Intervention Review

You have free access to this content

Medication review in hospitalised patients to reduce morbidity and mortality

  1. Mikkel Christensen1,*,
  2. Andreas Lundh2

Editorial Group: Cochrane Effective Practice and Organisation of Care Group

Published Online: 28 FEB 2013

Assessed as up-to-date: 10 AUG 2011

DOI: 10.1002/14651858.CD008986.pub2


How to Cite

Christensen M, Lundh A. Medication review in hospitalised patients to reduce morbidity and mortality. Cochrane Database of Systematic Reviews 2013, Issue 2. Art. No.: CD008986. DOI: 10.1002/14651858.CD008986.pub2.

Author Information

  1. 1

    Bispebjerg Hospital, Department of Clinical Pharmacology, Copenhagen, Denmark

  2. 2

    Rigshospitalet, The Nordic Cochrane Centre, Copenhagen, Denmark

*Mikkel Christensen, Department of Clinical Pharmacology, Bispebjerg Hospital, Bispebjerg Bakke 23, Copenhagen, 2400, Denmark. mch@dadlnet.dk.

Publication History

  1. Publication Status: New
  2. Published Online: 28 FEB 2013

SEARCH

 
Characteristics of included studies [ordered by study ID]
Gallagher 2011

MethodsRandomised controlled trial


Participants400 patients admitted via the emergency department under the care of a general medical physician at a tertiary medical centre at university hospital in Ireland. Median (IQR) age: intervention = 75 (71 to 80), control = 77 (71 to 82); 47% male; mean number of drugs: 7.7


InterventionsThe primary research physician applied STOPP/START criteria to the baseline data of patients in the intervention group to identify potentially inappropriate prescriptions and prescribing omissions. Research physician discussed with attending medical team and written communication within 24 hours. Team were not obliged to follow

No reporting of cointerventions


OutcomesThe primary outcome measures: MAI and the AOU index
The secondary outcome measures: mortality, frequency of general practitioner visits, hospital readmissions and falls

All outcomes had 6 months of follow-up


NotesFunding: The study was funded by the Health Research Board of Ireland, Clinical Research Training Fellowship number CRT/2006/029 


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskThe randomisation sequence was determined by an independently generated random numbers table using StatsDirect software, version 4.5 (StatsDirect; http://www.statsdirect.com). The random numbers table was retained, independent of the researchers, by a physician external to the study who assigned participants to groups using a sealed envelope system 

Allocation concealment (selection bias)Low riskGroup allocation was concealed from the research physician and participants until baseline data had been collected and inclusion criteria verified 

Blinding of participants and personnel (performance bias)
All outcomes
High riskStudy described as not blinded

Blinding of outcome assessment (detection bias)
Mortality (all-cause)
Low riskData collected by research physician who was aware of assignments, but will likely not influence assessment of mortality

Blinding of outcome assessment (detection bias)
Hospital readmissions (all-cause)
Low riskData collected by research physician who was aware of assignments, but will likely not influence assessment of hospital readmission

Blinding of outcome assessment (detection bias)
Adverse drug events
High riskData collected by research physician who was aware of assignments, may influence assessment of adverse drug events

Incomplete outcome data (attrition bias)
Mortality (all-cause)
Low riskMissing data not described. Called GP to obtain data. Registry data on mortality likely used

Incomplete outcome data (attrition bias)
Hospital readmissions (all-cause)
Low riskRegistry data and GP contact, no description of loss to follow-up, so data should be available

Incomplete outcome data (attrition bias)
Adverse drug events
Unclear riskFalls (=adverse events) obtained by telephone to patient's or GPs. No description on how many times patients could not be contacted. Lack of contact could be related to falls and GPs might not know if patients have fallen

Selective reporting (reporting bias)Low riskNo data on adverse events, but requires coding, so likely not assessed

Other biasLow riskNo evidence of other types of bias

Gillespie 2009

MethodsRandomised controlled trial


Participants400 patients admitted to two acute internal medicine wards at a university hospital in Sweden. Mean age: 87 years; 41% male; mean number of drugs: 8.0


InterventionsA comprehensive list of current medications was compiled on admission to complement that obtained in the ED, ensuring that the medication list received by the ward was correct. A drug review was performed, and advice was given to the patient’s physician on drug selection, dosages, and monitoring needs, with the final decision made by the physician in charge. Patients were educated and monitored throughout the admission process, and received discharge counselling.

