Optimizing Triage and Hospitalization in Emergency Patients
Optimizing Triage and Hospitalization in Emergency Patients
The overall hypothesis of this study is that an improved initial triage of patients at an early stage of ED admission with incorporation of the MTS, initial clinical parameters and vital signs, prognostic blood markers and the PACD score will improve patient triage and translate into more objective estimation of triage priority, need for hospitalization and post-acute care needs. In this initial study we aim to derive a three-part triage algorithm, which will subsequently be evaluated in a second randomized controlled trial.
To derive a three-part triage algorithm to better predict (a) treatment priority; (b) medical risk and thus need for in-hospital treatment; (c) post-acute care needs of patients at the earliest time point of ED admission in a large and unselected population of medical patients.
This is done by development of three algorithms for assessing:
(Enlarge Image)
Figure 1.
Patient assessment for improved triaging of initial triage priority (Figure B), need for in hospital treatment (Figure C) and care needs (Figure D). Figure A shows the current conventional approach.
This is a prospective, observational, multi-center, multi-national cohort study. Over the time course of 12 months, we will prospectively include all consecutive medical patients seeking ED care. As an observational quality control study, the Institutional review board (IRB) of the Canton of Aargau has approved the study and waived the need for informed consent (EK 2012/059).
We will conduct this study in an multi-center, multi-national inter-professional and interdisciplinary collaboration at the Kantonsspital Aarau (Switzerland) including the Medical University Department, the Emergency Department, the Center of Laboratory Medicine, and the Clinical Nursing Science Department, as well as the Clinical Trial Unit (CTU) of the University Hospital of Basel and the Institute of Nursing science of the University of Basel; as well as the Emergency Department, Hôpital Pitié-Salpétriêre in Paris (France) and the Morton Plant Hospital in Clearwater, Florida (USA). Depending on the availability, we will also include other clinical centers to validate our findings.
We will include all consecutive medical patients including patients with neurological admission diagnoses presenting to ED for medical reasons and follow them during the hospital course until hospital discharge. There will be no exclusions except for non-adult and non-medical patients.
We will record initial vital signs (i.e. blood pressure, respiratory rate and others) and clinical parameters (i.e. main complaint, initial diagnosis) in the ED and collect left over blood samples in all patients. Clinical information including socio-demographics and comorbidities, patient outcomes and nursing information using the "Selbstpflegeindex" (SPI) and the PACD will be assessed prospectively until hospital discharge using the routinely gathered information from the hospital electronic medical system used for coding of Diagnosis-Related Groups (DRG) codes. This already available information supports the reliable assessment of baseline characteristics including demographics, comorbidities, acute medical conditions requiring the ED visit and different patient outcomes including inhospital mortality, resource use in terms of admission to the intensive care unit, length of stay (LOS) in the hospital and overall costs. We will also collect information about care needs in case of transfer to another post-acute institution after hospital discharge.
We will contact all patients by phone interview 30 days after admission to evaluate vital and functional status, care needs at home, rehospitalisation rates, satisfaction with care, preparedness for discharge,quality of life measures using the EQ-5D questionnaire and EQ VAS among others.
We will assess clinical stability of patients daily during the medical rounds. We have developed an online computer-based stability assessment tool - called "Visitentool" – where patient's stability and readiness for hospital discharge must be entered daily on clinical rounds. Similarly to the MTS, this is done in five categories (medical red: not stable, orange: stabilizing, yellow: stable but elective procedure awaiting, green: stable, discharge possible, blue: terminal/palliation) (nursing red: biopsychosocial risk (PACD ≥8) and/or post-acute care need likely, orange: interventions planned, yellow: ready for discharge/transfer but delay, green: discharge/transfer possible, blue: terminal) (social red: social services involved/in process, orange: external placement done, yellow: definitive date set for external relocation with time lag, green: definite date for external relocation set, date corresponds to earliest possible date regarding clinical stability). Importantly, physicians, nurses and social workers assess clinical stability and readiness for discharge daily from their perspective with this online tool to better understand the time to medical stability and readiness for discharge, and to study delays in hospital discharge which will also be documented.
