wennberg international collaborative conference 'variation in excess cases of adverse events...
TRANSCRIPT
Varia%on in excess cases of adverse events amenable to health care: low value care with budgetary impact
Comendeiro-‐Malloe, Ridao-‐López M, Mar4nez-‐Lizaga N, Angulo-‐Pueyo E, García-‐Armesto S, Bernal-‐Delgado E on behalf of the Atlas VPM team
This is Micaela
BACKGROUND METHODS RESULTS QUESTIONS FOR DEBATE
BACKGROUND
Lower-‐value care (definiTon from paper in Berlin Spanish cases (from PSI paper) Rivard paper (copy & past)
DEFINITION OF VALUE
Underuse of effecTve intervenTons EffecTve intervenTons performed on non-‐eligible paTents IntervenTons with a more cost-‐effecTve alternaTve EssenTally ineffecTve intervenTons Low quality intervenTons Unsafe intervenTons
PSI
0.5 7.7 1.8 4.9 17.3
EB: 0.19 [CI:0.12;0.28]
Rivard
AIM
EsTmate the excess-‐cost ahributable to the appearance of postoperaTve PTE and/or DVT
METHODS
SETTING VARIABLES MAIN & SECONDARY ENDPOINTS ANALYSIS
Popula%on and seNng
• All paTents over 17, undergoing surgery in 2009 or 2010: 1.1 million paTents – CondiToned to paTents who survived – Excluded PTE/DVT as the cause of admission – Excluded stays below 2 days – Excluded admissions within MDC14: pregnancy, delivery and puerperium
• 50 hospitals with the largest surgical acTvity – 56% of the surgical acTvity in Spain
Variables
• Excess-‐cost – Excess length of stay
• Main predictor for excess-‐cost – Having or not postoperaTve PTE and/or DVT
• AlternaTve predictors for excess-‐cost – Age – Sex – ComorbidiTes – Hospital of treatment
Main & secondary endpoints
• Excess-‐LOS per hospital – Average – CondiToned to those exposed to the risk of having AE
• Excess-‐cost per hospital
Analysis
• StraTfied-‐descripTve • MulTlevel log-‐lineal -‐ average effect
• Kernel matching to determine average LOS-‐excess, condiToned to the risk of adverse event
Risk score
PaTents never have PTE/DVT
PaTents always have PTE/DVT
Risk of PTE/DVT
Subpop of paTents with a priori higher average-‐risk Subpop of paTents with a priori lower average-‐risk
Matching
PaTents never have PTE/DVT
PaTents always have PTE/DVT
Risk of PTE/DVT
Subpop of paTents with a priori higher average-‐risk Subpop of paTents with a priori lower average-‐risk
LOS No Event average (s.d.)
LOS event average (s.d.)
Differential*
n=1.064.836 n=7.777
PSI 12 or Adverse Event Overall cases 10,48 (13,89) 22,39 (24,87) 11,9
For each of the characteristics of the episodes Age From 18 to 39 7,6 (12,94) 30,24 (34,71) 22,64 From 40 to 64 9,7 (14,29) 24,44 (28,02) 14,74 65 or older 11,8 (13,69) 20,74 (21,91) 8,94
Gender Mail 11,07 (14,63) 22,86 (25,70) 11,79 Female 9,8 (12,96) 21,83 (23,85) 12,03
Comorbidities (Elixhauser) Pulmonary circulatory disease 15,54 (16,60) 23,56 (25,17) 8,02 Paralysis 22,02 (30,68) 40,73 (48,59) 18,71 Lymphoma 14,54 (15,72) 22,61 (25,71) 8,07 Cancer with metastasis 16,75 (16,39) 18,57 (16,52) 1,82 Metastasis without solid tumour 11,99 (14,50) 17,41 (14,54) 5,42 Coagulopathies 18,36 (21,30) 28,62 (29,43) 10,26 Weight loss 25,36 (26,82) 34,76 (31,27) 9,40
LOS in those with and without PTE/DVT
Hospital)Lenght)of)Stay)
Model1))Hospital)effect)(empty)model))
Model2))Hospital)effect)and)
Adverse)Event)
