- Question: Is nighttime extubation of adult mechanically ventilated patients associated with worse patient outcomes compared to those extubated during daytime hours?
- Findings: After reliably excluding palliative extubations and adjusting our analysis for a number of potential confounders, including routine postoperative and unplanned extubations, nighttime extubation was not associated with increased odds of reintubation at any time during hospitalization or in-hospital death.
- Meaning: In appropriately selected patients cared for with multidisciplinary extubation assessment, protocolized decision making, and expert airway assistance available 24/7, nighttime extubation is not associated with added harm and may improve resource utilization.
The decision to liberate a patient from mechanical ventilation (MV) is complex and requires bedside assessment to determine the adequacy of gas exchange, evidence of clinical improvement, whether positive pressure ventilation is needed, and the mental status is adequate. However, despite such clinical and laboratory assessments, as many as 1 in 5 patients will require reintubation during their hospital stay.1 Furthermore, patients who fail extubation suffer increased morbidity, longer hospital and intensive care unit (ICU) length of stay (LOS), more frequent discharge to long-term care or rehabilitation facilities, and higher hospital mortality.1–4
While several patient- and procedure-related factors are known to increase the likelihood of extubation failure, a 2016 study identified nighttime extubation (NTE), often defined between the hours of 7:00 pm and 7:00 am, as a risk factor for adverse outcomes in ICU patients who are mechanically ventilated.5 However, because that trial was retrospective, the association between NTE and the reported outcomes may have been confounded by patients who underwent palliative extubations, those on which limitations in care had been placed shortly after extubation, or those in whom extubation was unplanned.
At our institution, extubation decisions and outcomes for all patients who are liberated from MV are reviewed. In addition to the time of day at which the patient was extubated, we are able to identify potential confounders such as palliative and self-extubations. Thus, the primary aim of this study was to determine whether, when compared to daytime extubation (DTE), NTE is associated with greater odds of reintubation, longer hospital and ICU LOS, or mortality at a large, urban, teaching hospital.
Human Subjects Division approved this retrospective cohort study with a waiver of informed consent. The study site was an urban, university-affiliated, 413-bed level 1 trauma center serving a 5-state region in the Pacific Northwest. The hospital’s 88 adult ICU beds are geographically distributed among medical/cardiac, trauma/surgical, burn, and neuroscience patients. All units operate in a semiclosed fashion and are covered 24 hours a day by an intensivist-led team comprising an attending physician, fellow, senior resident, and junior resident. The Departments of Surgery, Anesthesiology, Internal Medicine, Neurology, and Emergency Medicine provide attending and trainee physician coverage. Respiratory therapists aid in the management of all ventilated patients, including completion of routine spontaneous breathing trials (SBTs) for those patients who qualify each morning.
Airway Management Model
The out-of-operating room airway management model in all areas of the hospital outside the emergency department (ward, ICU, radiology suites, and other remote locations) is anesthesiology based and includes responsibility for intubation and preextubation assessment in all high-risk patients. A preassigned anesthesia airway team comprising an anesthesiology trainee or certified registered nurse anesthetist and the attending anesthesiologist-in-charge of the operating rooms performs. For all intubations, the policy is that 2 operators should be present. Additionally, members of the department of respiratory therapy (RT) and bedside nurses familiar with emergency procedures attend all intubations. On the ward or in other remote locations, such as radiology or the outpatient clinic areas, dedicated “rapid response” nurses also attend and assist with intubations as part of a dedicated rapid response team.
To extubate any adult patient at our institution, an order must be entered via the computerized physician order entry system. For the system to allow order signature, a reintubation risk stratification checklist must first be completed. The reintubation risk stratification checklist is intended to identify patients who exhibit high-risk features for extubation failure/difficult reintubation (eg, history of difficult airway, restricted airway access, multilevel cervical spine surgery, neuropathology involving the posterior fossa, and body mass index [BMI] ≥45 kg/m2)1,2,6–11 (Supplemental Digital Content 1, Figure 1, http://links.lww.com/AA/C564). When patients are identified as high risk, an automatic page is sent to the airway team requesting a preextubation bedside evaluation. In addition, an electronic communication is sent to RT, which suspends the extubation process pending review of the patient and extubation plan with the airway team. In this way, airway managers primarily responsible for urgent/emergent reintubation are required to perform bedside consultation on all such patients. It is not mandatory that a member of the airway team be present during the actual extubation. Rather, the focus is on patient evaluation, formulation of a reintubation plan should it be required, and direct closed-loop communication with the ICU team, RT, and bedside nursing after the airway consultation has taken place. However, if sufficient concern exists after the multidisciplinary discussion occurs, a member of the airway team will typically remain in the ICU until clinical stability postextubation has been achieved. For patients without risk factors for extubation failure/difficult reintubation, RT proceeds to extubation without further physician intervention.