Cointerventions: Information about discharge medications (e.g. rationale for changes, therapeutic goals, and monitoring needs for newly commenced drugs) was communicated to the primary care physicians by the study pharmacists. A follow-up telephone call to patients two months after discharge was conducted


OutcomesThe primary outcome measure: Frequency of hospital visits (emergency department and readmissions (total and drug related))
The secondary outcome measure: Cost of hospital care
Mortality not stated, but also measured

All outcomes had 12 months of follow-up


NotesFunding: This study was funded by Uppsala County Council, University Hospital of Uppsala, Uppsala University, Apoteket AB, and Swedish Society of Pharmaceutical Sciences


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Unclear riskNo description of sequence generation

Allocation concealment (selection bias)Unclear riskClosed-envelope technique. Randomisation was performed in blocks of 20 (each block contained 10 intervention and 10 control allocations).  Unclear who included patients, but the 10 block arrangement and unblinding could make it possible to predict group

Blinding of participants and personnel (performance bias)
All outcomes
High riskStudy described as not blinded

Blinding of outcome assessment (detection bias)
Mortality (all-cause)
Low riskNot described if assessment was blinded, but when based on hospital records, it will likely not influence assessment of mortality

Blinding of outcome assessment (detection bias)
Hospital readmissions (all-cause)
Low riskThe two researchers responsible for analysing readmission data were blinded regarding the group to which the patients had been randomised 

Blinding of outcome assessment (detection bias)
Hospital readmissions (due to adverse drug events)
Low riskThe physician in charge of the patient was required to document in the medical record if readmissions were drug related. The physicians making this decision were blinded as to whether the patients were study participants. 

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (all-cause)
Low riskBased on national records

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (due to adverse drug events)
Unclear riskBlinded for drug related readmissions, but not described for emergency department contacts. Coded as unclear

Incomplete outcome data (attrition bias)
Mortality (all-cause)
Low riskDescribes 0 lost to follow-up

Incomplete outcome data (attrition bias)
Hospital readmissions (all-cause)
Low riskDescribes 0 lost to follow-up

Incomplete outcome data (attrition bias)
Hospital readmissions (due to adverse drug events)
Low riskDescribes 0 lost to follow-up

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (all-cause)
Low riskDescribes 0 lost to follow-up

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (due to adverse drug events)
Low riskDescribes 0 lost to follow-up

Selective reporting (reporting bias)Low riskNo data on adverse drug events, but likely not assessed. The registered protocol only specifies primary outcome, which is unchanged

Other biasLow riskNo evidence of other types of bias

Lisby 2010

MethodsRandomised controlled trial


Participants100 patients admitted to an acute ward of internal medicine at regional hospital in Denmark. Mean age: 79.2 years; male 39%; mean no. drugs: 10.2


InterventionsPharmacists and pharmacologist intervention was accomplished in two steps. First, a clinical pharmacist systematically collected information about the patients’ medication and second, the collected medical histories were discussed with a clinical pharmacologist (MD) according to the patients’ entire medical records including medical histories and laboratory test results. Discrepancies, inappropriate drugs, doses, routes, dosing schedules or inappropriate interactions between drugs would be described in an advisory note with recommendation for changes. Ward physicians were not obliged to follow these recommendations. No reporting of cointerventions


OutcomesThe primary outcome measure: Length of hospital stay (hours)

The secondary outcome measures: Time to first admission, readmissions, emergency department visits, visits to outpatient care clinic, general practitioner visits, specialists visits, after hours care, quality of life assessment, mortality

All outcomes had 3 months of follow-up


NotesFunding: ALIS, Amgros I/S


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskEligible patients were randomly assigned to either intervention or control by a computer-generated code

Allocation concealment (selection bias)Unclear riskAllocation concealment not described

Blinding of participants and personnel (performance bias)
All outcomes
High riskStudy described as not blinded

Blinding of outcome assessment (detection bias)
Mortality (all-cause)
Low riskStudy described as not blinded, but will likely not influence assessment of all-cause mortality

