Rapid Analysis of Mental Services (RAMS) An overview of correlates of high service utilisation in South-East Metropolitan Melbourne

1
Rapid Analysis of Mental Services (RAMS)An
overview of correlates of high service
utilisation in South-East Metropolitan Melbourne

• A/Prof Fabrizio Carinci
• Director, Centre for Health Systems Research
• February 2003

2
Rapid Analysis of Mental Services (RAMS)A phase
- Approch
1. explore correlates of high demand and
utilisation of health services both at the
individual and organisational level 2. build
user interfaces that would allow to simplify and
automate the procedure of rapid analysis of
mental services
3
Rapid Analysis of Mental Services (RAMS)SD phase
- Scaling down operational data
4
Rapid Analysis of Mental Services (RAMS)QT Phase
(Question Time)

• Main clinical question
• Identification of individual and structural
  characteristics associated to an increased rate
  of events in terms of the following outcome
  indicators
• abnormal length of stay
• readmissions
• excessive utilisation of mental services
  (frequent flyers)
• Population
• Southern Health mental health catchment area
• Time of observation
• 1/10/1999 - 30/9/2002

5
Rapid Analysis of Mental Services (RAMS)OD Phase
- Outcomes Definition
Analysis of the Literature Analyst/Statistician
(F.Carinci) Epidemiologist (E.Villanueva)
 
6
Rapid Analysis of Mental Services (RAMS) ADM
Phase - Analytical Data Management
 
7
RAMS Outcomes Datasets (SHCN Mental Services,
1/10/1999-30/9/2002)
Outcomes N Dataset Name Statistical Unit Time of Reference Observations per patient Model
Excessive Cumulative LOS HIGHLOS (TTLOSgt45) Cum. 3yrs LOS per subject higher than 45 days (75) 3,243 HIGHLOS Individual Subject Outcome condition met or last record Single Cox PH
High number of Contacts HIGHCONT (N_CONTgt100) More than 100 Contacts during the last 3 yrs (75) 3,243 HIGHCONT Individual Subject Outcome condition met or last record Single Cox PH
Revolving Door Patients REVOLVE (N_ADMgt3) Four or more Admissions during the last 3 years 3,243 REVOLVE Individual Subject Outcome condition met or last record Single Cox PH
Early Rehospitalisation EARLREHOSP Rehospitalisation within 28 days New Readmission READM Readmission at any time 6,283 READM Readmission / Last Record Previous Discharge to outcome condition or last record Multiple Cox PH / Andersen-Gill method (left truncation)
Excessive LOS ADMLOS (TLOSgt28) Admission longer than 28 days 6,494 ADMIS Admission Event Last available admission event per admission Multiple GEE / Logistic
New Long Stay NLS (LOSgt28) Consecutive in-hospital presence for more than 28 days 22,932 NLS Admission Event Admission Event Multiple GEE / Logistic
3,243 _FIRSRC_ Individual Subject First record Single
3,243 _LASTRC_ Individual Subject Last record Single
302,664 RAMSCOHO Any Record All records Multiple
8
RAMS catchment area (Melbourne Metro) individual
subjects
9
Demographic and clinical characteristics
10
Socioeconomic and service related characteristics

• Variable (Categories) _LASTRC_ HIGHLOS P(c2) HIG
  HCONT P(c2) REVOLVE P(c2)
• N () 3243 (100.0) 891 (27.5) 819 (25.3) 347
  (10.7)
• Employed
• Yes 344 (10.6) 49 (14.2) lt.0001 54
  (15.6) lt.0001 20 (5.8) .0020
• No 2899 (89.4) 842 (29.0) 765 (26.4) 326
  (11.2)
• Subordinate/Non Managerial
• Yes 2686 (82.8) 791 (29.4) lt.0001 745 (27.8)
  lt.0001 319 (11.9) lt.0001
• No 557 (17.2) 100 (17.9) 74 (13.1) 27 (4.8)
• Education 7-10 years
• Yes 2266 (69.9) 629 (27.8) 635 (28.1)
  lt.0001 279 (12.3) lt.0001
• No 977 (30.1) 262 (26.8) 184 (18.8) 67 (6.8)
• Assisted by a carer
• Yes 368 (11.3) 161 (43.7) lt.0001 130 (34.1)
  lt.0001 59 (15.9) .0006
• No 2875 (88.6) 730 (25.4) 689 (24.1) 287
  (10.0)

