be predictive of treatment response (i.e., discharge or meeting 50% predicted PEFR within 12 hours) began in phase I with calculations of crude incidence rate and hazard ratios (HRs).21 In phase II, a multivariable model of predictors was developed using the Cox proportional hazards model. Backward elimination procedures (PROC PHREG) were used for variable selection, with exclusion pvalues set at >0.05. The initial model was developed using the EDTU population. In phase III, the model was tested for stability on a group treated under a slightly different protocol (inpatients) and appropriate modifications were made to the model. Next, to illustrate the value of the prognostic variable identified in the model building step, re-
30
ASTHMA RECOVERY
McCarren et al. • PREDICTION OF RECOVERY FROM ACUTE ASTHMA
TABLE 1. Demographic and Clinical Characteristics of the
Selection of Predictive Variables
Emergency Diagnostic and Treatment Unit (EDTU) Group at Baseline Patient Characteristics*
Value
n
Gender—male Race—African American Age (years) Smoker Employed Hospitalization for asthma in past year
60% 89% 35 (26–44) yr 37% 41%
164 158 169 159 169
46%
170
10%
168
39%
170
14% 66%
169 149
68%
143
24 (20–26)
144
96 (88–108)
143
37% 36% 34% 40% 49 (33–68)
148 159 161 158 170
26 (21–30)
166
Index visit is relapse (2nd within 10 days) Three or more ED visits within past 6 months Systemic steroids >180 of previous 365 days Ever took prednisone Currently taking theophylline Respiratory rate at presentation (breaths/min) Heart rate at presentation (beats/min) Chest congestion duration ⱖ3 days Shortness of breath ⱖ3 days Wheezing duration ⱖ3 days Cough duration ⱖ3 days Dyspnea score (scale 0–100) Presenting PEFR (% predicted)
*Patient characteristics at presentation are shown. Values shown are the percentage of patients with the characteristic or the median (interquartile range) for the measure and the number of patients contributing to the estimate. PEFR = peak expiratory flow rate.
sult-specific likelihood ratios (LRs)22 were calculated, combining all patients (EDTU and inpatient). In the LR analyses reported, subjects lost to follow-up before meeting criterion were classified as not meeting criterion. This was done only after determining that the findings with these subjects excluded were similar.
RESULTS Of the 304 patients enrolled in the comprehensive cohort, 269 (88%) had sufficient data for this report. Baseline characteristics for the EDTU population are shown in Table 1. The outcome of the acute exacerbation is described by two recovery measures in Figure 1. By 12 hours, the probability of reaching 50% or higher predicted PEFR was 63%, while the probability of being discharged was 54%. The rate of discharge did not begin to rise until after three hours, reflecting an eligibility criterion for enrollment.
Phase I: Univariate Analysis Using the EDTU Sample. The EDTU sample was used for initial model building, and predictors were determined for each of the two outcomes in turn. We examined seven possible predictors, based on what was suggested in the literature: four presenting variables (PEFR, patient-reported dyspnea score, heart rate, and respiratory rate); one response variable (PEFR after the third treatment); and two historical variables associated with relapse in this population (three or more ED visits for exacerbations in the preceding six months and difficulty performing work or other activities as a result of physical health in the previous four weeks).12 The results of this first phase are shown in Table 2. Regardless of the outcome, the PEFR after the third treatment was the strongest predictor. Patients with PEFRs higher than 36% predicted (the median) after the third treatment were three to four times more likely (HRs: 3.1–4.2) than those with lower PEFRs to recover within 12 hours to the point of discharge or 50% predicted. A low presenting dyspnea score (indicating less dyspnea) was associated with a somewhat higher probability (HRs: 1.6–1.9) of achieving criteria than was a high dyspnea score. Presenting PEFR was predictive when the outcome was 50% predicted, but not for the outcome of discharge. No other variables were associated with the outcomes. Phase II: Multivariable Analysis Using the EDTU Sample. Next, a multivariable procedure was used to identify factors that were independently prognostic. Only subjects with data for all seven candidate variables were included (n = 135). For either outcome, only two of the seven variables, third-treatment PEFR and presenting dyspnea score, made independent contributions to predicting recovery. Presenting PEFR was no longer significant (HR: 1.2, p < 0.4). Phase III: Model Refinement Using the EDTU and Inpatient Samples. The HR estimates for the two-variable models are shown in Table 3, recalculated using the full EDTU data set for greater precision (EDTU columns). For either outcome, the third treatment PEFR was a stronger predictor than dyspnea score. In order to explore the generalizability of these predictors, we evaluated this two-variable model using the first 12 hours of data from similar subjects treated in a different setting, i.e., an inpatient unit. As previously reported, the baseline characteristics did not differ between those randomized to the EDTU and inpatient arms of the source study.1 For this analysis, only the outcome of 50% predicted PEFR within 12 hours could be used, as discharge assessments are not made in
31
ACADEMIC EMERGENCY MEDICINE • January 2000, Volume 7, Number 1
Figure 1. Survival curves (Kaplan-Meier) showing the probability of emergency diagnostic and treatment unit (EDTU) patients’ meeting criteria within 12 hours. Closed circles represent meeting 50% predicted peak expiratory flow rate (PEFR); open circles represent meeting discharge criterion. The numbers remaining at each time are shown below the graph.