Our institutional approach to liberating adults from MV is based on consensus guideline recommendations.12–14 All patients underwent a daily, protocolized ventilator-weaning assessment with subsequent assessments during the same 24-hour period at the discretion of the primary critical care team. Patients were considered eligible for an SBT if they met the following criteria: evidence of resolution or improvement of the underlying cause of respiratory failure, a minute ventilation (VE) <15 L/min, positive end-expiratory pressure ≤8 cm H2O, fraction of inspired oxygen <0.5, Pao2/fraction of inspired oxygen ≥150, pH ≥7.25 with intact respiratory drive, intracranial pressure <20 cm H2O, and hemodynamic stability without cardiovascular support. Sedation for MV was provided per standardized ICU protocol targeting a goal of 0 to −1 on the Richmond Agitation and Sedation Scale. All SBTs were performed on continuous positive pressure of 5 cm H2O without additional pressure support following a spontaneous awakening trial or during the provision of continuous light sedation.
The SBT was considered a failure if: respiratory rate exceeded 35 breaths/min; oxygen saturation <90% for >30 seconds; reduction in VE <75% of baseline during MV; a heart rate >140 bpm or a change >20% from baseline; a systolic blood pressure >180 or <90 mm Hg; sustained increase in anxiety, diaphoresis, or other clinical signs of respiratory distress; an intracranial pressure increase >20 cm H2O for >2 minutes; arrhythmia; pH ≤7.25; or a Paco2 increase ≥10 mm Hg. Additional factors considered by the primary critical care team before extubation after the patient had passed the SBT include the following: a requirement for tracheal suctioning more frequently than every 4 hours, an ineffective spontaneous cough, lack of a leak around the deflated tracheal tube cuff with a sustained manual inspiratory pressure of 30 cm H2O, and altered airway reflexes. Patient-specific factors, including additional MV weaning parameters, were also available on request and integral to each primary critical care team’s decision (not) to extubate eligible patients. Palliative extubations were exempted from this institutional protocol and thus did not require a structured note entry to the medical record. Although >50% of extubations typically occur by the late morning or early afternoon after a morning SBT, the decision to extubate patients at other times of day, including overnight, was entirely patient and provider specific, and was not otherwise regulated by the institution.
All patients ≥18 years of age requiring invasive MV via an endotracheal tube at the study site between July 1, 2015 and December 31, 2016 were included in this study. Patients <18 years of age and adult patients who underwent palliative extubation, received a tracheostomy without undergoing a trial of extubation, or who died before extubation were excluded from further analysis. For patients with multiple extubations during the same hospital encounter, only data from the initial extubation were assessed.
Decisions regarding extubation and the need for reintubation, if indicated, were made as part of routine clinical care by the primary critical care service caring for each mechanically ventilated patient. For each extubation, the patient’s respiratory therapist completes a structured procedure note that allowed prospective collection of data pertaining to the extubation process. Discrete fields in the note include the time, date, and unit in which the extubation occurred, whether it was planned or unplanned, pre- and post-SBT respiratory and gas exchange information, the extubation risk stratification level, and postextubation clinical evaluation details. Demographic variables (age, sex, height, weight, BMI, patient type [medical, surgical, trauma/burns, neuroscience], surgical status, and the Charlson comorbidity index [CCI]) and resource metrics (ventilator days, ICU and hospital LOS, and mortality) were captured by the hospital’s Caradigm Intelligence Platform (Caradigm USA LLC, Seattle, WA) and retrospectively assembled via a comprehensive structured query language-based reporting tool. Manual review of the electronic medical record was conducted to retrieve data unavailable through the Caradigm automated search and to verify data integrity.