Blinding of outcome assessment (detection bias)
Hospital readmissions (all-cause)
Low riskStudy described as not blinded, but will likely not influence assessment of readmission as based on registry data

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (all-cause)
Low riskStudy described as not blinded, but will likely not influence assessment of emergency department contacts as based on registry data

Incomplete outcome data (attrition bias)
Mortality (all-cause)
Low riskMissing data not described, all data readily obtainable from central registry

Incomplete outcome data (attrition bias)
Hospital readmissions (all-cause)
Low riskNo description of missing data, all data should be possible to obtain from registry

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (all-cause)
Low riskNo description of missing data, all data should be possible to obtain from registry

Selective reporting (reporting bias)Low riskNo data on adverse drug events, but requires coding, so likely not assessed

Other biasLow riskNo evidence of other types of bias

Lisby 2011

MethodsRandomised controlled trial


Participants108 patients admitted to an orthopaedic ward at regional hospital in Denmark. Mean age: 80.5 years; male 29%; mean no. drugs: 6.7


InterventionsSystematic medication review by a clinical pharmacist and a clinical pharmacologist. A clinical pharmacist obtained medication history through medical records, electronic prescribing system, registry of drug purchase and interview after ward physician had ordered in-hospital medication. Subsequently, the case was conferred with a clinical pharmacologist and an advisory note with recommendations for medication changes was prepared and handed directly to the physician responsible for the ward round. Ward physicians were not obliged to follow these recommendations. No reporting of cointerventions


OutcomesPrimary outcome measures: Time to first unscheduled physician contact (general practitioner, emergency department, ambulatory care or hospital) after discharge from the Orthopaedic Department
Secondary Outcome Measures: Admission time, time to first readmission, number of readmissions, emergency department visits, visits to outpatient care clinic, general practitioner contacts, If first contact to GP included medication issues, contacts to physicians outside working hours, medical specialists contacts, quality of life assessment, mortality

All outcomes had 3 months of follow-up


NotesFunding: The Health Insurance Foundation in Denmark


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskEligible participants were randomly assigned to either intervention or control by a computer-generated code

Allocation concealment (selection bias)Unclear riskNot described

Blinding of participants and personnel (performance bias)
All outcomes
High riskStudy described as not blinded

Blinding of outcome assessment (detection bias)
Mortality (all-cause)
Low riskNational and regional registries and contact to GP. Not described, but study described as not blinded. But will likely not influence assessment

Blinding of outcome assessment (detection bias)
Hospital readmissions (all-cause)
Low riskNational and regional registries and contact to GP. Not described, but study described as not blinded. But will likely not influence assessment

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (all-cause)
Low riskNational and regional registries and contact to GP. Not described, but study described as not blinded. But will likely not influence assessment

Incomplete outcome data (attrition bias)
Mortality (all-cause)
Low riskNo description of missing data, all data should be readily obtainable from registry

Incomplete outcome data (attrition bias)
Hospital readmissions (all-cause)
Low riskNo description of missing data, all data readily obtainable from central registry

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (all-cause)
Low riskNo description of missing data, all data should be possible to obtain from registry

Selective reporting (reporting bias)Low riskNo data on adverse drug events, but requires coding, so likely not assessed

Other biasLow riskNo evidence of other types of bias

Schnipper 2006

MethodsRandomised controlled trial


Participants178 patients admitted to the general medicine service at a university hospital in the United States of America. Mean age: 59.3 years; male 34%; median no. drugs: 8.0


InterventionsThe pharmacist intervention on the day of discharge consisted of several parts. First, discharge medication regimens were compared with preadmission regimens and all discrepancies were reconciled with the medical team’s help. Patients were screened for previous drug related problems, including non-adherence, lack of efficacy, and side effects. The pharmacist reviewed the indications, directions for use, and potential adverse effects of each discharge medication with the patient and discussed significant findings with the medical team.