To improve management of patients at the earliest time point of ED admission, we aim to develop a triage algorithm based on three distinct decision rules for (a) assessment of triage priority, (b) need for hospitalization and (c) post-acute care needs as shown in Figure 1. We therefore have three distinct main endpoints:
(Enlarge Image)
Figure 2.
Guidelines for adjudication of initial treatment priority with practical examples. The main question for adjudicators will be "under difficult circumstances, what is the maximum possible time that this patient would have been able to wait before being seen?" adapted from on a previous study [37].
Other endpoints will be defined as follows
All patient procedures are part of routine clinical care. Upon ED admission, a triage nurse will assess triage priority according to the MTS. Vital signs will be recorded and left over blood samples will be stored for later batch analysis of blood markers. The risk for post-acute care needs will be assessed with the PACD score per usual care. Patients will be reassessed daily during the hospital course for medical stability and readiness for discharge with an electronic tool as defined above (Visitentool). To assess patient outcomes, data from electronic medical records and from a patient quality questionnaire complemented with follow-up interviews at day 30 will be used. Below the detailed different steps of patient management are shown.
Step 1. Upon ED admission, all patients will be assessed by a designated triage nurse. MTS triage priority will be assigned based on the MTS as recommended. This will be entered into the clinical information system along with information about main complain, vital signs and clinical variables. The triage nurse will also assess the PACD on admission.
Step 2. In all patients, the triage nurse will perform a standardized blood draw for routine measurement of blood chemistry per usual care; left over samples will be aliquoted at the center of laboratory medicine and used for later batch analysis of biomarkers.
Step 3. Upon ED discharge, the attending ED physician will adjudicate a medical triage priority based on all medical results available at this time to all patients (high vs. low triage priority).
Step 4. Throughout the hospital stay, patients will be managed by physicians, nurses and social care in accordance to hospital guidelines according to the underlying medical condition. This will be at the discretion of the treating physicians, nursing and social worker staff, independent of the research team. During hospitalization, nursing scores will be collected per usual care and entered into the electronic medical system along with information about the planed care provided to patients after hospital discharge.
Step 5. All patients will be contacted 30 days after hospital admission for a telephone interview with a predefined questionnaire to assess vital and functional status, hospital readmission, as well as quality of life, care needs at home and satisfaction with care provided.
Left over blood samples of routinely collect blood tubes on admission will be immediatly centrifuged, aliquoted and frozen at -20C for later batch analysis of blood various biomarkers. The results of this analysis will not be available at the time of hospitalization of the patients and, thus, physicians and patients will be blinded to their results.
We will examine blood markers from different distinct biologic pathways as candidate biomarkers. Thus, we will assess markers of infection, inflammation, organ dysfunction, endothelial dysfunction, vasodilation/infection-control, stress hormones, cardiac dysfunction, nutrition, and kidney function, which all have been shown to predict adverse outcomes in different types of medical conditions (Table 1). Depending on the expected benefit from a literature research, the available funding and logistic support, we will decide which markers should be analyzed in the stored aliquots.
Within this study, we have several ancillary projects focusing on different aspects of patient care in this medical population.
First, we will look at costs from different perspectives, i.e. patient, society perspective, insurance perspective and hospital perspective. We will collect detailed cost data as well as resource use data. Based on the daily clinical assessment we will have good estimates how length of stay (LOS) could be reduced in patients without increasing their risk, i.e. at the time patients are classified as "medically stable" by the treating physician team. We will develop cost models using DRG reimbursements to evaluate the potential in savings.