11
Risk Model HIGHLOS
OutcomeCumulative .LOSgt45
Cox Regression Results
Variable
HR 95 CI
Age
0.99 0.99-1.00
Male
0.93 0.81-1.07
F1 Substance Abuse
0.74 0.61-0.89
F2 Schizophrenia
0.63 0.54-0.74
F3 Mood Disorder
0.78 0.67-0.91
F4 Neurotic
0.55 0.44-0.68
F9 Emotional
0.58 0.35-0.97
Dandenong
1.39 1.14-1.71
Acute Child
2.32 1.71-3.15
Aged Care
1.79 1.32-2.44
1-3 Contacts per month
1.48 1.22-1.80
gt3 Contacts per month
2.63 2.23-3.11
Average post-discharge delay 4-6 days
0.68 0.54-0.86
Average post-discharge delay 1 wk or more
0.69 0.58-0.81
Employed
0.59 0.44-0.79
Education 7-10 years
1.23 1.05-1.44
Occupation Subordinate/Non Managerial
1.62 1.31-2.01
Z51-Z65 Social Problems
0.54 0.36-0.81
Z72 Lifestyle/Addiction
1.53 1.09-2.14
Other
0.64 0.49-0.84
Dandenong 3175
1.26 1.01-1.57
Noble Park 3174
1.54 1.22-1.95
Bentleigh East 3165
1.67 1.19-2.33
Clayton South 3169
1.60 1.09-2.35
Cheltenham 3192
1.53 1.11-2.11
Doveton 3177
1.94 1.28-2.95
Hampton Park 3976
1.60 1.03-2.46
0.25
0.5
1.0
2.0
5.0
Decreased Risk
Increased Risk
HAZARD RATIO
12
Risk Model REVOLVE
OutcomeT.No.Adm.gt4
Cox Regression Results
Variable
HR 95 CI
Age
0.98 0.97-0.99
Male
1.02 0.81-1.28
F3 Mood Disorder
1.33 1.06-1.66
F4 Neurotic
1.40 1.07-1.83
F6 Personality Disorder
1.97 1.53-2.53
F7 Mental retardation
2.33 1.31-4.14
F9 Emotional
0.39 0.17-0.90
Average LOS 1-3 days
2.92 1.87-4.55
Average LOSgt3 days
4.48 2.82-7.12
Dandenong
1.40 1.09-1.81
Acute Child
0.50 0.28-0.88
1-3 Contacts per month
2.33 1.66-3.28
gt3 Contacts per month
6.27 4.67-8.42
Average post-discharge delay 1 wk or more
1.60 1.28-2.00
Occupation Subordinate/Non Managerial
1.70 1.14-2.53
Assisted by a carer
1.49 1.11-2.00
Continent of Birth Asia
0.60 0.37-0.97
Bentleigh East 3165
1.96 1.23-3.12
Cheltenham 3192
0.35 0.13-0.94
Doveton 3177
1.92 1.04-3.54
0.01
0.25
0.5
1.0
2.0
5.0
10.0
Decreased Risk
Increased Risk
HAZARD RATIO
13
Conclusions / Action lines

• RAMS has produced a model to scale down CMI and
  perform rapid analysis for population health
• Results of predictive modelling are relevant to
  the organisation of health services
• The longitudinal structure of the model may be
  applied at the patient level
• Validated concepts and statistical results may be
  incorporated into CMI to support integrated care
• RAMS carries forward the SPHIS model, now capable
  of including contacts in the community
• Areas of improvement
• CMI lacks important information on known
  predictors of patient outcomes (medicines,
  compliance) that should be incorporated in the
  database to be relevant for fully integrated care
• Overall health services utilisation is
  underestimated. Admissions for mental health
  patients not related to mental health services
  are not currently included.
• A direct consequence is that comorbidity is
  underreported. CMI does include few non-MH
  diagnoses.
• All the above factors may induce uncontrollable
  bias in all statistical analyses. For instance,
  the impact of chronic diseases, excessive use of
  medicines or non compliance is practically
  unknown.
• Action plan
• Finalisation of the project report
• Refinement and publication of the results in
  collaboration with the clinical experts
• Publication and update of all SPHIS/RAMS material
  on dedicated website
• Presentation of the model to users/DHS
• Preparation of drafts for further development
  (data linkage with VAED/PBS, state-level
  analysis, national and international comparative
  studies)

14
Links

• Centre for Health Systems Research
• http//www.med.monash.edu.au/healthservices/chsr
• The RAMS project
• http//www.med.monash.edu.au/healthservices/chsr/p
  rojects/rams