Cointerventions: A follow-up telephone call, where the pharmacist compared the patient’s self-reported medication list with the discharge list, exploring any discrepancies. The pharmacist also asked about medication adherence, possible adverse drug events (ADEs), and adherence with scheduled follow-up and laboratory appointments. Significant findings were entered into were communicated to the patient’s primary care physician


OutcomesThe primary outcome measure: Preventable ADE in patients
The secondary outcome measures: All ADEs (preventable or not), patient satisfaction, health care utilisation (readmission + ER contact), medication adherence, medication discrepancies

All outcomes had 30 days of follow-up


NotesFunding: This study was supported by the Division of General Medicine at Brigham and Women’s Hospital (BWH), Boston, Mass, the Fish and Anderson Fundsat BWH, and an unrestricted grant from the Merck Co Foundation, West Point, Pa. Dr Schnipper is supported by Mentored Clinical Scientist Development Award HL072806 from the National Heart, Lung, and Blood Institute, Bethesda, Md


Risk of bias

BiasAuthors' judgementSupport for judgement

Random sequence generation (selection bias)Low riskRandomisation was by a computer-generated algorithm

Allocation concealment (selection bias)Low riskSealed opaque envelopes opened only after patient consent was obtained

Blinding of participants and personnel (performance bias)
All outcomes
High riskNo blinding of the patients and staff

Blinding of outcome assessment (detection bias)
Hospital readmissions (all-cause)
Low riskOutcomes were assessed by research assistants and manuscript authors blinded to treatment assignment

Blinding of outcome assessment (detection bias)
Hospital readmissions (due to adverse drug events)
Low riskOutcomes were assessed by research assistants and manuscript authors blinded to treatment assignment

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (all-cause)
Low riskOutcomes were assessed by research assistants and manuscript authors blinded to treatment assignment

Blinding of outcome assessment (detection bias)
Hospital emergency department contacts (due to adverse drug events)
Low riskOutcomes were assessed by research assistants and manuscript authors blinded to treatment assignment

Blinding of outcome assessment (detection bias)
Adverse drug events
Low riskOutcomes were assessed by research assistants and manuscript authors blinded to treatment assignment

Incomplete outcome data (attrition bias)
Hospital readmissions (all-cause)
Unclear riskAll patients included in denominator, but it seems that 20 in intervention group and 18 in the control group were lost to follow-up

Incomplete outcome data (attrition bias)
Hospital readmissions (due to adverse drug events)
Unclear riskAll patients included in denominator, but it seems that 20 in intervention group and 18 in the control group were lost to follow-up

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (all-cause)
Unclear riskAll patients included in denominator, but it seems that 20 in intervention group and 18 in the control group were lost to follow-up

Incomplete outcome data (attrition bias)
Hospital emergency department contacts (due to adverse drug events)
Unclear riskAll patients included in denominator, but it seems that 20 in intervention group and 18 in the control group were lost to follow-up

Incomplete outcome data (attrition bias)
Adverse drug events
Unclear riskAll patients included in denominator, but it seems that 20 in intervention group and 18 in the control group were lost to follow-up

Selective reporting (reporting bias)Unclear riskNo deaths reported, some data should be available. For example by spouses of patient or hospital records

Other biasLow riskNo evidence of other types of bias

 
Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion

[no author] 2001Not a randomised controlled trial

Al Mazroui 2009Population did not meet inclusion criteria (included outpatients)

Al-Rashed 2002Intervention did not meet inclusion criteria (drug information)

Allen 1986Intervention did not meet inclusion criteria (geriatric team)

Bolas 2004Intervention did not meet inclusion criteria (medication reconciliation)

Burleson 2003Intervention did not meet inclusion criteria (medication history)

Burnett 2009Outcome did not meet inclusion criteria (medication appropriateness)

Cardinale 1993Not a randomised controlled trial

Ciechanover 1987Not a randomised controlled trial

Crotty 2004Intervention did not meet inclusion criteria (medication reconciliation)

Gattis 1999Population did not meet inclusion criteria (included outpatients)

Hellstrom 2011Not a randomised controlled trial

Kelly 2011Not a randomised controlled trial

Koehler 2009Intervention did not meet inclusion criteria (not medication review)

Lipton 1992Outcome did not meet inclusion criteria (medication appropriateness)

McMullin 1999Not a randomised controlled trial

Naughton 1994Intervention did not meet inclusion criteria (geriatric team)

