Top, One or more new psychiatric diagnoses by a psychiatrist. Bottom, New prescriptions for psychoactive medication during the 1 year of follow-up among the cohort of patients receiving mechanical ventilation with no 5-year prior history of any psychiatric diagnosis or prescriptions for psychoactive medications. For the overall diagnosis and drug analyses patients are censored at first diagnosis or drug. For individual diagnosis or drugs patients are followed up for each outcome.
eTable 1. ICD-10 codes for specific psychiatric diagnoses and Anatomical Therapeutic Chemical (ATC) codes for psychoactive medications
eTable 2. Comparison of patients receiving mechanical ventilation who were able to be matched (with hospitalized patients) and patients who were unmatched
eTable 3. Characteristics of the cohort of patients receiving mechanical ventilation, and matched hospital and population comparison cohorts who survived to hospital discharge (mechanical ventilation and hospital cohorts) with no psychiatric history in the five years prior to hospitalization (or index data for matched general population cohort).
eTable 4. Yearly adjusted hazard ratio for years 2 and 3 after hospital discharge comparing the risk of one or more new psychiatric illness and new prescriptions for psychoactive medication among patients receiving mechanical ventilation and comparison cohorts with no (5-year) prior history of any psychiatric diagnosis or prescription.
Wunsch H, Christiansen CF, Johansen MB, Olsen M, Ali N, Angus DC, Sørensen HT. Psychiatric Diagnoses and Psychoactive Medication Use Among Nonsurgical Critically Ill Patients Receiving Mechanical Ventilation. JAMA. 2014;311(11):1133-1142. doi:10.1001/jama.2014.2137
The relationship between critical illness and psychiatric illness is unclear.
To assess psychiatric diagnoses and medication prescriptions before and after critical illness.
Design, Setting, and Participants
Population-based cohort study in Denmark of critically ill patients in 2006-2008 with follow-up through 2009, and 2 matched comparison cohorts from hospitalized patients and from the general population.
Critical illness defined as intensive care unit admission with mechanical ventilation.
Main Outcomes and Measures
Adjusted prevalence ratios (PRs) of psychiatrist-diagnosed psychiatric illnesses and prescriptions for psychoactive medications in the 5 years before critical illness. For patients with no psychiatric history, quarterly cumulative incidence (risk) and adjusted hazard ratios (HRs) for diagnoses and medications in the following year, using Cox regression.
Among 24 179 critically ill patients, 6.2% had 1 or more psychiatric diagnoses in the prior 5 years vs 5.4% for hospitalized patients (adjusted PR, 1.31; 95% CI, 1.22-1.42; P<.001) and 2.4% for the general population (adjusted PR, 2.57; 95% CI, 2.41-2.73; P<.001). Five-year preadmission psychoactive prescription rates were similar to hospitalized patients: 48.7% vs 48.8% (adjusted PR, 0.97; 95% CI, 0.95-0.99; P<.001) but were higher than the general population (33.2%; adjusted PR, 1.40; 95% CI, 1.38-1.42; P<.001). Among the 9912 critical illness survivors with no psychiatric history, the absolute risk of new psychiatric diagnoses was low but higher than hospitalized patients: 0.5% vs 0.2% over the first 3 months (adjusted HR, 3.42; 95% CI, 1.96-5.99; P <.001), and the general population cohort (0.02%; adjusted HR, 21.77; 95% CI, 9.23-51.36; P<.001). Risk of new psychoactive medication prescriptions was also increased in the first 3 months: 12.7% vs 5.0% for the hospital cohort (adjusted HR, 2.45; 95% CI, 2.19-2.74; P<.001) and 0.7% for the general population (adjusted HR, 21.09; 95% CI, 17.92-24.82; P<.001). These differences had largely resolved by 9 to 12 months after discharge.
Conclusions and Relevance
Prior psychiatric diagnoses are more common in critically ill patients than in hospital and general population cohorts. Among survivors of critical illness, new psychiatric diagnoses and psychoactive medication use is increased in the months after discharge. Our data suggest both a possible role of psychiatric disease in predisposing patients to critical illness and an increased but transient risk of new psychiatric diagnoses and treatment after critical illness.