Second, within a subset of patients we will focus on psychological distress defined as negative psychological reaction which may pre-exist or develop in the context of an acute disease potentially involving a variety of affective, cognitive, and behavioral reactions, such as fear, sadness, anxiety, frustration, or non-compliance. In this ancillary project we aim to explore the prevalence and course of patients' psychological distress on ED admission and within the hospital stay. To measure psychological distress we will use several validated instruments including the Distress Thermometer (DT) and the positive and negative affect schedule (PANAS). Beside general distress our focus will particularly lie on anxiety and depression assessed with the Patient Health Questionnaire-4 (PHQ-4). Additionally we will explore the relation of psychological distress with health outcomes (mortality, comorbidity, health-related quality of life, LOS among other) 30 days after admission. Finally, we aim to further delineate the role of specific patient's psycho-social resources (personality, social support, age, sex, SES, medical diagnosis) with regard to distress and health outcomes.
The purpose of this study is to develop an improved triage tool based on three distinct algorithms for (a) estimation of treatment priority (model 1), (b) prediction of medical risk (model 2) and (c) risk of post-acute care needs (model 3). For this purpose we have defined three distinct binary endpoints (i.e. high vs. low triage priority, adverse medical outcome within 30 days, post-acute care need) for which independent prediction rules will be developed using a similar approach for each one. However, based on the published literature, different candidate parameters will be considered as predictors for inclusion into the models.
In brief, for each algorithm we will select a parsimonious set of parameters from a comprehensive list of candidates including vital signs, clinical/socio-demographic predictors, blood markers, the MTS and the PACD. For blood markers we will focus on proADM and urea as the most established prognostic markers; however, we will also consider other markers for completion based on the availability of routine data (Table 1). We will use multivariable logistic regression analysis and different selection techniques including stepwise regression, Lasso among others. We will also compare the non-parametric CART analysis to decide if a simpler algorithm would qualify. Improvements in the area under the receiver operating curve (AUC) and reclassification statistics will inform about the benefit of adding parameters to the model. We will apply split sample validation (training and validation set with a ratio of 1:1) and present goodness of fit statistics to assess robustness and internal validity. Based on these results, we will derive weighted admission risk scores for the three main models, which can be used for later decision making (Figure 1). We will also look at subgroups to investigate differences in performance between main diagnoses and socio-demographic factors (age, gender) by inclusion of interaction terms into the logistic models.
For our model 1 (treatment priority), we will use adjudicated initial triage priority as the endpoint of interest (low vs. high triage priority) as defined above. As the MTS is well established for this purpose, we will first investigate the ability of the MTS to identify high priority subjects. We will then investigate whether addition of clinical parameters, vital signs and blood markers improve the MTS using statistical approaches outlined above. In a second step, we will investigate the performance of the MTS in subgroups of patients, i.e. stratified by initial admission diagnosis (e.g. myocardial infarction, congestive heart failure, infection, falls, lung embolism), by main clinical complain (e.g. dyspnea, fever, cough, pain) and by age quartiles, we will include interaction terms to study whether the association of the MTS and/or biomarkers varies across subgroups (effect modification). If significant effect modification is found, we will adapt the risk score to certain admission diagnoses.
For our model 2 (adverse outcome within 30 days) we will focus on death or ICU admission as the main outcome, in accordance with established risk scores (such as the pneumonia severity index or the CURB65 score). In previous research we found that specific blood biomarkers (i.e. proADM) have very high prognostic accuracy in the range of clinical risk scores, and that this is true across different medical conditions. However, other "baseline" factors, such as age and comorbidities are likely providing prognostic information beyond that of blood markers. Thus, it is a promising approach to combine these factors in a combined risk model.
Our model 3 (post-acute care needs) will focus on care needs in patients after hospital discharge. The PACD score was developed for this purpose. However, the PACD focuses mainly on care needs of patients prior to hospital admission and availability of help in the home setting, but not as much on the current medical situation. It is therefore possible that addition of parameters reflecting the severity of disease (vital signs, blood markers) or the nutritional condition (blood markers) further improves its accuracy. We will therefore start with the PACD and investigate whether addition of other parameters significantly improves its accuracy as outlined above.