Study End Points and Definitions
The primary outcome of this study was reintubation at any time during the hospital encounter. Secondary end points included duration of MV, length of ICU and hospital stay, and in-hospital mortality. NTE was defined as extubation occurring between 7:00 pm and 6:59 am the following day. This time frame was chosen to mirror 2 previous studies on the subject.5,15 Our prespecified covariates included age, BMI, type of condition/injury, and CCI. Patients were grouped into 4 condition/injury classifications (medical, surgical, trauma/burn, and neuroscience) based on the primary admitting diagnosis as defined by the International Statistical Classification of Diseases and Related Health Problems, Ninth and Tenth Revisions, and the geographic location of the patient in the hospital.
The CCI is a quantitative score that reflects the burden of patients’ comorbid medical conditions. When combined with factors such as age, sex, and reason for admission, it has been reported to predict survival at 30 days and 1 year after discharge.16
Patients were considered to have undergone “routine” postoperative MV if, according to the electronic medical record, the patient was spontaneously breathing before surgery, returned from the operating room intubated and mechanically ventilated, and was subsequently extubated within 24 hours of return from the operating room.
Group comparisons between DTE and NTE were made using Pearson χ2 tests, Mann-Whitney U tests, or Student t tests for categorical, ordered categorical, or continuous data, respectively. The strengths of associations between DTE and NTE on the dichotomous outcomes (reintubation and mortality) were assessed using logistic regression models, adjusting for a priori defined potential confounders (age, sex, BMI, admission type, CCI, whether the extubation was routine, and if it was unplanned). Continuous outcomes (ICU and hospital LOS and days of MV) were summarized by cohort using medians and interquartile ranges. Mann-Whitney U tests were computed for unadjusted comparisons between the cohorts. To adjust for potential confounders (same set as in logistic regression models), linear regression analyses were performed on the log-transformed outcomes. The distributions of the residuals from the linear regression analyses were used to verify that the normality and constant variation assumptions held. After computation, the coefficients from the linear regression analyses were exponentiated and are reported as estimates of the ratios of geometric means between cohorts. Several sensitivity analyses were conducted to check for the robustness of the main findings. In the first sensitivity analysis, we excluded patients who were routine postoperative extubations (eg, admitted to ICU for extended anesthesia recovery) and/or unplanned extubations (eg, patients who self-extubated). In a second sensitivity analysis, we extended the definition of NTE to include extubations occurring between 5:00 pm and 6:59 am because the additional 2-hour window (5:00–6:59 pm) represents a time during which handoffs and house staff shift change are frequently occurring, potentially placing newly extubated patients at increased risk for complications. Data were collected over a consecutive 18-month period and included approximately 2000 qualifying extubations. This number allows approximately 80% power to detect a difference in mortality rates of 5% between the DTE and NTE groups (based on a mortality of 6.1% in the DTE and 11.2% in the NTE groups from Ref. 5) and assuming approximately 7 times more DTE procedures than NTE procedures. We considered a clinically significant margin between NTE and DTE for both outcomes, reintubation and mortality, to be >5%. A 2-sided α <.05 was considered statistically significant. Statistical analyses were performed using SPSS statistical software, version 19.0 (IBM Corp, Armonk, NY).
During the 18-month study period, 2903 patients underwent MV. Excluded patients were <18 years of age (n = 111), those who underwent tracheostomy without ever having an extubation attempt (n = 94), and those who underwent palliative extubation (n = 457). Thus, the final study cohort included 2241 adult patients who underwent ≥1 extubation. The number of patients who were extubated during each of the 24-hour time periods is shown in Supplemental Digital Content 1, Figure 2, http://links.lww.com/AA/C564. The patients (90.9%) (2037/2241) were extubated between the hours 7:00 am and 6:59 pm, and 204 of 2241 patients (9.1%) underwent NTE. Baseline characteristics of the study cohort, by time of extubation, are shown in Table 1. Patients who underwent NTE were younger, had a lower burden of medical comorbidities, and self-extubated more often than their daytime counterparts. Differences in primary and secondary outcomes between the DTE and NTE are shown in Table 2. After adjustment for predefined covariates, NTE was not associated with increased odds of reintubation (odds ratio [OR], 0.78; 95% confidence interval [CI], 0.43–1.41; P = .41) or in-hospital mortality (OR, 0.72; 95% CI, 0.28–1.84; P = .49). NTE was associated with reduced duration of MV, with a median duration of 2 days observed in the DTE cohort, and 1 day in the NTE (adjusted ratio of geometric means, 0.62; 95% CI, 0.54–0.7; P < .001). Similarly, NTE was associated with a reduced median ICU LOS (2 vs 4 days in the DTE cohort; adjusted ratio of geometric means, 0.65; 95% CI, 0.57–0.75; P < .001). Median hospital LOS was also lower for NTE patients (6 vs 13 days, adjusted ratio of geometric means, 0.64; 95% CI, 0.56–0.74; P < .001) (Table 3).