With recent advances in medical care, more patients are surviving critical illness,1 putting them at risk of subsequent morbidity.2,3 Critically ill patients are exposed to stress, including pain, respiratory distress, and delirium, all of which may impact subsequent mental health.4 Several studies of psychological sequelae of intensive care suggest that patients may develop a range of psychiatric disorders, including posttraumatic stress disorder, anxiety, and depression.3,5 However, these studies were limited by small sample size and lack of comparison groups.6 Other studies have assessed small subgroups of critically ill patients, such as elderly patients with severe sepsis or with acute respiratory distress syndrome.2,7 Recent data on older US patients with sepsis suggest that information regarding health status and trajectories prior to critical illness are essential for interpreting long-term outcomes.8 As yet, a good understanding of the burden of psychiatric illness prior to critical illness, as well as the magnitude of increased risk of psychiatric illness following critical illness remains elusive.
This study therefore had 2 aims: to examine the burden of psychiatrist-diagnosed psychiatric illness and all psychoactive medication prescriptions prior to nonsurgical critical illness and to examine whether risk of psychiatrist-diagnosed psychiatric illness and psychoactive medication use after nonsurgical critical illness was increased among patients with no prior psychiatric history.
This cohort study used linked data from Danish national population-based medical databases. We created a cohort of all adults (≥15 years) with a first-time nonsurgical admission to a Danish intensive care unit (ICU) requiring mechanical ventilation between January 1, 2006, and December 31, 2008; we chose to exclude surgical patients who required mechanical ventilation in an ICU because many are admitted as scheduled patients, receive transient mechanical ventilation for postoperative respiratory failure, and have significantly lower mortality in the year following hospitalization. Instead we focused on a high-risk group with unanticipated critical illness.9 Patients receiving mechanical ventilation were identified from the Danish National Registry of Patients database, which covers all Danish hospitals and has received mandatory reports on all ICU admissions since 2005.10 Mechanical ventilation was identified from the procedure code BGDA0, which is used for reimbursement. The accuracy of the code previously has been validated in this population.11 It does not include noninvasive ventilation, which has a separate code. We used the Civil Registration System to determine mortality in the year following hospital discharge (through December 31, 2009).12
If a patient had a surgical procedure other than endoscopy or minor surgery on the day of or 1 day prior to ICU admission, the ICU admission was considered surgical and therefore excluded. We obtained the primary diagnoses for hospital stays encompassing the ICU admission from the Danish National Registry of Patients and used these to create major disease categories. We also computed a Charlson Comorbidity Index score for each patient, based on national registry data on all inpatient and outpatient clinic diagnoses in the 5 years prior to ICU admission.13 Data on socioeconomic variables were obtained from the Integrated Database for Labor Market Research.14
The study was approved by the Danish Data Protection Agency (record number, 2009-41-3987). Ethical approval and informed consent are not required for register-based studies in Denmark.
The 2 main outcomes were any psychiatric diagnoses by psychiatrists and any redeemed prescriptions for a psychoactive medication (eTable 1 in the Supplement).15 Secondary outcomes included specific psychiatric disease categories and medication types. To obtain this information, the Danish National Registry of Patients and Civil Registration System data were linked with other national registries using the Danish Civil Registration Number, a unique identifier assigned to every Danish citizen at birth or immigration. Data were linked with information from the Danish Psychiatric Central Registry,16 which contains data on all hospital-based psychiatric outpatient clinic and inpatient contacts, including visits to emergency departments.17 Of note, there are no private psychiatric hospitals in Denmark and only a small number of psychiatrists in private practice. Data also were linked to the Danish National Prescription Registry, which contains individual-level data on all prescriptions filled at Danish pharmacies since 1995.15 This registry includes all prescribed medications other than those dispensed during hospitalizations.
The 2 matched comparison cohorts comprised (1) patients admitted to the same hospital in the same period as the patients receiving mechanical ventilation and (2) persons from the general population who were not in the ICU on or before the index date, which was the ICU admission date for the corresponding patient. We used the Danish National Registry of Patients to identify patients in the hospital comparison cohort. To create a balanced cohort, we selected 1 hospital patient for each patient receiving mechanical ventilation when possible, matched on the following criteria: (1) no stay in any Danish ICU from January 1, 2005, up to the time of and during the qualifying hospitalization, (2) hospitalization within 180 days of the corresponding patient’s ICU admission, and (3) admission to the same hospital department (non-ICU) in the same hospital. For the hospital comparison cohort, we did not match on age or sex. We used the Civil Registration System to sample the matched general population comparison cohort. Five general population cohort members were selected for each patient receiving mechanical ventilation, based on 4 criteria: (1) no previous ICU stay from January 1, 2005, to the index date, (2) same year of birth as the patient receiving mechanical ventilation, (3) same sex, and (4) alive on the index date.