We aim to include a total of at least 5000 patients over the course of 12 months, with expected rates for high treatment priority of 20% (n=1000), for adverse outcomes of 10% (n=500) and for post-acute care needs of 20% (n=1000). This will provide 50–100 degrees of freedom for each model (with 10 cases in the data set per degree of freedom in the statistical model), and thus high power for the calculation of the main multivariate models overall, in pre-defined subgroups and after inclusion of interaction terms.
Methods/Design
Overall Hypothesis and Research Aim
The overall hypothesis of this study is that an improved initial triage of patients at an early stage of ED admission with incorporation of the MTS, initial clinical parameters and vital signs, prognostic blood markers and the PACD score will improve patient triage and translate into more objective estimation of triage priority, need for hospitalization and post-acute care needs. In this initial study we aim to derive a three-part triage algorithm, which will subsequently be evaluated in a second randomized controlled trial.
Specific Aims
To derive a three-part triage algorithm to better predict (a) treatment priority; (b) medical risk and thus need for in-hospital treatment; (c) post-acute care needs of patients at the earliest time point of ED admission in a large and unselected population of medical patients.
This is done by development of three algorithms for assessing:
Treatment priority (high vs. low priority). This will be based on the MTS as the current state of the art tool, and other clinical variables and blood biomarkers (Figure 1B). This algorithm should help to correctly prioritize patients in a crowded ED setting and allocate resources to patients needing them first.
The overall 30 days medical risk based on different initial socio-demographic parameters, initial complaints, clinical parameters, vital signs and blood biomarkers across different medical conditions. This will help physicians to objectively estimate the need for inpatient treatment in patients and may improve site-of-care decisions (Figure 1C).
The risk for post-acute care needs, i.e. if patients need to be transferred to post-acute care institutions. This may improve early discharge planning (Figure 1D).
(Enlarge Image)
Figure 1.
Patient assessment for improved triaging of initial triage priority (Figure B), need for in hospital treatment (Figure C) and care needs (Figure D). Figure A shows the current conventional approach.
Study Design
This is a prospective, observational, multi-center, multi-national cohort study. Over the time course of 12 months, we will prospectively include all consecutive medical patients seeking ED care. As an observational quality control study, the Institutional review board (IRB) of the Canton of Aargau has approved the study and waived the need for informed consent (EK 2012/059).
Setting, Patient Population, Inclusion and Exclusion Criteria
We will conduct this study in an multi-center, multi-national inter-professional and interdisciplinary collaboration at the Kantonsspital Aarau (Switzerland) including the Medical University Department, the Emergency Department, the Center of Laboratory Medicine, and the Clinical Nursing Science Department, as well as the Clinical Trial Unit (CTU) of the University Hospital of Basel and the Institute of Nursing science of the University of Basel; as well as the Emergency Department, Hôpital Pitié-Salpétriêre in Paris (France) and the Morton Plant Hospital in Clearwater, Florida (USA). Depending on the availability, we will also include other clinical centers to validate our findings.
We will include all consecutive medical patients including patients with neurological admission diagnoses presenting to ED for medical reasons and follow them during the hospital course until hospital discharge. There will be no exclusions except for non-adult and non-medical patients.
Clinical Information and Assessment Outcomes
We will record initial vital signs (i.e. blood pressure, respiratory rate and others) and clinical parameters (i.e. main complaint, initial diagnosis) in the ED and collect left over blood samples in all patients. Clinical information including socio-demographics and comorbidities, patient outcomes and nursing information using the "Selbstpflegeindex" (SPI) and the PACD will be assessed prospectively until hospital discharge using the routinely gathered information from the hospital electronic medical system used for coding of Diagnosis-Related Groups (DRG) codes. This already available information supports the reliable assessment of baseline characteristics including demographics, comorbidities, acute medical conditions requiring the ED visit and different patient outcomes including inhospital mortality, resource use in terms of admission to the intensive care unit, length of stay (LOS) in the hospital and overall costs. We will also collect information about care needs in case of transfer to another post-acute institution after hospital discharge.