Sensitivity analyses excluding routine postoperative extubations (planned and unplanned), and extending the definition of NTE did not alter these results (Supplemental Digital Content 2, Tables 1–3, http://links.lww.com/AA/C565).
In this retrospective study of extubations, we found that patients extubated between 7:00 pm and 6:59 am were not at increased risk of reintubations, were mechanically ventilated for fewer days, had shorter ICU and hospital LOS, and had similar in-hospital mortality to those undergoing DTE. In addition, our results were not altered by sensitivity analyses excluding patients who underwent routine postoperative extubations and/or those who self-extubated and with an expanded definition of nighttime that included an “at risk” window toward the end of shift.
Our results are consistent with a previous study of adult patients who underwent NTE at 2 hospitals within a single tertiary academic medical center between July 2009 and May 2011.15 Among 2240 patients, NTE was not associated with increased odds of reintubation at 72 hours (OR, 0.7; 95% CI, 0.5–1.0) or in-hospital mortality (OR, 0.6; 95% CI, 0.3–1.0).
However, our results are in contrast to a large 2016 retrospective study of the Project IMPACT database (Society of Critical Care Medicine, Mount Prospect, IL), which included nearly 100,000 patients who were extubated at 1 of 165 participating ICUs in the United States between October 1, 2000 and March 29, 2009.5 Using matched-pairs analysis, the authors found that patients who underwent NTE after MV longer than 12 hours were at a significantly higher risk of reintubation or death during their hospitalization.
Several explanations may underlie the discrepancy between our results and those of 2 previous studies. First, our data are derived from patients cared for at a single academic medical center where (1) decisions regarding who is a candidate for liberation from the ventilator is highly standardized; (2) extubation risk assessment resulting in multidisciplinary discussion is required before every adult extubation; and (3) an anesthesia-based airway management team is continuously available and responsible for evaluating all high-risk extubation patients and all ICU intubations. Thus, candidates for NTE are carefully screened and expert rescue therapy is immediately available, possibly limiting the chances of adverse airway-related outcomes. Previous investigators reported NTE rates of 20%–30%,5,15 whereas NTE occurred in only 9.1% of our patients, potentially underscoring this point.
Although the lower occurrence of NTE in our study than in existing literature might limit the generalizability of our findings, our reported NTE rates may not differ significantly from previous investigations. Our institution does not perform cardiac surgery or have a dedicated cardiac surgical ICU (CSICU), so our data were derived from patients cared for in medical ICU, trauma-surgical ICU (SICU), or neuroscience ICU. In the 2016 report of Tischenkel et al,15 52% (1171/2240) of patients were cared for in their CSICU, with the remaining 48% (1069/2240) of patients cared for in the medical ICU, SICU, and mixed medical-surgical ICU. Excluding patients cared for in the CSICU, NTE occurred in 9.5% (102/1069) of patients, nearly identical to our rate. In the more recent report by Gershengorn et al,5 the number of dedicated CSICUs is not reported. However, among the included ICUs, 75% were SICU or mixed medical-surgical ICU. Given that 51.8% of the 19,666 patients undergoing NTE were admitted to ICU primarily with a cardiovascular system–related disease, we think it is likely that a proportion of these patients had undergone cardiac surgery and, if excluded, may reduce their reported NTE rate of 20.1%.