We conducted a 2-part analysis. For the first aim, we compared the prevalence of any psychiatric diagnosis by psychiatrists or psychoactive medication prescription for patients receiving mechanical ventilation with the prevalence within the matched hospital and general population cohorts, both 1 and 5 years before the ICU admission or index date. We examined both periods to determine whether specific changes occurred for patients receiving mechanical ventilation shortly before the critical illness. We first summarized patient characteristics for the patients receiving mechanical ventilation and comparison cohorts. We then determined the proportion of patients with 1 or more psychiatric diagnoses and the proportion with 1 or more redeemed prescriptions for a psychoactive medication (ie, the period prevalence). Finally, we examined differences among the cohorts using prevalence ratios (PRs) computed by a log binomial model, accounting for factors present before critical illness including age and sex, country of birth, and educational level.
For the second aim, we examined whether the risk of a new psychiatric diagnosis by a psychiatrist or new psychoactive medication prescription was increased in the first year following ICU admission. This analysis excluded patients receiving mechanical ventilation who did not survive to hospital discharge; further, it was restricted to patients and also matched comparison cohort members with no psychiatric diagnoses or redeemed prescriptions for a psychoactive medication during the 5 years prior to hospitalization or index date for the population cohort. We computed the cumulative incidence (risk) of any new psychiatric diagnosis or psychoactive medication prescription in the year following the date of hospital discharge or index date, accounting for the competing risk of death. We could not determine whether patients were discharged directly home or spent time in a nursing facility. However, discharge to a nursing facility is a relatively rare occurrence for Danish hospital patients.18,19 Individuals were censored after the first event (ie, first psychiatric diagnosis by a psychiatrist or prescription for a psychoactive medication, depending on the outcome). For the patients who received mechanical ventilation, we created cumulative incidence curves of each type of psychiatric diagnosis and medication prescription.
Using Cox proportional hazards models, we generated hazard ratios (HRs) comparing patients who received mechanical ventilation with matched hospital and general population cohort members, adjusting for available potential confounding factors (age, sex, Charlson score, country of birth, educational level, gross income, and job level at the time of hospitalization or index date) and accounting for the competing risk of death. The model for the comparison with the general population accounted for matching using a stratified Cox model. Because the 1-year models violated the proportional hazards assumption, we present the data on a quarterly basis after the hospitalization or index date, with cumulative incidence and Cox proportional hazards models for each quarter. We also examined the cumulative incidence and hazard ratio for each type of diagnosis and psychoactive medication prescription to assess differences in risk. Model fit was assessed using scaled Schoenfeld residuals and tests for time-dependency for each variable in the models.
We performed additional analyses for the second and third years of follow-up (eTable 4 in the Supplement). Follow-up data were available through December 31, 2009, so a full 3 years of follow-up was only available for patients with an ICU stay in 2006. Data for patients with later dates of care in an ICU were censored as appropriate. All analyses were performed using SAS version 9.2 (SAS Institute Inc). Two-sided significance testing was used throughout, with a P <.05 was considered statistically significant.
Between 2006 and 2008, 24 179 nonsurgical patients admitted to a Danish ICU received mechanical ventilation (Table 1). More than half were older than 65 years and 40.8% had no Charlson comorbidities. Half (52.5%) required vasopressor or inotropic support while in the ICU. Many patients (29.5%) had a hospital stay of 4 or more weeks.
During the 1 year before the index hospitalization, 2.9% of patients receiving mechanical ventilation had 1 or more psychiatric diagnoses from a psychiatrist (Table 2). This prevalence was higher than for the matched hospital comparison cohort during the same period: 2.3% (adjusted PR, 1.50; 95% CI, 1.34-1.69; P < .001) and compared with the matched general population cohort: 0.8% (adjusted PR, 3.69; 95% CI, 3.35-4.07; P < .001). The 5-year prevalence was also higher: 6.2% among patients receiving mechanical ventilation vs 5.4% for the matched hospital comparison cohort (adjusted PR, 1.31; 95% CI, 1.22-1.42; P < .001) and 2.4% for the matched general population cohort (adjusted PR, 2.57; 95% CI, 2.41-2.73; P < .001; Table 2).