Pope 2011Intervention did not meet inclusion criteria (medication review implemented after discharge)

Rainville 1999Intervention did not meet inclusion criteria (only heart failure medication reviewed)

Saltvedt 2005Outcome did not meet inclusion criteria (medication use)

Schmader 1997Not a randomised controlled trial

Schmader 2004Intervention did not meet inclusion criteria (geriatric team)

Scullin 2007Intervention did not meet inclusion criteria (complex pharmacist intervention not focused on medication review)

Smith 1996Not a randomised controlled trial

Spinewine 2007Not a randomised controlled trial (quasi-randomised trial, used alternate randomisation)

Stowasser 2002Intervention did not meet inclusion criteria (medication reconciliation)

Walker 2009Not a randomised controlled trial

 
Characteristics of ongoing studies [ordered by study ID]
ISRCTN08043800

Trial name or titlePharmacists' review of medicine during admission to hospital

MethodsRandomised controlled trial

ParticipantsInclusion Criteria:

  • Patients being admitted to an internal medicine ward


  • Age 18 years or more


  • Taking 4 types of medicine or more each day


  • Able to understand participant's information written in Danish


Exclusion criteria:

  • Patients transferred from other hospitals in the area


  • Dying or terminally ill patients


  • Patients being discharged within 48 hours from admission

InterventionsIntervention Group:
1. Review and use of patient's own drugs by clinical pharmacist
2. Clinical pharmacist taking secondary medication history
3. Medication review by clinical pharmacist
4. Entry of proposed prescriptions in the electronic medication system by pharmacist, ready for approval by doctor
The intervention takes place on the day the patient is admitted, and the duration of the intervention is approximately 1.5 hours

Control Group:
Standard care with no pharmacist involvement

OutcomesPrimary Outcome Measures:

Number of patients with in-hospital adverse drug events, detected by Adverse Drug Event Trigger Tool

Secondary Outcome Measures:

1. Length of hospital stay
2. Number of readmissions during the first year after admission
3. Direct cost for the hospital

Starting dateMarch 2009

Contact informationPrincipal Investigator: Trine R. H.  Nielsen, Region Zealand Hospital Pharmacy, Denmark

Noteswww.controlled-trials.com (accessed August 2011). Trial ID: ISRCTN08043800

NCT00844025

Trial name or titlePharmaceutical Care and Clinical Outcomes for the Elderly Taking Potentially Inappropriate Medication

MethodsRandomised controlled trial

ParticipantsInclusion Criteria:

  • Hospitalised patients aged at least 65 years
  • Taking at least six prescribed medicines regularly, including at least one potential inappropriate medication


Exclusion Criteria:

  • Patients who refused informed consent
  • Discharged before consent could be obtained
  • Cognitive impaired

InterventionsPatients in the intervention group will receive pharmaceutical care delivered by clinical pharmacist, which includes medication review, medication reconciliation, patient education and recommended actions. Patients randomised to usual care group will receive routine review of medication by ward-based pharmacist and nurse

OutcomesPrimary Outcome Measures: Number of unsolved drug related problems
Secondary Outcome Measures: Rate of adverse drug event during hospitalisation and number of potentially inappropriate medication

Starting dateFebruary 2009

Contact informationPrincipal Investigator: Liu Jen Wei, MS, Shin Kong Wo Ho-Su Memorial Hospital, Department of Pharmacy, Taipei, Taiwan

Noteswww.clinicaltrials.gov (accessed August 2011). Trial ID: NCT00844025  

 
Comparison 1. Primary outcome

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Mortality (all-cause)41002Risk Ratio (M-H, Fixed, 95% CI)0.98 [0.78, 1.23]

 
Comparison 2. Secondary outcomes

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Hospital readmission (all-cause)4956Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.88, 1.16]

 2 Hospital readmission (all-cause) - 3 months2206Mean Difference (IV, Fixed, 95% CI)-0.01 [-0.21, 0.20]

 3 Hospital Emergency Department contacts (all-cause)3574Risk Ratio (M-H, Fixed, 95% CI)0.64 [0.46, 0.89]

 4 Hospital Emergency Department contacts (all-cause) - 3 months2206Mean Difference (IV, Fixed, 95% CI)-0.05 [-0.17, 0.07]