We will contact all patients by phone interview 30 days after admission to evaluate vital and functional status, care needs at home, rehospitalisation rates, satisfaction with care, preparedness for discharge,quality of life measures using the EQ-5D questionnaire and EQ VAS among others.
Daily Assessment of Clinical Stability With the "Visitentool"
We will assess clinical stability of patients daily during the medical rounds. We have developed an online computer-based stability assessment tool - called "Visitentool" – where patient's stability and readiness for hospital discharge must be entered daily on clinical rounds. Similarly to the MTS, this is done in five categories (medical red: not stable, orange: stabilizing, yellow: stable but elective procedure awaiting, green: stable, discharge possible, blue: terminal/palliation) (nursing red: biopsychosocial risk (PACD ≥8) and/or post-acute care need likely, orange: interventions planned, yellow: ready for discharge/transfer but delay, green: discharge/transfer possible, blue: terminal) (social red: social services involved/in process, orange: external placement done, yellow: definitive date set for external relocation with time lag, green: definite date for external relocation set, date corresponds to earliest possible date regarding clinical stability). Importantly, physicians, nurses and social workers assess clinical stability and readiness for discharge daily from their perspective with this online tool to better understand the time to medical stability and readiness for discharge, and to study delays in hospital discharge which will also be documented.
Endpoints
To improve management of patients at the earliest time point of ED admission, we aim to develop a triage algorithm based on three distinct decision rules for (a) assessment of triage priority, (b) need for hospitalization and (c) post-acute care needs as shown in Figure 1. We therefore have three distinct main endpoints:
Initial triage priority adjudicated by two independent ED physicians. Similar to a previous study, the physicians will evaluate what the real degree of urgency ("Goldstandard") would have been, based on the ED data, results of diagnostic tests, and the final diagnosis. Specifically, the main question for the adjudicators will be "under difficult circumstances, what is the maximum possible time that this patient would have been able to wait before being seen?" with options of "patient could not wait", 10 minutes, 30 minutes, 90 minutes, or 3 hours. To further standardize the adjudication, we have developed examples as demonstrated in Figure 2. We will collapse the initial 5 priority categories into 2 categories (i.e. low [more than 10 min, class 3, 4 or 5] vs. high priority [less than 10 min, class 1 or 2]). The 2 adjudicators will answer this question in regard to a medical prognostic focus and to a patient comfort focus (i.e. pain). In case of disagreement, a third independent physician will review the case until consensus is reached.
Adverse 30 day outcome (death, intensive care unit admission or unplanned hospital re-admissions) within 30 days following ED admission.
Post-acute care needs immediately after hospital discharge. This will be defined as transfer of patients to a post-acute care institution (i.e. transition to a nursing home, rehabilitation center and others).
(Enlarge Image)
Figure 2.
Guidelines for adjudication of initial treatment priority with practical examples. The main question for adjudicators will be "under difficult circumstances, what is the maximum possible time that this patient would have been able to wait before being seen?" adapted from on a previous study [37].
Other endpoints will be defined as follows
Time to first physician contact as assessed in the nursing chart; we will investigate this endpoint stratified by patients' risk, i.e. we will compare time to first physician contact in high-triage-priority and low-triage-priority patients and stratified by different diagnoses.
Time to initiation of adequate medical therapy in predefined subgroups (e.g., antibiotic therapy for infections, door to needle time for myocardial infarction; early goal directed therapy in sepsis patients, pain relief medication in patients presenting with pain, blood pressure control in patients with a hypertensive crisis); we will further assess time to discharge from the ED to the ward.
Satisfaction with care, preparedness for discharge, need of care at home, functional status and quality of life as assessed in the day 30 telephone interview.