The period prevalence for any psychoactive medication prescription 1 year prior to the index hospitalization did not differ between the patients receiving mechanical ventilation and the matched hospital cohort. For the prior year, the period prevalence was 38.1% for patients receiving mechanical ventilation vs 36.9% for hospital patients (adjusted PR, 0.99; 95% CI, 0.97-1.01; P = .25). For the prior 5 years, there was no clinical meaningful difference in the period prevalence: 48.7% vs 48.8% (adjusted PR, 0.97; 95% CI, 0.95-0.99; P < .001; Table 2). The prevalence for the matched general population cohort was lower (22.9% for 1 year, adjusted PR, 1.57; 95% CI, 1.54-1.60; P < .001, and 33.2% for 5 years, adjusted PR, 1.40; 95% CI, 1.38-1.42; P < .001).
Among patients receiving mechanical ventilation who survived to hospital discharge, 9912 (59.2%) had no history of a psychiatric diagnosis from a psychiatrist or a prescription for a psychoactive medication in the five years prior to hospital admission. Characteristics of the patients receiving mechanical ventilation and comparison cohorts without prior psychiatric history are shown in eTable 3 in the Supplement. During the first year following hospital discharge, 15.7% died. The cumulative incidence for the year after hospital discharge among patients receiving mechanical ventilation, accounting for death, was 0.9% for receiving 1 or more new psychiatric diagnoses from a psychiatrist and 18.8% for receiving 1 or more prescriptions for psychoactive medications. The 2 most common types of new diagnoses were mood disorders and anxiety disorders (Figure). Among psychoactive medications, hypnotics were prescribed most frequently in the first few months after hospital discharge, whereas the frequency of new antidepressant prescriptions increased steadily over the first 12 months (Figure).
The patients receiving mechanical ventilation were found to be at increased risk of both new psychiatric diagnoses from a psychiatrist and prescriptions for psychoactive medications, although absolute risk differences for psychiatric diagnoses were small. Compared with the matched hospital cohort, the adjusted HR for a new psychiatric diagnosis in the cohort of patients receiving mechanical ventilation was highest in the first 3 months of the year following hospital discharge (cumulative incidence, 0.5% vs 0.2%; adjusted HR, 3.42; 95% CI, 1.96-5.99; P < .001), and showed no difference by the last 3 months (cumulative incidence, 0.1% vs 0.1%; adjusted HR, 0.78; 95% CI, 0.31-1.97; P = .59; Table 3). The risk was much higher for the patients receiving mechanical ventilation compared with the matched general population cohort in the first 3 months of the year following hospital discharge: 0.5% vs 0.02% (adjusted HR, 21.77; 95% CI, 9.23-51.36; P < .001), but again showed no difference by the last 3 months of the first year: 0.1% vs 0.04% (adjusted HR, 1.63; 95% CI, 0.66-4.04; P = .29).
The cumulative incidence of a new psychoactive medication prescription was increased in the cohort of patients receiving mechanical ventilation compared with the matched hospital cohort in the first 3 months following discharge: 12.7% vs 5.0% (adjusted HR, 2.45; 95% CI, 2.19-2.74; P < .001) but decreased to a similar risk by the last 3 months: 1.6% vs 1.5% (adjusted HR, 1.19; 95% CI, 0.88-1.61; P = .25). The risk compared with the matched general population cohort was very high for the first 3 months of the year following discharge: 12.7% vs 0.7% (adjusted HR, 21.09; 95% CI, 17.92-24.82; P < .001) and remained higher in the last 3 months: 1.6% vs 0.8% (adjusted HR, 2.08; 95% CI, 1.59-2.72; P < .001; eTable 4 in the Supplement reports data from years 2 and 3 after hospitalization).