inpatients within 12 hours. The inpatient group was somewhat less likely (p < 0.023) to reach 50% predicted within 12 hours (48%) compared with the EDTU patients (63%). In the inpatients, the PEFR after the third treatment was a significant predictor (HR = 9.0), but presenting dyspnea score was not (Table 3, Inpatient).
Since the dyspnea score was not a strong and robust predictor, we focused the remainder of the analysis on the early treatment variable. Furthermore, since the two recovery indicators showed good concordance, we have concentrated on the outcome of 50% predicted PEFR rather than discharge. This is the more objective, easily described
TABLE 2. Univariate Analysis of Characteristics Associated with Meeting Peak Expiratory Flow Rate (PEFR) or Discharge Criteria within 12 Hours in Emergency Diagnostic and Treatment Unit (EDTU) Patients Dependent Variables* Independent Variables
n
†Presenting PEFR >26% predicted †Presenting dyspnea score ⱕ49 (scale 0–100) †Presenting heart rate ⱕ98 beats/min †Presenting respiratory rate ⱕ24 breaths/min †PEFR after 3rd treatment >36% predicted Three or more ED visits in last 6 months Difficulties due to physical health
166 170 143 144 170 170 169
50% Predicted PEFR 1.9‡ 1.6‡ 1.1 1.0 4.2‡ 1.3 1.0
(1.3, (1.1, (0.7, (0.6, (2.7, (0.9, (0.6,
2.9) 2.4) 1.7) 1.6) 6.4) 2.0) 1.6)
Discharge 1.5 1.9‡ 1.2 1.3 3.1‡ 1.2 0.8
(0.9, (1.2, (0.7, (0.7, (1.9, (0.8, (0.5,
2.3) 3.0) 2.0) 2.3) 5.1) 1.9) 1.3)
*Hazard ratios (95% confidence intervals) are shown for each variable. Two different dependent variables (reaching 50% predicted PEFR or discharge criteria within 12 hours) were considered. †Dichotomized at median. ‡Factors significant (p < 0.05) in a univariate Cox proportional hazards model.
32
ASTHMA RECOVERY
McCarren et al. • PREDICTION OF RECOVERY FROM ACUTE ASTHMA
TABLE 3. Prognostic Factors Significant in a Multivariable Cox Proportional Hazards Model* Dependent Variables Independent Variables
50% Predicted PEFR EDTU (n = 170)
Discharge EDTU (n = 170)
50% Predicted PEFR Inpatient (n = 99)
4.3 (2.8, 6.7) 1.8 (1.2, 2.6)
3.2 (2.0, 5.3) 2.0 (1.2, 3.2)
9.0 (4.1, 19.4) 1.2 (0.7, 2.2)
Third-treatment PEFR >36% Presenting dyspnea score ⱕ49 (scale 0–100)
*Hazard ratios (95% confidence intervals) are shown for three two-variable models. The three models differed in their samples [emergency diagnostic and treatment unit (EDTU) or inpatient] and dependent variable of interest [reaching 50% predicted peak expiratory flow rate (PEFR) or discharge criteria within 12 hours].
endpoint that can be used with any population, regardless of admitting practices. Prediction Zones. Having identified a predictive factor for recovery that was robust enough to be predictive across different treatment protocols (EDTU and inpatient), we next used LRs to describe prediction zones in both types of patients combined. To stratify patients by probability of recovering to 50% predicted PEFR within 12 hours, we defined six levels of third-treatment PEFR and calculated range-specific LRs.22 Table 4 shows that patients in the strata with PEFRs less than 32% after the third treatment were very unlikely to meet PEFR recovery criterion, indicated by positive LRs much less than 1.0. Subjects with PEFRs in the higher strata (>40%) were much more likely to be those meeting criterion (LRs >1.0). For instance, a patient with a PEFR of 42% is twice as likely to be in the group that recovers to ⱖ50% predicted within 12 hours than to be in the group that fails to achieve that outcome. We collapsed the six levels into three to correspond to typical clinical decision making.23 The time to meet 50% predicted PEFR is illustrated for TABLE 4. Third-treatment Peak Expiratory Flow Rate (PEFR) Categories*
PEFR Range PEFR