 
Comparison 3. Subgroup analysis

Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size

 1 Mortality (all-cause)41002Risk Ratio (M-H, Fixed, 95% CI)0.98 [0.78, 1.23]

    1.1 Systematic medication review
1400Risk Ratio (M-H, Fixed, 95% CI)0.91 [0.51, 1.61]

    1.2 Nonsystematic medication review
3602Risk Ratio (M-H, Fixed, 95% CI)1.00 [0.78, 1.28]

 2 Hospital readmission (all-cause)4956Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.88, 1.16]

    2.1 Systematic medication review
1382Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.78, 1.37]

    2.2 Nonsystematic medication review
3574Risk Ratio (M-H, Fixed, 95% CI)0.99 [0.85, 1.17]

 3 Mortality (all-cause)41002Risk Ratio (M-H, Fixed, 95% CI)0.98 [0.78, 1.23]

    3.1 Low risk of bias
1400Risk Ratio (M-H, Fixed, 95% CI)0.91 [0.51, 1.61]

    3.2 High risk of bias
3602Risk Ratio (M-H, Fixed, 95% CI)1.00 [0.78, 1.28]

 4 Hospital readmission (all-cause)4956Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.88, 1.16]

    4.1 Low risk of bias
1382Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.78, 1.37]

    4.2 High risk of bias
3574Risk Ratio (M-H, Fixed, 95% CI)0.99 [0.85, 1.17]

 
Summary of findings for the main comparison. Medication review compared with standard care for hospitalised adult patients

Medication review compared with standard care for hospitalised adult patients

Patient or population: Hospitalised adult patients

Intervention: Medication review

Comparison: Standard care

OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No. of Participants
(studies)
Quality of the evidence
(GRADE)
Comments

Assumed riskCorresponding risk

Standard careMedication review

Mortality (all-cause)

1 year
Low risk populationRR 0.98 (0.78 to 1.23)1002
(4 studies)
⊕⊕⊝⊝

low2,3
NA

200 per 10001 196 per 1000
(156 to 246)

High risk population

400 per 10001 392 per 1000
(312 to 492)

Hospital readmission (all-cause)

1 year
Low risk populationRR 1.01 (0.88 to 1.16)956

(4 studies)
⊕⊕⊝⊝

low4,5
NA

300 per 10001 303 per 1000
(264 to 348)

High risk population

600 per 10001 606 per 1000
(528 to 696)

Hospital Emergency Department contacts (all-cause)

1 year
Low risk populationRR 0.64 (0.46 to 0.89)574
(3 studies)
⊕⊕⊕⊝
moderate6
Equal to a number

needed to treat

of 9 for the high risk

population and 28 for

the low risk population

100 per 10001 64 per 1000
(46 to 89)

High risk population

300 per 10001 192 per 1000
(138 to 267)

*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; NA: Not applicable; RR: Risk Ratio.

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

 1Risk in the high risk population of the control group was based on data from Gillespie 2009, the only trial with 12 months of follow-up (all outcomes). For the low risk population the 6 months of follow-up for Gallagher 2011 (mortality) and 3 months of follow-up for Lisby 2010 (hospital Emergency Department contacts) was extrapolated to give a 12 month risk. All trials had similar control group risks of Hospital readmissions and the low risk group was based on half of the risk of the high risk group.
2 The 'Risk of bias' assessments determined that Gillespie 2009, Lisby 2010 and Lisby 2011 had unclear risk of selection bias. Schnipper 2006 did not report on mortality and the risk of bias from selective outcome reporting was regarded as unclear (downgraded one category).
3 The point estimate was very close to one and the wide range of the CI included both a reduction and an increase in mortality (downgraded one category).
4 The 'Risk of bias' assessments determined that Gillespie 2009, Lisby 2010 and Lisby 2011 had unclear risk of selection bias (downgraded one category).
5 The point estimate was very close to one and the wide range of the CI included both a reduction and an increase in readmissions (downgraded one category).
6 The 'Risk of bias' assessments determined that Gillespie 2009, Lisby 2010 and Lisby 2011 had unclear risk of selection bias (downgraded one category).