By type of psychiatric diagnosis, new diagnoses of mood disorders were increased in the first 3 months after discharge in the cohort of patients receiving mechanical ventilation relative to the matched hospital cohort: 0.2% vs 0.1% (adjusted HR, 4.79; 95% CI, 1.94-11.81; P < .001). Anxiety disorders were markedly increased relative to the matched general population cohort 0.3% vs 0.01% (adjusted HR, 35.59; 95% CI, 10.43-121.40; P < .001) in the first 3 months following hospital discharge (Table 4). Differences between the cohort of patients receiving mechanical ventilation and the matched hospital cohort had resolved by 9 to 12 months but remained increased relative to the matched general population cohort. For categories of psychoactive medications, risk of a prescription for antipsychotics was highest for the patients receiving mechanical ventilation relative to the matched hospital cohort in the first 3 months: 1.1% vs 0.5% (adjusted HR, 2.62; 95% CI, 1.83-3.76; P < .001; Table 4). Compared with the matched general population cohort, the greatest increase was observed in prescriptions for hypnotics: 10.5% vs 0.4% (adjusted HR, 29.33; 95% CI, 23.81-36.12; P < .001).
We found that nonsurgical patients admitted to ICUs who received mechanical ventilation had a higher burden of psychiatric diagnoses from psychiatrists than either matched hospitalized patients or members of the general population in the 5 years preceding critical illness and had higher rates of psychoactive medication prescriptions compared with the general population. Patients receiving mechanical ventilation with no prior psychiatric history who survived to hospital discharge were at increased risk of new psychiatric diagnoses and psychoactive medication prescriptions. This increased risk was most pronounced in the first few months after the critical illness and resolved by the end of the first year of follow-up. Our data suggest both a possible role of psychiatric disease in predisposing patients to critical illness and an increased but transient risk of new psychiatric diagnoses and treatment after critical illness.
Our study provides key information extending findings from several small studies of intensive care survivors that observed a high prevalence of varying psychiatric disorders after hospital discharge.3,5,20 In particular, we provide detailed, unbiased information from the period prior to critical illness and for comparison groups allowing for robust risk estimates.6,8 Our findings are consistent with those of a study on the relationship between cognitive function and pneumonia, which observed that decreased cognitive function is associated with development of pneumonia and also appears to increase after pneumonia.21 Of note, there appeared to be an increased prevalence of psychiatric diagnoses and prescriptions for psychoactive medication among patients receiving mechanical ventilation relative to control groups in the year prior to the critical illness, compared with the 5-year period, suggesting that some deterioration in status may be occurring in the more immediate pre-ICU period.
In contrast to our findings, a study of patients with severe sepsis found a high burden of depressive symptoms in patients after the septic episode (measured on average almost a year later), but determined that the burden of depressive symptoms was similar in the period before illness.22 In our study, a large percentage of the critically ill population had no history of diagnosed depression or treatment for depression yet had substantially increased rates of this diagnosis and prescribed psychoactive medication in the year after the critical illness. However, the largest increase in risk occurred in the first few months after the critical illness, suggesting that the risk may be transient. An unexpected finding was a potentially decreased risk of a new psychiatric diagnosis or medication in the second or third year after hospital discharge. This may represent the “healthy worker” effect but warrants further investigation.23
A few studies have reported on medication use after an ICU stay.24,25 In their examination of US patients discharged from ICUs to long-term acute care facilities, Weinert24 found that 37% of patients without a prior history of psychiatric treatment were started on an antidepressant. It is notable that care patterns after hospital discharge in Denmark are likely very different from the United States, where a large proportion of critically ill patients are discharged from acute care hospitals to other care facilities.9 In Denmark, emphasis is on provision of nursing care in community-based environments and very few hospitalized patients subsequently receive care in nursing facilities.18,19 Moreover, prescriptions for patients who do require care in a nursing facility would still be prescribed by general practitioners and captured in our analysis.