Second, we reliably excluded palliative extubations and adjusted for unplanned extubation, accounting for 2 factors, potentially biasing results of other studies toward worse outcomes with NTE. If palliative extubations were not excluded (n = 456), our study cohort would have included 2697 patients. If an estimated 20% (n = 91) of palliative extubations were classified as NTE, our reported mortality for NTE would be 33.2%, a value more consistent with other studies for NTE. While it is possible that our sample size was too small to detect a clinically relevant difference in mortality between the DTE and NTE groups, we argue that not excluding palliative extubations can bias mortality estimates upward.5
Furthermore, in our cohort, NTE patients had 12% lower odds of reintubation than the DTE patients. While our data do not exclude a higher reintubation rate for the NTE group, they do limit the difference between NTE and DTE groups. At the extremes of our CI estimates, the NTE group would have at most a 4.2% increase in reintubation rates and a 4.4% incidence of mortality. These estimates suggest that, in our patient population, NTE rates of reintubation and mortality are unlikely to exceed those in DTE patients by a large margin.
Our study has limitations. First, our sample size for NTE is moderate. However, this is a limitation of most studies examining NTE given that many fewer extubations occur at night than during daytime hours, and we were able to adequately power our study to identify an effect of NTE. Second, those undergoing NTE in our study were younger and had fewer comorbidities than those extubated during the day. It is thus possible that residual confounding and selection bias may have contributed to our results. However, estimates of the patients’ burden of comorbid disease, reason for admission and type of ICU, as well as age and sex were similar to analyses in previous studies.5,15 Third, it is also possible that extubation of patients who are at increased risk of failure is delayed based on anesthesia availability, which in some cases may be the following morning. This may be particularly relevant when an extubation evaluation coincides with multiple operating room emergencies or when the airway team has concerns regarding an extubation that requires further planning or coordination such as might occur when it is felt the extubation should take place in the operating room with emergency surgical airway access immediately available. Unfortunately, the nature of our data does not allow us to know for certain how many extubations were delayed after the initial evaluation for any of these reasons.
Finally, we note that our experience with NTE and process for patient evaluation may differ from other hospitals, potentially decreasing the generalizability of our results. In addition to our process for identifying and vetting patients for extubation, we would add that >50% of our DTE patients were extubated between the hours of 7:00 am and 12:00 pm, likely using the results of the morning SBT as a trigger for possible extubation.
We adjusted our results by the CCI, which, like the Elixhauser17 and the mortality probability model at ICU admission-III18 scores, approximates illness severity based on the presence of comorbid disease at the time of admission, the onset of illness, and/or time of surgery.19,20 Although the CCI has been repeatedly validated as a mortality prediction tool for critically ill patients,17,21 it is not an ICU-specific severity-of-illness score and may be inferior to acute physiology-based illness severity indices.16,22,23
We did not find evidence of additional risk of reintubation or in-hospital mortality for NTE patients at a large, urban, university-affiliated trauma center. This relationship persisted after adjustment for age, BMI, medical comorbidity, and type of injury/illness, and after exclusion of palliative, unplanned, and routine postoperative extubations. NTE was also significantly associated with a shortened duration of MV and hospital LOS. Robust institutional extubation criteria supported by multidisciplinary extubation assessment, and continuously available anesthesiologists who provide 24/7 risk stratification and urgent/emergent airway coverage for out-of-operating room intubations may partially account for our findings. In care models similar to ours, NTE appears to be an effective approach to liberation from MV that may reduce resource utilization in appropriately selected adult patients.
Name: Kelly K. Everhart, MD, MS.
Contribution: This author helped collect and analyze the data, and prepare the manuscript.
Name: Sarah Khorsand, MD.
Contribution: This author helped analyze the data and prepare the manuscript.
Name: Nita Khandelwal, MD, MS.
Contribution: This author helped conduct the study and prepare the manuscript.
Name: Kelly E. Michaelsen, MD, PhD.
Contribution: This author helped analyze the data and prepare the manuscript.
Name: Charles F. Spiekerman, PhD.
Contribution: This author helped design the study, collect and analyze the data, and prepare the manuscript.
Name: Aaron M. Joffe, DO.
Contribution: This author helped design and conduct the study, collect and analyze the data, and prepare the manuscript.
This manuscript was handled by: Avery Tung, MD, FCCM.
1. Miu T, Joffe AM, Yanez NDPredictors of reintubation in critically ill patients. Respir Care. 2014;59:178–185.
2. Thille AW, Harrois A, Schortgen F, Brun-Buisson C, Brochard LOutcomes of extubation failure in medical intensive care unit patients. Crit Care Med. 2011;39:2612–2618.