Model)3)RiskAadjusted)by)
the)morbidity)characteristics)of)the)episodes)and)
hospital)Episode(characteristics))(β"coeff,"95%"CI)""
" " "
Constant) 7,37"(7,16"""7,59)" 7,33"(7,12"""7,55)" 3,83"(3,63"""3,90)"Psi12(PTE"post"o"TVP)" " 2,03"(1,99"""2,07)" 1,40)(1,38)))1,43))Age) " " "
From"18"to"39"years" " " AAA"From"40"to"64"years" " " 0,99"(0,86"""1,00)"
65"or"older" " " 1,04"(1,40"""1,57)"Gender) " " "
Mail" " " AAA"Female" " " 0,99"(0,99"""1,00)""
Comorbidity)(Elixhauser)" " " "Paralysis" " " 1,32"(1,30"""1,33)"
Lymphoma" " " 2,67"(1,08"""1,13)"Metastatic"Cancer" " " 1,36"(1,35"""1,38)"Coagulopathies" " " 1,12"(1,10"""1,14)"
Weight"loss" " " 1,58"(1,56"""1,60)"Amount)of)secondary)diagnosis)" " " 1,11"(1,10"""1,11)"
(Hospital(effect( " " "
Variance"of"level"hospital"(SE)" 0,11" 0,11" 0,13"
Average excess LOS
Model Excess LOS*
Average excess LOS (log-‐linear) 1.40
Average excess LOS conditioned to risk (overall) 1.75 Average excess LOS conditioned to risk (within hospital) 1.74 Min 1.2 Max 2.4 EQ 1.7 IQ 1.3 * Basal Hospital length of stay: 3,83 days
*Peiró-‐Moreno S, García-‐Petit J, Bernal-‐Delagado E, Ridao-‐López M, Librero-‐López J. “El gasto hospitalario poblacional, variaciones geográbicas y factores determinantes”. Presupuesto y gasto público 2007;49:193-‐209
Excess LOS condi%oned to same risk
QUESTIONS FOR DEBATE
RISK MATCHING IS JUST BUILT ON OBSERVABLE FACTORS TIME-‐DEPENDENT BIAS MISS-‐CLASSIFICATION OF THE EVENT
JUST OBSERVABLE
• Are we missing some variables at paTent-‐level that could determine differences in risk, beyond the already considered in the risk score?
• Since the esTmaTon of the risk-‐score has considered unobservable hospital-‐specific variables (mu), and the event has been defined as a paTent safety event (likely ahributable to hospital care) – are we miTgaTng the bias?
Time-‐dependent bias? In the opposite sense
Miss-‐classifica%on of events How to increase PPV?
IMPACT
Alarm performance Alert performance Average performance Good performance Excellent performance
Flagging hospitals beyond a threshold VariaTon in the adjusted-‐incidence of PTE/DVT
Performance relative position in terms of
ATET Hospitals Episodes AE-‐PSI12 Cost differential
Alarm/Alert (% total)
10 (20%)
229,792 (24.6%)
1,736 (22.3%)
€ 19,705,221.16 (29%)
Average (% total)
29 (58%)
515,595 (55.1%)
4,168 (53.6%)
€ 36,010,872.96 (53%)
Good/Excellent (% total)
11 (22%)
190,015 (20.3%)
1,873 (24.1%)
€ 12,264,628.15 (18%)
Total 50
(100%) 935,402 (100%)
7,777 (100%)
€ 67,980,722.3 (100%)
Impact on costs
*Peiró-‐Moreno S, García-‐Petit J, Bernal-‐Delagado E, Ridao-‐López M, Librero-‐López J. “El gasto hospitalario poblacional, variaciones geográbicas y factores determinantes”. Presupuesto y gasto público 2007;49:193-‐209
8.4
4,0
5,0
6,0
7,0
8,0
9,0
10,0
3,0 5,0 7,0 9,0 11,0 13,0 15,0 17,0 19,0
Risk adjusted hospital incidence of adverse events*1,000 surgeries
ATET (extra days) after TVP (psi12)
Quadrant I Quadrant II
Quadrant III Quadrant IV
6.7
8.4
Concilia%ng safety and costs Incidence of Adverse Events vs. Average effect on the exposed
PTE & DVT aWer surgery across countries
1,43 7,55 0,87 4,36 1,87
ANNEX: MODEL SPECIFICATIONS
Log-‐linear specifica%on
Risk-‐score matching specifica%on