Overall hospital costs as assessed by the electronic medical records.
Procedures and Management of Patients Throughout the Trial
All patient procedures are part of routine clinical care. Upon ED admission, a triage nurse will assess triage priority according to the MTS. Vital signs will be recorded and left over blood samples will be stored for later batch analysis of blood markers. The risk for post-acute care needs will be assessed with the PACD score per usual care. Patients will be reassessed daily during the hospital course for medical stability and readiness for discharge with an electronic tool as defined above (Visitentool). To assess patient outcomes, data from electronic medical records and from a patient quality questionnaire complemented with follow-up interviews at day 30 will be used. Below the detailed different steps of patient management are shown.
Step 1. Upon ED admission, all patients will be assessed by a designated triage nurse. MTS triage priority will be assigned based on the MTS as recommended. This will be entered into the clinical information system along with information about main complain, vital signs and clinical variables. The triage nurse will also assess the PACD on admission.
Step 2. In all patients, the triage nurse will perform a standardized blood draw for routine measurement of blood chemistry per usual care; left over samples will be aliquoted at the center of laboratory medicine and used for later batch analysis of biomarkers.
Step 3. Upon ED discharge, the attending ED physician will adjudicate a medical triage priority based on all medical results available at this time to all patients (high vs. low triage priority).
Step 4. Throughout the hospital stay, patients will be managed by physicians, nurses and social care in accordance to hospital guidelines according to the underlying medical condition. This will be at the discretion of the treating physicians, nursing and social worker staff, independent of the research team. During hospitalization, nursing scores will be collected per usual care and entered into the electronic medical system along with information about the planed care provided to patients after hospital discharge.
Step 5. All patients will be contacted 30 days after hospital admission for a telephone interview with a predefined questionnaire to assess vital and functional status, hospital readmission, as well as quality of life, care needs at home and satisfaction with care provided.
Blood Draws and Candidate Biomarkers
Left over blood samples of routinely collect blood tubes on admission will be immediatly centrifuged, aliquoted and frozen at -20C for later batch analysis of blood various biomarkers. The results of this analysis will not be available at the time of hospitalization of the patients and, thus, physicians and patients will be blinded to their results.
We will examine blood markers from different distinct biologic pathways as candidate biomarkers. Thus, we will assess markers of infection, inflammation, organ dysfunction, endothelial dysfunction, vasodilation/infection-control, stress hormones, cardiac dysfunction, nutrition, and kidney function, which all have been shown to predict adverse outcomes in different types of medical conditions (Table 1). Depending on the expected benefit from a literature research, the available funding and logistic support, we will decide which markers should be analyzed in the stored aliquots.
Ancillary Projects
Within this study, we have several ancillary projects focusing on different aspects of patient care in this medical population.
First, we will look at costs from different perspectives, i.e. patient, society perspective, insurance perspective and hospital perspective. We will collect detailed cost data as well as resource use data. Based on the daily clinical assessment we will have good estimates how length of stay (LOS) could be reduced in patients without increasing their risk, i.e. at the time patients are classified as "medically stable" by the treating physician team. We will develop cost models using DRG reimbursements to evaluate the potential in savings.
Second, within a subset of patients we will focus on psychological distress defined as negative psychological reaction which may pre-exist or develop in the context of an acute disease potentially involving a variety of affective, cognitive, and behavioral reactions, such as fear, sadness, anxiety, frustration, or non-compliance. In this ancillary project we aim to explore the prevalence and course of patients' psychological distress on ED admission and within the hospital stay. To measure psychological distress we will use several validated instruments including the Distress Thermometer (DT) and the positive and negative affect schedule (PANAS). Beside general distress our focus will particularly lie on anxiety and depression assessed with the Patient Health Questionnaire-4 (PHQ-4). Additionally we will explore the relation of psychological distress with health outcomes (mortality, comorbidity, health-related quality of life, LOS among other) 30 days after admission. Finally, we aim to further delineate the role of specific patient's psycho-social resources (personality, social support, age, sex, SES, medical diagnosis) with regard to distress and health outcomes.