3. Demling RH, Read T, Lind LJ, Flanagan HLIncidence and morbidity of extubation failure in surgical intensive care patients. Crit Care Med. 1988;16:573–577.
4. Epstein SK, Ciubotaru RL, Wong JBEffect of failed extubation on the outcome of mechanical ventilation. Chest. 1997;112:186–192.
5. Gershengorn HB, Scales DC, Kramer A, Wunsch HAssociation between overnight extubations and outcomes in the intensive care unit. JAMA Intern Med. 2016;176:1651–1660.
6. Rady MY, Ryan TPerioperative predictors of extubation failure and the effect on clinical outcome after cardiac surgery. Crit Care Med. 1999;27:340–347.
7. Menon N, Joffe AM, Deem SOccurrence and complications of tracheal reintubation in critically ill adults. Respir Care. 2012;57:1555–1563.
8. Sagi HC, Beutler W, Carroll E, Connolly PJAirway complications associated with surgery on the anterior cervical spine. Spine (Phila Pa 1976). 2002;27:949–953.
9. Cavallone LF, Vannucci AReview article: extubation of the difficult airway and extubation failure. Anesth Analg. 2013;116:368–383.
10. Kuhn JE, Graziano GPAirway compromise as a result of retropharyngeal hematoma following cervical spine injury. Clin Spine Surg. 1991;4:264–269.
11. Howard R, Mahoney A, Thurlow ACRespiratory obstruction after posterior fossa surgery. Anaesthesia. 1990;45:222–224.
12. Collective Task Force Facilitated by the American College of Chest Physician, the American Association for Respiratory Care, and the American College of Critical Care Medicine. Evidence-based guidelines for weaning and discontinuing ventilatory support. Respir Care. 2002;47:69–90.
13. Girard TD, Alhazzani W, Kress JP, et alATS/CHEST Ad Hoc Committee on Liberation From Mechanical Ventilation in Adults. An official American Thoracic Society/American College of Chest Physicians Clinical Practice Guideline: liberation from mechanical ventilation in critically ill adults. Rehabilitation protocols, ventilator liberation protocols, and cuff leak tests. Am J Respir Crit Care Med. 2017;195:120–133.
14. Ouellette DR, Patel S, Girard TDLiberation from mechanical ventilation in critically ill adults: an official American College of Chest Physicians/American Thoracic Society Clinical Practice Guideline: inspiratory pressure augmentation during spontaneous breathing trials, protocols minimizing sedation, and noninvasive ventilation immediately after extubation. Chest. 2017;151:166–180.
15. Tischenkel BR, Gong MN, Shiloh ALDaytime versus nighttime extubations: a comparison of reintubation, length of stay, and mortality. J Intensive Care Med. 2016;31:118–126.
16. Christensen S, Johansen MB, Christiansen CF, Jensen R, Lemeshow SComparison of Charlson comorbidity index with SAPS and APACHE scores for prediction of mortality following intensive care. Clin Epidemiol. 2011;3:203–211.
17. Ladha KS, Zhao K, Quraishi SAThe Deyo-Charlson and Elixhauser-van Walraven comorbidity indices as predictors of mortality in critically ill patients. BMJ Open. 2015;5:e008990.
18. Higgins TL, Teres D, Copes WS, Nathanson BH, Stark M, Kramer AAAssessing contemporary intensive care unit outcome: an updated Mortality Probability Admission Model (MPM0-III). Crit Care Med. 2007;35:827–835.
19. Charlson M, Szatrowski TP, Peterson J, Gold JValidation of a combined comorbidity index. J Clin Epidemiol. 1994;47:1245–1251.
20. Charlson ME, Pompei P, Ales KL, MacKenzie CRA new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383.
21. Stavem K, Hoel H, Skjaker SA, Haagensen RCharlson comorbidity index derived from chart review or administrative data: agreement and prediction of mortality in intensive care patients. Clin Epidemiol. 2017;9:311–320.
22. Needham DM, Scales DC, Laupacis A, Pronovost PJA systematic review of the Charlson comorbidity index using Canadian administrative databases: a perspective on risk adjustment in critical care research. J Crit Care. 2005;20:12–19.
23. Poses RM, McClish DK, Smith WR, Bekes C, Scott WEPrediction of survival of critically ill patients by admission comorbidity. J Clin Epidemiol. 1996;49:743–747.