Although relative increases in psychiatric diagnoses from psychiatrists and prescriptions in our cohort of patients receiving mechanical ventilation were similar, the absolute percentage of patients with psychiatric diagnoses was very low compared with the proportion of patients receiving psychoactive medications—both before and after the index hospitalization. The most likely reason for the large discordance between diagnoses and medication use is that many patients were diagnosed and treated for psychiatric problems, such as depression, by general practitioners without evaluation by a psychiatrist. Studies from many countries confirm that the majority of mental illnesses are diagnosed and treated without referral to specialists.26,27 A study from the Netherlands, which has a social structure and culture similar to Denmark’s, found that only 6% of episodes of mental illness involved a referral to mental health services or social work,26 consistent with our observed discrepancy in rates of medication prescription vs specialty-confirmed psychiatric diagnoses. Alternatively, some patients may have been seen by a private psychiatrist, so the encounters were not captured in our data. However, there are no private psychiatric hospitals in Denmark and few private psychiatrists. A final possibility is that some of these medications were prescribed without any formal diagnosis or were used to treat other problems such as sleep disorders,28 pain,29 or physical problems2,30 that are not covered by standard psychiatric diagnoses. In particular, patients often experience sleep disorders after hospitalization,31,32 and are frequently prescribed new medications specifically for sleep.33 However, studies of sleep after acute illness suggest that many of these patients also have concurrent psychiatric conditions.31,32
Our study has a number of limitations. First, we only assessed the prevalence of diagnoses and prescriptions for the 5 years prior to critical illness. Although this assessment allowed us to examine the recent burden of psychiatric diagnoses and psychoactive medication use, it does not provide a comprehensive assessment of any individual’s full psychiatric history. In particular, the age of onset of psychiatric illness may have occurred earlier in life, for example, with a diagnosis of depression at age 20 or 30 years followed by a period of quiescence.34,35 However, our approach provided a consistent look-back period for all patients, with availability of high-quality data during a period of little change in either diagnostic criteria or available medications for treatment. Although the life-time prevalence of psychiatric diagnoses has been reported to be twice as high as the 12-month prevalence,34,35 we found a similar increase in prevalence comparing 12 months with 5 years, suggesting that we may have captured the majority of patients with a history of psychiatric diagnoses.
Second, we chose to compare critically ill patients with a general population cohort that did not exclude people who may have been hospitalized. However, this potential exposure in the general population group allowed us to be conservative in our estimates of differences because it biases results toward the null hypothesis. Third, survivors of critical illness continue to have a relatively high risk of death 6 months to 1 year after hospital discharge.9 The difference in risk of death for patients receiving mechanical ventilation vs the comparison groups creates the potential for survivor bias. Fourth, we also lacked specific information on severity of the critical illness in the hospital or medications administered during the hospitalization (including psychoactive medications that may have been initiated as part of hospital care). However, we did have data on Charlson comorbidities, which are accurately captured in the Danish registries,36 as well as on length of stay, which together may represent a proxy for illness severity.37
Fifth, due to the complexity of data on diagnoses and prescriptions, we chose to analyze the data in terms of incident diagnoses and medication prescriptions. As some patients may have received multiple diagnoses or medications concurrently, we may have underestimated the total burden of psychiatric illness. Finally, we did not address the question of duration of medication use and therefore cannot confirm whether patients were being treated transiently with psychoactive medications, or for more prolonged periods of time.
Despite these limitations, our study provides important data on the burden of psychiatric illness among patients who experience critical illness requiring mechanical ventilation, as well as on the risks of psychiatric diagnoses and treatment with psychoactive medications in the year following ICU discharge. Discharge planning for these patients may require more comprehensive discussion of follow-up psychiatric assessment and provision of information to caregivers and other family members regarding potential psychiatric needs. Although the absolute risks were low, given the strong association between psychiatric diagnoses, such as depression, and poor outcomes after acute medical events, such as myocardial infarction and surgery,38,39 our data suggest that prompt evaluation and management of psychiatric symptoms may be an important focus for future interventions in this high-risk group. Prescriptions for psychoactive medications without concomitant care by psychiatrists also highlights the need for further research to understand patterns of psychoactive medication needs and use among survivors of critical illness.
Among nonsurgical patients with critical illness receiving mechanical ventilation, there was an increased prevalence of prior psychiatric diagnoses compared with hospitalized controls and with the general population but no significant increase in the prevalence of psychoactive medication use prior to admission compared with hospitalized controls. Patients who received mechanical ventilation without psychiatric history had an increased risk of new psychiatric diagnosis and new use of psychoactive medication, primarily in the first few months following hospital discharge.
Corresponding Author: Christian F. Christiansen, MD, PhD, Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Alle 43-45, 8200 Aarhus N, Denmark (firstname.lastname@example.org).
Author Contributions: Dr Christiansen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Wunsch, Christiansen.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Johansen.
Obtained funding: Christiansen, Sørensen.
Administrative, technical, or material support: Sørensen.
Study supervision: Christiansen, Angus, Sørensen.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported by grant 271-05-0511 from the Danish Medical Research Council, the Clinical Institute at Aarhus University, and the Department of Clinical Epidemiology’s Research Foundation at Aarhus University Hospital.
Role of the Sponsor: Funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: Dr Angus, Associate Editor for JAMA, was not involved in the evaluation or decision to publish this article.