Statistical Considerations and Sample Size
The purpose of this study is to develop an improved triage tool based on three distinct algorithms for (a) estimation of treatment priority (model 1), (b) prediction of medical risk (model 2) and (c) risk of post-acute care needs (model 3). For this purpose we have defined three distinct binary endpoints (i.e. high vs. low triage priority, adverse medical outcome within 30 days, post-acute care need) for which independent prediction rules will be developed using a similar approach for each one. However, based on the published literature, different candidate parameters will be considered as predictors for inclusion into the models.
In brief, for each algorithm we will select a parsimonious set of parameters from a comprehensive list of candidates including vital signs, clinical/socio-demographic predictors, blood markers, the MTS and the PACD. For blood markers we will focus on proADM and urea as the most established prognostic markers; however, we will also consider other markers for completion based on the availability of routine data (Table 1). We will use multivariable logistic regression analysis and different selection techniques including stepwise regression, Lasso among others. We will also compare the non-parametric CART analysis to decide if a simpler algorithm would qualify. Improvements in the area under the receiver operating curve (AUC) and reclassification statistics will inform about the benefit of adding parameters to the model. We will apply split sample validation (training and validation set with a ratio of 1:1) and present goodness of fit statistics to assess robustness and internal validity. Based on these results, we will derive weighted admission risk scores for the three main models, which can be used for later decision making (Figure 1). We will also look at subgroups to investigate differences in performance between main diagnoses and socio-demographic factors (age, gender) by inclusion of interaction terms into the logistic models.
For our model 1 (treatment priority), we will use adjudicated initial triage priority as the endpoint of interest (low vs. high triage priority) as defined above. As the MTS is well established for this purpose, we will first investigate the ability of the MTS to identify high priority subjects. We will then investigate whether addition of clinical parameters, vital signs and blood markers improve the MTS using statistical approaches outlined above. In a second step, we will investigate the performance of the MTS in subgroups of patients, i.e. stratified by initial admission diagnosis (e.g. myocardial infarction, congestive heart failure, infection, falls, lung embolism), by main clinical complain (e.g. dyspnea, fever, cough, pain) and by age quartiles, we will include interaction terms to study whether the association of the MTS and/or biomarkers varies across subgroups (effect modification). If significant effect modification is found, we will adapt the risk score to certain admission diagnoses.
For our model 2 (adverse outcome within 30 days) we will focus on death or ICU admission as the main outcome, in accordance with established risk scores (such as the pneumonia severity index or the CURB65 score). In previous research we found that specific blood biomarkers (i.e. proADM) have very high prognostic accuracy in the range of clinical risk scores, and that this is true across different medical conditions. However, other "baseline" factors, such as age and comorbidities are likely providing prognostic information beyond that of blood markers. Thus, it is a promising approach to combine these factors in a combined risk model.
Our model 3 (post-acute care needs) will focus on care needs in patients after hospital discharge. The PACD score was developed for this purpose. However, the PACD focuses mainly on care needs of patients prior to hospital admission and availability of help in the home setting, but not as much on the current medical situation. It is therefore possible that addition of parameters reflecting the severity of disease (vital signs, blood markers) or the nutritional condition (blood markers) further improves its accuracy. We will therefore start with the PACD and investigate whether addition of other parameters significantly improves its accuracy as outlined above.
We aim to include a total of at least 5000 patients over the course of 12 months, with expected rates for high treatment priority of 20% (n=1000), for adverse outcomes of 10% (n=500) and for post-acute care needs of 20% (n=1000). This will provide 50–100 degrees of freedom for each model (with 10 cases in the data set per degree of freedom in the statistical model), and thus high power for the calculation of the main multivariate models overall, in pre-defined subgroups and after inclusion of interaction terms.
