NAHM Methodology

National Audit of Hospital Mortality

 

Aim of National Audit of Hospital Mortality

National Audit of Hospital Mortality analyses HIPE data for purpose of identifying and trending mortality in acute hospitals in Ireland.

  • Identify potential areas for improvement.
  • The aims of this audit are to

    • Understand and improve the quality of hospital based mortality data
    • Stimulate reflection on the quality of overall patient care

     

    Methodology for NQAIS National Audit of Hospital Mortality

  •  aspects of the quality of care.
  • Variation in mortality patterns may be explained by:

    • random (statistical) variation;
    • differences in patient characteristics;
    • issues of data quality;

    Similar statistical methodologies are used internationally to account for the first two factors (see below). Data quality depends on the accuracy and depth of coding done locally. Quality of care is a potential explanation for differences when the other factors have been taken into consideration.

     

    Inclusion Criteria

    Data routinely collected by the hospital inpatient enquiry (HIPE) system and provided by the Healthcare Pricing Office (HPO) of the Health Service Executive is used in the model.

  • Palliative Care – any hospital episode with the ICD-10-AM code Z515 (palliative care) occurring in either the principle or secondary diagnosis fields are included in this analysis.  This was introduced in November, 2016.
  • Hospital discharge episodes are grouped using the Agency for Healthcare Research and Quality (AHRQ) Clinical Classification Software (CCS). This system collapses all ICD-10 diagnostic codes to 260 clinically meaningful groups therefore allowing adequate numbers of deaths within groups for statistical analysis. The following predictor variables are included in the model:

    • Age on admission
    • Gender
    • Admission type
    • Admission source
    • Number of emergency admissions in the previous 12 months
    • Co-morbidities – using the Charlson Index (modified). The Charlson co-morbidity score consists of 17 medical conditions with attached weights that predict the risk of death within one year on a population basis.
    • Indicator of Deprivation

     

    Exclusion Criteria

    Maternity and Day Case discharges are EXCLUDED.

     

    Calculation of expected deaths per hospital

  • Individual diagnosis level indicates an unusual pattern for a precise diagnosis such as pneumonia or acute myocardial infarction.
  • National Audit of Hospital Mortality allows each hospital to compare itself to the national norm (but not to other hospitals) at three levels:

    • All diagnosis” summary level simply indicates that an unusual mortality pattern has been detected somewhere within the vast range of diagnostic groupings.
    • Diagnostic category” level signals an unusual pattern within a broad “specialty” such as respiratory.

    A unique logistic regression equation is computed for each diagnostic group with 40+ deaths and a crude rate is used if <40 deaths.

    The expected number of deaths is calculated by summing the predicted number of deaths per CCS group for each institution. The observed number of deaths per institution is extracted from HIPE discharge data. The hospital standardised mortality ratio (SMR) is calculated using the equation:

    SMR =Observed number of deaths    X   100
    Expected number of deaths

    The SMR indicates the ‘difference from average’.   Confidence intervals (95%, 99.8%) are computed around each SMR value.  These confidence intervals/control limits indicate its statistical significance having adjusted for the numbers involved. The confidence intervals are wide if the numbers are small, and narrow if the numbers are large.

     

    CUSUM analysis

    SMR is supplemented by Cumulative Sum (CUSUM) analysis.    Where the SMR aggregates  mortality patterns across a defined time period, CUSUM analyses the sequence of discharges as a continuous time series and looks for evidence that the actual mortality pattern may have changed significantly from expected, even over a short time period. The CUSUM control chart is a statistical control chart which detects sequential changes in the difference between the cumulative observed deaths and the cumulative risk-adjusted expected deaths over time – even when relatively small numbers of cases are involved.

     

    Deployment

    NAHM is now live in the following 44 acute hospitals nationally:

    Bantry General HospitalCork
    Beaumont HospitalDublin
    Cappagh National Orthopaedic HospitalDublin
    Cavan General HospitalCavan
    Connolly HospitalDublin
    Cork University HospitalCork
    Croom HospitalLimerick
    Ennis HospitalClare
    Galway University HospitalsGalway
    Letterkenny University HospitalDonegal
    Louth County HospitalLouth
    Kilcreene Regional Orthopaedic HospitalKilkenny
    Mallow General HospitalCork
    Mater Misericordiae University HospitalDublin
    Mayo University HospitalMayo
    Mercy University HospitalCork
    Merlin Park University HospitalGalway
    Midland Regional Hospital, MullingarWestmeath
    Midland Regional Hospital, PortlaoiseLaois
    Midland Regional Hospital, TullamoreOffaly
    Monaghan General HospitalMonaghan
    Naas General HospitalKildare
    Nenagh HospitalTipperary
    Our Lady of Lourdes Hospital, DroghedaLouth
    Our Lady’s Children’s Hospital, CrumlinDublin
    Our Lady’s Hospital, NavanMeath
    Portiuncula University HospitalGalway
    Royal Victoria Eye and Ear HospitalDublin
    Roscommon County HospitalRoscommon
    Sligo University HospitalSligo
    South Infirmary Victoria University HospitalCork
    South Tipperary General HospitalTipperary
    St Columcille’s HospitalDublin
    St James’s HospitalDublin
    St John’s HospitalLimerick
    St Luke’s General HospitalKilkenny
    St Michael’s HospitalDublin
    St Vincent’s University HospitalDublin
    Tallaght HospitalDublin
    Temple Street Children’s University HospitalDublin
    University Hospital KerryKerry
    University Hospital LimerickLimerick
    University Hospital WaterfordWaterford
    Wexford General HospitalWexford